Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Franchising as a share contract : an empirical assessment Lafontaine, Francine 1988

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-UBC_1988_A1 L33.pdf [ 18.23MB ]
Metadata
JSON: 831-1.0098068.json
JSON-LD: 831-1.0098068-ld.json
RDF/XML (Pretty): 831-1.0098068-rdf.xml
RDF/JSON: 831-1.0098068-rdf.json
Turtle: 831-1.0098068-turtle.txt
N-Triples: 831-1.0098068-rdf-ntriples.txt
Original Record: 831-1.0098068-source.json
Full Text
831-1.0098068-fulltext.txt
Citation
831-1.0098068.ris

Full Text

FRANCHISING AS A SHARE CONTRACT: A N EMPIRICAL ASSESSMENT by FRANCINE L A F O N T A I N E B A A . Ecole des Hautes Etudes Commerciales, 1980 M.Sc. Ecole des Hautes Etudes Commerciales, 1982 A Thesis Submitted in Partial Fulfillment of the Requirements for the Degree of D O C T O R OF PHILOSOPHY in The Faculty of Graduate Studies Department of Economics We accept this thesis as conforming to the required standard The University of British Columbia August 1988 © Francine Lafontaine, 1988 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of The University of British Columbia Vancouver, Canada DE-6 (2/88) ABSTRACT Contractual arrangements have been the subject of a substantial body of eco-nomic research. In particular, economists have sought an explanation for the exis-tence of share contracts. Under this kind of contract, two or more parties share in the output of the production process. These contracts present a problem to economists because they imply more than one residual claimant. Thus incentives are diluted and inefficiency is expected to result. But this type of contract has existed for centuries and continues to be used today. Why is that if they are inefficient? The answer is that under conditions of uncertainty and imperfect information, share contracts can be preferable to fixed-wage (vertical integration) or fixed-rent (market transac-tion) agreements. In fact, many explanations for the existence of share contracts and their coexistence with fixed-wage and rental arrangements are found in the theoretical literature. While the theoretical literature on the subject of share contracts has flourished over the last decade, empirical analyses of these models has lagged behind. This thesis aims to rectify the situation somewhat. More precisely, recent advances in the theoretical literature are applied to the analysis of franchise contracts. An em-pirical model of franchising based on profit-maximizing behavior is developed which makes it possible to examine whether the factors theorists have suggested as poten-tial explanations for share contracts are relevant when it comes to explaining what one observes in the context of franchising, and whether their effects are consistent with predictions from the various theories. Both the contract mix, i.e. franchisors' decisions concerning the proportion of stores they want to operate and franchise, and the terms of the franchise contract, fixed and variable fees, are examined. In order to carry out the analysis, data on a cross-section of 548 individual fran-chisors in 1986 were gathered. These franchisors are involved in a variety of business activities in the U.S., such as Fast-food Restaurants, Business Aids and Services, Construction and Maintenance, and Non-food Retailing. Censoring problems arise from the fact that a number of franchisors in the sample franchise all of their outlets. Also, some firms require no variable or no fixed fee. For these reasons, the maximum likelihood Tobit estimator is used. Empirical work in an area such as this, where theories rely on concepts that are not easily quantifiable, can hardly provide unambiguous answers about the validity of n the theories. Nevertheless, the following results emerge from the empirical analysis. First, the effect of risk, measured either by the proportion of discontinued outlets or by the variance of sales in the sector, is found to be the opposite of what pure risk-sharing and one-sided hidden-action models would predict. Second, firms resort to franchising more often when monitoring downstream operators becomes costlier, and use it proportionately less when the value of the inputs they themselves provide increases. This is consistent with two-sided hidden-action models. Results with respect to capital-market-imperfection arguments are rather inconclusive. It appears that franchising relaxes some form of constraint franchisors face in trying to expand their operations, since they use it more when they are growing faster, but whether this is a financial constraint remains unclear. The explanatory power of the model is greater with respect to the proportion of franchised stores than it is for any of the two fees. Thus, in response to changes in the exogenous variables considered here, franchisors, who have a choice between modifying the terms of their franchise contract or changing the proportion of stores they want to franchise, tend to do mostly the latter. Contrary to what one would have expected on a theoretical basis, the observed royalty rates and franchise fees are not negatively correlated in this data set. Com-bined with the fact that the model is less satisfactory relative to the fees, this suggests that there are considerations in the determination of the royalty rate and the fran-chise fee that have not been taken into account in the theories. One possibility in the case of the fixed fee is that it may include the price of services provided by the fran-chisors. It also appears that franchisors use input sales as another means to extract rent from franchisees. This may contribute to the lack of correlation between the two fees. Finally, the equation for the franchise fee was derived under the assumption that all remaining surplus at the downstream level, given the royalty rate, should be extracted through the franchise fee. The lack of relationship between the fees could be an indication that this assumption is incorrect, and that there are in fact rents left at the downstream level. This would be consistent with the existence of queues of potential franchisees in many franchise chains. m TABLE OF CONTENTS A B S T R A C T ii T A B L E OF CONTENTS iv LIST OF TABLES viii LIST OF FIGURES xi A C K N O W L E D G E M E N T xii I INTRODUCTION 1 II W H Y DOES FRANCHISING EXIST: IMPLICATIONS OF T H E CON-T R A C T U A L A R R A N G E M E N T LITERATURE 6 1. Introduction 6 2. The Nature of Franchising 6 3. An Overview of the Theoretical Literature on Share Contracts . . 10 3.1 Pure Risk-Sharing Models 11 3.2 One-Sided Hidden-Action Models 18 3.3 Two-sided Hidden-Action Models 19 3.4 Capital-Market-Imperfection Arguments 23 3.5 Self-Selection and Screening Models 24 4. Conclusion 27 III A N EMPIRICAL MODEL OF FRANCHISING 29 1. Introduction 29 2. An Overview of the Empirical Literature on Share Contracts . . . 29 - 2.1 The Empirical Literature on Sharecropping 29 2.2 Some Empirical Papers on Franchising 34 3. The Empirical Model 38 iv 3.1 Franchisors' Choices of Contractual Mix 40 3.2 Extension to the Contractual Design 46 4. Conclusion 51 IV T H E DATA: SOURCES AND CHARACTERISTICS 53 1. Introduction 53 2. The Sample of Franchisors 53 3. Some Interesting Descriptive Statistics 59 3.1 The Number of Outlets 59 3.2 The Use of Franchising 62 3.3 About the Contractual Design: Royalty Rates and Franchise Fees 65 3.4 A Comparison of the Two Samples 74 3.5 Measures of Geographical Dispersion and of Franchisors' Contri-bution 75 3.6 Some Relevant Sectoral Data 78 3.7 Alternative Measures of Risk 81 4. Conclusion 88 V T H E ECONOMETRIC SPECIFICATION AND RESULTS 90 1. Introduction 90 2. The Definition of the Variables 90 2.1 Measuring Risk or IU 91 2.2 Measuring Franchisees' Supervision Costs or i/ 92 2.3 Measuring the Franchisor's Contribution or IT 93 2.4 Measuring the Franchisors' Capital Constraint or IK 94 v 2.5 Some Other Relevant Variables 95 3. The Model Specification 98 3.1 Functional Form 98 3.2 Estimation Technique 100 3.3 Covariance Terms 103 3.4 Simultaneity Problems 104 4. The Empirical Results 104 4.1 The Effect of Risk, IU 110 4.2 The Effect of Franchisee Supervision Costs, 7/ I l l 4.3 The Effect of the Franchisor's Contribution, i j 113 4.4 The Effect of Franchisors' Capital Constraints, IK 117 4.5 The Effect of Input Sales and Royalty Rates 120 4.6 Some General Comments 120 5. Conclusion 123 VI RESULTS F R O M VARIOUS SUBSAMPLES OF FIRMS . 127 1. Introduction 127 2. Age Cohorts 128 3. Size Cohorts 136 4. Sectoral Effects 141 4.1 The Effect of Sectoral Dummy Variables 143 4.2 Differences Among Sectors 147 5. Conclusion 152 VII CONCLUSION 155 BIBLIOGRAPHY 158 vi SOURCES OF DATA 163 APPENDIX A: DESCRIPTIVE STATISTICS FOR T H E SAMPLE OF 890 FRANCHISORS 164 APPENDIX B: CORRELATION MATRIX OF T H E VARIABLES . . 170 APPENDIX C: HISTOGRAMS FOR T H E T H R E E D E P E N D E N T VARI-ABLES 172 APPENDIX D: ESTIMATION RESULTS EXCLUDING LIMIT OBSER-VATIONS 175 APPENDIX E: ESTIMATION RESULTS FOR T H E NUMBER OF FRANCHISES AND COMPANY-OPERATED STORES 180 APPENDIX F: T H E LONG T E R M T R E N D IN FRANCHISING . . . 183 APPENDIX G: RESULTS FOR GROUPS OF EIGHT COHORTS . . 185 vii LIST OF TABLES 3.1 Expected Effects of Indices on q//Q in the Various Models 51 4.1 Coverage of the Sample of 548 American Franchisors 58 4.2 Number of Outlets for the 548 Franchisors 60 4.3 Proportion of Franchised Outlets for the 548 Franchisors 63 4.4 Royalty Rates for the 548 Franchisors 66 4.5 Franchise Fees for the 548 Franchisors 68 4.6 Royalty Rates and Franchise Fees Within Size and Age Cohorts . . . . 72 4.7 Measures of Geographical Dispersion and of Franchisors' Contribution . 76 4.8 Some Relevant Sectoral Data 79 4.9 Alternative Measures of Risk 86 5.1 Descriptive Statistics for the 548 Franchisors 97 5.2 Proportion of Franchised Outlets Under a Linear Specification 106 5.3 Royalty Rates and Franchise Fees Under a Linear Specification 107 5.4 Proportion of Franchised Outlets Under a Partially Logarithmic Specifica-tion 108 5.5 Royalty Rates and Franchise Fees Under a Partially Logarithmic Specifi-cation 109 5.6 Expected and Observed Effects of the Indices on q//Q 124 6.1 The Proportion of Franchised Stores Within Age Cohorts 130 6.2 The Variable Fee Within Age Cohorts 131 6.3 The Franchise Fee Within Age Cohorts 132 6.4 The Proportion of Franchised Stores Within Size Cohorts 137 6.5 The Variable Fee Within Size Cohorts 138 viii 6.6 The Franchise Fee Within Size Cohorts 139 6.7 The Effect of Sectoral Dummy Variables Under a Linear Specification 144 6.8 The Effect of Sectoral Dummy Variables Under a Partially Logarithmic Specification 145 6.9 The Proportion of Franchised Outlets in Various Sectors 148 6.10 The Variable Fee in Various Sectors 149 6.11 The Franchise Fee in Various Sectors 150 A . l Number of Outlets for the 890 Franchisors 165 A.2 Proportion of Franchised Outlets for the 890 Franchisors 166 A.3 Royalty Rates for the 890 Franchisors 167 A.4 Franchise Fees for the 890 Franchisors 168 A. 5 Royalty Rates and Franchise Fees Within Size and Age Cohorts for the 890 Franchisors 169 B. l Correlation Matrix of the Variables (Sample of 548 Franchisors) . . . . 171 D. l OLS Regressions Without Limit Observations : Proportion of Franchised Stores - Linear 176 D.2 OLS Regressions Without Limit Observations : Proportion of Franchised Stores - in Log 177 D.3 OLS Regressions Without Limit Observations : Royalty Rates and Fran-chise Fees - Linear 178 D. 4 OLS Regressions Without Limit Observations : Royalty Rates and Fran-chise Fees - in Log 179 E. l OLS Regressions for the Number of Franchises 181 E. 2 TOBIT Regressions for the Number of Company-Operated Outlets . . 182 F. l The Proportion of Franchised Stores Within Eight Age Cohorts . . . . 186 ix F.2 The Royalty Rate Within Eight Age Cohorts 187 F.3 The Franchise Fee Within Eight Age Cohorts 188 F.4 The Proportion of Franchised Stores Within Eight Size Cohorts . . . . 189 F.5 The Royalty Rate Within Eight Size Cohorts 190 F.6 The Franchise Fee Within Eight Size Cohorts 191 x LIST OF FIGURES 1 Contractual Choices as Self-selection Mechanisms 26 2 The Trade-Off Between the Two Decision Variables, r and qf/Q . . . . 49 C . l Distribution of the Percentage of Franchised Outlets n=548 173 C.2 Distribution of Royalty Rates and Franchise Fees, n=548 174 F . l The Long Term Trend in Franchising 184 XI ACKNOWLEDGEMENT I would like to thank Mukesh Eswaran, Margaret Slade and Ken White for their continuous technical and moral support. As advisors for this thesis, they showed much patience and skill. I would also like to thank John Cragg, Ashok Kotwal and Hugh Neary for their helpful comments and advice. Most of all, I am indebted to Robert Picard; he and I both know why. I also want to acknowledge financial support from the Social Sciences and Hu-manities Research Council of Canada. xn CHAPTER I Introduction Contractual arrangements have been the subject of a substantial body of eco-nomic research. In particular, economists have sought an explanation for the existence of share contracts. Under this kind of contract, two or more parties share in the output of the production process. Examples include sharecropping, patent licensing, joint ventures, and franchising. These contracts present a problem to economists because they imply more than one residual claimant. Incentives are therefore diluted and inefficiency is expected to result. But this type of contract has existed for centuries and continues to be used today. Why is that if they are inefficient? The answer is that under conditions of uncertainty and imperfect information, share contracts can be preferable to fixed-wage (vertical integration) or fixed-rent (market transac-tion) agreements. In fact, many explanations for the existence of share contracts and their coexistence with fixed-wage and rental arrangements are found in the theoretical literature. While the theoretical literature on the subject of contractual choice has flourished over the last decade, empirical analyses of these models has lagged behind. This thesis aims to rectify the situation somewhat. More precisely, recent advances in the theoretical literature are applied to the analysis of franchise contracts. An empirical model of franchising, based on profit maximizing behavior, is developed. It will make it possible to examine whether the factors theorists have suggested as potential explanations for share contracts are relevant when it comes to explaining what one observes in the context of franchising, and whether their effects are consistent with predictions from the various theories. Because so little has been done in this area, this work is necessarily exploratory and descriptive. But such an analysis is useful in at least three different respects. First, it increases our understanding of franchising, which in and of itself is worth studying given its importance in the retailing sector and the service industries. Sec-ond, if share contracts exist for similar reasons whether in agriculture, patent licensing or the service sector, this analysis should further our understanding of this type of contract in general. Finallj', it indicates what factors seem to be especially relevant in the context of franchising, and it also suggests questions that deserve further at-1 tention. In that sense, it contributes to future theoretical and empirical research in the area. Franchising offers a rare opportunity to assess theories concerning firms' contrac-tual decisions.Franchisors tend to operate a certain number of stores and to franchise the others. In the first case, a manager, hired under a fixed-wage contract, is re-sponsible for the outlet.1 In the second case, the franchisee generally pays royalties based on sales or profits, which means that most franchise contracts are de facto share contracts. Franchisors mix these two types of contracts in varying proportions. For example, in 1986, McDonald's franchised 76.4% of its 9060 stores, and Burger King franchised 82% of its 4635 outlets. Consequently, one question that can be ad-dressed empirically is: What determines the proportion of stores each firm chooses to franchise? In other words, what are the factors that make them decide between fixed-wage and share contracts. The second advantage that arises from the use of data on franchising stems from the fact that each franchisor generally uses the same franchise contract, i.e. the same royalty rate and same franchise fee with all of its potential franchisees at any point in time. For that reason, it is possible to obtain information concerning the terms of the franchise contract for individual franchisors. Since the terms of the optimal share contract are determined endogenously in most theoretical models, one can then see if the models have some explanatory power with respect to the chosen royalty rates and franchise fees across franchisors. In order to carry out the empirical analysis, data on a cross-section of 548 individual franchisors in 1986 were gathered. These franchisors are involved in a variety of business activities in the U.S., such as Fast-food Restaurants, Business Aids and Services, Construction and Maintenance, and Non-food Retailing. Empirical work in an area such as this, where theories rely on concepts that are not easily quantifiable, can hardly provide unambiguous answers about the validity of the theories. Nevertheless, the following results emerge from the empirical analysis. 1 In reality, it is not necessary that the manager be paid a fixed wage. His compensation can be based on the outlet's performance to a large extent. What is necessary is that his contract not give him as much incentives to work as the franchise contract does. As pointed out by Goldberg (1982) and Brickley and Dark (1987), since the manager can not appropriate increases in the resale value of the outlet due to his good management, his contract necessarily gives him less incentives than that of the franchisee who gets those benefits when he sells his outlet. 2 First, franchisors' propensity to use franchising is non-decreasing in the amount of risk they face in their sector, measured either by the proportion of discontinued outlets or by the variance of sales in the sector. Similarly, royalty rates decrease with these measures of risk. Such a result is inconsistent with the risk-sharing argument for share contracts unless one is ready to assume that the franchisors are more risk-averse than their franchisees. It is also not consistent with one-sided hidden-action models, i.e. models where risk-neutral franchisors use franchising to provide insurance to their risk-averse franchisees, while at the same time giving them a type of contract that incites them to work. Of course these interpretations depend on the capacity of the measures to capture exogenous risk, as opposed to the variability that is due to moral hazard. And even if they do, in models with asymmetric information, increases in riskiness automatical!}' compound the unobservability problem. The more risk there is, the more difficult it is to assess peoples' behavior. This confuses issues and could affect the observed effect of risk on the proportion of franchised stores as well as the royalty rates. Still, if increased risk leads to a greater reliance on franchising due to the greater difficulty of evaluating franchisees' performance for example, the results imply that indeed incentive issues overwhelm risk-sharing considerations in the determination of the contract mix and of the terms of the contract. It is also found that franchisors tend to use franchising more when the cost of supervising dowstream operators increases due to increased geographical dispersion. Thus the notion that they use franchising contracts for their incentive properties vis-a-vis franchisees is supported by the data. This is in agreement with Brickley and Dark (1987) who found that outlets that are close to monitoring centers are more likely to be company-operated than those that are farther away. On the other hand, franchisors tend to franchise less when the value of their own inputs increases. Combined, these results lend empirical support to two-sided hidden-action models, i.e. those that posit that share contracts arise from market imperfections and the resulting need to provide incentives to both parties to the contract. Finally, with respect to received explanations for share contracts, it is found that franchisors use franchising more during periods of rapid expansion, which is consistent with the idea that they view franchisees as a source of capital. However, the proportion of franchised stores tends to decrease with the amount of capital required, which casts some doubt on a capital-market-imperfection explanation for 3 franchising. Thus franchisors may use franchising to relax some form of constraint that they face during periods of expansion, but the constraint need not be financial. In general, the explanatory power of the empirical model is satisfactory with respect to franchisors' decisions concerning the contractual mix, i.e. the proportion of stores they choose to franchise. Interestingly, however, the explanatory power of the variables is much lower in general for both the royalty-rate and the franchise-fee equations. Theoretical models of franchising such as those found in Rubin (1978), Blair and Kaserman (1982) and Mathewson and Winter (1985), analyse a single franchisor - franchisee pair. Comparative-static results in this case necessarily center on the terms of the franchise contract because these are the franchisor's only control variables. In a context where franchisors have many outlets, and can therefore choose to franchise some and operate others, they have an additional instrument at their disposal. And it is found here that franchisors often adjust to changes in the exogenous variables by modifying the proportion of stores they decide to franchise rather than changing the terms of their franchise contract. One explanation for this may He in the fact that each franchisor uses the same "average" share contract for all of its franchisees. Consequently, the franchise contract is not meant to be responsive for example to outlet-specific variables; some of the adjustments for differences among outlets or among outlet operators are handled instead through the choice of the type of contract. This behavior suggests either the existence of a significant cost associated with the development of franchise contracts, and/or the possibility that franchisors use other means to differentiate outlets and franchisees. For example, as indicated by Caves and Murphy (1976), the sale of inputs at a price greater than marginal cost can serve as another way for franchisors to extract rent from franchisees. Franchisors could specify different levels of input requirements depending on the profitability of the various outlets. However, the franchise contract is generally the same across franchisees in a given chain in this respect as well. Another variable franchisors have control over is the density of stores. They may be able to render various locations more similar by adjusting the number of stores in each area as a function of the level of demand. Finally, in the theoretical models, the fixed fee is generally assumed to be chosen so as to extract whatever rents may be left downstream given the royalty rate. Thus one would expect a negative correlation between the two fees. Also in the reduced-form equations, the same variables should explain both types of fees, although their 4 effect on each should be of opposite sign. However, this is not what is found in these data, suggesting that there is not necessarily a trade-off between the two fees. In other words, the fixed fee is not determined entirely once the variable fee is known. This could result from the inclusion in the fixed fee of amounts that represent a payment for services rendered by the franchisor, for example training. It may also be interpreted as an indication that some rents are left with the franchisees. No real surprises were found at the empirical level when the estimations were done separately for groups of firms defined on the basis of their size, measured in number of outlets, their age, measured by the number of years they have been in business, or the sector in which they operate. In general, the results that were obtained on the sample of franchisors as a whole were confirmed within subsamples of firms. The thesis is organized as follows. In the next chapter, I provide a brief de-scription of franchising, followed by an overview of the main explanations for share contracts that are found in the theoretical literature. In Chapter 3, the existing em-pirical literature on share contracts is briefly reviewed. Then, the empirical model is developed. This is done first with respect to firms' choices concerning the contract mix, and then extended to include their decisions concerning the terms of the fran-chise contract. The data used in the empirical analysis are described in Chapter 4. The econometric specification and general results obtained on the whole sample of franchisors are found in Chapter 5. Chapter 6 is concerned with estimating the model within various subsamples of firms defined on the basis of the number of years they have been in operation, the number of outlets in the franchise chain, and the sector in which the firm operates. Finally, concluding remarks are found in Chapter 7. 5 CHAPTER II Why Does Franchising Exist: Implications of the Contractual Arrangement Literature 1. Introduction This chapter contains first a brief definition and description of what franchising entails. This includes the presentation of a unifying framework to facilitate the discussion of the theoretical models in the context of franchising. The review of the theoretical literature on share contracts follows in Section 3. Most of the models discussed in this section were developed to explain the existence of sharecropping in agriculture and its coexistence with fixed-wage and fixed-rent agreements. These theories are classified here among five major categories: pure risk-sharing models, one-sided hidden-action models, two-sided hidden-action models, explanations based on capital market imperfections and, finally, self-selection and screening models. Each of these is discussed in turn, and then applied to the case of franchising. Concluding comments are found in Section 4. 2. The Nature of Franchising Franchising is a growing phenomenon: in Canada, sales through franchised outlets increased by 133% between 1976 and 1981.1 The Association of Canadian Franchisors estimates that they have grown by at least 10% each year since then. In the United States, they grew by 72% in the first half of the 1980's. Franchising now accounts for approximately 40% of Canadian and about 33% of American retail business.^ Technically, a franchise agreement is a contractual arrangement between two in-dependent firms, whereby the franchisee pays the franchisor for the right to sell the franchisor's product and/or the right to use his trademark at a given place and for a certain period of time. The U.S. Department of commerce classifies franchises ac-cording to what the main component of the transaction is. "Product and Tradename 1 Statistics Canada, Franchising in the Canadian Economy, 1976-1981. See ITS. Department of Commerce (1986). 6 Franchising", also referred to as "Traditional Franchising", is characterized by fran-chised dealers who "concentrate on one company's product line and to some extent identify their business with that company" ^ . Examples of this type of franchising are Car dealers and Gasoline Service stations. In "Business Format Franchising" on the other hand, the relationship between franchisor and franchisee "includes not only the product, service, and trademark, but the entire business format itself — a marketing strategy and plan, operating manuals and standards, quality control, and continu-ing two way communication"^. Examples of this include Restaurants, Business and Employment services, and Real Estate. Most of the growth in franchising since the 1950's has been in this latter category. In most Business Format franchises, the payment made by the franchisee takes the form of a fixed fee up front, the franchise fee, and royalties that are proportional to sales or sometimes to profits. In product and tradename franchises, franchisees do not pay royalties, but they are bound to buy their inputs from their franchisors. When downstream firms produce with a fixed-proportion technology, or in those cases where they buy all of their inputs from a single manufacturer, they can not substitute away from this manufacturer's products. In those circumstances, input mark-ups become equivalent to royalties (or a tax) based on output. That is, under fixed proportions, one can write q = min(x,y), so that in equilibrium, q = x = y. Therefore, the upstream firm can obtain the same equilibrium price and quantity and the same amount of revenue by taxing either q or x and/or In traditional franchising, downstream firms simply buy goods from their fran-chisors and resell them. Thus the conditions under which input mark-ups are equiva-lent to royalties on output generally hold in this type of franchise relationship. Con-sequently, these Product and Tradename franchisors, as well as Business Format franchisors, receive some form of variable payments from their franchisees. However, the value of these variable payments as a percentage of sales is unobservable in the case of product and tradename franchises. For this reason, my analysis concentrates on Business Format franchisors. All of the franchisors studied here belong to this ^ U.S. Department of Commerce, 1987, p. 1 ^ U.S. Department of Commerce, 1987, p. 3 ^ These however remain different from royalties on sales unless the price is also controlled by the franchisor. If it was, assuming a tax on output of t dollars, and a sales price of p, the equivalent royalty rate would simply be t/p. 7 category. Because franchising generally entails variable payments, it can be taken as a type of share contract similar to sharecropping, joint ventures, and patent licensing. It is, however, possible that a contract involving only fixed fees would be included in the franchising statistics. These types of contracts are not the norm. For example, in their survey of fast-food franchises in 1971, Ozanne and Hunt found 88% of franchisors required between 1 and 18% of sales to be paid in royalties. The remaining franchisors were the major suppliers of their franchisees and used input mark-ups. Similarly, of the 548 Business Format franchisors studied here, only 36 require no variable payments from their franchisees. In this paper, the term franchising will be used in the sense of share contract, keeping in mind that in some cases fixed-rental agreements are included in this category. In addition to monetary aspects, franchising arrangements often entail stipula-tions as to the type of assistance the franchisor will provide to the franchisee (site selection, training, accounting, ongoing guidance, etc.) and the manner in which the franchisor expects the franchisee to conduct the business. One may find clauses con-cerning hours of operation, prices, recruiting, cleanliness, etc. Typically, the contract also contains numerous termination clauses such that the franchisor can terminate the contract at will. It is the existence of these constraints on the franchisee's con-duct that makes such an arrangement resemble an employer-employee relation and casts some doubt on the "independence" of the franchisee. Yet, franchise agreements He somewhere between market and intrafirm transactions; if in some respects they resemble labour contracts, it remains that each agreement entails the participation of two independent firms. In the next section, I will be presenting an overview of the theoretical literature on share contracts. As a unifying framework for the various models, I assume that an upstream manufacturer derives some monopoly power from a tradename. This firm can open its own retail outlets (fixed-wage contract) or it can sell the right to sell its products and to use its tradename to independent retailers (franchising). ® Assume It is worth pointing out that in reality, all franchised outlets are not necessarily individually owned and operated. In other words, it is possible for a single franchisee to be responsible for more than just one store in a chain. For example, master franchises for whole geographical areas are sometimes granted to single individuals or firms. Because outlets that are franchised in this way are not distinguished from individually owned franchises in the data, possible 8 also that, for a given p, demand at the retail level can be written as Xi = f(T,li) + 6i (2.1) for each of n franchisees. / is increasing and concave in its two arguments, T represents the value of the trademark, which is the franchisor's responsibility, while stands for local inputs provided by the franchisee. These would include such things as the franchisee's managerial inputs and local advertising. 9i is a random variable that is i.i.d. with mean 0 and variance O~Q for all franchisees. This formulation implies that uncertainty is independent of input levels. But because of this uncertainty, it is not possible for the franchisor to infer the level of l{ given T and the ex-post level of demand. Similarly the franchisee can not infer T from his knowledge of A"; and l{. Thus there is potentially a two-sided moral hazard problem. Note that if the 6{ were perfectly correlated, the franchisor would be able to infer a ranking for the level of Z; provided by franchisees from the observed ranking of the X{. He could then use this information in devising an optimal franchise contract. In general, this would entail including some measure of other agents' performance in the contract for any one agent.7 With the 9{S independently distributed, no such ranking is possible. In that case, the optimal contract for each agent will depend on his output alone and not on anybody else's.** Observed franchise contracts are a function only of the sales level of the outlet, X{. They do not depend on any of the Xj. This can be interpreted to mean that the assumption of independence of the 6{ is reasonable in this context. The franchise contract is identical for all the franchisees of a given franchisor. It entails the payment of a fixed fee F, which is paid only once for the duration of the contract, and of royalties on sales r, where 0 < r < 1? The expected profits of the ramifications of this additional level of organization are not addressed here. As long as the effect of the franchise contract on the master franchisee is such as to make each one of his franchises benefit from the incentives he receives in a way that approaches the effect of individual franchise contracts, this should not present any problem at the empirical level. 7 See for example Lazear and Rosen (1981), Holmstrom (1982), Nalebuff and Stiglitz (1983) and Mookherjee (1984) on the problem of a single principal facing many agents. R See Mookherjee (1984) for a proof that subject to the constraint that agents play Nash among themselves, if the random variables in the agents' output function are independent, the optimal contracts are also independent in the sense that agent i:s payment scheme should not be made dependent on agent j's output. ^ The main reason for assuming an identical contract for all franchisees at this point is that this is what one observes in reality. Why this is so is not addressed here. 9 franchisor, II, and of each franchisee, iti, assuming that all outlets are franchised, are given by1^ n n = ^ [ r . p . A ' i - r F ] - C r ( n ) (2.2) Ti = (1 - r)p • Xi - C(Xi) - Ci{U) - F (2.3) where n is the total number of outlets in the chain, CT represents the cost of developing and upholding the value of the tradename, C(X{) stands for the costs of the retailing activity, and Ci is the cost of providing local inputs. Assuming an unlimited (infinitely elastic) supply of potential franchisees, franchisors choose the contractual structure and design the contracts in order to maximize their expected profits (or utility if they are risk-averse) subject to the participation constraint of franchisees. Competition among potential franchisees should ensure that they get only their opportunity level of utility, u. 3. An Overview of the Theoretical Literature on Share Contracts In the literature on sharecropping, it has been established that in a perfectly competitive economy, the choice between a fixed wage, a fixed-rental or a share contract would be irrelevant to the question of efficiency.11 Consequently, in such an economy, there would be nothing systematic about the contractual structure. This result of course is akin to that which has been established in the vertical integration literature about the equivalence, in terms of the upstream firm's profits, of fixed fees and/or royalties to vertical integration.12 The upstream firm could choose randomly which outlet it wishes to control which way. Nothing could explain the contractual structure in this case. It is clear that upon relaxing some of the assumptions of the perfectly competitive model in agriculture, one can gain some insight into the circumstances which would favor, for example, sharecropping over ^ Since the fixed fee is for the duration of the contract, if Xi here is interpreted as yearly sales, F would refer only to that portion of the fixed fee attributable to a year. If F represents the whole fixed fee, Xi would have to refer to total sales over the duration of the contract, which is 14 years on average according to the U.S. Department of Commerce. 1 1 See Reid (1977) p. 404. 1 2 See for example Blair and Kaserman (1980), Dixit (1983), Gallini and Winter (1983) and Mathewson and Winter (1983) on this subject. 10 other types of arrangement, and thus develop some theory of the determinants of contractual structures. Similarly, it is necessary to relax assumptions of perfect information concerning demand and costs in order to differentiate modes of vertical control from the upstream firm's point of view. Explanations for the existence of share contracts and their coexistence with fixed-wage and rental contracts that are found in the theoretical literature on contractual arrangements can be classified among five major categories. ^ I will discuss these in turn in the context of franchising. 3.1 Pure Risk-Sharing Models The possibility of risk-sharing that is embodied in the sharing rule was first put forward by Cheung (1969) as an explanation for the existence of sharecropping. When production is uncertain, the use of a fixed-wage contract makes the landlord bear all risk. Fixed-rent contracts on the other hand imply that the tenants have to bear all risk. With a share contract, both parties get fluctuations in their revenues when production varies, but the fluctuations are smaller. / / both are risk averse, they gain from this. In Cheung's (1969) model, share contracts are assumed to have higher transaction costs than either fixed-wage or fixed-rental agreements. Hence share contracts are observed when the benefits from risk-sharing outweigh the higher transaction costs. The higher the risk, the higher the gains from sharing and the higher the probability that a share contract will be chosen. The coexistence of all three types of contracts in agriculture would result from differences in risk among crops and/or regions as well as differences in risk aversion. Cheung's position has been criticised on the basis of both theoretical and empir-ical arguments. At the theoretical level, Stiglitz (1974) and"Reid (1976) showed that the same risk-dispersion effect could be achieved for both the landowner and tenant by a combination of fixed-rental and fixed-wage agreements. In other words, if each tenant works part of his time under a fixed-wage arrangement and another part of his time under a fixed-rent contract, and the landowner rents out part of his land under a fixed rent and cultivates another part using hired labour, they both obtain the same For insightful surveys of this literature, that has mostly been developed in order to explain the existence of sharecropping, see for example Binswanger and Rosenzweig (1984), Jaynes (1984) and Singh(1987). 11 reduction in risk than if they used a single share contract. In these circumstances, given the assumed higher transaction costs of the share contract, it would not be chosen. Since then, Newbery and Stiglitz (1979) have argued that with additional independent sources of risk, or with economies of scale in agricultural production such that subdividing lots becomes inefficient, sharecropping might again be chosen as an insurance policy. However, Jaynes (1984) and Eswaran and Kotwal (1985) point out that the assumption of higher transaction costs for the share contract does not hold when one takes supervision costs into account. Thus the argument of a trade-off be-tween transaction costs and risk-sharing lacks credibility. Assuming equal transaction costs for all three contracts, sharecropping becomes optimal whenever some risk is involved just as long as it cannot be spread through a mixture of fixed-wage and rent contracts. That is unless some of the agents are risk neutral in which case they should bear all the risk. In other words we would observe fixed-rental (wage) agreements only when the landlord (tenant) is risk neutral. The amount of "riskiness" would have no effect on the contractual choice except in so far as one would go from certainty to uncertainty. At the empirical level, the evidence is inconclusive. Cheung (1969), Higgs (1973), Huang (1974), and to some extent Bardhan (1977) present results that are consistent with a positive relationship between risk and the use of sharecropping. On the other hand, Rao (1971), Reid (1973), and Chao (1983) present evidence to the contrary. In the context of franchising, one can imagine that the uncertainty of demand might also make some form of risk sharing appealing. The fact that the "franchise way" can reduce the riskiness of going into business for the franchisee is advanced as a major reason for its popularity. A recognizable name which decreases the uncertainty of demand is usually a central part of the franchise package. But the franchisee could pay a fixed lump-sum fee for this input. To justify the use of a share contract, the risk sharing argument has to be that the franchisor and the franchisee want to spread the remaining risk. With a risk-neutral franchisor and a risk-averse franchisee, and given that each party is assumed to be perfectly able to observe the other's behavior, the optimal contract would be a fixed-wage one, with the franchisor insuring the franchisee com-pletely. Thus risk aversion by franchisees is not sufficient for the emergence of share contracts. Both parties must be risk-averse. Then, assuming they are not able to spread risk by mixing fixed-wage and fixed-rental agreements, they both benefit from 12 the insurance that arises from the use of a share contract. Stiglitz (1974), for example, develops this argument. Similarly, in his survey of the principal-agent lit-erature, Rees (1985) provides a discussion of the optimal riskrsharing argument for share contracts. There are many problems with this type of explanation in the context of fran-chising. First, one has to wonder about how reasonable it is to assume franchisors are risk-averse. These are often public firms that have access to financial markets. Only those franchisors that do not have such access, for example small firms with a sin-gle owner-operator, could reasonably be assumed to be risk-averse. Thus one would expect only small firms to use share contracts, while large public firms would choose to vertically integrate, i.e. offer only fixed-wage contracts to downstream operators. But this is not the case. Second, even if one accepts the notion that franchisors and franchisees are all risk-averse, then share contracts would be chosen over fixed-wage and fixed-rent contracts as soon as erg > 0. Since there is always some amount of uncertainty present, share contracts should arise most of the time. Thus these models do not really explain the coexistence of the three types of contracts. In particular, they can not explain why franchisors generally mix franchise and fixed-wage contracts. If franchisees facing a particular franchisor all have identical attitudes with respect to risk, the optimal strategy for the franchisor will be to use a single share contract with all franchisees rather than to spread risk through a combination of fixed-wage and share arrangements. This can be shown straightforwardly. With perfect information the franchisor can enforce the level of k he prefers, Given a known level of T , demand can be rewritten as Xi=x + 6i, i = l , . . . , n (2.4) assuming l± = I* so that f(T,l*) = x is the same for all outlets. Also at this point, Ci(lf) is known and fixed. Hence assume the costs of distribution, including the cost ^ In a survey published in 1980 by Entrepreneur Magazine, 35% of the franchisors who answered specifically indicated they did not allow absentee-ownership of their franchised outlets. In Venture's "Franchisor 100", 38 out of the 100 franchisors reviewed indicated that the franchise had to be the owner's main job. Thus it is true that franchisees are not always able to mix contracts. 13 of providing I*, are given by C{x + Bi) + Ci{V) = c-{x + 6l) + FC (2.5) where c is the constant marginal cost of distribution, and FC represents plus any other fixed costs of distribution. With perfect information, the franchisor knows c and FC so the profits given in (2.2) and (2.3) is rewritten as n n = 53[a-(p-c)( a ! + fli)-rF]-CT (2.6) *i = {1 - a){p - c){x + 6i) - F - FC (2.7) where a represents a royalty rate on "gross" profits whereas r was based on sales.^ Competition among potential franchisees should ensure that they get only their opportunity level of utility, u. In the absence of risk, this would imply zero profits, so that the franchisor would demand a fixed fee FQ = (1 — a)(p — c)x — FC. With some positive amount of risk, the fixed fee F will have to be smaller than FQ. In fact, (Fo — F) will measure the risk premium that the franchisee requires in order to be compensated for the variance in his revenue. This premium can be written as:1*' Pi = |-Var7ri (2.8) where p is the coefficient of absolute risk aversion which is assumed to be the same for all franchisees, and VarTTi = Var[(l - a)(p - c)(x + 0;) - F - FC} (2.9) = (1 - a)2(p - c)2*l (2.10) Since this variance is the same for all franchised outlets, so is the risk premium: P = P-.{l-a)\p~cfal (2.11) ^ Keeping a sales royalty formulation would complicate the algebra without providing more insight. Basically, with royalties on sales, the franchisee bears the burden of the variance in costs on his own. The risk sharing argument cannot justify this. 1 6 See Deaton and Muelbauer (1980), p. 399. 14 Assume that the franchisor chooses to mix contracts, i.e. he vertically integrates with Tii outlets and uses franchise contracts with the others. His total profits are then ni n = £ [ ( p - c ) ( * + * 1 - ) - F ' < 7 ] + £ [a(P-c)(x + 6j) + F}-CT. (2.12) ,7=711+1 t=l For any amount of risk he will have to bear, the franchisor will choose a and n\ so as to maximize his expected profits, where17 E{U) = raj [(p - c)x - FC] + (n - n i ) [ a ( p - c)x + F] - C T But F = (1 - a)(p - c)x -FC - P , so that (2.13) E(U) = m[(p- c)x - FC} + {n - ni)[(p - c)x - F C - P} - C T (2.14) = n[{p-c)x-FC}-(n-n1)P-CT- (2.15) The variance of the franchisor's profit is given by T i l Var(H) = " c)2*} + £ (a2(p - c)2<rj) i=l j=ni+l = (P ~ C)2<T9 [nl + a2in ~ nl)) (2.16) (2.17) using the assumption that the 6{ are independently distributed. Along any constant variance curve for the franchisor there is a trade-off between the number of company-operated outlets, n i , a * id the share parameter, a. Hence, da dn-\ _ - d V a r n / d n ! V « n W d V a r l l / d a ( I - " 2 ) 2a{n — ni) (2.18) Consequently, dEU dn-i _ dEU. dEU da Varn=<r2 = -fa{p-c?crl{l-a? (2.19) (2.20) Note that CT here is assumed to be unaffected by the proportion of franchised versus company-operated stores, i.e. the firm's choice of n.j. It can be allowed to depend on the total number of outlets n but since that is fixed a priori, CT would be as well. This is reasonable assumption to make here since perfect information implies there are no supervision costs included in CT • 15 which is < 0 for all a s.t. 0 < a < 1. For a given amount of risk borne by the franchisor, his profits are monotonically decreasing in n\. Thus the franchisor will choose rii = 0, i.e. he will not mix contracts but will franchise all his outlets.1** This is not surprising. Given that all retailers are identical in the sense that they all have the same aversion to risk, treating some of them differently by offering them a wage contract and transfering more risk onto the others cannot be optimal. Clearly, the fact that the optimal strategy for the manufacturer is to use strictly share contracts remains true whatever amount of risk is present at the demand level. As long as cr^ ^ 0, it is worth it for the franchisor to use share contracts only. Thus, as noted previously, risk-sharing arguments are unsatisfactory when it comes to explaining the coexistence of the different types of contracts. Given that most franchisors actually do mix fixed-wage and franchise contracts, one could say that at a crude level, risk-sharing models of franchising are rejected by the data. And indeed, it is necessary to extend the model in order to make any sense of this mixing of contracts in the context of risk-sharing. First, one must allow heterogeneity in the preferences of potential franchisees. But this is not sufficient. From the theory, we know that with such heterogeneity, the franchisor will find it in his best interest to use only share contracts again, except that he will use a different share contract, i.e. a different r, for each type of franchisee he faces. Thus it is necessary to introduce an additional constraint on the franchisor's behavior in order to get the result that he will mix fixed-wage and franchise contracts. That constraint is that each franchisor uses only one share contract, as they actually do. 1 9 In these circumstances, for those potential franchisees that are very risk-averse, the franchisor will prefer to use his fixed-wage contract, and for those that are not so risk-averse, he will opt for his franchise contract. How many stores will be franchised or company-operated will depend on his choice of r and on the distribution of the risk-In his paper, Martin (1987) affirms that "portfolio choice theory would imply that a mixture of operations should dominate either 100 percent franchising or 100 percent company opera-tion"^.1]. This is incorrect. Since the franchisor controls the share parameter, he can and should alter the amount of risk he faces by modifying this parameter. As long as the franchisor and all his franchisees are risk-averse, share contracts with all of them will be preferable to a mixture of fixed-wage and share contracts on the basis of a pure risk-sharing argument. The optimal share parameter may vary from one franchisee to another if they do not all have the same risk-aversion parameter. 1 9 The question that arises next of course is why franchisors use the same share contract with all of their franchisees. This will be discussed further later on. 16 aversion parameters in the population of potential franchisees. It would also depend on the amount of risk present at the outlet level. Given r, greater uncertainty would mean that the value of franchising to the franchisor would decrease if franchisees are generally more risk-averse than the franchisor is. This is because franchisees will require increases in risk-premia that will be greater than the risk-premium the franchisor would find sufficient to compensate him for the increased riskiness, given r. In particular, this will be true of the marginal franchisees, i.e. those that were indifferent between the fixed-wage and the franchise contract with a royalty rate r. The franchisor will now prefer to hire these marginal franchisees under the fixed-wage contract. Thus given r, the franchisor should rely on franchising less when faced with more risk in this extended version of the pure risk-sharing model. To summarize, it is clear that under the assumption that all potential franchisees have the same degree of aversion to risk, a franchisor's optimal policy, if he is also risk-averse, will be to offer an identical share contract to all his franchisees whenever some risk is present. Thus increases in risk would have no effect on the observed contract mix, all firms being fully franchised. Similarly, when some heterogeneity in the preferences of potential franchisees is allowed for, franchisors would still opt for a contractual structure that would entail only franchised outlets. However in this case, the optimal share parameter would vary across franchisees. Under this type of scenario, increases in riskiness would lead to a reduction in r for all the franchisees if they are more risk-averse than the franchisor.2^ But the firms would remain 100% franchised. It is only when franchisors face heterogeneity in the preferences of potential franchisees and are constrained to use a single share contract that one can expect to find franchisors mixing fixed-wage and share contracts. It is also only under such circumstances that changes in the amount of risk at the outlet level can affect the optimal contract mix. Given that he is less risk-averse that his franchisees, the franchisor will find it in his best interest to bear a greater proportion of the risk whenever the level of risk increases. This can be achieved as in the previous case by an increase in r. For a given r however, it could also result from a reduction in the proportion of franchised stores. See Stiglitz (1974). 17 3.2 One-Sided Hidden-Action Models This second class of models assumes risk-neutral franchisors and risk-averse franchisees. In the pure risk-sharing model, both parties were assumed to be able to observe the other's behavior perfectly. Here, T is fixed ex ante and franchisees can observe it, but franchisors can not observe how much l{ franchisees provide. And as was mentioned previously, given the presence of the random term in the demand equation, the franchisor cannot infer how much effort the franchisee puts into the provision of l{ from his knowledge of Xi and T. Thus, there is a moral-hazard or hidden-action problem on the franchisee side. The franchisor will want to provide his franchisee with a contract that gives him the most incentives to work, i.e. a fixed-rent contract. However, because the franchisee is risk-averse, and the franchisor risk neutral, it is not optimal to have the former bear all the fluctuations in sales that result from the presence of 9{, assuming ag > 0. Share contracts emerge in these circumstances as a compromise between the need to provide the franchisee with insurance and the need for a contract that gives him incentives to work. An example of this explanation for the existence of share contracts is found in Stiglitz (1974). This is also the structure of the basic principal-agent problem surveyed by Rees (1985).21 Clearly, this argument also relies on the idea that franchisees cannot mix contracts, i.e. work for wages part of the time, and under a fixed-rent contract for another part, as a means to obtain insurance. Note that these models do not allow for any monitoring on the part of the franchisor. Mathewson and Winter (1985) show that with any positive amount of non-noisy monitoring, a franchisor can use a forcing contract, i.e. impose such a large penalty on a franchisee caught cheating that truth telling becomes the dominant strategy for the franchisee. The incentive aspect of a share contract becomes worthless in this situation. The franchisor opts for a fixed-wage contract knowing he can enforce the level of U he chooses. This assumes however that these large penalties can be collected. If there exists a binding constraint on franchisee wealth, share contracts 91 There is nothing in the modelling of the principal-agent problem to suggest that the optimal contract should be a simple linear share contract. Often the optimal payment schedule is found to be non-linear and/or discontinuous. However, if one restricts the analysis to the set of linear contracts, as is done for example in Stiglitz (1974), share contracts are then found to be preferable to fixed-wage and rental contracts. Since franchisors use linear share contracts in general, I will concentrate on those here. 18 become optimal again, even in the presence of non-noisy monitoring. One problem with these models is that, like pure risk-sharing models, they lead to a contractual structure that entails only share contracts. Fixed-rent contracts are used only when there is no risk in production. Fixed-wage contracts are observed only in those cases where supervision is costless or there is no binding constraint on the franchisee's wealth. Thus the coexistence of the three types of contracts in reality is not explained by these models either. Nor is the tendency of franchisors to mix fixed-wage and franchise contracts. Consequently, one could say again that these models are rejected by the data on franchising. As in the previous section, if one allows for heterogeneity in franchisees's prefer-ences with respect to risk, and constrains franchisors to use a single franchise contract, then franchisors would mix contracts in the context of these models as well. Given r, in some cases franchisees would be so risk-averse that the benefit in terms of incen-tives that would arise from the use of the share contract would be offset by the high risk-premia they would require. Thus a franchisor would find it in his best interest to hire these people under a fixed-wage contract. In these circumstances, increases in risk given r would lead to more weight on the insurance component of the trade-off and should therefore lead to a greater reliance on company-operated stores. Again future references to one-sided hidden-action models should be taken as references to this modified version of the models. 