UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Three essays on R&D and technological progress Zhang, Yimin 1989

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

Item Metadata

Download

Media
831-UBC_1989_A1 Z48.pdf [ 6.19MB ]
Metadata
JSON: 831-1.0098182.json
JSON-LD: 831-1.0098182-ld.json
RDF/XML (Pretty): 831-1.0098182-rdf.xml
RDF/JSON: 831-1.0098182-rdf.json
Turtle: 831-1.0098182-turtle.txt
N-Triples: 831-1.0098182-rdf-ntriples.txt
Original Record: 831-1.0098182-source.json
Full Text
831-1.0098182-fulltext.txt
Citation
831-1.0098182.ris

Full Text

T H R E E E S S A Y S O N R & D A N D T E C H N O L O G I C A L P R O G R E S S By Yimin Zhang A THESIS S U B M I T T E D IN P A R T I A L F U L F I L L M E N T O F T H E R E Q U I R E M E N T S F O R T H E D E G R E E O F D O C T O R OF P H I L O S O P H Y in T H E F A C U L T Y OF G R A D U A T E STUDIES C O M M E R C E A N D BUSINESS A D M I N I S T R A T I O N We accept this thesis as conforming to the required standard T H E U N I V E R S I T Y OF BRITISH C O L U M B I A February 1989 © Yimin Zhang, 1989 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. Commerce and Business Administration The University of British Columbia 2075 Wesbrook Place Vancouver, Canada V6T 1W5 Date: Abstract This thesis investigates several aspects of research and development (R&D) and techno-logical progress. In Chapter 2 of the thesis, major stylized facts about R&D and other economic activities established in the existing empirical literature are reviewed and sum-marized. The chapter also undertakes some original empirical work. Specifically, the role of firm size in the determination of R&D intensity is examined, and the market evalua-tion of intangible assets of knowledge is investigated. Both relationships are estimated using alternative formulations and a set of pooled cross-section and time-series data for the U.S. In Chapter 3, the relationship between the R&D investment decisions and the finan-cial decisions of a firm is studied from the perspective of managerial behaviour. A model of managerial agency costs is set up and the relationships between agency costs, debt financing, R&D investment, and the technological state of the firm are analyzed. Com-parative statics analysis shows that firms with different technologies may choose different levels of financial leverage and R&D investment. In particular, firms possessing superior technology tend to invest more on R&D projects and use less debt than firms with "nor-mal" technology. The chapter also presents some empirical evidence that supports this relationship. In Chapter 4, a general equilibrium model is set forth to study the economic conse-quences of technological progress in the manufacturing sector. Empirical evidence has shown that the service sector has been growing rapidly relative to manufacturing, signify-ing a significant shift in economic structure in several countries, particularly the U.S. This n essay shows that this structural shift can be partly explained by fast growing produc-tivity in the manufacturing sector. Also, the impact of the structural shift, particularly on capital investment, is examined within a general equilibrium framework. Different assumptions, with regard to the capital intensiveness of the service sector and with re-gard to the type of technological change, yield different predictions. Empirical evidence reveals that past technological changes have been principally labour saving and that the service sector in the U.S. economy is likely to be relatively capital intensive. i n Table of Contents Abstract ii List of Tables vi List of Figures vii Acknowledgement viii 1 Introduction 1 2 Some Stylized Facts about R&D 7 2.1 Introduction 7 2.2 Some Stylized Facts in the Empirical Literature 8 2.2.1 Patents and R&D 9 2.2.2 Composition of R&D and Productivity Growth 11 2.2.3 R&D and Financial Markets 12 2.2.4 Market Structure and R&D 13 2.2.5 R&D Strategy 15 2.2.6 International Invention, R&D and Foreign Trade 16 2.3 Further Discussions on R&D and Financial Markets 17 2.4 A New Investigation of R&D and Financial Markets 21 2.5 Further Discussions on R&D and Market Structure 23 2.6 Concluding Remarks 26 2.7 Appendix to Chapter 2 31 i v 3 R & D Investment, Technology and Financial Structure 42 3.1 Introduction . . 42 3.2 The Model 47 3.2.1 R&D and Technological Progress 47 3.2.2 Financial Structure of the Firm 48 3.2.3 Manager's Utility Function and Agency Costs 49 3.2.4 The Optimization Problem 57 3.3 Solution of the Model 60 3.3.1 The Conditions of Optimization 60 3.3.2 Solutions for Different Current Technologies 62 3.4 Empirical Evidence 66 3.5 Conclusion 76 4 Economic Consequences of Technological Progress 78 4.1 Introduction . 78 4.2 The Model 85 4.3 General Predictions of the Model 92 4.4 Numerical Simulations 107 4.5 Accelerator Models of the Investment Function 120 4.6 Conclusion 126 Bibliography 130 v List of Tables 2.1 Regression of Tobin's Q 22 2.2 R&D and Firm Size, I 25 2.3 R&D and Firm Size, II 27 3.4 Grouping of the Data 70 3.5 Regression of R&D 73 3.6 Regression of Debt 74 4.7 Simulation Results for Model Economy, I 109 4.8 Simulation Results for Model Economy, II I l l 4.9 Estimated Equations of Investment Function 124 vi List of Figures 4.1 Nonresidential Fixed Investment 82 4.2 Volumes of Goods and Services 83 4.3 Prices of Goods and Services 84 4.4 Predicted Goods : Neutral 112 4.5 Predicted Services : Neutral 113 4.6 Predicted Investment : Neutral 114 4.7 Predicted Capital Stock : Neutral 115 4.8 Predicted Goods : Labour Saving 116 4.9 Predicted Services : Labour Saving 117 4.10 Predicted Investment : Labour Saving 118 4.11 Predicted Capital Stock : Labour Saving 119 vii Acknowledgement I am deeply indebted to the members of my supervisory committee, Professors James A. Brander, Maurice Levi, and Ashok Kotwal, as well as Professor Tae H. Oum for their valuable support and encouragement. I also wish to thank CIDA (Canadian International Development Agency) program coordinator Grace Wong and English tutor Mackie Chase for their help. Financial sup-port from CIDA is gratefully acknowledged. V l l l Chapter 1 Introduction Technological progress has dramatically changed modern economies. Firstly, productiv-ity growth continuously raises national income. Secondly, the quality of consumer goods and services constantly improves. Thirdly, new products and services appear regularly. All these consequences of technological progress have tremendously benefited consumers. While there is hardly any doubt that intensive research and development (R&D) ac-tivities significantly contribute to technological progress, many questions regarding the economics of R&D, technological progress and other related economic issues remain un-settled. Despite the fact that the economic literature on these topics is quite large and is rapidly growing, more research work is called for to clarify many issues currently under dispute. Here, I report my studies of some aspects of R&D and related issues in the form of three essays. The first essay (Chapter 2 of this thesis) presents major stylized facts found in the existing empirical literature regarding R&D and other economic activities and re-examines some issues using alternative models. The second essay (Chapter 3) studies the interrelationship between a firm's financial decisions, its R&D investment decisions, and its state of technology. The third essay (Chapter 4) analyzes the overall economic consequences of technological progress in manufacturing industries as a result of intensive R&D investment. The first essay, titled "Some stylized facts about R&D", reviews part of the existing empirical literature and summarizes major stylized facts regarding R&D expenditure, patents, productivity, market structure, international trade, etc. The aim of this essay 1 Chapter 1. Introduction 2 is to give a broad picture of the current knowledge about the subject and to serve as a basis for further research. In addition, after the major stylized facts are presented, some empirical issues are re-examined. The issues include the role of firm size in determining R&D investment, and the evaluation of firms' intangible assets by financial markets. Since most existing studies are based on cross-sectional observations, the use of a pooled cross-section time-series data set in this essay provides additional information concerning these important issues. According to the relationship between R&D and other economic activities, the major stylized facts can be classified into several categories. For example, some authors studied the relationship between R&D and patents and found that there was a general statistical relation between the number of patents and R&D input. Hence patents seem to be a good indicator of innovative output, which is intrinsically difficult to measure. Some other authors examined the composition of R&D and productivity growth. They found that basic research and long term R&D projects contributed more significantly to productivity growth, and that large firms and firms in very competitive industries contributed more than proportionally to basic research. Furthermore, in some papers, attention is focused on R&D and financial markets with the general conclusion that the rate of return on R&D capital is higher than the rate of return on conventional capital. In other work, the subject of study is the structure of output markets. This is a subject of considerable controversy although empirical tests have concentrated on hypotheses concerning the role of firm size and industrial concentration as determinants of firms' R&D investments. Finally, there are many studies reporting empirical findings regarding firms' R&D strategies, i.e., innovative versus imitative, and regarding international diffusion of innovation and the role of foreign trade. The findings provoke interest for further research. After the major stylized facts are presented, the remainder of the first essay is devoted to a re-examination of some of the issues. Since the reviewed empirical work on R&D and Chapter 1. Introduction 3 market structure is based on cross-sectional observations, a re-examination of this issue using a pooled cross-section time-series data set yields additional evidence regarding the role of firm size. Another issue, namely the creation of intangible capital of knowledge and its evaluation by financial markets is re-examined with added consideration of financial leverage. The results reveal that financial leverage is an important factor for the market evaluation of a firm's intangible assets created by R&D investment. Hence, financial decisions are closely interrelated with a firm's R&D investment decisions. The second essay, titled "R&D investment, technology, and financial structure", stud-ies the relationship between a firm's financial decisions and R&D investment decisions from the perspective of managerial behaviour. A model of managerial agency costs is set up, in which a manager makes all the investment and financial decisions for a firm. The perquisites the manager extracts from the job form the agency costs to the shareholders of the firm. These costs can be reduced by debt financing. It is found that when the manager's utility is based on both the perquisites of the job and the monetary income from the firm's residual profit, it is to the advantage of both the manager and the share-holders of the firm to use some debt financing. However, bankruptcy costs will prevent the manager from using too much debt. On the other hand, the manager will also make decisions concerning the firm's R&D investment. For a financially leveraged firm, the manager must carefully compute the risk of bankruptcy for the firm. If bankruptcy occurs before the firm's R&D projects can render any benefit, the R&D investment will be a sheer loss because of the intangible nature of R&D capital. Hence, R&D expenditure will be constrained by the firm's ability to repay its debt obligation. The manager will make both financial decisions and R&D investment decisions conditional on the firm's current state of technology. Taking into consideration that the firm's state of technology in the future depends on the outcome of current R&D investment, the manager's simultaneous decisions on R&D and finance Chapter 1. Introduction 4 indeed become optimal in a multiperiod setting. Empirical tests are carried out for four U.S. industries. The results show that positive correlations exist between firms' R&D investment and state of technology. They also show that negative correlations exist between firms' debt level and state of technology. One interpretation is that when a firm possesses a superior technology, enjoying a higher profit, its financial condition becomes more secure and the firm will invest more in R&D to exploit potential profits into the future. At the same time, the firm will reduce its debt level to minimize the risk of bankruptcy. This behaviour is certainly not surprising because a firm aiming at future prosperity should be expected to avoid the risk of early bankruptcy which will ruin all of its future prospects. Therefore, the model presented in this essay also offers an alternative explanation to the stylized facts established in the empirical literature that financial leverage is negatively correlated with firms' R&D expenditure. While the second essay focuses on R&D investment decisions from the point of view of a single firm, the overall economic consequences of technological progress in the manufac-turing sector is studied in the third essay titled "Economic consequences of technological progress". It is well established that R&D has greatly contributed