3.3 Two-sided Hidden-Action Models The third and most interesting explanation of share contracts relies on market imperfections in general, and more specifically on "hidden-action" problems on the part of both parties. The advantage of these types of models is that not only can they explain why share contracts occur, but they can account for their coexistence with other contractual forms as well as for temporal and spatial variations in the contractual structure. Again, the models are based on Alchian and Demsetz's (1972) notion that when it is difficult to assess the marginal contribution of an input in a production process, and supervision is costly, one way to make sure the owner of the input has the incentives to provide the efficient level of this input is to make him or her a residual claimant. However, here, more than one agent is responsible for the provision of such inputs. One way to give them both incentives to cooperate and 19 self-monitor is to give them both a share of the output. Thus in such models sharing occurs strictly as a result of incentive problems. Preferences with respect to risk are not involved. In fact, both parties are assumed to be risk-neutral. An example of this view, as it applies to sharecropping, is found in Reid (1977). Eswaran and Kotwal (1985) have formalized this approach in the context of sharecropping. They assume that the agricultural production process requires two inputs for which markets are imperfect, namely supervision and management. Both inputs can be provided by either the tenant or the landlord, but the latter is more efficient in management and the former in supervision. Under a share arrangement, both parties will be allowed to specialize in the activity they are most efficient at. Un-der a fixed-wage contract, landlords must provide both supervision and management, whereas the tenant has to take care of both of these under a fixed-rental agreement. Eswaran and Kotwal then show that the landlord opts for a share contract whenever the gains from having each party specialize in the activity they are best at outweigh the losses that result from the disincentive effects associated with sharing. Which type of contract maximizes the landlord's profits depends on the importance of both super-vision and management in the production process. It also depends on the strength of the landlord and tenant's comparative advantages. Depending on the circumstances, an}r of the three types of contracts may be optimal in the sense that it maximizes the landlord's profits. The coexistence of the three types of contracts in this model is explained by the use of different technologies (i.e. various crops) that affect the importance of the two unmarketed inputs. Changes in the contractual structure occur as a result of technological change. What is interesting is that the model will hold for any two inputs for which markets are imperfect as long as they can only be made available in combination with their owner's time. Both Reid (1977) and Eswaran and Kotwal (1985) view sharecropping as a partnership and emphasize the need for ongoing cooperation between the tenant and the landlord as the basic reason for the use of share contracts. It is worth pointing out that the transactions-cost analysis proposed by Murrell (1983), while different in flavor from this type of model, is consistent with it. In both cases, one compares the properties of the three types of contracts under the constraints imposed by the production process and then determines which one is 20 best. In the case of franchising, assume, as was discussed previously, that the franchisor can not observe or infer the level of l{ provided by franchisees. But now the franchisees can not observe the value of T either (nor the value of any managerial assistance that the franchisor may provide). Rubin (1978) for example points out that the value of the tradename over the duration of the franchise contract depends non-trivially on the ex-post behavior of the franchisor, i.e. on how much effort he will put into advertising and in the monitoring of established franchisees to make sure they do not shirk on quality. Under fixed-rent contracts, and given incomplete contracting in advertising and monitoring, the franchisor would choose the ex-post level of these activities , and hence of T, to maximize the profits he would derive from future sales of franchises. In this maximization the franchisor need not consider the effect his choice of T will have on the performance of all previously established franchisees. Without a type of contract that offers the upstream firm with the right incentives, T will not be chosen optimally in the eyes of the franchisees. Unless they can get some form of assurance that the value of the tradename will be upheld, franchisees will not agree to a fixed-rent contract. Note that a franchisor can show he has incentives to maintain the value of the tradename either by operating a certain proportion of outlets himself, and/or by using share contracts. We now have a bilateral moral-hazard problem. In these circumstances, both the franchisees and the franchisors will need contracts that give them incentives to cooperate and self-monitor. As in Eswaran and Kotwal (1985), share contracts may be chosen by franchisors depending on the importance of l{, T, and possibly the managerial assistance given by franchisors in the downstream demand equation (or production process). This of course will vary according to the kind of product sold. For example, some types of goods require much more personalized services (computer sales for example) than others (fast food). This affects the importance of l{. The choice of contract will also depend on how costly it is to monitor l{ and T . 2 2 For example the buy backs of the early 70's, which occurred mainly in urban areas, could be due to the decrease in monitoring costs associated with an increase in the density of stores in.those areas. The trend towards more franchised units in recent years could be the result of increased distance between outlets which would increase If U can easily be monitored, then the franchisor becomes as efficient as the franchisee in providing it through a hired manager. 21 monitoring cost. Better established reputations could also explain this tendency to franchise more: they reduce the need for the franchisor to guarantee his performance by operating his own stores.2"' This kind of model has the advantage of being able to account for two stylized facts one finds in the context of franchising. The first is the very infrequent use of fixed-rent contracts. The second is the fact that franchisors only offer two out of the three types of contracts at any point in time. More precisely, they have a single franchise contract that is usually a share contract, although in some cases it is a fixed-rent one, and then they operate some of their stores directly under a fixed-wage contract. The argument in both cases hinges on the input of the franchisor, T. In Eswaran and Kotwal (1985), under a fixed-rent contract, the tenant must provide both unmar-keted inputs (supervision and management in their case). But here, the franchisee can never provide the tradename component of T. The franchisor can provide both li and T if he hires a manager for an outlet. In that case, a fixed-wage agreement is used. Or the franchisor provides T and the franchisee l{, in which case a share contract arises. Since T necessarily comes from the franchisor, as long as it is unob-servable, the contract must give the franchisor incentives to maintain its value. This makes fixed-rent contracts costly. Thus, these contracts will tend to be observed only when the provision of T can easily be monitored. Since that will not be the case in general, one will find mainly share and fixed-wage contracts in franchising. In addition, for a given franchise system, either T is or it is not easily monitored. If it isn't, this firm will have to design its franchise contract as a share contract, while if it is, its franchise contract should be a fixed-rent agreement. Thus we will find each franchisor using a franchise agreement that is either a share or a fixed-rent contract, but never both at the same time.2'* 23 24 The U.S. Department of Commerce establishes these trends in "Franchising in the Economy". A question that arises then is why would a firm that opts for a fixed-rent contract ever want to operate stores directly, i.e. under a fixed-wage contract, if its choice of contract reflects the fact that there are no incentive problems on the franchisor's side ? Do they in fact operate significantly less stores than other franchisors do? Results from Chapter 5 suggest that they do. 22 3.4 Capital-Market-Imperfection Arguments The traditional explanation put forward to explain the existence and the use of franchising has to do with franchisors facing a binding capital constraint, and resorting to franchising as a means of obtaining capital. For example, Caves and Murphy (1976) argue that before the tradename is well established or at times of rapid expansion, it is difficult for a franchisor to raise all the capital he needs.2^ Thus he recruits franchisees who invest their own capital in their outlet. Jaynes (1984) puts forth a similar argument in the case of cropsharing: "The landowner who finds a tenant with a plow and possibly his own team of bullocks may find that offering a share tenancy to this agent is a cheaper way of obtaining this capital than having to risk some portion of his own property as collateral for a loan that would allow him to lease or purchase all the necessary capital inputs himself and to hire wage labor." (p. 60) There are several difficulties with such an explanation for franchising. First, if the franchisor's need for capital was the main explanation for the existence of franchising, one should not observe well-established firms using franchising with their new outlets, except perhaps during rapid expansions. According to Caves and Murphy (1976), as the company matures, one would find a trend toward more company operation. It has also been suggested that if capital constraints were the main reason for the existence of franchising, franchisors should ultimately buy back all of their outlets.26 Franchising would simply be a transitional state. But while a tendency towards more company-owned outlets seems to have been observed in the early 70's, according to the U.S. Department of Commerce this trend has reversed itself since the mid 70's. A second problem with this explanation for franchising is that franchisors often provide financing to their franchisees.27 This of course is not consistent with an 2 ^ In the Ozanne and Hunt survey, this need for capital, and the fact that franchisees who own their businesses would manage the outlets better, were the two reasons generally given by franchisors for choosing this arrangement. The fact that 18% of them said they would ultimately prefer to have no franchised outlets is consistent with a binding capital constraint argument. 2 6 See for ex. Hunt (1973), Martin (1987), and Scott (1987). 2 7 In the Entrepreneur's 1986 survey, 228 out of the 968 franchisors who responded said they provide direct financing. 23 explanation for franchising that relies on a binding capital constraint for franchisors. Finally, as Rubin(1978) points out, investing in a single outlet is clearly much riskier than investing in a portfolio of shares from all outlets in a chain. Hence a risk-averse franchisee would require a higher return on his investment in a single outlet, implying that the franchisor could obtain cheaper capital by offering shares of all its outlets to its store managers. Brickley and Dark (1987) use the same argument. However, this ignores the possibility of moral hazard on the part of these managers. If there is an incentive problem at the downstream level, with a portfolio of shares from all stores, we get a standard prisoner's dilemma. Every retailer benefits only marginally from increasing his own /{, so they all set it too low. Knowing this, retailers might well demand a higher rate of return on portfolios, even if they are less risky, than they would for capital invested in a single store that they operate themselves. Hence it is possible that the upstream firm would benefit from cheaper capital through franchising. This would be especially valuable to those firms facing a binding capital constraint. Note however that the existence of a non-observable l{ is central to the argument. In general then, if it allows franchisors to get cheaper capital during periods of expansion, franchising may be more attractive than company-operation at those times and hence firms may use it more. But their need for capital is not necessary to get franchises. It is worth pointing out that an explanation of franchising based on franchisors' capital requirements should lead to fixed-rent contracts since that would guarantee more capital to franchisors faster. In addition, so far, the incentive problems are only on the franchisees' side, and by themselves, those lead to fixed-rent contracts. For share contracts to emerge, incentive problems on both sides, or risk-aversion of the franchisees combined with non-observable are still necessary. With respect to the terms of the share contract, a binding capital constraint for the franchisor should increase the probability that the contract stipulates a large fixed payment and low royalties. 3.5 Self-Selection and Screening Models The fifth class of explanations for share contracts is based on asymmetric infor-mation concerning labor quality and the resulting need for the principal to screen 24 potential agents. The latter have differing abilities with respect to the provision of their input. Thus there is a hidden characteristic problem. Examples of such mod-els in the case of sharecropping include Hallagan (1978) and Newbery and Stiglitz (1979). The existence of different types of contracts in these models forces down-stream operators to reveal information about themselves. The most capable choose a fixed-rent contract and the least capable, a wage contract. Those in between opt for a share contract. An advantage of this kind of model is that it naturally leads to the coexistence of the three types of contracts. In the context of franchising, this would imply that store operators have differ-ential abilities with respect to the provision of Assume that their ability level is in the closed interval [0,L]. If each operator devotes all his or her time to this activity, and there are no additional costs associated with l{, the amount of /{ will vary directly with these people's ability level. Assuming that the franchisees are risk-neutral, they care about their expected revenue. If C(/(T,/;)) — c • f(T,l{) -f FC, the franchisee's expected revenue under the three types of contracts would be given by E{Rf) = w (2.21) E(R?) = {p-c)f(T,li)-FC-FR (2.22) E(R?) = (1 - a)(p - c)f(T,li) - F C - F s (2.23) where w stands for fixed-wage, R stands for fixed-rent and S stands for share contract. Naturally, the fixed fee associated with the fixed-rent contract would be larger than that of the share contract, i.e. FR > Fs- The resulting revenue functions, assuming that / ( T , 0) = 0, are illustrated in Figure l . 2 8 Both Rf and Rf are concave in for all 0 < a < 1, given that / is concave in Zi. Individuals with the ability to provide l{ < A will opt for a fixed-wage contract since their revenues under that scheme are greater than under any of the other arrangements. With U = A, they are indifferent between a fixed-wage or a share arrangement. With the ability to give A < l{ < B, an individual prefers a share contract and naturally for l{ > B, one chooses a fixed rent contract. A more realistic screening model in the case of franchising would have to take into account the unobservability of the value of the franchise as well as that of the potential 2 8 See Hallagan (1978). 25 l<=0 l.=L Figure 1: Contractual Choices as Self-selection Mechanisms franchisee. Most members of this "industry" emphasize the need for prospective franchisees to shop around before getting involved with a particular franchisor. The information asymmetry is two-sided. Allen (1982) proposes a model where both land quality and tenants' abilities are unobservable. An adaptation of this model might be more appropriate for the study of franchising. While such self-selection models are interesting, it is difficult to believe that they capture the whole story. For one thing, these models, like those discussed in the previous sections, imply that each franchisor should use a variety of share contracts. Indeed, one would think that franchisors should offer a whole array of contracts, i.e. many share contracts, from those with low royalty rates and high franchise fees all the way to contracts with high royalty rates and low franchise fees, in addition to fixed-wage and fixed-rent contracts. This would allow downstream operators to self-select as finely as possible. Yet, as was mentioned earlier, franchisors offer two types of contracts only, a franchise and a fixed-wage contract. 26 Given the average length of franchise contracts, it is also difficult to believe that a franchisor's goal in offering them is to screen potential franchisees. According to the U.S. Department of Commerce, franchise contracts last for almost 15 years on average. If the various contracts were used so that franchisees would reveal information about their types, one would not expect them to last for such long periods of time. Another problem with self-selection models is that the proportion of franchised stores in a given chain is not the result of a decision the franchisor makes. Instead, it is determined exogenously and depends simply on the distribution of abilities in the population and the random draw of "potential franchisees" of this firm. Assuming that n is large enough and that ability levels are uniformly distributed on [0,L], one would find a number of fixed-wage contracts proportional to the distance OA in the diagram, while the quantity of share contracts would be proportional to AB and that of fixed-rent contracts, to BL. The only thing that could affect these proportions across firms would be differences in technology, i.e. in general, differences in the type of activity the firms are involved in. These would affect the importance of k in the production process and thus shift the revenue curves depicted in Figure 1. But the direction in which the proportions of the three types of contracts would vary given a change in the importance of l{ is undetermined. Self-selection models are not very satisfactory when it comes to explaining tem-poral and spatial variations in the contractual mix. Similarly, they can not explain variations in the actual terms of the share contract since these are exogenously given in such models. In the empirical analysis, both the contract mix (i.e. the proportion of franchised stores) and the terms of the share contract (royalty rates and franchise fees) are assumed to be chosen optimally by each franchisor. Thus the empirical work concentrates on the first four types of models and not on the latter. 4. C o n c l u s i o n The aim of this chapter was to provide a general description of franchising. It was also meant to establish the fact that models developed to explain the existence of sharecropping have natural extensions to other types of share contracts, notably fran-chising. Because the models were developed in the context of single principal-agent pairs, they all lead to a single dominant contract, a result that remains unchanged if these models are modified to incorporate multiple principal-agent pairs. As was 27 noted, this was the case for the standard versions of the pure risk-sharing and the one-sided hidden-action models. It was necessary to extend the models in order to get the result that franchisors can operate some of their outlets directly and franchise others. In addition, in theoretical work authors tend to isolate those factors they feel are most relevant to the problem at hand and they abstract from the others. This is generally done for tractability. But in reality, there is no reason for a single explanation of share contracts to dominate all others. While incentive issues might be very important, how much a firm decides to rely on franchising may also depend on its financial constraints and on the amount of uncertainty it faces. The various explanations of the existence of share contracts are not mutually exclusive and thus should not be treated as such empirically. By allowing interactions among the various factors suggested by the theories, it is possible to develop a framework to analyse firms' propensity to franchise. Optimal risk-sharing models and principal-agent models based on "hidden action" rather than "hidden characteristics" can all be used to explain variations in the contract mix chosen by franchisors. So can imperfect-capital-market arguments. Similarly, all of these have implications concerning the terms of the share contract. They are therefore the models on which the empirical model and analysis will be based. In the next Chapter, I provide a brief review of the existing empirical literature on share contracts, followed by a description of the empirical model. 28 C H A P T E R III A n Empirical Model of Franchising 1. Introduction The purpose of this Chapter is to develop an empirical model for franchising that will allow testing of some implications of the theoretical models just discussed. In the next section, I briefly review some of the existing empirical literature on share contracts. I begin with those empirical papers that have been written in the area of sharecropping. Then I discuss the few papers that have examined some aspects of franchising. . In Section 3, I develop a model aimed at explaining firms' tendency to use franchising which combines the first four types of models discussed in Chapter 2. Next, the model is extended to take into account the fact that franchisors not only choose the contract mix, but also the terms of the franchise contract, i.e. the royalty rate and franchise fee. Section 4 contains a few concluding remarks. 2. A n Overview of the Empirical Literature on Share Contracts 2.1 The Empirical Literature on Sharecropping There exists a relatively large body of empirical literature on the subject of sharecropping. One issue that has received a certain amount of attention at the empirical level is that of the allocative efficiency of sharecropping contracts. In other words, do they lead to different input/output intensities than those that are observed under fixed-rent and fixed-wage contracts?1 Since I am interested in those circumstances under which share contracts arise as the more efficient, albeit second best, arrangement, I will not discuss this literature. A number of other empirical papers have focused on the contract mix and the contractual terms in agriculture. The aim of the following discussion is to provide a general idea of the questions that authors have addressed in such studies as well 1 See for example Shaban (1987) and the references therein. 29 as the problems they encountered. In general, it is worth mentioning that authors have not tried to discriminate among precisely specified models in their empirical work. They have tended to insist on the complementarity of various explanations of sharecropping, rather than seeing them as mutually exclusive models. Thus they mainly searched for patterns that would be consistent with these explanations. The first question that has been addressed empirically in this area concerns the effect of risk on landlords' tendency to use sharecropping. According to the argument suggested by Cheung (1969), there should be a positive relationship between the extent of sharecropping, as a proportion of all contract types, and the amount of exogenous risk in production. Under the pure risk-sharing argument presented in this thesis, it was noted that share contracts should prevail irrespective of the amount of risk present when all potential tenants have the same preferences with respect to risk, and/or when landlords are able to modify r in those cases where tenants' attitudes vis-a-vis risk are allowed to differ. Thus no systematic positive relation between risk and sharecropping is to be expected in either of these two cases. If potential tenants' attitudes with respect to risk are heterogeneous, but landlords are constrained to use a single share contract, then increases in risk, assuming the landlord is less risk-averse than the tenant, would increase the tenant's relative insurance need. This would lead to a shift away from fixed-rent contracts and toward more share and more fixed-wage contracts. Reductions in uncertainty would mean less fixed-wage contracts, and more fixed-rent and more share contracts. Thus no clear correlation, either positive or negative, should be expected between the extent of sharecropping as a proportion of all three types of contracts and the amount of risk in this case either. As was mentioned in Chapter 1, the empirical evidence on the subject remains inconclusive. Cheung (1969), Higgs (1973), Huang (1974), and Bardhan (1977) present evidence of a positive correlation between risk and the use of sharecropping, while Rao (1971), Reid (1973) and Chao (1983) for example find no such support for Cheung's hypothesis. One must however exercise some caution in interpreting these results. Early studies on the effect of risk, for example, Cheung (1969), Higgs (1973), Reid (1973) and Huang (1974), compared the extent of sharecropping in some geographical regions or for different crops to some measure of riskiness, either the variance or the coefficient 30 of variation of yields. Rao (1971) used the variance of profits rather than that of yields. Given the paucity of the data, they could only observe these aggregate measures and were not able to control for other possible effects. Hence, as Huang (1974) points out, Rao's measure of riskiness is more likely to embody "risk associated with decision-making" by the tenant and not just exogenous risk. Thus the negative correlation he observes between uncertainty and the extent of sharecropping as opposed to fixed-rent contracts should not necessarily be interpreted in the context of an explanation for share contracts based on risk. Rather it provides evidence that the use of fixed-rent contracts as opposed to sharecropping increases when the managerial talents of tenants become more important. This is clearly consistent with those models that involve some hidden action problem on the part of tenants. Under the risk-sharing argument used in this thesis, if one concentrates on the extent of share contracts as a proportion of fixed-rent and share arrangements only, as Higgs (1973) does, increases in risk should lead to a greater incidence of share contracts. And this is indeed what he finds. Reid's (1973) result of no significant correlation is consistent with this given that he observes acreage under sharecropping as a proportion of total acreage. For the same reason, Bardhan's (1977) result is also consistent with the risk-sharing argument used here. In addition, contrary to earlier studies where authors were unable to control for other effects due to the lack of data, Bardhan (1977) controls for the labor-intensity of crops, the wage level of laborers, and the extent of credit provided by landlords to tenants in his regressions. Chao's (1983) result, hoAvever, which is based on micro-level data, contradicts both Cheung's hypothesis and the pure risk-sharing model described in this thesis. He obtains his result by calculating the average coefficient of variation of rents actually collected over 20 to 30 years by landlords who had two different types of contracts that they used simultaneously on different plots of land. Those were "pure sharecropping" contracts, where landlords and tenants each received 50% of the yields, and "modified sharecropping". The latter was a transitive type of contract between a pure-share and a fixed-rent contract since it involved a nominal rent, fixed at the level of 50% of the yields in the good years, which the landlord could reduce during "less-than-perfect" years. Chao (1983) found that these average coefficients of variation were not different enough to suggest that there existed differences in risk between the two types of plots. 31 Other issues that have been addressed in the empirical literature on sharecrop-ping include the effect of changes in the tenants' exogenous wage rate, the effect of technological change and the effect of supervision costs on contractual choice. In terms of the effect of the exogenous wage rate, as Binswanger and Rosenzweig (1984) point out, there is some consensus to the effect that declining real wages should lead to changes in the contract mix and/or the contract terms that yield either a lower income, or more risk, or more work for the tenants. This is true given the production technology, and the amount of labor used. Thus Clay (1976) reports a case in which declining real wages lead to a shift from harvest share payments for the tenants to payment in cash. There was a shift away from share contracts toward fixed-wage contract. More importantly, this shift resulted in lower real incomes for the tenants. However, in his study of the evolution of the contractual structure in China, Chao (1983) finds that decreases in real wages resulted in a shift from sharecropping to fixed-rent contracts. In his general equilibrium framework, this is explained by the use of more labor-intensive techniques in agriculture as wages fall. More labor-intensive technologies imply more supervision costs, and thus a greater need to provide tenants with contracts that incite them to work, hence a greater tendency to use fixed-rent contracts. Thus, holding the technology fixed, decreases in real wages tend to shift contractual choices toward more fixed-wage contracts. But if one allows for changes in the method of production, the resulting increases in supervision costs would lead to a greater reliance on fixed-rent contracts. With respect to technological change, the empirical consensus seems to be that land-augmenting technological change such as irrigation, and the use of High Yielding Varieties of grains (HYV), generally lead to a shift away from fixed-rent contracts and toward fixed-wage ones. Thus Rao (1975) reports that as a result of the introduction of H Y V and other technological advances, the total area under tenancy, and especially that under share tenancy, declined substantially in the sixties in India. Similarly, Day (1967) finds that sharecropping contracts were replaced by fixed-wage contracts in the post-bellum South as a result of mechanization. These results (as well as Chao's result) are consistent with an explanation of contractual choices based on transactions costs in general, and on supervision costs in particular. By reducing the need for labor and/or labor supervision relative to other inputs such as know-how and capital, mechanization and land-augmenting technological change reduce the importance of supervision costs. Thus they diminish 32 the need to provide tenants with contracts that give them incentives to work: This explanation for contractual choices can also account for the observed tendency of absentee landlords to lease out their land under fixed-rent contracts.2 Since they are not in a position to supervise their tenants, they must provide them with contracts that give them incentives to work. Some empirical studies on sharecropping have focused on the effect of supervision costs directly. These include Alston and Higgs (1982), Alston, Datta and Nugent (1984), and Datta, O'Hara and Nugent (1985). In the first two, the authors use a data set based on a survey of 22 cotton plantations in Georgia undertaken by the U.S. Census Bureau in 1911. Because of the homogeneous nature of the sample, risk and technology can be assumed to be constant across these plantations. Since information about the number of supervisors was collected, the effect of supervision costs can be tested directly. Their results, although inconclusive in some cases, are consistent with the hypothesized relations between supervision costs and contractual choice. In general then, the empirical studies give some support to the notion that as the role of the tenant becomes more important, or harder to supervise, there is a shift towards contracts that give more incentives to work to the tenant. As is illustrated by the discussion thus far, authors of empirical work on share-cropping have focused on explaining the contract mix. The contractual design, i.e. the determination of the sharing rule, has received little attention in the empirical literature. Yet the theoretical models have implications for the terms of the share contract as well as for the contractual structure, if not more so. Because they focus on a single principal-agent pair, most of these models must concentrate of the share parameter as the principal's or the landlord's main control variable. Thus this issue should also be addressed empirically. A major reason for the lack of empirical work on the sharing rule is that it has been found to remain relatively constant, at 50% of gross output, within and across regions, as well as through time. Chao (1983), for example, reports that the division of output was done on a 50-50 basis for more than 2000 years in China, and that, despite a steady increase in population and the corresponding decrease in real wages. In fact, this constancy of the share parameter around one half is one of the four stylized facts Newbery and Stiglitz (1979) suggested theories concerning sharecropping should be See for example Jodha (1984). 33 able to explain. But if the terms of the share contract do not vary, one can not find factors that determine them. Hence the lack of attention in the empirical literature. But while 50-50 splits are common, contract terms do vary within some regions. With the advent of more detailed data sets, and more precisely information on individual plots or contracts, authors have begun to observe these differences. For example, Roumasset (1984) observes that landowners' shares vary substantially in the Philippines. He also finds that these shares are positively correlated with land quality, and negatively related to the relative price of labour. In semiarid tropical India, Jodha (1984) found that the terms of tenancy contracts were fairly flexible. The tenant's share varied between 50 to 75 %. In addition, the rules governing the sharing of input costs and the allocation of by-products varied widely. Thus the common 50-50 split of output can hide many differences in the final shares attributable to each party. Because detailed information concerning individual contract terms is not easy to come by, there are only a few such studies. Similarly, with respect to the contract mix, most of the work has been done on the basis of aggregate data due to the unavailability of micro-level data. Wrhile some micro data sets have now become available (see for example Alston and Higgs (1982) and Chao (1983)), they remain too rare (and often too small) to permit a thorough empirical investigation of the abundant theoretical literature on sharecropping. It therefore appears natural to broaden the scope of these theories to other types of share contracts, as was done in Chapter 2 in the case of franchising, and then test them within these new contexts. If share contracts do exist for similar reasons in different sectors of the economy, these tests should prove useful in enhancing our understanding of these types of contracts in general. While this approach has not been used before, some empirical work has been done on franchising. I now turn to a discussion of these papers. 2.2 Some Empirical Papers on Franchising The main issue that has been addressed in the empirical literature on franchising is whether or not this is just a transitional stage for firms on their way to becoming fully-vertically-integrated chains. This question was first posed by Oxenfeldt and Kelly (1969). In a sense, this has also been an issue in the sharecropping literature. Many have argued that as markets develop, sharecropping becomes unnecessary, and so with time, it should disappear. 34 Hunt (1973) first examined this issue at an empirical level. Based on the data collected in the Ozanne and Hunt (1971) survey, he was able to show that at an aggre-gate level, the proportion of company-operated outlets in fast-food chains increased between 1960 and 1971. In addition, on the basis of chi-square tests on disaggregated data, he found that large chains were more likely than small ones to be increasing the proportion of units they operate directly. He found the same kind of pattern relative to the age of the franchise, i.e. older chains tended to increase their proportion of company-operated stores. Caves and Murphy (1976) also examined this issue. Using aggregate data pub-lished in Franchising in the Economy, they found a trend toward more company-ownership in franchising as a whole. They also noted that reversions to company-ownership, i.e. franchisors buying back franchised outlets to operate them directly, were important numerically. Finally, two recent empirical papers on franchising also investigate the long-term trend in the proportion of outlets franchisors choose to operate. These are Martin (1987) and Scott (1987). Martin (1987) uses data on 772 franchisors classified among sixteen different sectors. Within each of these sixteen sectors, using a logit model, he calculates a franchising lifecycle function, i.e. how the proportion of company-owned outlets varies as a function of the number of years since the firms began franchising. Using these, he derives long-run proportions of company operated outlets for firms in each sector. He then uses these to test alternative explanations for franchising, including the capital-market-imperfection argument. His test in this case is based on the notion that ultimately, this explanation leads to 100% company-operated chains. On the other hand, he suggests that if franchising is used because it provides a better incentive system, then mature chains should be 100% franchised. Finally, he contends that under a portfolio hypothesis, firms should choose an interior solution. As was shown in Chapter 2 however, this statement is mistaken.^ His results indicate that none of the proportions of company-owned outlets converge to one. Thus he rejects the capital-market-imperfection argument and the notion that franchising is a transitory phenomenon. He can not reject his other two hypotheses, i.e. the notions that the chains will ultimately become lOOor that they will converge to some interior solution. ^ See the discussion pp. 13 to 15. 35 Scott (1987) also argues that under the capital-market explanation, firms should reduce their reliance on franchise contracts as they grow and become more mature, but that if franchising is used because it is an efficient incentive device, there should be no such trend toward company-operation as the firm matures. He uses aggregate data from the U.S. Department of Commerce to estimate, using generalized least squares, how the proportion of franchised units varies across sectors as a function of, among other things, the average age and size (number of outlets) of firms in the industry, as well as the average start-up cash and the total investment required. He finds positive coefficients for both the age and size variables (although the coefficient of size was insignificant), and negative signs for the cash and investment requirement, all of which go against the capital-market explanation for franchising. Three other empirical studies, i.e. Goldberg (1982), Brickley and Dark (1987) and Norton (1988), have examined conditions under which firms choose to use fran-chising. They address this problem in a transactions or an agency-costs framework. In his thesis, Goldberg (1982) uses data on 25 companies involved in the restau-rant business between 1960 and 1979.4 Using a logit model, he relates the proportion of company-operated stores to two measures of the extent of the externality outlets impose on each other in a chain by free-riding on the value of the trademark. The higher the value of the tradename, the higher the gains from free-riding and the more costly franchising should be. These two measures of the value of the tradename, i.e. the company's age and the number of outlets, are expected to increase the proportion of company-operated stores. He finds that firms' tendency to own stores increases with their age given the number of outlets, but decreases with the number of stores given the age. Thus depending on which measure of "maturity" is preferred, the re-sults are consistent or not with the hypothesis he proposed. He also finds that firms tend to increase the proportion of stores they operate themselves after going public, a result which is consistent with a capital-shortage explanation for franchising. He how-ever cautions against this interpretation stating that the causality is unclear in that case. Firm-specific effects are found to be an important determinant of franchisors' tendency to own stores. In addition, a dummy variable that was given a value of one when the restaurant was classified as a "coffee shop" and zero otherwise had a signif-icant positive effect on the proportion of company-owned stores. He interprets this 4 For those companies in the sample that were not in existence in 1960, the data start later. In no cases did it start later than 1969. 36 as an indication that when repeat sales are rare, the externality outlets can impose on each other is greater, and thus company operation becomes preferable. Finally he finds some support for the notion that increased government legislation affects firms' choices with respect to franchising. Norton's (1988) analysis of the benefits of franchising focuses on the differences in agency costs between franchised and vertically-integrated stores. He tests his hypotheses about the circumstances that favor franchising by estimating a logit model where the dependent variable measures the incidence of franchising in three sectors of the U.S. economy, notably Restaurants and Lunchrooms, Refreshment Places and Motels and Tourists Courts. The data he uses are those published in the 1977 U.S. Census of Retailing and in the 1977 Census of Service Industries on an aggregate basis for each State in 1977. There is however a problem with the way he calculates his dependent variable. It is described as the number of franchise holders as a proportion of the total number of establishments in the sector in each State. Given this dependent variable, many of the theoretical arguments he presents have no a priori reason to hold. This is because factors that favor franchising in his discussion also favor the existence and the number of independent retailers in an industry. He argues for example that increased labor intensity leads to greater monitoring costs, so that firms will find it in their best interest to hire someone who specializes in monitoring in each outlet. To prevent this monitor from shirking, they will want to make his revenues depend on his performance. This is what a franchise contract does. Thus he expects to find a higher proportion of franchised outlets in those industries and States where the ratio of employees to sales is greater. But independent retailers also qualify as an efficient way to organize production in these circumstances. If their number grows faster than that of franchised outlets, then the proportion of franchised outlets overall in the industry in the State would be reduced.^ Consequently, the interpretation he gives of his empirical results is problematic. Finally, Brickley and Dark (1987) concentrate their empirical work on the effect of monitoring costs, repeat business and initial capital requirements on franchisors' decision to operate or franchise each individual outlet. They obtained information from 36 franchisors concerning, for each outlet, whether it is franchised or owned, and ^ While most of the arguments he gives would tend to favor both franchising and independent operators simultaneously, his argument concerning the value of the trademark would tend to favor chains in general, i.e. both franchised and vertically integrated ones. 37 the distance to the nearest headquarters. They found that on average, outlets that were controlled by the franchisor were significantly closer to headquarters than those that were franchised. If distance increases monitoring costs, this result supports the notion that firms opt for franchise contracts, i.e. contracts that provide more incentives to the outlet operator and thus reduce the need for monitoring, in those cases where monitoring costs are high. The authors also partition a larger sample of firms (143) into two groups on the basis of whether or not they operate in an industry characterized by repeat business. They define nonrepeat sectors as those that serve mainly transient populations, i.e. Restaurants, which are mainly fast-food outlets, Car Rental stores, and finally Motels and Hotels. They find that the average proportion of franchised stores is significantly greater for the group of firms that operate in the repeat business sectors than for the other group, thus confirming their hypothesis. In addition, they test the notion that due to inefficient risk-bearing, franchising should be used less when the initial capital requirement is high. They also find some support for this argument. In general, these empirical studies on franchising, like those that have been done on sharecropping, have been constrained by a lack of available data. In most cases, authors have used aggregate data, and in those cases where information concerning individual franchisors was collected, the number of firms and/or the type of infor-mation was quite limited. In some cases, the hypotheses being tested, and the way in which they related to the theoretical literature, were not well defined. Finally, these papers have all addressed the question of the contractual mix. None has inves-tigated the terms of the franchise contract. In the next section, I present a model of franchising, based on profit maximisation, that relates franchisors' choices concerning the contract mix and the terms of the franchise contract to the theoretical literature discussed in Chapter 2. 3. The Empirical Model In this section, I develop the empirical model that will allow me to test predictions from the first four types of theories of share contracts presented in Chapter 2. This is done first with respect to firms' decisions about the contractual mix, and is then extended to take into account their choices concerning the design of the franchise contract. 38 In developing their models, theorists can choose those aspects of a problem they feel are most relevant, and abstract from other considerations they think are not as important. This makes their problems more tractable. Empirical work can contribute to this process by indicating which factors may be relevant and what questions deserve some further attention. As was seen in the previous chapter, many different types of models of share contracts are found in the theoretical literature. They all point to different factors that should influence the incidence of share contracts. As Singh (1987) points out in his review of the literature on sharecropping: "Having gone through the various models (...), we do not "pick a winner". This is because we do not think that there is a single explanation, no matter how ingenious or complicated, of the exis-tence of share contracts or sharecropping." (p. 4) Similarly, the goal of this empirical analysis is not to "pick a winner". Rather its purpose is to examine whether the various factors theorists have examined as potential reasons for the existence of share contracts seem to be relevant when it comes to explaining the use of a particular type of share contract, namely the franchise contract. There is no reason to expect these explanations to be mutually exclusive. Thus whereas in the theoretical literature these factors are treated in isolation, I want to allow interactions among them to see whether their effects on firms' tendency to use franchising, and on the way they design their franchise contracts, is consistent with the various theories. For that reason, the empirical model will embody simultaneously all the factors suggested by the first four types of theories discussed in Chapter 2. In this way, other effects will be controlled for when the empirical results are contrasted with predictions from the various models. It is important to point out again that while franchise arrangements are mainly share contracts, some franchisors choose to use a fixed-rent contract, so that the roy-alty rate r is nil in some cases. No firm ever employs the three types of contracts simultaneously. Empirically then, fixed-wage contracts (company-operated outlets) will be compared to a spectrum of share arrangements that include fixed-rent con-tracts as a special case.6 Any of these gives more incentives to the franchisee, and A test will be performed to determine whether or not franchisors that opt for a fixed-rent contract rather than a share contract should be treated separately from the others. 39 less so for the franchisor, than a fixed-wage contract would be. They also reduce the amount of risk faced by the franchisor. 3.1 Franchisors' Choices of Contractual Mix One of the peculiarities of franchising exploited here is that the majority of franchisors operate "mixed systems", i.e. both company-operated and franchised units. As noted by Caves and Murphy (1976), the fact that franchisors were willing to give their views on the optimal number of company-operated outlets in the Ozanne and Hunt (1971) survey indicates that they consider this to be a choice variable. This proportion of franchised outlets varies significantly across franchisors. Consider an upstream firm that produces a homogeneous product under a trade-name. As a consequence, this firm possesses some market power. It can distribute units of this good either through company-operated or through franchised stores. Consumers do not differentiate the two types of outlets so there is a single demand for this firm's product, given by the sum of the individual demands at the outlet level. It will depend on the price at which the good is sold, p, which is assumed to be the same across all outlets of a given franchisor, on the value of the tradename T, and on the provision of local inputs Z{. Franchised outlets all operate under the same linear contract with a share parameter equal to r, and a franchise fee of F? In other words, the franchisees pay an amount given by vp-Xi+F = r-P-(f(T,h) + 6l) + F (3.1) to the franchisor. The upstream firm is the one who chooses F and r. Given that the two types of contractual arrangements lead to differences in the amount of risk the franchisor faces, in the amount of supervision he must perform, in the cost of capital, etc., the sale of one unit of the good will contribute differently to the franchisor's profits (or utility) depending on the contractual arrangement under which it is sold. One can think in terms of a shadow price pc associated with each of the qc units sold through company-operated outlets, where qc = ^ " I j A'j. There would be a different shadow price p/ for the qj units sold through franchised outlets, 7 Again, since the fixed fee is for the duration of the contract, which lasts for almost 14 years on average, if Xi is interpreted as yearly sales for example, F would refer only to that portion of the fixed fee that is attributable to a year. 40 where <?/ = 53" = n i A'i- In general, these shadow prices will vary as a function of the circumstances each franchisor faces. An Illustration: In order to provide some intuition about what these shadow prices might represent, as an illustration, consider a case where there is non-noisy monitoring at the outlet level. Thus assume that the franchisor can impose his optimal level of for i = 1... ra, where n is the total number of outlets, but it requires that he expands more resources monitoring company managers than franchisees. In other words, there is a supervision cost function Cs = Cs(qf,qc) where qj and qc refer to the number of units sold through franchised and through company-owned stores respectively. If supervision costs are the only costs the firm incurs to uphold the value of its tradename, Cs = CT from Chapter 2. For simplicity, assume that where 82 < S\ so that the marginal supervision cost of units sold through franchised stores is less than that associated with units sold through company-owned stores. The value of 81 and of 82 would depend on how difficult it is to monitor the provision of as well as on the importance of this input in the retailing process. Assume that there are no other differences between franchised and company-operated outlets, i.e. neither party cares about risk, the value of the tradename is known, and franchising offers no advantages in terms of the cost of capital. Given that the /{ are fixed by the franchisor, total demand Q — Xi = qf + qc becomes a simple function of p. Its inverse is given by P{Q). For ease of exposition, assume also there are no fixed costs associated with retailing, (FC in Chapter 2) and the fixed fee F is set to 0. The franchisor's profits can be written as Cs = SQ + 81 * ql + 82 * qj (3.2) (P(Q) -c)-qc+r- P{Q) • qf - (80 + 8l * q2c + 82 * qj) - C{Q) (3.3) 'c and r [ P'(Q) • qf + P(Q)} + qcP'(Q) - 282qf - C. (3.5) dqf 41 In these circumstances, the shadow prices p c and pf would be denned as P c = P'{Q) [ qc + rqf} + P(Q) - c - 2Siqc (3.6) and P f = P'(Q) [ q c + r q f } + rP(Q) - 252qf. (3.7) Thus, both shadow prices would depend on qf and on qc. In addition, since the £'s depend on how difficult it is to monitor l{, the shadow prices would depend on this as well. The franchisor's problem could then be rewritten as max [pfqf + pcqc - C(qf + qc)] • (3.8) If .9c The solution to this problem would give a profit function U = U(Pf,pc,w) (3.9) where w is a vector of input prices. The General Case: In general, as was discussed previously, franchises may be used for many reasons in addition to possible differences in the cost of supervising downstream operators. If share contracts are used because of their risk-sharing property, as in the first type of model discussed in Chapter 2, the shadow value of units sold through franchised outlets will include a positive risk premium for the franchisor. This is because he faces less risk when he uses franchising than when he uses vertically integrated outlets. And the higher the amount of risk present, the higher this risk premium: thus pf would increase relative to p c . 8 However, the risk-averse franchisees will also require a higher risk-premium when faced with more uncertainty. And pf would decrease as the risk premium demanded by the franchisees increased. The net effect of increased risk on pf is therefore indeterminate. If the franchisor is less risk-averse than the franchisees, which is the usual assumption in this type of models, the increase in his risk-premium given an increase in o~e would be smaller than the increase in the franchisee's premium, and thus pf would fall as the variance of 6 increases. As a result, franchisors would use franchising less in those R ° It is also true that the more risk-averse the franchisor, the higher this risk premium would be. I will ignore this in order to concentrate on testable implications of the theoretical models. 42 cases where the amount of exogenous risk is greater. If the franchisor was the more risk-averse party, the opposite would result. This is consistent with Stiglitz (1974). He shows that the share of output paid to the landlord should increase as a result of an increase in risk if the landlord is less risk-averse than the tenant. As will be seen later on, in the framework used here, this is equivalent to saying that pf will decrease, given r, whenever risk increases. This is true of course if one assumes that the franchisor is less risk-averse than the franchisee. In the context of one-sided hidden-action models, share contracts are used as a compromise, providing both insurance and incentives to franchisees. In this case, increased risk, controlling for the level of supervision costs, puts more weight on the insurance side of the trade-off. Here, more risk would clearly decrease pj relative to Pc-But in such models, share contracts are also used to provide franchisees with incentives to work harder. Then, any factor that increases the value of l{ in the downstream production process, or makes it more difficult to monitor, increases the value of a contract that gives more incentives to the downstream operator. Thus pj increases relative to pc. An example of this is given by Rubin (1978). He discusses the fact that it is more costly for a manufacturer to monitor the behavior of distributors as the outlets get farther apart and farther away from him. Hence he argues that increased distance should lead to a greater reliance on franchise contracts. On the other hand, factors that increase the importance of the franchisor's contribution in the downstream production process, namely T, and possibly some management expertise, would make it more important to provide him with incentives in the Eswaran and Kotwal (1985) framework. In this case, pc increases with the value of T and the cost of maintaining its value. Therefore if the tradename is very central to the agreement, or as its value increases, one would expect more control by the franchisor. Similarly, pc should increase relative to p/ when the managerial assistance of the franchisor is very valuable. In Goldberg (1982) as in Brickley and Dark (1987), an increase in the value of the tradename was also expected to reduce a firm's tendency to use franchising. In their transactions costs framework, this was due to the fact that franchisees would be more inclined to free-ride on the tradename, by offering lower quality products and thus increasing profits, if the value of the tradename was greater. Because store managers & 43 would not have the same incentives to do this, assuming that their revenues do not depend on profits as much as franchisees's do, franchisors would tend to operate more outlets with hired managers. This is similar to the argument presented by Mathewson and Winter (1985) to explain the buy-backs of the early 70's. This explanation of the effect of the value of the tradename can easily be reconciled with the one derived from two-sided hidden-action models if one considers that the franchisor's role in maintaining the value of his tradename includes policing and other activities aimed at controlling franchisee free-riding. The value of franchising may also increase when upstream firms face a binding capital constraint if it allows franchisors to get access to cheaper capital. The tighter the constraint, the more py would go up relative to pc. Hence, under a capital-market-imperfection explanation of franchising, one may expect pf to be relatively larger in new and in rapidly expanding chains. It should also be greater when the amount of capital required to open an outlet is large. The share parameter in the franchise contract, r, would also affect how valuable franchising will be to the franchisor. For example, the value of the risk premium included in py depends on the royalty rate. Similarly, the incentives provided to both sides by the share contract, and thus the value of using such a contract, are a function of r. Finally, as was illustrated above, given that the franchisor faces a downward sloping demand curve, and that there may be non-linearities involved (as in the case of supervision costs), the value of selling one more unit through a franchised or a company-operated store would be negatively related to the two quantities qj and qc. These arguments can be summarized in the following general functional specifi-cations for pf and p c : where IU is an index that represents the level of risk or uncertainty, I\ is an index that stands for the costliness of monitoring the retailer's provision of l{, represents the importance of the franchisor's inputs, namely T, and IK stands for the franchisor's need for capital. As in (3.9), the franchisor's profit function can be written as Pf = Pf(Iu,Il,lTjK,r,qf,qc) and Pc = Pc{Iu, IIJTJK, If, 9 c ) (3.10) (3.11) n = n ( p / , p c , w ) (3.12) 44 where w is the vector of input prices. Assuming that these input prices are the same for all firms, we have n ( p / , p c , ™ ) = U{pf,pe,w) = TLx{pf,pc). (3.13) As is shown in Diewert (1982), Hotelling's lemma can be applied to this profit function to give 9/ = -QZ~ ~ ff(I«JlJT,lK,r,qf,qc) (3.14) 9c = - Q — = fc{IuJhh jK,r,qf,qc). (3.15) OPc Solving for q^ and q*c, we have 9/ = 9f(Iu,Ii,lT,lK,r) and (3.16) qc =9c(h,Il,lT,lK,r) (3.17) i.e. the quantity of the good sold each way depends on the amount of uncertainty, the importance of l{ and T in the retailing process, the franchisor's need for capital and the royalty rate. The firm's propensity to use franchising is defined as qf/Q and is given by 9/ 9/ 9f{IuJl,lT,lK,r) 7^7 = , . t = 7 T - T — H =9f{Iu,Il,lT,lK,r) (3-18) Q 9f+9c 9f(-) + 9c(-) Assuming that there are no systematic differences in size between franchised and company-operated stores, q*f/Q* can be measured as the percentage of franchised stores in each franchise chain. This formulation of the franchisor's problem assumes that somehow the param-eters of the franchise contract are exogenously given. This might be the case if there were "industry standards" that the franchisor felt constrained by, for example. In order to see if the variable fee does influence the choice of contract mix, equation (3.18) will be estimated below. However, in general, the franchisor is in a position to choose the terms of the contract. The following discussion endogenizes the choice of r and F. 45 3.2 Extension to the Contractual Design Most theoretical work on contractual choice takes for granted that the decision-maker (landlord or franchisor) can choose the parameters of the share contract, here F and r, in such a way as to maximize the value to himself of this contract. And these parameters have been found to vary systematically from one franchisor to another, even within sectors. As was mentioned, Ozanne and Hunt found royalty rates as low-as 1% and as high as 18% in the fast-food industry in their 1971 survey. Stiglitz (1974) shows that under the pure risk-sharing argument, the share of output paid to the landlord should increase (decrease) as a result of an increase in risk if the landlord is less (more) risk-averse than the tenant. In Eswaran and Kotwal (1985), the landlord's problem is presented as a two-stage one that is solved recursively. The landlord first determines the optimal share-contract parameters as functions of all of the variables in the model. These include the importance of the inputs provided by the tenant and the landlord, which vary with the type of technology used, as well as the extent of their differential abilities in providing these inputs, their opportunity. wages, and the monitoring technology. Then the franchisor steps back and compares the outcome from this optimal share contract to the outcomes from the fixed-rent and the fixed-wage contract. He chooses the one that maximizes his expected income. Finally, as was mentioned in Chapter 2, an explanation for franchising that relies on imperfect capital markets would generally imply a low royalty rate and a high franchise fee.9 Given the assumption of an infinitely-elastic supply of potential agents, in most of the models, F serves, given r, as a way of extracting whatever surplus is left at the downstream level after the franchisee has achieved his opportunity level of utility. Under this interpretation, F is simply a function of r. But since r depends on the variables in the models, F is ultimately also a function of these variables. Thus, we 9 In the context of their model, the fixed-rent contract is not equivalent to a share contract with r = 0, F > 0, and the fixed-wage contract to r = 1, F < 0. If they were, the two-stage procedure would not be justified. But with their fixed-wage contract, the landlord provides both unmarketed inputs and in the fixed-rent contract, the tenant does. Hence they are not simply special cases of the share contract, the latter having both the tenant and the landlord specializing in the activity for which they have a comparative advantage. In the data on franchising however, fixed-rent contracts are taken to be a special case of franchise contracts, with r = 0. 46 have the following reduced form equations for the two fees: r = r(Iu,Ii,IT,IK) (3.19) and F = F(Iu,It,IT,IK). Substituting (3.19) into (3.16) and (3.17) gives (3.20) GF(IU,IHIT,lK) (3.21) GC(IU,II,ITJK) (3.22) so that Qf[h,Il,lT,lK)- (3.23) Q* In this case, we have equations (3.19) and (3.20) to estimate together with (3.23). Since the few theoretical papers that have been written on franchising, i.e. Rubin (1978), Blair and Kaserman (1982), and Mathewson and Winter (1985), do treat r and F as decision variables of the franchisor, some of their results concerning the choice of contractual parameters can be tested here. For example, Blair and Kaserman (1982) argue that a greater value of the brand name implies that future demand relies heavily on the franchisor's future behavior. Their model predicts that the optimal level of the variable fee should be larger, and thus the fixed fee lower, for these franchisors than for those whose brand name is not as valuable. In fact, anything that increases the franchisor's tendency to shirk is expected to lead to more reliance on the variable fee, and a correspondingly lesser reliance on the fixed fee. Anything that increases the probability that the franchisee will cheat would have the opposite effect. This is consistent with Rubin (1978) and Eswaran and Kotwal (1985). Since Ji and IT are meant to capture exactly those factors that affect both parties's incentives, the estimation of (3.19) and (3.20) provides an empirical test of these results. The estimation of (3.20) also provides an implicit test of the assumption that is maintained in deriving it, i.e. the idea that no rents are left downstream and that franchisees are held at their opportunity level of utility. It is important to notice again that the theoretical models on which the construc-tion of (3.19)and (3.20) is based generally imply a different optimal share contract for each principal and agent pair. This is clear for example in the risk-sharing argument 47 since the optimal share is a function of the principal and agent's risk-aversion parame-ters.1^* It is also quite obvious in the Eswaran and Kotwal (1985) model. The optimal share contract is chosen as a function of the individuals' abilities, the importance of their inputs, etc. Thus a different optimal share contract would obtain whenever the individuals under consideration differ. Yet this is not what is observed in franchising. Each franchisor establishes a single franchise contract, and uses it with all of its fran-chisees.11 As was mentioned in Chapter 2, one of the puzzles authors have tried to address in the sharecropping literature has been the relative constancy of the share parameter around 50%. Here, share contracts vary across franchise chains, but the same contract is used for all the franchisees of a given franchisor. One way to explain this, given existing theories, would be to assume that the outlets and outlet operators a firm faces are similar enough that different share contracts are not necessary. This is clearly unsatisfactory. The use of a single share contract by franchisors could also be rationalized if differences in the parameters of the problem tended to give relatively similar results in terms of optimal share, as in Eswaran and Kotwal (1985). The fact that firms belonging to the same franchise sector offer different share contracts goes against both of these arguments. Another way to explain this behavior would be to suppose that the development of a franchise contract is not costless. So while franchisors may feel it worthwhile to develop different contracts for very different types or sizes of businesses, it is easily conceivable that it may not be worth doing for each franchisee. Thus the franchisor chooses an "average" optimal share contract for all of its franchisees, and a fixed-wage contract. As was noted in the context of the pure risk-sharing and the one-sided hidden-action models in Chapter 2, the fact that each franchisor's franchise contract would be an "averaged" one would in general affect the value of selling units through franchises, pf, and would alter the optimal choice of contract mix. In other words, some of the adjustments for differences among outlets or among outlet operators would then be done through the company-own or franchise decision rather 1 0 See Stiglitz (1974) and Rees(1985). 1 1 There are some exceptions, i.e. some franchisors offer various contracts to franchisees. These are usually associated with different types or sizes of businesses. For example, restaurant franchisors may offer a different type of contract to a franchisee for a full fledged restaurant than for a small satellite operation. These are referred to as "different franchise packages". When this happened in the data, the average royalty rate or franchise fee demanded was used on the basis that these should apply to the average type of outlet. 48 Figure 2: The Trade-Off Between the Two Decision Variables, r and q//Q than modifications of the terms of the contract. If one takes for granted the notion that F is determined as a residual, given r, the franchisor controls two major variables with respect to his contractual relations with his franchisees: the proportion of franchised stores and the share parameter, r. Since the same explanatory variables are found in (3.19) and (3.23), it is clear that in response to differences in any of these variables, either r and/or the proportion of franchised stores can be modified. Thus franchisors can trade off one for the other. This is illustrated in Figure 2. Assuming, for example, that all outlets are equally risky, the total amount of risk the franchisor faces will be proportional to the shaded area, given that he has chosen r* and q*f/Q* on the graph. A franchisor that would find a higher degree of risk more profitable to him could achieve it by choosing a higher royalty rate with q*j/Q* the same, or by reducing q//Q for the same r*, or through any combination of the two. Hence any movement in the directions shown by the arrows would achieve this goal.12 12 This assumes that starting from a given point, r* and q~f/Qa firm that would now want to bear more of the risk, will not choose to move in a direction that will decrease this risk on one 49 In terms of individual franchisees, of course, changes in r are not equivalent to changes in qf/Q- But in terms of the amount of risk the franchisor must contend with, there is clearly a trade-off. And the franchisor is the one who chooses the contract mix as well as the terms of the contracts here. Similarly, from the point of view of the franchisor, the amount of incentives provided to downstream operators in the chain is proportional to the unshaded area. In order to increase this "total amount of incentives" downstream, movements in any direction opposite to the arrows will be sufficient, i.e. an increase in qf/Q and/or a decrease in r. If the former is chosen, a greater number of store operators will have a kind of contract that incites them to work more. In the latter case, those who were already getting a share of the output will now get an even larger share, which means their contract will incite them to work even more than before. This relationship between contract mix and contract terms has also been noticed in the study of sharecropping. For example, Binswanger and Rosenzweig (1984) remark that "It is not known why, for example, in certain areas adjustments typically take the form of changes in contract terms whereas else-where terms remain unaltered but there are shifts in the relative importance of different types of contracts."(p.31) Also, Chao's (1983) explanation for the constancy of the terms of the sharecropping contract, at 50-50 for over 2000 years in China, was based on the argument that landlords and tenants adjusted to change by modifying the amount of land and labour they provided under share contract, rather than by changing the contract itself. Table 3.1 summarizes the expected effect of each of the indices on qf/Q under the various theoretical models discussed in Chapter 2. It is important to point out again that the various explanations for share contracts put forth in the literature are not mutually exclusive. This can be seen here in the fact that there are no conflicts among the signs of the various indices across models. The effects of the same indices on r and F are readily inferred from the table. The previous discussion implies that with endogenous fees, increases in r and reductions in of the two instruments and then move in the direction that increases the risk it faces using the other instrument in such a way as to more than compensate for the reduction in risk induced by the first movement. In any case, this would not be possible in a context where r and qf jQ were treated independently. 50 Table 3.1 Expected Effects of Indices on qf/Q in the Various Models Model In h IT IK Pure Risk-sharing One-sided Hidden-action — + Two-sided Hidden-action + — Capital-market Imperfection (-) + (-) + * Assuming the franchisee is the more risk-averse party. Note: Parentheses indicate that one and/or the other effect should occur. q//Q are to some extent equivalent ways for the franchisors to deal with differences in the indices. Thus those things that would tend to increase qf/Q should reduce r, and vice versa. In addition, F is taken to be a residual given r, so that those circumstances that would lead to a lower royalty rate should also increase the optimal value of F. Consequently, the effects of the indices are expected to be in the same direction for qf/Q and F, and in the opposite direction for r. 4. Conclusion From the brief review of the existing empirical literature on share contracts presented here, it is clear that there is still much to be done in terms of evaluating the abundant theoretical literature on this subject. Firstly, theoretical models have tended to focus on single principal-agent pairs and as a result, the comparative statics results that arise from this literature concern mainly the variable fee, r, and, by extension, the fixed fee F. Due to the unavailability of data on contract terms, empirical work on the other hand has focused on the observed contract mix. In reality, franchisors make decisions with respect to both the terms of the share contract, and the extent to which they want to utilize share contracts. This should be taken into account in the theoretical as well as the empirical literature. In addition, it was seen that the way in which the hypotheses being tested relate to the theoretical literature was not always well defined. The intention behind the 51 development of the empirical model in the second part of this Chapter was to provide a more systematic framework with which one could interpret the estimation results to come. Finally, it is clear that the limited amount of data that have been available has imposed major constraints on empirical work in this area. Thus one of the tasks involved in this thesis was the construction of a new micro level data set. In other words, information on individual franchisor's proportion of franchised outlets, royalty rate and franchise fee, as well as data on factors that are likely to affect these according to the theories, had to be gathered. Given that this type of data set has not been available before, I provide a fairly detailed description of it in the next Chapter. 52 C H A P T E R TV The Data: Sources and Characteristics 1. Introduction The first part of this Chapter is concerned with the way in which the final sample of 548 U.S. franchisors to be studied in the empirical analysis was arrived at. It also describes the main sources of data that were used. The next section contains a rather detailed statistical summary of the data. Both the type of information that has been obtained at the level of the individual franchisor, and the relevant data that are only available on a sectoral level from the U.S. Department of Commerce, are discussed. This is intended to provide information on the sample of franchisors studied here, as well as some stylized facts about franchising in general. Concluding remarks are found in Section 4. 2. The Sample of Franchisors Traditional product franchising, such as auto and truck dealerships, gasoline stations and soft-drink bottlers, in which the franchisor-franchisee relationship is limited to the sale of a given product under the franchisor's tradename, is still very widespread. According to the U.S. Department of Commerce, it accounted for 74% of all sales done through franchising in 1985 (but only 26% of all franchising employment). However most of the growth that has occurred since the 1950's has been attributable to business-format franchising, i.e. those franchises in which franchisors offer a complete system of operation and ongoing support to franchisees. This is probably why most surveys on franchising focus on format franchising. The following data having been compiled from such surveys, they are also restricted to business-format franchising. Because traditional product franchises do not in general involve the payment of a franchise fee up front or the payment of royalties, they could not have been included in this study in any event.1 1 Instead, as was discussed previously, the franchisor is the sole supplier to his franchisees, so he can rely on input mark-ups as a source of revenue. Franchisees are not capable to substitute away from these inputs, therefore no distortions result from mark-ups in this case. 53 The data set used here in fact consists of a cross-section of business-format franchisors in 1986. The main source of these data was Entrepreneur Magazine's "1986 Franchise 500" survey which was published in January 1987. It provides data on 1110 franchisors.2 Exclusions from this data base were made as follows. Seven firms were excluded because they reported no outlets in 1986. Further, 40 reported only one outlet, either company-operated or franchised. Since such firms did not really qualify as "chains", they were removed from the sample. An additional 23 firms were discarded because of inconsistencies in the reported data on the number of outlets, the fees, etc. Three more companies were excluded because they were not really engaged in franchising as defined here."* Of the remaining 1037 franchisors, 22 based their royalties on something other than their outlets' sales or gross revenues. Since I was interested in analysing contractual design using precisely these data, all such firms were eliminated: . they do not have a royalty rate that can be compared to that of the other firms in the sample. There remained 1015 firms of the original 1110 after these exclusions. It is worth noting that 15 of the 22 firms just mentioned extracted royalties on a per unit of output basis. As was mentioned previously, this is equivalent to levying royalties on sales if the price of output at the downstream level is controlled by the franchisor. Assuming a per unit fee of £ dollars, and a fixed price of p, the equivalent royalty rate is simply tjp. Thus these firms calculate their royalties on something which is similar to sales. Only the remaining 7 franchisors based their royalties on a notion of profits. That such an overwhelming majority of firms, i.e. 1001 out of 1008, ask for a percentage of sales or a royalty based on output rather than a proportion of profits is an interesting empirical fact. A survey published in 1985 by the Association of Canadian Franchisors found a similar regularity. 87% of the franchisors in their sample assessed royalties on the basis of outlets' sales; less than 9% based them on profits; the others calculated royalties on the basis of commissions (i.e. another notion of total revenues), purchases (i.e. input sales), or on the number of stores operated by the franchisee. The theoretical models discussed previously cannot account for ^ This number excludes the four companies which were surveyed but were subsidiaries of other surveyed firms. Their outlets were already accounted for by the parent company. They are Sunshine Polishing Systems, MicroAge Canada, Baskin-Robbins Ice Cream Silcorp and Minimaid/Minimenage. ^ Those were USBL, a sports franchise, Western Temporaries and Foremost Liquor Stores. 54 such constancy in franchisors' behavior: royalties on sales and royalties on profits are equivalent in the models. Firms that reported no franchised outlets in 1986, i.e. those that were wholly company-operated chains, were excluded as well. Only 29 of those were included in the survey, 6 of which would have remained in the final sample once additional information had been gathered. So even though it would have been interesting to have fully-company-operated firms in the sample, they were very underrepresented here and I felt it best to exclude them altogether. Two firms were also added to the sample.4 These firms were not seeking franchisees in 1986 and so did not respond to Entrepreneur's survey. They were put back in the sample since data were available for them in the "1985 Franchise 500" and other sources. Complete information concerning royalty rates, franchise fees, advertising fees and franchised and company-operated outlets in 1984, 1985 and 1986 was available for 913 of the resulting 988 franchisors.5 For each of these 913 firms, the survey also provided information about the year the}' had begun their operations and the year they had begun franchising. Whether or not franchisees are required to have previous experience in the business and whether the franchisor provides financing to his franchisees was indicated as well. However, 23 of those 913 firms were found to have no franchise fee and no royalty rate, even taking into account future fixed payments and advertising fees. Some of these were contacted by phone. At Lindal Cedar Homes for example, I was told that the company was not really involved in franchising. They were looking for dealers and franchise surveys were a good way to get exposure. They added that in order to be able to demand either a franchise fee or royalties, they would have to go through the paper work required by the various State laws on franchising in the U.S. and they did not feel it was worth it. In other words, there are costs associated with becoming a franchisor. At Champion Auto Stores, I was told the firm made money by selling inputs to its franchisees who were not, however, required to buy from it. Antitrust laws in general do not allow such requirements.6 Assuming that these firms get all of their revenues from dealers through the use of 4 These were Burger King and Ken's Pizza. 5 The 1986 and 1987 Franchise Annuals were used to get some of this information when it was missing in the survey. 6 See Brickley and Dark (1987). 55 input mark-ups, do they belong in the sample of franchisors studied here? Under the conditions discussed in Chapter 2, the sale of inputs at a price higher than marginal cost can be equivalent to royalties on sales. However these conditions are rather stringent. In particular, it is necessary for the downstream firm to produce with a fixed-proportions technology or for the upstream firm to control 100% of the inputs used at the downstream level. Any possibility of substitution at the downstream level would lead to distortions that would make input mark-ups less efficient than royalties on sales or profits.7 Consequently one would expect royalties on sales to be preferred by upstream firms. In addition, to get the equivalence between royalties based on output and roy-alties on sales, it is necessary that the price at the downstream level be controlled by the franchisor. In his analysis of 172 franchise agreements, 60% of which were for fast-food franchises, Udell (1972) found that 28% of contracts contained a clause stipulating that the "franchisor controls the prices franchisee may charge". Ozanne and Hunt (1971) found direct price control clauses in only 14% of the 121 fast-food agreements they examined. However, Caves and Murphy (1976) discuss the possi-bility that franchisors often control franchisees' prices through the price information they include in their advertising. Thus the conditions that are necessary for input mark-ups to be equivalent to royalties on sales may hold in some cases, but in general, input mark-ups will not be equivalent to royalties on sales.8 Consequently, it is not clear whether these firms' franchisees can be considered to be operating under a share contract, and if they are, what the parameters of that contract would be. Since the purpose of this study is to analyse the use and design of share (or fixed-rent) contracts as opposed to fixed-wage contracts, it was decided to exclude them from the sample. Finally, additional information on the number of States in which each franchisor has outlets, the number of outlets outside the home country, and the length of the initial training period was found in Entrepreneur Magazine's 1987 Franchise Yearbook 7 See J.M. Vernon and D.A. Graham (1971), R. Schmalensee (1973) and F.R. Warren-Boulton (1974) among others on the related question of the variable proportions incentive for vertical integration. R 0 Of course, if there is a cost involved in the use of royalties, as indicated by Mr. Lindal of Lindal Cedar Homes, this inefficiency may not be sufficient to insure that we do not observe input mark-ups rather than royalties. Champion Auto Stores is a case in point. 56 and the United States Department of Commerce's Franchise Opportunities Handbook for 1985 and 1986. Because of coverage problems, this further reduced the sample size from 890 to 548 franchisors. Table 4.1 gives a general idea of the coverage of this final sample of 548 fran-chisors. To produce its yearly publication, Franchising in the Economy, the U.S. De-partment of Commerce mails questionnaires to all known American business-format franchisors. In 1984 they received replies from 1942 franchisors. In 1985, 2090 fran-chisors responded. In both cases it was estimated that these franchisors accounted for 99% of all business-format franchising sales and establishments in the U.S. The distribution of all the franchisors surveyed, by sector and number of establishments, is published in Franchising in the Economy. Using the 1985 numbers as a descrip-tion of the whole U.S. franchisor population, it is therefore possible to calculate the proportion of this population that is accounted for in the final sample of 548 firms. The first column in Table 4.1 presents the number of franchisors included in the sample (N). The sectors, numbered 1 to 14, are those defined by the U.S. Department of Commerce in Franchising in the Economy. All franchisors in the sample were classified by the author in one of these 14 sectors; the Department of Commerce does not provide lists of the firms included in each of its sector. The total number of franchisors in the U.S., and the proportion of U.S. franchisors included in this sample are found in the second and third column respectively. Note that fewer size cohorts are found in some sectors than in others. In a few cases this is due to the way the data were published in Franchising in the Economy. In other cases size cohorts were combined by the author in order to eliminate those for which the number of franchisors in the sample was nil. Table 4.1 establishes clearly that the sample of firms studied here is biased towards large firms in the sense that a higher proportion of those are included in the sample. In fact, the proportion of firms in the sample tends to grow steadily as the size of the chain increases. This is not too surprising given the sources of the data. The probability that Entrepreneur is aware of the existence of a franchisor, and the probability that firms will respond to such a survey could easily grow with the size of the firm. The same is true for the ancillary sources of information that were used to complement the data set. The fact that firms with only one outlet were eliminated from the sample certainly contributed to this bias as well. 57 Sectors and s i z e cohorts Total U.S. N Franch. % 1- AUTOMOTIVE PRODUCTS 45 172 26, .2 0 - 10 OUTLETS 6 57 10. ,5 11 - 50 OUTLETS 19 57 33 .3 51 - 150 OUTLETS 12 30 40. .0 151 - 500 OUTLETS 5 15 33. .3 501 - 1000 OUTLETS 1 4 25. .0 10O1 + OUTLETS 2 9 22. .2 2- BUSINESS AIDS AND SERVICES 96 421 22. .8 0 - 10 OUTLETS 16 166 9 .6 11 - 50 OUTLETS 36 150 24. .0 51 + OUTLETS 44 105 41 . 9 3- CONSTRUCTION & MAINTENANCE 50 183 27, .3 0 - 10 OUTLETS 12 50 24 ,0 11 - 50 OUTLETS 12 80 15. .0 51 - 150 OUTLETS 7 28 25. .0 151 - 500 OUTLETS 11 16 68. 8 501 + OUTLETS 8 9 88. .9 4- CONVENIENCE STORES 8 27 29. .6 o - 50 OUTLETS 2 11 18. .2 51 - 150 OUTLETS 3 5 60. .0 151 - 1000 OUTLETS 2 8 25. ,0 1001 + OUTLETS 1 3 33. .3 5- EOUC PRODUCTS & SERVICES 29 75 38. ,7 0 - 10 OUTLETS 8 30 26. 7 11 - 50 OUTLETS 12 25 48. 0 51 - 150 OUTLETS 2 12 16. ,7 151 - 500 OUTLETS 5 5 100. 0 501 + OUTLETS 2 3 66. 7 6- RESTAURANTS 121 470 25. 7 0 - 10 OUTLETS 13 141 9. 2 11 - 50 OUTLETS 39 191 20. 4 51 - 150 OUTLETS 28 68 41. 2 151 - 500 OUTLETS 19 40 47. 5 501 - 1000 OUTLETS 13 17 76. 5 1001 + OUTLETS 9 13 69. 2 7- HOTELS. MOTELS & CAMPGROUNDS 10 43 23. 3 0 - 150 OUTLETS 2 30 6. 7 151 - 500 OUTLETS 5 8 62. 5 501 + OUTLETS 3 5 60. 0 Sectors and si z e cohorts Total U.S. N Franch. % 8- LAUNDRY & DRY CLEANING 3 18 16 .7 0 - 150 OUTLETS 1 14 7 . 1 151 + OUTLETS 2 4 50 .0 9- RECREATION AND TRAVEL 11 43 25 .6 0 - lO OUTLETS 2 14 14 .3 11 - 150 OUTLETS 5 19 26 .3 151 + OUTLETS 4 10 40 .O 10- AUTO & TRUCK RENTALS 6 26 23 . 1 0 - ISO OUTLETS 2 12 16 .7 151 - 1000 OUTLETS 2 9 22 .2 1001 + OUTLETS 2 5 40 .0 11- EQUIPMENT & TOOL RENTAL 1 31 3 .2 12- NON-FOOD RETAILING 99 329 30 . 1 0 - lO OUTLETS 13 108 12 .0 11 - 50 OUTLETS 33 1 10 30 .0 51 - 150 OUTLETS 25 62 40 .3 151 - 500 OUTLETS 20 30 66 .7 501 - 1000 OUTLETS 5 lO 50 .0 1001 + OUTLETS 3 9 33 .3 13- FOOD RETAILING - NON-CONV. 52 177 29 .4 0 - 10 OUTLETS 9 62 14 .5 11 - 50 OUTLETS 20 61 32 .8 51 - 150 OUTLETS 11 28 39, .3 151 - 500 OUTLETS 6 17 35 .3 501 • OUTLETS 6 9 66. .7 14- MISCELLANEOUS 17 75 22. .7 0 - 10 OUTLETS 1 34 2. .9 11 - 50 OUTLETS 7 26 26. .9 51 - 150 OUTLETS 2 6 33, .3 151 - 500 OUTLETS 6 6 100. .0 501 - 1000 OUTLETS 1 3 33. 3 ALL SECTORS 548 2090 26. .2 0 - 10 OUTLETS 80 686 11. ,7 11 - 50 OUTLETS 186 761 24. .4 51 - 150 OUTLETS 117 319 36. ,7 151 - 500 OUTLETS 99 204 48. .5 501 - 1000 OUTLETS 36 67 53. ,7 1001 + OUTLETS 30 53 56. .6 There is no reason to expect this bias in the sample to affect in a systematic way the distribution of any of the main variables of interest here, i.e. the proportion of franchised outlets, the royalty rate and the franchise fee. The observed stratification is in terms of exogenous variables only. The descriptive statistics that follow however are dependent upon the sample itself, which with 548 firms, represents 26% of the whole population of franchisors. 3. Some Interesting Descriptive Statistics This section contains a statistical summary of all the data used in this research. First, the data available for each franchisor in the sample of 548 firms are described with the aid of Tables 4.2 to 4.7. For all these tables, except Table 4.6, individ-ual franchisors were again grouped into the 14 major franchising sectors denned in Franchising in the Economy. Information for subsectors with a reasonable number of firms is provided as well. However these subsectors are not meant to account for all the firms in the original main sector. Data that could only be obtained on a sectoral basis from the U.S. Department of Commerce, and that were found to be relevant to this study, are presented in Tables 4.8 and 4.9. 3.1 The Number of Outlets Table 4.2 gives a description of the distribution of the total number of outlets, either company-operated or franchised, among franchisors in the various sectors. It also indicates the total number of outlets and franchisors (N) covered by the sample. Not surprisingly, the franchisor with the most outlets is found in the Hamburger category, and it is McDonald's. This is also the category with the largest average number of outlets. The smallest average numbers of outlets per franchisor are found in the sectors of Tune-up and Educational and Training programs.9 On average, franchisors in the sample had about 273 outlets in 1986. This is rather large and clearly affected by the presence of very large chains in the sample. The median number of outlets is about 54. This table also provides information on how long, on average, these firms have been in business (column 7) and how long they have been franchising (column 8). 9 As was previously discussed, all firms with only one outlet were eliminated from the sample, which explains why the minimum number of outlets is two. 59 Table 4.2 : Number of O u t l e t s f o r the 54B F r a n c h i s o r s Sectors TOTAL OUTLETS OUTLETS OUTLETS OUTLETS YEARS YEARS % TIME OUTLETS in 1986 In 1986 In 1986 In 1986 In BUS. FRANC. Not N i n 1986 Mini mum Maximum Mean St. Dav. Mean Mean FRANC. 1- AUTOMOTIVE PRODUCTS 45 7407 3 20O6 164 .6 392 .2 19 .7 11 .5 0.321 PARTS AND SERVICES 10 2462 3 1706 246 .2 521 .3 34 .6 14 .5 0.422 BRAKES, MUFFLERS ft SHOCKS 7 2957 5 2006 422 .4 742 . 1 15 .9 13 .7 0. 199 TRANSMISSION 6 44 1 7 142 73 .5 55 .9 18 .8 14 .5 O. 189 TUNE-UP 1 1 670 9 322 60 .9 90 . 1 9 .8 7 . 7 0.225 2- BUSINESS AIDS ft SERVICES 96 19593 3 6597 204 . 1 705 .4 14 .9 9 .5 0. 336 ACCOUNTING AND COLLECTION 19 2354 4 885 123 .9 212 .5 15 .7 10 .5 0.324 EMPLOYMENT SERVICES 20 1396 3 439 69 .8 1 10 .5 20 .6 13 . 4 0.390 PRINTING AND COPYING 14 4126 17 1081 294 .7 373 .2 14 .7 1 1 . 1 0.271 REAL ESTATE 8 9285 13 6597 1160 .6 2255 .8 14 .5 9 . 1 0.226 3- CONSTRUCTION ft MAINTENANCE 50 12922 2 3365 258 .4 530 .8 14 .8 9 .8 O. 294 MAID SERVICES 8 811 4 320 101 . 4 115 O 9 .0 6 .4 0.292 CONSTRUCTION 13 1253 9 4O0 96 .4 124 .4 13 .8 7 8 0.322 HOME IMPROVEMENT ft REPAIRS 11 1036 2 810 94 .2 238 .3 8 .8 6 .2 0.283 CARPET CLEANING 9 7948 150 3365 883 . 1 986 .3 27 . 1 19 . 1 0.249 4- CONVENIENCE STORES 8 2682 29 1276 335 .2 453 .6 25 .2 22 .4 0. 174 5- EDUCATIONAL PRODUCTS ft SERVICES 29 5727 3 3319 197 .5 618 . 1 14 .9 7 .6 0.389 EDUCATIONAL ft TRAINING PROGRAMS 12 773 4 273 64 4 88 .8 17 .6 9. 4 0.416 HEALTH AIDS ft SERVICES 17 4954 3 3319 291 4 800 .3 12 .9 6 4 0.370 6- RESTAURANTS 121 54295 3 9060 448 . 7 1273 . 1 19 .8 13 . 3 0.299 FAST FOOD 84 48143 3 9060 573 . 1 1507 .0 19 .9 13 0 O. 307 - CHICKEN 1 1 1 1334 4 8200 1030 .4 2419. 6 21 .6 12. 8 0.363 - MEXICAN 10 3158 10 2292 315 8 704 .6 17 .4 12 8 0.251 - HAMBURGER 9 18970 39 9060 2107 8 3097. .2 30 .2 20 8 0.266 - PIZZA 22 6236 11 340O 283 .4 726. 3 18 . 1 1 1 . 4 0.328 - SUBMARINES 13 1730 5 BOO 133 . 1 213. .3 17 3 12. 2 0.300 TABLE SERVICE 37 6152 5 915 166 3 221 . 6 19 .7 14 . 1 0.280 - STEAKHOUSES 6 2102 44 623 350. 3 267. 2 23 3 2 1 . 8 0.067 - ITALIAN 6 637 10 320 106 2 1 13. 8 20 .2 13. 3 0.298 - FULL MENU 11 2118 5 915 192. 5 288. 4 17. .9 13. 1 0.347 7- HOTELS, MOTELS ft CAMPGROUNDS 10 5298 63 1672 529. 8 456. 1 27. 5 21 . 9 0. 181 8- LAUNDRY ft DRY CLEANING 3 1281 46 1084 427. 0 57 1 . 4 23 .0 17 . O 0. 338 9- RECREATION ft TRAVEL 1 1 1097 6 34 1 99. 7 119. 1 1 1 . 5 9 . 0 0.282 TRAVEL AGENCIES 4 627 21 273 156. 7 104. 3 9. 7 8. 7 0.098 lO- AUTO ft TRUCK RENTALS 6 5905 58 3320 984 . 2 1240. 8 14 . 5 12 . 2 0.204 11- EOUIPMENT ft TOOL RENTAL 1 373 373 373 373. 0 0. 0 42. 0 24. 0 0.429 12- NON FOOD RETAILING 99 14635 2 1283 147 . 8 229. 1 17. 8 10. 0 0.345 CLOTHING ft SHOES 16 2019 6 486 126. 2 151 4 19. 6 12. 6 0.304 FURNITURE ft ACCESSORIES 9 1106 24 313 122. 9 126. 1 17. .7 10. 6 0. 350 ART SUPPLIES 8 869 6 323 108. 6 106 . 5 18. 6 10. 5 0.272 COMPUTER PROOUCTS 8 1403 7 576 175. ,4 187 . 2 6 .5 4 . 9 0.261 VIDEO RENTAL 7 1569 14 703 224. 1 286. 1 5 4 4. 6 0. 134 13- FOOD RETAILING - NON CONVENIENCE 52 15323 2 4894 294 . 7 825. 5 21 . 3 13. 4 0.326 DONUT SHOPS io 3201 23 1512 320. 1 496. 6 29. .4 18. 5 0.382 ICE CREAM PARLORS 17 10373 9 4894 610. 2 1351 . 5 25 5 17 . 9 0.258 SPECIALTY FOOD SHOPS 9 737 2 495 81 . 9 158. 8 19. 0 1 1 . 4 0.399 14- MISCELLANEOUS 17 3246 7 863 190. 9 234 . 8 12 2 8. 7 0.231 BEAUTY SALONS 12 3084 7 863 257. 0 252. 7 14 .7 10. 5 0.202 ALL SECTORS 548 149780 2 9060 273. 3 784. 2 17 .8 11 . 4 0.316 Firms in Parts and Services, those in the Hamburger sector as well as the only firm involved in Equipment and Tool Rental have been in existence for more than 30 years on average. The oldest firm in the sample has been in business for 110 years; it is called Ben Franklin. The sectors in which firms have been involved in franchising for more than 20 years on average include Convenience stores, Hamburger Restaurants, Steakhouses, Hotels, Motels and Campgrounds and Equipment and Tool Rentals. The newcomers to franchising, Video Rental and Computer Product stores, are also the "youngest" in terms of years in business. In most cases, firms were in business for quite a few years before they got involved in franchising. The last column of Table 4.2 gives the proportion of total time in business during which the firms have not been involved in franchising. It may be interpreted as an indicator of the difficulty or ease with which firms are able to adapt their operations to franchising. The smaller this proportion is, the faster the firms in the sector began to use franchising, which would imply that there was not much of a cost involved in adapting the business to franchising. When this proportion is large, it could indicate for example that it took more time to develop the concept or the franchise package, suggesting that the adaptation of this firm to franchising was more costly. On average, firms in the sample did not franchise for the first 31.6% years that they have been in business. This proportion is lowest for Travel Agencies and Steakhouses (9.8% and 6.7% respectively) and highest for Educational Products and Services (41.6%), Parts and Services (42.2%) and again for the firm in Equipment and Tool Rental (42.9%). The previous discussion would imply that the latter are harder to franchise than the former. It is worth pointing out however that 31.6% on average seems quite high. This goes against the notion that firms use franchising as a means to obtain capital. If this were the case, one would expect firms to start using franchising as soon as possible. As just noted, the delay may be explained at least in part by the need to develop the "system". But a lag of 6.4 years on average seems long in the context of an explanation of franchising as a response to capital constraints on the part of franchisors. Firms would be expected to be more "anxious" than that to get involved in franchising if they were waiting for this to expand their operations. The main purpose of Tables 4.3, 4.4, and 4.5 will be to show how the percentage of franchised outlets, the royalty rates and the franchise fees, which are the main variables of interest here, vary across and within the franchising sectors defined by 61 the U.S. Department of Commerce. 3.2 The Use of Franchising Table 4.3 is mainly concerned with franchisors' propensity to use franchising. From it, one finds that although the average proportion of franchised outlets is always quite high in these chains,, it still varies significantly from 56% in the Fast-Food Restaurants that specialize in serving chicken to 98.3% in the Carpet Cleaning and the Travel Agency businesses. On average, franchisors in this sample franchised 82.7% of all their outlets in 1986. According to the U.S. Department of Commerce's survey, 78.8% of all outlets in the same sectors (i.e. excluding traditional franchise sectors) were franchised in 1985. The lowest proportion of franchised outlets in this sample is found in the sector of Non-Convenience Food Retailing. More precisely, this record is held by a firm called Winchell's Donut House. The maximum proportion of franchised outlets is 100% in a majority of sectors. In fact, 117 of the 548 franchisors in this sample franchise all of their outlets. In the initial sample of 890 firms, 199 were wholly franchised. This contradicts the finding of Martin (1987). He obtained data on company-owned and franchised outlets for 772 franchisors from The 1985 Franchise Annual and found that only three of these firms had no company-owned outlets. This discrepancy can be explained, I believe, by the way in which he selected his 772 firms from the 3000 plus surveyed by The 1985 Franchise Annual. It appears that the Franchise Annual , in its description of franchise chains, does not indicate the number of company-operated outlets when this number is zero. Since he needed data for each firm on the number of franchised and company-owned outlets, as well as years in business and years involved in franchising, he would have eliminated all of those for which this information was not available. If this is correct, then most of the wholly-franchised chains would have been excluded from Martin's sample.1^ !0 This was verified by comparing data for wholly-franchised chains in the Entrepreneur survey with that from the Franchise Annual. More precisely, firms that operate both types of outlets will usually have the following information: Number of Units: Company Owned: x, Franchised: y. Those that have only franchised outlets either have: Number of Units: z or Number of Units: Franchised: z and nothing else. It is worth pointing out that some of the franchisors included in my sample are quite proud of the fact that they are 100% franchised: Century 21, for example, requires that franchisees advertise the fact that "Each office is independently owned and operated". 62 Table 4.3 : P r o p o r t i o n of F r a n c h i s e d O u t l e t s f o r the 548 F r a n c h i s o r s S e c t o r s 1- AUTOMOTIVE PRODUCTS PARTS ANO SERVICES BRAKES, MUFFLERS ft SHOCKS TRANSMISSION TUNE-UP 2- BUSINESS AIDS ft SERVICES ACCOUNTING AND COLLECTION EMPLOYMENT SERVICES PRINTING AND COPYING REAL ESTATE 3- CONSTRUCTION & MAINTENANCE MAID SERVICES CONSTRUCTION HOME IMPROVEMENT ft REPAIRS CARPET CLEANING 4- CONVENIENCE STORES 5- EDUCATIONAL PRODUCTS ft SERVICES EOUCATIONAL ft TRAINING PROGRAMS HEALTH AIDS ft SERVICES 6- RESTAURANTS FAST FOOD - CHICKEN - MEXICAN - HAMBURGER - PIZZA - SUBMARINES TABLE SERVICE - STEAKHOUSES - ITALIAN - FULL MENU 7- HOTELS. MOTELS * CAMPGROUNDS 8- LAUNDRY ft DRY CLEANING 9- RECREATION S TRAVEL TRAVEL AGENCIES 10- AUTO ft TRUCK RENTALS 11- EQUIPMENT & TOOL RENTAL 12- NON FOOD RETAILING CLOTHING * SHOES FURNITURE ft ACCESSORIES ART SUPPLIES COMPUTER PRODUCTS VIDEO RENTAL 13- FOOD RETAILING - NON CONVENIENCE DONUT SHOPS ICE CREAM PARLORS SPECIALTY FOOD SHOPS 14- MISCELLANEOUS BEAUTY SALONS ALL SECTORS % FRANC. % FRANC. % FRANC. % FRANC. X EXP. % FIN. N Minimum Max 1mum Mean St. Dev. Mean Mean 45 0 .368 1 .000 0 .837 0 . 198 0 . 133 0 .200 10 0 .368 1 .000 0 .846 0 .211 0 .000 0 .200 7 0 .400 1 .000 0 . 749 0 .243 0 .286 0 . 143 6 0 .857 1 .000 0 .962 0 .053 0 . 167 0 .000 1 1 0 .444 1 .000 0 .791 0 .234 0 . 182 0 . 182 96 o .088 1 .000 0 .899 0 . 187 0 .406 0 .437 19 0 .500 1 .coo 0 .929 0 . 127 0 .737 0 .474 20 0 .088 1 coo 0 .780 O .311 O .200 0 . 500 14 0 .672 1 .000 0 .947 0 .095 0 .071 0 .429 8 0 .824 1 .ooo 0 .978 0 .062 0 .750 0 .750 50 0 .444 1 coo 0 .924 O . 134 0 . 180 0 . 320 8 0 .750 1 .coo 0 .913 0 . 107 0 . 125 0 .250 13 0 .444 1 .coo 0 .918 0 . 151 0 .385 0 . 154 1 1 0 .500 1 .ooo 0 .880 0 . 194 0 .273 0 .364 9 0 .921 1 .ooo 0 .983 0 .032 0 .000 0 .667 8 0 .688 O .992 0 840 0 .094 0 . 125 0 250 29 0 .345 1 ooo 0. 817 0. . 176 0 .414 0 .207 12 O .345 1 .ooo 0. .762 0. .234 0 .250 0 333 17 0 .667 1 ooo 0. 855 0. . 114 0 .529 0 . 118 121 0. . 158 1 .000 0. 719 0. .239 0 .421 0 .058 84 0. , 158 1 .000 0. 715 0. 243 0 393 0. 036 1 1 0. 225 0 .992 0. 559 0. 259 0. 636 0. OOO 10 0 185 1 ooo 0. 737 0. 279 0. 500 0. OOO 9 0. . 158 0 .981 0. 677 0. 284 0. .556 0 OOO 22 0. 200 1 .000 0. 732 0. 226 0. .364 0. ,045 13 0 596 1. .000 0. 886 0. 149 0. 077 0. 154 37 0. 159 0 995 0. 727 0. 233 0. 486 0. 108 6 0. 159 0 .995 0. 719 0. 366 0. 500 0. 167 6 0. 572 0 992 0. 834 0. 169 0. 333 0. 167 11 0. 263 0 .944 0. 658 0. 174 0 636 0. OOO 10 0. 421 1 .000 0. 835 0. 190 0. 500 0. 2CO 3 0. 689 1 .000 0. 846 0. 156 0. 333 0 COO 1 1 0. 300 1, OOO 0. 860 0. 212 0. 091 0. 182 4 0. 952 1. COO 0. 983 0. 021 0. 000 0. 250 6 0. 948 1, OOO 0. 982 0. 021 0. 167 0. 667 1 0. 815 0 .815 0. 815 0. OOO 0. COO 0. 000 99 0. 103 1. 000 0. 834 0. 216 0. 242 0. 152 16 0. 103 1. COO 0. 787 0. 272 0 062 0. 062 9 0. 167 1. OOO 0. 718 0. 302 0. 444 0. 111 8 0. 825 1. OOO 0. 909 0. 070 0. COO 0 250 8 0. 714 1 COO 0. 936 0. 094 0. 625 0. OOO 7 0. 583 1. OOO 0. 902 0. 161 0 143 0. 143 52 0. 016 1 000 0. 800 0. 251 0. 250 0 077 10 0. 016 1. OOO 0. 820 0. 306 0. 10O 0. 300 17 0. 633 1. ooo 0. 907 0. 117 0. 235 0. 059 9 0. 111 0. 983 0. 6CO 0. 299 0. .222 0. OOO 17 0. 429 1 000 0. 866 0. 165 0. 000 0. 353 12 0. 429 1. 000 0. 895 0. 161 0. 000 0. 417 548 0. 016 1. 000 0. 827 0. 215 0. 297 0. 210 When asked what the ideal proportion of company-operated outlets would be, fast-food franchisors gave an average response of 42% in the Ozanne and Hunt (1971) survey [p.82]. Here, fast-food franchisors operate only 28% of their outlets. Either franchisors in this sample are quite far from what they consider ideal, or more likely, this preferred proportion has decreased since 1971. Still, fast-food franchisors tend to control directly more outlets than franchisors in other sectors do. In addition, Table 4.3 contains information on the proportion of franchisors that require their franchisees to have previous experience in the business (%EXP, column 6). 30% of the firms in the sample require such experience. This is especially frequent in sectors that demand special expertise, such as Accounting and Collection, and Real Estate. There are quite a few sectors, notably Parts and Services, Carpet Cleaning, Travel Agencies, Art Supplies stores and Beauty Salons, for which experience is not required by any of the franchisors.11 The last column, entitled %FIN, indicates the proportion of franchisors that provide direct financing to their franchisees, as opposed to simply aiding them in obtaining financing from other sources or giving them no assistance at all. Direct financing is found mostly in the Business Aids and Services sectors, and in such things as Carpet Cleaning, Auto and Truck Rental and Beauty Salons. However it is quite rare in the Restaurant sectors and subsectors among others. In total, 21% or 115 of the 548 franchisors offered financial assistance to their franchisees. Ozanne and Hunt (1971) found that a majority of the fast-food franchisors they studied preferred to enroll franchisees who had no previous experience in the business [p.124-125]. 3.3 About the Contractual Design: Royalty Rates and Franchise Fees A description of royalty rates in each sector is found in Table 4.4. Royalty rates here are meant to capture all those fees that represent the sharing component of the contract. Advertising fees stated separately as a certain percentage of sales were therefore added to reported royalties to generate the royalty rates used in this study.12 The highest royalty rate for this sample is found in the Health Aids and Services sector: it is that of Jazzercise, at 25%. In general however, the highest royalty rates are found in the sector of Automotive Products, where average royalties are at 9.2%, with two subsectors, Brakes, Mufflers, and Shocks, and Tune-up, with average rates over 10%. On the other hand a minimum rate of 0% is found in many sectors. The "% fixed" column indicates the proportion of franchisors that do not require their franchisees to pay any percentage of their sales. The next column gives the number of firms with royalty rates greater than 0. The last column indicates what the average royalty rate per sector becomes when only these last firms are included in the calculation. The exclusion of fixed-fee contracts for the calculation of average royalties is justified in the Eswaran and Kotwal (1985) model. Depending on whether a fixed-fee or a share contract is chosen, the tenant and the landlord provide different inputs. Therefore, in this model, fixed-fee contracts are not simply a special case of share contracts where the variable payments would be nil. Consequently, they should be separated from "real" share contracts. ^ In all, there are only 37 franchisors out of the 548 that have a zero royalty rate. Consequently, on the whole sample, the average royalty rate is not much affected by whether or not these firms are included in the calculation: it goes from 6.5% to 7.0%. But some sectors are more affected by this than others. In particular, while these In addition, royalty rates for travel agencies presented some problem. Some reported them on the basis of total commission. This is the relevant notion of total revenues for the outlets in this industry, and it is also used as the basis for royalties in the real estate industry. Others it seems, including International Tours, which is in the final sample, and Uniglobe, which is not, stated their royalty rate on the basis of total "Travel Sales" in the Entrepreneur survey. I found from a few articles that commissions in this industry are approximately 10% of sales value. And indeed, the royalty rate reported directly by Uniglobe when they were contacted was 10 times that given in the Entrepreneur Survey, and so was the rate they reported in Retailing & Services Guide 87. For these reasons, the royalty rate used for these firms was 10 times the rate indicated in the Entrepreneur Survey. Whether firms that use a fixed-fee contract in this sample behave differently from those that opt for a share contract will be tested later. 65 Table 4.4 : Royalty Rates for the 548 Franchisors ROYALTY ROYALTY ROYALTY ROYALTY X FIXED N ROYALTY Sectors exc. exc. fixed N Mini mum Maximum Mean St. Dev. Mean fixed Mean 1- AUTOMOTIVE PRODUCTS 45 0.0 19.0 9.2 3.5 0.022 44 9.4 PARTS AND SERVICES 10 3.0 10.0 6.9 2.5 0.000 10 6.9 BRAKES, MUFFLERS ft SHOCKS 7 9.0 19.0 11.9 3.5 0.000 7 11.9 TRANSMISSION 6 6.0 13.0 9. 1 2.3 0.000 6 9. 1 TUNE-UP 11 7.0 16.5 10.7 2.9 0.000 11 10.7 2- BUSINESS AIDS & SERVICES 96 O.O 18.0 6.6 3.9 0. 125 84 7.5 ACCOUNTING AND COLLECTION 19 0.0 17.0 6.3 4.6 0. 158 16 7.5 EMPLOYMENT SERVICES 20 0.0 10.0 7.2 2.4 0.050 19 7.6 PRINTING AND COPYING 14 4.5 8.0 6.5 1.3 0.000 14 6.5 REAL ESTATE 8 0.0 14.0 5.4 4.8 0. 125 7 6.2 3- CONSTRUCTION ft MAINTENANCE 50 0.0 20.0 6.2 4.4 0.200 40 7.8 MAID SERVICES 8 0.0 10.0 6.2 2.9 0. 125 7 7. 1 CONSTRUCTION 13 0.0 10.0 4.2 3.4 0.308 9 6.0 HOME IMPROVEMENT ft REPAIRS 11 0.0 10.0 5. 1 3.8 0. 182 9 6.2 CARPET CLEANING 9 0.0 17.0 6.6 5.0 0.222 7 8.4 4- CONVENIENCE STORES 8 0.0 15.0 5.9 5.4 0. 125 7 6.8 5- EDUCATIONAL PRODUCTS 8 SERVICES 29 0.0 25.0 7.5 4.5 0.069 27 8.0 EDUCATIONAL & TRAINING PROGRAMS 12 2.0 11.0 8.5 2.6 o.ooo 12 8.5 HEALTH AIDS ft SERVICES 17 0.0 25.0 6.8 5.4 0. 118 15 7.7 6- RESTAURANTS 121 1.0 15.5 6.6 2.2 0.000 121 6.6 FAST FOOD 84 2.0 15.5 7.0 2.2 0.000 84 7.0 - CHICKEN 11 4.0 10.0 7.6 1.8 0.000 11 7.6 - MEXICAN 10 2.0 10.0 5.7 2.8 O.OOO 10 5.7 - HAMBURGER 9 3.0 15.5 B.2 3.2 0.000 9 8.2 - PIZZA 22 4.0 10.0 6.7 1 .B 0.000 22 6.7 - SUBMARINES 13 5.0 13.0 7.4 2.3 o.ooo 13 7.4 TABLE SERVICE 37 1.0 10.0 5.7 1.9 0.000 37 5.7 - STEAKHOUSES 6 2.0 6.8 4.5 1.8 0.000 6 4.5 - ITALIAN 6 5.0 10.0 6.7 2.1 0.000 6 6.7 - FULL MENU 11 3.5 8.0 6.4 1.6 0.000 11 6.4 7- HOTELS, MOTELS ft CAMPGROUNDS 10 4.0 10.0 6.3 1.6 0.000 10 6.3 8- LAUNDRY ft DRY CLEANING 3 0.0 1.0 0.3 0.6 0.667 1 1.0 9- RECREATION & TRAVEL 11 0.0 10.0 5.8 2.9 0.091 10 6.4 TRAVEL AGENCIES 4 0.0 9.0 5.5 3.9 0.250 3 7.4 10- AUTO ft TRUCK RENTALS 6 3.5 9.0 6.8 1.9 0.000 6 6.8 11- EQUIPMENT ft TOOL RENTAL 1 2.7 2.7 2.7 0.0 0.000 1 2.7 12- NON FOOO RETAILING 99 O.O 16.5 5.7 3.0 0.040 95 5.9 CLOTHING ft SHOES 16 2.5 11 .0 5.6 3.0 0.000 16 5.6 FURNITURE ft ACCESSORIES 9 2.0 8.0 5.3 1.8 0.000 9 5.3 ART SUPPLIES a 1 .0 8.0 5.7 2.9 0.000 8 5.7 COMPUTER PRODUCTS 8 0.0 9.0 6.1 3.0 0. 125 7 6.9 VIDEO RENTAL 7 3.5 8.0 5.4 1.8 0.000 7 5.4 13- FOOD RETAILING - NON CONVENIENCE 52 0.0 9.9 6.0 2.4 0.038 50 6.2 DONUT SHOPS 10 0.0 9.9 7.4 2.9 0. 100 9 8.2 ICE CREAM PARLORS 17 0.0 9.0 5.2 2.6 0.059 16 5.6 SPECIALTY FOOD SHOPS 9 2.0 8.0 5.4 2.0 0.000 9 5.4 14- MISCELLANEOUS 17 0.0 15.0 6.9 3.7 0. 118 15 7.8 BEAUTY SALONS 12 0.0 15.0 7. 1 3.7 0.083 11 7.8 ALL SECTORS 548 0.0 25.0 6.5 3.4 0.066 511 7.0 Note: Royalty rates include the a d v e r t i s i n g fee franchisees pay to franchisors when this fee i s spe c i f i e d as a % of sales or gross revenues. Averages were used when ranges were given. "fixed-fee" contracts are relatively frequent in the Business Aids and Services and the Construction and Maintenance sectors, it is interesting to see that none of the 121 restaurants in the sample use this type of contract. In fact, none of the firms that operate with such a fixed-fee contract are found in those sectors that Brickley and Dark (1987) classify as "non-repeat" types of activities, i.e. Restaurants, Hotels and Motels, and Auto and Truck Rentals. This is consistent with the fact that the downstream operator's role is not as important in these kinds of non-repeat businesses. Franchise fees are the main focus of Table 4.5. In this case, as these fees must represent the "fixed" component of the contract, the present value of all future fixed payments specified in the contract was added to the initial franchise fee to generate the actual measure of franchise fee used in this thesis. In order to calculate the present value of such future payments, it was necessary to know what time horizon was covered by the contract. Individual data on the length of contracts were not available for all of the 548 franchisors. However the U.S. Department of Commerce publishes information from which the average duration of contracts in each of the franchising sectors can be calculated (see Table 4.8). Consequently, the discounting was done over the calculated average length of franchise agreements in the sector in which the individual firm was classified. A nominal discount rate of 10% was used since there was no indication in the data that the amounts of future fixed payments are adjusted to account for inflation.14 This computation was performed only 31 times, since only 31 of the 548 franchisors asked for future fixed pa3rments. The average fixed payment was $464. per month. Franchise fees in this sample vary from 0 to as much as 286.2 thousand dollars, with the latter found in the "Health Aids and Services" category,1^ as the highest royalty rate was. Only 7 firms in the sample have no franchise fee. The average franchise fee is largest in the "Laundry and Dry Cleaning" and "Travel Agencies" categories. The former is also the one with the smallest average royalty rate (with 2 out of 3 firms having no royalties at all). Since franchise fees and royalty rates are two means by which the franchisor can extract rent from its franchisees, one would 1 4 The resulting franchise fees were almost the same whether a 10% or a 5% discount rate was used. ^ This fee of 286.2 is for a company called MedStop. It is generated by adding up a franchise fee of $55,000 and future payments of $2,500 per month discounted at a rate of 10% for the average duration of contracts issued in 1985 in this sector (14.5 years, see Table 4.8). Clearly the actual number obtained is sensitive to this procedure and the numbers used. 67 Table 4,5 : Franchise Fees for the 54B Franchisors F. Fee F. Fee F. Fee F. Fee Cor. Spearman Cor. Spearman Sectors N Minimum Maxlmum Mean St. Dev. (F.r) Rank Cor. (F'.r') Rank Cor 1- AUTOMOTIVE PRODUCTS 45 0.0 79.5 19.7 15.0 -0.05 0.08 -0.05 -0.02 PARTS AND SERVICES 10 0.5 52.9 15.5 14.3 0.44 0.33 0.41 0.35 BRAKES. MUFFLERS & SHOCKS 7 10.0 22.5 15.8 5.3 0.30 -0.04 0.32 0. 