to productivity growth in manufacturing industries. As a result, while producing more commodities for an aver-age consumer, the manufacturing sector actually employs a smaller share of labour than before. The service sector, in contrast, has been experiencing rapid growth, absorbing increasing amounts of productive resources. This signifies a structural shift in the econ-omy. The economy has been turning from producing more goods to producing more services. Presumably, if the service sector employs a capital/labour ratio which is differ-ent from the capital/labour ratio of the manufacturing sector, the overall capital/labour ratio of the economy will change markedly as services outgrow manufacturing. This will inevitably affect business investment in the economy. Chapter 1. Introduction 5 Traditionally, manufacturing has been considered capital intensive relative to ser-vices. Therefore, the relative growth of the service industries should, other things equal, decrease the overall capital/labour ratio of the economy, and have a negative effect on business investment. However, empirical evidence does not show a decline in investment, which makes one speculate that services may be in fact more capital intensive. Here a distinction is made between capital installation and capital consumption. Service indus-tries rely heavily on buildings such as business offices, stores, warehouses etc. While the depreciation of buildings is slow, the original costs of construction are high. Hence, in terms of capital consumption (chiefly through depreciation) services can still be regarded as labour intensive; while in terms of capital installation, services may well be capital intensive. Based on these arguments, the analysis of economic consequences of the structural shift from manufacturing to services is evidently complicated. The essay sets forth a general equilibrium model of a three-sector economy consisting of manufacturing, services and capital investment. The analysis shows that the noted structural shift in the economy can be (at least partially) attributed to technological progress in manufacturing industries as a result of active R&D. The overall impact of such technological progress varies in the long-run, depending on the nature of the technological change. For example, a labour saving technological change in manufacturing will encourage investment, whereas a capital saving technological change will discourage investment. Moreover, a neutral technological change will stimulate investment in the short-run but will eventually retard investment in the long-run. The empirical section of this essay carries out illustrative simulations based on the model economy to draw a complete picture of the dynamic process underlying the struc-tural change caused by technological progress. The results reveal that in certain cases, the predictions of the model depend on the assumptions regarding the relative capital Chapter 1. Introduction 6 intensiveness of service sector. In addition, the impact of technological change upon business investment is also examined within the framework of a conventional accelerator model of investment. It is found that disaggregating total output enhances the explana-tory power of the accelerator model, and that the overall trend in business investment can be segregated into two parts, the effect of structural shift in the economy and the effect of fundamental technological change. On the whole, these three essays show that R&D related economic issues provide a rich field for further research. Past studies have been very fruitful; however, with constantly improving data and econometric techniques, future work will certainly be even more rewarding. Chapter 2 Some Stylized Facts about R & D 2.1 Introduction There is a large body of literature devoted to the subject of research and development (R&D) and related economic issues. Despite the fact that this literature covers a wide range of issues, many questions regarding the role of R&D in various economic activities and the relationships between R&D and other economic variables remain unanswered. Admittedly, there are many intrinsic difficulties in the study of innovative activities. Nev-ertheless, due to the fundamental importance of such activities in advancing technology, which is the major driving force of modern economies, studies on industrial R&D and other relevant economic activities have never stopped and more work in this area is called for. To give a broad picture of the current knowledge about the subject, which serves as a basis for further research, this essay reviews part of the existing empirical literature and summarizes major stylized facts regarding R&D expenditure, patents, productivity, market structure, and international trade and invention etc. Although some of them are not in unanimous agreement, these empirical findings provide a background of facts, from which one may draw inductive inferences, make deductive statements, or raise new questions. After the major stylized facts are presented, some interesting issues will be re-examined. As is well known, in the study of the determination of a firm's R&D intensity, there has long been a controversy regarding the role of firm size. Although the existing empirical 7 Chapter 2. Some Stykzed Facts about R&D 8 studies have yielded different results, they are commonly based on cross-sectional ob-servations. We will look at the time-series dimension and test the relationship between R&D intensity and firm size using a pooled cross-section time-series data set. The re-sults generally show that the relationship between R&D intensity and firm size varies among different industries. While there are exceptions, the size effect seems to be more important in time-series than in cross-section. Another issue, namely, the creation of the intangible capital of knowledge and its evaluation by the financial markets, will also be examined using a model which is an alternative to that employed in some literature. The results reveal that the firm's financial leverage is an important factor to consider when the market evaluates a firm's intangible assets which are created by R&D investment. The essay is organized as follows: section 2 reviews the major stylized facts concerning R&D and provides some interpretations; section 3, section 4, and section 5 discuss and re-examine some issues regarding R&D, financial markets, and firm size; section 6 contains concluding remarks; a summary of major stylized facts appears in an appendix at the end. 2.2 Some Stylized Facts in the Empirical Literature Obviously, R&D is related to many important economic issues, hence, the major empirical findings can be classified into several categories according to the relationships between R&D and other economic activities. We shall begin with the issue of the measurement of innovative output, then examine the composition of R&D, productivity growth, rate of return on R&D, firms' R&D strategies, market structure, and evolution of technology. Finally , we will look at the issue of international invention and foreign trade. Chapter 2. Some Styhzed Facts about R&D 9 2.2.1 Patents and R & D R&D has been widely recognized as economic investment, however, the fundamental output of such investment is knowledge, which is not always marketable and so cannot always be measured in monetary terms. Besides, the intangible nature of knowledge often prevents any direct measurement, monetary or nonmonetary, of R&D output. This makes the assessment of R&D investment very difficult. However, with the recent availability of patent data, economists have been able to use the number of successfully applied patents as a measure of R&D output. It is quite obvious that a successful application of a patent should never be the sole purpose of any R&D activity. In other words, should there exist a production function describing R&D, the output variable would never be successful patenting. Furthermore, not all R&D output can be patented. For example, the law of energy conservation, the result of numerous failed experiments to seek some machine which could work without consuming net energy, could never possibly be patented. Other natural and physical laws or mathematical algorithms cannot be patented either. The philosophy behind this restriction is that the costs of monopolization of basic discoveries in science are too high. Hence, the number of patents is at best an indirect measure of innovative output. Nevertheless, economists do find a statistical relationship between the number of patents and R&D input, and some conclude that patents do seem to be a good indicator of innovative output (See, for example, Pakes and Griliches, 1984), which is reassuring because a good measure of innovative output will be highly valuable in some policy assessment. It appears that the patent-R&D relationship is statistically significant and strong in cross-section data, but weakly significant in time series data (Bound et al, 1984, Pakes and Griliches, 1984). Also, cross-sectionally, small firms tend to patent more per R&D dollar than large firms, and larger firms have a nearly constant ratio of patenting to Chapter 2. Some Stylized Facts about R&D 10 R&D. Taken at face value, this would suggest general constant returns to scale with initial decreasing returns to scale. However, in fact, many small firms do not carry out any R&D projects at all; therefore, those that do sustain R&D (mostly in high-tech industries) must do it exceptionally well. Hence the initial higher patenting-R&D ratio may be the result of a biased sample of small firms. On the other hand, it is found in time series that the number of patents per R&D dollar consistently decreases in time. Whether this is caused by a change in firms' propensity to patent, or an increase in R&D costs due to general installation of more expensive equipment, or the exhaustion of technical opportunity is unclear. Although not perfect, the patent rate provides us with a convenient measure of in-novative output which makes empirical tests of some hypotheses possible. For example, there have been two conflicting theories about technological evolution. One is Kuznets' view (Kuznets, 1930) that declining technological opportunity over the industry life-cycle leads to a decline in innovative activity; the other is Schmookler's view (Schmookler, 1966) that when an industry matures, larger market shares make the rate of return on R&D investment higher and therefore provide more incentive to innovate. Taking patent rate as the measure of innovative activity, Gort and Wall (1986), and Scherer (1982) confirm both the life-cycle hypothesis that declining technological opportunity retards innovation, and the demand-pull hypothesis that larger market share contributes to in-novation. They find that the net effect on the time path of R&D is the result of the above two offsetting factors and cannot be specified c priori. Jaffe (1986) makes very interesting use of patent data to examine supply side ef-fects, namely, technological opportunity and R&D spillovers, on the rate and direction of R&D. He classifies patents and R&D into a technological space and shows that R&D productivity of a certain firm is increased by the R&D of the firm's "technological neigh-bour". That is, there are positive technical spillovers among firms employing similar Chapter 2. Some Styhzed Facts about R&D 11 R&D portfolios. On the other hand, the technological neighbour's R&D is shown to have a negative effect on the profits and market values of firms with low R&D investment which are not actively involved in R&D competition and therefore unable to take ad-vantage of other firms' R&D spillovers. Also, Jaffe finds that firms are adjusting their technological composition in response to changing technological opportunities. 2.2.2 C o m p o s i t i o n of R & D a n d P r o d u c t i v i t y G r o w t h There are many ways of classifying R&D projects, e.g., basic research v.s. applied re-search, short term v.s. long term, more risky v.s. less risky, etc. Mansfield (1981) reports that the composition of R&D differs substantially both interfirm and interindustry. In-terfirm, percentages of basic research and long term projects increase with firm size, i.e., large firms contribute more than proportionally to basic research. Interindustry, the percentages of basic research and long term projects decrease with the four-firm concen-tration ratio, i.e., the less competitive the industry is, the less the proportion of R&D which is allocated to basic research. As related to the productivity growth, Mansfield (1980) and Griliches (1986) conclude that while overall R&D contributes significantly to productivity growth, the basic research component does so even more strongly. In view of this, they worriedly report that many industries have recently changed R&D compo-nents, resulting in cuts in the proportion of basic research, coinciding with the recent productivity slow down. Griliches (1986) also looks into the share of government financed R&D and finds it rather unstable. In fact, total R&D expenditure in U.S. industries (in real terms) peaked around 1968, dropped in the early 1970's and recovered somewhat later. However, as a percentage of sales, privately financed R&D kept almost constant, while federally financed R&D fluctuated. Moreover, it is shown that privately financed R&D expenditure has a significantly larger effect on private productivity than federally financed R&D. Combining Chapter 2. Some StyHzed Facts about R&D 12 all these facts may suggest that government financed R&D is more likely aimed at non-basic research, short term and safe projects. If this is true, one would have reason to question the wisdom of such a policy. On the other hand, in comparison to the rest of the world, Evenson (1984) finds that the importance of the United States in world patenting has declined whereas the importance of Japan, West Germany, and planned economies has risen, perhaps revealing changing patterns in comparative advantages. 2.2.3 R & D and Financial Markets . Since most industrial R&D activity is undoubtedly an economic investment, the funda-mental problem for a firm in determining its optimal R&D investment is to estimate the rate of return on an R&D project. For a project manager, this may not be particularly difficult on an ad hoc basis. However, for econometricians whose major interest is not in any individual project, the task of estimating the rate of return on R&D proves to be quite complex. Some early estimates yield a rate of return on R&D as high as 30-40 percent. The fact that such high rates of return did not cause an "R&D rush" has been a puzzle. Pakes and Schankerman (1984) argue that when the research gestation lag and the rate of technological obsolescence are taken into account, which would shorten the profitable lifetime of R&D capital while prolonging its investment horizon, the rate of return on R&D is considerably reduced. Based on their estimates, the rate of return on R&D would not exceed twice the rate of return on conventional capital. Mohnen et al (1986) recently estimated the rate of return of R&D for manufacturing industries in the U.S., Japan, and Germany. They begin with a model of production structure, from which the demand for factor input is derived and then the rates of return on quasi-fixed factors, one of which is R&D capital, are estimated. By this approach, they find that the rate of return on R&D for the three countries is about 10-15%, a return higher than Chapter 2. Some StyHzed Facts about R&D 13 the rate of return on conventional capital, yet quite reasonable. As evidenced by the ability of bringing a positive rate of return, R&D investment is sometimes said to have created "intangible" capital. Naturally, this sort of capital should affect the market valuation of the firm. Griliches (1984) and Pakes (1984) take up this issue, studying the market reaction to firms' R&D investment and patenting. Their major findings are that only unanticipated changes in R&D expenditure are significantly correlated with changes in firms' market value and that patenting, being unanticipated in nature, always affects the firms' market value. On the other hand, Long and Malitz (1985) examine the relation between financial leverage and R&D investment. Their theory predicts that optimal financial leverage should be negatively correlated with the amount of intangible assets since agency costs of debt increase with intangible assets due to higher monitoring costs. They take firms' R&D investment along with advertising expenses as the proxies for intangible assets and find that financial leverage and R&D investment indeed bear a significantly negative correlation. 2.2.4 Market Structure and R & D Market structure and R&D is a subject which has been undergoing much debate. Kamien and Schwartz (1975) made an extensive survey of the ample literature in the field. This essay only focuses on presenting major stylized facts. The empirical tests on the subject are largely focused on two sets of hypotheses, namely, the role of firm size and the role of industrial concentration as determinants of firms' R&D investments. The findings are by no means consistent. For example, Hamberg (1964) finds that R&D intensity, defined as the ratio of R&D expenditure to sales, decreases weakly with firm size, and Comanor (1967) obtains a similar result. Scherer (1965) finds that the relationship between R&D intensity and firm size is nonmonotonic. Specifically, R&D intensity first increases with Chapter 2. Some Stylized Facts about R&D 14 firm size but, after reaching a threshold, it decreases with firm size in most industries. One exception is the chemical industry in which R&D intensity increases with firm size throughout. Scherer's finding is generally cited in the literature as the inverted-U relationship. However, Bound et al (1984) finds a relationship rather like an upright-U. In other words, both small firms and large firms do R&D more intensively than medium sized firms. Pavitt et al (1987) obtain similar results for firms in the U.K. Apparently, their results contradict the results of earlier work. Bound et al give three reasons for this contradiction. Firstly, earlier work used relatively small samples consisting of larger firms. Secondly, they measure R&D by expenditure whereas earlier work used the number of R&D employees. Finally, the relationship may have changed over time since the earlier work used data 15 to 20 years older than their data. On the other hand, Cohen et al (1987) suggest that firm size has no direct effect. Cohen et al argue that firm size is only superficially correlated with R&D intensity; what really determine R&D intensity are the factors underlying the industrial differences in technological opportunity, appropriability conditions, etc. Based on the responses from a survey, Cohen et al try to quantify these underlying factors. They construct variables like closeness to science, importance of external source of knowledge (e.g. knowledge obtained from government research units, suppliers of material and equipment, and users of the output), industry maturity, appropriability conditions, etc. They find that, by itself, firm size has some effects on R&D intensity. However, when the other quantified variables controlling industrial differences are included, firm size is no longer significant. A similar situation also arises in the empirical tests for the role of industrial concen-tration in determining R&D intensity. Earlier literature (e.g. Scherer, 1967; 1980, p.437) established an inverted-U relationship between R&D intensity and seller concentration. The maximum R&D intensity appears at the four firm concentration level of 50 to 60. Chapter 2. Some Styhzed Facts about R&D 15 Recently, Scott (1984), and Levin et al (1987) show that when other variables controlling industrial and f i rm differences in technological opportunity, appropriability conditions, etc. are taken into account, seller concentration does not remain significant. 2.2.5 R & D Strategy There are basically two kinds of strategies in R&D activity, namely, innovative and imitative. For example, an R&D project that will bring a new product into the market is innovative, while an R&D project that tries to imitate a certain product already existing in the market is imitative. Generally speaking, imitative R&D is cheaper than innovative R&D both in terms of monetary costs and the time required. Mansfield et al (1981) obtained the estimates that the ratio of the costs of imitation R&D to the costs of innovation R&D for acquiring a new technology is around 0.65 and the ratio of the time required for completing the imitation to the time needed for the original innovation is around 0.70. Of course, there always exists some trade-off between imitation costs and imitation time. The above estimates are based on unpatented innovations. For the case of patented innovation, it is shown that while patenting cannot completely eliminate imitation, it does increase the imitation costs. Mansfield et al also reveal a surprising but strong result that industrial concentration is significantly related to the ease of imitation. From a simple regression, the ratio of imitation costs to innovation costs alone explains four-firm concentration ratios with an R-square of 0.60. This suggests that if a firm finds it easy to imitate other firms' innovations, the firm is likely to find itself in an industry characterized by a lack of concentration. From a different approach, Link and Neufeld (1986) test the relationship between the innovativeness of a firm and the variables characterizing market structure and find that firm size and market share are significantly and positively related to innovation whereas industrial concentration is negatively but not significantly related to innovation. Chapter 2. Some Stylized Facts about R&D 16 2.2.6 International Invention, R & D and Foreign Trade The importance of international invention is manifested by a high degree of foreign patent-ing in most of the world's economies. For example, in the U.S. 30-40% of the patents are granted to foreign applicants (Evenson, 1984). While this high degree of foreign patent-ing is a worldwide phenomenon, there are other characteristics of international invention that vary widely among countries. Evenson (1984) argues that only innovative outcomes of R&D done overseas would be patented at home, whereas most outcomes of adaptive R&D would be patented at the local level. Hence, for any country, the ratio of patents granted abroad to patents granted at home would be a measure of adaptiveness in the nature of R&D done by that country. Finding that this ratio in industrialized countries is 10 times higher than that in developing economies, Evenson concludes that R&D in developing economies is mainly adaptive rather than innovative. Pakes and Griliches (1984) report that the number of patents per R&D dollar for U.S. industries, with the exception of drug firms, consistently decreases with time. Evenson (1984) finds this negative trend prevailing in almost all industrialized countries. More-over, the number of patents per scientist and engineer has also declined over time in industrialized countries. These trends seem to reveal that it has become increasingly difficult to make innovations. Another trend reported by Evenson is that total invest-ment in R&D as a percentage of total GDP declines over time for most industrialized countries. It seems plausible that this negative trend in R&D investment, relative to GDP, is the economic consequence of the negative trend in R&D productivity. If this is true, we can reasonably predict that technological progress is slowing down at present and will continue to slow down in the future. Another issue in international invention is the relation between R&D and foreign trade. Mansfield, Teece et al (1979) find that the percentage of R&D carried out overseas Chapter 2. Some Stylized Facts about R&D 17 by U.S. firms is positively related to the percentage of sales from foreign subsidiaries, but negatively related to the percentage of firms' sales from exports, confirming that export of goods is a substitute for export of technology. In another study, Mansfield, Romeo et al (1979) also find that foreign sales play a significant role in determining firm's optimal investment in R&D. It is reported that firms having a larger percentage of foreign sales, especially R&D intensive ones, count more on foreign revenues to returns on R&D projects, although the amount of product R&D is much more significantly related to foreign sales than the amount of process R&D. Thus it can be concluded that foreign sales stimulate domestic R&D. On the other hand, Hughes (1986) tries to test the reverse relation, i.e., the role of R&D in determining exports. Hughes uses U.K. data and reports that both the level of R&D measured by annual expenditures and the gap in R&D, i.e., the difference between U.K. and foreign levels of R&D expenditures, have positive effects on exports. In view of the stimulation that exports put on domestic R&D, Hughes concludes that it is a simultaneous relation that exists between R&D and exports. The major stylized facts from the empirical literature reviewed in this section are summarized in the Appendix. 