11 TRANSMISSION S 15.0 25.0 21.1 3.8 0.20 0.03 0.24 0. 14 TUNE-UP 11 15.0 79.5 24.2 IB.9 0.00 -0.39 -0.27 -0.32 2- BUSINESS AIDS & SERVICES 96 0. 1 136.0 21.7 17.9 0. 11 0. 12 0. 11 0. 12 ACCOUNTING AND COLLECTION 19 0. 1 71.0 20.9 18.8 0.42 0.33 0.45 0.39 EMPLOYMENT SERVICES 20 5.0 50.0 21.5 12.8 0.24 0.07 0.25 0. 15 PRINTING AND COPYING 14 7.5 42.5 28.9 11.5 0.48 0.54* 0.47 0.51 + REAL ESTATE 8 5.0 19. 1 10.5 4.6 0. 15 0.26 0. 10 0. 14 3- CONSTRUCTION & MAINTENANCE 50 0.0 69.0 16.3 12. 1 -0.21 -0. 10 -0. 18 -0. 1 1 MAID SERVICES 8 7.5 31.1 13.4 7.6 -0.73 -0.02 -0.80 -0.76* CONSTRUCTION 13 5.0 69.0 19.5 16.7 -0.37 -0.34 0.30 0.27 HOME IMPROVEMENT & REPAIRS 1 1 3.4 56.9 18.3 14.8 0. 17 0.37 0.27 0.35 CARPET CLEANING 9 5.8 18.0 12.8 4.1 0.42 0.58+ 0.72 0.70* 4- CONVENIENCE STORES 8 5.0 25.0 12.8 7.3 -0.24 -0.26 0.08 0. 17 5- EDUCATIONAL PRODUCTS tt SERVICES 29 0.5 . 286.2 32.9 50.6 -0. 12 -0.08 -0.04 -0.01 EDUCATIONAL & TRAINING PROGRAMS 12 7.5 35.0 22.7 8.4 0.28 0. 17 0.33 0. 17 HEALTH AIDS & SERVICES 17 0.5 286.2 40. 1 65.6 -0. 11 -0. 12 0.09 -0.23 6- RESTAURANTS 121 5.0 87.5 21.2 12.2 -0.02 0. 11 -0.01 0. 12 FAST FOOD 84 5.0 87.5 18.9 10.6 0. 15 0.27* 0. 16 0.25* - CHICKEN 11 10.0 30.0 21.4 6.7 0.04 -0. 16 -0.03 -0.02 - MEXICAN lO 6.0 35.0 16.4 7.8 0.67 0.81**' 0.52 0.44 - HAMBURGER 9 10.0 87.5 31.9 22.9 -0. 18 -0.33 0.03 -0. 15 - PIZZA 22 5.5 25.0 15.6 5.0 -0.01 -0.22 0.40 0.26 - SUBMARINES 13 5.0 39.0 15.8 8.6 0. 18 0.35 0.21 0.30 TABLE SERVICE 37 7.5 64.5 26.4 14. 1 -0. 14 -0.02 -0. 15 -0.06 - STEAKHOUSES 6 15.0 50.0 26.3 13.0 0.73 0.54 0.90 0.77 + - ITALIAN 6 10.0 25.0 14.6 5.6 0.94 0.94* 0.97 0.94* - FULL MENU 11 18.0 50.0 32.3 10.2 -0.27 -0.27 -0.28 -0.47 7- HOTELS. MOTELS & CAMPGROUNDS 10 12.5 50.0 24.7 10.6 -0.23 -0.45 -0.32 -0.42 8- LAUNDRY & DRY CLEANING 3 42.2 66.5 52.5 12.6 0.96 1.00 n.a. n.a. 9- RECREATION & TRAVEL 11 6.0 127.5 34.3 37 .O -0.61 -0.30 -0.77 -0.71* TRAVEL AGENCIES 4 20.5 127.5 49. 1 52.3 -0.96 -0.80 n.a. n.a. 10- AUTO & TRUCK RENTALS 6 5.0 28.3 IB.7 10.2 -0. 15 0.31 -0.19 -0.37 11- EQUIPMENT & TOOL RENTAL 1 20.0 20.0 20.0 n.a. n.a. n.a. n.a. n.a. 12- NON FOOD RETAILING 99 0.0 112.5 19.9 16.9 0.21 0.32** 0.22 0.32** CLOTHING 8 SHOES 16 0.0 35.0 13.6 9.8 -0.01 0.00 0. 13 0. 10 FURNITURE a ACCESSORIES 9 0.0 50.0 23.5 17.3 0. 14 0.33 0.33 0.30 ART SUPPLIES 8 6.0 34 .0 17.2 10.5 0.74 0.67+ 0.77 0.86** COMPUTER PRODUCTS 8 5.0 45.0 24.9 13.8 0.87 0.90** 0.83 0.71 + VIDEO RENTAL 7 0.0 29.9 15.3 8.9 0.77 0.82* 0.69 0.39 13- FOOD RETAILING - NON CONVENIENCE 52 0.0 164.5 20.0 21.9 -0. 19 0.30* -0. IB 0. 11 DONUT SHOPS 10 5.0 40.0 22.0 10.9 0.50 0.35 -0. 19 -0.02 ICE CREAM PARLORS 17 0.0 164.5 24.8 36.8 -0.49 -0.07 -0.51 -0.31 SPECIALTY FOOD SHOPS 9 7.5 30.0 15.2 7.4 0.55 0.68* 0.63 0.83** 14- MISCELLANEOUS 17 5.0 99.0 26.0 23.2 -0.25 0. 17 -0. 19 -0. 19 BEAUTY SALONS 12 12.5 99.0 29.7 24.4 -0.63 -0.42 -0.69 -0.76** ALL SECTORS 548 0.0 286.2 21.5 20.3 -0.04 0. 10* -0.04 0. 10* oo Note: Franchise fees are In thousands of U.S. d o l l a r s . They Include the present value of a l l future payments s p e c i f i e d for the duration of the contract (using the average length of agreements In Table 4.8). The discount rate used Is 10%. Averages were used when ranges were given. F' and r' are error terms from regressions of F and r on number of outlets and number of years in business. Two-sided tes t s : ** : .01 l e v e l . * : .05 l e v e l , + :.10 le v e l . expect this kind of negative correlation between the two. However the average royalty rate for "Travel Agencies", at 5%, is below the overall average of 6.6%, but not as much as one could have expected it to be. Similarly the lowest average franchise fee of $10,500 is found in the "Real Estate" sector which does not have a high average royalty rate. In fact at 5.4%, it is below the overall average. A positive correlation between franchise fees and royalty rates could be observed in some cases. This is because a franchisor may be able to require higher fees in general when the value of his tradename is greater than that of another franchisor. In particular, if the brandname is more valuable, one expects the terms of the contract or the contract mix to be adjusted so that the franchisor is given more incentives to uphold its value. Thus r may increase. But a more valuable tradename would also make individual stores more profitable. Consequently, the franchisor could also increase F. In such a case, both r and F could be larger for one firm than for the other, and a positive relationship between the fees would be observed. Note that if the franchisor with a more valuable tradename chose a larger value for only one of the two fees, then one would find no correlation between F and r. The last four columns of Table 4.5 were compiled to investigate this correlation further. The simple correlation coefficient between the franchise fee and the royalty rate is presented in the first of those columns. The second column gives the Spearman rank correlation coefficient between the two fees. The advantage of the latter is that it is possible to test whether or not this correlation coefficient is significantly different from zero without having to make any assumptions on the distribution of the two variables. ^ For the whole sample, the simple correlation coefficient is negative but very small. Given the size of the sample, one may assume that the two fees are jointly normally distributed. Under that assumption, it is possible to test whether this correlation coefficient is significantly different from zero, and it is not. 1 7 The rank correlation on the other hand is positive and although quite small, given the size of the sample, it is significantly different from zero at the 0.05 level using a two-tail test. Correlation coefficients could be calculated for only 13 of the 14 main sectors. 1 ^ For n < 30, Spearman rank correlation tables are available for example in Gibbons (1976). For n > 30, we have that z = Ry/n — 1 ~ iV(0,1), where R is the rank correlation coefficient. 1 7 Given normality, under H(,:p — 0, we have ry/N — 2/y/\ — r2 ~ tjv-2 where r refers to the correlation coefficient. The calculated t-statistic in this case was 0.93, which is not significant. 69 It is interesting to see that the simple correlation coefficients are positive for only 3 of those sectors, while the rank correlations are positive for 8 of them. Only Non-Food Retailing and Non-Convenience Food Retailing have significant rank correlation coefficients on the basis of two-sided tests. In both cases the association between the two fees is positive. Similarly, in all the subsectors for which a significant association is observed between the two fees, it is a positive one. Because the explanation for a positive relation, or a lack of relation, between the fees depends on some notion of growth in the value of the tradename, it was thought that if this effect was taken out of the data, the negative correlation might be observed. Consequently, both the franchise fees and the royalty rates were regressed on the two variables that can serve as proxies for the value of the tradename, i.e. the number of outlets and the number of years since each firm started its operations. The correlation coefficients reported in the last two columns of Table 4.5 were calculated using the error terms from these regressions. Since three degrees of freedom were necessary in order to estimate these regressions, and because there are only three observations in the Laundry and Dry Cleaning sector, these new correlation coefficients could not be estimated for that sector. As a result of this procedure, both the simple and rank correlation coefficient for the whole sample remain unchanged. The simple correlation coefficients are reduced or made more negative in 17 of the 45 sectors for which all coefficients were calculable. In most cases though, the change is minimal. Only two coefficients go from positive to negative, while six go from negative to positive. Similarly, four rank correlation coefficients go from positive to negative, but another four go the other way. Significant negative rank correlations can now be found in one main sector, Recreation and Travel, and two subsectors. But in general, this attempt to take the effect of growth out of the two fees in order to uncover the negative relationship one expects to find between them was not very successful. The regressions fared so badly that this result is not too surprising.181 also tried controlling for the amount of inputs sold by franchisors, in addition to years in business and number of outlets, and also separately, and again found no correlation. Thus no clear relationship between franchise fees and royalty rates emerges from all this.1 9 The F-statistics were generally insignificant so that the hypothesis that both slope coefficients were nil could not be rejected in most cases. 1 9 The same results were obtained when the fixed fee that was used was divided by the duration of 70 It could be argued that this lack of correlation is simply a reflection of the little importance of the fixed fee as a proportion of total revenues collected by the franchisor from each franchised store. However, using aggregate data on average sales per outlet and on the average duration of franchise contracts, as published by the U.S. Department of Commerce, one can estimate the proportion of the franchisor's revenues that is attributable to the fixed fee. Total revenues to the franchisor over the duration of the contract are given by the franchise fee plus the variable fee multiplied by average sales and by the number of years contracted for. The franchise fee represents 13.7% of these total revenues and is therefore clearly not negligible.2^* Consequently, this argument can not explain the observed lack of correlation between the two fees. A final attempt was made to elucidate this relationship in Table 4.6. Instead of separating firms on a sectoral basis, size cohorts were generated. The 548 observations were sorted by the number of outlets they had, and the cutoff points for each decile were located. Because many firms often had an identical number of outlets, it was not possible to generate cohorts of 55 firms each, but the size cohorts in Table 4.6 were chosen to get as close as possible to that. As usual, N indicates the number of franchisors in each category. Because the value of the tradename can also be associated with the number of years a firm has been in business, age cohorts were generated as well. In this case however, the number of firms with the same number of years was too large. It made it impossible to get 10 age cohorts of similar size. For that reason, only 6 of them, with about 90 franchisors each, were constructed. The second column of Table 4.6 describes how the proportion of franchised stores varies across cohorts. For both size and age cohorts, this proportion seems to go up, reach a maximum around the median, and then maybe go down again. With respect to the fees, what transpires from this table is that royalty rates do seem to grow as the size of the chain, and possibly the number of years in business, increases. As both the size of the chain and the number of years in business can be used as proxies for the value of the tradename, this is consistent with predictions from the contract given in Table 4.8, to give a measure of the yearly fixed fee. This is not surprising since these durations do not vary much across sectors. 9ft The average sales figures used for this computation actually represent average yearly sales per outlet whether franchised or company-operated. Average sales per franchised outlet tend to be smaller than the overall average. Thus the proportion of revenues attributable to the fixed fee as calculated here may underestimate the true value. 71 Table 4.6 : Royalty Rates and Franchise Fees Within Size and Age Cohorts Size cohorts % Franc. Royalty Royalty F . Fee F . Fee Cor(F.r) Spearman Rank Cor N Mean Mean St. Dev. Mean St . Dev. (F.r) 1- 1 < OUTLETS < 8 58 0.697 6.6 2.9 16. 1 8.4 0.34 0.32* 2- 9 < OUTLETS < 15 51 0.730 6.6 2.8 25.9 40. 3 0.12 0.24+ 3- 16 < OUTLETS < 24 56 0.830 6. 1 3.0 20.9 15.4 0. 15 0. 11 4- 25 < OUTLETS < 35 52 0.791 6.4 3.2 19.2 13.5 0.00 0. 17 5- 36 < OUTLETS < 54 58 0.870 6.0 3.3 23.4 20.0 -0.03 0. 19 G- 55 < OUTLETS < 85 56 0.885 6.2 3.4 22.3 19.5 0.00 0.26+ 7- 86 < OUTLETS < 151 54 0.874 6.0 3.4 20.2 14.3 -0.02 -0.06 8- 152 < OUTLETS < 299 55 0.884 7.2 3.8 20.9 17.6 -0.35 -0.24+ 9- 300 < OUTLETS < 620 55 0.860 7. 1 3.2 23.3 13.4 0.01 O.05 10- 621 < OUTLETS < 9060 53 0.849 7.4 4.7 23. 1 25.5 -0.28 -0.06 Age cohorts 1- 1 < YEARS IN BUS. < 6 87 0. .841 6 .2 3 . 1 20. ,5 19 .4 0. .07 0. 18+ 2- 7 < YEARS IN BUS. < 9 91 0. ,827 6 .4 3. .2 23. 7 32. .4 -0. 13 -0. 09 3- 10 < YEARS IN BUS. < 13 91 O. .832 6 .5 3. .2 21 . 9 14. .2 -0. 05 0. 21* 4- 14 < YEARS IN BUS. < 17 83 0 .888 6 .9 4 .5 20. . 1 13 .5 -0, .08 -0. .00 5- 18 < YEARS IN BUS. < 27 lOO 0, .794 6 .5 3. . 1 20. . 1 13 .9 -O. .04 0. .05 6- 28 < YEARS IN BUS. < 170 96 0. .794 6. .7 3. .4 22. .5 21 . 9 0. .01 0. .22* Note: Two-sided test s : ** sign, at the .01 l e v e l . * sign, at the .05 l e v e l , + sign, at the .10 l e v e l . the two-sided hidden action models. Franchise fees seem to remain about the same, or grow a little as the number of outlets in the chain gets larger. The variability of royalty rates increases a little with the number of outlets, as one would expect. However this is not the case for franchise fees. As the number of years in business increases, neither fee's standard deviation goes up. Turning our attention to the correlation between the two fees, it is interesting to find that it is positive for small size chains, and then it becomes negative for chains with a large number of outlets. The way in which mean fees vary across size cohorts tends to indicate that both fees could grow simultaneously. However the simple correlation coefficients suggest that large chains must trade off one type of fee for the other, while small chains do not have to. According to the Spearman rank correlations there is a significant positive relationship between the two fees for small and medium sized firms. The relationship is negative and significant in one of the larger size cohort. The others show no relationship at all. The pattern is somewhat different for age cohorts. A significant positive rela-tionship between royalty rates and franchise fees is found in both the "youngest" and "oldest" age cohort. No age cohort shows a significant negative correlation. There is therefore no trade off between the fees even for firms that have been in business for a very long time. The average fees found for each age cohort seem relatively constant, indicating that on average firms do not tend to increase either fee as the number of years in business increases. This can be taken to mean that years in business do not provide a good proxy to the value of the tradename. Within each age cohort firms with relatively "stronger" tradenames are still able to demand higher fixed and variable fees. In general, one can only conclude that there is no simple relationship between the two fees. In most sectors and in many of the size and age cohorts, no significant relation was found at all. In a majority of the cases where there was a significant association between the two fees, it was a positive, not a negative one. However, within size classes there was some indication, from the simple correlation coefficient, of a trade off between franchise fees and royalty rates for large but not for small chains. Royalty rates did seem to increase with the size of the chain. In fact, the pattern may be U-shaped. Franchise fees remained relatively constant as the number of outlets gets larger. Neither fee really increased with the number of years in business. 73 In their model, Mathewson and Winter found that "Monitoring increases for an established firm in response to the increased temptation for the franchisee to free ride on the brand name" [p. 520]. They argued that this result was consistent with 01 observed buy-backs in North America. If we allow for a certain "critical" size or years in business for a firm to be defined as established in their model, their result is also consistent with what we observe here, i.e. the fact that the average proportion of franchised outlets goes down slightly once firms have reached the median size or age. Rubin's explanation of buy-backs was based on the notion that monitoring costs per outlet go down as density increases. To make this consistent with what is found in Table 4.6, one would also have to invoke a notion of "critical density" from which monitoring costs would start decreasing. Otherwise, given that density increases with the number of outlets, one would expect the average proportion of franchised outlets to decrease monotonically with the size of the chains. 3.4 A Comparison of the Two Samples For comparison purposes, Tables A . l to A.5 in appendix A show the same statistics for the same variables as Tables 4.2 to 4.6, but for the sample of 890 franchisors. As one would expect, the main difference between the two sets of tables shows up in the total number of outlets covered and in the number of outlets per chain. The latter goes up in the sample of 548 firms: many small franchisors were ehminated because they were not covered in the other surveys from which the additional data were obtained. The average number of years in business and years franchising are also slightly larger in the sample of 548 for the same reason. Most of the other numbers are quite similar across the two sets of tables. It was the need for additional data that resulted in the exclusion of firms, most of them relatively small, and the reduction of the sample size from 890 to 548 firms. This does not seem to have introduced any systematic bias, especially for the dependent variables. In particular it is worth noting that the average proportions of franchised outlets in most sectors in Table A.2 are very similar to those found in Table 4.3. Similarly, according to Tables 4.3 and 4.4, the average franchise contract was given by r = 6.5% and F = 21,500. In Tables A.2 and A.3, one finds an average r of 2 1 As was mentioned previously, this trend seems to have reversed itself. 74 6.5% and an average F of 21,700. The two are almost the same. In their survey, the Association of Canadian Franchisors (ACF) found an average royalty rate of 5% and an average advertising fee of 2%, resulting in an average r of about 7%. Since the survey of the A C F was done using data for 1984, and is restricted to a sample of Canadian franchisors that accounts for only about 10% of all franchised sales in Canada, it is interesting that the numbers are so similar. 3.5 M e a s u r e s o f G e o g r a p h i c a l D i s p e r s i o n a n d o f F r a n c h i s o r s ' C o n t r i b u t i o n As was discussed previously, additional data were collected for the sample of 548 franchisors. These include measures of geographical dispersion of outlets, which make the monitoring of franchisees more costly, as well as the amount of training provided by each franchisor. The latter serves as a measure of franchisors' contribution. These were available for 1985 only.2** They are described in Table 4.7. Here, States refers to the number of States each franchisor has outlets in. Excluding Equipment and Tool Rental which contains only one observation, the average number of States covered by franchisors is highest in the Carpet Cleaning category, and then for Hotels, Motels and Campgrounds. The smallest average number of States, 5.5, is found in the category of Italian Restaurants. The next two columns provide information on the number and the proportion of foreign outlets in each chain. Both the number of states and the proportion of foreign outlets are meant to capture the degree of geographical dispersion of each franchise chain. However establishing the proportion of foreign outlets was not as straightforward as one would expect. The survey produced by Entrepreneur asks franchisors for the number of franchised and company-operated outlets they have worldwide. But in many cases, the American corporation is legally distinct from its foreign counterparts and does not report foreign outlets as its own. The empirical model used here requires each data point to represent the decision-maker, i.e. the entity that chooses the optimal combination of company operated and franchised outlets as well as the optimal contract. If the foreign firms are distinct in the sense that they can make these decisions with respect to outlets in their jurisdiction, then The standard error of F however is larger for the sample of 890 than that of 548, which is not surprising. n o The existence of a lag between these variables and my dependent variables has the advantage of getting around some possible simultaneity. 75 Table 4.7 : Measures of Geographical Dispersion and of Franchisors' Contribution Sectors STATES FOREIGN ^FOREIGN TRAINING GROWTH CAPITAL OUTLETS OUTLETS In In NEEDED WEEKS OUTLETS ($000) N Mean Mean Mean Mean Mean Mean 1- AUTOMOTIVE PRODUCTS 45 12.2 13.0 0.019 3.0 0.231 60.8 PARTS AND SERVICES 10 16.4 9.5 0.006 3. 1 0.252 48.5 BRAKES, MUFFLERS S SHOCKS 7 16.9 68.9 0.036 3.6 0.090 84.4 TRANSMISSION 6 17.0 0.3 0.002 3.8 0.077 77.9 TUNE-UP 11 7.3 0.0 0.000 2.B 0.224 66.3 2- BUSINESS AIDS 8 SERVICES 96 17.7 12.5 0.022 2.2 0.300 38.3 ACCOUNTING AND COLLECTION 19 14.3 0.7 0.013 2. 1 0.332 16. 1 EMPLOYMENT SERVICES 20 15.3 0.2 0.002 3. 1 0. 184 32.5 PRINTING AND COPYING 14 21.5 23.5 0.028 3.7 0.364 75.0 REAL ESTATE 8 22.5 93.5 0.065 0.7 0.320 24.9 3- CONSTRUCTION & MAINTENANCE 50 19.8 26.9 0.062 1.8 0.292 21.8 MAID SERVICES 8 10.4 18.5 0. 124 2. 1 0.421 15.2 CONSTRUCTION 13 22.0 3.8 0.060 1.9 0.369 33.3 HOME IMPROVEMENT 8 REPAIRS 11 • 11.5 11.1 0.04B 2.0 0.360 18.9 CARPET CLEANING 9 40.4 112.7 0.083 1.6 0.083 14.5 4- CONVENIENCE STORES 8 10.9 1.9 0.002 2.7 0.040 49.5 5- EDUCATIONAL PRODUCTS & SERVICES 29 12.0 11.2 0.054 2.0 0.312 39.8 EDUCATIONAL 8 TRAINING PROGRAMS 12 10.5 2.8 0.068 2.4 0.270 32.3 HEALTH AIDS & SERVICES 17 13. 1 17. 1 0.043 1.6 0.342 45. 1 6- RESTAURANTS 121 13.3 40.3 0.024 5.5 0. 138 202. 1 FAST FOOD 84 13.2 56. 1 0.022 4.8 0. 143 170.0 - CHICKEN 11 16.0 171.4 0.027 5.1 0. 164 258.7 - MEXICAN 10 11.7 5.0 0.037 3.3 0.149 135.4 - HAMBURGER 9 24.7 279.8 0.089 7.6 0. 147 339.2 - PIZZA 22 9.4 3.5 0.009 5.0 0.159 148.7 - SUBMARINES 13 8.2 0.2 o.ooo 3.4 0. 126 59.3 TABLE SERVICE 37 13.5 4.5 0.028 6.9 0. 128 275.0 - STEAKHOUSES 6 25.2 13.8 0.033 7.9 0.016 300.2 - ITALIAN 6 5.5 O.O O.OOO 4.5 0. 109 156.7 - FULL MENU 11 12.8 6.3 0.033 10. 1 0. 162 429.7 7- HOTELS. MOTELS & CAMPGROUNDS 10 38.6 40.6 0,077 1.4 0. 135 444.9 8- LAUNDRY 8 DRY CLEANING 3 23.0 0.0 o.ooo 4.0 0.291 86.7 9- RECREATION & TRAVEL 11 19.6 4.2 0.041 1.6 0.345 108.7 TRAVEL AGENCIES 4 27.7 0.7 0.033 1.7 0.216 41.2 10- AUTO 8 TRUCK RENTALS 6 37.0 545.7 0.255 1.1 0. 167 45.5 11- EQUIPMENT & TOOL RENTAL 1 48.0 0.0 0.000 2.4 -0.111 255.0 12- NON FOOD RETAILING 99 16.3 8.2 0.035 2. 1 0.257 84 .6 CLOTHING & SHOES 16 17.3 8.0 0.026 2.1 0. 146 87.5 FURNITURE & ACCESSORIES 9 16. 1 21.6 0.082 2.2 0.201 110.6 ART SUPPLIES 8 19.0 3.2 0.020 3.3 0. 176 52. 1 COMPUTER PRODUCTS 8 30.0 27 .4 0.069 2. 1 0.477 93.9 VIDEO RENTAL 7 13.7 0.0 0.000 1.5 0.476 80.2 13- FOOD RETAILING - NON CONVENIENCE 52 12.5 38.5 0.026 2.9 0.263 84.4 DONUT SHOPS 10 11.5 41.5 0.048 4.9 0.084 98.2 ICE CREAM PARLORS 17 15. 1 93.0 0.052 2. 1 0.294 79.6 SPECIALTY FOOD SHOPS 9 13.8 0.9 0.002 3.7 0.093 75.2 14- MISCELLANEOUS 17 15.2 6.8 0.019 1.4 0.384 40.9 BEAUTY SALONS 12 18.2 9.6 0.023 1.4 0.222 42.8 ALL SECTORS 548 15.9 27.4 0.034 2.9 0.242 97.4 one should treat these firms as separate entities and not consolidate the outlets. Assuming that whether or not a firm claims the outlets as its own indicates something about the amount of control it feels it has over them, I have used the data as given in the survey. In cases where it was clear that all foreign outlets had been excluded from the reported number of franchised and company-operated units, and consequently the American firm only was the data point, the number of foreign outlets was given a value of zero. Not surprisingly, Auto and Truck Rental and then Hamburger franchisors are found to have the largest average number of foreign outlets. However, franchisors in the Auto and Truck Rental and then those in Maid Services are the ones that have the largest proportion of foreign outlets. On average, franchisors in this sample operated 27 outlets outside of the U.S., representing 3.4% of all their outlets. Training, in the fifth column, stands for the length of the initial training period and is given in number of weeks. It also varies a lot on average, from a low of 0.7 weeks in the Real Estate business to a high of 10.1 weeks for Full Menu Restaurants. Overall, franchisors provide around 3 weeks of training to their franchisees. Two other variables were obtained for the sample of 548 firms and are presented in Table 4.7. First, there is a measure of growth of the chain, given by half the difference in the logarithm of the number of outlets in 1986 and in 1984. The mean value for this variable in the various sectors is found in the sixth column.2 5 Not too surprisingly, the fastest growing sectors during this period are those of Video and Computer stores. Excluding Equipment and Tool Rental, which contains a single firm that experienced negative growth between 1984 and 1986, one finds that there has been growth in all franchising sectors. The slowest growth is found in the Steakhouses sector. At the firm level, in total, 62 of the 548 firms experienced negative growth between 1984 and 1986. Another 34 firms had zero growth during that period. All 94 For example, A & W clearly indicates that its Canadian counterpart is a distinct entity and so it does not report any Canadian outlets. And while the American firm franchises almost all its outlets (620 out of 630), A & W Canada operates 28% of its stores, i.e. 82 out of 295. The franchise fees and the royalty rates (not counting the advertising rate which is unavailable for the Canadian company) are also different for the two franchisors: F = 20000 and r — 4% for the American corporation, F = 25000 and r is between 2 and 2.75% for the Canadian Company. 2 5 Firms that had no outlets in 1984 were eliminated in the construction of the sample of 548 firms because this variable was not well defined in those cases. 77 the other chains show positive growth. The resulting average growth rate for all the firms in the sample is estimated to be around 25%, which is very high. When it was calculated on the basis of actual percentage changes in the number of outlets between 1984 and 1985, and then between 1985 and 1986, the average of these two gave an estimated growth rate of 45%. The fact that this number is large is probably due to the presence of small firms in the sample for which the addition of one or two outlets represents a very high increase in proportion to the size of the chain. The last variable presented in Table 4.7 represents the amount of capital needed to start one of their franchises according to each franchisor. Whenever a range was given in the survey, the lower bound was used as it represents the minimum amount required to start the business. It is an indication of the minimum capital constraint the firm faces when it seeks to expand. As can be seen in Table 4.7, Hotels, Motels and Campgrounds, as well as Restaurants, tend to require much more capital on average than the other types of franchised businesses. Over the whole sample, franchisors estimated the minimum capital requirement to be around $100,000. 3.6 Some Relevant Sectoral Data Useful data were found on an aggregate basis in the U.S. Department of Com-merce's "Franchising in the Economy". Again, the most recent available data were for 1985. Some of these are presented in Table 4.8. The first column, entitled Total Outlets, mainly serves to give some notion of the coverage of the samples of fran-chisors used in this paper. This is done here in terms of outlets as opposed to number of franchisors as in Table 4.1. However, the numbers are not directly comparable to those found in the first columns of Tables 4.2 and A l . One reason for this are the different years: data in Table 4.8 are for 1985, whereas all previous tables contained 1986 data. More importantly, total outlets in the U.S. Department of Commerce's publication seem to refer to U.S. outlets only. Tables 4.2 and A l were compiled using worldwide outlets. The second column presents total sales in the U.S. for each sector in millions of dollars. Restaurants remain the most important business-format franchising sector both in terms of outlets and in terms of sales. This accounts for the fact that 121 out of the 548 franchisors studied here operate in this sector. Column 3 gives the amount of average sales per outlets; this is a proxy for the average size of outlets 78 Table 4.8 Some Relevant S e c t o r a l Data T o t a l Total Average Mean Franchisees Franchisees' Discontinued O u t l e t s {%) O u t l e t s Sales Sales Agreement purchases purchases Average 1984- 1985 (5) (1) (1) per Length per o u t l e t per o u t l e t Total Co--Owned Franc. o u t l e t (In years) <K$) (% s a l e s ) O u t l e t s (2) (3) (4) (4) (U.S. 85) (MS 85) <K$ 85) (U.S. 85) (84 -85) (84--85) 1- Automotive Products 36472 10658 .6 292 15. 1 64 .65 30 .27 2 .95 1 .28 3.20 2- Business Aids and S e r v i c e s 49834 11970 . 1 240 12. 9 2 . 12 1 .01 4 .03 0 .60 4.55 Accounting and c o l l e c t i o n 2081 168 .6 81 13. 1 n . a. n . a. n . a. n. . a. n.a. Employment S e r v i c e s 4831 2732 . 1 566 12. 9 n . a. n . a. n . a. n . a. n.a. P r i n t i n g and Copying 451 1 919 8 204 18. 6 n . a. n .a. n . a. n, . a. n.a. Tax P r e p a r a t i o n S e r v i c e s 8 147 427 .3 52 8. O n . a. n . a. n . a. n . a. n.a. Real E s t a t e (*) 13862 4618 . 1 333 7. 7 n . a. n. .a. n. . a. n. . a. n.a. Mi s e e l l a n e o u s 16402 3103 .3 189 13. 4 n . a. n. . a. n . a. n . a. n.a. 3- C o n s t r u c t i o n and Maintenance 17482 4066 .7 233 13. 6 10 .55 7. . 11 3 .55 0. 77 3.66 4- Convenience Stores 15141 10839 . 1 716 11 . B 32 .40 5, .02 2 .92 2 .98 2.83 5- Educational Products & Serv. 8170 767. .5 94 14 . 5 6 .36 7 .85 4 .38 4 . 50 4.35 6- Restaurants 73892 47678 4 645 16. 8 26 . 15 4 .47 2 .41 1 , .81 2 .69 Chicken 8720 4118. 9 472 n.a I. n . a. n, . a. n. . a. n, . a. n.a. Hamburgers ft Franks 30563 23407. .4 766 n. 8 i. n . a. n, . a. n . a. n . a. n.a. P i z z a 14 174 6193. 9 437 n. a I. n . a. n. . a. n. . a. n, . a. n.a. Mex1 can 4 125 240O. 0 582 n. a i . n . a. n. . a. n . a. n. . a. n.a. Seafood 2423 1213. 0 501 n.a I. n . a. n. . a. n. . a. n. a. n.a. Pancakes. W a f f l e s 176 1 1101 . 9 626 n. a i. n . a. n. . a. n . a . n, . a. n.a. Steak 9466 8562 . 4 905 n.a I. n. . a. n. a. n. . a. n. a. n.a. Sandw iches 2660 680. 9 256 n.a i. n. . a. n. a. n, .a. n. . a. n.a. 7- H o t e l s , Motels. Campgrounds 7490 14770. 6 1972 17. 1 3 18 0. 20 4 . 18 3. 28 4 . 33 8- Laundry and Dry C l e a n i n g 2345 303. 2 129 16. 5 2. . 10 1 . 93 2. .71 0. 92 2.83 9- R e c r e a t i o n and Travel 7816 2318. 0 297 13. 7 0. 45 0. 25 1. 84 0. 39 1.91 10- Auto-Truck Rental 1 1228 5685. 6 506 18. 1 0 27 0. 11 1. 95 0. 45 2.36 1 1- Equipment Rental 2547 668. 7 263 9. 7 23. 18 10. 75 4. .24 1 . 77 4.84 12- Non-Food R e t a i l i n g 45120 20571. 3 456 13. 5 110. 27 27. 69 3. .41 1 . 83 3.97 13- (Non-Conv.) Food r e t a i l i n g 18682 1O08O. 5 540 12. 7 50. 69 10. 70 2. 25 1 . 48 2.41 14- Mi s e e l l a n e o u s 5470 942. 3 172 14 . 7 8. 34 5. 23 2. 05 3. 08 1 .90 T o t a l 301689 141320. 5 468 14. 4 30. 63 9. 49 3. 05 1 . 80 3.40 (*) Gross commissions. Source: U.S. Dept. of Commerce, F r a n c h i s i n g i n the Economy. 1985-1987. Jan. 1987 (except as s p e c i f i e d ) . (1) Table 1. To t a l excludes t r a d i t i o n a l f r a n c h i s i n g here. These numbers are not d i r e c t l y comparable to those found i n the f i r s t columns of Tables 1 and Al f o r two reasons: f i r s t , these are f o r 1985. not 1986. Second, t o t a l o u t l e t s in " F r a n c h i s i n g In the Economy" seem to be r e s t r i c t e d to U.S. o u t l e t s . Tables I and A1 r e f e r to worldwide o u t l e t s . S t i l l a comparison of these columns provides some idea of the coverage of the samples In terms of o u t l e t s . (2) Table 19. (3) Compiled from t a b l e 28; perpetual assumed to be 30 years on average, others excluded. (4) Tables 1 and 16 from F r a n c h i s i n g In the Economy. 1985-87. and Tables 1 and 14 from F r a n c h i s i n g In the Economy. 1984-86. (5) Tables 1 and 22 from F r a n c h i s i n g In the Economy. 1985-87. and Tables 1 and 20 from F r a n c h i s i n g In the Economy. 1964-86. in a sector. Hotels, Motels and Campground franchises are found to be the largest types of operation, on average, while Tax Preparation franchises are the smallest. The fourth column contains data on the average length, in years, of contracts issued in 1985 in each sector. This information was used to calculate the present value of all future fixed payments agreed to in the contract and thus to generate F. Columns 5 and 6 provide information on franchisees' average purchases from franchisors. In column 5, the value of these inputs is given in thousands of dollars. For each sector, it is calculated as the total amount bought by franchisees from their franchisor, divided by the total number of franchisees. In column 6, they are given in proportion to the average sales of franchisees in the sector.2^ In both cases the averages for 1984 and 1985, rather than just the number for 1985, were used in order to reduce the effect an unusual year might have on the measure. Since inputs sold at a price higher than marginal cost can be a substitute for royalties under certain conditions, this variable can be used to see whether firms that are already involved in franchising, and thus have incurred the setup costs discussed at Lindal Cedar Homes, will use input mark-ups as a substitute for royalties. Note however that this information tells us whether or not franchisors sell a lot of inputs to their franchisees, but it says nothing about whether or not these inputs are sold at a price greater than marginal cost; yet the latter is central to the argument. If franchisees buy a lot from franchisors because they benefit from "volume purchasing", then franchisees' purchases would not serve as a substitute way to extract rent. One should not expect any negative effect of this variable on royalty rates in that case.27 The last three columns of Table 4.8 indicate the average proportion of outlets that 9fi Consequently, they are not equal to column 5 divided by column 3 since the latter represents average sale per outlet, whether franchised or not. Also, column 3 refers to 1985 only, whereas column 6 is calculated as the average of inputs/sales for 1984 and 1985. 97 The Ozanne and Hunt (1971) survey revealed that most franchisees do not believe that franchisors' "volume purchasing" results in lower prices for them. 47% of franchisees who had to buy from their franchisors believed they were paying higher prices than they could have obtained on their own. 25% of them however believed they payed lower prices. Also 41% of franchisees said their franchisors received kickbacks from approved suppliers, while 39% said they did not. The rest did not know. In the survey done by the Retail Council of Canada and published by Info Press in May 1986, 30.8% of franchisees said that buying power was the most beneficial service provided by their franchisor. Advertising was a close second with 28.8% of the votes. The other possibilities were education (11%), store supervision (4.8%), unique or special product (15.9%) and accounting (8.7%). 80 have been discontinued in each sector, on average, in 1984 and 1985. Because of the negative effect it has on their tradename, franchisors are known to resist closing down outlets. When a store is not doing well, if the franchisor feels it has any potential, he will generally prefer to take it over rather than discontinue it . 2 8 In particular, if the poor performance is due to bad management on the part of the franchisee, the franchisor will not discontinue it. For that reason, the proportion of discontinued outlets can be interpreted as a measure of exogenous risk as opposed to endogenous risk or moral hazard on the part of the franchisee. It represents the probability that an outlet will be closed down and therefore provides a measure of "down-side" risk for the sector. The first of the last three columns in Table 4.8 refers to total outlets in the sector, whether company-owned or franchised. Educational Products and Services (which includes Health and Diet Services), Equipment Rental, Hotels, Motels and Campgrounds and finally Business Aids and Services are found to be the riskiest franchising sectors. Franchises in the Recreation and Travel and the Auto-truck Rental businesses are the least likely to be discontinued. The last two columns give respectively the proportion of company-owned and franchised outlets that have been discontinued in each sector. This break-down between company-owned and franchised outlets depends on franchisors' decisions and would therefore be endogenous in a theoretical model. Still it is worth pointing out that in a majority of sectors, i.e. eleven out of fourteen, the proportion of franchised outlets that have been discontinued is larger, and often much larger, than the proportion of company-owned outlets that met with the same fate. This lends support to the notion that franchisors use franchising to reduce the amount of risk they have to face. Here, it seems, they tend to franchise the riskiest outlets. 3.7 Alternative Measures of Risk While a bankruptcy notion provides one good way to think about riskiness, uncertainty really arises from the presence of 9{ in the demand function, i.e. the fact that Xi = f{T,li) + 0i. (4-1) See for example Thompson (1971), p.34. 81 Thus one may want to look for measures of <JQ directly. Assuming perfect information about both T and li, it is clear that Var(*0 = VarfA'i), (4.2) i.e. riskiness can be measured directly by observing the variance of sales at the outlet level. There are two ways one could think of getting a measure of the variance of theta. First, for each franchisor, one could try to obtain data on the sales level of all outlets in a given year, and then calculate their variance. This of course assumes that given T , the optimal level of local inputs, , is the same for all outlets in the chain. Otherwise the variation in would show up in the estimated variance of sales and thus overstate the variance of 9{. Since the variation in 7? could be different from one chain to the other, this would bias our measure making it smaller for those franchisors whose do not vary much. Similarly, even if the l\ are constant across outlets, assuming that the l{ vary because of moral hazard, the estimated variance of A'j will again overstate the variance of 9. In this case however, one may be able to assume that the possibility for moral hazard is the same for all firms in a given sector so that the ranking of firms within an industry would be unaffected by this bias in the measure.29 Another and better way to think about measuring risk is to look at how demand at the retail level fluctuates through time. In this case, for a given franchisor, one would want to observe the sales level of every outlet in the chain over a given number of years. Assuming that each set of observations on an outlet's sales is a random draw from a population whose variance is crj, then an unbiased estimate of erg for the franchisor would be given by 5 | = i k i n L f i (4.3) n where n is the number of outlets in the chain , and Sf is the variance of an outlet's sales through time. One advantage of this method is that there is no need to assume 2 9 In particular, assuming that f { T , k ) — a - T + b - k , the variance of Xi is V a r [ a - T + b - l i + 8 i ] which assuming that T is constant, as is done implicitely here since moral hazard is assumed only on the side of the franchisee, implies that Var(A'i) = b2 • Var(Jj) + var(0j) if ^ is independent of li so that the estimated variance of sales simply contains an extra term which is the same for all firms in a sector. 82 l\ to be the same across outlets in this case. It isn't necessary to assume that T and Z{ are constant through time for a given outlet either. If we assume instead that both T and Z? are growing, it suffices to look at how the sales level of the outlet varies around a trend. What is needed is that the variance of this error term be the same through time and across the franchisor's outlets. Because some outlets can be much larger than others, the latter may be a problem. One would then want to look at the variance of some normalized error, for example e/y, around a trend. Again, given 11, if l{ can vary through time due to moral hazard, the variance of sales Sf will overestimate the variance of 0; and thus will be biased upwards. An advantage of this method is that it can easily be used to measure "sectoral" risk. Since variations in both T and /,* are controlled for, one can think of the remaining variance as the same for all outlets in a sector.**0 Then the estimate of <TQ for the sector would be Si = (4-4) where N is the total number of outlets in the sector, and S2 is still the variance of an outlet's normalized sales around a trend. Again, however, the caveat concerning the effect of moral hazard applies. Unfortunately, neither of these two methods is really possible since figures for individual outlet's sales are unavailable. Even data on total sales for one year for each of the 548 franchisors in the sample could not be obtained. However, aggregate data on average sales per outlet for each franchising sector are published over a number of years in various issues of the U.S. Department of Commerce's "Franchising in the Economy", a yearly publication."*1 Using these data on average sales per outlet, it is possible to obtain a measure of variability by fitting a trend and then calculating the variance of a normalized error term around this trend. This procedure assumes however that Zt- and T follow approximately the same trend in all outlets of a given sector. In addition, the variance o n Note that we have already assumed such a notion of sectoral risk by saying that outlets in the same sector all face the same probability of being discontinued. Data for subsectors in the restaurant category for the years 1977 to 1982 were in fact taken from the National Restaurant Association's Franchise Restaurants: a Statistical Appendix to Food Service Trends. Their source was "Franchising in the Economy" for various years. Nominal sales level from the published data were transformed to constant 1986 U.S. dollars using the Consumer Price Index. 83 of average sales per outlet, is given by N J N2 N N-l N ^ 2 + 2 E E C ! ° v ( A ' i . ^ - ) i = l i = l j=i+l (4.5) where a2 — Var(A',). If one assumes that the AVs are independent from each other (in Chapter 2, this was assumed to be the case for all outlets of a given franchisor, here, it must be true for all outlets in a given sector), and that they all have the same variance, o~2x, then V a r ( l > V A ^ =*x/N- (4-6) Given an estimate of the variance of average sales per outlet obtained using time series data, S2t, one could then get an estimate for a2^, i.e. N S\ = Vn^Xi/N) • N = S2e< • N. (4.7) t=i But TV varies over time, so that the correction implied by (4.7) isn't that straigth-forward. Note that the same problem would have arisen in the calculation of (4.4). The most appealing way to handle this here would be to use the average value of N over the time period, N. If the A'j's are not independent from each other, i.e. if the sales level of one outlet in a sector is dependent upon the sales level of other outlets in the sector, we are back to (4.5). In this case, if all the correlations can be assumed to be positive, we have that P2x<Sl-N (4.8) so that we have an upper bound on what a2 can be.**2 Finally, if one assumes that If some of the correlations are negative, which may be the case for example for outlets that, are close to each other or for outlets in competing chains, then it becomes possible that. J V - l AT 2 E H Vov(Xi.Xj) <0, (4.9) i = l j = i + l so that <r\ > S*t • N (4.10) and we cannot, think of S% • N as an upper bound anymore. Since the A*,'s refer to all outlets in the sector in the U.S., I will assume that overall, the sum of the correlations would be positive. 84 ( sales are perfectly correlated across all outlets, which, as was mentioned in Chapter 2, is not very consistent with the fact that franchise contracts are independent contracts, then from (4.5), Si = V a r ( ^ k ^ i ) = _ L [N . 4 + { N _ 1 ) N . A 2 X ] = *x (4-11) i.e. the variance of average sales would be the same as the variance of each outlet's sales. Since correlation coefficients cannot go beyond 1, S2 clearly gives a lower bound on o\. The interpretation of measures of risk based on aggregate data, as we have here, thus depends critically on the assumptions one makes about correlations. If the average number of outlets, N, was similar across sectors, the difference between (4.7) and (4.11) would vanish. Using either one of these, one would obtain the same ordering of sectors' risk, and in fact, relative magnitudes would be the same. If N varies a lot between sectors, as is the case here, then if the A"'s are independent, i.e. (4.7) is correct, but o\ is estimated using (4.11), the "riskiness" of sectors with a large number of outlets is underestimated. If the correlations are positive and large but (4.7) is used, the riskiness of sectors with a large number of outlets on average is overestimated. Finally, for any intermediate scenario in terms of correlations, which are the most realistic, differences in N between sectors will not be the only thing influencing the bias of the measures. The "amount of correlation" between outlets' sales will also tend to vary across sectors and this will affect the validity or the bias of any measure that is actually used. In these circumstances, the best strategy is to do some sensitivity analysis. Since we can obtain upper and lower bounds for <rx, one can ask how the ordering of sectors varies when one uses one or the other. The lower bounds for <rx, i-e. the calculated standard errors of average sales per outlet, are presented in the first three column of Table 4.9. The next three columns of Table 4.9 provide the average number of outlets, the number of company-operated and the number of franchised outlets respectively. These show clearly how different Nc and Np tend to be. They are used in conjunction with the first three columns to generate estimates for the upper bound of ax, i-e Set • N, which are shown in the last three columns of Table 4.9. For all these sets of columns, the first of the three refers to all outlets, whether franchised or not, whereas 85 Table 4.9 A l t e r n a t i v e Measures of Risk S e c t o r s ST. DEV. (AV. SALES) AV. NUMBER OF OUTLETS ST. DEV.(AV. SALES)*SQRT(N) N T o t a l Co-owned Franc. Total Co-owned Franc. Total Co-owned Franc. St. Dev. St. Dev. St. Dev. mean mean mean 1- Automotive Products 9 0 .041 0 066 0 .032 40621 4355 36266 a .3 4 .4 6 . 1 2- Business A i d s and S e r v i c e s 9 0 .087 0. 039 0 . 100 43183 6241 36942 18. , 1 3 . 1 19 2 Accounting and c o l l e c t i o n 9 O .084 O. , 124 0 .068 2593 43 2550 3 .3 0 8 3 .4 Employment S e r v i c e s 9 0 .079 0 065 0, .096 4267 1304 2963 5. 2 2 .3 5. 2 P r i n t i n g and Copying 9 O .047 0. ,07 1 0 .046 3237 122 3115 2 .7 0 .8 2 .6 Tax P r e p a r a t i o n S e r v i c e s 9 0 .093 0. 111 0 .078 8894 4171 4723 8. 8 7 .2 5. ,4 Real E s t a t e (*) 9 0 . 109 0. 243 O . 122 15025 254 14771 13 .4 3 .9 14 8 Mi s e e l l a n e o u s 9 o. . 198 0. 167 0 , 174 9167 346 8821 19 .0 3 . 1 16 .3 3- C o n s t r u c t i o n and Maintenance 9 0. 126 0. 467 0, . 142 15261 587 14673 15. 6 11 .3 17 .2 4- Convenience Stores 9 0. 035 0. 060 0 047 14963 9225 5738 4 .3 5 .8 3 .6 5- Educational Products ft Serv. 9 o. 1 19 0. 153 0 . 139 4499 579 3920 8 0 3 .7 8 .7 6- Restaurants 9 0. 035 0. 039 0 .036 62680 19159 43521 8. 8 5 .4 7 .5 Chicken 9 0. 067 0. 053 0 106 7574 2420 5865 5 8 2 .6 8 1 Hamburgers ft Franks 9 0. ,047 0. 062 0 .050 27507 5664 21844 7 .8 4 .7 7 4 P tzza 9 0. 035 0. 061 0. 025 10O75 3652 6424 3. ,5 3 .7 2 .0 Mexican 9 0. 071 0. 130 0. .030 3017 1325 1693 3 9 4 .7 1 .2 Seafood 9 o. 071 0. 117 0. 061 2382 1129 1253 3, 5 3 .9 2 .2 Pancakes. W a f f l e s 9 0. 077 0. 120 0. 070 1538 448 1090 3. .0 2 .5 2 .3 Steak 9 0. 044 O. 045 o. ,056 8893 4131 4752 4 1 2 .9 3 .9 Sandwiches 9 0. .055 0. 080 0. 055 1704 341 1364 2 .3 1 .5 2 .0 7- H o t e l s , Motels. Campgrounds 9 0. 054 0. 055 0. 079 6577 1018 5559 4 . 4 1 .8 5. 9 8- Laundry and Dry C l e a n i n g 9 0. , 185 0. 282 0. 192 2908 88 2820 10. .0 2 .6 10 .2 9- R e c r e a t i o n and Travel 9 o. 191 0. 752 0. 102 5624 158 5465 14 .3 9 .5 7 .5 10- Auto-Truck Rental 9 0. , 101 0. 072 0. 087 8866 2095 6772 9 .5 3 .3 7 .2 11 - Equipment Rental 9 0. 072 0. 153 0. 068 1933 264 1669 3 2 2 .5 2 .8 12- Non-Food R e t a i l i n g 9 o. 104 0. 156 0. 101 38452 11858 26593 20 .4 17 .0 16 .5 13- (Non-Conv.) Food r e t a i l i n g 9 0. 128 o. 206 o. 164 15550 1792 13758 16 O 8 . 7 19 . 2 14- Ml s e e l l a n e o u s 9 0. 252 0. 294 0. 262 3527 339 3188 15. .0 5 .4 14 8 Tot a l 9 0. 035 0. 048 0. 036 264640 57760 206880 18. 0 11 .5 16 .4 Notes: The standard d e v i a t i o n s a re those of e r r o r terms around trends. N-9 i n a l l cases because the time s e r i e s data used Is from 1977 to 1985. the second concentrates on company-owned outlets and the third is for franchised outlets only. Again the breakdown between company-operated and franchised outlets is endogenous, but it is informative. Concentrating on the first 3 columns and on the 14 main sectors for which data concerning discontinued outlets were available, it is interesting to find that company-owned outlets are the ones that have experienced the most variation in average sales in 11 of the 14 sectors. This is exactly the opposite of what occurred in the case of discontinuations. Given the large differences in numbers of outlets however, aet • y/~N shows a pattern that is closer to that found in the last three columns of Table 4.8. In particular, it is the variance of sales of franchised outlets, once adjusted in this way, that is greater than that of companj'-operated outlets in the majority of sectors, i.e. in 11 out of 14 cases. Despite the important changes in ranking that seem to result from the adjustment based on the number of outlets however, the rank correlation coefficient calculated on the upper and the lower bound versions, across the 14 sectors, is .56, i.e. positive and significant. Thus it seems the ranking of sectors is affected by this adjustment for N but there remains some relation between the two. A measure of risk based on bankruptcies should be positively correlated with measures based on the variance of sales since increased variance of sales can be expected to lead to more bankruptcies under imperfect capital markets. Yet, rank correlation coefficients calculated between the discontinuation rates and the upper and lower bound measures of the variance of sales yielded negative, although insignificant, values (—.37 and —.31 respectively). Thus the ranking of sectors according to their degree of riskiness will be very different depending on the measure used. It is important to point out again that the variance of sales calculated here represents the variance of 6 if one assumes perfect information or no moral hazard on the part of the franchisee (or store manager) or on the part of the franchisor. If /,• and T vary across outlets, then the variance of sales will embody these variances as well as that of 6{. The component of &x that is attributable to moral hazard is indistinguishable from exogenous risk so that it is not possible to disentangle the two. Since the moral hazard problem should be more acute in some types of businesses than in others, the variance of sales will overestimate the variance of 8{ even more in these sectors than in others. Thus there is not only a bias upwards in the measure, but the extent of the bias depends on the extent of moral hazard present. For this reason, and given franchisors' reluctance to close down outlets, measuring risk by the 87 proportion of discontinuations may be preferable to relying on the variance of sales. At least, in that case, the effect on the measure of risk of moral hazard on the part of franchisees is partially controlled for. 4. C o n c l u s i o n The purpose of this Chapter was to describe how and where the data used in this study were obtained and also give a statistical summary of these data. The following few points have emerged from the latter. On average, franchisors in this sample had about 273 outlets worldwide, but the median number of outlets is about 54. In most cases firms were in business at least a few years before they began franchising. The amount of time during which firms have not been franchising as a proportion of total time in business is about 32%. This implies a relatively long lag of 6.4 years on average before existing firms began franchising. It seems to go against the explanation of franchising that is based on franchisors' need for capital. The proportion of franchised outlets tends to be high, with an average of 82.7%. Many firms, 117 out of 548, franchise all their outlets. Specialty food shops and restaurant franchisors in general tend to operate more outlets than franchisors in other sectors do. Most franchisors do not require any experience from their franchisees. Most of them do not provide any direct financial help to their franchisees either. The average royalty rate (including the advertising rate) of franchisors in the final sample of 548 firms is 6.5%, and the average fixed fee (including discounted future fixed payments) is $21,500. 37 franchisors in this sample operate with a fixed-fee contract. Only 7 of the 548 franchisors do not demand a franchise fee. No clear correlation, either positive or negative, could really be established between the magnitudes of these two types of fees across sectors. There seemed to be some trade-off between the two fees for large chains but not for small ones. Average royalty rates did increase across size cohorts but franchise fees did not. Across age cohorts, neither fee grew. The average franchisor in the sample was found to operate outlets in 15.9 States, and to have 16.8 outlets in other countries, representing 3.4% of all its outlets. It also offered 2.9 weeks of training to its franchisees. The average amount of capital required to open an outlet as indicated by franchisors was around $100,000. Using the difference between the logarithm of the number of outlets to calculate the growth 88 rate, one finds that on average, franchise chains in this sample have grown at a rate of about 25% a year between 1984 and 1986, which is very large. When the growth rate was calculated using the proportional change in the number of outlets between 1985 and 1986, and then between 1984 and 1985, and averaging the two, the resulting growth rate was even greater, i.e. 45%. Finally, risk, as measured by the proportion of discontinued outlets, was greater for franchised than for company-owned outlets in most main sectors (11 out of 14). The exceptions were Convenience Stores, Educational Products and Services, and the Miscellaneous category. This gives some support to the notion that franchisors use franchising to reduce the amount of risk that they themselves face. Alternative measures of risk, based on the variance of sales, were discussed as well. Because of possible correlation among the sales level of outlets however, no simple characteriza-tion of risk could be established in this way. But it was possible to calculate an upper and a lower bound for the variance of sales in the sectors defined by the U.S. Depart-ment of Commerce. Both of these were found to lead to a very different ranking of sectors in terms of riskiness than that implied by the discontinuation rates. This is somewhat surprising and it suggests that caution should be exercised in interpreting results derived from all these measures. 89 C H A P T E R V The Econometric Specification and Results 1. Introduction In this Chapter, the exact specification of the model to be estimated is discussed. This includes the description of how each of the indices will be measured, which is found in Section 2, and the choice of functional form and of estimation technique in Section 3. The empirical results concerning the choice of optimal contract mix and the terms of the share contract for the whole sample of franchisors are presented in Section 4. The effects of each of the indices described in Chapter 3 are examined in turn. This section also contains some general comments about the estimation results. Finally, Section 5 summarizes the findings. 