2.3 Further Discussions on R & D and Financial Markets The preceeding section presents the major stylized facts found in the existing empirical literature. In some areas such as international invention, the existing empirical work is mostly descriptive in nature. A theoretical framework has been lacking. In some other areas, e.g. market structure and R&D, extensive empirical tests are performed centering around a few well known theoretical arguments. In other areas, e.g. R&D and financial markets, both empirical and theoretical work is just beginning. These are the areas which Chapter 2. Some Styhzed Facts about R&D 18 demand further investigation. In one of the first attempts, Griliches (1984) establishes a model linking the firm's R&D and patents with the ratio of the firm's market value and its real assets. The market value of the firm is assumed to be the sum of the firm's real assets and its intangible assets consisting of the stock of profitable knowledge which is accumulated through R&D investments. Specifically, let V = q(A + K), (2.1) where V is the market value of the firm, A is the current value of the firm's conventional assets, K is the current value of the firm's intangible "stock of knowledge", and q is the current market valuation coefficient of the firm's assets. Dividing V by A gives the ratio V/A, which is commonly called Tobin's Q. Taking logarithms leads to lnQ =hij =ln<? + ln(l + ^ ) , (2.2) The value of Q represents the ratio between the two valuations of the same physical asset, viz. V and A. Specifically, V is the market valuation —the market price for exchanging existing assets—and A is the replacement cost—the price for newly produced commodities. Tobin and Brainard explain why V should deviate from A (Tobin et al, 1977, p.237): The various physical assets of a business enterprise are often designed, installed, and used in complex combinations specific to the technology. It is costly or impossible to detach and move individual assets or to apply them to alternative purposes. The valuation of the business as a whole as a going concern is generally much more relevant than the separate valuations of the assets on used goods markets. Chapter 2. Some StyHzed Facts about R&D 19 Nevertheless, it is conceivable that the equilibrium value for Q should be unity. If Q is greater than unity for some firm, the construction of a similar business would be profitable as well. Hence Q being greater than unity would stimulate investment, resulting in expansion or new entry. Similarly, if Q is less than unity, this would discourage investment. However, the construction of a complete new plant is often costly and time-consuming. For existing firms, the equilibrium state would dictate that the marginal value of Q equals unity. That is, the ratio of the increment of market valuation of the investment to the cost of the investment is unity. This does not necessarily imply that the average Q is unity since for firms with monopoly power or non-constant returns to scale, the average value of Q can be quite different from the marginal value of Q. Since Q plays a significant role in determining business investment, there have been many empirical studies on this subject (See, for example, Tobin et al, 1977, Brainard et al, 1980, Lindenberg et al, 1981). However, given the presence of intangible assets, the valuation of Q would be somewhat ambiguous. For example, consider two firms with similar plant and equipment but assume that only one of them carries out R&D programs. Then the market values of the two firms would differ by the amount of the intangible assets created by the R&D investment whereas the replacement values of the physical capital of the two firms would be the same because an R&D project cannot be replaced in the commodity market. Hence, Equation (2.2) actually represents a decomposition of Q. Specifically, the first term at the right is the variable q which, as defined in (2.1), represents the ratio between the firm's market value and its intrinsic value. The latter is the sum of the reproduction costs of the firm's physical assets and the value of its intangible assets. The second term at the right of Equation (2.2) captures the discrepancy between the two ratios, Q and q, which is ignored in most of the studies regarding Tobin's Q. Chapter 2. Some Stylized Facts about R&D 20 It is clear that when intangible assets are non-trivial, the equilibrium value of Q need not equal un^r. In fact, Q should exceed unity. As the intangible assets of knowledge are not readily obtainable from the commodity markets, a Q-value greater than unity would not necessarily attract new entry. A mere duplicate of any physical plant could not enjoy the same market value as the existing one carrying on an intangible capital and thus would not attract new entry. However, to the extent that physical assets and intangible assets are substitutable, the equilibrium state will be reached if the marginal value of q equals unity. Hence, theoretically, in determining the equilibrium in investment, q is more fundamental than Q. However, in reality, the intangible assets are often unobservable, which greatly impairs the measurement of q. Nevertheless, if we are mainly interested in the market reaction to the firm's R&D investment, we need not be too concerned about the direct measurement of q. As in Griliches (1984), the term mq can be expressed as In qit = rrii + dt + uit, (2.3) where mi is the firm dummy, dt is the time dummy and un is the random disturbance. Also, the unobserved K, the stock of knowledge, was replaced by a distributed lag term of past R&D expenditures, i.e., K = Y ^hRt-h • Finally, approximating ln(l + ^|-) by ^ leads to the equation (See Griliches, 1984): InQ =ex\nQ_1 +m + d+ ^hh^-h + M ; (2.4) where InQ- i is added mainly to control serial correlations. It is found that InQ- i is highly significant and that current R&D is positive and sig-nificant while past R&D is not significant. It appears that past R&D is being adequately reflected in lagged values of Q. Chapter 2. Some Styhzed Facts about R&D 21 2.4 A New Investigation of R & D and Financial Markets As discussed above, Tobin's Q represents the market valuation of the firm. Essentially, the market value of the firm's capital is the value of the expected (discounted) flow of future profits, therefore uncertainty inevitably exists regarding the future cashflows. Moreover, for a levered firm, the possibility of bankruptcy would also impose a deduction in the firm's market value due to the possible outlay of bankruptcy costs. Hence, in evaluating Q, the firm's risk should be taken into account. However, in the formulation of Equation (2.4) with time series data, should the firm's risk change over time, the firm dummy would not be able to capture this effect. Accordingly, in the following, the firm dummy in Equation (2.4) is replaced by the debt/equity ratio as the proxy for the firm's financial risk. Also, <5_i is suppressed, while a first order autocorrelation among disturbances is allowed, which gives the alternative equation below In Qt = a + + 02 I n ~ + «t , (2-5) where ut = put_x + et, E{eiej) = 0, i^j, and £ ( 3 3 ) = cr2. The data used here to estimate Equation (2.5) are from Compustat Industrial Annual tape sampling period 1968 to 1987. The equation is estimated for four industries, namely, drugs (SIC codes 2830-2839), machinery (SIC codes 3520-3560), computers (SIC codes 3680-3689), and instruments (SIC codes 3810-3890). Estimation is also done for the pooled four industry data. In this case, an industry dummy is added in the equation. There are 222 firms listed in these industries; however, the requirement that each firm should have a continuous data record for at least 12 years reduces the data to a sample of 127 firms totaling 2127 observations. The Q measure is constructed according to the Chapter 2. Some Styhzed Facts about R&D 22 Table 2.1: Regression of Tobin's Q Industry A (S.E.)* ft (S.E.)* R-Square Drugs 2.45 (0.95) -0.52 (0.02) 0.93 Machinery 2.18 (0.71) -0.34 (0.01) 0.82 Computers 1.33 (0.61) -0.40 (0.01) 0.84 Instruments 1.52 (0.68) -0.44 (0.01) 0.89 Pooled 1.85 (0.35) -0.41 (0.01) 0.89 * Standard Error Tobin and Brainard (1977) procedure, and the obtained sample average of Q is 1.16. The results of the regressions are reported in Table 2.1. It is clear that both R&D expenditures and debt/equity ratios are significant for all four industries. While the coefficients of R&D are consistently positive showing that the intangible stock of knowledge does contribute to the market value of the firm, viewed from a considerably higher coefficient on R&D, the drugs industry seems to be more productive in R&D than the other three industries. Also, the coefficients of D/E are Chapter 2. Some Styhzed Facts about R&D 23 consistently negative, revealing that financial leverage has a negative effect on Q. This may reflect the costs of financial leverage, e.g., bankruptcy costs, managerial agency costs, etc., or the general risk-aversion of the markets. 2.5 Further Discussions on R & D and Marke t Structure As mentioned earlier, both theoretical and empirical work on R&D and financial mar-kets are just beginning, yet there have been many studies on R&D and output market structure. The latter subject is mainly concerned with the role of firm size or indus-trial concentration in the determination of R&D intensity. While the definitions of R&D intensity and firm size are generally common in the literature, the empirical reports regarding the relation between the two are quite inconsistent. For example, Hamberg (1964) and Comanor (1967) find that the relationship is monotonically decreasing. How-ever, Scherer (1965) reports that the relation has the shape of an inverted-U, whereas Bound et al (1984) recently found that the relation looks like an upright-U. Contrary to all, Cohen et al (1987) suggests that there is no direct effect of firm size at all. In spite of the conflicting reports, these studies have a common feature that the empir-ical tests are performed on cross-sectional data. This makes one wonder what additional information might be extracted from the time-series dimension, which motivates the fol-lowing experiments with the pooled cross-section time-series data from Compustat data tape. According to Bound et al (1984), the relationship between R&D intensity, defined as the ratio of R&D expenditures to the sales, and firm size, represented by sales, can be tested by the regression equation In R = a + A In S + /52(ln S)2 + f33 In A + £ fcA + e, z where R represents R&D expenditures, S stands for sales, A stands for gross plant—a Chapter 2. Some Stylized Facts about R&D 24 measure of the capital, Di is a set of industry dummies, and e is the random disturbance. The results of Bound et al (1984) show that /3i > 0 and /32 > 0, which means a U-shaped relationship between InR and In 5. In other words, this implies that both small firms and large firms do R&D more intensively than medium sized firms. In addition, Bound et al also find (3% > 0, and they interpret this as evidence of the complementarity between R&D intensity and capital intensity (capital intensity is denned as the ratio of capital to sales). Now, for the four industries mentioned before, we test the same relationship using the pooled cross-section time-series data. Conceivably, sales and capital generally grow together over time for individual firms. Hence, to avoid coUinearity between S and A, we will drop A from the equation (this collinearity may not be present in cross-sectional data because different firms may use technologies requiring different capital intensity). Furthermore, we will examine the R&D-intensity-firm-size relationship within each of the four industries, thus, the industry dummy is not needed. However, to capture possible permanent firm-specific factors cross-sectionally, a set of firm dummies is added to the equation, which is to be compared with the equation without the firm dummies. Specifi-cally, the following two alternative regression equations will be estimated and compared: ]nR = a + p1]nS+p2QnS)2 + e, (2.6) In R = a + ft In S + 32{ln S)2 + £ 7 iD { + e, (2.7) i where Di is a set of firm dummies. The results of the estimation for the four industries are. reported in Tables 2.2 and 2.3. For the first regression, since the logarithms of the R&D intensity can be expressed as R In — = In R — In 5 , Chapter 2. Some Stylized Facts about R&D 25 Table 2.2: R&D and Firm Size, I In R = a + 0x In S + /32(ln Sf + e Industry Pi (S.E.)* 02 (S.E.)* R-Square Drugs 0.94 (0.12) 0.008 (0.010) 0.87 Machinery 0.94 (0.13) 0.007 (0.011) 0.79 Computers 0.95 (0.07) 0.005 (0.006) 0.93 Instruments 0.67 (0.07) 0.034 (0.007) 0.89 * Standard Error Chapter 2. Some Styhzed Facts about R&D 26 and since the joint hypothesis H0:81 = l and fa = 0 will be rejected for instruments industry but cannot be rejected for the other three at the 5% confidence level, it appears that we can be confident that size has an effect upon R&D intensity for the firms in instruments industry, but not for the firms in the other three industries. Yet, the inclusion of firm dummies drastically alters the results as revealed in Table 2.3. Now, except for drugs industry, the joint hypothesis H0 : Pi = 1 and 02 = 0 will be rejected, suggesting that when inter-firm differences are accounted for by the dummy variables, the effect of size becomes important in explaining firms' time path of R&D intensity. No doubt, in both Tables 2.2 and 2.3, the most conspicuous case is instruments industry, which reveals a strong U-shaped relation between R&D intensity and firm size. Specifically, for a particular firm in instruments industry, its R&D intensity first decreases and then increases as it grows from small to large. 2.6 Concluding Remarks This essay reviews the empirical studies on R&D-related issues and presents the major styhzed facts found in the literature. There is indeed a wide range of economic issues, upon which empirical studies regarding R&D have been performed. These studies reveal that R&D has had a profound impact on various economic activities. As is often the case in empirical work, different formulations or different data sets may give rise to different results. In this essay, some issues related to R&D, firm size, Chapter 2. Some Stylized Facts about R&D 27 Table 2.3: R&D and Firm Size, II In R = a + pi In S + /32(ln Sf + 7 i D i + e Industry Pi (S.E.)* P2 (S.E.)