2. The Definition of the Variables Two ways of characterizing each franchisor's behavior have emerged from the discussion of the empirical model. The first is given simply by ^ =qf{Iu,IltIT,lK,r) (5.1) where qf/Q is the proportion of units sold through franchised stores. The royalty rate r in this case is assumed to be exogenously given. The second version endogenizes the terms of the contract, i.e. both the royalty rate r and the franchise fee F so that we have ^ =Qf(IuJl,lT,IK) (5.2) r = r(Iu,Il,IT,IK) (5.3) F = F(Iu,Il,IT,IK). (5.4) As was mentioned previously, qf/Q can be measured by the proportion of fran-chised outlets, under the assumption that the size of franchised outlets is the same 90 as that of company-operated outlets. Quantifying things such as exogenous risk, supervision costs, and the importance of the inputs provided by the two parties to a contract, presents many challenges. This certainly contributes to the relative scarcity of empirical work in areas where these notions are prevalent in the theories. One has to rely on proxies and hope that the assumptions that are made to establish a relationship between the proxy and the concept to be measured are reasonable. Clearly, one must also be cautious in interpreting the results. Still I believe this is a worthwhile exercise. It is the only way to examine how theories based on concepts that are not easily quantifiable withstand confrontation with empirical facts. 2.1 Measuring Risk or Iu From Chapter 4, risk can be measured in three alternative ways. First, the average proportion of outlets discontinued in 1984 and 1985 in the sector in which the firm operates will be used to reflect the probability of bankruptcy. Second, the calculated value of the standard error of average sales per outlet in the sector will be used: it represents a lower bound for o~x- The third measure of risk is that given by the standard error of average sales adjusted for the number of outlets, which provides an upper bound on ax- The advantages and drawbacks of each of these were discussed in Chapter 4. As can be seen from Table 3.1, increases in risk should have a negative effect on franchisors' propensity to use franchising under the strict risk-sharing argument. That of course assumes that the franchisor is less risk-averse than the franchisee. 1 In fact, average sales per franchised and per company-operated outlet are given in Franchising in the Economy on an aggregate basis for the 14 main sectors and 14 subsectors denned in that publication. These are found to be systematically different in almost all sectors, with the average sales per franchised outlet usually smaller than those of company-operated outlets. Thus the proportion of franchised stores will in general overestimate q//Q. A correction for this was attempted. For those firms that franchise all of their outlets, qf/Q is obviously equal to 100%, so no correction was done. But for the others, the proportion of franchised stores was multiplied by the ratio of average sales per franchised outlet (q~j) to average sales per outlet (Q) found in the sector in which the firm is classified. Since (qf /q~f)/(Q/Q) is equal to the proportion of franchised stores, this correction provided a measure of qf/Q. Because only aggregate data on q~f /Q were available for this correction, it proved to be unsatisfactory. In general, the qualitative results remained the same. It simply introduced an additional "sectoral" component in the dependent variable. For that reason, it was not used. 91 Franchising may also be used by risk-neutral franchisors as a means of providing insurance as well as incentives to risk-averse franchisees in a model based on one-sided hidden action. Increased risk, given the costs of supervision, tips the balance in favor of more company ownership. Thus in both types of models, risk is expected to reduce the tendency of firms to rely on franchising. 2.2 M e a s u r i n g F ranch i sees ' S u p e r v i s i o n C o s t s o r Ii The cost of supervising franchisees and the importance of /; are proxied by a few different variables. First, measures of geographical dispersion are used as a means to capture increases in monitoring costs. These include the proportion of foreign outlets in each franchise chain, and the number of States each franchisor has outlets in. An interesting measure of the importance of the downstream operator's inputs (i.e. 82/Si in Eswaran and Kotwal (1985)) would have been the value added per unit of output at the outlet level. Unfortunately, no data are available on this. However, data on average sales per outlet and on inputs sold by franchisors to franchisees can be found on a per sector basis in the U.S. Department of Commerce's Franchising in the Economy. Thus a measure of the scope of the franchisees' responsibilities is given by the average sales per outlet minus inputs sold by franchisors per outlet, as a proportion of average sales per outlet, i.e. (av.sales — inputs)fav.sales., for the sector in which the firm belongs.^ In addition, a measure of the managerial skills required of the franchisee is provided by the average sales per outlet in the sector. This is because the extent to which the franchisee will be called upon to manage rather than simply participate in the day-to-day operations will increase as the size of the outlet increases. The franchisee gets to specialize in management and supervision activities, which are the type of activities that are difficult to evaluate. Since the average sales volume should be correlated with the size of the outlet, the importance of the "managerial" role of the outlet operator should increase with the level of sales. Inputs here refer to franchisees purchases from franchisors as given in Table 4.8, column 5. Average sales refer to franchises only, contrary to the information found in the third column of the same table. Again this information was obtained in Franchising in the Economy, 1984-86 and 1985-87. The measure of the franchisee's jurisprudence was calculated as one minus the proportion of inputs to average sales. Obtained values for 1984 and 1985 were averaged. 92 Finally, a dummy variable that denotes whether or not the franchisor requires potential franchisees to have previous experience in the business is taken as another indicator of the importance of franchisees' inputs. All of these variables are such that they increase as the importance of l{, or the cost of monitoring its provision, goes up. Given that in Table 3.1 J; is expected to increase franchisors' use of franchising under both the one-sided and the two-sided hidden action models, these variables should all have a positive effect on qj/Q in order to give support to these types of models. 2.3 Measuring the Franchisor's Contribution or IT The franchisor's role in business-format franchising is twofold. First, he provides a tradename and sees to it that its value is preserved or enhanced. This includes ongoing advertising as well as the monitoring of both franchisees and outlet managers. Second, he assists his franchisees in starting up and managing their businesses. As a way to measure the latter, the number of weeks of initial training specified in the franchise contract will be used. It gives an indication of the amount of knowledge the franchisor has and tries to communicate to his franchisee. For the former, one would want to measure the amount of resources the franchisor puts into maintaining the value of the tradename. Thus one would want to know such things as the franchisor's advertising budget or the number of people the firm hires specifically to supervise franchisees and outlet managers. Unfortunately, these were not available. For that reason, measures of the value of the tradename, rather than measures of the effort franchisors put into its maintenance, are used in what follows. Since the tradename is of greater value the larger the number of outlets displaying it, the number of outlets in the chain will be used as a measure of the value of T, and thus of the franchisor's contribution, IT- Because franchisors are supposed to use franchising less the more valuable their tradename under the two-sided hidden action models, this variable should have a negative effect on qf/Q in order to give some support to these models. Note that as a consequence of the portfolio effect, franchisors face less risk per outlet the more outlets they have. For that reason, as the chain expands, the risk premium to the franchisor included in pf would be reduced. On the basis of the risk-sharing explanation then, franchisors may use franchising less the more outlets they have. Both of these interpretations of the effect of the number 93 of outlets lead to a negative sign and therefore it will not be possible to distinguish them empirically. On the other hand, it has been argued that franchise chains may contain a few company-operated stores for historical reasons. In other words, franchisors would open a few company-operated stores at the beginning of their operations in order to develop their concept. But from then on, all additional outlets would be franchised. If this was the way in which franchised chains evolve, the proportion of franchised stores in a chain at any point in time would simply reflect the development stage of a given franchisor. All of them ultimately would become almost fully franchised. In this case, empirically, one would observe a positive correlation between the number of outlets in a chain and the proportion of franchised stores. The value of the tradename is also assumed to be larger for well-established franchisors. Thus the number of years in business is used as another proxy for IT- Finally, the proportion of total time in business during which the firm did not franchise (i.e. the number of years in business minus the number of years since it began franchising as a proportion of the number of years in business) is taken as an indicator of how costly the development of the franchise package was. If this proportion is high, it is an indication that getting involved in franchising took a long time, i.e. it was costly to this firm. The franchisor's inputs are more valuable the higher these variables are, and therefore, from Table 3.1, these should have a negative effect on qf/Q if two-sided hidden-action models are to be supported by the data. 2.4 Measuring the Franchisors' Capital Constraint or IK Franchisors' need for capital is assumed to be especially severe in the early years of the franchise's development and during periods of rapid expansion. The more binding this constraint is to the franchisor, the more it would be expected to resort to franchising as a means of obtaining capital. Thus the number of years in business, which was taken as a measure of the value of the tradename, also provides information on how binding the franchisor's capital constraint may be. The larger the number of years in business, the less binding one would expect it to be, and thus this variable should have a negative effect on qf/Q under a capital-market-imperfection argument for franchising. This was also the case when it was taken as a measure of the value 94 of the tradename. These two interpretations of the effect of the number of years in business are indistinguishable empirically. Growth in the total number of outlets in the last two years provides another measure for 1%. It is a proxy for franchisors' desired growth. In addition, the amount of capital required to start a franchise according to franchisors gives an idea of how binding the franchisor's capital constraint might be. The greater a franchisor's desired growth is, and the higher the cost of starting each outlet, the more this franchisor would need to rely on franchising to obtain capital if capital-market-imperfection arguments are to explain franchising. Finally, information about whether or not franchisors provide financing to their franchisees will be utilized. Clearly, the franchisor-capital-constraint hypothesis can-not be used to explain the use of franchising when in fact it is the franchisor who provides financing to his franchisees. Thus one would expect that growth, and possi-bly years in business, will have no effect on the propensity to use franchising of those franchisors that provide financing. As an intercept dummy, financing should reduce firm's propensity to use franchising if the capital-market-imperfection explanation for franchising is correct. Compared to other firms, those that provide financing have one less reason to franchise outlets. However, the fact that they do provide financing to franchisees suggests that they are eager to get franchisees for reasons other than those suggested by the capital-market-imperfection argument. Thus it would not be too surprising if this dummy variable had a positive coefficient. 2.5 Some Other Relevant Variables The value of the variable fee is included in the estimation of equation (5.1). This is because this formulation of the model was based on the assumption that the terms of the contracts are determined exogenously or that they are "quasi-fixed" by the time the franchisor decides on qf/Q. In this case, higher variable fees are expected to have a negative effect on qf/Q as the franchisee-incentive effect and the risk premium for the franchisor are smaller the higher r is. Thus all else equal, franchisors will find franchising less attractive or profitable as r increases (i.e. pf is lower when r is larger), and they will resort to company-owned outlets relatively more often and we n^ve „ , _ • « g a S a (5,) 95 This implies that the priors given in Table 3.1 for the qf/Q equation are made even stronger when r is endogenized. Because the indices are expected to affect r and qf/Q in opposite directions, we have, for example in the case of Iu, d(qf/Q) = d(qf/Q) | d(qf/Q) dr dlu dlu dr dlv u (5.6) for the model where r is endogenized. With JJ^ > 0 and — < 0, ^9Jj^ is clearly negative, as Jj is. In the same way, the expected effects of the other indices are only reinforced by the endogenization of r. The sale of inputs to franchisees at a price greater than marginal cost can be a substitute for royalties. In fact, as was noted previously, the two are perfect substitutes when downstream firms use the franchisor's inputs in fixed proportion to output and the upstream firm controls the downstream price. Since these sales can be interpreted as an additional source of revenue for the franchisor, one would want to treat them as an additional decision variable for the franchisor. However, data on the value of these sales in each franchise chain are not available. Only aggregate data could be obtained. For this reason, rather than treat these sales as an additional dependent variable, they will be introduced in the estimated equations in order to control for their potential effect on the two fees as well as on qf/Q. Given that these sales are an additional source of revenues for the franchisor, as their value increases, one can expect the royalty rate and the franchise fee to be reduced. Given the trade-off between qf/Q and r discussed previously, if increased input sales are equivalent to higher royalties, the effect of these sales on qf/Q should be negative. Note however that the value of these sales gives no information on whether or not the inputs are sold at a price greater than marginal cost. Thus the observed effect of the variable should be interpreted with care. Summary descriptive statistics for all the variables over the whole sample are presented in Table 5.1. For those variables that were calculated on the basis of aggregate data, means and standard errors are therefore weighted as a function of the number of firms from the sample that belong in each sector. Appendix B contains the correlation matrix for these variables. Histograms for the three dependent variables are found in Appendix C. 96 T a b l e 5.1 D e s c r i p t i v e s t a t i s t i c s f o r the 548 f r a n c h i s o r s V a r i a b l e N Mean S t . Dev. Minimum Maximum % F r a n c h i s e d 548 82 .75 21 . .55 1 . 62 100. ,00 Av. % D i s c o n t i n u e d (1) 548 3 . 13 0. .75 1 . 83 4. 38 A d j . S t . D e v . ( A v . S A l e s ) (1) 548 1 1 . .75 6. .50 2. .27 20. ,39 S t . D e v . ( A v . S a l e s ) (1) 548 0. . 10 0. ,05 0. 03 0. ,25 F o r e i g n O u t l e t s (%) 548 3. .36 10. .25 0. .O 96. .63 Number of S t a t e s 548 15. .95 16. 29 1 . 00 50. ,00 ( A v . S a l e s - I n p u t s ) / A v . S a l e s (%) 548 88. .96 10. .82 69. .73 99. .89 Av. S a l e s / O u t l e t (1) 548 3 . 94 2. .80 0. .52 19. .28 F r a n c h i s e e E x p e r i e n c e 548 0. .30 0. 46 0. .0 1 . 00 Weeks o f t r a i n i n g 548 2. .95 2. .55 0 .0 19. .00 O u t l e t s i n 1986 (100's) 548 2. .73 7 . 84 0 .02 90. .60 % Time Not F r a n c h i s i n g 548 31 .55 27 .22 0 .0 96. .30 Y e a r s i n B u s i n e s s 548 17. .82 14 . .41 3 .00 1 10. OO Growth i n o u t l e t s 548 0. .24 0. .33 -0. .28 1 . 87 F r a n c h i s o r F i n a n c i n g 548 0 .21 0. .41 0, .0 1 . 00 C a p i t a l R e q u i r e d ($K) 548 97. .45 172. .91 0. .50 2000, .00 F r a n c h i s o r I n p u t s ($K) (1) 548 38 .39 38 .67 0, .27 1 10, .27 V a r i a b l e Fee (%) 548 6. .54 3. .42 0, .0 25. ,00 ( 1 ) : Data a v a i l a b l e o n l y 1n a g g r e g a t e form (per s e c t o r ) from the U.S. Department Commerce's F r a n c h i s i n g In the Economy. 1984-1986 and 1985-1987. Means and s t a n d a r d d e v i a t i o n a r e t h e r e f o r e weighted as a f u n c t i o n of the number of f i r m s i n the sample which b e l o n g t o each s e c t o r . 97 3. The Model Specification 3.1 Functional Form The theoretical models do not provide any information about functional form. There are two main options to consider in this case. The first would be to assume a typical functional form for the profit function (3.12), and then assume one for r, Pf and pc in (3.19), (3.10), and (3.11), and ultimately for the way each index relates to the proxies used to measure it. Functional forms for g|, q* and qj/Q* in (3.16), (3.17) and in (3.18), or in (3.21), (3.22) and (3.23) could be derived from these. For example, assume that the upstream production process involves two inputs with prices u>\ and W2- As was discussed previously, these input prices are assumed to be the same for all firms in the sample. Using a quadratic profit function, and normalizing with respect to 1L2, we have il(pc,pf,wi,w2) = Ii(pc,pf,wy) = aipc + a2pf + a 3 w i + pupl + (322pj + WnPcPf + l\pcw\ + i2PfW\ + 73«'i (5.7) where pc — pc/w2, Pf = Pf/u>2, a n d W\ = w\/w2- Since input prices are fixed, w\ — TZ>I . Under these conditions, using Hotelling's lemma, we get dU Of = ^ = OL2 + 2022Pf + 2/3i2TJC + 72^1 (5.8) and an qc = -x— = <*i + 2/?i ipc + 2012P/ + 71W1. (5.9) OPc Now, assuming a linear form for pf, pc and r, one can write Pf = ao + a\Iu + a2Ii + a 3 / j - f a^Ix + 0 5 9 / + aetfc (5.10) Pc = bo + bilu + b2Ii + b3lT + b±Ix + hqf + b6qc. (5-11) It can be shown, after manipulations, that the resulting q^ and q*c would be linear functions of the indices. Thus, if one assumes also a linear relationship between the 98 indices and the proxies used to capture them, then and qAc become hnear functions of the proxies as well.'* And so does r by assumption. The alternative approach would be to specify a functional form for the proportion q//Q directly, as was done for r in the previous scenario, since it is known to be an arbitrary function of the indices. In that a first approximation, one would want to use a linear form for consistency with the first approach. Note that functional forms for the r and F equations are chosen arbitrarily in both cases. The main advantage of the first method is the fact that it is derived in a way that is consistent with economic theory. However, it leads to simple functional forms for qc and qf, but not for qf/Q. In fact, q*f/(qf - f q*c) is a very nonlinear function of the proxies under this scenario. This functional form for qf/Q would be very costly to estimate within a Tobit framework, especially given the large number of explanatory variables. The theoretical models discussed in Chapter 2 have implications concerning qf/Q rather than qf and qc. Thus a major advantage of the second approach is that it allows simple functional forms to be used for the qf/Q equation which is clearly of more interest theoretically than are the qf and qc equations. At a statistical level, because qf/Q is the dependent variable in that case, the second approach has the added advantage of reducing heteroscedasticity problems.4 While in an ordinary least squares framework heteroscedasticity simply leads to a loss of efficiency, in a Tobit model, it leads to inconsistency of the estimates.5 Since there is no strong reason to believe that the functional forms that were assumed in the first approach are correct, the proportion of franchised stores as well as the two fees will be estimated using simple linear and logarithmic specifications within the Tobit model. Still, assuming again that franchised and company-operated Note however that the constant terms in the q~ and the q"j equations would be functions of 71 and 72, which one would expect to be different across sectors. I will come back to this in Chapter 6. 4 Whether or not there was any heteroscedasticity problem in the estimation of equations (5.1) to (5.4) was tested using the method suggested by Maddala (1983), i.e. by introducing cr, = a + f3Xi rather than <?i = a for all i = 1... n in the likelihood function, and then testing 0 = 0. In this context, the Xi are any of the explanatory variables the variance of the error term might be related to. Using both the number of outlets and the number of years in business, the hypothesis that the /3's were not different from 0 could not be rejected in all four cases. Thus there was no heteroscedasticity in these cases. However, disturbances from the qf and the qc equations were found to be heteroscedastic. 5 See Maddala (1983), 178-182. 99 stores are of the same size on average, qf and qc can be approximated by the number of franchised stores, nf, and the number of company-operated outlets, nc. Thus results obtained for ny and nc under a linear specification (and corrected for heteroscedasticity when appropriate) are reported in Appendix E for comparison purposes. Since the number of franchised outlets is never nil in this sample by its very nature, it is not necessary to use a limited dependent variable framework to estimate the nf equation. However, whenever the proportion of franchised stores is at its limit, this implies that there are no company-operated stores. Thus the Tobit estimator must be used for the nc equation. 3.2 Estimation Technique Given measures for each of the indices, the estimation procedure must take into account the fact that many firms included in the sample, i.e. 117 out of 548, franchise all of their outlets. Similarly, 37 firms use a fixed-rent contract (i.e. r — 0), while only 7 rely on a pure share contract (F = 0). Thus, for all the equations to be estimated, there is some degree of censoring, i.e. observations on the dependent variables that take on limit values. This censoring problem is clearly illustrated in. the histograms, in Appendix C. In fact, since all the firms that choose not to franchise are excluded from the sample, it could be argued that there is also a truncation problem in the estimation of (5.1) and (5.2). I will treat the decision firms make to franchise or not as a separate issue from the one studied here. This is because I believe that to most firms in the economy, franchising is not really an issue. In addition, as was pointed out by the managers of some of these firms, there are costs associated with getting involved in franchising. These include the development of the franchise package, and of the disclosure statements required by law in most States in the U.S. Because of these, firms have to invest first in franchising before they can determine the value of franchised versus company-operated outlets. Only then are they in a position to take the decisions analyzed here. The observed distribution of the proportion of franchised stores gives support to this notion. The average proportion of franchised stores is quite large, i.e. 82.6%, indicating that when firms use franchising, they tend to use it a lot. This is also seen in the histogram for the proportion of franchised stores presented in Appendix 100 C, in figure CI. Very few franchisors tend towards 0 franchised stores. Thus even if a correction for truncation were included in the estimation, the difference would be minimal empirically given that so few firms are close to the lower limit. In the empirical work on share contracts described in Chapter 3, the depen-dent variable was often denned as the proportion of land under sharecropping or the proportion of franchised or of company-operated stores. Those who took into con-sideration the limited nature of this type of variable used a log-odds formulation to get around this problem. In other words, they estimated a logit model where the proportion they observed was interpreted as a statement of probability. In the case of the proportion of franchised stores for example, this would mean that a franchisor's outlets are viewed as a group of observations. For each such group, the probability that an outlet will be franchised is estimated by the actual percentage of franchised outlets. In the logit model, it is the logarithm of the ratio of the probability of be-ing franchised and the probability of not being franchised that is then related to the explanatory variables. In the empirical model presented in Chapter 3, the proportion of franchised stores is derived as the result of the franchisor's decision-making process in a context where all outlets are assumed to be identical. This proportion could have been interpreted as the probability that an outlet from a given chain will be franchised here as well. One advantage of this interpretation is that it allows for differences among outlets. In this case, pf and pc depend on the values of the indices measured at the outlet level. Then, for each outlet, pf can be compared to pc and on that basis, the franchisor decides whether to operate or franchise the outlet. In this framework, the proxies used for the indices for each franchisor would represent averages for all of their outlets. And one would observe that the probability that an outlet be franchised in a given franchise chain would depend on these averaged indices. The equations to be estimated wTith respect to the proportion of franchised stores would be the same as those derived in Chapter 3. In their paper, Brickley and Dark (1987) found that outlets that are physically close to monitoring units have a higher probability of being franchised. That would lend support to this interpretation of the indices as averages. Despite the fact that the proportion of franchised stores could be interpreted as a probability, the logit model can not be used to get around the censoring problem. This is because there are many observations at the limit. Authors who have used the logit model to avoid issues related to the boundedness of the dependent variable were 101 either using aggregate data, which substantially reduces the probability that one gets a limit observation, or, like Martin (1987), had a sample of firms that did not include firms that are 100% franchised. Since this sample includes such firms, the maximum likelihood Tobit estimator is used. Consequently, the estimated model in the case of equation (5.2) is given by (gf/Q)i = Qf [(iuh Vih(hh (iK)i] + » i i if RHS < IOO = 100 otherwise. (5-12) Equation (5.1) is the same except for the inclusion of r on the right hand side. In the case of the two fees, the limit observations are at zero and we have ri=r[(Itl)i,{Il)1J{IT)i,(IK)i}+u2i i f R H S X ) = 0 otherwise (5.13) and Fi = F[(Iu)i, (/,),-, {IT)i, (Ijc)i} + u3i if RHS > 0 = 0 otherwise. (514) The model can be estimated by maximum likelihood. The likelihood function is composed of two parts to allow for observations at the limit and observations off of it. Assuming normally distributed errors, and that the limit for the dependent variable is a lower limit at 0, the logarithm of the likelihood function would be written as LLF = log(l + £ logfa/v) (5.15) o 1 wThere Y^ 0 and refer to summations over all the limit observations and over all the non-limit observations respectively. $i and (j>i are respectively the cumulative density function and the density function of the standard normal distribution, evaluated at Zi/a, i.e. - i = e - ' / 2 ^ (5.16) -co ' & = - L c - * ( £ ) 2 (5.17) 102 where Z{, in the case of equation (5.14) for example, is given by Z , = F [ ( J U ) , , ( / , ) , , ( / r M W , (5.18) which under a linear functional form simply becomes Z{ = a0 + a i / u . + a2Ii{ + a3ITi + aiIKi. (5.19) When there is an upper bound on the dependent variable, and it is different from zero, as in the case of the proportion of franchised stores, the problem can be rewritten in such a way as to give it a lower bound at zero. This is done by multiplying (5.12) by minus 1, and then subtracting the limit value from the resulting equation. For comparison purposes, although these are known to be biased in a limited dependent variable context, the OLS estimates when limit observations are excluded were calculated as well. These can be found in Appendix D. 3.3 Covariance Terms Equations (5.2), (5.3), and (5.4) will be estimated separately here. Given that there can be trade-offs between r and qf/Q and between r and F, one could argue that the estimation technique should allow for non-zero covariances among the error terms of the three equations. This is not done in this case for the following reasons. First, while there is theoretically a trade-off between r and F, no clear correlation was found between these two variables in Chapter 4. Similarly, from Appendix B, the correlation coefficient between the proportion of franchised stores and the variable fee r is found to be -0.07. The rank correlation is -0.02, which is not significant at any reasonable confidence level. Thus it is not clear that one should expect the error terms from the equations for qf/Q, r and F to be correlated. More importantly, in a non-limited dependent variable context, one incorporates nonzero covariance terms in the estimation stricly to gain efficiency. In addition, there is no more efficiency gain from doing this if the equations all contain the same explanatory variables, as is the case here. Whether or not the introduction of nonzero covariances also loses its efficiency effect when the explanatory variables are the same across the equations in a Tobit context is unknown. Probably, it does not. But since there is very little censoring in the case of the two fees, and since the Tobit 103 estimator reduces to OLS under normality when there are no limit observations, one would expect that the efficiency gain to be had by introducing nonzero covariances here would be fairly small. On the other hand, the computations would become significantly more complex and costly if one was to allow for such covariances. Thus I do not believe it is warranted in this case. Finally, at a technical level, assuming that r and F follow a joint normal dis-tribution, truncated at 0 in both cases, specifically allows for firms that have a zero franchise fee and a zero royalty rate. Although some firms in the Entrepreneur sur-vey were found to do this, they were excluded from the sample on the argument that they were not really business-format franchisors. In other words, no fixed fee and no variable fee was argued to be inconsistent with what franchising is. 3.4 Simultaneity Problems Many of the proxies used to measure each of the indices could be thought of as endogenous to decisions concerning the contractual mix or the contract design. For example, one could think that geographical dispersion and growth may increase for those firms that use franchising more. Similarly, the proportion of discontinued outlets may depend on the contract design. Past values were used for all explanatory variables except the number of outlets, the number of years in business, the proportion of time during which the firm was not franchising, the fees and of course the dummy variables. In other words, the number of States each franchisor has outlets in is in fact measured as the number of States the firm had outlets in in 1985 rather than 1986. The same is true of the proportion of foreign outlets. Similarly, growth and the proportion of discontinued outlets for example are measured over two previous years. Thus the variables used here can be interpreted as instruments for the proxies, rather than the proxies themselves. Still, in interpreting the empirical results, one should keep in mind that there may be some endogeneity problems. 4. The Empirical Results Results from the estimation of equations (5.1) and (5.2), which are concerned with firms' propensity to use franchising, are presented in Tables 5.2 and 5.3. Table 5.2 contains those results that were obtained assuming a linear relationship between 104 the proportion of franchised stores and the explanatory variables. However, in Table 4.6, it was found that the proportion of franchised stores, as well as the franchise fee and the royalty rate, may be nonlinearly related to the number of years in business and the number of outlets in the chain. In order to allow for some curvature, regression results were calculated using the natural log of those independent variables for which it is well defined, which includes both the number of years in business and the number of outlets. These are presented in Table 5.3. The qualitative nature of the results remains basically unaltered.1^ The results that pertain to the terms of the franchise contract are given in Tables 5.4 and 5.5. Again, the first of these contains estimates obtained using a hnear functional form within the Tobit model, whereas Table 5.5 is concerned with those results that were derived using the logarithmic form for the explanatory variables whenever possible. As was mentioned above, Appendix D provides tables that correspond to Tables 5.2 to 5.5, but in this case the equations are estimated by Ordinary Least Squares after excluding the limit observations. Although these estimates are known to be biased in general, it is interesting to find that they are not very different from the Tobit estimates. This is not too surprising however given that the amount of censoring is relatively small, especially in the case of the fees. It is also worth noting that the Tobit results in Tables 5.2 to 5.5 are not sig-nificantly affected by the removal of firms with a very large number of outlets, or number of years in business, or number of weeks of training, or an unusually high capital requirement or variable fee. Thus the results presented in these tables are not due to a few firms having a disproportional effect on the parameters. All of the results presented here were obtained under the assumption that fixed-rent contracts (i.e. those that have r = 0) are just a special case of franchise contracts. In other words, franchisors choosing a fixed-fee contract are assumed to not behave differently from those that opt for a share contract when it comes to the proportion of stores they decide to franchise, and to the factors that affect this decision. A test ^ The linear version of the qj/Q equation was reestimated with the square of both the number of years in business and the number of outlets among the regressors to allow non-linearities for these two variables in a different way. Again the effect of the number of outlets was negative, and the effect of the number of years in business was positive. In both cases, the square of the variable had the opposite sign from the variable itself, indicating that the slopes were decreasing in absolute value. The shapes for the qj/Q equation suggested by this version were consistent with those implied by the partially logarithmic version for both variables. 105 Table 5.2 : Proportion of Franchised Outlets under a Linear Spec i f i c a t i o n % Franch. % Franch. % Franch. % Franch. % Franch. % Franch. Av. % Discontinued 3 .60 ** 3 .69 ** (2 .61) (2 .67) St. Dev. of Theta 0 .39 * 0 .40 * (2 .02) (2 . 10) St. Dev. of Av. Sales 1 . 15 2 .51 (0 .06) (0 .12) Foreign Outlets 16 .83 17 .35 + 16 .02 16 .51 16 .96 17 .53 + (1 .63) (1 .67) (1 .54) (1 .58) (1 .63) (1 .68) Number of states 0 .33 ** 0 .33 ** 0 .33 ** 0 .33 ** 0 ,36 ** 0 .36 ** (4 .44) (4 .44) (4 43) (4 .43) (4 .77) (4 .77) (Av.Sales-Inputs)/Av.Sales 0 .00 0 .06 -o .22 -0 . 16 -0 . 13 -o .07 (0 .01) (0 .27) (-0 .94) (-0 .72) (-0 .55) (-0 .29) Av.Sales/ outlet ($100K) -1 .06 ** -1 .08 ** -0 .87 * -0 .88 * -1 .06 * -1 .06 * (-2 .72) (-2 75) (-2 . 17) (-2 .18) (-2. .53) (-2 .54) Franchisee Experience -1 .21 -1 .30 -0 .60 -0 .68 -0 .55 -0 .64 (-0 57) (-0 .61) (-0 .28) (-0 .32) (-0 .26) (-0 .30) Weeks of Training - 1 . .93 ** -1 .93 ** -1 .93 ** -1, .93 ** -2 . 17 ** -2 . 17 ** (-4 77) (-4. .77) (-4. .68) (-4 .68) (-5 . 19) (-5 19) Outlets 1n 1986 (100's) -0. .29 * -0. .32 * -0. .31 * -0 .34 * -0, .33 * -0 .36 * (-2. .03) (-2. 26) (-2. 19) (-2 .44) (-2. .30) (-2 57) % Time Not Franchising -0, .33 ** -0 .34 ** -0 .32 ** -0 .33 ** -0 .32 ** -0 .33 ** (-7. 89) (-8. 14) (-7. .75) (-8 .00) (-7 .63) (-7, .88) Years in Business 0. . 12 0. . 14 + 0. . 12 0, . 14 + 0. . 12 0, . 13 (1 .55) (1. 72) (1 55) (1 .73) (1. 43) (1 .62) Growth 1n out l e t s 6. ,94 * 7. 46 * 6. 19 + 6. .70 * 6. ,83 * 7. .40 * (2. 14) (2 .33) (1. .90) (2. .08) (2 .09) (2 .29) Capital Required -0. .02 * -0. ,02 * -0. ,02 ** -0. .02 ** -0. .02 ** -0 .02 ** (-2. .56) (-2. 56) (-2. ,64) (-2. .64) (-2. .69) (-2 .69) Franchisor Financing 4. 64 + 4. 40 + 5. 45 * 5. .23 * 5. ,57 * 5. .32 * (1. 90) (1. 81) (2. 25) (2. 16) (2. ,29) (2. ,20) Franchisor Inputs ($K) -0. 02 -0. .00 -0. 12 + -0. . 10 -0. .06 -0. .04 (-0. 30) (-0. 05) (-1. 66) (-1. .49) (-0. 88) (-0. 64) Variable Fee -0. 31 -0. 31 -0. .35 (-1. 07) (-1. 08) (-1. 21) Constant 89. 69 ** 81 . ,44 ** 118. 72 ** 111. . 11 ** 114. ,05 ** 105. ,51 ** (3. 61) (3. 45) (5. 10) (5. .02) (4. 88) (4. 74) Limit Observations 117. 00 117. 00 117. 00 117. .00 117. 00 1 17 ,00 Non-Llm1t Observations 431. 00 431. 00 431. 00 431. 00 431 . 00 431. 00 Standard error of Estm. 20. 59 20. 61 20. 63 20. 65 20. 67 20. ,70 At Mean X(I), E(Y) 82 . 99 82. 97 82. 97 82. 96 82. 93 82. 92 Log Likelihood Function -2007. 63 -2008. 20 -2009. 00 -2009. .58 -2011. ,04 -2011. .77 Squared Cor(Y, E(Y)) 0. 35 0. 35 0. 35 0. 35 0. 35 0. 35 LR test ( a l l slope coef.=0) 221 . 06 ** 219. 92 ** 218. 32 ** 217. 16 ** 214. 24 ** 212. 78 ** ** s i g n i f i c a n t at the 0.01 l e v e l , * s i g n i f i c a n t at the 0.05 le v e l , + s i g n i f i c a n t at the 0.10 level Table 5.3 : Proportion of Franchised Outlets under a P a r t i a l l y Logarithmic Specification Log Av. % Discontinued Log Adj. St. Dev.(Av.Sales) Log St. Dev.(Av.Sales) Foreign Outlets (%) Log Number of States (Av.Sales-Inputs)/Av.Sales Av.Sales/ outlet ($100K) Franchisee Experience % Time Not Franchising Weeks of Training Log Outlets in 1986 Log Years in Business Growth in Outlets Franchisor Financing Log Capital Required Log Franchisor Inputs ($K) Variable Fee Constant Limit Observations Non-Limit Observations Standard error of Estm. At Mean X(I) . E(Y) Log Likelihood Function Squared Cor(Y, E(Y)) LR test ( a l l slope coef.=0) % Franch. X Franch. % Franch. % Franch. % Franch. % Franch 6 .29 6 . 39 ( 1 .32) ( 1 .34) 1 . 16 1 .29 (0 .74) (0 .83) -0 .69 -0 .44 ( -0 .35) (-0 .22) 12 . 30 12 .69 1 1 .52 1 1 .82 12 .17 12 .51 ( 1 .22) ( 1 .26) ( 1 . 14) ( 1 .17) (1 .20) ( 1 .23) 3 .93 ** 4 . 14 *' 3 .92 ** 4 .11 4 .11 ** 4 .31 (3 .38) (3 .62) (3 .34) (3 .55) (3 .50) (3 .71) -0 . 14 -0 . 12 -0 .22 -0 .20 -0 .25 -0 .24 (-0 .81) ( -0 .70) (- 1 .34) (-1 .22) ( - 1 .63) (-1 .52) -0 .96 *' -0 .94 * -0 .91 * -0 .89 * -0 .94 * -0 .91 (-2 .63) ( -2 .58) ( -2 .50) ( -2 .44) (-2 .52) (-2 .44) - 1 .23 - 1 .27 -0 .84 -0 .89 -0 . 78 -0 .82 ( -0 .58) (-0 .60) ( -0 .40) ( -0. 42) (-0 .37) ( -0 .39) -0 . 34 *• -0 . 34 " -0 . 34 '• -0 . 34 •* -0 . 33 ** -0 .34 (-7 .71) ( -7 .91) (-7 .72) (-7. .93) ( -7 .63) ( -7 .85) - 1 .75 ** - 1 .75 ** - 1 .72 '* - 1 . 72 ** -1 .81 ** -1 .80 ( -4 .23) (-4 .22) ( -4 .13) (-4 .11) (-4 .22) ( -4 . 18) -0 . 35 -0 .59 -0 .46 -0 .69 -0, .53 -0 . 77 ( -0 .34) (-0 .60) (-0 .45) ( 0 .70) (-0 .52) ( -0 .78) 4 . 18 * 4 .40 ' 4 .52 * 4 .75 * 4 .29 ' 4 .55 (2 .24) (2 .37) (2 .41) (2 .55) (2 .29) (2 .45) 8 . 13 * 8 .63 * 8 .06 * 8 .53 ' 8 . 14 • 8 .65 (2 . 33) (2 .50) (2. .31) (2, .47) (2, .34) (2 .51) 2. .87 2, .66 2, .98 2, .78 2, .90 2 .68 ( 1 .  16) ( 1 .08) ( 1 . . 20) ( 1. . 13) ( 1 , . 17) ( 1 , .09) -3, . 53 •* -3, .58 '* -3 .69 ** -3 .73 ** -3, .78 " -3 .84 ( -3 .87) (-3, .93) ( -4. 09) (-4 . 14) (-4, .21) ( -4 .27) - 1 . .41 - 1 . , 26 -2, ,31 + -2, . 18 + -2 .36 + -2, .21 ( -0. 99) (-0. 89) ( - 1 . ,86) (-1. .77) ( - 1 . ,90) ( - 1 . .79) -0. 28 -0. 27 -0. ,30 ( -0. 99) (-0. 95) ( - 1 . 05) 108. .88 ** 103. ,90 ** 121 . ,94 *' 116. ,72 " 126. .97 ** 122. .61 (4. ,55) (4. 44) (5. 87) (5. ,83) (6. .58) (6. .51) 117. 00 117. 00 117. 00 117. 00 117. 00 117. .00 431 . 00 431 . 00 431 . 00 431 . 00 431 . 00 431 . 00 20. 28 20. 30 20. 29 20. 31 20. 28 20. .31 83. 07 83. 06 83. 07 83. 06 83. 06 83. 05 1999. 69 -2000. 18 -2000. 29 -2000. 73 - 2000. 50 -2001. ,05 0. 37 0. 37 0. 37 0. 37 0. 37 0. 37 236. 94 •* 235. 96 *• 235. 74 •• 234. 86 *• 235. 32 ** 234. 22 ** s i g n i f i c a n t at the 0.01 l e v e l , * s i g n i f i c a n t at the 0.05 level, + s i g n i f i c a n t at the 0.10 level Table 5.4 : Royalty Rates and Franchise Fees under a Linear Specification Variable Fixed Variable Fixed Variable Fixed Fee Fee Fee Fee Fee Fee Av. % Discontinued -0 .31 0 . 34 (- 1 . 38) (0 .26) Adj. St. Dev.(Av.Sales) -0 .07 * 0 .09 (-2 . 17) (0. .49) St. Dev.(Av.Sales) -5 . 37 39 .22 * ( - 1 .61) (2 .03) Foreign Outlets (%) -1 . 15 0 .08 -0 .95 -0 .17 - 1 .09 -0 . 18 (-0 .77) (0 .01) (-0 .64) (-0. .02) ( -0 .73) (-0 .02) Number of States -0 .00 0 .08 -0 .00 0 .08 -0 .00 0 .06 (-0 .43) (1 .20) (-0 .24) (1 . 14) ( -0 .40) (0 .92) (Av.Sales-Inputs)/Av.Sales -0 .20 ** -0 .13 -0 . 18 * * -0 . 16 -0 . 18 ** -0 .26 (-5 .50) (-0 .61) (-4 .75) (-0 .76) (-4 .71) (-1 .21) Av.Sales/ outlet ($100K) 0 .07 -0 .44 0 .04 -0 .40 0 .03 -0 . 17 (1. . 12) (-1 .21) (0 . 56) (-1 .06) (0, .51) (-0, .45) Franchisee Experience 0. . 26 1 . .69 0, .21 1 .75 0. . 18 1 , 90 (0 .76) (0, .85) (0 .61) (0 .89) (0, .53) (0, .97) Weeks of Training 0. .04 0, . 77 * 0 .02 0 .81 * 0. .02 1 , .00 • (0. .59) (2, .04) (0 .25) (2 .08) (0, .37) (2 .58) Outlets in 1986 (100's) 0. .09 *• -0. , 17 0 .09 •• -0 . 17 0. .09 ** -0. . 14 (4, . 17) (-1. ,28) (4, .20) (-1 .28) (4, . 19) (-1. . 12) % Time Not Franchising 0. .03 ** -0. .01 0, .03 •* -0, .01 0. ,02 " -0. ,01 (3 .90) ( -0, 33) (3, .91) (-0. .33) (3 .81) (-0. .34) Years in Business -0. .04 •* -0. 03 -0. .04 *' -0, ,03 -0. .04 ** -0. ,03 ( -3. 34) (-0. ,42) (-3, .41) (-0. .40) (-3. .34) (-0. .37) Growth in Outlets - 1 . ,67 *' 4. , 16 -1, . 56 ** 4. .02 -1 . 59 ** 3. 52 (-3. .21) (1. .39) (-2 .99) (1. .33) (-3, .04) (1. .17) Capital Required 0. .00 0. 02 ** 0. .00 0. .02 " 0. .00 0. .02 * (0, ,07) (3. .70) (0, .09) (3. .70) (0 . 14) (3. .71) Franchisor Financing 0. 82 ' 5. ,84 ** 0. .77 " 5. . 88 *• 0. ,74 ' 5. 93 • (2. 15) (2. ,64) (2. 05) (2. .68) (1. ,97) (2. .71) Franchisor Inputs ($K) -0. 05 *• -0. 05 -0. 04 ** -0. 07 -0. 05 ** -0. 08 (-5. 13) (-0. 88) (-3. 53) (-1. 07) (-4. 57) (- 1 . ,32) Constant 27. 07 ** 28. 52 24. ,12 31 . 98 24. 31 *• 36. 84 + (7. 06) ( 1 . 31) (6. 73) ( 1 . 57) (6. 76) ( 1 . 81) Limit Observations 37. 00 7. 00 37. 00 7 . 00 37. 00 7. 00 Non-Limit Observations 511 . 00 541 . 00 511 . 00 541 . 00 511 . ,00 541 . 00 Standard error of Estm. 3. 40 19. 84 3. 40 19. 83 3. 40 19. 76 At Mean X(I) , E(Y) 6. 48 22. 77 6. ,48 22. 77 6. ,48 22. ,76 Log Likelihood Function -1400. 60 -2391. 20 - 1399. 21 -2391. , 12 - 1400. 27 -2389. 18 Squared Cor(Y, E(Y)) 0. 12 0. ,06 0. ,13 0. ,06 0. , 12 0. ,07 LR test ( a l l slope coef.=0) 66. 10 •* 36. 72 ** 68. ,88 ** 36. ,88 ** 66. 76 ** 40. 76 • oo o ** s i g n i f i c a n t at the 0.01 l e v e l , * s i g n i f i c a n t at the 0.05 level, + s i g n i f i c a n t at the 0.10 level Table 5.5 : Royalty Rates and Franchise Fees under a P a r t i a l l y Logarithmic Specification Variable Fixed Variable Fixed Variable Fixed Fee Fee Fee Fee Fee Fee Log Av. % Discontinued -0 .42 1 .47 (-0 .53) (0 . 33) Log Adj. St. Dev.(Av.Sales) -0 . 68 ** 1 .54 (-2 .67) ( 1 .07) Log St. Dev.(Av.Sales) -0 .99 ** 3 .89 * (-3 .04) (2 .13) Foreign Outlets (%) -1 .21 0 .09 -0 .85 -0 .92 -0 .91 - 1 . 18 (-0 .80) (-0 .01) (-0 .57) (-0 .11) (-0 .61) (-0 . 14) Log Number of States -0 .79 *• 0 .67 -0 .72 ** 0 .52 -0 .69 ** 0 .29 ( -4 . 19) (0 .65) (-3 .80) (0 .50) (-3 .66) (0 .28) (Av.Sales-Inputs)/Av.Sales -0 .07 ' -0 . 19 -0 .09 ** -0 .17 -0 .06 * -0 .24 + (-2 .55) (-1 . 13) (-3 .27) (-1 . 10) (-2 .35) (-1 .65) Av.Sales/ outlet ($100K) -0 .05 -0 .51 -0 .05 -0 .49 -0 .09 -0 .34 ( -0 .78) (-1 .51) ( -0 .90) (-1 .46) (-1 .49) (-0 .98) Franchisee Experience 0 . 12 0 .63 0. . 13 0 .66 0 .09 0 .75 (0 .35) (0 .32) (0, .37) (0 .34) (0 .26) (0 .39) Weeks of Training 0 .02 0 .56 -0 .01 0 .62 -0 .04 0 .79 * (0 .30) (1 .47) (-0, .11) (1 .60) (-0 .53) ( 1 .98) Log Outlets in 1986 0. .87 '* -0. . 26 0. .85 •* -0 .23 0 .82 ** -0 .09 (5, .26) (-0. .29) (5. .20) (-0, .26) (5, .06) (-0 . 10) % Time Not Franchising 0. .02 *• -c. .01 0. ,03 ** -0 .01 0. .03 '* -0 .01 (3. .48) (-0 .22) (3. .71) (-0 .31) (3 .71) (-0 .37) Log Years in Business -0. .83 ** -0. .80 -0. ,94 •* -0, .55 -0. .92 ** -0. .44 (-2. .70) (-0. .47) (-3. 05) (-0, .32) ( -3. 03) (-0 .25) Growth in Outlets - 1 . .77 *• 5. ,07 -1 . ,74 ** 4. .99 -1 . 72 *" 4, .84 ( -3. 07) (1. .58) ( -3. 03) (1. .56) (-3. .00) ( 1 , .52) Log Capital Required 0. 23 5. ,09 ** 0. ,21 5. . 14 *• 0. .21 5. . 18 * ( 1 . 58) (6. ,06) (1. 42) (6. .20) (1. .49) (6, ,29) Franchisor Financing 0. 83 * 6. ,90 ** 0. 81 * 6. .96 ** 0. .80 * 7. ,01 * (2. 10) (3. 10) (2. 05) (3. ,13) (2. 05) (3. ,17) Log Franchisor Inputs ($K) -0. 52 * -2. 63 ' -0. 47 * -2. 82 * -0. 47 ' -2. ,81 * (-2. 22) (-2. ,01) (-2. 31) (-2. ,47) (-2. .30) (-2. ,47) Constant 13. 99 ** 24. 02 16. 44 ** 20. , 15 10. ,28 ** 37. 99 * (3. 70) ( 1 . 14) (5. 20) ( 1. 15) (3. 54) (2. ,34) Limit Observations 37. 00 7. 00 37. 00 7. 00 37. 00 7. 00 Non-Limit Observations 511. 00 541 . 00 511 . 00 541 . 00 511 . 00 541 , 00 Standard error of Estm. 3. 46 19. 40 3. 44 19. ,38 3. 43 19. 32 At Mean X(I) , E(Y) 6. 49 22. 67 6. 48 22. 67 6. 48 22. ,66 Log Likelihood Function -1408. 38 -2378. 97 - 1404. 95 -2378. 45 -1403. 90 -2376. 76 Squared Cor(Y, E(Y)) 0. 09 0. 11 0. 11 0. 11 0. 11 0. 11 LR test ( a l l slope coef.=0) 50. 54 *' 61 . 18 *" 57. 40 ** 62. 22 ** 59. 50 ** 65. 60 * o ** s i g n i f i c a n t at the 0.01 l e v e l , * s i g n i f i c a n t at the 0.05 level, + s i g n i f i c a n t at the 0.10 level of this assumption was performed by doing separate regressions for firms that use a share contract and for those that use a fixed-rent contract. The hypothesis that the coefficients are the same for the two groups could not be rejected. However, 43% of the firms choosing a fixed-fee contract franchised all of their outlets, as opposed to 24% overall in the sample (117 out of 548). In addition, the average proportion of franchised stores in a chain from this group is 93.9% as opposed to 82.7% for the whole sample, and 81.9% for those firms using a share contract only. This is consistent with the argument that firms that opt for fixed fee contracts tend to have relatively little monitoring problems on the franchisor side. Consequently, they operate very few stores if any.7 In the remainder of this chapter, I will discuss, in turn, the empirical results concerning the effect of each of the indices on firms' propensity to use franchising and on the terms of the franchise contracts. General comments with respect to the estimation results are found in the last part of the following section. Whether one looks at results from the linear or the partially logarithmic func-tional form, they do not differ much across the two versions (5.1) and (5.2). The introduction of the variable fee among the regressors does not really affect any of the other coefficients. It also does not explain firms' choices of contract mix in any significant way. Consequently, there will be no need to distinguish these two versions in general in the discussion of results. 4.1 The Effect of Risk, Iu In Tables 5.2 through to 5.5, risk is measured as the proportion of discontinued outlets in the first two columns, and then by the adjusted standard deviation of sales around a trend, i.e. an upper bound on o\\-, in the next two columns, and by the unadjusted standard deviation of sales, or lower bound on o~x, in the last two columns. Despite the fact that they are very different in their ordering of sectors in terms of riskiness, these measures of risk are all found to have a positive effect on franchising or no effect at all.** They also systematically have a negative effect on the variable fee and a positive effect on the fixed fee, although these are again not always measured with enough precision to make them significantly different from 0. Thus it seems that 7 See Section 3.3, Chapter 2. ^ The one case where the effect is negative, it is clearly not significant. 110 as the amount of risk franchisors face increases, they do not choose to franchise less. They also do not increase the royalty rate but rather reduce it. This is an interesting result. As was discussed earlier, under the pure risk-sharing argument, and assuming that franchisors are less risk-averse than franchisees, they should decrease their use of franchising, and/or increase r, as the amount of uncer-tainty increases. And since franchisors are generally larger and better established than their franchisees, one would expect them to be less risk-averse. In addition, franchisors have many outlets. Being more diversified in this way, the average risk per outlet is less to them. Consequently, it is difficult to really believe that fran-chisors are more risk-averse than their franchisees.^ In general then, it appears that the predictions concerning the effect of risk in the pure risk-sharing models of share contracts are not borne out in these data. Risk was also supposed to have a negative effect on firms' tendency to use franchising, and a positive effect on their royalty rate, under one-sided hidden-action models. Thus the type of explanation that says franchisors use franchising to provide both insurance and incentives to their franchisees is also not supported by these data. Of course these interpretations depend on the capacity of the measures to capture exogenous risk, as opposed to the variability that is due to moral hazard. But even if they do, in models with asymmetric information, increases in riskiness automatically compound the unobservability problem. The franchisor's capacity to evaluate the franchisees' performance depends critically on the variance of 9. When there is only a little exogenous risk, the franchisor can assess franchisees' behavior to a large extent. But when uncertainty in the environment is great, they can not. This confuses issues and coidd affect the observed effect of risk on the proportion of franchised stores as well as the royalty rates. In particular here, results obtained with respect to the risk variables could be reinterpreted to say that increased unobservability of franchisees implies a greater reliance on franchising. Still this implies that incentive problems overwhelm risk-sharing arguments in the context of franchising. 4.2 T h e Effect of F ranch i see S u p e r v i s i o n C o s t s , // Under the assumption that franchisors with only a few outlets are more likely to be more risk-averse than their franchisees, one can examine this issue further by dividing the sample of firms into subgroups based on the number of outlets they have. This is done in Chapter 6. I l l In Table 5.2 as in Table 5.3, the regression results are consistent with the idea that as outlets get farther apart, it becomes more difficult to monitor outlet operators and franchising, which gives more incentives to downstream operators than wage contracts do, becomes more appealing.1^ Franchisors' choices of qf/Q increase significantly with both geographical dispersion variables. This is also consistent with Brickley and Dark's (1987) results at the outlet level. Thus one may conclude that geographical dispersion increases franchisors' propensity to franchise. Results concerning the terms of the contract are also consistent with the notion that as supervision becomes costlier, franchisors will want to use a contract that gives more incentives to outlet operators. The royalty rate, which is the portion of the franchisee's revenue he pays to his franchisor, is systematically reduced as the geographical dispersion of outlets increases. In most cases, the fixed fee is increased consequently, however this effect is never significantly different from zero. The variable that is interpreted as a measure of the scope of the franchisee's responsibilities, or of the amount of jurisdiction he has in his store, i.e. (Av.sales — inputs)/Av.sales, has very little effect on the proportion of franchised stores. As a measure of the importance of this variable should have a positive effect on qf/Q- As it turns out, in some cases its coefficient is negative, and in some cases it is positive, but it is always measured too imprecisely to be clearly different from zero. On the other hand, while the effect of this variable on the use of franchising is basically nil, its effect on the royalty rate is clearly and significantly negative. Given that franchisors can opt for an increase in qf/Q or a reduction in r, the results support the notion that as the scope of the downstream operator's responsibilities increases, the franchisor chooses a franchise contract that provides franchisees with more incentives. However, the effect of this variable on F is always negative, though not significantly different from zero in general. This again seems to contradict the notion that there is a trade-off between the two fees. Finally, both the experience and the average sales per outlet variable have a significant negative effect on qf/Q. They also have a correspondingly positive effect on the variable fee in most cases, though again this is not measured with enough precision to really be distinct from zero. Under the hypothesis that the cost of supervising managers increases the more important their role is, these variables should 1 0 See Rubin (1978). 112 have had a positive influence on the proportion of franchised outlets, and a negative one on the variable fee. One way to rationalize the fact that we observe the opposite can be found in Eswaran and Kotwal (1985). All landlords are assumed to have access to an elastic supply of tenants at a given opportunity wage, u. Landlords' opportunity income is denoted by v. In the model, increases in the importance of the tenant's work increase the tendency to use share or even fixed-rent contracts, given v/u. This is the result on which the expectations concerning the effect of the experience and the average sales per outlet variable were based. However, reductions in v/u, all else equal, tend to increase the use of fixed-wage contracts and increase the optimal share paid to the landlord. Here, franchisors require franchisees to have previous experience in their field of business. One would expect that, compared to the whole population of potential franchisees, experienced people command a higher opportunity wage u. Similarly, in sectors where the outlets are fairly large, the managers' responsibilities are increased and their opportunity wages would be larger than average. The problem arises from the fact that there is no variable in the estimated equations to control for probable differences in the opportunity wage of potential franchisees. Thus variables that were intended to measure differences in the value of the franchisees' inputs, given u, really measure differences in franchisees' market value. The empirical results in Tables 5.2 to 5.5 provide some evidence to the effect that increases in u, i.e. reductions in v/u, do lead to a greater reliance on fixed-wage contracts. On the other hand, geographical dispersion is not correlated with u. These variables actually measure differences in monitoring costs and their negative coeffi-cients can be taken to mean that for a given u, higher supervision costs lead to more franchises. In the same way, since the size of the outlet is controlled for when one calculates (Av.Sales — inputs)/Av.sales, it is more likely to be fixed, and one actually observes the effect of changes in the importance of the franchisee's inputs when using this variable. 