* R-Square Drugs 1.06 (0.14) 0.002 (0.011) 0.97 Machinery 1.46 (0.17) -0.051 (0.016) 0.95 Computers 0.91 (0.09) 0.013 (0.008) 0.98 Instruments 0.52 (0.06) 0.055 (0.007) 0.97 * Standard Error Chapter 2. Some Stylized Facts about R&D 28 and financial markets are re-examined using a pooled cross-section time-series data set adopted from the Compustat data tape and some interesting results are obtained. With the availability of patent data, in addition to the cumulated R&D data, the recent surge of empirical studies has been very fruitful. Due to the complexity of inno-vative activities, the current findings in this field are still mostly preliminary. Although the studies on the relationship between R&D and market structure seem to be largely guided by a few theories, many stylized facts found in the rest of the literature are ba-sically descriptive. For example, it is reported by several authors that the productivity of R&D as measured by patent per R&D dollar or patent per scientist and engineer has been declining for most industries in most countries. The cause of this negative trend is, at present, undetermined. Also, it is found that basic research and long term R&D projects contribute more to productivity growth than applied research and short term R&D projects. Government financed R&D is mostly non-basic research and short term, yet the impact of the reported fluctuations of government financed R&D on economic growth is by no means clear. Although the expanding literature on R&D has provided us with a large collection of stylized facts, it appears that the field of R&D economics in general still calls for more intensive work. In the existing literature, one frequently encounters statements, describ-ing work in an area, as "helping fill the gap" (Mansfield, 1980, 1984, for example). This clearly reveals that the intensity of study in the related area has been quite inadequate. Perhaps the most striking feature of the existing literature on R&D is the lack of analysis of two aspects. The studies reported in the rest of this thesis have been motivated by these missing analyses. Firstly, most existing studies on R&D ignore the financial aspect of R&D projects. R&D projects facing economic agents, such as firms in the industries, are likely to be selected by some investment criterion. For example, if a project does not yield a positive Chapter 2. Some Styhzed Facts about R&D 29 return, possible government subsidy is accounted for, the project may well be put aside. Furthermore, for some risky R&D projects, substantial economic benefits may only be realized long after the initial investment. For a firm facing such projects, the possibility of bankruptcy should be seriously considered before deciding to take the long term, risky projects of R&D. Therefore, it is clear that innovative activities should not be separated from investment behaviour in the economic study of R&D. In Chapter 3 of this thesis, an integrated model is set up to examine the interrelation between firms' R&D investment decisions and financial decisions. Indeed, the analysis in Chapter 3 illustrates that finance is an important factor influencing firms' decisions about R&D activities. Secondly, the existing literature lacks a macroeconomic analysis of R&D. Although, as reviewed in this chapter, many authors report on the causal link between R&D invest-ment and productivity growth, both at firm level and industrial level for manufacturings, the overall economic consequences of this productivity growth have not been examined. This is a significant omission. Since R&D is an innovative activity, often subsidized by government, the issue of desired level of R&D is naturally related to government policy. Clearly, firm level or industrial level analyses are not adequate for guiding government policy. Motivation for the analysis carried out in Chapter 4 of this thesis is to provide an economy level study which is of practical importance in guiding government policy. Chapter 4 constructs a general equilibrium model which establishes structural links be-tween manufacturings and other sectors in the economy and attempts to predict the overall impact of productivity growth in the manufacturings. The analysis shows that the marked boom in the service sector of most industrialized economies can be explained, at least partially, by the productivity growth in the manufacturings. On the whole, economic studies of R&D have progressed rapidly. That we find more questions than answers from the existing literature is no reason for being pessimistic. Chapter 2. Some Stylized Facts about R&D 30 With continuing accumulation of raw data and ever improving econometric techniques, future studies of R&D will surely be rewarding, on R&D will surely be rewarding. Chapter 2. Some Styhzed Facts about R&D 2.7 A p p e n d i x to Chapter 2 Summary of M a j o r Stylized Facts Relation Reference Styhzed Facts I. Patents [4] [57] and R&D *There is a statistically signifi-cant relationship between R&D and patent application. The relation is strong in cross-section, weak but still significant in time-series. *Cross-sectionally, small firms tend to patent more per R&D dollar then large firms, and larger firms have a nearly constant ratio of patenting to R&D. *Except drug firms, number of patent per R&D dollar consistently decreases in time. II. Composition of R&D and Productivity [46] *Composition of R&D (i.e. basic re-search, long-term, ambitious, risky etc.) differs substantially inter-er 2. Some Stylized Facts about R&D Growth industry and interfirm. * Percentage of basic research and long term project increases with firm size, interfirm. *Percentage of basic research and long term project decrease with 4-firm concentration, interindustry [45] [21] *R&D contributes significantly to productivity growth. *Basic research component does so even more strongly. *Privately financed R&D expenditure have a significanly larger effect on private productivity than feder-ally financed R&D. *Many industries have changed R&D components recently, resulting in cuts in the proportion of basic research. Chapter 2. Some Styhzed Facts about R&D III. Government [21] *U.S. total R&D expenditure in indus-Financed try peaked around 1968, dropped in R&D early 1970's and recovered somewhat in late 1970's. *As percentage of sales, privately financed R&D kept constant, while federally financed R&D fluctuated. [47] * Government financed R&D is slightly complementary to private R&D. IV. Rate of Return [53] *From a model of production struc-on R&D and ture, derive the demand for factor Market Value input and estimate the rate of re-of Firms turn on quasi-fixed factors for manufacturing industry in three countries. *Rate of return on R&D ranges 10-15%, higher than the rate of return on conventional capital. [58] *Taking account of rate of obsoles-Chapter 2. Some Stylized Facts about R&D cence, research gestation lag con-siderably reduces the private rate of return on R&D. *(ROR on R&D) < 2*(R0R on capital) [20] [55] *R&D investment create "intangible" capital, so should show up in the valuation of the firm by market. ^Unanticipated change in R&D expen-diture and patenting are signifi-cantly correlated with change in firm's market value. [41] *Theory predicts that agency costs of debt increase with amount of intangible assets. *Using R&D as a proxy for intangible assets, find that firms' financial leverage is negative^' related to R&D expenditure. V. Market [13] [23] *R&D intensity decreases weakly in er 2. Some Stylized Facts about R&D Structure and R&D [61] [4] [60] [12] [62] [63] [38] [66] firm size. *R&D intensity and firm size have an inverted-U relationship for most industries except chemicals. *R&D intensity and firm size have an U-shaped relationship. * Standing alone, firm size has some effects on R&D intensity. * After controlling for technological opportunity, appropriability etc. firm size is no longer significant. *R&D intensity and industrial con-centration have an inverted-U re-lationship. * After controlling for industrial and firm difference in technological opportunity etc. the seller con-centration ratio does not remain significant. Chapter 2. Some Stylized Facts about R&D VI. Evolution [19] [64] of Technology *Declining technological opportu-nity over the industry life cycle contribute to a decline in innova-tive activity (measured by patent rate). (Confirming Kuznets view) *The demand-pull hypothesis of inno-vation is also substantial. (Con-firming Schmookler view) *The net effect on the time path of R&D expenditure cannot be specified a priori, which is the largely off-setting results of above two factors. [26] * Supply side effects on the rate and direction of R&D, i.e. technological opportunity and R&D spillovers. *R&D productivity is increased by the R&D of "technological neighbour". ^Neighbour's R&D lowers the profits and market value of the low-R&D-Chapter 2. Some Styhzed Facts about R&D intensive firms. * Firms are shown to adjust the tech-nological composition in response to technological opportunity. *The importance of U.S. in world patenting declines and the impor-tance of Japan, W. Germany, and planned economies rises, revealing changing patterns in comparative advantage. *High degree of foreign patenting in most of the world's economies, (e.g. 30-40% in U.S.) *The ratio of patents granted abroad to patents granted home (a measure of adaptiveness in the nature of R&D) of industrialized countries is 10 folds higher than developing economies, revealing that the latter is mainly adapting rather than in-novating. VII. International [17] Invention, R&D and Foreign Trade Chapter 2. Some Styhzed Facts about R&D 38 *Patents per scientist and engineer and patents per R&D dollar decline for almost all the industrialized countries, revealing exhausting of technological opportunity? *Total R&D/GDP declines for most industrialized countries. [50] *Percentage of R&D carried out over-seas is significantly related to the percentage of sales from foreign sub-sidiaries, but significantly nega-tively related to percentage of firm's sales from exports. [48] * Firms having larger percentage of foreign sales count more on foreign revenue to returns on R&D project. *R&D intensive firms usually count a higher percentage on foreign sales to returns on R&D. Chapter 2. Some Stylized Facts about R&D *So, foreign sales stimulate domestic * Amount of product R&D is much more significantly related to the foreign sales than process R&D. determining exports using U.K. data. *Both the level of R&D (measured in annual expenditure) and the gap in R&D (the difference between U.K. and foreign levels of R&D expenditure) have positive effects on exports. *There is a simultaneous relation bet-ween R&D and exports. R&D. [25] Test the role of R&D expenditure in VIII. R&D Strategy [49] On average, the ratio of imitation Innovative costs to innovation costs = .65 v.s. and of imitation time to innovation Imitative time = .70. [49] *Patented innovation cannot elimi-Chapter 2. Some Styhzed Facts about R&D nate imitation, but it increases imitation costs. * Four-firm concentration ratio is significantly related to the easi-ness of imitation. (4-firm concen-tration ratio is explained by the ratio of imitation cost to innova-tion cost alone at R-square = .60) [40] *Test the relation Strategy[l=nnnovate 0=imitate] =F(Size,MarketShare,Concentration) *Size and MarketShare are signifi-cantly positive. ^Concentration is negative but not significant. IX. R&D and *In Oct. 1974, FASB issued FAS2, Accounting which mandates that all R&D be Standard treated as expense, while before that many firms capitalized and amortized R&D costs. Chapter 2. Some Stylized Facts about R&D 41 *Dukes et al report that there was no change in firms' R&D outlay as affected by FAS2. [24] *Horwitz et al report that there was a cut in R&D outlay for small sized and high-tech firms after FAS2. [16] Chapter 3 R & D Investment, Technology and Financial Structure 3.1 Introduction There has been a rapidly expanding literature dealing with firms' R&D activity. In most of the literature, R&D projects are characterized as "high risk projects" and the riskiness of an R&D project is commonly modeled by the random outcome of winning or losing an R&D "race". The goal of such a race is to achieve a particular patent or, more abstractly, to gain a prize which only rewards the winner of the race (e.g. Grossman and Shapiro, 1986, Klette and de Meza, 1986). In general, such models are somewhat like a cost/benefit analysis for a specific project as they focus on the relationship between the expenditure pattern and the payoffs of an R&D project. On the other hand, there are only a few studies discussing the relationship between the financial decisions and R&D investment decisions of a firm. According to Myers (1977), any firm can be evaluated based on its real assets and its growth opportunities for future profits. The growth opportunities are called "intangible assets", interpreted as having no liquidation value. Myers (1977) points out that the shareholders of a financially levered firm will have incentive to maximize equity value, rather than firm value, by making suboptimal investment decisions. The loss due to suboptimal investment, termed agency cost of debt, varies to the extent that the debtholders can monitor the investment decisions of the firm. Since it is more difficult to monitor the investment made in growth opportunities, a firm with a high level of intangible assets will not be able to support a 42 Chapter 3. R&D Investment, Technology and Financial Structure 43 high level of debt. Examining the importance of agency costs in the determination of corporate debt, Long and Malitz (1985) used R&D expenditure and advertising expenses as the proxies for the value of a firm's intangible assets. Indeed, they found a general negative correlation between a firm's R&D expenditure and the firm's debt level. In a recent study of the determinants of capital structure, Titman and Wessels (1988) identified eight attributes which affect the financial decisions of a firm. In addition to growth opportunities, R&D contributes to two other attributes, namely, non-debt tax shields and uniqueness of the firm's products. On one hand, since R&D can be expensed before tax, it creates non-debt tax shields, which substitute for the tax shields of debt. On the other hand, a firm's R&D enhances the uniqueness of the firm's product, which makes bankruptcy more costly to the firm's employees and customers (See also Titman, 1984). Therefore, all these three attributes, which are positively loaded with R&D, are found to bear negative correlation to firms' debt levels. Although the above studies offer various explanations of the negative correlation between firms' financial leverage and firms' R&D investment, they do not address the question of how firms decide how much to invest in R&D since firms' R&D investment is treated as exogenously determined. This essay differs from previous studies in that both R&D investment decisions and financial decisions are considered simultaneously. The motivation for this essay is based on the belief that the uncertainty associated with the duration and the payout of most R&D projects is larger than the uncertainty of other types of business investment. Before a firm decides to embark on a long term R&D project, the firm must carefully assess its financial security in the short term. Inversely, when the firm makes its financial decisions, the outcome of its R&D investment should also be taken into account. Therefore, in this essay, the firm's decisions concerning the use of funds (R&D expenditure) and concerning the source of funds (capital structure) will be considered as intrinsically related. Chapter 3. R&D Investment, Technology and Financial Structure 44 Kim and Maksimovic (1987) have derived an interesting link between R&D and debt. They work with a production function of two factor inputs, namely, capital and labour, and assume R&D activity as capital augmenting technical progress. They find that the marginal value of capital augmenting R&D is larger to the shareholders of an all equity firm than to those of a levered firm because the shareholders of the unlevered firm will find capital augmenting valuable in more states than the shareholders of the levered firm do. Their findings offer an alternative interpretation of the negative correlation between firms' R&D expenditure and debt level. However, in this essay, we shall not restrict R&D to be a factor of production serving as capital augmenting (or labour augmenting). Instead, R&D as a driving force to the general technological progress is considered as a strategic investment for the future. Accordingly, this essay avoids the assumption of a specific production function and does not require R&D to be concurrently connected with production. Yet, firms' R&D decisions are still related to their decisions on the optimal debt level. As we are looking into the capital structure decisions, we cannot avoid asking why firms use particular amounts of debt. There have been many answers to this question. Modigliani and Miller (1958) first proved that under certain ideal circumstances, capital structure does not matter at all. Other answers have associated benefits and costs to debt such that an interior solution would exist for the problem of seeking an optimal debt level. These answers have highlighted tax shield effects of debt, direct or indirect bankruptcy costs, agency costs of debt and agency costs of equity, etc. (Modigliani and Miller, 1958, 1963, Kraus and Litzenberger, 1973, Jensen and Meckling, 1976, Williams, 1987, and Titman and Wessels, 1988, among others). Admittedly, the question of optimal capital structure is itself the central theme of a branch of financial literature and is still unsettled; one cannot be sure which particular answer would be the ultimate or even cor-rect answer. However, since our primary concern in this essay is the interaction between Chapter 3. R&D Investment, Technology and Financial Structure 45 R&D investment and capital structure, rather than the pure issue of the choice between debt and equity, we select theories of managerial agency costs (Jensen and Meckling, 1976) to represent the basic trade-off in financial decisions in our problem. Embodying any other theories (e.g. tax shields v.s. bankruptcy costs) would not materially change our results. In our model, a manager makes all the investment and financial decisions for a firm. The perquisites the manager extracts from the job in the firm constitute the agency cost to the shareholders of the firm; this cost can be reduced by debt financing. We find that when the manager's utility is based on both the perquisites of the job and the monetary income from the firm's residual profits, it may be to the advantage of both the manager and the shareholders of the firm to employ debt. However, indirect bankruptcy costs will discourage excessive use of debt. As such we lay down the basic financial structure of the firm on which we then build up the consideration of R&D investment and obtain an integrated model. When we combine financial decisions with R&D investment decisions, we find that the state of technology plays an important role. Naturally, the subjects of R&D and technology are always associated for the simple reason that R&D is a powerful impetus to advancing technology. Hence, the most common link relating R&D and technology is in a form of R&D production function (Griliches, 1986, Mansfield, 1980, Pakes and Griliches, 1984). However, in this essay, we will be emphasizing a reverse link, namely, the impact of firms' state of technology upon firms' R&D expenditure. The mechanism of this reverse link can be best explained from a financial point of view. When a financially leveraged firm makes its R&D investment decisions, it must carefully compute its risk of bankruptcy. If bankruptcy occurs before the R&D project can render any benefit, the investment will be a sheer loss. Hence, the firm's R&D expenditure is constrained by its ability to repay its debt obligation. When the firm possesses a superior technology, Chapter 3. R&D Investment, Technology and Financial Structure 46 it can naturally expect a higher revenue as the technology is applied to production. Consequently, the firm's financial ground will become more secure and more capable of supporting R&D expenditure to exploit future potential. The above argument reveals a dependence of R&D investment upon the state of technology. This sort of dependence is tested based on U.S. industrial data and the results are encouraging. Although this essay treats R&D decisions and financial decisions as interrelated, it does not involve the structure of output market. This omission is a sacrifice to the desire to keep the exposition of the relation between R&D decisions and financial decisions as simple as possible. In the literature, some analysis of the relations between financial structure and output market structure has already been provided. Brander and Lewis (1986, 1988) have offered an analysis on the relationship between firms' financial decisions and output decisions, and Brander and Spencer (1989) have also established a model incorporating the effects of firms' ownership on the relations between financial structure and output market structure. As to the analysis of the relation between R&D and the output market, it is by no means scarce in the literature (See for example Loury, 1979, Lee and Wilde, 1980, Dasgupta and Stiglitz, 1980a, 1980b, and Brander and Spencer, 1983). The remainder of the essay is organized into four sections. Section 2 presents a model incorporating both the R&D investment decision and the financial decision; Section 3 discusses the solution to the model; Section 4 is devoted to the testing of empirical implications and Section 5 gives a brief conclusion. Chapter 3. R&D Investment, Technology and Financial Structure 47 3.2 The Model 3.2.1 R&D and Technological Progress Consider technological progress of the following type: while the production cost remains constant in a certain industry, the quality of the product of that industry improves. This type of technology can be represented by a quality index of the product, Q. Suppose the rate of technological progress is p, i.e., if the best product of the industry in period t has a quality index of Qt, the best product in period i + 1 will have a quality index of Qt+i = (1 + / - O Q t - Thus, the technology of the industry is advancing exponentially. Suppose the diffusion of technology takes one period of time, hence the best technology used by any firm in the industry in period i will be known to all of the firms in the same industry in period i + 1. In other words, every firm in the industry will be able to produce products of quality Qt in period i + 1. However, any firm can invest in an R&D project in period i to improve future technology. If the project is successful, the firm will have a technology which can produce a product of quality Qt+i = (1 + p)Qt hi the next period. If the project fails, the firm can only produce a product of quality Qt in period i + 1. Similarly, in period i + 2, most firms will select the best technology of period i + 1 and produce goods of quality Qt+i • Only those firms whose R&D projects succeeded in i + 1 can enjoy producing goods of a better quality, i.e., Qt+2 — (1 + ^ Qt+i- As a result, in any period, there are always two technologies available in the industry, one is the old technology known to all firms, the other is the improved new technology known only to those firms with successful R&D projects in the last period. When the new technology becomes known to all of the firms in the next period, the new technology then becomes old. This technological progress is an ongoing process. On the other hand, suppose the relative prices accurately reflect the quality changes as the technology progresses such that in any period, the equilibrium price of goods Chapter 3. R&D Investment, Technology and Financial Structure 48 produced by the new technology is (l+p)p(9). The goods produced by the old technology, however, will be priced at p(8), where 8 is the state of nature representing general demand condition. 3.2.2 Financia l Structure of the F i r m Consider a firm having a fixed capacity of producing q units of output each period with constant production cost. Let cq denote this production cost and I denote the R&D expenditure, the total investment in period t is then cq + L t • Suppose the firm is controlled by a manager who makes the decisions on the level of R&D investment and the financial structure of the firm. The amount of debt the manager decides to borrow to finance the investment is determined at the beginning of each period and is denoted by B; the rest of the investment, cq + It — B, will be financed by equity either through using retained earnings or issuing new shares. The total amount of equity capital is denoted by S. At the end of each period, the firm obtains sales revenue { (1 + p)qp{0), if the firm possesses the new technology, qp(9), if the firm has only old technology and repays debt obligation which has a face value of D. . If R > D, after repaying debt, the manager claims some administrative costs and then distributes the rest of the revenue to the shareholders of the firm. However, if R < D, the firm will declare bankruptcy. As a result, the debtholders will take over the firm, taking all the revenue and firing the manager. There are no tax or direct bankruptcy costs. Chapter 3. R&D Investment, Technology and Financial Structure 49 3.2.3 Manager's Utility Function and Agency Costs The question of why firms use particular amounts of debt has long been discussed but is not yet settled. Since the main interest of this essay is the correlation between debt and R&D expenditure, rather than the prediction of the absolute level of debt, we will not try to give a comprehensive discussion of the determinants of firms' capital structure. Instead, a simple model of managerial agency costs is used to insure that an interior solution to optimal capital structure exists. Given the existence of the optimal debt level, the interdependencies of R&D, technology, and capital structure then can be analyzed in a more general framework, in which the role of agency costs becomes inessential as far as we can agree that there are both costs and benefits associated with debt financing. In our specific model of agency costs, the basic costs of debt financing are the indirect bankruptcy costs consisting of foregone investment, while the basic benefit is the reduc-tion in managerial agency costs. In particular, the agency costs take the form of the perquisites to the manager, which reduce distributable income to the shareholders of the firm. Admittedly, for large firms, the extent of extravagance of the management has little consequence on shareholders' income and hence the costs of management perquisites may be negligible. However, for small firms, this sort of agency cost can be important and pertinent to their financial decisions. Therefore, it is reasonable to assume that the fol-lowing analysis regarding agency costs is specific to small business. Yet as noted before, once the existence of an interior solution to optimal capital structure is established, the analysis of general interrelations between R&D investment, capital structure, and state of technology would not be limited by firm size. Bearing this in mind, we now proceed with the formulation of our model of agency costs. Suppose the limited foresight of the manager allows him to be concerned about his utility in only two periods, namely, in the current one and the next one. The manager's Chapter 3. R&D Investment, Technology and Financial Structure 50 utility function is U{NX, M l 9 N2, M2) = U(NU Mx) + r^-U(N2, M2), (3.1) where Ni is the nonpecuniary compensation of the job, and Mi the monetary income, in period i. U(-, •) is assumed to be concave and 17(0,0) = 0. (3.2) The term nonpecuniary compensation refers to the firm's expenses used to provide the manager with a comfortable office environment, business travelling, and other perquisites of the job. However, the satisfaction derived from the pride of being a manager, or the feelings of being important and decisive do not count in the utility function (3.1) which is based on material consumption. Therefore, iV can be claimed as administrative costs against the revenue after debt repayment. Accordingly, the distributable equity income is R{9)-D-N, and