4.3 The Effect of the Franchisor's Contribution, IT Increases in the value of the franchisors' inputs should have a negative effect on qf/Q and a positive influence on r in order to give some support to two-sided 113 hidden-action models. The number of weeks of initial training, which proxies the amount of assistance provided by the franchisor, has a significant negative effect on qf/Q in Tables 5.2 and 5.3. With respect to the terms of the contract, this variable generally has a positive effect on the royalty rate, as expected, although the coefficient is never really different from zero. While differences in (av. sales - inputs)/av. sales affected mainly r, and not qf/Q, we have the opposite in this case. Franchisors rely on changes in their contract mix rather than in the variable fee when the amount of training they provide varies. Increases in the number of weeks of training is one of the few variables that really affects the fixed fee in a significant way in most cases, and its effect is positive. This is not what was expected from Table 3.1. However, given that a certain portion of the fixed fee is often required by franchisors in exchange for training, the fact that F increases with this variable is not too surprising. It does imply however that caution must be exercised when interpreting results based on the fixed fee. For example, the inclusion of this type of fee for service in F could certainly contribute to the lack of correlation between F and r observed in Chapter 4. The number of outlets in the chain in 1986, which proxies the value of the tradename, also has the effect on qf/Q and on r one would have predicted on the basis of two-sided hidden-action models. As was mentioned previously, this effect could also be attributed to a reduction in the amount of risk per outlet faced by the franchisor as the number of outlets increases. However, it is clearly inconsistent with the notion that the proportion of franchised stores one observes at any point in time simply reflects the development stage of a firm that will ultimately become completely franchised. The proportion of total time in business during which the firm has not been in-volved in franchising was taken to be positively correlated with the cost of developing the franchise package. It is found to have a significant negative effect on qf/Q, and a significant positive effect on r, as expected. Thus the higher this proportion, the higher is the cost of getting involved in franchising to the firm, and the lower are Pf and qf/Q- It could be argued instead that the effect of this variable reflects the fact that there are costs associated with changing qf/Q, so that a firm that has been operating for a long period of time before getting involved in franchising necessarily will have a low value of qf/Q. It is true that there are adjustment costs associated with changing the contract mix, and that they may be partially captured by this 114 variable. It is also true that franchisors sometimes do buy back a number of outlets or franchise company-operated stores in a way that significantly alters their propor-tion of franchised stores within relatively short periods of time. More importantly, this adjustment cost argument can not account for the significant positive effect this variable has on the royalty rate. But the interpretation of this variable as a proxy for the value of the tradename can. The last variable used to reflect the value of the tradename is the number of years a firm has been in business. However, this variable has a significant positive effect on qf/Q in most cases and a correspondingly negative effect on the royalty rate r. This variable also tends to reduce F but in this case the coefficients are never really different from zero. Whether it is interpreted as a measure of the value of the tradename, in the context of two-sided moral hazard, or as an indicator of the franchisor's need for capital, its effect on qf/Q (r) should have been negative (positive).11 In order to rationalize this result in a two-sided hidden-action framework, one could argue that an increased number of years in business implies a higher opportunity wage for the franchisor so that v/u rises and there is a tendency to move away from fixed-wage contracts, i.e increase the number of franchised outlets, as a result. But then a larger number of outlets should imply a higher value for v/u as well, and therefore have a positive effect on qf/Q- One reason may be that the portfolio effect implies a negative coefficient for the number of outlets variable but not for the number of years in business variable. Combined with the positive effect implied by the higher value of v in the two-sided hidden-action framework, the net effect may be negative or nil for "outlets" but positive for "years". Another, and I believe more satisfactory, explanation for this may have to do with reputation. Note that if there was no possibility of franchisor free-riding, as in Section 3A of Mathewson and Winter (1985), then the royalty rate ((1 — /) in their model) would be larger for new than for established firms. In other words, there would be a negative relationship between years in business and royalty rates. In this model, this is due to ^ Since this variable was meant to capture the value of the tradename, the number of years in business rather than the number of years franchising was used. This is because the tradename can certainly increase in value as long as the firm is operating, whether or not it franchises. However, the number of years since the firm began franchising was found to have the same effects on the dependent variables as the number of years in business did. 115 the fact that when the brandname is new, franchisees have little incentives to free-ride on it given that it is of little value. As the franchisor becomes better established, the value of the tradename increases, and free-riding becomes more and more attractive to franchisees. It is necessary to increase franchisees' share of their outlets' profits in order to increase their incentives and prevent them from free-riding on the now more valuable brand name. Hence the negative relationship between royalty rates and years in business in this setting. This result, which contradicts those of two-sided moral-hazard models, depends crucially on the hypothesis that the franchisor can not free-ride. One way that the franchisor can guarantee his future behavior, i.e. assure franchisees that he will not free-ride, is by establishing a reputation. This however is not allowed for in two-sided hidden-action models. 0 Thus assume that a franchisor, through the years, can establish a reputation with respect to the value of its tradename. In other words, holding the number of outlets fixed, a greater number of years in business gives franchisees assurances that the value of the tradename will be maintained in the future. High royalty rates and company-owned outlets would no longer be necessary as a guarantee to franchisees that the franchisor will not free-ride. Consequently, royalty rates could be reduced and the proportion of franchised stores could increase as the number of years in business increases. And given the franchisee-incentive effect, they would. It is possible that a large number of outlets may not provide an equivalent guarantee to franchisees. This is because the cost to the franchisor of upholding the value of the tradename increases with the number of outlets he must oversee. At the same time, the benefits he can expect from the future sale of outlets decrease as the number of existing outlets increases. Both of these effects make defaulting more attractive to the franchisor as the number of outlets grows, given the number of years in business. Still, it is interesting to see that depending on how one decides to measure the value of the tradename, either as a function of how long it has existed or how prevalent it is, one gets very different results. One can only infer from this that there is no simple way to characterize long term trends in franchising. In particular, firms that are involved in franchising do not seem to tend toward either franchising all of their outlets or toward 100% company-owned chains. This can be seen in Appendix F which shows how the predicted values for the proportion of franchised outlets of the "average" franchisor in this sample change as the number of outlets and the number 116 of years in business are allowed to vary from zero to their maximum values. This graph was obtained using the coefficients of the partially logarithmic version of the model with risk measured by the proportion of discontinued stores and the variable fee excluded from the estimation, i.e. Table 5.3, column 2.^ Whether it is true that the number of years in business provides a better measure of reputation than the number of outlets does is, to an extent, debatable, but this would explain the empirical results obtained here. These issues will be examined further in the next chapter where results based on age and size cohorts of firms will be discussed. 4.4 T h e Effect o f F r a n c h i s o r s ' C a p i t a l C o n s t r a i n t s , IK As was already mentioned, the coefficient on the number of years in business variable being positive in the qf/Q equations, it does not support the notion that newer firms use franchising more because of their need for capital. One reason may well be that until they have established their tradename, it is difficult for them to attract franchisees and their capital. However, the coefficient on the growth variable is positive and significantly different from zero, indicating that firms do use franchising more during periods of expansion. This result should not be interpreted simply to mean that small firms use franchising more, despite the fact that measured growth tends to be quite large for relatively small franchise chains. From Table 4.6, it is clear that the proportion of franchised outlets tends to be smaller on average in small firms than in larger ones. In addition, the effects of the growth variable on the royalty rate and the franchise fee are also consistent with the capital-market-imperfection explanation of franchising. The effect of the amount of capital required according to franchisors in order to open one of their outlets has a significant negative effect on the proportion of franchised stores. It does not seem to significantly influence the choice of the variable fee, but it does have a significant positive effect of F. The latter result is clearly consistent with the notion that firms use franchising to get access to capital, but the negative effect on qf/Q is not. Both Brickley and Dark (1987) and Scott (1987) 19 Goldberg also obtained opposite effects on the proportion of franchised stores for these two variables. However, in his case, qj /Q increased with the number of outlets, and decreased with the number of years in business. 117 obtained a similar negative relationship between the use of franchising and the amount of capital required. Consequently, arguments based on imperfect capital markets can explain the positive relation one gets between F and the amount of capital required, but they can not account for the negative relationship between qf/Q and capital requirements. Note however that Brickley and Dark (1987) expected this negative correlation between the two variables. They argue that higher capital requirements imply more investment risk to the franchisees, who then demand higher risk-premia. Thus fran-chising becomes less attractive to franchisors. They also suggest that high capital requirements probably imply that the franchisee must invest a lot in firm-specific as-sets. Consequently, he or she is more vulnerable to the franchisor's post-contractual behavior. This in turn increases contracting costs, reducing pf. While this argument can explain the negative relation observed between the proportion of franchises and capital requirements by relating it to risk rather than capital-market arguments for franchising, it is difficult to rationalize the positive correlation between F and the amount of capital required in this case. One way for the franchisor to reduce the total initial investment the franchisee must make is to ask for a smaller fixed fee. This way the risk-premia required by franchisees would be reduced. Yet franchisors do not do this. One possibility for the positive relationship between the fixed fee and capital requirements in the context of the argument used by Brickley and Dark would be that the amount required to start a franchise already includes F. In other words, there might be a causality problem. But the correlation coefficient between the two variables is only .16. Also, when the franchise fee is subtracted from the amount of capital required, this amount is negative for 121 of the 548 franchisors. Thus there is not much evidence that the amount of capital required, as reported by franchisors, includes F.1"* Consequently, arguments based on risk can account for the negative relationship observed between the use of franchising and capital requirements, but not for the existence of a positive relation between F and the amount of capital required. The fact that firms provide financing to their franchisees seems to have a positive effect on qf/Q- If franchisors generally used franchising as a means to get access to capital, one would not observe firms providing capital to their franchisees. If any were ^ According to the survey description, it is not supposed to either. 118 to do that anyway, one would expect these to use franchising less since they do not exploit one of the benefits from franchising. The fact that this dummy variable has a positive effect on qf/Q suggests that those firms that provide financial assistance to their franchisees find franchising especially profitable for reasons that have nothing to do with a need for capital. It is also possible that the fact that a franchisor is willing to invest his own capital in the outlet may reduce the amount of risk as perceived by franchisees. In this case the risk-premia required by franchisees would be smaller, and thus pf would go up relative top c . Consequently, franchisors who provide financing would use franchising more. An intercept dummy does not allow for any differences in the effect of other variables, such as growth and years in business, between firms that do and firms that do not provide financing. However, when separate sets of regressions were performed for these two types of firms, it was found that the coefficients were not significantly different across the two groups for any of the four equations (5.1) to (5.4).^ Still it is interesting to notice that the proportion of firms that provide financial assistance to their franchisees and that franchise all of their outlets is much higher, at 38.0%, than the 21.4% found for the whole sample (and 17.6% for firms that do not provide financing). Franchising has to be a very attractive option to these firms for this to be the case. One can only draw very tentative conclusions from this last set of results. The effect of growth on the use of franchising, and the positive correlation between the fixed fee and the amount of capital required, give some support to the notion that franchising may be used to obtain capital. But the evidence is not clear, especially given the negative effect of the amount of capital on the use of franchising, and the fact that franchisors who provide financing to their franchisees still seem to find franchising very attractive. The positive effect of growth could well be interpreted as a sign that franchisors use franchising to relax some form of constraint they are facing, but it need not be a financial one. For example, it could be a constraint on ^ The likelihood ratio test values were 19.34, 19.9, 15.16 and 28.42 respectively for each of the four equations for the linear case. Thus the coefficients for the fixed fee were found to be different in this case. However, when some of the explanatory variables were expressed in logarithmic form, which is the preferable form in the case of the fixed fee, the likelihood ratio test values were 19.88, 20.06, 13.32 and 22.58. None of these are significant at the .05 level. 119 their time. It has also been suggested that the recruitment of franchisees may be less costly to firms than that of managers. During expansion periods, franchisors could resort to franchising more often if this was the case.^ Whatever the reason, in the trade literature, one often finds testimony from franchisors to the effect that they decided to franchise because they wanted to expand faster. If franchising allows them to attain this higher desired growth level, it must relax some form of constraint as compared to expansion through company-operated stores. 4.5 The Effect of Input Sales and Royalty Rates As was mentioned previously, from Tables 5.2 and 5.3, it is clear that when they are assumed to be exogenously given, royalty rates do not contribute much to the explanation of the proportion of franchised stores. The coefficient is always negative, as predicted, but it is never measured with enough accuracy to distinguish it from zero. This suggests that a model in which r is assumed to be exogenously given to the firm is unsatisfactory. However, the amount of inputs franchisors sell to franchisees does have a negative and generally significant effect on q//Q. Since it can be interpreted as a substitute for royalties, it is consistent with the notion of a trade-off between variable fees and firms' use of franchising. The effect of this input variable on both fees is negative and generally significantly so. This supports the notion that these sales are a substitute fee. They represent another component of the fee structure in the franchise agreement, one which is not as clearly spelled out and not as easily measured as the other two. This provides another reason for the lack of correlation between the reported royalty rates and franchise fees. However, as was noted in Chapter 4, even when this variable is controlled for, the rank correlation between the two fees remains insignificant. 4.6 Some General Comments This section contains some general comments concerning the empirical results. The first thing to note is that the variation in the proportion of franchised stores across firms is explained to a certain extent by differences in the exogenous variables in Tables 5.2 and 5.3. From Tables 5.4 and 5.5 however, it is clear that the contract 1 5 See for example Norton (1988). 120 terms are much less responsive to the same explanatory variables. The Hnear version seems to be more appropriate for the variable fee and the version with some of the explanatory variables in logarithmic form fares better with respect to the fixed fee. But in both cases, the explanatory power of the estimated equations remains low. It appears that franchisors choose to adjust to differences in risk, supervision costs, capital constraints, etc., principally by changing their degree of reliance on franchising rather than by modifying the terms of their franchise contract. Because theoretical models of franchising typically concentrate on a single fran-chisor-franchisee pair, their comparative statics results center on the terms of the franchise contract. But franchisors that operate many stores have an additional instrument at their disposal. They can and do modify the proportion of stores they choose to franchise. One explanation for the fact that franchisors modify qf/Q rather than r and F may lie in the fact that each franchisor tends to use a single contract with all of its franchisees. Consequently, those contracts are not meant to be responsive, for example, to outlet-specific variables. Thus outlets that are further away may, more often than not, be franchised, as Brickley and Dark (1987) found, and so qf/Q will increase with geographical dispersion. But a franchisor will not offer different franchise contracts according to distance, so the same variables would not have much effect on r and F. The fact that each franchisor relies on a unique franchise contract, and adjusts, at least in part, through their choice of contract mix, suggests the existence of a significant cost associated with the development of franchise contracts. It may also reflect the fact that franchisors have other means of extracting rent from franchisees beside the franchise fee and the royalty rate. One of these other control variables the franchisor has, i.e. the sale of inputs to franchisees by franchisors, has been taken into account at the empirical level. However, since input requirement clauses found in franchise contracts are generally the same for all franchisees in a given chain, they may imply greater fees for all, and in particular greater variable fees, but they do not really allow the franchisor to differentiate among outlets. Another control variable the franchisor may use to reduce his need for a variety of franchise contracts is the location and density of stores in a given geographical territory. The firm generally decides how many outlets will be allowed to operate in any area. It can choose the density of stores so that most outlets are given an equivalent market. In this way, it 121 can make the outlets similar enough that they may not warrant different contracts. Despite the fact that the terms of the contract are not explained as well as q//Q is, one finds that some of the variables do have a certain effect on these two fees. The likelihood ratio tests for the whole regressions are significant, so that the hypothesis that the slope coefficients are equal to zero is clearly rejected. As was pointed out in the previous discussion, in the vast majority of cases, and especially in those cases where the coefficients are measured with enough precision to be clearly different from zero, the sign of the coefficients in the q//Q equation are the opposite of those in the equation for r. This is as expected. Exceptions include the dummy variable on financing, that increases simultaneously qf/Q, r and F, and the franchisor input sales variable which has a negative effect on all three dependent variables. The latter result is very consistent with the notion that the sale of inputs by the franchisor is a substitute for royalties. As for the effect of financing by the franchisor, as far as the fees are concerned, it could be due to the fact that they may include a portion of the remuneration that the franchisor gets for his capital. Finally, it is noteworthy that those few variables that significantly affect the franchise fee F are generally different from those that explain r. Equation (5.4) was derived from the notion that F is simply a function of r since it exists to extract whatever rents would be left downstream, given r. The fact that different variables seem to explain r and F casts some doubt on this assumption. In other words, the reduced form (5.4) may not be appropriate. This would be the case for example if there were other things besides the indices found in (5.4) which affected the determination of the fixed fee. An example of this is provided by the positive effect found for the number of weeks of training. According to the derivation of (5.4), training should have a negative effect on F. But a certain portion of F is often required by franchisors in exchange for this training. The price of other services provided by franchisors may also be included in F. Since I have no information on the value of those services, it is not possible to take them explicitly into account. In general, the results obtained here and the lack of correlation between the two fees found in Chapter 4, suggest that the two fees are not necessarily closely related. This could be the case if F was calculated on the basis of a fee for service as was just discussed. It would also result if the assumption that the franchise fee extracts all surplus downstream was incorrect. In other words, there might be rents left at the downstream level. Mathewson and Winter (1985) found that the optimal franchise 122 contract, in a principal-agent framework where the franchisee has limited wealth, would yield positive expected rents to franchisees. They interpret the existence of queues of potential franchisees for such chains as McDonald's and Burger King as evidence that there are rents left downstream. Testimonies of franchisees from such chains also seem to support this notion. Yet it is difficult to imagine that these companies would be unable to devise a contract that would extract all the rent from their franchisees if they wanted to. If there really are rents left downstream by major franchising firms, it must be because they find it in their best interest to leave them there. Rents being left at the downstream level could be explained by some type of efficiency wage argument. In other words, franchisors would allow franchisees to get some of the rents so that the threat of losing this stream of future earnings would act as a disciplinary device. Franchisees might also get some rents if they have some bargaining power. However, since contracts are generally offered by franchisors to franchisees on a take it or leave it basis, this last explanation lacks credibility. 5. Conclusion In this chapter, the way in which the empirical model was operationalized, in terms of choices of variables, estimation technique, etc. was described. Then the empirical results obtained using the whole sample of 548 franchisors were presented. In terms of the theoretical models discussed in Chapter 2, results are summarized in Table 5.6. As can be seen from the table, the observed effects of the various measures of risk were not consistent with predictions from either the pure risk-sharing and the one-sided hidden-action models of share contracts. Therefore one can say that these types of models are not well supported by these data. As was pointed out however, these interpretations depend on the capacity of the measures to capture exogenous risk, as opposed to the variability that is due to moral hazard. In addition, even if they did, increases in riskiness automatically compound the unobservability problems in models with asymmetric information. This confuses issues and can affect the observed effect of risk on the proportion of franchised stores as well as the royalty rates. But if increased risk leads to a greater reliance on franchising because of the greater difficulty the franchisor has in evaluating his franchisees' performance, the results imply that in fact incentive issues dominate risk-sharing considerations in the 123 Table 5.6 Expected and Observed Effects of the Indices on qjjQ Model In IK Pure Risk-sharing * One-sided Hidden Action — + Two-sided Hidden Action + — Capital-Market Imperfection (-) + (-) + Observed effects of Indices + + — ? * Assuming the franchisee is the more risk-averse party. Note: Parentheses indicate that one and/or the other effect may occur. determination of the contract mix. In other words, risk itself certainly has an effect on firms' choices, but this is not the effect that would have arisen in the context of a risk-sharing type of argument. Although there were a few wrong signs for some of the proxies, the empirical results were found to be generally consistent with two-sided hidden-action models, suggesting that there really are incentive issues on both sides. Franchisors recognize this, and choose both the contract mix and the terms of the franchise contract in a way that reflects this. As for capital-market-imperfection arguments for franchising, only very tentative conclusions can be drawn here. It seems to be clear that firms use franchising more when they want to grow faster. Consequently, one can say that franchising allows franchisors to relax some form of constraint on their growth. Some of the results are consistent with the notion that this constraint might be financial, while others are not. In addition, the following few points have emerged from the empirical analysis. First, it is apparent that the variable fee, when it is taken to be exogenously given to the firm, does not contribute to the explanation of the firms' choice of contract mix. At the same time, it was found that the variables that affect firms' decisions regarding the proportion of franchised stores also have some effect on the variable fee. And there is some trade-off between the two in the sense that variables have opposite effects on qf/Q and r. Consequently, it is clear that the variable fee is in fact a choice 124 variable for the franchisor and should be treated as such empirically. However, given that the proxies used here can account significantly better for variations in qf/Q than for differences in r, it appears that firms prefer to adjust to differences in risk, supervision costs, etc. by modifying their contract mix rather than the terms of their franchise contract. Existing theoretical work on franchising has focused on single franchisor-franchisee pairs. In this case, the franchisor's only control variable is the share parameter. Hence, the comparative-static results derived in these papers are concerned with r only. The fact that firms use the proportion of franchised stores as an instrument suggests that theoretical work that would address the question of firms' choices of contract mix, as well as their decision concerning r, would be useful. It was also found that the franchise fee tends to be explained by variables that are often different from those that affect the royalty rate significantly. Or variables were found to have a positive or a negative effect on both fees. In the case of the input sales variable, and possibly the financing dummy variable, the same effect on both fees was expected or can be explained. But since the equation for the fixed fee was derived from the assumption that it is simply an inverse function of r, one would have expected the two fees to be affected mainly by the same variables, and in opposite directions in general. In reality, the fact that they are not is very consistent with the lack of correlation that was found in Chapter 4.^ One possible reason for this lack of correlation between the two fees is the existence of another way for franchisors to extract rent from franchisees, namely through input sales. And this variable was found to have a negative effect on both fees, supporting the notion that it is used as an alternative to the royalty rate and the fixed fee. But some of the results, for example the one concerning the number of weeks in training, suggest that there may be other factors influencing F beside those that influence r. In this case, equation (5.4) would be misspecified. Equivalently, it would not be possible to infer equation (5.4) from (5.3) if there were any rents left at the downstream level. Both of these scenarios could explain the observed lack of relationship between the observed royalty rates and franchise fees. 1 fi While the regression results are affected by the choice of functional form, and could reflect a nonlinear relation between the two fees that is not taken into account in the estimation, the rank correlation coefficients allow for nonlinearities. The results in Chapter 4 are more robust in that sense. 125 Finally, because this question has been discussed often in the literature on franchising, it is worth pointing out that the empirical results suggest that franchised chains do not tend toward either 100% franchised or 100% company-operated. ^ The results in this chapter were derived under the assumption that all firms involved in franchising behave in a similar way in the sense that they were treated as if they all belong to the same population. This is to an extent reasonable given that people who are involved in franchising often refer to it as an industry. In addition, by examining first the sample as a whole, it was possible to assess the effect of some variables, such as risk for example, for which only aggregate measures are available. But as was seen in Chapter 4, these firms are in fact different in many respects. First, they operate in various types of businesses. Also, some firms have been in business for a long time, while others are just getting started. Similarly, there are important differences in the size of the franchise chains. Since a lot of the variability in the proportion of franchised outlets and in the terms of franchise contract remains unaccounted for so far, it is worth exploring these differences. In the next Chapter, results obtained for various subsamples of firms, based on these criteria, are discussed. See Appendix F. 126 C H A P T E R V I Results from Various Subsamples of Firms 1. Introduction In this chapter, the effect of differences among franchisors in terms of the number of years in operation, the size of the franchise chain and the type of business the firm is involved in, are explored. The sample of firms is divided into subgroups on the basis of these criteria so that one can ask whether the behavior of the firms across the various subgroups is significantly different. Given the size of the initial sample and the number of categories to be made, it will be necessary to examine age cohorts, and then size cohorts, and finally sectoral differences separately. Since it was found in Chapter 5 that (5.1) is not a good characterization of franchisors' behavior, i.e. that the royalty rate should not be treated as if it was exogenously given to the firm, only (5.2), (5.3) and (5.4) will be estimated here. Also, each of the three measures of risk was found to have a similar effect on the proportion of franchised stores and on the terms of the franchise contract as the others. For that reason, and also because this measure was found to have more desirable properties than the other two in Chapter 4, only those results obtained with the proportion of discontinued of outlets will be presented here. Finally, across age and size cohorts, the proportion of franchised outlets and the franchise fee were systematically better approximated by the version of the model where some of the explanatory variables are entered in logarithmic form. This was the case whether one used the likelihood function value or the correlation between the actual and the predicted values for the dependent variable as the criterion for goodness of fit. In the case of the variable fee however, the linear form was preferred. This is consistent with what was found in Chapter 5. In order to reduce the amount of tables, only those results obtained by using the relevant functional form for each case will be presented in the next two sections. Again, the conclusions are unaffected by this. In the sectoral analysis, even the royalty rate will be better explained by the version of the model where some of the variables are entered in logarithmic form. Thus, this form will be used in general in that section. 127 2. A g e C o h o r t s If there is anything to the notion of a life-cycle for firms involved in franchising, as many have argued, one would expect different behaviors from a group of recently established firms than from a group of firms that have been in operation for a number of years. For example, it has been suggested that new firms have a greater need for capital. They are therefore more likely to resort to franchising as a means to get access to capital than established firms are. A second issue that may imply differences among age cohorts is that of adjust-ment costs. Given the average delay of 6.4 years calculated in Chapter 4 between the time a firm starts its operation and the time it gets involved in franchising, firms that have been in business for 1 to 8 years have generally just begun franchising. Thus they are more likely to be in the process of adjusting qf/Q than other firms are. In other words, the proportion of franchised stores that is observed for these franchisors may not be their optimal one. This could affect the results from the estimation of (5.2) for the group of youngest firms in various ways. A third and last issue that one would expect to lead to differences among age cohorts has to do with reputation effects. In Chapter 4, the proportion of franchised stores appeared to be nonlinearly related to the number of years in business, increasing at first, and then going down. In Chapter 5, when other effects were controlled for, the number of years in business was found to have a positive and generally significant effect on the use of franchising. This result is inconsistent with the notion that years in business is a proxy for the value of the tradename in a two-sided hidden-action model. Thus it was argued that firms may be capable of establishing a reputation with respect to the value of their tradenames as the number of years they have been operating increases. Well-established franchisors would not need to operate stores or require as large a variable fee in order to reassure potential franchisees as to the future value of the tradename, contrary to recently established firms. In other words, one of the benefits franchisors derive from operating outlets would disappear as the number of years a firm has been in business increases, and pc would decrease as a result. In order to assess whether or not firms behave differently depending on how many years they have been in business, the sample was divided into four age cohorts.^ Then ^ Very similar results were obtained when the number of years since a firm began franchising 128 each of these was divided again into two groups, for a total of eight age cohorts. This subdivision scheme and the chosen number of cohorts are clearly arbitrary. There is nothing fundamental in the data to suggest appropriate cut-off points that would represent particular stages in the development of a franchise chain. The decision to create a maximum of eight cohorts was dictated mainly by degrees of freedom considerations. In addition, in order to maximize the degrees of freedom within each group, the number of firms was equalized across cohorts as much as possible. However, because many firms had been in business for the same number of years, it was not possible to create groups of exactly 137 and 68 or 69 firms respectively. The advantage of creating the groups in this manner is simply that it makes it possible to test whether there are significant differences among the initial four groups. Then one can determine whether further refinement of the cohorts is warranted simply by testing the coefficients from the two groups within each initial age cohort. This gives information on exactly which of the four initial cohorts are the least homogeneous. In other words, by creating initial cohorts and then subdividing these, the second set of cohorts is nested within the first. • Because these are easier to read and they show the same general patterns, I have chosen to present results based on the four age cohorts in the following tables. Table 6.1 shows results from the estimation of equation (5.2) for each of the age cohorts. Table 6.2 contains the results from the estimation of the variable-fee equation, while results for the fixed-fee equation are presented in Table 6.3. For each of the dependent variables, it was found, on the basis of a likelihood ratio test, that the coefficients were significantly different across the four age cohorts. However, in the case of the variable fee, the differences were significant at the 5% confidence level, not so at the 1% confidence level. Not only do the coefficients vary significantly across the cohorts, but different variables are found to have a significant effect on the contract mix and on the terms of the contract in the various groups. There are, however, very few significant sign reversals among the cohorts. The coefficient of the number of outlets variable is the exception in the case of q//Q, as is the coefficient on the number of states variable in the fixed-fee equation. In Appendix G, Tables G . l to G.3 are equivalent to Tables 6.1 to 6.3, but was used to generate the cohorts. 129 T a b l e 6.1 : The P r o p o r t i o n of Franchisee! S t o r e s w i t h i n Age C o h o r t s Number o f Years 1n B u s i n e s s : 1 t o 8 9 to 13 14 t o 22 23 + % F r a n c h . % Franch. % F r a n c h . % Franch. Log. Av. % D i s c o n t i n u e d 13.47 1 .84 5, .62 -13 .99 ( 1 .62) (0 .22) (0. .51) (-1 25) F o r e i g n O u t l e t s (%) 0.06 0 .54 0. . 14 0. .08 (0.36) (1 .32) (0. 64) (0 41) Log. Number of S t a t e s 2.37 4 .69 + 1 . 74 8 . 6 0 * * (1.16) (1 .90) (0. .76) (3 .75) ( A v . S a l e s - I n p u t s ) / A v . S a l e s -0.71 + 0 .06 0. .01 -0, . 12 (-1.71) (0 .22) (0. .04) (-0. .36) A v . S a l e s / o u t l e t ($100K) 2.31 -1 . 12 -0. .58 -0 .96 (1 .63) (-1 .25) (-0, .93) (-1 .61) F r a n c h i s e e E x p e r i e n c e 2.32 -1 .37 -7. .59 + 0 .03 (0.60) (-0 .32) (-1. ,79) (0.01) Weeks o f T r a i n i n g -2.42 + -1, .64 * -0. .68 -2. . 18 ** (-1.97) (-2 .05) (-0. .85) (-3 .42) Log. O u t l e t s i n 1986 (100's) 3.09 + 2 .21 -0. .43 -6 .97 ** (1.68) (0 .99) (-0. .22) (-3 71) % Time Not F r a n c h i s i n g -0.39 ** -0, .20 * -0. ,39 ** -O, .53 ** (-3.96) (-2. .49) (-4. . 13) (-6. ,54) Log. Y e a r s i n B u s i n e s s 8.78 -16 .08 -32. .48 * 16 .00 ** (1.19) (-1 .14) (-2. 50) (2. .73) Growth In O u t l e t s 10.27 * 8 .27 14. ,31 19. .33 (2.13) (1. 27) (1. ,43) (1. 61) Log. C a p i t a l R e q u i r e d -1 .49 -1 . 38 -7. 16 ** -5. 08 * (-1.03) (-0 .73) (-3. ,63) (-2. .54) F r a n c h i s o r F i n a n c i n g 2.83 15. .53 ** 2. 55 -1 . ,71 (0.60) (2. 81) (0. 56) (-0. 37) Log. F r a n c h i s o r I n p u t s ($K) -6.66 * 1 . 16 -3. 38 -1 . 07 (-2.03) (0. .50) (-1. 18) (-0. 36) C o n s t a n t 143.22 ** 122. .11 * 220. ,93 ** 88. ,51 + (2.80) (2 48) (3. 86) (1. 78) L i m i t O b s e r v a t i o n s 35.00 28. .00 37. 00 17. ,00 No n - L i m i t O b s e r v a t i o n s 111.00 96. 00 104 . OO 120. OO S t a n d a r d e r r o r of Estm. 18.53 17 . 75 19. 45 18 . 47 At Mean X ( I ) . E(Y) 84.52 84. .09 86. 22 79. 28 Log L i k e l i h o o d F u n c t i o n -509.86 -433. .51 -479. 69 -533. 78 Squared C o r ( Y , E ( Y ) ) 0.43 0. 47 0. 42 0. 56 ** : s i g n , a t the 0.01 l e v e l , • : a t the O, .05 l e v e l , + : a t the 0. 10 l e v e l LR t e s t : HO: The c o e f f i c i e n t s a r e the same a c r o s s c o h o r t s H1: They a r e d i f f e r e n t LR s t a t i s t i c » 86.68, s i g n i f i c a n t a t the .01 l e v e l . 130 Table 6.2 : The Variable Fee within Age Cohorts Number of Years in Business: 1 to 8 9 to 13 14 to 22 23 + Variable Variable Variable Variable Fee Fee Fee Fee Av. % Discontinued -0 .26 -0, .80 * -1 .00 + 0 .23 (-0 .59) (-2 .07) (-1 .85) (0 .50) Foreign Outlets (%) -0. .01 0, .02 -0 .06 + -0 .03 (-0. .37) (0 ,80) (-1 .82) (-1 .03) Number of States -0 .04 -0 .04 0 .02 0 .02 (-1 47) (-1, .25) (0 .76) (0 .97) (Av.Sales-Inputs)/Av.Sales -0. .05 -0, .32 ** -0, .20 * -0, . 15 * (-0. ,50) (-5. .17) (-2, ,35) (-2 .28) Av.Sales/ outlet ($100K) 0. .02 -0, . 18 0 .09 0 .06 (0. 07) (-0, 88) (0, .81) (0 .64) Franchisee Experience 0. .88 0, ,31 -0, . 14 0, .09 (1. .38) (0, .50) (-0 17) (0 .14) Weeks of Training 0. . 12 0, . 14 -o. .07 -0 .02 (0. ,63) (1. , 22) (-0, .40) (-0, . 15) Outlets in 1986 (100's) 0. ,21 0. 26 0, . 15 ** 0 .08 ** (0 77) (1, 42) (2 64) (3 . 10) % Time Not Franchising -0. ,01 0. .03 ** 0, .03 + 0, .04 ** (-0. .89) (2. 79) (1, .69) (3 .43) Years in Business 0. 26 -0. , 15 -0, .03 -0 . 10 ** (1. 15) (-0. 77) (-0 . 19) (-4 .64) Growth in Outlets -0. 62 -3. ,06 ** 1, .96 -1 , .97 (-0. 78) (-3 , 12) (1, .06) (-0 .99) Capital Required 0. .00 0. .00 -0 .00 0 .00 (0, 02) (1. 56) (-0 . 19) (0 .08) Franchisor Financing 0. ,27 1 . ,59 * 0, .40 0, .94 (0. 36) (2. 11) (0. .49) (1, .32) Franchisor Inputs ($K) -0. ,02 -0. ,08 ** -0, .05 * -0 .03 (-0. 61) (-4. 19) (-2 . 19) (-1 .36) Constant 10. ,56 40. ,80 ** 28. ,03 ** 21 , .50 ** (1. 18) (6. 17) (3. .20) (3, .06) Limit Observations 12. 00 7. ,00 1 1 , ,00 7 , .00 Non-Limit Observations 134. oo 117. 00 130. .00 130, ,00 Standard error of Estm. 3. 25 2. 69 3. ,83 3. ,08 At Mean X(I). E(Y) 6. 26 6. ,36 6. .57 6 .73 Log Likelihood Function -364. 40 -291 . 69 -373. ,03 -339. .09 Squared Cor(Y, E(Y ) ) 0. 08 0. 33 0. , 17 0. .25 ** : sign, at the 0.01 le v e l , * at the 0.05 1evel, + : at the 0.10 1evel LR test : HO: The c o e f f i c i e n t s are the same across cohorts H1: They are d i f f e r e n t LR s t a t i s t i c = 64.78, s i g n i f i c a n t at the .05 l e v e l . 131 T a b l e 6.3 : The F r a n c h i s e Fee w i t h i n Age C o h o r t s Number of Years In B u s i n e s s : Log. Av. % D i s c o n t i n u e d F o r e i g n O u t l e t s (%) Log. Number of S t a t e s ( A v . S a l e s - I n p u t s ) / A v . S a l e s A v . S a l e s / o u t l e t ($100K) F r a n c h i s e e E x p e r i e n c e Weeks o f T r a i n i n g Log. O u t l e t s i n 1986 (100's) % Time Not F r a n c h i s i n g Log. Y e a r s i n B u s i n e s s Growth In O u t l e t s Log. C a p i t a l R e q u i r e d F r a n c h i s o r F i n a n c i n g Log. F r a n c h i s o r I n p u t s ($K) Co n s t a n t L i m i t O b s e r v a t i o n s Non-L1m1t O b s e r v a t i o n s S t a n d a r d e r r o r of Estm. At Mean X ( I ) , E(Y) Log L i k e l i h o o d F u n c t i o n Squared Cor(Y, E ( Y ) ) 1 t o 8 F i x e d Fee 9 t o 13 F i x e d Fee 14 t o 22 F i x e d Fee 23 + F i x e d Fee -11 .01 1 1 , . 26 -1 , .75 6 .31 (-1 54) (0 .95) (-0. .30) (0 57) 8 . 10 17 .91 -7, .21 -17 .75 (0 .55) (0 .70) (-0 76) (-0 .89) 4, .56 * 0 .66 -0, .71 -4 .23 + (2 .55) (0 .20) (-0 .58) (-1 .87) -0. .38 0 .01 -0, .42 * -o .07 (-1, .07) (0, .04) (-2, 19) (-0 .20) -1 .37 -2 .07 -0 . 16 0 . 14 (-1 .15) (-1, .59) (-0, .48) (0 .24) 0. .34 7, 38 -0. .36 -2 .82 (0. , 10) (1. .22) (-0. . 15) (-o 73) 0, .42 0 .24 0. .44 1 . 03 (0. 42) (0 21) (0. .96) (1, .63) -1. .26 -0. ,07 0. .05 1 . 81 (-0 79) (-0. .02) (0. 05) (0, .99) -0. .05 -0. ,01 0. .01 -0. .09 (-0. .66) (-0. 09) (0. 21) (-1. 24) 4. 10 -6. ,50 -2. .53 0. 38 (0. ,65) (-0. ,33) (-0. ,36) (0. .07) 4 . ,21 4. 49 8. .76 11. .49 (1. 01) (0. .48) (1. 61) (0. 95) 5. ,07 ** 7. ,60 ** 4. ,25 ** 4. ,57 * (4. 06) (2. 79) (4. 12) (2. 29) 2 . 53 3. ,44 4. 19 + 16 . 50 ** (0. 65) (0. 47) (1. 67) (3. 52) -4. , 17 -1 . 93 -4. 31 ** -0. 02 (-1. 48) (-0. 58) (-2. 78) (-0. 01) 47. ,41 3. ,50 61 . , 17 * 6. 01 (1. 09) (0. 05) (2. 02) (0. 12) 1 . OO 1 . ,00 3. OO 2. 00 145. 00 123. OO 138. OO 135. 00 16. 51 26. 04 1 1 . 27 18 . 92 22. 13 25. 39 19. 76 23. 17 613. 48 -576. 07 -533. 49 -590. 40 0. 27 0. 1 1 0. 23 0. 17 : a t the 0, ,05 l e v e l , + : a t the 0. 10 l e v e l LR t e s t : HO: The c o e f f i c i e n t s a r e the same a c r o s s c o h o r t s H1: They a r e d i f f e r e n t LR s t a t i s t i c = 131.06, s i g n i f i c a n t at the .01 l e v e l . 132 the results are based on eight cohorts. They also show the likelihood ratio tests to determine whether the coefficients of each pair of cohorts are significantly different from those obtained for each of the initial four cohorts. In the case of the proportion of franchised stores and of the variable fee, the subdivision of the initial cohorts is found to be really warranted only in the case of firms that have been in operation between 14 and 22 years. For the franchise fee however, this pattern is reversed. The coefficients for the two subgroups are the same within this age cohort, but different for all others. Comparing the results with those obtained in Chapter 5, one finds that they are generally very consistent. It is clear that focusing in this way on a few age cohorts increases substantially the explanatory power of the model with respect to the contract mix as well as the terms of the franchise contract. Still, it remains true within these groups that the proportion of franchised outlets is better explained by differences in the explanatory variables than the two fees are. Using the correlation between the expected and the observed value of the dependent variable as the goodness of fit criterion, the only exception to this for the variable fee is found in Table G.2, in the group of firms that have been in business for 9 to 10 years. In the case of the fixed fee, an exception is found in the first of the eight age cohorts in Table G.3. In general, the subdivision of the sample into a number of age cohorts does not give much support to the notion that newly established firms, more than others, use franchising in order to get access to capital. In Tables 6.1 and 6.2, it is true that growth has a significant positive effect on young firms' propensity to franchise. It also has a negative effect on their royalty rate, especially for firms between 9 and 13 years in operation. This is consistent with the notion that they need franchising to grow, which could be because it increases their access to capital. However, growth also has a large effect on firms' propensity to use franchising in the other cohorts, even though these are not significant. In Table G . l , growth is also found to have a large positive effect on the proportion of franchised outlets of two of the oldest groups of firms. Older firms are the ones that significantly reduce their use of franchising when the amount of capital required to open an outlet increases. This negative effect of the capital-requirement variable on the proportion of franchised stores was interpreted in Chapter 5 as evidence against the capital-market-imperfection explanation for franchising. The effect of this capital variable is generally negative for groups of 133 young firms as well, but in this case it is not significantly different from zero. In order to really give support to the notion that newly established firms use franchising as a source of capital, this coefficient should have been positive. Still, an insignificant effect for these firms is better than a significant negative effect when it comes to capital-market-imperfection arguments for franchising. However, firms in all age cohorts demand higher fixed fees when the required amount of capital is greater, not just younger firms. In this respect, all firms behave similarly and in a manner that is consistent with all of them using franchising as a source of capital. Finally, the fact that franchisors offer financing tends to increase the use of franchising, as well as both fees, across most age cohorts. In the case of qf/Q and r, this effect is clearly different from zero in the group of firms that have been in operation between 9 and 13 years. Given that it was already clear from the results in Chapter 5 that these firms find franchising especially appealing for reasons that have nothing to do with their capital requirements, this is not surprising. What is more surprising is that the proportion of firms that offer financing to their franchisees is not lower in the group of youngest firms than in the group of oldest firms. More precisely, these proportions are 21.9%, 16.1%, 23.4% and 21.8% respectively for each of the four age cohorts in Tables 6.1 to 6.3. Thus firms that have been in business for 9 to 13 years are the least likely to provide financing to their franchisees. Those that have been in business for 8 years or less provide financing as frequently as those that have been around for 23 years or more. This could be interpreted to mean that newly established firms have more difficulty recruiting franchisees and are therefore unable to use franchising as a source of capital. Once they become more credible, i.e. after having been in business for a number of years, they can attract franchisees who will be willing to invest their own capital. Later on, they have access to capital through other means. This could explain the fact that the proportion of firms that offer financing is lowest in the second age cohort. However, those firms that do offer financing in this age cohort certainly increase their use of franchising as a result. Overall, the evidence with respect to the capital-market explanation for fran-chising remains inconclusive. There is no strong support for the notion that newly established firms behave very differently from others in that respect, although some of the results are consistent with this interpretation. The positive effect of the number of outlets on qf/Q is the one surprising result that is found for the group of most recently established firms that may be explained 134 by the notion that these firms have not yet adjusted to their optimal proportion of franchised stores. If it is true that such firms have just begun franchising, it is likely that increases in their number of outlets will imply increases in the proportion of franchised stores simply because the new outlets will be franchised one. However, is it worth pointing out that the variable given by the proportion of time a firm has been in business without franchising has a significant negative effect on qflQ in all age cohorts, and in particular in the group of oldest firms. It also has a significant positive effect on the royalty rates of well-established firms. The same was true when the cohorts were generated on the basis of the number of years firms have been involved in franchising rather than the number of years in business. If this variable did capture adjustment costs as was discussed in Chapter 5, given that oldest firms (and even more those firms that have been involved in franchising the longest) are most likely to be in equilibrium in terms of their contract mix, one would expect it to have a much lower effect on the proportion of franchised stores of well-established franchisors than on that of the more recent ones. This is not the case here. This result, combined with the large positive effect of this variable on the royalty rate, in Chapter 5 as well as here, gives little support to the notion that this variable simply reflects the fact that firms that did not franchise for a large proportion of their time in business have not yet adjusted their proportion of franchised stores fully. But it is consistent with the notion that this variable measures the value of the franchisor's inputs in the context of two-sided hidden-action models. The other main difference that was expected across age cohorts had to do with reputation. It is clear from Tables 6.1 and 6.2 that among the group of most-established firms in this sample, a further increase in the number of years in business leads to a significant increase in the firm's tendency to use franchising, as well as a significant reduction in the royalty rate. This is consistent with Mathewson and Winter (1985) assuming that in this group of firms, the value of the tradename can be taken as fixed ex ante. In table G . l , the same variable has a significant positive effect on the use of franchising for the first age cohort as well, but not the corresponding negative effect on the royalty rate. This result may also be attributed to the fact that such firms have not yet achieved their preferred proportion of franchised stores, just like the positive effect of the number of outlets on qf/Q in this cohort was. Thus, one would observe them in a disequilibrium position with respect to qj/Q, but not when it comes to the terms of the contract. 135 It seems clear that the effect of the number of years in business on q//Q and on r in Chapter 5 were due mainly to the last age cohort. In that sense, the explanation based on reputation that was given in Chapter 5 to explain the effect of the number of years in business variable is supported here. Before firms establish this reputation however, increases in the number of years in business should have the kind of effect dictated by two-sided hidden-action models. In other words, q//Q should be reduced as the number of years in business increases in the other two age cohorts, as it is. The variable fee on the other hand is really unaffected by this variable. The number of outlets in the chain has a significant negative effect on the use of franchising within the group of oldest firms, and a positive effect on their royalty rates. This again points to the fact that given the number of years in business, increases in the number of outlets does not contribute to the establishment of a reputation. As was mentioned previously, this may be due to the increased supervision costs the franchisor faces when the number of outlets he must oversee increase. The other proxies for IT also continue to affect q//Q and r in ways that are consistent with two-sided hidden-action models. Thus even if firms establish reputations through the years, increases in the value or the cost of their contribution continue to imply a smaller reliance on franchising and a larger variable fee so that the franchisor will still have the right incentives. 3. Size Cohorts In this section, differences among groups of firms defined on the basis of the total number of outlets they had in 1986, are examined. Again the sample of 548 franchisors was first divided into four groups, each of which was subsequently split in two. Tables 6.4 to 6.6 give the results for the four size cohorts for the proportion of franchised outlets, the variable fee and the fixed fee respectively. Here as well, the likelihood ratio tests indicate that the coefficients are significantly different across the four size cohorts for all three dependent variables. Again, across the cohorts, results are generally consistent with those from Chapter 5. Dividing the sample among these size cohorts also contributes significantly to the explanatory power of the equations, as opposed to no subdivision at all. Tables G.4 to G.6 present the results based on the eight size cohorts, as well as the likelihood ratio tests indicating whether or not the coefficients are significantly 136 T a b l e 6.4 : The P r o p o r t i o n of F r a n c h i s e d S t o r e s w i t h i n S i z e C o h o r t s Number o f O u t l e t s : 2 - 2 0 2 1 - 5 3 54 - 196 197 + % F r a n c h . % Fr a n c h . % Fr a n c h . % Fr a n c h . Log. Av. % D i s c o n t i n u e d 17. .63 5. .04 3. .28 14. ,02 + (1 54) (0. 48) (0. .36) (1 ,94) F o r e i g n O u t l e t s (%) -10, .60 -28. .68 18. .71 27. .82 + (-0. .37) (-0. 75) (1. .39) (1. 73) Log. Number of S t a t e s 2. .87 4. ,23 + 3. , 10 5. .04 + (1. .05) (1. .95) (1. .63) (1. 84) ( A v . S a l e s - I n p u t s ) / A v . S a l e s -0. .29 0, .50 -0. . 11 -0. .31 (-0, .68) (1. 18) (-0 .38) (-1 .06) A v . S a l e s / o u t l e t ($10OK) 1, .31 -0, . 12 -1 . . 15 -1 , . 13 * (0 .90) (-0, .08) (-1 , 64) (-2 .56) F r a n c h i s e e E x p e r i e n c e 4. .46 -8 ,36 + -2. .07 -2 . 15 (0. .96) (-1 84) (-0. 57) (-0 .62) Weeks of T r a i n i n g -3. .24 ** -3, .76 ** -0. .31 -1 , .48 * (-2, .95) (-3, 77) (-0, .45) (-2 52) Log. O u t l e t s i n 1986 (100's) 4, .56 -3, . 15 -5. .90 -2, . 15 (1 .28) (-0, .46) (-1. .32) (-1 .02) % Time Not F r a n c h i s i n g -0 .50 ** -0. .24 * -0. .28 ** -0, .38 ** (-5. .54) (-2 .50) (-3 .66) (-5 12) Log. Vears i n B u s i n e s s 15. .20 ** 1 . 42 -1 . .99 8 .49 * (3 52) (0. .35) (-0, .63) (2 .60) Growth i n O u t l e t s 1 1 . .28 + 2. .96 0. .58 12 .61 (1 .82) (0 .44) (0 .09) (1 . 16) Log. C a p i t a l R e q u i r e d -0 .97 -2 .35 -3 .86 * -6 .24 ** (-0 .53) (-1, .15) (-2. 43) (-3 .69) F r a n c h i s o r F i n a n c i n g 16 .92 ** 8 .91 -2 . 17 -7 .09 + (3 14) (1 .57) (-0 .51) (-1 87) Log. F r a n c h i s o r I n p u t s ($K) -0 .32 2 .63 -2 .01 -3 .76 + (-0 .09) (0 .69) (-0 .88) (-1 .79) C o n s t a n t 77, .55 43 .65 129 .29 ** 115 .38 ** (1 .38) (0 .77) (3 .35) (3 . 10) L i m i t O b s e r v a t i o n s 25 .00 31 .00 34. .00 27, .00 Non-Lim1t O b s e r v a t i o n s 111 .00 106 .00 103 .00 111 .00 S t a n d a r d e r r o r o f Estm. 20 .57 20 .37 16 .97 16 .23 At Mean X ( I ) , E(Y) 74, .80 83 .43 88 .20 87 .02 Log L i k e l i h o o d F u n c t i o n -514 .85 -494 .57 -461 .40 -481 .07 Squared Cor(Y, E ( Y ) ) 0 .43 O .36 0 .40 0 .57 ** : s i g n , a t the 0.01 l e v e l , * : a t the 0.05 l e v e l , + : at the 0.10 l e v e l LR t e s t : HO: The c o e f f i c i e n t s a r e the same a c r o s s c o h o r t s H1: They a r e d i f f e r e n t LR s t a t i s t i c * 96.58, s i g n i f i c a n t at the .01 l e v e l . 