the rate of return on equity is P = m - S D - N - 1 . (3.3) Suppose the manager's monetary income depends on the above rate of return (this will be the case if the manager's salary is related to his performance or the manager himself holds some shares of the firm), i.e., M = M(p) with M'(•) > 0. For simplicity, assume M"(-) = 0 Chapter 3. R&D Investment, Technology and Financial Structure 51 and M ( - l ) = 0. Hence, the manager will get zero income in the face of bankruptcy. In addition, if the firm becomes bankrupt in period 1, the manager will have trouble finding a job in period 2 due to his loss of reputation in the job market, therefore we assume his income in period 2 will also be zero. The distribution of R(8) is known to both the manager and the debt-holders. Ac-cordingly, the firm's debt with face value D can be fairly priced as B(D). Also, the firm's investment decisions and financial decisions are made simultaneously; the manager can-not change the firm's investment project once the debt has been issued. Thus, in this model, there is no conflict of interest between debtholders and shareholders. However, there is an agency cost of equity, which takes the form of nonpecuniary compensation to the manager. Here, it should be noted that a part of the managerial expenses is essential to the proper management of the firm, and we assume that this part is already included in the production costs. Therefore, N only represents the expenses for the unnecessary, luxury items which in principle could be avoided, and is a pure cost to the shareholders. Now, we prove that the firm can reduce this agency cost by using debt financing. Consider the following one period utility of the manager given D and S, (3.4) Without loss of generality, assume R'(-) > 0. The manager chooses N based on the solution to the problem V* = max V(N,8), which is characterized by the first order condition dU dU dM 8N + dM dN = o, if e > a Chapter 3. R&D Investment, Technology and Financial Structure 52 or dU if 6 > a (3.5) dN dM S The first term on the left of (3.5) is the marginal utility of increasing nonpecuniary compensation and the second term represents the marginal utility of monetary income which is lost when the nonpecuniary compensation is increased. Hence, (3.5) determines the optimal amount of N for the manager in nonbankrupt states. Clearly, in bankrupt states, J V is identically zero. For nonbankrupt states, i.e., when 6 > a, the second order condition NN in" ML 4- II" 'NM g is always satisfied if the usual assumptions uNN<o, KM>o MM (3.6) ^MM < 0 are imposed. Differentiating the first order condition in (3.5) with respect to D gives dN* •TJ" _ 77" 3D M'p I dN' R- D - N* OS NM^~ V~dD + 1 + S dD » ( M ' P 'dN' , R-D + 1 + — 5 J V * os 3D + dU M'p dS dM S2 dD = 0. which leads to dN' ~dD 1 dU M'p dS ~ A dM S2 dD 1 f R-D-N*dS' + A 1 + — W M' TJ" P _ TJ" UNM g u™ 'M'p MM where A is the expression defined in (3.6). If the magnitudes of U^M and UMM are small, in other words, the marginal utility of money is diminishing slowly and is not much affected by the nonpecuniary compensation, then the second term on the right hand side of the above equation can be ignored. Henceforth, we will have dN" dD <0, Chapter 3. R&D Investment, Technology and Financial Structure 53 which demonstrates that debt financing can reduce the agency cost. Intuitively, when the firm increases debt, the rate of return on equity will get more leverage, which incidently enlarges the manager's marginal rate of substitution between his nonpecuniary compensation and monetary income. In other words, any saving on managerial expenses would contribute more to shareholders' rate of return for a lever-aged firm than for an unleveraged firm due to leverage. Since the manager's monetary income is a function of the rate of return on equity, higher financial leverage will make higher monetary payoff for the manager at the same amount of sacrifice of nonpecu-niary compensation. Consequently, the manager will be induced to move part of the firm's residual profit (profit after debt repayment) from his nonpecuniary compensation to the distributable equity income for the shareholders. However, it should be noted that although shareholders realize the merits of debt financing, they have no direct control over the firm's financial structure. What is important is the manager's personal attitude toward debt financing, to which we turn next. From Equation (3.5) and the budget constraint S + B(D) = cq + 1, we see that TV and M are functions of R, D, and I. Hence, the manager's utility function (3.1) can be expressed in the form of an indirect utility function V(D,, h, R1{61)) + -L-V{D2, h, R2{02)) (3.7) I + 77 with the properties: VJ < 0, VJt> 0, and -j^EV > 0. (3.8) Also, following (3.2), V = 0, i£R<D. (3.9) Chapter 3. R&D Investment, Technology and Financial Structure 54 Vj < 0 states that R&D investment is a pure cost in the current period since any benefit of R&D investment can only be realized in the next period. > 0 simply says that high revenue will bring high income, and -§j=jEV > 0 implies that debt financing has some advantages even from the manager's own point of view, which we will prove below. Following (3.4), the expected one period utility given D and S is EV(N,6) = I" U{0,0)f{6)d6 Jo L 'OO + I u N,M W)-D-N_; f{6)de The comparative static effect of V* with respect to D is then Using(3.5), _d_ 3D Am too EV = U(0,0)f(a)— + du + 8MM> dN dM S 1 _ R{6) -D- N*(6) dS S S2 dD HD f{e)de tf(o,o)/(«)£. d r — EV = - / dD L dU M'p dM S R{6) — D — N*{9) dS dD + 1 f(e)de. From the budget constraint S + B — cq + I, S = cq + I — B . So and dS_ dD dS_ dB dS_dB_ ~dB~dD ~~ - 1 dB dD 1+r Chapter 3. R&D Investment, Technology and Financial Structure 55 where r is the marginal interest rate of debt which depends on D but is determined before 6 is revealed. Thus dD EV 1 /°° dU M'p dM~S~ Since — f 1 +r r Ja R(6) -D- N*(8) (1+r) f(8)dd. R(6) - D - N*(6) f(6)d6 is the expected rate of return on equity, which should be higher than the expected rate of return on debt, we have 'R(0) - D - N"(8) f Ja (1+r) f(6)dd > 0. Notice that ^ > 0, M'p > 0, if a stronger condition dU , M dM p R(8) — D — N*(8) (1+r) f(6)d8 > 0 is imposed, we will have dD EV > 0, (3.10) (3.11) which is to be demonstrated. The intuition behind (3.10) can be explained as follows. The manager decides on his desired level of nonpecuniary compensation such that he is indifferent to the choice of spending one more dollar on his job or receiving additional monetary income attributable by distributing that dollar to shareholders and thereby raising the rate of return on equity. If the firm increases debt, the manager's marginal utility of nonpecuniary income will not be affected. But the marginal utility of monetary income resulting from a marginal deduction in managerial expenses will be enlarged > because financial leverage raises the rate of return on equity. Hence, the manager will gain total utility if the firm issues some debt and he shifts some of his expenses on the job to the distributable equity income so Chapter 3. R&D Investment, Technology and Financial Structure 56 as to obtain a higher monetary payoff. However, in some bad states, increasing the firm's debt may actually hurt the manager if the resulting rate of return on equity is lower than the interest rate for debt. Over all, the manager's expected utility may still increase as stated in (3.11), thus favoring financial leverage. Previously, we proved that debt financing was beneficial to the shareholders for it reduces the agency cost. Now we have shown that the manager himself will also be interested in issuing debt. However, it should be noted that the above proofs do not lead to the conclusion that the firm will be fully financed by debt. (3.11) is stated in terms of one period utility while the manager will decide on the amount of debt so as to maximize his two period utility. In fact, as we shall see later, the fear of losing his job in the second period in the face of bankruptcy will prevent the manager from using too much debt in the first period. In summary, this section analyzes the impact of financial leverage upon the agency costs of equity. Conceivably, in the real world, there are many other factors affecting managerial decisions. For example, if there is a market for corporate control, the man-ager may exercise extra caution in making his decisions for fear of takeover. In this case, the manager may unwillingly reduce his consumption of perquisites in order to improve managerial efficiency. This implies that, in reality, the manager's choice of the opti-mal level of perquisites and debt is far more complicated than presented in this model. However, to incorporate all the real-world factors in one model would make the model unmanageable. Besides, the primary interest in this essay is not the pure issue of corpo-rate finance. Therefore, we abstract from other mechanisms of discipline influencing the manager and concentrate on the interaction between firms' financial decisions and R&D investment decisions. Chapter 3. R&D Investment, Technology and Financial Structure 57 3.2.4 The Optimization Problem Suppose the manager chooses Di, I\, conditional on today's technology so as to maximize while in the next period he will choose D2, h, conditional on the future technology. Given that the relevant state variable, whether or not the manager possesses the new technology, takes on only two values, there will be at most two optimal strategies corresponding to the two states, namely, (-D,^) a n d {D_,L)- In any period t, if the manager possesses the new technology, he chooses D and / , otherwise he chooses D_ and / . Thus, in seeking the optimal strategy of period 1, (Di,Ii), the optimal strategy of period 2, (D2,I2), can be thought of as exogenous, depending only upon the technology in that period. Suppose the probability of obtaining new technology in period 2, IT, depends on the R&D expenditure in period 1, with the property his expected two period utility (3.12) TT'(I) > 0, TT"(I) < 0 (3.13) and 7r'(J) is bounded. The manager's expected utility is then v(D2, i2, R2(e2)) / ( W ^ ) ^ A . Using (3.9), Chapter 3. R&D Investment, Technology and Financial Structure 58 (3.14) where R{8) = (1 + p)qp(6), R(8) = qp(6), R(a) = D , R(a) = D,and EV = EV = J~v(D,I,R{8))f(e)de, /OO V(D,I,R(0))f(O)de. (3.15) Having derived the expected utility of the manager, we can finally formulate the manager's optimization problem as It is clear that, according to (3.14), the manager's expected two period utility depends on the two choice variables, the amount of debt, D, and the amount of R&D investment, I. The investment I constitutes a cost at the beginning, lowering the expected utility in the first period, but may bring higher return later, raising expected utility in the second period. Conversely, the debt D enhances the expected utility in the first period at the cost of an increasing possibility of bankruptcy later. The question of why debt is beneficial (in the first period) has been a subject of continuous dispute. The model presented in this section offers an explanation which may pertain to only small firms. However, whatever the real explanation is, we do observe large firms issuing debt. This indicates that financial leverage indeed has some merits for both small and large firms. Upon the completion of the description of our model, one may ask the following question: what distinguishes R&D from other investment projects? From the social point of view, output of R&D activity bears characteristics of public goods. Therefore, R&D- is distinctively associated with issues of appropriability, monopoly power, patent race, preferrable taxation, etc. However, from the financial point of view, any distinction max $(D,J) . D,I (3.16) Chapter 3. R&D Investment, Technology and Financial Structure 59 between R&D and other investment projects should be boiled down to the pattern of cashflow brought about to the firm. In terms of cashflow, R&D projects are more uncertain than other types of invest-ment. Firstly, the variance in the payoff from an R&D project is relatively large because outcomes from experimental activities are generally less predictable. Secondly, the pay-off horizon from R&D is relatively long and has large variance because the completion date of an R&D project is usually unforeseeable. The nature of R&D presented in our model captures these two aspects of high uncertainty. Specifically, the payout of R&D in our model is either zero or a specific amount (which may be fairly large) with non-trivial probability assigned to zero payout. Furthermore, the time when any firm's R&D project succeeds is random, and the earliest possible time will be in the next period, with nontrivial probability assigned to the case where the firm does not succeed after a large number of periods. Admittedly, our model does not distinguish R&D from other investment projects if they are characterized by the same pattern of cashflow. Therefore, the analysis developed in this essay can be extended to other investment projects provided that the cashflow of these projects is associated with high uncertainty both in cashflow size and in payoff horizon. Similarly, from the financial point of view, our previous assumption regarding the product quality improvement nature of R&D is also nonessential. In the case of cost reduction R&D, which associates new technology with lower production costs, our analysis will still be valid as long as the R&D project can yield positive profits to the firm. Chapter 3. R&D Investment, Technology and Financial Structure 60 3.3 Solution of the Mode l 3.3.1 