137 T a b l e 6.5 : The V a r i a b l e Fee w i t h i n S i z e C o h o r t s Number of O u t l e t s : 2 - 20 21 - 53 54 - 196 197 + V a r i a b l e V a r i a b l e V a r i a b l e V a r i a b l e Fee Fee Fee Fee Av.% D i s c o n t i n u e d ) -0 .91 * 0, .26 0 .20 -0, .40 (-2 41) (0, .65) (0. .38) (-0 .82) F o r e i g n O u t l e t s (%) -0, .03 -0. .04 -O, .00 -0, .05 + (-0. .66) (-0, 73) (-0. .08) (-1 .67) Number of S t a t e s 0 .06 -o . 16 ** -0 .07 ** -0 .01 (0 79) (-3, .10) (-2. .68) (-0 .21) ( A v . S a l e s - I n p u t s ) / A v . S a l e s -0. .24 ** -O, .24 ** -0. . 1 1 -0 .23 * (-3. .96) (-3. .51) (-1. .38) (-2 .52) A v . S a l e s / o u t l e t ($100K) 0. .01 0 . 10 0. . 12 0 . 1 1 (0. .07) (0. 47) (0. 74) (1 .21) F r a n c h i s e e E x p e r i e n c e 0. .54 -0. .07 -0. .46 0, ,44 (0. .93) (-0. 12) (-0. 63) (0 .60) Weeks of T r a i n i n g 0. ,04 -0. .08 -0. ,01 0, .08 (0. ,27) (-0 .55) (-0. 07) (0 .63) O u t l e t s 1n 1986 (100's) -1. ,87 -0. .75 0. 99 0 .09 ** (-0, .37) (-0. 27) (1. 17) (3, .28) % Time Not F r a n c h i s i n g 0. .01 0, .03 * 0. ,02 0, .05 ** (0. 85) (2. 27) (1. 61) (3 .40) Y e a r s i n B u s i n e s s -0, .01 -0, .05 + -0. .03 -0, .09 ** (-0. 21) (-1. .92) (-1. , 13) (-3, .51) Growth i n O u t l e t s -0. .23 -2. .39 ** -1. .83 -0. .68 (-0. 31) (-2. 74) (-1. 47) (-0, .32) C a p i t a l R e q u i r e d -0. OO 0. OO - -0. OO 0 .00 (-0. .00) (0. ,61) (-0. 33) (0. 39) F r a n c h i s o r F i n a n c i n g -0. .07 0. ,24 1. .53 + 1. 29 (-0. 11) (0. ,33) (1. 84) (1. 66) F r a n c h i s o r I n p u t s ($K) -0. ,07 ** -0. ,08 ** -0. ,02 -0. ,05 + (-3. 67) (-3. 77) (-1. 12) (-1. 95) C o n s t a n t 32. ,90 ** 31 . ,25 ** 15. 83 * 30. ,50 ** (5. 3 D (4. 46) (2. ,00) (3. 15) L i m i t O b s e r v a t i o n s 6. 00 7 . 00 13. 00 1 1 . 00 Non-L1m1t O b s e r v a t i o n s 130. .00 130. ,00 124. ,00 127 . ,00 S t a n d a r d e r r o r o f Estm. 2. 70 2. 92 3. 60 3. 66 At Mean X ( I ) , E(Y) 6. 51 6. ,00 6 . 17 7 . 20 Log L i k e l i h o o d F u n c t i o n -322. .33 -332. .93 -351 . , 12 -358. ,56 Squared Cor(Y, E ( Y ) ) O. , 17 0. 23 0. , 12 0. ,24 ** : s i g n , a t the 0.01 l e v e l * a t the 0.05 l e v e l , + : a t the 0.10 l e v e l LR t e s t : HO: The c o e f f i c i e n t s a r e the same a c r o s s c o h o r t s H1: They a r e d i f f e r e n t LR s t a t i s t i c = 71.32, s i g n i f i c a n t a t the .01 l e v e l . 138 T a b l e S.6 : The F r a n c h i s e Fee w i t h i n S i z e C o h o r t s Number o f O u t l e t s : 2 - 20 21 - 53 54 - 196 197 + F i x e d F i x e d F i x e d F i x e d Fee Fee Fee Fee Log. Av. % D i s c o n t i n u e d 0. .47 7 . 23 -8 .90 -0. .08 (0. .04) (0. .93) (-1. .08) (-0. 01) F o r e i g n O u t l e t s (%) 35. .02 -23. .23 6. .90 -5. 19 (1 .05) (-0. .85) (O. 64) (-0. 35) Log. Number o f S t a t e s 4 .71 -1 , .39 0 .58 -0 .09 (1. .56) (-0. .88) (0. 34) (-0. .03) ( A v . S a l e s - I n p u t s ) / A v . S a l e s -o. . 17 -0. .37 -0. .59 * 0. .35 (-0. 36) (-1. .19) (-2. .23) (1 . 12) A v . S a l e s / o u t l e t ($100K) -3, .07 + 1, . 14 -0. .49 -0. .00 (-1 91) (1 .07) (-0, 79) (-0. .00) F r a n c h i s e e E x p e r i e n c e 5. .09 -6. .58 + 1 , .60 0. 81 (0. .98) (-1, .96) (0, .49) (0. ,22) Weeks of T r a i n i n g 0. .87 0. .64 0. .23 0. ,56 (0. 71) (0. .88) (0 37) (0. .89) Log. O u t l e t s In 1986 (100's) -0 68 0. .60 -3. .01 -0. ,95 (-0. . 17) (0. 12) (-0. 75) (-0. 43) % Time Not F r a n c h i s i n g 0. .08 -0. , 11 0. .04 0. 00 (0. 87) (-1. 61) (0 .56) (0. ,07) Log. Years i n B u s i n e s s -1. .79 2 .81 -3. .48 -2. .40 (-0, .38) (0 .96) (-1 . 19) (-0. 71) Growth 1n O u t l e t s 1. .90 8 .56 + 7. 36 -3. .55 (0. 27) (1. 71) (1. 25) (-0. .32) Log. C a p i t a l R e q u i r e d 5. .81 ** 5. .91 ** 5. .90 ** 1 , .98 (2. 77) (3. .93) (4. .07) (1. 16) F r a n c h i s o r F i n a n c i n g -2 .81 12. , 20 ** 8 . 88 * 10 .51 ** (-0. .49) (2. ,99) (2. 34) (2. 66) Log. F r a n c h i s o r I n p u t s ($K) -2. .05 -4. . 13 -6. .03 ** 1 . ,36 (-0. .49) (-1. .49) (-2. .80) (0. 61) C o n s t a n t 20. 48 27. .62 83. ,39 * -16. 29 (0. .32) (0. 66) (2. .38) (-0. 42) L i m i t O b s e r v a t i o n s 2. OO 1 . 00 1 . OO 3. OO Non-L1m1t O b s e r v a t i o n s 134 . 00 136. .00 136 . 00 135. .00 S t a n d a r d e r r o r of Estm. 23. 80 15. .53 16 . 1 1 18 . 01 At Mean X ( I ) , E(Y) 22 78 22. ,06 22. .67 22. 69 Log L i k e l i h o o d F u n c t i o n -616. .48 -566. .99 -572. , 13 -584. 91 Squared Cor(Y, E ( Y ) ) 0. . 18 0. ,25 0. ,24 0. 08 ** : s i g n , a t the 0.01 l e v e l , * : a t the 0.05 l e v e l , + : a t the 0.10 l e v e l LR t e s t : HO: The c o e f f i c i e n t s a r e the same a c r o s s c o h o r t s H1: They a r e d i f f e r e n t LR s t a t i s t i c = 76.92. s i g n i f i c a n t a t the .01 l e v e l . 139 different across each pair of cohorts that constitute one of the four initial groups. Subdividing the sample further in this way is found to be especially worthwhile in the case of the fixed fee and the proportion of franchised stores. Because there should exist some form of relationship between the number of outlets a firm has and the number of years it has been operating, those issues that were discussed in the previous section could apply here as well. Surprisingly though, the number of firms that were common to an age cohort and its corresponding size cohort was relatively low. For example, only 64 firms belong to both the oldest age cohort and the largest size cohort, which is less than half the number of firms in each of these groups. In all other cases, the numbers were even lower, i.e. 60, 42 and 51 respectively for the first, second and third cohorts. Still, one might expect firms that have a large number of outlets to also be less likely to need capital than smaller firms are. Similarly, large firms may have established a reputation as to the value of their tradename. Finally, one may argue that small firms do not have the kind of flexibility that would allow them to operate their preferred proportion of stores. In other words, there is a higher probability that small firms will, at any point in time, be off of their optimal path, in terms of the contract mix. Larger firms, on the other hand, would generally have achieved their preferred contract mix. Consequently, the types of results that were expected in the case of age cohorts may also be observed across size cohorts. And indeed, with respect to the growth variable, the number of years in business, the number of outlets, the proportion of time not franchising and the amount of capital required, the results are quite similar whether one looks at size or age cohorts. In that sense, it does not really matter whether one determines how established a firm is on the basis of the number of years in business or as a function of the number of outlets it has. However, the patterns seem clearer in the latter than in the former, suggesting that it is the number of years in business that is the more relevant classification scheme with respect to these issues. There is, in addition, one issue that is specific to size cohorts. It is related to the risk-sharing argument and was discussed briefly in Chapter 5. Firms that operate very few outlets are more likely than large firms to be more risk averse than their franchisees, especially in view of the fact that risk per outlet decreases as the number of outlets in a chain increases. As a result, on the basis of pure risk-sharing arguments, one may expect firms with only a few outlets to use franchising more, and reduce r, when risk increases, whereas other firms should reduce their proportion of franchised 140 stores and increase r when there is more exogenous risk. From tables 6.4 and 6.5, and even more so from Tables G.4 and G.5, it is clear that this is not what happens. The effect of risk on qf/Q is positive in most size cohorts, including the largest one. Interestingly, that was not the case in the group of oldest firms, where the effect of the proportion of discontinued outlets on qj/Q was negative, although not significantly so. However, the portfolio argument does not apply in that context. The effect of risk on the royalty rate is mostly negative although in this case, it is only measured with enough accuracy to be different from zero in the cohort comprised of the smallest chains. Consequently, the fact that risk had a non-negative effect on the use of franchising as well as a negative effect on the royalty rate in Chapter 5 can not be explained away by attributing it to small firms' risk aversion. Franchisors that have a large number of outlets behave in the same way as those who have only a few in that respect. 4. Sec to ra l Effects Firms that operate in different types of businesses face different constraints that may influence their choices with respect to both the contract mix and the terms of the franchise contract. An example of this is found in Brickley and Dark (1987). They argue that firms involved in nonrepeat sales businesses will need to company-operate more stores because of possible free-riding by franchisees. In the context of two-sided moral-hazard models, the argument would be that the role of the downstream operator is less important in nonrepeat businesses than it is in repeat ones. Hence, it is not as necessary to give incentive-compatible contracts to the outlet managers in that case. More generally, one would expect differences in technologies, whether production technologies at the downstream level, or monitoring technologies, to lead to different behavior on the part of franchisors. Similarly, risk and franchisees opportunity wages may be similar within sectors, but different across them. With respect to the fees, franchisors involved in different sectors may face more or less competition from other franchisors in the recruitement of franchisees. This was not taken into account so far because theoretical models assume an infinite supply of potential franchisees. But if franchisors compete more in certain sectors than in others in order to get franchisees, this may be reflected in the fees. In fact one would expect to find significant sectoral 141 differences in the fees in this case. Some of the differences between sectors were captured in Chapter 5 by variables such as the measures of risk, the average size of outlets, and franchisor's sale of inputs to franchisees. This is because information for these variables was only available on a sectoral basis. By introducing these sectoral variables directly in the equations, one allowed, for example, for risk to have an effect on qf/Q, r and F. But the way in which the other variables related to the franchisor's decision variables was unaffected by this. In that sense, those variables measured on an aggregate basis can capture some of the sectoral effects, but only in a way that is similar to what sectoral dummy variables can achieve. Using the functional form approach described in Chapter 5, Section 3.2, the constant terms in the qj and g* equations that result from the calculations would be functions of 7 i u h and of 72^1- But 71 and 72 would normally be different across sectors due to differences in the production process. In these circumstances, sectoral variations due to technological differences would affect mainly the constant terms in the qj and q*c equations. One question one might want to ask then is whether actual sectoral dummy variables should be introduced in the estimation of equations (5.2) to (5.4), given that these already include sectoral variables. In fact, this would be one way to control for sectoral variations directly which may make it possible to separate sectoral effects related to differences in technology, for example, from the effect of those variables that were measured on an aggregate basis. Alternatively, sectoral effects can be introduced in the equations as in the two previous sections, i.e. by dividing the sample of firms into various groups depending on the type of operation they have, and estimating each equation separately for each group. This would allow those things that are related to the sector in which each firm operates to affect not only the constant term, but also the way in which each of the proxies relates to qf/Q, r and F. However, in that case, it is impossible to leave any of the variables that were measured on an aggregate basis in the equations. Thus the effect of risk can no longer be taken into account as it was in Chapter 5. But by using a sectoral measure for risk, it was already assumed to be the same for all the firms in a sector. In that case, it becomes irrelevant as an explanation for observed differences in qf/Q, r and F among firms involved in the same type of business. Similarly, it was argued that average sales per outlets may reflect differences 142 in the opportunity wages of potential franchisees. If outlets are large in a certain type of business, franchisee's managerial skills become more important. Thus their opportunity wage may increase. Within a sector however, the variance in the size of outlets should be much smaller than it is across all types of businesses. For that reason, within sectors, one may expect potential franchisees', opportunity wages to be relatively constant. Again this was implicitely assumed to be the case in Chapter 5 since the size of outlets was proxied by average sales per outlets measured on a sectoral basis. Finally, as was noted above, differences in the technologies, be it production technologies at the downstream level or monitoring technologies, are minimized when one concentrates on firms in a single type of business. Thus by separating franchisors from different sectors, one allows differences in risk, in production technology, in monitoring technology, in franchisee's opportunity wages, etc. to affect the way in which the remaining variables influence firms' choices of qj/Q, r, and F. Of course, all these arguments assume that the sectoral classification used here, which is based on the sectors defined by the U.S. Department of Commerce, generates homogeneous groups of firms. Whether the constant term is the only coefficient in equations (5.2) to (5.4) that is significantly affected by sectoral effects, or whether different equations are warranted for each sector, is examined in what follows. The use of intercept dummy variables is discussed first. 4.1 The Effect of Sectoral Dummy Variables Table 6.7 contains results from the estimation of equations (5.2), (5.3) and (5.4) under a linear functional form. In Table 6.8 results obtained when some of the explanatory variables are in logarithmic form are presented. For each dependent variable, in both tables, the first column refers to a version where sectoral dummy variables on the intercept have been introduced in addition to the variables that were already measured on an aggregate basis. Only those sectors for which the number of observations was large enough to allow separate regressions later on are represented by dummy variables in these tables. When dummy variables for all the sectors or subsectors defined by the U.S. Department of Commerce were used, the hypothesis that the coefficients of the supplementary dummy variables were 143 Table 6.7 : The Effect of Sectoral Dummy Variables Under a Linear Specification % Franch. % Franch. Variable Variable Fixed Fixed Fee Fee Fee Fee Av. X Discontinued 6. 88 -0. 97 -0. .99 (1. 24) (-1. 08) (-0. . 19) Outlets outside home {%) 18. 72 + 18. .46 + -0. 79 -0. .77 -2. ,35 -1. 38 (1. 78) (1. 77) (-0. 53) (-0. 52) (-0. ,27) (-0. 16) Number of states 0. 31 ** 0. 30 *• 0. 00 0. ,00 0. ,06 0. 07 (4. 03) (3. 99) (0. 36) (0. ,30) (0. .92) (1. 06) (Av.Sales-Inputs)/Av.Sales 2. 40 -0. 17 -0. .55 (1. 3D (-0. 57) (-0. .32) Av.Sales per outlet ($10OK) -1 . 76 * 0. 15 -0. 43 (-2. 39) (1. 29) (-0. 63) Franchisee Experience -1. 16 -1 . 24 0. 10 0. . 10 1. ,78 1. ,55 (-0. 54) (-0. 58) (0. 28) (0. ,28) (0. ,91) (0. 79) Weeks of Training -1. 74 «* -1. .95 *» -0. 04 -0, .02 1. .11 1 . .06 ** (-3. 86) (-4. 42) (-0. 62) (-0. .33) (2. .67) (2. .60) Outlets in 1986 (100's) -0. 29 * -0. 30 * 0. 08 ** 0. .09 *» -0. , 14 -0. , 15 (-2. 06) (-2. 10) (3. 74) (3. 84) (-1. ,11) (-1. 14) % Time Not Franchising -0. 34 ** -0. 33 •* 0. 03 ** 0. ,03 •* -0. ,02 -0. 01 (-8. 25) (-8. 09) (4. 04) (3. .97) (-0. .48) (-0. 28) Years In Business 0. 14 + 0. 13 -0. 04 ** -0. ,04 ** -0. 03 -0. ,05 (1. 76) (1. 60) (-3. 44) (-3. .42) (-0. .40) (-0. ,70) Growth in number of outlets 7. 18 * 7 . 08 * -1 . 54 ** -1 . ,54 ** 3. ,47 3. 82 (2. 24) (2. 20) (-2. 97) (-2. 96) (1. ,17) (1. .29) Capital Required -0. 01 * -0. 02 •*. 0. 00 0. ,00 0. ,02 ' 0. ,02 »» (-2. 29) (-2. 95) (0. 11) (0. .25) (3. 57) (3. 42) Franchisor Financing 3. 43 3. 42 0. 89 « 0. ,91 * 6. ,04 1 ** 6. , 18 «* (1. 39) (1. 38) (2. 31) (2. ,35) (2. ,73) (2. ,78) Franchisor Inputs ($CO0) 0. 14 0. 01 -0. 50 (0. 37) (0. 10) (-1. 39) Automotive Products 55. 26 6 . 11 -0. 69 3. ,44 »* 6. ,89 -5. 24 (1. 45) (1. 40) (-0. 11) (4. 87) (0. .19) (-1. .30) Business Aids. Services -6. 15 13. 52 ** 2. 59 0. ,27 -6. 62 -4 , 47 (-0. 51) (3. 54) (1. 34) (0. .44) (-0. 60) (-1. ,30) Construction, Maintenance 8. 09 11 . 67 »« 0. 78 -0. .04 -10. . 16 -8. ,35 » (0. 79) (2. 71) (0. 47) (-0. .05) (-1. ,07) (-2. , 15) Educational Services -6. 17 2. 80 3. 08 1. .37 + 4 . 64 8 . 74 + (-0. 39) (0. 57) (1. 20) (1. 71) (0. 31) (1. .89) Restaurants 0. 02 0. 28 0. 40 0. .71 -0. .90 -7. ,73 • (0. 00) (0. 07) (0. 37) (1. . 12) (-0. ,15) (-2. . 12) Non-Food R e t a i l i n g 39. 87 3. 85 -3. 98 -0. .20 32. 56 -5. , 18 (1. 18) (1. 07) (-0. 72) (-0. ,34) (1. 03) (-1. ,55) Non-conv. Food R e t a i l i n g 14. 54 1 . 06 -1 . 49 0, .28 11 . 58 -4. ,57 (1. 07) (0. 25) (-0. 67) (0, .42) (0. ,91) (-1. 17) Constant -151. 81 90. 22 •* 23. 90 5, ,82 •* 81 . 46 18 , 84 •« (-0. 85) (18. 86) (0. 82) (9. .33) (0. ,49) (5. ,44) Limit Observations 117. 00 117. 00 37. OO 37. ,00 7. 00 7. .00 Non-Limit Observations 431 . 00 431 . 00 511 . 00 511, .00 541 . 00 541 , .00 Standard error of Estm. 20. 42 20. 57 3. 37 3. .38 19. ,45 19. .55 At Mean Values of X(I). E(Y) 83. 09 83. 04 6. 48 6 .48 22. 69 22, .71 Log Likelihood Function -2001. 84 -2005. 05 -1394. 94 -1395. .85 -2380. ,57 -2383, .26 Squared Cor(Y. E(Y)) 0. 36 0. 35 0. 14 0 . 14 0. , 10 0, .09 \ / \ / \ / Likelihood r a t i o tests 6. 42 1 . .82 5. 38 s i g n i f i c a n t at the 0.01 l e v e l . * s i g n i f i c a n t at the 0.05 level, , + s i g n i f i c a n t at the 0. 10 level 144 Table 6.8 : The Effect of Sectoral Dummy Variables Under a P a r t i a l l y Logarithmic Specification % Franch. % Franch. Variable Variable Fixed Fixed Fee Fee Fee Fee Log. Av. % Discontinued 23 .66 -4 . ,83 + -6. 95 (1 .51) (-1. 90) (-0. 48) Outlets outside home (X) 14 .99 13. .95 -0. 92 -0 .77 -1 . 47 -0. 91 (1 .46) (1 .38) (-0. 61) (-0 .52) (-0. 17) (-0. 11) Log. Number of states 3 .92 ** 3. ,81 ** -0. ,70 ** -0 .68 •* 0. ,79 0. 82 (3 .36) (3. ,28) (-3. .79) (-3 .72) (0. 75) (0. 79) (Av.Sa1es-Inputs)/Av.Sales 2, .22 0. 05 1. 39 (0 .92) (0. 12) (0. 62) Av.Sales per outlet ($1O0K) -1. .93 •* 0. , 18 -0. 49 , (-2 .70) (1. 61) (-0. 76) Franchisee Experience -1 , 27 -1. ,56 -0. 02 -0, .01 0. 78 0. 54 (-0. ,60) (-0. 73) (-0. 05) (-0 .03) (0. 40) (0. 28) Weeks of Training -1. .60 »• -1. ,86 ** -0. 09 -0 .06 0. 85 • 0. 81 (-3 .56) (-4. ,21) (-1. 25) (-0 .89) (2. ,07) (2. 01) Log. Outlets In 1986 (10O's) -0. .43 -0. 37 0. 90 ** 0, .89 ** -0. 27 -0. 31 (-0. .43) (-0. 37) (5. 61) (5, .50) (-0. 30) (-0. 35) X Time Not Franchising -0. .34 ** -0. 34 «* 0. 03 ** 0. ,03 ** -0. 01 -0. 00 (-7. .94) (-7. 74) (4. 16) (4 11) (-0. 22) (-0. 09) Log. Years In Business 4. .32 * 3. 91 • -1. 03 ** -1 , .03 *» -0. 86 -1. 24 (2. .32) (2. 11) (-3. 43) (-3 .47) (-0. 50) (-0. 73) Growth In number of outlets 8. .51 * 8. 22 * -1. 86 *» -1 , .84 ** 4 . 66 4. 58 (2. ,47) (2. 37) (-3. 34) (-3 ,28) (1. 47) (1. 44) Log. Capital Required -3, .40 ** -3. ,65 ** 0. 25 + 0 .25 + 5. ,05 *» 4 . 89 (-3 .66) (-3. 94) (1. 69) (1. .75) (5. ,95) (5. 81) Franchisor Financing 2. 28 2. ,32 0. 90 • 0 .93 * 7. 37 •* 7. 51 (0. .93) (0. 94) (2. 35) (2. 41) (3. 34) (3. 41) Log. Franchisor Inputs ($000) 0. ,13 0. 40 1 . 04 (O. .03) (0. 57) (0. 26) Automotive Products 55, ,91 6. ,70 5. 37 3 .65 ** 28. ,97 -5. 82 (1. .00) (1. 56) (0. 59) (5. 25) (0. 55) - (-1. 48) Business Aids, Services -11. . 10 11. 66 ** 4. 31 * 1 .03 + -2. 18 -2. 15 (-0. ,95) (3. 02) (2. 30) (1, .70) (-0. 21) (-0. 62) Construction, Maintenance 1 , .87 a. 95 * 3. 11 + 0 .73 1. 04 -4 . 76 (0. . 18) (2. ,04) (1. 82) (1. 06) (0. 11) (-1. 22) Educational Services -10. .96 1. ,73 5. 87 * 2 .03 * 18. 60 10. 47 (-0. .70) (0. ,35) (2. 28) (2 .54) (1. 27) (2. 30) Restaurants 1. OS 0. 47 0. 28 0. , 74 -10. 30 + -9. 54 (0. .17) (0. 12) (0. 28) (1 19) (-1. ,79) (-2. 69) Non-Food R e t a i l i n g 47. 88 4 . 81 1 . 73 -0, .07 26. ,77 -5. 69 (0. .98) (1. 35) (0. 22) (-0 . 13) (0. 59) (-1. 74) Non-conv. Food R e t a i l i n g 19. , 13 1 . ,98 -0. 48 0 .39 0. 92 -6 . 22 (1. 56) (0. 49) (-0. 24) (0 .58) (0. 08) (-1. 64) Constant -137 , .02 90. ,67 ** 7 . 05 8 .74 »* -125. ,08 3. 51 (-o. .58) (12. , 12) (0. 18) (7, .72) (-0. 57) (0. 56) Limit Observations 117. ,00 117. 00 37. 00 37 .00 7. 00 7. 00 Non-Limit Observations 431 . ,00 431 . 00 511 . 00 511 , .00 541 . 00 541 . 00 Standard error of Estm. 20, . 14 20. 33 3. 33 3 .34 19. , 11 19. 16 At Mean Values of X(I), E(Y) 83. . 15 B3. , 1 1 6. 48 6 .48 22 . 61 22 . 62 Log Likelihood Function -1995, .03 -1999. , 16 -1388. 80 -1390 .77 -2370. ,74 -2372. 23 Squared Cor(Y. E(Y)) 0. .38 0. 37 0. 16 0 . 16 0. , 13 0. 13 \ / \ / \ / Likelihood r a t i o tests 8.26 3.94 2.98 ** s i g n i f i c a n t at the 0.01 l e v e l , * s i g n i f i c a n t at the 0.05 le v e l , + s i g n i f i c a n t at the 0.10 level 145 zero could not be rejected at the .05 confidence level. In the Tables, the base case is given by the miscellaneous sector defined in Franchising in the Economy, plus all other sectors that have too few observations to form an acceptable subsample. In the second column, the variables that are only available on an aggregate basis were removed. This was done in order to determine whether the latter variables captured most of the intersectoral effects or not. The likelihood ratio tests given at the bottom of the tables indicate clearly that when sectoral dummy variables are introduced in the estimated equations, the effect of the variables that are measured on an aggregate basis becomes insignificant. However, using the results obtained in Chapter 5, one finds that given that variables measured on a sectoral basis are included in the equation, the sectoral dummy variables still contribute significantly to the estimated equation in the case of the two fees.^  Thus it appears that the variables that are measured on a sectoral basis are able to capture differences among sectors as much as sectoral dummy variables can in the case of firms' propensity to use franchising, but unable to capture all these intersectoral differences with respect to the terms of the franchise contracts. Naturally, the coefficients of the dummy variables are very different whether one looks at a version where the variables measured on an aggregate basis are included or not. These two groups of variables are interrelated in a way that makes the interpretation of individual coefficients problematic. Also the coefficients on the dummy variables are different in the two tables, i.e. depending on which functional form is used. One obtains a slightly better fit for all three dependent variables when some of the explanatory variables are entered in logarithmic form. From Table 6.8 then, using the version where variables measured on an aggregate basis are excluded, since these were found to be insignificant as a group when sectoral dummy variables were included, it appears that when a firm is involved in Business Aids and Services, or in the Construction and Maintenance sector, it tends to use franchising more. This is consistent with what was found in Table 4.3. Of the eight sectors defined here, these are the two with the largest average proportion of franchised outlets in that table. Similarly, the fact that a firm belongs to the sector of Automotive Products or of ^ The values of the likelihood ratio test statistics are 12.72, 11.32, and 21.26 respectively for the qf/Q, r and F equations in the linear version, and 10.3, 39.16, and 16.46 in the partially logarithmic version. These are to be compared to a chi-square value of 14.07 for a significance level of .05. 146 Educational Services tends to have a positive effect on its royalty rate. Operating in the sector of Educational Services also has a positive effect on a firm's franchise fee. For those in the Restaurant business, the effect on the fixed fee is negative. This last result is not confirmed in Table 4.5 in the sense that this sector does not display an unusually low average franchise fee. 4.2 Differences Among Sectors The explanatory power of the equations is not improved very much by the introduction of sectoral dummy variables. Thus it may be the case that subdividing the sample in a number of groups on the basis of the sectors in which the firms operate may be a better way to capture intersectoral differences. In Tables 6.7 and 6.8, all three dependent variables were better approximated by the version where some of the explanatory variables were entered in logarithmic form. This also generally occurred when the equations were estimated separately for subsamples based on the sector of operation. For that reason, only the results obtained under this functional form will be discussed in what follows. As can be inferred from the previous tables, there are 8 main sectors with a suf-ficient number of observations to allow separate regressions. These are Automotive Products and Services, Business Aids and Services, Construction and Maintenance, Educational Services, Restaurants, Non-Food Retailing, Non-Convenience Food Re-tailing and a miscellaneous category. In the case of Restaurants, it would have been possible to create separate subsamples for fast-food places and restaurants with table service. However, when this was done, no significant differences were found between the two groups. Thus no distinction will be made in the following discussion. Again, the miscellaneous category contains the Miscellaneous sector defined by the U.S. Department of Commerce, as well as all the other main sectors with too few obser-vations. Results concerning the proportion of franchised stores, the royalty rate and the franchise fee for these eight sectors are found in Tables 6.9 to 6.11 respectively. Dummy variables are clearly significant when variables measured on an aggrea-gate basis are removed from the equations. For that reason, the likelihood-ratio tests given at the bottom of these tables were calculated taking for granted that different intercepts are warranted. These tests indicate that the slope coefficients are differ-ent as well. This is especially true in the case of the two fees, more so than for the 147 T a b l e 6.9 : The P r o p o r t i o n of F r a n c h i s e d O u t l e t s In V a r i o u s S e c t o r s S e c t o r s : Automotive B u s i n e s s C o n s t r u c t i o n Educ. R e s t a u r a n t s Non-Food Non-Conv. M i s c . P r o d u c t s S e r v i c e s Maintenance S e r v i c e s R e t a i l i n g Food % F r a n c h . % Fr a n c h . % Franch. % Franch. % Franch. % Franch. % Franch. % Franch. F o r e i g n O u t l e t s (%) -12 .77 -27 .36 19 .29 -0 .93 45 .74 + 14. .47 -0. .56 3. 32 (-0 34) (-0 .54) (1 .03) (-0. .02) (1 .69) (0. 76) (-0. .01) (0. 18) Log. Number of S t a t e s 4. .95 4 .55 2 . 16 7 .68 1 . 01 6. 11 * -2. . 15 5. 28 + (1. . 14) (1 .29) (0 .62) (1 .36) (0. .37) (2. 34) (-0. .60) (1. 99) F r a n c h i s e e E x p e r i e n c e -1 .01 6. .04 -5. ,23 23 .52 ** -6 .48 2. .97 -0. .99 -2. 14 (-0, 11) (1 .06) (-0, .82) (3. .05) (-1 .49) (0. .53) (-0 . 15) (-0. 34) Weeks of T r a i n i n g -4, .02 * -3. .86 + 0 .45 3. .92 -1 . 53 ** -0 .62 -4. .50 * 1 . 13 (-2 .05) (-1 .80) (0, .27) (1 .39) (-2 72) (-0. 38) (-2. .63) (0. 72) Log. O u t l e t s 1n 1986 (100's) -1 .35 4. .71 4, ,42 + -3 .36 -2. .30 -1. .65 -1 , . 10 -0. 71 (-0 .29) (1 .39) (1. .77) (-0. ,89) (-1 . 13) (-0. 64) (-0. 31) (-0. 28) % Time Not F r a n c h i s i n g -0 .41 * -0 .23 + 0, .02 -0. .04 -0 .45 ** -0 .32 ** -0 .61 ** -0. 28 •* (-2 .56) (-1 .67) (0 .15) (-0 25) (-4. 81) (-3 . 16) (-4 .54) (-2 73) Log. Years 1n B u s i n e s s 14. .20 + 0 .47 1. . 15 -5 07 6. .32 -0. .71 20 .70 ** -o. 89 (1 .99) (0. .08) (0 24) (-0, .55) (1. .44) (-0. 18) (3. . 17) (-0. ,21) Growth 1n O u t l e t s 12. .73 7. .05 -3. ,74 -3 .78 -0. .52 13. .83 22. .94 * 0. 07 (0. .76) (0. 77) (-0. .54) (-0 ,36) (-0. .05) (1. 49) (2. .02) (0. 01 ) Log. C a p i t a l R e q u i r e d -8. . 14 * 0 .21 -3. .33 -3, 68 -3. .92 + -4 . 47 * -1 .57 -1. 83 (-2 19) (0 .08) (-1. 24) (-1 . 19) (-1 .82) (-2. 14) ("O .40) (-0 92) F r a n c h i s o r F i n a n c i n g -0. .24 3. .41 -7. .01 2 .91 4, . 18 -7. .79 14. .09 6. . 1 1 (-0. .03) (0. .58) (-1. .35) (0 .36) (0. 52) (-1. 24) (1 18) (1. ,22) C o n s t a n t 92. .30 * 101 . .83 ** 101 . ,99 ** 75 .45 ** 94. .65 ** 100. ,62 ** 61 . .98 * 86. 85 ** (2. 72) (4. .69) (5. .76) (3 54) (5 .67) (5. 88) (2. .39) (5. ,58) L i m i t O b s e r v a t i o n s 1 1 .00 43, .00 19. .00 5. .00 7. .00 21 . 00 5 .00 6 . ,00 Non-Limit O b s e r v a t i o n s 34. .00 53 .00 31 . ,00 24. .00 114. .00 78 .00 47 .00 50. .00 S t a n d a r d e r r o r o f Estm. 17. .43 23. .01 13. .23 13 .94 19. .43 20. .49 18 .91 14 .00 At Mean X ( I ) , E(Y) 84. .65 90. .67 93 ,25 82. .66 71 , 96 83. .31 79 .75 86 .32 Log L i k e l i h o o d F u n c t i o n -154. .05 -267 .07 -135. .83 -101 .08 505. .94 -363 .78 -209 . 1 1 -207 .21 Squared Cor(Y, E ( Y ) ) 0. .47 0. .31 0 .47 0. .54 0, .39 0 .37 0 .54 0 .34 ** s i g n i f i c a n t a t the 0.01 l e v e l , * s i g n i f i c a n t at t h e 0.05 l e v e l , + s i g n i f i c a n t at the 0. 10 l e v e l LR t e s t : HO: The c o e f f i c i e n t s a r e the same a c r o s s c o h o r t s H1: They a r e d i f f e r e n t LR s t a t i s t i c = 110.18, s i g n i f i c a n t a t the .05 l e v e l . T a b l e G.10 : The V a r i a b l e Fee In V a r i o u s S e c t o r s S e c t o r s : Automotive B u s i n e s s C o n s t r u c t i o n Educ. R e s t a u r a n t s Non-Food Non-Conv. M i s c . P r o d u c t s S e r v i c e s Maintenance S e r v i c e s R e t a i l i n g Food V a r i a b l e V a r i a b l e V a r i a b l e V a r i a b l e V a r i a b l e V a r i a b l e V a r i a b l e V a r i a b l e Fee Fee Fee Fee Fee Fee Fee Fee F o r e i g n O u t l e t s (%) -0 .35 -9 .90 -1 .20 7 .64 4 .48 + -2 .42 0 . 17 0 . 26 (-0 .05) (-1 .40) (-0 27) (0 .83) (1 .71) (-0 .93) (0 .03) (0 .07) Log. Number of S t a t e s -0 .55 -0 .29 -2 .66 * 1 . 10 -0 .94 ** -0, .38 0. . 16 -0 .75 (-0 .71) (-0 .56) (-2 .47) (0 .81) (-3 .43) (-1 .08) (0 .37) (-1 .27) F r a n c h i s e e E x p e r i e n c e 3 .36 * -0 .43 - 1 . 15 1 .38 -0 . 11 -0 .43 0 .53 -0, .54 (2 .03) (-0. 51) (-0 .59) (0 .80) (-0 .24) (-0, .56) (0. .68) (-0 .38) Weeks of T r a i n i n g -0 .32 0 . 15 0 .36 0 .63 -0 .07 -0 .01 0 .25 -0 .49 (-0 .92) (0 47) (0 81) (0 .96) (-1 19) (-0 .06) (1 .26) (-1, .36) Log. O u t l e t s i n 1986 (100's) 0 .89 1. .07 * 1 . 13 0. .51 1, . 12 ** 0. .94 * 0. .41 0. .74 (1 .07) (2. .26) (1 .48) (0 57) (5 .43) (2. .57) (0. .98) (1. 27) % Time Not F r a n c h i s i n g 0 .02 0. .08 ** -0 .01 0 .04 0 .01 0. .03 * 0. ,02 0. .01 (0 74) (3. .89) (-0 15) (1. .10) (1 .65) (2. .43) (1. .62) (0. .44) Log. Years 1n B u s i n e s s -1 .59 -1 . ,85 + 1 .94 -3. .20 -0. .77 + -0. 70 -1. 41 + -3. 18 ** (-1 27) (-1. .93) (1 ,39) (-1 .44) (-1. 73) (-1. 29) (-1. .94) (-3. . 14) Growth i n O u t l e t s -2. .98 -3. 27 * 0. .45 -0. .75 -1. .43 -0. 89 -0. 86 -4 . ,83 * (-1 .01) (-2. ,22) (0, 21) (-0. .30) (-1, 3 D (-0. ,70) (-0. .65) (-2 41) Log. C a p i t a l R e q u i r e d 0. .36 1 . ,29 ** 0. .40 -1. .83 * -0. .04 0. .09 0. .40 0 98 * (0. 57) (3. 14) (0. .50) (-2. 39) (-0 17) (0. 33) (0. 86) (2. 10) F r a n c h i s o r F i n a n c i n g -0. .26 -0. 26 2. .66 - 1. .42 0 .58 1. 27 -o. 03 1 . 84 (-0. 18) (-0. ,30) (1 67) (-0. 72) (0. 73) (1. 52) (-0. .02) (1. 62) C o n s t a n t 14. .30 * 6. 74 * 5. .59 16. ,63 ** 10. ,93 ** 7. .46 ** 6. .66 * 13. ,34 ** (2. 41) (2. 16) (1, .36) (3. 24) (6. 94) (3. 45) (2. 19) (4. 04) L i m i t O b s e r v a t i o n s 1. .00 12. 00 10. .00 2, 00 0. ,00 4. 00 2. 00 6. ,00 Non-L1m1t O b s e r v a t i o n s 44. 00 84. 00 40. 00 27. 00 121 . 00 95. 00 50. 00 50. 00 S t a n d a r d e r r o r o f Estm. 3. 23 3. 84 4. 33 3. 42 1. ,97 2. 88 2. 28 3. 24 At Mean X ( I ) , E(Y) 9. 16 6. 42 5. 90 7. 41 6. 62 5. 65 5. 94 5. 84 Log L i k e l i h o o d F u n c t i o n 115. 75 -245. 29 -125. 31 -74. 17 248. 26 -241 . 07 -114. 96 -135. 65 Squared Cor(Y, E ( Y ) ) 0. 18 0. 22 0. 31 0. 48 0. 26 0. 15 0. 15 0. 26 ** s i g n i f i c a n t a t the 0.01 l e v e l * s i g n i f l e a n t at the 0.05 l e v e l , + s i g n i f i c a n t at the 0. 10 l e v e l Note: When t h e r e a r e no o b s e r v a t i o n s a t the l i m i t , the TOBIT e s t i m a t o r s i m p l i f i e s to OLS under n o r m a l i t y . LR t e s t : HO: The c o e f f i c i e n t s a r e the same a c r o s s c o h o r t s H1: They a r e d i f f e r e n t LR s t a t i s t i c = 180.58, s i g n i f i c a n t a t the .01 l e v e l . T a b l e 6.11 : The F r a n c h i s e Fee i n V a r i o u s S e c t o r s S e c t o r s : Automotive B u s i n e s s C o n s t r u c t i o n Educ. R e s t a u r a n t s Non- -Food Non-Conv. Mi s c . P r o d u c t s S e r v i c e s Maintenance S e r v i c e s R e t a i 1 i n g Food F i x e d F i x e d F i x e d F i x e d F i x e d F i x e d F i x e d F i x e d Fee Fee Fee Fee Fee Fee Fee Fee F o r e i g n O u t l e t s (%) 49 .24 + -31 . 10 -2 .04 -74. .33 11 .59 11 .07 -11, .60 2 .71 (1 .90) (-1 . 14) (-0 . 19) (-0 47) (0. 79) (0 .78) (-0, .23) (0. ,11) Log. Number of S t a t e s -2 .50 4 .28 * 3 .51 8. ,23 1. .23 -1 . 09 0. .85 4. , 17 (-0 .87) (2 .07) (1 .38) (0. .35) (0. .80) (-0 .56) (0 .26) (1 ,01) F r a n c h i s e e E x p e r i e n c e 2 .97 -2 .00 9 .63 * 18. ,00 -0. .96 0 .51 -0. .42 0. ,56 (0 .49) (-0 .58) (2 .06) (0. 61) (-0. .39) (0. 12) (-0.07) (0. .06) Weeks of T r a i n i n g -0 .91 0 . 19 -2 .30 * -1 . 10 0 .85 ** 2 .06 + 0, .94 5. .30 * (-0, .71) (0 . 14) (-2 .05) (-0. .10) (2. .68) (1 .69) (0, .64) (2. . 17) Log. O u t l e t s i n 1986 (100's) -3, .00 -0 .00 -1 . 87 -8. 88 -0. .97 0 .27 0 .82 -1 . ,66 (-0. 97) (-0.00) (-1 .02) (-0. .58) (-0. 83) (0, 14) (0. ,26) (-0. .41) % Time Not F r a n c h i s i n g -0 . 16 0 .03 -0 .05 0. 1 1 0. 01 0, .02 -0. .01 -0. ,01 (-1 . 48) (0 .39) (-0. 62) (0. 19) (0. 16) • (0. 28) (-0. ,08) (-0. ,09) Log. Years i n B u s i n e s s 2 .98 -2. .88 -3. .80 -0. 48 1. .39 -1 .64 1. .51 -8 .05 (0. .64) (-0 .76) (-1 .13) (-0. 01) (0. .55) (-0 .56) (0 .28) (-1 .20) Growth i n O u t l e t s 1 . 01 4. .98 - 1 . .24 6. 66 9. 82 -3 . 15 5. .54 5. .42 (0.09) (0 .85) (-0. 25) (0. 15) (1. ,59) (-0. .46) (0. 55) (0 .41) Log. C a p i t a l R e q u i r e d 7. .27 ** 6 .23 ** 0. .56 8. 71 3. ,94 ** 5. .21 ** 0. .99 2. 91 (2. ,87) (3 87) (0. .29) (0. 71) (3. 21) (3. 42) (0. 28) (0. 92) F r a n c h i s o r F i n a n c i n g 14 . 40 * 7 .67 * 7, . 14 + -2. .05 4. .53 7. .99 + 47. . 13 ** -1 . 30 (2. .49) (2 18) (1 .82) (-0.06) (1. .02) (1. 75) (4 .02) (-0. . 16) C o n s t a n t -10. .91 -4 . 47 16. ,95 + -24. 77 -10. ,54 -0. .04 3. .62 15. ,46 (-0. ,51) (-0 37) (1. 74) (-0. 32) (-1. 19) (-0.00) (0. 16) (0. .73) L i m i t O b s e r v a t i o n s 1 . 00 0. .00 1 . ,00 0. 00 0. 00 4. ,00 1 . ,00 0. ,00 Non-L1m1t O b s e r v a t i o n s 44. .00 96. .00 49.00 29. 00 121 . OO 95. .00 51 . .00 56. .00 S t a n d a r d e r r o r of Estm. 12. 00 15, .64 10. 69 58. 71 11. , 14 15. .68 17. .27 22. .86 At Mean X ( I ) , E(Y) 19. 67 21 , .71 16. .49 32. 91 21 . 22 20. ,34 20. .93 26. .06 Log L i k e l i h o o d F u n c t i o n -172. 23 -394. .39 -186. 77 -152. 34 -457. 61 -400. 81 -218. 82 -248. ,58 Squared Cor(Y, E ( Y ) ) 0. .36 0. .32 0. 24 0. 14 0. 24 0. 20 0. .39 0. , 19 ** s i g n i f i c a n t a t the 0.01 l e v e l , * s i g n i f i c a n t a t the 0.05 l e v e l , + s i g n i f i c a n t a t the 0.10 l e v e l Note: When t h e r e a r e no o b s e r v a t i o n s a t the l i m i t , the TOBIT e s t i m a t o r s i m p l i f i e s to OLS under n o r m a l i t y . LR t e s t : HO: The c o e f f i c i e n t s a r e the same a c r o s s c o h o r t s H1: They a r e d i f f e r e n t LR s t a t i s t i c = 281.36, s i g n i f i c a n t a t the .01 l e v e l . proportion of franchised stores. In this last case, coefficients are found to be different across sectors, but just barely. One possible explanation for the importance of the sector in the determination of the terms of the contract would be the existence, within sectors, of some form of standard franchise contract that franchisors would tend to adopt. As was pointed out above, the degree of competition franchisors face in the recruiting of franchisees in any sector may also explain the presence of these significant sectoral differences. Another explanation in the case of the royalty rates may have to do with differ-ences in costs structures in various sectors. Royalty rates are calculated on the basis of sales. If profits, as a proportion of sales, vary significantly across sectors, royalty rates should reflect this. And to an extent, it seems that they do. With respect to the fixed fee, it was noted previously that the lack of relationship between r and F might be due to the inclusion of certain fees in F. For example, it could include payment for training provided by franchisors. But fees for services are more likely to be similar across firms involved in the same type of business than across sectors. Thus, part of the total variation in F could be due to differences in these fees, which are at least partially controlled for when firms in different sectors are treated separately. This could explain the importance of the intersectoral effects for the fixed fee. According to Brickley and Dark (1987), the franchised restaurant industry is a nonrepeat type of business. As was noted above, these businesses should tend to use franchising less on average in a two-sided hidden-action framework as well. And indeed, in Table 6.9, one finds that very few restaurants are 100% franchised. The expected value of the proportion of franchised stores, at the mean values of the explanatory variables, is also significantly lower in this sector than in the others. This confirms what was established in Chapter 4, Table 4.3, i.e. that of the eight sectors studied in this section, the Restaurant industry is the least prone to use franchising. In general, the results obtained within sectors are again fairly consistent with what was observed on the sample as a whole. There are only a few significant sign reversals. For example, the number of outlets tends to increase rather than decrease franchisors' tendency to use franchising in the Construction and Maintenance sector and in the Business Aids and Services industry. Why this is so is unclear. It may be due to the presence of many relatively small or young firms in these sectors. The 151 proportion of foreign outlets has a positive rather than negative effect on the variable fee of franchisors in the Restaurant business. But it also has a significant positive effect on q//Q in that sector, which is what hidden-action models would predict. The amount of capital required significantly affects this fee positively in some sectors, and negatively in another. Under a capital-market-imperfection argument, the effect of this variable on r should be negative. However, given that this type of explanation for share contract has not received much support in these data in general, this result is not very surprising.^ Finally, the number of weeks of training offered by franchisors has a significant negative effect on the franchise fee in the Construction and Maintenance industry. These are the only cases where significant sign reversals occur. Finally, despite the importance of sectoral differences in the case of the two fees, the proportion of franchised outlets is still better explained in general by the variables included here than the terms of the contract are. The fixed fee in the sector of Business Aids and Services is the only exception to this. On the basis of the squared correlation between the observed and the expected values of the dependent variable, the amount of variation in F and q//Q that is accounted for in the regressions is about the same. 5. Conclusion It is clear from the previous section that sectoral dummy variables can not account for all the sectoral effects. The subdivision of the sample according to the type of activities firms are involved in fared significantly better. However, none of the three classification schemes used here, on the basis of the number of years in business, of the number of outlets, or of the sector in which the firm operates, clearly dominates the others. In all cases, the explanatory power of the equations was significantly enhanced by focusing on various groups of firms. But these are not nested versions of each other so that usual tests cannot be performed to determine which type of classification is best. Simply on the basis of the sum of the values of the likelihood function when eight groups are used, size cohorts are preferred in the case of q//Q. The variable fee on the other hand is better explained when one ' As was noted previously, according to Brickley and Dark (1987), the amount of capital required should have a negative effect on qf/Q and a positive effect on r. This is exactly the opposite of what the capital-market-imperfection explanation for franchising suggests. They derive this expectation from an agency cost argument. 152 concentrates on individual sectors. This gives some amount of support to the notion that differences in profit rates across sectors may account for some of the variance in royalty rates. However, the differences in the sum of the likelihood function values are relatively small across the three types of classification casting some doubt on this interpretation. Finally, size cohorts are preferred again in the case of the fixed fee, but here, the difference is really minimal. These results suggest that market power considerations are not really an im-portant factor in the determination of the two fees. In other words, the degree of competition firms face when trying to recruit franchisees does not affect the fees that much. If they did, sectoral differences would have accounted for more of the variation in the fees than size or age differences. That is unless one assumes that firms of similar size compete for franchisees across sectoral boundaries. In general, there were no real surprises in the results obtained on subsamples of firms. Mostly, what was found in Chapter 5 was confirmed here. With respect to capital-market explanations for franchising, there is no clear evidence that recently established firms, or small firms, use franchising more than others. There was how-ever some indication that these firms have not necessarily achieved their optimal proportion of franchised stores yet. Reputation effects were found in the group of oldest and the group of largest firms. In fact, whether one measures how established firms are on the basis of years in business or of their number of outlets, one obtains consistent results in general. Thus in both cases, increases in the number of years in business, controlling for other effects, lead to an increase in qf/Q and a reduction of r. This is contrary to predictions from the two-sided hidden-action models where reputation is not allowed for. Further increases in the number of outlets lead to a reduction in qf/Q and/or an increase in r, indicating that a large number of outlets for these firms does not enhance their reputation. This was explained on the basis of increased supervision costs to the franchisor. Note also that these opposite effects for these two variables, which also appeared in Chapter 5, suggest that there is no simple long term trend in franchising toward either 100% franchised or 100% company-operated chains. The last result that was noted in the case of size cohorts is the fact that the non-negative effect of risk on qf/Q found in Chapter 5, and its negative effect on r, cannot be attributed to small firms only. It is true that these are more likely than 153 large firms to be more risk-averse than their franchisees. The observed effects of risk on q//Q and r would then be consistent with the risk-sharing models for such firms. But the same effects were observed for the group of largest firms which should be less risk-averse than their franchisees. Thus the results are still not consistent with predictions from pure risk-sharing models. Finally, firms in the Restaurant business were found to resort to franchising significantly less on average than firms in other sectors. This is consistent with Brickley and Dark's (1987) notion that firms involved in nonrepeat businesses tend to operate more stores to reduce the effects of franchisee free-riding. As was noted above, this result is also consistent with two-sided hidden-action models. 154 C H A P T E R V I I Conclusion This thesis was meant to provide an empirical assessment of theories concerning share contracts in the context of the franchising phenomenon. Most of the results were outlined in the introduction. First, franchisors' propensity to use franchising was found to be non-decreasing in the amount of risk they face in their sector, measured either by the proportion of discontinued outlets or by the variance of sales in the sector. Similarly, royalty rates decreased with these measures of risk. This is not consistent with predictions from either the pure risk-sharing or the one-sided hidden-action models of share contracts. However it was noted that increased riskiness makes it more difficult for the franchisor to evaluate his franchisees' performance. This confuses issues and would affect the observed effect of risk on the proportion of franchised stores as well as the royalty rate. If increased risk leads to a greater reliance on franchising due to the greater difficulty of evaluating franchisees' performance, the results imply that indeed there is a role for risk in the determination of the contract mix and of the terms of the contract, but this role is not that which is implied by risk-sharing models of share contracts. Franchisors resort to franchising more often when monitoring downstream oper-ators becomes costlier, and use it proportionately less when the value of the inputs they themselves provide increases. This is consistent with the notion that there exist two-sided hidden-action problems in franchising. Results with respect to capital-market-imperfection arguments are rather incon-clusive. Franchisors' propensities to franchise were found to increase during periods of rapid expansion, but they decreased as the amount of capital required to open a new outlet increased. While the first of these results lends support to the notion that franchisors use franchising in order to obtain capital when they need it most, the second does not. Thus it appears that franchising relaxes some form of constraint franchisors face in trying to expand their operations, but whether this is a financial constraint remains unclear. With respect to the terms of the share contract, the explanatory power of the model was found to be much better for the proportion of franchised stores than for any of the two fees. This remained true in general within subgroups of firms defined 155 on the basis of the number of years they had been in business, or the number of outlets they have, or the sector in which they operate. Thus, in response to changes in the exogenous variables considered here, franchisors, who have a choice between modifying the terms of their franchise contract or changing the proportion of stores they want to franchise, tend to do mostly the latter. The lack of correlation between the royalty rate and the franchise fee, and the fact that the model is less satisfactory relative to these fees, suggests that there are considerations in the determination of F and r that have not been taken into account here or in the theories. One such consideration was discussed with respect to the franchise fee, i.e. the possibility that it includes the price of services provided by the franchisors. The fact that franchisors use input sales as another means to extract rent from franchisees may also contribute to the lack of correlation between the two fees. Finally, equation (5.4) was derived under the assumption that all remaining surplus at the downstream level, given r, should be extracted through F. The lack of relationship between the fees could be an indication that this assumption is incorrect, and that there are in fact rents left at the downstream level. This would be consistent with the existence of queues of potential franchisees in many franchise chains. Testimonies of existing franchisees in important chains also tend to corroborate the notion that there are rents left downstream in such chains. The analysis carried out in this thesis suggests interesting directions for both empirical and theoretical research in this area. First, at the empirical level, one issue that could not be addressed here due to the nature of the data set, concerns the way in which franchisors' choices of contract mix and contract terms evolve through time. This requires access to panel data, which I am now in the process of gathering. It will allow me to see if the results obtained here hold in a more dynamic context. Since franchise contracts are just one type of share contract, a second interesting empirical issue is whether results obtained in this thesis are specific to franchising, or whether they apply in other contexts. One way to verify this for example would be to examine issues involved in patent licencing contracts. Very little has been done in that area. At the theoretical level, I believe there are many issues that would warrant some attention. First, as was pointed out previously, existing theoretical work on franchis-ing has focused on single franchisor-franchisee pairs, thereby restricting franchisors 156 to a single control variable, namely the royalty rate. Yet, it is clear empirically that franchisors adjust the proportion of stores they franchise in response to differences in risk, supervision costs, etc. Consequently, theoretical work that would address the question of firms' choices of contract mix as well as their decisions concerning the terms of the contract would be quite useful. One stylised fact about franchising that could also be examined at the theoretical level is the tendency of franchisors to offer the same franchise contract to all their potential franchisees at a point in time. They do not discriminate according to the individual involved, or even depending on location. One way to rationalize this was discussed in this thesis. Franchisors might have other instruments at their disposal that may allow them to discriminate among franchisees (e.g. input sales) and/or to make various locations relatively similar (density of stores) so that they do not need different franchise contracts. Finally, the fact that franchisors require a proportion of sales rather than profits from their franchisees cannot really be explained in the context of the models discussed here. It has implications in terms of the objectives of the two parties that may explain a lot of the conflicts that arise between franchisors and franchisees. In that sense, it might be an interesting issue to explore. 157 B I B L I O G R A P H Y Alchian, A. and Demsetz, H., (1972), "Production, Information Costs, and Economic Organization", American Economic Review, 62, 777-795. Allen, F., (1982), "On share contracts and screening", Bell Journal of Economics, 13, 541-547. Alston, L.J., Datta, S.K. and J.B. Nugent, (1984), "Tenancy Choice in a Competitive Framework with Transactions Costs", Journal of Political Economy, 92, 1121-1133. Alston, L.J. and R. Higgs, (1982), "Contractual Mix in Southern Agriculture since the Civil War: Facts, Hypotheses and Tests", Journal of Economic History. 42, 327-355. Bardhan, P.K., (1977), "Variations in Forms of Tenancy in a Peasant Economy". Journal of Development Economics, 4, 105-118. Bardhan, P.K. and T.N. Srinivasan, (1971), "Sharecropping Tenancy in Agriculture: A Theoretical and Empirical Analysis", American Economic Review, 61(1), 48-64. Binswanger, H.P. and M.R. Rosenzweig, (1984), "Contractual Arrangements, Em-ployment, and Wages in Rural Labor Markets: A Critical Review", in Contrac-tual Arrangements, Employment, and Wages in Rural Labor Markets in Asia, H.P. Binswanger and M.R. Rosenzwieg (eds), Yale University Press. Blair, Roger D. and David L. Kaserman, (1980), "Vertical Control with Variable Proportions: Ownership Integration and Contractual Equivalents", Southern Economic Journal, April, 1118-1128. Blair, Roger D. and David L. Kaserman, (1982), "Optimal Franchising", Southern Economic Journal, October, 494-504. Brickley, J.A. and F.H. Dark, (1987), "The Choice of Organizational Form: The Case of Franchising", Journal of Financial Economics, 18, 401-420. Caves, Richard E., Crookell Harrold and J. Peter Killing, (1983), "The Imperfect Market for Technology Licences", Oxford Bulletin of Economics and Statistics, 45(3), 249-267. 158 Caves, Richard E. and William F. Murphy II, (1976), "Franchising: Firms, Markets, and Intangible Assets", Southern Economic Journal, 42, April, 572-586. Chao, K., (1983), "Tenure Systems in Traditional China", Economic Development and Cultural Change, 31, 295-314 Cheung, S.N.S., (1969), The Theory of Share Tenancy, Chicago: University of Chicago Press. Clay, E.C., (1976), Institutional Change and Agricultural Wages in Bangladesh, Bangladesh Development Studies, 4, 423-440. Datta, S.K., D.J. O'Hara and J.B. Nugent, (1985), Choice of Agricultural Tenancy in the Presence of Transactions Costs, mimeo. Davidson, W.H. and D.G. McFetridge, (1984), "International Technology Transac-tions and the Theory of the Firm", Journal of Industrial Economics, 32(3), 253-264. Daj?, R., (1967), "The Economics of Technological Change and the Demise of the Sharecropper", American Economic Review, 57, 427-449. Deaton, A. and J. Muellbauer, (1980), Economics and Consumer Behaviour, Cam-bridge University Press. Diewert, W.E. , (1982), "Duality Approaches to Microeconomic Theory", in Handbook of Mathematical Economics, Vol. II, edited by K.J. Arrow and M.D. Intriligator, North-Holland Publishing. Dixit, A., (1983), "Vertical Integration in a Monopolistically Competitive Industry", International Journal of Industrial Organization, 1, 63-78. Eswaran, M. and A. Kotwal, (1985), " A Theory of Contractual Structure in Agricul-ture", American Economic Review, June, 75(3), 352-367. Gallini, N. and R.A. Winter (1983), "On Vertical Control in Monopolistic Competi-tion", International Journal of Industrial Organization, 1,275-286. Gibbons, J.D., (1976), Nonparametric Methods for Quantitative Analysis, Holt, Rine-hart and Winston. Goldberg, J.D.,(1982), A Transaction Costs Analysis of Franchising with Empirical Applications to the Restaurant Industry, Ph.D. Dissertation, University of Penn-sylvania. 159 ' Hallagan, W., (1978), "Self-Selection by Contractual Choice and the Theory of Sharecropping", Bell Journal of Economics, 9, 344-354. Higgs, R., (1973), "Race, Tenure and Resource Allocation in Southern Agriculture, 1910", Journal of Economic History, 33, 149-169. Holmstrom, B., (1982), "Moral Hazard in Teams", Bell Journal of Economics, 13, 324-340. Huang, Y., (1974), "Risk, Entrepreneurship and Tenancy", Journal of Political Economy, 81, 1241-1244. Hunt, S., (1973), "The trend Toward Company-Operated Units in Franchise Chains", Journal of Retailing, Summer, 3-12. Jaynes, G.D., (1984), "Economic Theory and Land Tenure", in Contractual Arrange-ments, Employment, and Wages in Rural Labor Markets in Asia, H.P. Bin-swanger and M.R. Rosenzwieg (eds), Yale University Press. Jodha, N.S., (1984), "Agricultural Tenancy in Semiarid Tropical India", in Contrac-tual Arrangements, Employment, and Wages in Rural Labor Markets in Asia, H.P. Binswanger and M.R. Rosenzweig (eds), Yale University Press. Lazear, E.P. and S. Rosen, (1981), "Rank-Order Tournaments as Optimum Labour Contracts", Journal of Political Economy, 89, 841-864. Maddala, G.S., (1983), Limited-Dependent and Qualitative Variables in Econometrics, Cambridge University Press. Martin, R.E., (1987), Franchising and Risk Management, mimeo, Louisiana State University, May. Mathewson. G.F. and R.A. Winter, (1983), "Vertical Integration by Contractual Restraints in Spatial Markets", Journal of Business, 56, 497-517. Mathewson. G.F. and R.A. Winter, (1984), "An Economic Theory of Vertical Re-straints", Rand Journal of Economics, 15, 223-234. Mathewson. G.F. and R.A. Winter, (1985), "The Economics of Franchise Contracts", Journal of Law and Economics, 28, October, 503-526. , Monteverde, K. and D.J. Teece, (1982), "Supplier Switching Costs and Vertical Integration in the Automobile Industry", Bell Journal of Economics, 13, Spring, 206-213. 160 Mookherjee, D., (1984), "Optimal Incentive Schemes with Many Agents", Review of Economic Studies, 51, 433-446. Murrell, P., (1983), "The Economics of Sharing: a Transactions Cost Analysis of Contractual Choice in Farming", Bell Journal of Economics, 14, 283-293. Nalebuff, B. and J. Stiglitz, (1983), "Prices and Incentives: Towards a General Theory of Compensation and Competition", Bell Journal of Economics, 14, 21-43. Newbery, D.M.G. and J.E. Stiglitz, (1979), "Sharecropping, Risk Sharing and the Importance of Imperfect Information", in Risk, Uncertainty and Agricultural Development, edited by Roumasset J., Broussard J. and I. Singh, SEARCH-ADC Publication. Norton, S.W., (1988), "An Empirical Look at Franchising as an Organizational Form", Journal of Business, 61, 197-217. Oxenfeldt A.R. and A.O. Kelly, (1969), "Will Successful Franchise Systems Ultimately Become Wholly-Owned Chains?", Journal of Retailing, 44, 69-87. Rao, C.H.H., (1971), "Uncertainty, Entrepreneurship, and Sharecropping in India", Journal of Political Economy, 79, 578-595. Rao, C.H.H., (1975), Technological Change and Distribution of Gains in Indian Agriculture, Delhi: Institute of Economic Growth. Rees, R, (1985), "The Theory of Principal and Agent: Part 1 and Part 2", Bulletin of Economic Research, 37. Reid, J.D. Jr, (1973), "Sharecropping as an Understandable Market Response: The Post-Bellum South", Journal of Economic History, 33, 106-130. Reid, J.D. Jr, (1976), "Sharecropping and Agricultural Uncertainty", Economic Development and Cultural Change, 24, 549-576. Reid, J.D. Jr, (1977), "The Theory of Share Tenancy Revisited — Again", Journal of Political Economy, 85(2), 403-407. Roberts, M.J., (1984), "Testing Oligopolistic Behaviour", International Journal of Industrial Organization, 2, 367-383. Roumasset, J., (1984), "Explaining Patterns in Landowner Shares: Rice, Corn, Co-conut and Abaca in the Philippines", in Contractual Arrangements, Employment, 161 and Wages in Rural Labor Markets in Asia, H.P. Binswanger and M.R. Rosen-zweig (eds), Yale University Press. Rubin, H., (1978), "The Theory of the Firm and the Structure of the Franchise Contract", Journal of Law and Economics, 21, April, 223-233. Scott, F.A. Jr., (1987), Vertical Control Via Franchising, mimeo, Universitjr of Kentucky, Feb. Shaban, R.A.,(1987), "Testing between Competing Models of Sharecropping", Jour-nal of Political Economy, 95, 893-920. Singh, N.,(1987), Theories of Share Cropping, Discussion paper #155, University of California-Santa Cruz, June. Schmalensee, R., (1973), "A Note on the Theory of Vertical Integration", Journal of Political Economy, 81, 442-449. Spence, M. , (1973), "Job Market Signalling", Quarterly Journal of Economics, 87, 355-379. Stiglitz, J.E., (1974), "Incentives and Risk-Sharing in Sharecropping", Review of Economic Studies, 41, 219-255. Thompson, D.N., (1971), Franchise Operations and Antitrust. Heath Lexington Books. Tobin, J, (1958), "Estimation of Relationships for Limited Dependent Variables", Econometrica, 26, p. 24-36. Vaughn, C.L., (1979), Franchising, Its Nature, Scope and Advantages, 2nd and revised edition, Lexington Mass: Lexington. Books. Vernon J.M. and D.A. Graham, (1971), "Profitability of Monopolization by Vertical Integration", Journal of Political Economy, 79, 924-925. Warren-Boulton, F.R., (1974), "Vertical Control with Variable Proportions", Journal of Political Economy, 82, 783-802. White, K.J., (1978), "A General Computer Program for Econometric Methods — Shazam", Econometrica, January, 239-240. Williamson, 0., (1971), "The Vertical Integration of Production: Market Failure Considerations", American Economic Review, May, 112-123. 162 S O U R C E S O F D A T A Association of Canadian Franchisors, (1985), Franchising: The Current Situation, Toronto. Entrepreneur Magazine, Annual Franchise 500, Jan. 1987, 15, 98-183; Jan. 1986, 14, 55-133; and Jan. 1985, 13, 55-110. Entrepreneur Magazine, Franchise Yearbook, 1987, 1; 1987-88, 2. Groupe Promexpo Inc., Retailing and Services Guide 87, 1987. Info Press Inc, The Franchise Annual, 1986; 1987, ed. by Edward L. Dixon. International Monetary Fund, (1987), International Financial Statistics, X L , June. National Restaurant Association Research and Information Services, (1984), Fran-chise Restaurants: A Statistical Appendix to Food Service Trends, Feb. Ozanne, U.B. and S.D. Hunt, (1971), The Economic Effect of Franchising, U.S. Sen-ate, Select Committee on Small Business, Washington, Gov. Printing Office. Retail Council of Canada, (1986), Franchisee Attitude Survey, published in The Info Franchise Newsletter, Info Press Inc., 10, May. Statistics Canada, Franchising in the Economy 1976-1981, mimeo. Udell, G.G. (1972), "The Franchise Agreement", Cornell Hotel and Restaurant Administration Quaterly, 13, Aug., 13-21. U.S. Department of Commerce, Franchise Opportunities Handbook, compiled by Andrew Kostecka, Oct. 1985; Nov. 1986. U.S. Department of Commerce, Franchising in the Economy, 1982-84, Jan. 1984; 1983-85, Jan. 1985; 1984-86, Jan. 1986; 1985-87, Jan. 1987, prepared by Andrew Kostecka. Venture, (1985), The Franchise 100, A Look at Returns After the First Two Years, Sept., 40-46. Venture, (1986), The Franchise 100, Nov., 52-57. 163 A P P E N D I X A : D E S C R I P T I V E S T A T I S T I C S F O R T H E S A M P L E O F 890 F R A N C H I S O R S 164 Table A.I : Number of O u t l e t s f o r the 890 F r a n c h i s o r s Sectors 1- AUTOMOTIVE PRODUCTS PARTS AND SERVICES BRAKES. MUFFLERS ft SHOCKS TRANSMISSION TUNE-UP 2- BUSINESS AIDS ft SERVICES ACCOUNTING AND COLLECTION EMPLOYMENT SERVICES PRINTING AND COPYING REAL ESTATE 3- CONSTRUCTION & MAINTENANCE MAID SERVICES CONSTRUCTION HOME IMPROVEMENT ft REPAIRS CARPET CLEANING 4- CONVENIENCE STORES 5- EDUCATIONAL PRODUCTS ft SERVICES EDUCATIONAL ft TRAINING PROGRAMS HEALTH AIDS ft SERVICES 6- RESTAURANTS FAST FOOD - CHICKEN - MEXICAN - HAMBURGER - PI22A - SUBMARINES TABLE SERVICE - STEAKHOUSES - ITALIAN - FULL MENU 7- HOTELS. MOTELS & CAMPGROUNDS 8- LAUNDRY ft DRY CLEANING 9- RECREATION ft TRAVEL TRAVEL AGENCIES 10- AUTO ft TRUCK RENTALS 11- EQUIPMENT ft TOOL RENTAL 12- NON FOOD RETAILING CLOTHING ft SHOES FURNITURE ft ACCESSORIES ART SUPPLIES COMPUTER PRODUCTS VIDEO RENTAL 13- FOOD RETAILING - NON CONVENIENCE DONUT SHOPS ICE CREAM PARLORS SPECIALTY FOOD SHOPS 14- MISCELLANEOUS BEAUTY SALONS ALL SECTORS TOTAL OUTLETS OUTLETS OUTLETS OUTLETS YEARS YEARS X TIME OUTLETS i n 1986 In 1986 In 1986 In 1986 In BUS. FRANC. FRANC. N I n 1986 Minimum Max 1mum Mean St. Dev. Mean Mean 71 9670 2 2006 136.2 320.2 18.9 11.0 0 .334 13 2602 3 1706 20O. 1 460. 1 31.2' 14 .O O . 396 7 2957 5 2006 422.4 742. 1 15.9 13.7 0 . 199 9 596 7 142 66.2 48.2 19.6 . 13.7 0 . 267 17 1352 4 398 79.5 113.3 9. 1 6.9 0 . 245 147 21822 2 6597 148.4 576.9 13.7 8.5 0 .349 26 2684 4 885 103.2 184.2 13.3 8.8 0 .311 33 2029 3 439 61.5 98.9 19.7 12.3 O .388 22 4917 4 1081 223.5 332.8 12.7 9.0 0. 284 10 9464 13 6597 946.4 2040.1 15.4 9.8 0 . 242 86 15981 2 3365 185.8 425.7 15.0 9.5 0 .319 10 978 4 320 97.8 107.2 8.8 6.3 0 .284 22 1458 4 4O0 66.3 102.2 12.6 8. 1 0 .287 17 1557 2 810 91.6 210. 1 9.0 5.6 0 332 13 8197 22 3365 630. 5 896.9 24.7 16.4 0 295 13 3844 6 1276 295.7 369.9 25.3 18.2 O. 217 47 6522 2 3319 138.8 490. 1 14.0 8.0 0. 383 21 1237 2 273 58.9 81.1 18. 1 11.1 0. 391 26 5285 3 3319 203.3 653.5 10.7 5.5 0. 376 201 70282 2 9060 349.7 1079.5 20. 1 12.9 O 319 140 61260 2 9060 437.6 1275.7 19.8 12.5 0. 324 15 1 1807 4 82CO 787 . 1 2087.6 22.0 15.4 0. 272 16 3519 10 2292 219.9 562 .9 20.3 14.6 o 274 13 20033 2 9060 1541.0 2686.1 29. 1 20.2 0. 251 37 13882 6 5571 375. 2 1055.3 18. 1 11.9 o. 315 17 2059 5 800 12 1.1 191.1 17.9 11.8 o. 337 61 9022 3 1268 147.9 233.9 20.8 13.8 0. 308 8 2302 1 1 623 287.7 258.2 23. 1 20.5 0. 130 1 1 1064 10 320 96 . 7 87 . 3 18.7 12.8 o. 315 22 4 130 5 1268 187.7 322.9 23. 1 14.3 0. 353 18 6646 2 1672 369.2 399.5 24.3 18.9 o. 205 8 1334 2 1084 166.7 373.8 10.5 7.9 0. 252 20 1746 6 341 87.3 105.4 11.1 8.2 0. 249 8 1 108 21 308 138.5 108.6 11.4 7.7 0. 158 10 6620 11 3320 682.0 1013.7 14.6 12.4 0. 216 1 373 373 373 373.0 O.O 42.0 24.0 0. 429 163 19229 2 1283 118.0 205. 1 16.0 9.2 0. 329 24 2361 5 466 98.4 129.8 17.6 11.0 0. 267 11 1 178 21 313 107 . 1 118.3 19.5 11.3 0. 353 12 937 6 323 78 . 1 96.3 15.0 8.3 0. 321 15 1601 2 576 106.7 153.5 6.6 4.7 0. 291 15 2107 3 703 140. 5 212.7 5. 1 4. 1 0. 167 77 17120 2 4894 222 . 3 686.9 17.7 11.1 0 326 14 3329 4 1512 237 .8 435.2 22.8 14.4 0. 338 23 1 1021 9 4894 479.2 1 177.8 21.6 15.7 0. 244 12 866 2 495 72 . 2 137.9 17.6 10. 1 0. 440 28 3572 2 863 127.6 200. 1 10.4 6.8 0. 322 19 3389 4 863 178 .4 226.7 12. 1 8.2 0. 285 890 184960 2 9060 207.8 657. 1 16.7 10.5 0. 324 Table A.2 : P r o p o r t i o n of F r a n c h i s e d O u t l e t s f o r the 548 F r a n c h i s o r s Sectors 1- AUTOMOTIVE PRODUCTS PARTS AND SERVICES BRAKES, MUFFLERS S SHOCKS TRANSMISSION TUNE-UP 2- BUSINESS AIDS ft SERVICES ACCOUNTING AND COLLECTION EMPLOYMENT SERVICES PRINTING ANO COPYING REAL ESTATE 3- CONSTRUCTION ft MAINTENANCE MAID SERVICES CONSTRUCTION HOME IMPROVEMENT & REPAIRS CARPET CLEANING 4- CONVENIENCE STORES 5- EDUCATIONAL PRODUCTS ft SERVICES EDUCATIONAL 6 TRAINING PROGRAMS HEALTH AIDS ft SERVICES 6- RESTAURANTS FAST FOOD - CHICKEN - MEXICAN - HAMBURGER - PIZZA - SUBMARINES TABLE SERVICE - STEAKHOUSES - ITALIAN - FULL MENU 7- HOTELS. MOTELS ft CAMPGROUNDS 8- LAUNDRY & DRY CLEANING 9- RECREATION ft TRAVEL TRAVEL AGENCIES 10- AUTO & TRUCK RENTALS 11- EQUIPMENT ft TOOL RENTAL 12- NON FOOD RETAILING CLOTHING ft SHOES FURNITURE ft ACCESSORIES ART SUPPLIES COMPUTER PRODUCTS VIDEO RENTAL 13- FOOD RETAILING - NON CONVENIENCE DONUT SHOPS ICE CREAM PARLORS SPECIALTY FOOD SHOPS 14- MISCELLANEOUS BEAUTY SALONS ALL SECTORS X FRANC. X FRANC. X FRANC. X FRANC. X EXP . X FIN N Minimum Maximum Mean St . Dev. Mean Mean 71 0 .333 1 .ooo 0 .841 0 .200 O .211 0 .211 13 0 .368 1 .000 0 .867 0 . 191 0 .077 0 .231 7 0 .400 1 .000 0 . 749 0 .243 0 .286 0 . 143 9 0 .857 1 ooo o .975 0 .046 O .111 0 . OOO 17 0 . 444 1 .000 0 .786 0 .230 0 .353 0 . 176 147 0 .088 1 .ooo 0 .872 0 . 194 0 .388 0 .408 26 0 500 1 .000 0 .908 0 . 162 0 .615 0 .423 33 0 .088 1 .000 0 .794 0 .283 0 . 182 0 .424 22 0 .667 1 .000 0 .940 0 . 100 0 .091 0 .409 lO 0 . 396 1 ooo 0 .922 0 . 193 0 . 70O 0 .700 86 0. 444 1 .000 0 921 0 . 134 0 .221 0 .384 10 0 . 750 1 .ooo 0 .92 1 0 .097 0 . ICO 0 , 40O 22 0 . 444 1 .ooo 0 .903 0 . 165 0 . 409 0 . 136 17 0. .500 1 .000 0 .907 0 . 168 0 .235 0 .353 13 0 .831 1 .OOO 0 . 970 0 .050 0 .077 0 .769 13 0 . 114 1 .OOO 0 .806 0 .227 O .077 0 .231 47 0 , 345 1 .000 0 .805 0 . 175 0 .511 0 . 170 2 1 O . 345 1 .ooo o . 774 0 . 223 0 .381 0 . 238 26 0 .571 1 .000 0 .830 0 . 123 0 .615 0 . 1 15 201 0. 054 1 .000 0 .717 0 .252 0 .398 0 .060 140 0 . 1 18 1 OOO o. .7 13 0 .247 0. .357 0 ,043 15 0. 225 1 OOO 0 655 0 278 0 600 0 OOO 16 0. 185 1 .000 0. 771 0 .233 0. .437 0. .000 13 0. 158 0 .988 0 641 0. 290 0. 615 0. OOO 37 0. 200 1 .000 0 746 0. 216 0. .324 0. 054 17 0. 596 1 .000 0 882 0 . 137 0 059 0. 176 61 O 054 1 .OOO 0 725 0. .264 0. 492 0 098 8 0. 159 0 .995 0 753 0 .317 0 .500 0 125 1 1 0. 495 1 .000 0. 801 0. 187 0 .364 0. . 182 22 o. 054 O .988 0 671 0. 285 0 591 0. OOO 18 0. 321 1 ooo 0 819 0. 218 0. 444 0. 111 8 o. 500 1 .000 0. 832 0. 166 0. 250 0. 125 20 O. 300 1 OOO o. 870 0 200 0. 250 0. 250 8 0. 952 1 .000 0. 981 0. 017 0. 125 0. 125 10 0. 364 1 .000 0. 924 0. 198 0. 300 o. 500 1 0. 815 0 .815 o. 815 o. OOO 0. OOO o. OOO 163 0. 103 1 .000 0. 831 0. 216 0. 239 0. 129 24 0. 103 1. OOO 0. 796 0. 240 0. 167 o. 083 1 1 0. 167 1 OOO o. 737 0. 274 0. 364 0. 091 12 0. 520 1 .000 0. 884 0. 131 0. 000 0. 167 15 0. 500 1 .000 0. 878 0. 148 0. 667 0. OOO 15 o. 263 1 ooo o. 860 0. 213 o. 133 o 067 77 0. 016 1 .000 0. 801 0. 244 0. 208 0. 104 14 o. 016 1. OOO 0. 864 0. 266 0. 07 1 0. 357 23 0. 633 1. .000 0. 916 0. 105 0. 217 0. 043 12 0. 1 11 0. .983 0. 590 0. 301 0. 167 0. 083 28 o. 154 i ooo 0. 724 o. 277 o. 036 o. 321 19 0. 154 1. 000 0. 727 0. 308 0. 000 0. 316 890 0. 016 1. 000 0. 816 o. 224 0. 303 0. 204 Table A.3 : Royalty Rates for the 548 Franchisors ROYALTY Sectors N Minimum 1- AUTOMOTIVE PRODUCTS 71 0.0 PARTS AND SERVICES 13 0.0 BRAKES. MUFFLERS & SHOCKS 7 9.0 TRANSMISSION 9 6.0 TUNE-UP 17 7.0 2- BUSINESS AIDS & SERVICES 147 0.0 ACCOUNTING AND COLLECTION 26 0.0 EMPLOYMENT SERVICES 33 O.O PRINTING AND COPYING 22 1.5 REAL ESTATE 10 0.0 3- CONSTRUCTION & MAINTENANCE 86 0.0 MAID SERVICES 10 0.0 CONSTRUCTION 22 0.0 HOME IMPROVEMENT & REPAIRS 17 0.0 CARPET CLEANING 13 0.0 4- CONVENIENCE STORES 13 0.0 5- EDUCATIONAL PRODUCTS 8. SERVICES 47 0.0 EDUCATIONAL & TRAINING PROGRAMS 21 2.0 HEALTH AIDS & SERVICES 26 0.0 6- RESTAURANTS 201 0.0 FAST FOOD 140 0.0 - CHICKEN 15 4.0 - MEXICAN 16 2.0 - HAMBURGER 13 3.0 - PIZZA 37 3.5 - SUBMARINES 17 5.0 TABLE SERVICE 61 1.0 - STEAKHOUSES 8 2.0 - ITALIAN 11 5.0 - FULL MENU 22 3.5 7- HOTELS, MOTELS & CAMPGROUNDS 18 1.0 8- LAUNDRY & DRY CLEANING 8 0.0 9- RECREATION & TRAVEL 20 0.0 TRAVEL AGENCIES 8 0.0 10 - AUTO 6 TRUCK RENTALS 10 3.5 11 - EQUIPMENT & TOOL RENTAL 1 2.7 12 - NON FOOD RETAILING 163 0.0 CLOTHING & SHOES 24 2.5 FURNITURE & ACCESSORIES 11 2.0 ART SUPPLIES 12 1.0 COMPUTER PRODUCTS 15 0.0 VIDEO RENTAL 15 0.0 13 - FOOD RETAILING - NON CONVENIENCE 77 0.0 DONUT SHOPS 14 0.0 ICE CREAM PARLORS 23 0.0 SPECIALTY FOOD SHOPS 12 2.0 14 - MISCELLANEOUS 28 0.0 BEAUTY SALONS 19 0.0 ALL SECTORS 890 0.0 Note: Royalty rates here Include the adve r t i s i n g fee f i s s p e c i f i e d an a % of sales or gross revenues. ROYALTY ROYALTY ROYALTY FIXED ROYALTY N exc. f i x e d exc. Maximum Mean St. Dev. ' Mean Mean fix e d 19.0 8 .8 3. 8 0.056 9. ,3 67 10.0 6 .0 3. .5 0. 154 7. . 1 1 1 19.0 11 .9 3. ,5 0.000 11. .9 7 13.0 9 . 1 2. .4 0.000 9 . 1 9 16.5 10 .7 2. 9 0.000 10, .7 17 18.0 6 .5 3. 8 0. 116 7, .4 130 17.0 6 .2 4. ,0 0. 115 7, .0 23 14.0 7 .2 2. .6 0.030 7 .4 32 10.0 6 .7 1 .9 0.000 6, .7 22 14.0 5 .6 4 . 3 0. 100 6 .3 9 20.0 6 .5 3. .9 0. 140 7 .7 73 10.0 e .9 2 9 0. 100 7 .7 9 10.0 4 .6 3, .5 0.273 6 .4 16 10.5 6 .4 3 .6 0. 118 7 .7 14 17.0 6 . 1 4. .6 0.231 7 .9 lO 15.0 5 .8 5 .0 0. 154 6 .8 11 25.0 7 .2 3 .7 -0.043 7, .5 45 11.0 8 . 1 2 .4 0.000 8 . 1 21 25.0 e .5 4, .4 0.077 7 . 1 24 15.5 6 .5 2 . 1 0.005 6 .5 20O 15.5 6 .7 2 . 1 0.007 6 .8 139 10.0 7 .2 1 .9 0.000 7 .2 15 10.0 5 .7 2 .4 0.000 5 .7 16 15.5 8 .0 2. .8 0.000 8 .0 13 10.0 6 .7 1 .8 0.000 6 .7 37 13.0 7 .4 2. . 1 0.000 7 .4 17 10.0 5 .8 1 .9 0.000 5 .8 61 6.8 4 .3 1 .6 0.000 4 .3 8 10.0 7 .3 2 .0 0.000 7 .3 11 9.0 5 .9 1 .6 0.000 5 .9 22 10.0 5 .4 2. .0 0.000 5 .4 18 7.0 3 .5 2. .9 0.250 4 .7 6 12.0 5 .9 3 .5 0. 150 7 .0 17 10.0 5 .0 4 .3 0.375 8 .0 5 10.0 7 .2 1 .9 0.000 7 .2 lO 2.7 2 .7 0 .0 o.ooo 2 .7 1 16.5 5 .8 3 . 1 0.074 6 .3 151 1 1 .0 5 .6 2. .6 0.000 5 .6 24 12.0 6 . 1 2. .6 0.000 6 . 1 11 10.0 6 .2 2 .8 O.OOO 6 .2 12 10.0 5 .6 2. .9 0. 133 6 .4 13 10.0 5 .3 2 .3 0.067 5 .7 14 11.0 6 .3 2 .3 0.026 6 .5 75 10.0 7 .4 2 .6 0.071 8 .0 13 9.0 5 .3 2. .5 0.043 5, .6 22 11.0 6 .2 2. ,5 0.000 6 .2 12 15.0 7 .4 3 .4 0.071 7 .9 26 15.0 7 .7 3. .6 0.053 8 . 1 18 25.0 6 .5 3. .3 0.066 7 .0 830 r -ranchlsees pay to franchisors when this fee Averages were used when ranges were given. Table A.4 : Franchise Fees for the 548 Franchisors F. Fee F. Fee F. Fee F. Fee Cor(F.r) Spearman COR(F'.r') Spearman Sectors N Minimum Maximum Mean St. Dev. Rank Cor. Rank Cor. 1- AUTOMOTIVE PRODUCTS 71 0.0 424.5 36, .3 66. .4 -0.21 -0.09 -0.22 -0.06 PARTS AND SERVICES 13 0.5 283.0 42, ,7 76. 8 -0.61 -0. 12 -0.69 -0.39+ BRAKES. MUFFLERS & SHOCKS 7 10.0 22.5 15, .8 5. ,3 0.30 -0.04 0.32 0. 1 1 TRANSMISSION 9 14.4 137.8 32, ,6 39. .6 -0.33 -0.28 -0.41 -0.07 TUNE-UP 17 10.0 195.9 34. . 1 44. 6 -0.26 -0. 12 -0.42 -0.35+ 2- BUSINESS AIDS & SERVICES 147 0.0 136.0 20, .2 16. 8 0.03 -0.03 0.02 0.08 ACCOUNTING AND COLLECTION 26 0.0 71.0 18, . 1 16, .9 0.43 0.38* 0.46 0.40* EMPLOYMENT SERVICES 33 0.0 50.0 20, .4 12. ,0 0.09 0.09 0. 12 0.09 PRINTING AND COPYING 22 1.5 42.5 24, .4 12. .0 0.07 0.02 0.03 0.05 REAL ESTATE 10 5.0 19. 1 10, . 1 4. . 1 0. 13 0.38 0. 10 0.27 3- CONSTRUCTION & MAINTENANCE 86 0.0 69.0 16, .4 13, .9 -0.06 -0.01 -0.05 -0.02 MAID SERVICES 10 6.5 31 . 1 12 . 1 7, .2 -0.77 -0.24 -0.79 -0.64* CONSTRUCTION 22 1.5 69.0 19. 2 17. , 1 -0.04 -0.02 0.34 0.44* HOME IMPROVEMENT & REPAIRS 17 0.0 56.9 18. 8 14. ,0 0. 1 1 0.21 0.21 0. 19 CARPET CLEANING 13 5.8 28.5 14, a 6 .2 0.13 0.41+ 0.37 0.39+ 4- CONVENIENCE STORES 13 0.0 30.0 14. . 1 B. .9 -0.33 -0.35 -0.47 -0.44+ 5- EDUCATIONAL PRODUCTS & SERVICES 47 0.0 286.2 27, .7 40, ,7 -0.07 -0.06 -0.00 0. 1 1 EDUCATIONAL & TRAINING PROGRAMS 21 0.0 54.0 23 . 1 11. .5 0.26 0.24 0.31 0.25 HEALTH AIDS & SERVICES 26 0.5 286.2 31 , 5 53. 9 -0.09 -0.06 0.03 -0.06 6- RESTAURANTS 201 2.5 87.5 19. ,7 10. .6 0.05 0. 18** 0.05 0. 16* FAST FOOD 140 2.5 87.5 17. .9 9. .3 0.20 0.29** 0.20 0.27** - CHICKEN 15 10.0 30.0 19. 1 7. , 1 0.30 0.23 0.32 0. 19 - MEXICAN 16 6.0 35.0 16. 7 6. ,7 0.64 0.73** 0.51 0.66** - HAMBURGER 13 10.0 87.5 28. .7 19. .9 -0.08 0. 13 0.04 0.06 - PIZZA 37 2.5 28.8 15, .5 5. .8 0. 15 0.06 0.25 0. 18 - SUBMARINES 17 5.0 39.0 15, .2 8. .3 0. IB 0.38+ 0. 14 0.27 TABLE SERVICE 61 5.4 64.5 23. .9 12, .3 -0.07 0.03 -0.07 0.03 - STEAKHOUSES 8 10.0 50.0 23. .5 12. .5 0.76 0.72* 0.81 0.69* - ITALIAN 11 10.0 25 . 2 16. .8 5 .5 0.85 0.84** 0.75 0.79** - FULL MENU 22 5.4 50.0 27. .4 10 5 O.OB 0. 15 0.03 -0.06 7- HOTELS. MOTELS 8 CAMPGROUNDS 18 6.0 50.0 24. 8 12. .9 0.21 0.24 0.20 0.24 8- LAUNDRY & DRY CLEANING B 10.0 66.5 30. ,9 19. .6 -0.80 -0.74* -0.77 -0.36 9- RECREATION ft TRAVEL 20 6.0 127.5 35, .3 29, O -0.41 -O. 16 -0.41 -O. 15 TRAVEL AGENCIES 8 20.5 127.5 50. .4 34, .8 -0.48 -0.24 -0.60 -0.57+ 10- AUTO & TRUCK RENTALS 10 3.5 30.0 17, .4 10, .2 0. 10 0.35 0. 10 0.27 11- EOUIPMENT & TOOL RENTAL 1 20.0 20.0 20, .0 n.a. n.a. n.a. n.a. n.a. 12- NON FOOD RETAILING 163 0.0 112.5 19, .9 16, .9 0. 13 0.29** 0. 14 0.28** CLOTHING & SHOES 24 0.0 35.0 13, 8 8. .7 0. 13 0.15 0.21 0.20 FURNITURE & ACCESSORIES 11 0.0 60.0 26. .0 19, .3 0.06 0. 10 -0.02 0. 13 ART SUPPLIES 12 6.0 34.0 18. 5 10. .0 0.79 0.75** 0.82 0.76** COMPUTER PRODUCTS 15 5.0 45.0 21 . 3 11. ,5 0.55 0.44 + 0.49 0.48* VIDEO RENTAL 15 0.0 57.3 20. .3 15. . 1 -0.36 -O.OO -0.40 0. 18 13- FOOD RETAILING - NON CONVENIENCE 77 0.0 164 .5 18. .7 IB. .4 -0. 18 0.25* -0. 15 0.09 DONUT SHOPS 14 0.0 40.0 19 .6 11, . 1 0.48 0.34 0.34 0.31 ICE CREAM PARLORS 23 0.0 164.5 22 .4 31. .7 -0.46 -0. 10 -0.4B -0.35+ SPECIALTY FOOD SHOPS 12 7.2 30.0 13. 8 6. .9 0. 18 0.50+ 0.09 0. 17 14- MISCELLANEOUS 28 5.0 99.0 22. . 1 19. .0 -0.20 0. 10 -0. 13 -0.04 BEAUTY SALONS 19 5.8 99.0 24. ,6 20. .6 -0.49 -0. 16 -0.48 -0.51* ALL SECTORS 890 0.0 424.5 21. ,7 25 .9 -0.03 0.11** -0.03 0.10** oo CO Note: Franchise fees are In thousands of U.S. d o l l a r s . They Include the present value of a l l future payments s p e c i f i e d for the duration of the contract (using the average length of agreements In Table 4.8). The discount rate used is 10%. Averages were used when ranges were given. F' and r' are error terms from regressions of F and r on number of outlets and number of years in business. Two-sided t e s t s : ** : .01 l e v e l , * : .05 l e v e l . + : . 10 l e v e l . Table A.5 Royalty Rates and Franchise Fees Within Age and Size Cohorts for the 890 Franchisors Size cohorts % Franc. Royalty Royalty F . Fee F . Fee Cor(F,r) Spearman Rank Cor N Mean Mean St. Dev. Mean St. Dev. (F,r) 1- 1 < OUTLETS < 8 136 0.699 6.4 2.7 16.8 10.4 0.08 0.21* 2- 9 < OUTLETS < 15 95 0.755 6.5 2.9 21.8 31.3 -0.00 0. 11 3- 16 < OUTLETS < 24 95 0.814 6.3 3. 1 21.8 18.8 0. 18 0.25* 4- 25 < OUTLETS < 35 86 0.793 6.4 3.1 18. 1 12.0 0.06 0. 18+ 5- 36 < OUTLETS < 54 92 0.855 6.3 3.4 23.0 18. 1 -0. 16 0. 10 6- 55 < OUTLETS < 85 93 0.907 6.2 3.3 25.2 33. 1 -0. 17 0.20+ 7- 86 < OUTLETS < 151 79 0.854 6.2 3.4 21.4 23.7 0.05 0.07 8- 152 < OUTLETS < 299 84 0.866 7.0 3.5 24.4 46.8 -0.04 -0.22* 9- 300 < OUTLETS < 620 67 0.882 7.5 3.2 26. 1 25.9 0.01 0.02 10- 621 < OUTLETS < 9060 63 0.815 7. 1 4.5 22. 1 23.9 -0.24 0.02 Age cohorts 1- 1 < YEARS IN BUS. < 6 191 0. 809 6 .3 3 .0 21 .8 27. .7 -0. . 12 0. , 15* 2- 7 < YEARS IN BUS. < 9 151 0. ,809 6 .6 3 .3 22 .6 27 .2 -0. . 12 -0. .06 3- 10 < YEARS IN BUS. < 13 137 0. .813 6 .5 3, .2 23 .7 36 .9 0 .07 0. .21* 4- 14 < YEARS IN BUS. < 17 118 0. .873 6 .7 4. .2 21 . 0 23. .7 0. .01 0. , 14 5- 18 < YEARS IN BUS. < 27 142 0. 808 6 .7 3. .0 19. .7 13. .5 -0. .02 0. 03 6- 28 < YEARS IN BUS. < 170 151 0. 793 6. .6 3. .2 21 . 2 20. ,3 -0. 02 0. . 19* CO Note: The same lower and upper bounds as In Table 5 were used to generate cohorts here. For that reason, the number of franchisors per cohort isn't even close to being the same across cohorts. This was done in order to f a c i l i t a t e comparisons between the two tables. Two-sided t e s t s : ** sign, at the .01 l e v e l , * sign, at the .05 l e v e l , + sign, at the .10 l e v e l . A P P E N D I X B : C O R R E L A T I O N M A T R I X O F T H E V A R I A B L E S 170 Correlation Matrix of the Variables (Sample of 548 franchisors) X Franchisee! 1.0 Av. X Discontinued (1) 0.19 1.0 Adj. St.Dev.(Av.Sales) (1) 0.15 0.17 1.0 St.Dev.(Av.Sales) (1) 0.17 0.11 0.65 1 .0 Foreign Outlets ( X ) 0. 14 0.02 0.04 0 .03 1.0 Number of States 0.28 0.09 0.04 0 .04 0.24 1 .0 <Av.Salee-Inp.)/Av.Sales ( X ) -0.00 -0.01 -0.46 0 . 18 0.03 0 .07 1.0 Av. Sales / Outlet (1) -0 16 -0. 18 -0. 17 -0 .36 0.05 0 . 15 0.06 Franchisee Experience -0.10 0.07 -0.07 -0 .05 0.02 0 .09 0.14 X Time Not Franchising -0.37 0.07 0.05 0. .01 -0.14 -0 .30 -0.05 Weeks of training -0.32 -0.31 -0.34 -0. 37 -0.08 -0 .01 0. 12 Outlets In 1986 (lOO's) 0.04 -0.08 -0.O8 0. 08 0.24 0. 50 0.09 Years In Business -0.07 -0.08 -0. 10 -0. 18 0.06 0 27 -0.04 Growth 1n outlets 0.03 0.09 0.14 0. 18 -0.05 -0. 23 -0.01 Franchisor Financing 0. 17 0.21 0.00 0. 10 0.08 0. 13 0. 11 Capital Required <$K> -0.25 -0.17 -0.17 -0. 22 0.03 0. 06 0.08 Franchisor Inputs <tK) (1) -0.05 -0.10 0.52 -0. 18 -0.03 -0. 07 -0.91 Variable Fee ( X ) -0.07 0.01 -0. 12 -0. 10 -0.01 0. 02 -0.05 Fixed Fee ($K) -0.04 0.02 -0.04 0. 05 0.01 0. 04 0.05 X Franc. X Disc. Adj. St. Dev. Av. Sales St. Dev. Av. Sales X For-eign States Sales-Inputs/ Sales 0.10 1.0 -0.08 0.05 1.0 0.23 0.12 0.01 1.0 0.12 0.12 -0.18 0.08 1.0 i—i 0.20 0.00 0.21 0.20 0.26 1.0 r— i—i -0.14 0.01 0.24 -0.18 -0.14 -0.38 1.0 -0.12 0.02 -0.07 -0.12 0.05 0.01 0.02 1.0 0.42 0.25 0.02 0.33 0.11 0.14 -0.08 -0.17 1.0 0.11 -0.10 0.05 -0.04 -0.07 0.08 -0.02 -0.16 -0.00 1.0 -0.05 0.04 0.08 0.01 0. 14 -0 02 -0.06 0.09 -0.01 -0.05 1.0 -0.00 0.09 -0.01 0.10 -0.01 -0.01 0.06 0.10 0.16 -0.07 -0.04 Av. Sales F'ee Exp. X Time not Weeks Outlets of in Train- 1986 Years in Bus. Growth F'or F1n. Capi-tal needed F'or Var. Inputs Fee Franc. Ing A P P E N D I X C : H I S T O G R A M S F O R T H E T H R E E D E P E N D E N T V A R I A B L E S 172 Figure C . l Distribution of the percentage of franchised outlets n=548 25 - i 0 10 20 30 40 SO 00 70 80 90 100 % F r a n c h i s e d 173 Figure C .2 Distribution of royalty rates n=548 10 IS 20 Royal ty Rates 25 30 Distribution of Franchise Fees n=548 0 20 40 60 60 100 120 F r a n c h i s e Fees ( T h o u s a n d s of U.S. Dol lars ) 174 A P P E N D I X D : E S T I M A T I O N R E S U L T S E X C L U D I N G L I M I T O B S E R V A T I O N S 175 T a b l e D.1 : OLS R e g r e s s i o n s Without l i m i t O b s e r v a t i o n s -- P r o p o r t i o n of F r a n c h i s e d S t o r e s -- L i n e a r Av. % D i s c o n t i n u e d S t . Dev. of T h e t a S t . Dev. of Av. S a l e s F o r e i g n O u t l e t s (%) Number o f s t a t e s ( A v . S a l e s - I n p u t s ) / A v . S a l e s A v . S a l e s p e r o u t l e t ($100K) F r a n c h i s e e E x p e r i e n c e Weeks o f T r a i n i n g O u t l e t s 1n 198G (100's) % Time Not F r a n c h i s i n g Years i n B u s i n e s s Growth i n number of o u t l e t s C a p i t a l R e q u i r e d F r a n c h i s o r F i n a n c i n g F r a n c h i s o r I n p u t s ($000) V a r i a b l e Fee C o n s t a n t Number o f Obs. R-Square A d j . R-Square F- T e s t % F r a n c h . % Franch. % Franch. 0 . 18 0 .30 (0 . 13) (0 .23) 0 . 11 (0 .57) 15 .64 16 .41 15 .40 (1 47) (1 .54) (1 .45) 0 .37 ** 0 .37 ** 0 .37 ** (5 11) (5 . 10) (5 .02) 0, .07 0 . 14 0, .04 (0 .32) (0 64) (0 19) -0. .76 * -0 .77 * -0 ,71 + (-2, .03) (-2 .05) (-1 .87) -3. .05 -3, .09 -3. .02 (-1 .48) (-1 .50) (-1. 47) -1 . 61 ** -1. .61 ** -1. .56 ** (-4. .33) (-4. .32) (-4. 13) -0. .34 * -0. .37 ** -0. ,33 * (-2. .44) (-2. 71) (-2. 43) -0. .30 ** -0. .30 ** -0. ,30 ** (-7. .60) (-7. 87) (-7. 66) 0. . 19 * 0. .20 ** 0. , 19 * (2. .50) (2. 72) (2. 51) 10. .40 ** 11 . ,05 ** 10. , 15 ** (3. .40) (3. .66) (3. 29) -0. ,01 * -0. ,01 * -0. 01 * (-2. ,39) (-2. 41) (-2. 38) 0. 22 -O. 13 0. 29 (0.09) (-0.05) (0. 11) 0. 00 O. ,02 -0. 02 (0. 00) (0. 31) (-0. 27) -0. 37 -0. 37 (-1. 27) (-1. 26) 82. ,91 ** 73. ,04 ** 85. 40 ** (3. 40) (3. 16) (3. 81) 431 . 00 431 . 00 431 . 00 0. 34 0. 34 0. 34 0. 32 0. 31 0. 32 14. 22 15. 10 14 . 25 % Franch. % Franch. % Franch 0. 12 (0.62) -11 .59 -10 .46 (-0 .58) (-0 .52) 16. 16 15 .79 16 .58 (1.52) (1 .48) (1 .56) 0.37 ** 0 .38 ** 0 .38 (5.02) (5 . 18) (5 . 18) 0. 1 1 0 . 10 0 . 16 (0.48) (0 42) (0 72) -0.72 + -0 .83 * -0 .83 (-1.88) (-2 . 10) (-2 11) -3.05 -3, .09 -3 . 12 (-1.48) (-1 .50) (-1 51) -1.56 ** -1. .70 ** -1 , .69 (-4.13) (-4. .43) (-4 41) -0.37 ** -0. .35 * -0 38 (-2.71) (-2. .50) (-2 .80) -0.30 ** -0. .30 ** -0, .30 (-7.91) (-7. .62) (-7 .88) 0.20 ** 0. , 19 * 0 .20 (2.73) (2. ,49) (2, 71) 10.78 ** 10. ,63 ** 1 1 , 28 (3.54) (3. 45) (3 .70) -0.01 * -0. ,01 * -0, .01 (-2.41) (-2. 41) (-2. 43) -0.02 0. 21 -0. 1 1 (-0.01) (0.08) (-0. .04) -0.00 0. .00 0. .02 (-0.03) (0.06) (0. 34) -0. 39 (-1. .31) 76.60 ** 82. 93 ** 73. 83 (3.59) (3. 70) (3. 46) 431.00 431 . 00 431 . 00 0.34 0. 34 0. 34 0.32 0. 32 0. ,31 15. 13 14. 25 15. , 12 ** s i g n i f i c a n t a t the 0.01 l e v e l , * s i g n i f i c a n t a t the 0.05 l e v e l , + s i g n i f i c a n t a t the 0.10 l e v e l Table D.2 : OLS Regressions Without l i m i t Observations -- Proportion of Franchised Stores -- in Log % Franch. % Franch. % Franch. % Franch. % Franch. % Franch. Log Av. % Discontinued -0 .98 -0 .87 (-0 .22) (-0 . 19) Log Adj. St. Dev.(Av.Sales) 0 .08 0 . 18 (0 .05) (0 • 12) Log St. Dev.(Av.Sales) -1 .61 -1 .34 (-0 .84) (-0 .70) Foreign Outlets (%) 9 .50 10 . 13 9 .54 10 . 12 9 .93 10 .53 (0 .92) (0 .98) (0 .92) (0 .98) (0 .96) (1 .02) Log Number of States 4 .07 ** 4 .37 ** 4 .05 ** 4 .33 ** 4 .23 ** 4 .51 ** (3 .58) (3 .91) (3 .52) (3 .83) (3 .67) (3 .97) (Av.Sales-Inputs)/Av.Sales -0 .07 -0 .05 -0 .05 -0 .03 -0 .05 -0 .03 (-0 .43) (-0 .29) (-0 .33) (-0 . 17) (-0 .34) (-0 . 19) Av.Sales/ outlet ($100K) -0 .78 * -0 .74 * -0 .79 * -0 .75 * -0 .85 * -0 .80 * (-2 .25) (-2 . 15) (-2, .30) (-2 .20) (-2 42) (-2 .29) Franchisee Experience -3. .85 + -3 .86 + -3 .92 + -3 .92 + -3, .94 + -3 .93 + (-1. 87) (-1 .87) (-1 91) (-1 91) (-1 .93) (-1 .92) Weeks of Training -1 . 66 ** -1 .65 ** -1. .65 ** -1 .64 ** -1 . 74 ** -1 .72 ** (-4. 34) (-4 31) (-4. .29) (-4. .25) (-4. 41) (-4 .34) Log Outlets in 1986 0. . 12 -0, .20 0. . 15 -0, . 17 0. 07 -0, .26 (0. 12) (-0 21) (0. . 15) (-0. . 18) (0.07) (-0 27) % Time Not Franchising -0. ,29 ** -0, .30 ** -0. ,29 ** -0. ,30 ** -0. ,29 ** -0 ,30 ** (-7. 04) (-7, .23) (-7. 01) (-7. 21) (-6. 94) (-7, . 15) Log Years in Business 5. ,11 ** 5. .40 ** 5. ,08 ** 5. ,37 ** 4. ,95 ** 5, .27 ** (2. 85) (3 .02) (2. 84) (3. ,02) (2. 77) (2, 97) Growth in Outlets 1 1 . 81 ** 12. .47 ** 11 . ,79 ** 12. ,44 ** 11 . 91 ** 12. ,58 ** (3. 59) (3. 83) (3. 58) (3. 81) (3. 62) (3. 86) Log Capital Required -2. 20 * -2, ,31 * -2. , 16 * -2. 27 * -2. 23 * -2. ,35 * (-2. 38) (-2. 52) (-2. 36) (-2. 49) (-2. 46) (-2. 58) Franchisor Financing -1 . 00 -1 . 36 -1 . 00 -1 . 35 -1 . 14 -1 . 49 (-0. 39) (-0. 54) (-0. 39) (-0. 53) (-0. 45) (-0. 59) Log Franchisor Inputs ($K) -0. 68 -0. 50 -0. 54 -0. 37 -0. 59 -0. ,40 (-0. 51) (-0. 37) (-0. 46) (-0. 31) (-0. 49) (-0. 34) Variable Fee -0. 39 -0. 39 -0. 41 (-1. 35) (-1. 35) (-1. 43) Constant 93. 69 ** 87. 20 ** 90. 23 ** 83. 53 ** 87. 26 •* 81 . 38 ** (4. 17) (3. 97) (4. 58) (4. 37) (4. 81) (4. 60) Number of Obs. 431 . 00 431 . 00 431 . 00 431 . 00 431 . OO 431 . 00 R-Square 0. 35 0. 35 0. 35 0. 35 0. 35 0. 35 AdJ. R-Square 0. 33 0. 33 0. 33 0. 33 0. 33 0. 33 F-Test 15. 14 16. 06 15. 14 16. 06 15. 21 16. 11 ** s i g n i f i c a n t at the 0.01 l e v e l , * s i g n i f i c a n t at the 0.05 l e v e l , + s i g n i f i c a n t at the 0.10 level T a b l e D.3 : OLS R e g r e s s i o n s Without l i m i t O b s e r v a t i o n s -- R o y a l t y Rates and F r a n c h i s e Fees -- L i n e a r V a r i a b l e F i x e d V a r i a b l e F i x e d V a r i a b l e F i x e d Fee Fee Fee Fee Fee Fee Av. % D i s c o n t i n u e d -0 (-0 .08 .40) 0 (0 .43 .33) S t . Dev. of T h e t a -0 (-0 .00 .10) 0 (0 . 10 .56) S t . Dev. of Av. S a l e s -0 (-0 .49 .17) 39 (2 .47 * .03) F o r e i g n O u t l e t s (%) -1 .76 1 .71 -1 .75 1 .41 -1 , .75 1 .44 (-1 .41) (0 .19) (-1 .40) (0 .16) (-1 .40) (0 .16) Number of s t a t e s 0 .00 o .07 0 .OO 0 .07 0 .OO 0 .06 (0 .44) (1 .10) (0 .39) (1 .04) (0 .40) (0 .83) ( A v . S a l e s - I n p u t s ) / A v . S a l e s -0 . 19 ** -o . 13 -0 . 19 ** -0, . 17 -0 . 19 ** -0 .26 (-6 .39) (-0.61) (-6 25) (-0 .79) (-6 .14) (-1 21) A v . S a l e s per o u t l e t ($100K) -0 .03 -0 .47 -0 .04 -0 .42 -0 .04 -0, .20 (-0 .66) (-1 .28) (-0 67) (-1, 11) (-0, .68) (-0 .52) F r a n c h i s e e E x p e r i e n c e 0 .43 1 , .55 0 .42 1, .62 0, .41 1 .78 (1 .48) (0. 77) (1. .44) (0. .82) (1. .44) (0 .90) Weeks of T r a i n i n g -0 .02 0 .70 + -0 .02 o .73 + -0 .02 0 .92 * (-o 36) (1. 83) (-0. .30) (1. .88) (-0. 32) (2 .36) O u t l e t s i n 1986 (100's) 0. .08 ** -0. , 16 0. .08 ** -0 . 16 0. ,08 ** -0 . 14 (4. .33) (-1. .20) (4. .39) (-1 19) (4. 37) (-1 .04) % Time Not F r a n c h i s i n g 0. .01 ** -0. .02 0. .01 ** -0. 02 0 .01 ** -0, .02 (2. 72) (-0. .65) (2. .69) (-0. 65) (2. ,70) (-0. .66) Ye a r s i n B u s i n e s s -0. 03 * 0. .00 -0. ,03 * 0. .00 -0. 03 * 0. .01 (-2. .50) (0. .01) (-2. .49) (0. 03) (-2. ,49) (0.07) Growth i n number of o u t l e t s -1 . 37 ** 4 . 97 -1 . ,37 ** 4. 80 -1 . ,37 ** 4. ,30 (-3. 04) (1. 64) (-3. 02) (1. 58) (-3. 02) (1. 42) C a p i t a l R e q u i r e d -0. 00 0. 02 ** -0. 00 0. 02 ** -0. 00 0. ,02 ** (-0. 49) (3. 68) (-0. 47) (3. 68) (-0. 47) (3. 69) F r a n c h i s o r F i n a n c i n g 0. 52 5. 68 * 0. 51 5. 74 ** 0. 51 5. 79 ** (1. 61) (2. 55) (1. 57) (2. 60) (1. 57) (2. 63) F r a n c h i s o r I n p u t s ($000) -0. 06 ** -0. 04 -0. 06 ** -0. 06 -0. 06 ** -0. 07 (-6. 69) (-0. 69) (-5. 98) (-0. 95) (-6. 60) (-1. 14) C o n s t a n t 26. 89 ** 28. 36 26. 38 ** 32. 58 26. ,35 ** 37. 07 + (8. 57) (1. 29) (8. 97) (1. 58) (8. 95) (1. 80) Number o f Obs. R-Square A d j . R-Square F - T e s t 511.00 541.00 0.17 0.06 0.15 0.04 7.19 2.47 511.00 541.00 0.17 0.06 0.14 0.04 7.17 2.49 511.00 541.00 0.17 0.07 0.14 0.04 7.17 2.78 ** s i g n i f i c a n t a t the 0.01 l e v e l , * s i g n i f i c a n t at the 0.05 l e v e l , + s i g n i f i c a n t a t the 0.10 l e v e l T a b l e D.4 : OLS R e g r e s s i o n s Without l i m i t O b s e r v a t i o n s -- R o y a l t y Rates and F r a n c h i s e Fees -- In Log Log Av. % D i s c o n t i n u e d Log A d j . S t . D e v . ( A v . S a l e s ) Log S t . D e v . ( A v . S a l e s ) F o r e i g n O u t l e t s (%) Log Number of S t a t e s ( A v . S a l e s - I n p u t s ) / A v . S a l e s A v . S a l e s / o u t l e t ($K) F r a n c h i s e e E x p e r i e n c e Weeks o f T r a i n i n g Log O u t l e t s In 1986 % Time Not F r a n c h i s i n g Log Years i n B u s i n e s s Growth i n O u t l e t s Log C a p i t a l R e q u i r e d F r a n c h i s o r F i n a n c i n g Log F r a n c h i s o r I n p u t s ($K) C o n s t a n t Number o f Obs. R-Square A d j . R-Square F - T e s t V a r i a b l e F i x e d V a r i a b l e F i x e d V a r i a b l e F i x e d Fee Fee Fee Fee Fee Fee -0 .08 1 .93 (-0 .12) (0.44) -0 .28 1 .82 (-1 .27) (1 .26) -0 .55 * 4, .23 * (-2 .01) (2 .30) -2 .08 + 1 .53 -1 .92 0 .54 -1 , .89 0, .36 (-1 .65) (0 . 18) (-1 .52) (0 .06) (-1 .50) (0 .04) -0 .76 ** 0 .65 -0 .73 ** 0 .47 -0 .70 ** 0 .25 (-4 .84) (0 .62) (-4 .63) (0 .45) (-4 .45) (0, .23) -0 .07 ** -0 .20 -0 .07 ** -0 . 18 -0, .06 ** -0, .26 + (-2 .77) (-1 .19) (-3 .33) (-1 15) (-2, .90) (-1 .78) -0 . 13 * -0 .52 -0 . 13 * -0. .50 -0, . 15 ** -0, .33 (-2 .54) (-1 .53) (-2 .58) (-1. 47) (-2. .94) (-0 97) 0 .41 0 .50 0 .42 0 .55 0. .41 0. .66 (1 .40) (0 .26) (1 .46) (0. ,28) (1 42) (0 .34) 0 .00 0 .48 -0 .01 0 .55 -0 .03 0 .72 + (0 01) (1 25) (-0. . 19) (1. 41) (-0. .56) (1 81) 0. .95 ** -0, .26 0. .94 ** -0. .23 0. .92 ** -0. .09 (6. .95) (-0. .29) (6. .91) (-o. ,26) (6. 74) (-0. . 10) 0. .01 * -0. .02 0. ,02 * -0. .02 0. .02 ** -0. 03 (2. ,44) (-0. .50) (2. ,56) (-0. 61) (2. 61) (-0. 67) -0. .57 * -0. .37 -0. ,61 * -0. 06 -0. 62 * 0. 04 (-2. .16) (-0. 21) (-2. .31) (-0. 03) (-2. ,36) (0. .02) -1 . .52 ** 5. .63 + -1 . .50 ** 5. 52 + -1 . ,49 ** 5. .34 + (-3 04) (1. 74) (-3. 00) (1. 71) (-2. .99) (1. 66) -0. , 13 5. .07 ** -0. . 14 5. , 13 ** -0. 15 5. , 17 ** (-1. 02) (6. .04) (-1. 15) (6. 20) (-1. 18) (6. 30) 0. 37 6. .88 ** 0. 36 6. 95 ** 0. 35 7. ,01 ** (1. 09) (3.07) (1. 06) (3. 11) (1. 05) (3. 14) -0. 62 ** -2. .41 + -0. 61 ** -2. 66 * -0. 61 ** -2. 65 • (-3. 15) (-1. 83) (-3. 58) (-2. 32) (-3. 57) (-2. 32) 19. 21 ** 22. .43 20. 43 ** 18. 53 17. 46 ** 39. 32 * (6. 01) (1. 04) (7. 50) (1. 01) (6. 79) (2. 28) 511. 00 541 . 00 511. OO 541 . 00 511 . 00 541 . OO 0. 15 0. 10 0. 15 0. 10 0. 16 0. , 1 1 0. 13 0. 08 0. 13 0. 08 0. 13 0. 09 6. 25 4. 30 6. 38 4. 41 6. 58 4. 70 ** s i g n i f i c a n t a t the 0.01 l e v e l . * s i g n i f i c a n t a t the 0.05 l e v e l , + s i g n i f i c a n t a t the 0.10 l e v e l A P P E N D I X E : E S T I M A T I O N R E S U L T S F O R T H E N U M B E R O F F R A N C H I S E S A N D C O M P A N Y - O P E R A T E D S T O R E S 180 T a b l e E.1 : OLS R e g r e s s i o n s f o r the Number of F r a n c h i s e s F r a n c h i s e s F r a n c h i s e s F r a n c h i s e s F r a n c h i s e s F r a n c h i s e s F r a n c h i s e s i n 1986 i n 1986 i n 1986 i n 1986 i n 1986 i n 1986 Av. % D i s c o n t i n u e d 10 .26 10 .48 (1 • 51) (1 .53) S t . Dev. of T h e t a 1 .54 1 .58 (1 .54) (1 .51) S t . Dev. of Av. S a l e s 47 .66 51 .62 (0 .54) (0 .57) O u t l e t s o u t s i d e home (%) 1 .27 + 1 .28 + 1 .22 + 1 .23 + 1 .25 + 1 .27 (1 .94) (1 .95) (1 .89) (1 .90) (1 .95) (1 .96) Number of s t a t e s 0 .65 0 .65 0 .62 0 .62 0 .70 0 .71 (1 .16) (1 .17) (1 .10) (1 .11) (1 .30) (1 .31) ( A v . S a l e s - I n p u t s )/Av.Sales 0 .24 0 .40 -0 .48 -0 .35 -0 .24 -0 .09 (0 .18) (0 .35) (-0 .33) (-0 .28) (-0 .17) (-0 .07) A v . S a l e s p e r o u t l e t ($100K) -2 .31 -2 .35 -1 .54 -1 .55 -2 .01 -2 .03 (-1 .43) (-1 .45) (-0 .87) (-0 .88) (-1 .15) (-1 . 16) F r a n c h i s e e E x p e r i e n c e -19 .82 -20 .05 -17 .93 -18 . 1 1 -17 .55 -17 .74 (-1 .52) (-1 .53) (-1 .39) (-1 .39) (-1 .38) (-1 .39) Weeks of T r a i n i n g -7 .49 * -7 .51 * -7 .20 + -7 .20 + -7 .89 * -7 .90 ("2 .01) (-2 .02) (-1 .83) (-1 .84) (-1 .97) (-1 .98) O u t l e t s i n 1986 (100's) 82 .78 ** 82 .71 ** 82 .74 ** 82 .67 ** 82 .70 ** 82 .62 (21 .21) (21 .32) (21 .33) (21 .44) (21 .28) (21 .40) % Time Not F r a n c h i s i n g -0 .55 * -0 .57 * -0 .54 * -0 .56 * -o .52 * -0 .54 (-2 48) (-2 .44) (-2 .40) (-2 .35) (-2 .31) (-2 .29) Y e a r s i n B u s i n e s s 0 . 15 0 . 18 0 . 17 0 .20 0 . 13 0 . 16 (0 .24) (0 .28) (0 26) (0 .30) (0 .20) (0 .25) Growth i n number of o u t l e t s 4 . 14 5 .39 1 .73 2 .79 3 .33 4 .64 (0 .32) (0 .40) (0 13) (0 .20) (0 .25) (0 .34) C a p i t a l R e q u i r e d -0 .04 -0 .04 -0 .04 -0 .04 -0 .05 -0 .05 (-1 .09) (-1 .09) (-1 11) (-1 11) (-1 .13) (-1 .13) F r a n c h i s o r F i n a n c i n g 10. .08 9 .47 11 .82 11 .29 12 .54 11 .93 (0 .70) (0 71) (0 81) (0 .82) (0 .85) (0 .85) F r a n c h i s o r I n p u t s ($0O0) -o. OO 0 .04 -0 .35 -0 .32 -0 . 13 -0 .09 (-0 .01) (0 . 14) (-0 87) (-0 87) (-0 .39) (-0 .29) V a r i a b l e Fee -0 .80 -0. .72 -0 .88 (-0. .38) (-0 .35) (-0 .41) C o n s t a n t -3 .73 -25 . 15 82 .76 65 . 19 70 .00 48 .60 (-0. .03) (-0 .26) (0 .63) (0. .60) (0 .54) (0 .45) Number of Obs. 548. OO 548 .00 548. .00 548. .00 548. .00 548 .00 R-Square 0. ,97 0 .97 0. 97 0. 97 0 .97 0. .97 AdJ. R-Square 0. ,97 0. .97 0. ,97 0. 97 0. .97 0 .97 F - T e s t 1046. . 12 1 122 .43 1046. .71 1 123. 16 1043. .05 1119 .03 ** s i g n i f i c a n t a t the 0.01 l e v e l , * s i g n i f i c a n t at the 0.05 l e v e l , + s i g n i f i c a n t a t the 0.10 l e v e l Note: C o r r e c t e d f o r h e t e r o s c e d a s t i c i t y u s i n g White's (1980) h e t e r o s c e d a s t i c - c o n s i s t e n t c o v a r i a n c e m a t r i x . Table E.2 : TOBIT Regressions for the Number of Company-Owned Outlets Co-owned Co-owned Co-owned Co-owned Co-owned Co-owned 1n 1986 in 1986 in 1986 in 1986 1n 1986 in 1986 Av. % Discontinued -0 .57 " -0. 60 ** (-2 .70) (-2. 79) St. Dev. of Theta -0 .06 ' -0 .06 • (-2 .08) (-2 . 10) St. Dev. of Av. Sales -5, .27 •• -4 .90 (-6 .27) (-3 .43) Outlets outside home (%) -0 .46 -0. 49 -0 .51 -0 .54 -0 .07 -0 .06 (-0 . 14) (-0. 16) (-0 . 16) (-0 .17) (-0 .02) (-0 .02) Number of states -0 .22 *• -0. 22 •• -0 .21 *• -0 .21 '* -0 .22 ** -0 .22 (-3 .93) (-4. 15) (-3 .79) (-3 .88) (-3 .88) (-3 .30) (Av.Sales-Inputs)/Av.Sales -.0 .00 -0. 01 0 .04 ** 0 .03 ** 0 .04 ** 0 .03 ( -o . 16) (-0. 52) (3 .75) (2 .58) (4 .83) (2 .53) Av.Sales per outlet ($100K) -0 .05 -0. 05 -0 .00 0 .01 -0 .02 -0 .01 (-0 .42) (-0. 44) (-0. .04) (-0 .06) (-0 .19) (-0 .05) Franchisee Experience 0 .27 0. 30 0 . 14 0 .22 0, . 15 0 .23 (0 .72) (0. 79) (0, .37) (0 .59) (0, .39) (0 .63) Weeks of Training 0 .02 " 0. 02 ** 0 .02 • 0 .02 • 0. .02 •• 0 .02 (3 . 14) (3. 04) (2 .51) (2 .33) (2. .64) (2 .36) Outlets in 1986 (100's) 0 . 19 * 0. 19 • 0 . 18 * 0 . 18 • 0 . 19 • 0 . 19 (2 .09) (2. 03) (1 .99) (1 .88) (2, .11) (1 .99) % Time Not Franchising 13 .34 •• 13. 32 •* 13 . 14 •• 13 .09 '* 13, .31 •• 13 .27 ( 13 .53) ( 16. 93) (11 .75) (14 .35) (14, . 18) (6 .04) Years in Business -0 .04 * -0. 04 • -0 .04 • -0 .04 • -0, .05 " -0 .04 (-2 .33) (-2. 26) (-2 .57) (-2 .39) (-2, .66) (-2 .32) Growth in number of outlets 0 .04 0. 06 0, .04 0 .07 0, .02 0 .03 (0 .08) (0. 12) (0 .09) (0 . 12) (0. .04) (0 .05) Capital Required -1 .36 •• -1. 35 •*• -1, . 56 •• -1 .55 •• -1. ,54 •• -1 .52 (-3 . 15) (-3. 16) (-3, .52) (-3 .43) (-3. .53) (-3 .37) Franchisor Financing 0 .00 0. 00 0, .00 0 .00 0. ,00 0 .00 (1 . 12) (1. 14) ( 1, .08) (1 .11) (1. .16) (1 .24) Franchisor Inputs ($000) -0 .01 -0. 01 0. .01 + 0 .01 0, .01 0 .00 (-1 .04) (-1. 41) (1. ,88) (1 .47) (1. .28) (0 .63) Variable Fee 0 .02 0, .05 0. .06 (0 .41) (1. . 18) ( 1. 31) Constant 2, .68 • 3. 27 " -2. .85 •* -1, .67 • -3. ,26 •* -1 .81 (2. .55) (3. 32) (-3. ,73) (-1, .76) (-4. 46) (-1 .72) Standard Error of Estm.: Alpha 0. , 10 0. 10 0. 08 0. ,07 0. 10 0, .11 (0. 81) (0. 72) (0. 55) (0. ,47) (0. 70) (0, ,62) Beta 22. , 70 *' 22. 72 22. ,84 *• 22. ,89 " 22. 80 *" 22. ,80 (31. 38) (28. 74) (25. 87) (25. ,29) (30. 57) ( 19. 00) Limit Observations 117. .00 117 .00 117. 00 117. ,00 117. 00 117 .00 Non-Limit Observations 431. ,00 431 .00 431. ,00 431. 00 431. ,00 431. .00 Log Likelihood Function -1865. ,21 -1865 .28 -1866. 12 -1866. 77 -1B67. 19 1867. .98 " significant at the 0.01 level, * significant at the 0.05 level, • significant at the 0.10 level Note: Corrected for heteroscedasticity using the method suggested by Maddala (1983), p. 178-182 A P P E N D I X F : T H E L O N G T E R M T R E N D IN F R A N C H I S I N G 183 Figure F . l The Long Term Trend in Franchising 184 A P P E N D I X G : R E S U L T S F O R G R O U P S O F E I G H T C O H O R T S 185 Table G.1 : The Proportion of Franchisee! Stores within Eight Age Cohorts Number of Years in Business: 1 -• 5 6 • • 8 9 - 10 % Franch. % Franch. % Franch. Log. Av. % Discontinued 28. ,90 * 7. .21 20. ,29 (2. .30) (0, .59) (1, ,15) Foreign Outlets (%) 148. 31 + 8 .04 74. ,46 ( 1 . 98) (0, .46) (1, .32) Log. Number of States 1 . 08 1, .36 1. .92 (0. .35) (0. 49) (0. .44) (Av.Sales-Inputs)/Av.Sales 0. 37 -0 . 76 -0. .56 (0. .46) (-1. .53) ( -1. .15) Av.Sales/ outlet ($100K) -1 . 26 2. .11 0. .62 (-0. 45) (1. .23) (0. ,27) Franchisee Experience 3. ,21 -1 . 34 2. 06 (0. .49) (-0. . 28) (0. ,29) Weeks of Training -3. . 16 -2 .37 + -2 .26 (-1. .25) (-1 , .67) ( -1. ,51) Log. Outlets in 1986 (100's) 4, .76 3 .85 -1 . 28 ( 1 . ,56) ( 1 .63) ( -0. 29) % Time Not Franchising -0. . 17 -0, .41 *' -0. .44 * (-1. ,00) (-3 .26) ( -3. .38) Log. Years in Business 26 .22 + 4, .42 95 . 19 (1. .79) (0 .22) (1. .59) Growth in Outlets 7 . 24 11 , .48 13, .93 (1. 26) (1, .54) ( 1 . .59) Log. Capital Required 0. .39 - 1 , .47 -0. .37 (0. , 17) (-0 .81) ( -0. . 12) Franchisor Financing 13. .30 + -2 .95 17. .47 + (1. .88) (-0 .46) (1. .81) Log. Franchisor Inputs ($K) 7 .90 -10 .07 * -1, .81 (1. , 14) (-2 .56) ( -0. .46) Constant -30. .87 178 .49 * - 85 .80 (-0. , 30) (2 .63) ( -0. 57) 11 - 13 14 - 16 17 - 22 23 - 32 32 + % Franch. % Franch. % Franch. % Franch. % Franch. -8. ,67 30. .93 * -21 . 93 - 17. , 38 -27 . ,85 (-0. .94) (2 .35) (-1 . 16) (-1. ,18) (-1. .64) 42. ,22 34 .50 2. ,79 16. ,88 -20. .17 (0. .62) (1 .43) (0. .08) (0. ,65) (-0. .59) 5. ,94 • -2, .54 9. ,04 * 7. , 12 * 7. .56 ' (2. .04) (-1 .13) (2. .25) (2. ,21) (2. .23) 0. 32 0 .09 -0. .56 -0. . 33 -0. .07 (0. .92) (0 .31) (-0 .74) (-0. .73) (-0. .14) -0 .75 -0 .94 0 .05 -0. ,08 - 1 . .48 + (-0. ,75) (-1. .45) (0. .05) (-0. 09) (-1. .86) -3. . 10 - 1 , .50 - 16. 31 * -3. , 35 2 .40 (-0. ,58) (-0 .36) (-2. .33) (-0. ,68) (0. 43) -0. 39 -1 .17 0 . 10 - 1 . .58 * -3 .09 * (-0. ,43) (-1 .09) (0. .09) (-2. ,02) (-3. .04) 3. .29 2 .31 -6 .53 * -7. .11 '* -3 . 38 ( 1 . ,28) ( 1 .09) (-2. .05) ( -2. 71) (-1. . 19) -0. , 19 + -0, . 18 -0. .48 " -0. .41 ** -0. .56 ' (-1. 73) (-1 .67) ( -3. 35) (-3. .45) (-5. . 16) -33 26 -35 .94 -25 . 15 -6. .90 25. .65 * (-1. ,00) (-0 .95) (-0. .90) (-0, .28) (2 .43) 5. .90 -19 . 15 37 .27 * -2 .11 49 .37 * (0. ,61) (-1 .47) (2 .66) (-0. . 12) (2 .88) -4. . 16 -2 .60 -6. .68 * -7 . 15 ** -4, .96 (-1. .51) (-1 .07) (-2. .35) (-2 .73) (-1 , .49) 8 .33 13 .55 * 1 .27 1 .67 -4 .26 (1 .19) (2 .21) (0 .20) (0 .28) (-0 .56) 1 . 15 -2 .39 -9 .36 0 .21 -4 .62 (0. .43) (-0 .93) (-1 .62) (0 .05) (-0 .98) 158, .81 + 179 .54 276 .45 • 186 .53 * 80 .98 ( 1 . 71) ( 1 .64) (2 . 15) (2 .03) ( 1 .02) Limit Observations 13 Non-Limit Observations 41 Standard error of Estm. 15 At Mean X(I) , E(Y) 85 Log Likelihood Function -179 Squared CorCY, E(Y)) 0 Likelihood r a t i o tests 00 22.00 15.00 13.00 00 70.00 37 .00 59.00 20 18.57 16.36 15.83 81 84.40 87.46 82.54 53 -320.67 -164.15 -258.04 47 0.52 0.51 0.63 / \ / 19.32 22.64 + 19, .00 18. .00 7. .00 10, .00 46 .00 58 .00 62 .00 58 .00 13 .40 20, .31 16. .43 18 .00 89 .79 84 .41 79 .59 79 .56 - 196 .08 -268 .21 -266 .72 -257 .40 0 .52 0 .53 0 .49 0 .74 \ / \ / 30.80 ** 19.32 ** s i g n i f i c a n t at the 0.01 l e v e l , * s i g n i f i c a n t at the 0.05 level, + s i g n i f i c a n t at the 0.10 level Table G.2 : The Royalty Rate within Eight Age Cohorts Number of Years in Business: 1 - 5 6 8 9 - 10 11 - 13 14 - 16 17 - 22 23 - 32 32 + Var iable Variable Var iable Var iable Var iable Variable Var iable Var iable Fee Fee Fee Fee Fee Fee Fee Fee Av. % Discontinued 0 19 0 11 -1 00 -0 51 -0 34 -0 82 0 50 -0 51 (0 31) (0 18) (-1 49) (-1 07) (-0 32) (-1 38) (0 85) (-0 67) Foreign Outlets (%) -4 40 -1 62 4 93 * -13 88 -14 13 * -2 61 -2 98 -3 58 (-0 35) (-0 54) (2 24) (-1 17) (-2 18) (-0 89) (-0 81) (-0 56) Number of States -0 11 + -0 03 0 03 -0 06 0 06 -0 03 0 01 0 04 (-2 02) (-0 81) (0 73) (-1 53) (1 35) (-0 84) ' (0 52) ( 1 28) (Av.Sales-Inputs)/Av.Sales 0 04 -0 09 -0 27 *• -0 31 ** -0 18 -0 29 + 0 13 -0 22 * (0 27) (-0 85) (-3 83) (-3 37) (-1 53) (-2 00) (-1 50) (-2 23) Av.Sales/ outlet ($100K) 0 71 0 04 0 15 -0 46 -0 05 0 22 + 0 14 0 01 (1 44) (0 12) (0 51) (-1 52) (-0 23) (1 79) (0 92) (0 10) Franchisee Experience -0 82 1 25 -0 66 1 29 1 29 -1 78 + -0 71 1 14 (-0 71) ( 1 59) (-0 82) (1 46) ( 1 01) (-1 84) ( -0 86) ( 1 19) Weeks of Training -0 88 + 0 35 0 21 0 03 0 09 -0 11 -0 04 -0 04 (-1 85) (1 52) (1 46) (0 22) (0 27) (-0 62) (-0 29) (-0 26) Outlets in 1986 (100's) 1 22 * 0 11 -0 68 + 0 52 ' 0 07 0 28 * 0 13 *• 0 01 (2 08) (0 34) (-1 99) (2 27) (0 89) (2 64) (3 62) (0 35) % Time Not Franchising -0 05 + -0 01 0 04 ** 0 04 * 0 05 0 01 0 04 + 0 04 * (-1 77) (-0 51) (2 98) (2 18) (1 54) (0 74) ( 1 99) (2 65) Years in Business 0 73 0 45 0 75 -0 63 0 70 0 27 0 12 -0 14 * (1 08) (0 93) (1 17) (-1 31) (0 94) (1 41) (-0 81) (-4 20) Growth in Outlets -0 12 -0 83 -2 35 * -3 93 * 3 74 2 30 -1 99 -2 33 (-0 11) (-0 67) (-2 40) (-2 17) (0 95) (1 34) (-0 71) (-0 85) Capital Required 0 01 + -0 00 -0 00 0 00 + 0 00 0 00 0 00 0 00 (1 96) (-0 49) (-0 57) (1 75) (0 11) (0 45) (0 40) (0 67) Franchisor Financing 2 15 + -0 55 -0 02 2 16 + 2 28 -0 19 0 31 1 88 + (1 84) (-0 55) (-0 02) (1 94) (1 40) ( -0 23) (0 35) ( 1 71 ) Franchisor Inputs ($K) -0 01 -0 02 -0 08 ** -0 07 ' -0 05 -0 06 + -0 03 -0 03 (-0 32) (-0 62) (-3 61) (-2 49) (-1 44) (-1 74) (-1 17) (-1 12) Constant -1 43 11 95 27 85 "* 45 39 •• 11 55 30 81 * 19 59 * 31 24 ' (-0 09) (1 07) (3 03) (4 28) (0 69) (2 09) (2 11) (2 86) Limit Observations 5 00 7 00 3 00 4 00 6 00 5 00 1 00 6 00 Non-Limit Observations 49 00 85 00 49 00 68 00 59 00 71 00 68 00 62 00 Standard error of Estm. 2 90 3 24 2 00 2 83 4 37 2 77 2 77 3 05 At Mean X(I) , E(Y) 5 76 6 54 6 30 6 41 7 11 6 06 7 10 6 34 Log Likelihood Function -127 67 -230 16 -107 20 -172 62 -177 85 - 178 89 -167 27 - 164 24 Squared Cor(Y, E(Y)) 0 20 0 10 0 52 0 36 0 28 0 20 0 29 0 36 \ / \ / \ / \ / Likelihood r a t i o tests 13 14 23 74 + 32 60 " 15 16 ** s i g n i f i c a n t at the 0.01 l e v e l , * s i g n i f i c a n t at the 0.05 level, + s i g n i f i c a n t at the 0.10 level Table G. 3 The Franchise Fee within Eight Age Cohorts Number of Years in Business: 1 to 5 6 to 8 F i xed Fixed Fee Fee Log. Av. % Discontinued 1 . 08 -23 .24 (0. 09) (-2, .40) Foreign Outlets (%) -32. . 79 9. .07 (-0. .48) (0 .62) Log. Number of States 7. 03 * 2. .53 (2. .26) (1 . 11) (Av.Sales-Inputs)/Av.Sales 0. 35 -0. .84 (0. ,44) (-2. 06) Av.Sales/ outlet ($100K) 0. 49 -1. 98 (0. 17) (-1 . 48) Franchisee Experience - 1 . 01 -0. 16 (-0. 16) (-0. .04) Weeks of Training -0. 86 1 . 99 (-0. 36) (1. 76) Log. Outlets in 1986 (100's) -3. 15 1 . 15 (-1. 09) (0. 59) % Time Not Franchising -0. 10 -0. 02 ( -0. 58) (-0. 24) Log. Years in Business 19. 47 -3. 73 (1. 32) (-0. 23) Growth in Outlets 8. 72 3. 12 ( 1 . 50) (0. 51) Log. Capital Required 9. 16 *" 3. 16 (3. 83) (2. 21) Franchisor Financing 18. 05 * -5. 42 (2. 60) (-1. 07) Log. Franchisor Inputs ($K) -2. 13 -6. 93 (-0. 32) (-2. 19) Constant -86. 54 138. 33 ( -0. 88) (2. 55) Limit Observations 1 . 00 0. 00 Non-Limit Observations 53. 00 92. 00 Standard error of Estm. 15. 90 15. 67 At Mean X(I) , E(Y) 21 . 32 21 . 85 Log Likelihood Function -223. 16 375. 53 Squared Cor(Y, E(Y)) 0. 56 0. 26 \ / Likelihood r a t i o tests 29.58 • 9 to 10 11 to 13 14 to 16 17 to 22 23 to 32 32 + Fixed F i xed Fixed F ixed Fixed F i xed Fee Fee Fee Fee Fee Fee 38. 33 -8 .73 -15 .61 6 .07 11 .30 6 .42 (1.12) (-1 .14) (-1 .31) (0 .87) (0 .95) (0 .32) 15. 76 184 . 16 ' * 5 .00 -7 .03 - 17 .09 -21 .89 (0.40) (3 .44) (0 .27) (-0 .62) ( -0 .81) (-0 . 53) 0.94 -2 .96 -2 .53 0 .94 0 .72 -7 . 14 + (0.12) (-1 .28) (-1 .26) (0 .55) (0 .28) ( - 1 .77) 0.14 -0 . 14 -0 .56 + -0 .24 -0 . 19 -0 .01 (0.15) (-0 .47) (-1 .99) (-0 .82) (-0 .50) (-0 .01) -2.61 0 .85 -0 . 32 0 .27 0 . 32 -0 . 75 (-0.60) ( 1 .00) (-0 .53) (0 .67) (0 .43) ( -0 .79) 11.50 -3 .42 2 .43 -5. .01 + 5 .07 -9 .88 (0.84) (-0 .81) (0 .63) (-1 .69) ( 1 .27) ( - 1 .46) -1.15 1 .85 • 1 . 15 0 . 25 1 .40 * 0 .79 (-0.38) (2 .44) (1 . 12) (0. .54) (2. .20) (0 .67) 0. 18 3 .54 + 1 .20 -0. ,02 -0, .55 1 .90 (0.02) ( 1 .71) (0 .68) (-0 .02) (-0, .26) (0 .57) 0.07 -0 .23 * -0, . 17 + 0. . 14 * -0, .00 -0. .22 + (0.31) ( -2 .61) (-1, .76) (2. .31) ( -0. 00) (-1, .93) -74.43 -0. .64 -6, .93 15. ,85 14. .98 -11, .23 (-0.66) (-0 .02) (-0 .21) (1, .28) (0, .73) (-0, .88) -2.07 11 .68 28. .54 * -2. , 11 -1 , .90 23, . 12 (-0.12) ( 1. 46) (2. .40) (-0. .37) (-0. ,13) (1, . 14) 10.83 + 1 . ,35 2. .29 4. 10 • * 2. 98 10. .52 * (1.79) (0. .60) ( 1 . , 15) (3. 31) (1. ,47) (2. . 55) 0.25 -5. .79 -3. 32 5. 76 * 13. , 52 ** 24. .72 * (0.02) (-1. .02) (-0. .62) (2. 05) (2. ,81) (2. .63) - 1 . 36 -3. .86 + -6. , 54 ** -2. 27 0. ,31 -0 . 93 (-0. 18) (-1. .72) (-2. .70) (-1 . 04) (0. 10) (-0. 17) 104.71 55. ,20 120. 86 -29. 63 -50. 07 42. 89 (0.37) (0. 74) ( 1 . .25) (-0. 59) (-0. 67) (0. 45) 1 .00 0. 00 0. 00 3. 00 0. ,00 2. 00 51 .00 72. 00 65. 00 73. 00 69. 00 66. 00 34.82 13. 41 13. 13 9. 28 13. 75 22. ,15 28. 14 22. 34 20. 48 18. 92 21 . 41 24. 28 253.90 -280. 66 -251 . 06 -269. 96 -270. 29 -299. 68 0.22 0. 37 0. 37 0. 29 0. 29 0. 29 \ / \ / \ / 83. 02 " 24 .94 + 40. 86 " s i g n i f i c a n t at the 0.01 le v e l , * s i g n i f i c a n t at the 0.05 le v e l , + s i g n i f i c a n t at the 0.10 level Table G.4 : The Proportion of Franchised Stores within Eight Size Cohorts Number of Outlets: Log. Av. % Discontinued Foreign Outlets (%) Log. Number of States (Av.Sales-Inputs)/Av.Sales Av.Sales/ outlet ($100K) Franchisee Experience Weeks of Training Log. Outlets in 1986 (100's) % Time Not Franchising Log. Years in Business Growth in Outlets Log. Capital Required Franchisor Financing Log. Franchisor Inputs ($K) Constant Limit Observations Non-Limit Observations Standard error of Estm. At Mean X(I), E(Y) Log Likelihood Function Squared Cor(Y, E(Y)) Likelihood r a t i o tests 2 - 10 11 • 20 21 • 32 33 - 53 54 - 91 92 - 196 197 - 490 491 + % Franch. % Franch. % Franch. % Franch. % Franch. % Franch. % Franch. % Franch. 49.76 *• -8. 89 -8. .45 19. 67 -4. 46 1. 27 -0. 11 33.98 * (3.47) (-0. ,51) (-0. .54) (1. 61) (-0. 38) (0. 10) (-0. 02) (2.70) 37 .82 -37. .73 -38. .20 -6. 40 36. 90 11. 42 34. 49 46.50 * (0.80) (-1. 00) (-0. 86) (-0. 09) (1. 38) (0. 74) (1 . 44) (2.04) -2.33 3. .35 8. . 15 * 2. 23 1. 12 5. 32 + 7. 70 + 2.63 (-0.61) (0. 80) (2. .52) (0. 85) (0. 47) ( 1. 96) ( 1. 92) (0.69) 0.84 - 1 . 51 * -0. .34 1. 37 * -0. 45 0. 43 -0. 09 -0. 50 ( 1 .57) (-2. , 19) (-0. .57) (2. 50) (-1. 34) (0. 99) (-0. 26) (-1.21) -0.78 5. 44 + 3 .00 -2. 94 + -0. 95 -3. 31 * -1 . 06 * - 1 .09 (-0.51) ( 1 . 74) ( 1 . .22) (-1. 71) (-1. 28) ( -2. 15) ( -2. 14) (-1.56) 0.56 6. 07 -8. ,50 -7. 97 2. 92 -5. 19 0. 76 -1 .94 (0.10) (0. 82) (-1. .29) (-1. 49) (0. 64) (-1. 07) (0. 18) (-0.38) -2.35 + -4. 81 •* -1 . ,64 -6. 43 ** 0. 56 -0. 35 - 1 . 66 + -0.99 (-1.74) (-2. .93) (-1. .17) (-4. 97) (0. 68) (-0. 34) (-1. 76) (-1.31) 8.92 - 1 . 66 -6. . 19 37 . 77 * 6. 53 6. 24 8. 96 -2. 37 (1.60) (-0. 11) (-0. .31) (2. 44) (0. 50) (0. 66) ( 1 . 14) (-0.71) -0.31 '* -0. ,80 ** -0. , 19 -0. 26 * -0. 22 * 0. 31 *' -0. 19 * -0.61 * (-3.11) (-4. 63) (-1. .32) ( -2. 30) (-2. 37) (-2. 97) ( -2. 25) (-5.09) 10.06 * 24. 60 ** 2. 27 4. 51 2. 17 -4. 93 7. 93 + 9.24 + (2.21) (2. 72) (0. .40) (0. 88) (0. 57) (-1. 14) (1. 94) ( 1 .90) -1 .75 22. 51 * 3. .05 11 . 22 3. 97 -2. 43 13. 21 13 . 28 (-0.22) (2. 28) (0. .34) (1. 28) (0. 51) (-0. 25) (1. 04) (0.76) -0.07 -1. 28 -6. .02 + 3. 77 -2. 86 -6. 34 " -7. 67 ** -7.45 * (-0.03) (-0. 53) (-1 . ,97) (1. 53) (-1. 28) (-2. 91) (-3. 56) (-2.74) 28.36 ** 9. 12 11 . 91 10. 15 3. 67 -6. 84 -7. 06 -10.96 + (4.10) ( 1 . 16) ( 1 . 26) (1. 66) (0. 66) (-1. 16) ( - 1 . 59) (- 1.85) 11.35 * - 13. 20 + -6. 40 10. 76 * -5. 37 + 2. 92 -3. 36 -1.31 (2.47) (-1. 96) (-1. .15) (2. 20) (-1. 89) (0. 89) (-1. 26) (-0.44) -70.66 219. 67 * 144. 97 + -51 . 25 163. 43 ** 88. 57 91 . 33 + 1 18 . 30 * (-1.03) (2. 21) ( 1 . 70) (-0. 69) (3. 17) (1. 62) (1. 99) (2.18) 12.00 13. 00 18. 00 13. 00 19. 00 15. 00 12. 00 15.00 68.00 43. 00 53. 00 53. 00 50. 00 53. 00 57. 00 54.00 18.36 18. 75 20. 27 16. 10 14. 22 16. 27 13. 14 16.48 71 .29 80. 59 83. 41 84. 92 89. 93 87. 13 88. 76 86. 11 -304.61 - 196. 36 -247. 92 -231 . 78 -217. 74 -233. 01 -233. 98 -234.57 0.46 0. 57 0. 44 0. 53 0. 37 0. 59 0. 61 0.66 \ / \ / \ / \ / 27. 76 * 29. 74 * 21 . 30 25. 04 ' OO s i g n i f i c a n t at the 0.01 l e v e l , * s i g n i f i c a n t at the 0.05 le v e l , + s i g n i f i c a n t at the 0.10 level Table G.5 : The Royalty Rate within Eight Size Cohorts Number of Outlets: 2 - 1 0 1 1 - 2 0 21 - 32 33 - 53 5 4 - 9 1 92 - 196 197 - 490 491 + Variable Variable Variable Variable Variable Variable Variable Variable Fee Fee Fee Fee Fee Fee Fee Fee Av.% Discontinued) -1 . 43 *' -0, .26 0 .89 -0 .24 0. .51 -0 .24 -0. .26 -0 .95 (-2 .86) ( -0 .51) ( 1 .65) (-0 .43) (0 .68) ( -0 .32) ( -0. 47) ( - 1 .11) Foreign Outlets (%) 1 , .89 -4, .64 -10 .94 • 17 .29 5 .62 -3 .57 -7. .91 -5 .49 (0 .29) ( -1 .02) (-2 .03) (1 .42) (0. .95) (-1 .40) (-1. .42) (-1 .48) Number of States 0. . 37 ** 0, .01 -0, .26 ** -0. . 10 -0. .07 -0, .09 '* -0. ,05 0 .06 (2, .72) (0 .07) (-3 . 12) (-1 .40) (-1, .43) (-2 .81) ( - 1 . . 15) ( 1 .28) (Av.Sales-Inputs)/Av.Sales -0. . 34 •* -0, .21 * -0 . 25 ** -0 .24 * -0, .07 -0 . 39 ** 0 .21 * -0 .30 + ( -3 .85) ( -2 .70) (-3 .08) (-2 .17) (-0. .74) ( -2 .86) ( -2, .04) ( - 1 .89) Av.Sales/ outlet ($100K) 0. .24 -0, .51 -0 .23 0 .30 0, . 19 0 .47 0, .05 0 . 19 ( 1 . .02) (-1, . 38) (-0, .78) (0 .99) (0. .88) ( 1 .40) (0. .41) ( 1 . 36) Franchisee Experience -0. .07 1 . 05 0 .46 -0. . 34 -0. . 70 -0 .29 0, 99 0 .04 ( -0. 09) (1. .20) (0, .58) (-0, .35) (-0. .68) (-0, .31) (1. .06) (0 .04) Weeks of Training -0, . 19 0, .34 + -0, . 19 0. . 16 0. . 14 -0, .13 0 .23 0 .04 ( - 1 . .00) ( 1 , .74) (-1, . 17) (0 .65) (0, ,75) ( -0, .70) ( 1 . , 14) ( -0 .27) Outlets in 1986 (100's) -3. .35 0. , 19 -9, .17 -0, .04 3. ,51 2 .43 + 0 89 0 .09 '* ( -0. 24) (0, .02) (-1, .00) ( -0, .01) (0. 88) ( 1 , .79) (1, ,47) (3 .01) % Time Not Franchising 0. 02 0. ,01 0, .01 0. .04 • 0. ,05 * 0, .01 0. ,03 + 0 .08 •* (1 .33) (0, .42) (0, .71) (2, .17) (2. ,32) (0, .30) (1. .71) (3 .42) Years in Business 0. .00 0. .00 -0, .04 -0. .05 -0. ,06 -0, .01 -0. ,01 -0 . 16 •' (0. . 12) (0. .05) (-1, .48) (-1, .23) (-1 . 61) (-0 . 14) (-0. . 17) (-4 .48) Growth in Outlets 0, .97 -0, .85 -1. .80 + -3 .33 * - 1 . ,59 -1, .29 2. .31 -4 .74 (0. ,89) ( -0, 89) (-1, .76) (-2. . 35) (-1 . ,06) (-0, .67) (0, .92) (-1 .29) Capital Required 0, .01 0, .00 0. .01 -0. .00 -0. ,00 0 .00 -0, .00 0 .00 (1. ,43) (0. ,24) (1. 49) (-0. .84) (-1. .04) (0. .44) (-1. .34) ( 1 .47) Franchisor Financing 0. .43 -1. . 13 0 .60 1. .38 -0. , 39 3. .56 •* 1 . 54 1 .51 (0. 50) (-1. ,23) (-0. .57) (1. 43) (-0. 31) (3 .29) ( 1 , .54) ( 1 .27) Franchisor Inputs ($K) -0. .10 " -0. .05 + -0. .08 -0. .08 * -0. .03 -0, .08 ' -0. .03 -0 .08 + (-3. .78) (-1. .79) (-3 .21) (-2. 32) (-1. .02) ( -2 .25) (-1, .20) ( - 1 .75) Constant 42. , 12 ** 28. .65 •* 34. .49 *" 30. .85 ** 9. ,58 41 , .76 *' 25. .42 * 37 .87 * (4. ,65) (3. .58) (4. .02) (2. .75) (0. 99) (2. .96) (2, .39) (2 . 17) Limit Observations 4. .00 2. .00 3 00 4. .00 6. ,00 7 .00 4, .00 7 .00 Non-Limit Observations 76. ,00 54. .00 68. .00 62. .00 63. ,00 61 , .00 65, .00 62 .00 Standard error of Estm. 2 60 2. 34 2 .59 2. .90 3 38 3 .30 3 . 17 3 .61 At Mean X(I), E(Y) 6. ,55 6. 45 6. , 12 5. 86 6. .23 6 .08 7 .34 7 .00 Log Likelihood Function - 186. 09 125. ,67 - 165. 22 -158. .96 - 173. 26 - 167. 64 - 172. ,21 - 174 .82 Squared Cor(Y. E(Y)) 0. 28 0. ,29 0. 38 0. .25 0. . 13 0 .36 0, .25 0 . 39 \ / \ / \ / \ / Likelihood r a t i o tests 21 . 14 17 .50 20.44 23.06 + ** s i g n i f i c a n t at the 0.01 l e v e l , * s i g n i f i c a n t at the 0.05 level, + s i g n i f i c a n t at the 0.10 level Table G.6 : The Franchise Fee within Eight Size Cohorts Number of Outlets: 2 - 10 11 - 20 21 - 32 Fixed Fixed Fixed Fee Fee Fee Log. Av. % Discontinued -6 . 16 23. 62 24. 70 * (-0. 87) (0. 83) (2. 60) Foreign Outlets (%) 4 . 39 -21 . 23 -23. 40 (0. 18) (-0. 33) (-0. 90) Log. Number of States 4. 97 * 8. 41 -3. 48 + (2. 64) ( 1 . 30) (-1. 85) (Av.Sales-Inputs)/Av.Sales -0. 55 * 0. 39 -0. 15 ( -2. 04) (0. 36) (-0. 43) Av.Sales/ outlet ($100K) -0. 54 -11. 12 * 1 . 85 (-0. 71) (-2. 33) ( 1 . 36) Franchisee Experience 2. 31 16. 48 -6. 75 • (0. 79) (1. 35) (-1. 71) Weeks of Training 1 . 13 + 2. 77 -0. 68 ( 1 . 67) ( 1 . 07) (-0. 81) Log. Outlets in 1986 (100's) -4. 16 -35. 44 -19. 81 (-1. 49) (-1. 43) (-1. 67) % Time Not Franchising -0 . 01 0. 14 -0. 28 ' (-0. 19) (0. 57) (-3. 24) Log. Years in Business - 1 . 96 7. 50 2. 77 ( -0. 88) (0. 53) (0. 81) Growth in Outlets 2. 86 6. 22 2. 86 (0. 69) (0. 39) (0. 54) Log. Capital Required 2. 10 + 9. 61 • 7. 70 * (1. 73) (2. 27) (4. 25) Franchisor Financing 2. 26 -11. 03 8. 94 + (0. 71) ( -0. 87) ( 1 . 72) Log. Franchisor Inputs ($K) -3. 44 5. 75 -2. 06 (-1. 52) (0. 56) (-0. 64) Constant 61 . 99 + -163. 89 -39. 06 (1. 79) (-1. 04) (-0. 78) Limit Observations 0. 00 2. 00 1 . 00 Non-Limit Observations 80. 00 54. 00 70. 00 Standard error of Estm. 9. 46 32. 27 12. 52 At Mean X(I) , E(Y) 16. 63 28. 87 19. 83 Log Likelihood Function -284. 97 -265. 47 -277. 24 Squared Cor(Y, E(Y)) 0. 29 0. 36 0. 39 \ / \ Likelihood r a t i o tests 132 .08 " 24 '* s i g n i f i c a n t at the 0.01 l e v e l , * s i g n i f i c a n t at the 0.05 le v e l , 33 - 53 54 - 91 92 - 196 197 - 490 491 + Fixed Fee -1.86 (-0.14) -8.01 (-0.10) -2. 10 (-0.75) -0.38 (-0.66) 0. 16 (0.09) -6.81 (-1.16) 1.39 (1.09) -0.96 (-0.06) -0.01 (-0.08) 5.99 ( 1.10) 16.89 + (1.75) 4.85 + (1.81) 14.79 ' (2.30) -4.81 (-0.92) 32.83 (0.41) 0.00 66.00 18.41 23.60 -277.39 0.31 / Fixed Fixed Fee Fee 21. ,05 + -29. .53 (1. .77) (-2, .66) 9 . 40 11 , .98 (0. ,36) ( 1 . .02) -0. ,98 3 . 15 (-0. 40) ( 1 . 39) -0. .25 -0 .62 (-0. 73) (-1 .62) -1. .82 • 0 .08 (-2. 58) (0, .06) -3. , 76 4, .66 ( -0. 82) ( 1 . 13) -0. 69 0, .66 (-0. .81) (0 .75) -23. .51 + 2 .37 (-1. ,80) (0, .29) -0. .07 0, .07 (-0. ,80) (0. .78) 0. ,92 -6, .26 (0. ,23) (-1, .63) 16. ,90 ' -1 , .37 (2. .24) (-0, .16) 13. ,69 ** 0, .97 (6. ,48) (0, .55) 21. 70 ** 0. .04 (4. ,05) (0, .01) -3. .50 -6 .62 (-1. , 19) (-2, .26) -23. 88 125 .53 (-0. 48) (2, .57) 0. .00 1 .00 69. .00 67 .00 15. .19 14 .70 22 .75 21 , . 78 277, .17 -276, . 16 0 .51 0 .28 \ / 37.6 ** Fixed Fixed Fee Fee 6. , 33 -12 .09 (1. .07) (-0. .80) - 19 . .88 -8 . 1 3 (-1. .07) (-0. .34) 3. .55 - 3. 52 ( 1 . . 15) ( -0, .71) 0. . 29 0, .26 ( 1 . ,05) (0, .51 ) -0. . 30 -0 .02 (-0. .79) (-0 .02) 4. .84 -2, , 32 ( 1 . .52) ( -0 . 38) 1 , .35 + 0 . 17 ( 1 , . 82) (0 . 18) 12 .10 + -3 .66 (1. .98) (-0, .88) 0, .09 -0 .05 (1. .50) (-0, . 39) -9, .48 ** 2, .86 ( -3 .02) (0 .49) - 18. 95 * 12 . 38 ( 2 .07) (0 .57) 1 ,62 1 .42 (1. ,08) (0, .45) 10. . 16 * * 9 .66 (2, .89) ( 1 . 37) 1 .81 1 .03 (0, .87) (0 . 28) -26, .43 13 . 29 (-0 .74) (0 .20) 2 .00 1 .00 67 .00 68 .00 10 .54 21 .73 20 .92 24 . 32 -254 .93 -306 .89 0 . 39 0 . 11 \ / 46. 18 ** s i g n i f i c a n t at the 0.10 level 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0098068/manifest

Comment

Related Items