The Conditions of Optimization In the preceding section, we have set the manager's objective function, i.e., the com-bined expected two period utility function. The manager chooses the optimal level of R&D investment and the amount of debt financing simultaneously so as to maximize his expected utility as stated in problem (3.16). The first order conditions of this problem are ^=fv^.f(9)de-^^{^EV + (l-^EV}=0, (3.17) and <9$ t00 1 - F(a) r — i V;.f(6)d6+ l + Wir'I[EV-EY]=0. (3.18) The first term in the right-hand side of Equation (3.17) is the marginal reduction in agency costs achieved by debt financing, whereas the second term is the marginal loss of utility in the second period due to possible bankruptcy in the first period. If the firm becomes bankrupt in the first period, the manager will be jobless in the second period; it is this threatening situation that prevents the manager from using too much debt even though there are no direct bankruptcy costs to the firm. Since TV'J > 0, the magnitude of the possible loss increases with J; hence the optimal debt must be negatively related to I to keep Equation (3.17) satisfied. Intuitively, when the firm invests more in R&D, the expected future income will increase. Accordingly, the firm will be more reluctant to take debt obligations which mature before the realization of future income. It is also interesting to look at Equation (3.18), which is basically the balancing of costs and benefits of an R&D project. While the two terms in the right-hand side of (3.18) have opposite signs, namely, f Ja v;-f{6)de < o, Chapter 3. R&D Investment, Technology and Financial Structure 61 it can be shown that and That implies ^m*>[EV-EY\>0.. > 0 - , d [ l - F ( a ) dq\ 1+7] 7 F / EV - EV] 1 > 0 . hence, there may exist a q such that when q < q, This reveals the importance of firm size in R&D activity. Intuitively, since the reward for a successful R&D project takes the form of enjoying a higher selling price for the firm's product, if the size of the firm is too small, the realized benefit of obtaining new technology will be very limited and hence the firm may be better off not doing any R&D at all. Suppose the firm under consideration has a capacity larger than the threshold size, then the first order conditions (3.17) and (3.18) have an interior solution. The second order partial derivatives of $ are listed as follows: •fc> = r K» • / ( » ) * - ^ " ^ i l / ( a W B [*EV + (1 - ,)EV\ , Jo. 1 + Tj L J Ki = J~Vgrf(6)d9-y^4[EV-EY. 1 - F(a) _„ *// = />/W(*)cW + — f i ^ Ja 1 + t] EV -EV] . While the signs of VpD, V^j, V}'j are ambiguous, all of the second terms on the right-hand side of the above expressions have definite signs. The magnitudes of those second Chapter 3. R&D Investment, Technology and Financial Structure 62 terms are determined by the difference between EV and EV_, which is brought about by different technologies. If EV — EV_ is too small, the R&D project will not be worth taking at all. Suppose EV — EV_ is sufficiently large that the second terms outweigh the first terms, we will have $ ^ < 0 , $ ? 7 < 0 . (3.19) Also, * w < 0 , (3.20) i.e., increasing debt reduces the marginal benefit of R&D investment due to the increased risk of bankruptcy. Further, we require that KD*'1I ~ $DI2 < 0 (3.21) to ensure that the second order conditions are satisfied. Under these conditions, the solution to equations (3.17) and (3.18) will be the optimal strategy of the manager given the technology known to him. 3.3.2 Solutions for Different Current Technologies It is clear that the solution to the optimization problem (3.16) depends on the technology possessed by the firm at the beginning of the current period. Specifically, f (TJ,!) , if J1! = new, [ (D,I), if r x = o l d . Yet, the current state of the firm's technology affects the manager's expected utility only through the firm's revenue. In fact, D , m , (l+f*)9P(*), if r i = n e w : qp{6), if = old Chapter 3. R&D Investment, Technology and Financial Structure 63 Therefore, letting R1(8) = (l+T)qp(8) (3.22) and taking r as the parameter determining the optimal solution, we can examine the impacts of the current technology on the solution by comparing the case of r = p with the case of r = 0. Substituting (3.22) into (3.14), the expression for 4>, and differentiating (3.17) and (3.18) with respect to r yields the following equation DL where Generally, ^ DT ^DD *DI /•oo l V^R.R'Tf(8)de-Vj)(a).f(a)a'T  a [f'(a)a'Ta'D + / ( a ) a £ T 1+7/ / V!R-R'Tf{8)de-Vl{a).f{a)a'T Ja irEV +(1-TT)£?Z , 1+7? TT} \EV - EV V^R>0, Vj'R>0, Vj(a)<0, VV(a) = 0, i ? ; > 0 , a'T < 0 , /'(a)«r«i> + / ( a ) a £ T < 0 , (3.23) (3.24) (3.25) therefore, we have > 0. (3.26) The solution to (3.23) is K (3.27) (3.28) Chapter 3. R&D Investment, Technology and Financial Structure QA where ^ — ^DD^II *DI Applying (3.19), (3.20), (3.21), and (3.26) to the expressions (3.27) and (3.28), we see that the signs of D'T and l'r depend on the sign of $£>T. Specifically, if then D'r < 0 , and I'T > 0 , (3.29) which implies D < D and 7 > I. That is, when the firm has the new technology (the result of the successful R&D project of the previous period), the firm will invest more on its current R&D project and, at the same time, reduce its debt. Intuitively, if the firm has the new technology at the beginning, which will bring a higher expected revenue than the old technology would in the first period, the manager will be more interested in securing his future income (due to diminishing marginal utilily of his first period income). Consequently, he will invest more in the R&D project, and use less debt in the current period. However, if Kr >0 , the signs of D'T and I'r cannot be determined since in this case D'T is a negative term plus an ambiguous term while I'r is a positive term plus an ambiguous term. Yet, it still can be concluded that D'T and I'T cannot be both negative. Rewriting (3.27) and (3.28) as A d>£7 Chapter 3. R&D Investment, Technology and Financial Structure 65 anc / ' = — $ ' / T A ~ ^ DD we see that D'T < 0 would imply ^DT ^DI > ^DI Applying the second order condition (3.21) yields IT which implies I'T > 0. Thus, it cannot be true that both D'T and I'r are negative. In other words, the optimal level of either the R&D investment or the firm's debt (or both) when the firm possesses the new technology must be higher than that which is achieved when the firm has only the old technology. In general, D<D,m'i>L if ( i ^ £ i ) > i | - . <»•*» D>D and I> I if ^ - > [ ^ f 1 ) > , (3-31) D > D and I < I if $DD > f ^ „ DIJ . (3.32) (3.30) and (3.31) state that when the firm is successful in its past R&D, it will continue putting large investment in R&D. Since a higher investment in R&D will bring a greater chance of new success, the high level of R&D spending is likely to be sustained. Conse-quently, the firm is likely to become a technological leader in the industry. Gn the other hand, if the firm fails in its past R&D project, it will reduce its level of current R&D investment. This will put the firm in a position of disadvantage in the R&D race, and so the firm will likely become a technological follower. Chapter 3. R&D Investment, Technology and Financial Structure 66 If (3.32) holds instead, it will rule out the above process of classifying R&D leaders and followers. (3.32) predicts that the firm will reduce its R&D effort after success in its past R&D and will increase effort after failure. This makes a firm's continuing success in the R&D race less likely. Similarly, comparative statics with respect to firm size shows that large firms tend to invest more in R&D than small firms do, which implies that the R&D race favors large firms. 3.4 Empirical Evidence From the previous discussion of the solutions to the optimization problem, it is clear that two factors are crucial in determining the optimal R&D investment and the financial structure of the firm, namely, the firm's size and the state of its technology. While it is widely accepted that firm size plays an important role in determining R&D investment, the dependence of R&D upon the state of technology has not been recognized. Here, we should distinguish such a link between R&D expenditure and technological state from the relation between technological outcome and the R&D investment. The latter is often referred to as the production function of R&D and is quite a different subject from the effect referred to in this essay. In Section 3, the general solution to our model is expressed in the following fashion where both yD,I) and (D_,L) depend on the firm's output, q. Clearly, we can rewrite this solution as follows: new , old Di = Di(Ti), Chapter 3. R&D Investment, Technology and Financial Structure 67 Since we have assumed q to be constant, it appears that the technological state, T, solely determines the optimal level of D and I. However, in reality, most firms change their capacity from time to time; therefore, the functional forms of D and I should explicitly incorporate the effects of firm size, represented by q. Consequently, the general form of the solution can be written as Dt = D(qt,Tt,t) K ' (3.33) It = I(qt,Tt,t). In the following, we derive an empirically testable relationship from (3.33). Our primary interest is to test the hypothesis that Tt is important in explaining It and Dt • Taking differences from (3.33) and using a first-order approximation yields dD dD d Dt-Dt-! « —(qi-qt^) + —(Tt-Tt.1) + — (3.34) dl dl dl It-It-i ~ -^(qt - qt-x) +-^(Tt - Tt^) + ~ (3.35) which suggests the following regression equations ADt = b0 + bxAqt + b2ATt + et (3.36) AIt = c0 + cxAqt + c2ATt+vt (3.37) Since et and vt capture the omitted terms of higher orders of Aqt and ATt, they are nat-urally correlated to Aqt and ATt. This indicates that the ordinary least square estimates of fc's and c's will be biased, hence it may be useful to use instrumental variables in place of qt and Tt. A good instrumental variable for the firm's output, q, might be the firm's total assets, A. If A is a good measure of the firm's capital, and if the firm's capital is roughly proportional to its output (as with a Leontief technology), then we should have: qt = aAt (3.38) Chapter 3. R&D Investment, Technology and Financial Structure 68 with a being constant. In search of an instrumental variable for T, the state of technology, we first observe that T, by its nature, will be fully revealed by the price function of the firm's output, which is (l+rp)p(8), where r = 1 represents the new technology while r = 0 represents the old technology. Multiplying and dividing (1 + Tp)p(8) by q, we obtain (l+rp)p(e)q = R(8) q q where R(0) is sales. Further, using (3.38) gives . 1R(0) a A Next, we divide the ratio of the firm's total sales and total assets by the industrial average of that ratio to extract the effect of the random general demand condition, p(0). Specifically, let n denote the number of the firms in the industry, and let where, and in all following occasions, the first of a double subscript designates the firm. Define f =Qi / — = [1 + Titp)p{6)a = 1 + run /, 4 m * A* I Vt J E"=i(l + rjtp)p{6)a I E"=i(l + w ) ' { ' > The average state of technology ^ E ( i + w ) n3 = l can be roughly regarded as constant, hence, the normalized ratio of sales and assets, Ta, as defined in (3.40) can be used as an instrumental variable for the firm's state of technology. Chapter 3. R&D Investment, Technology and Financial Structure 69 Replacing qt and Tt in (3.36) and (3.37) by aAu and Tu yields ADit = /30+faAAit+!32Afit+eit (3.41) A/ft = f0 + li&Ait+72Afit+uit, (3.42.) where en and Ujt are assumed to be i.i.d. random disturbances and are uncorrelated with AAit and ATit. There is one more point we must discuss before proceeding with the empirical test. As is mentioned at the beginning of this section, the state of technology, T, may be related to R&D investment through some R&D production function. Apparently, this relation is in the reverse direction to the causal link established in (3.41) and (3.42), thus, it may undermine the testability of (3.42). However, since an R&D project is generally time consuming, there is a lag between investment in R&D and the acquisition of new technology. Accordingly, the unknown production function of R&D can be reasonably formulated as Tt = G{It-1,(t), (3.43) where ( t is a random factor. From this, we obtain AT t = Tt- T t_! = G(It-i, Ct) - G{h-2, Ct-i)- (3.44) Clearly, It-\ and It-2 are predetermined variables, and generally ( t and £t-i are uncor-related with et or ut. Therefore, the possibility of contemporaneous correlation between ATt and AIt need not worry us. Now, we are in a position to test the relations in (3.41) and (3.42). We use annual data from four U.S. industries, which are most frequently studied in R&D literature, to fit the regression equations. The data are adopted from Compustat Industrial Annual data tape. The firms in the sample are those listed under the SIC codes 3680 to 3689 (Computer industry), 2830 to 2839 (Drugs and Pharmaceutical Preparation industry), Chapter 3. R&D Investment, Technology and Financial Structure 70 Table 3.4: Grouping of the Data Industry Name SIC Codes No. of firm selected Total No. of observations Computer 3680-3689 31 507 Drugs 2830-2839 23 431 Machinery 3520-3560 45 727 Instruments 3810-3890 28 462 Total 127 2127 3520 to 3560 (Machinery industry), and 3810 to 3890 (Instruments industry). There are 222 firms in these categories; however, only the firms having at least 12 continuous annual observations are selected. This is because a long time series of observations is highly desirable to test the difference equations (3.41) and (3.42). The sampling period is from 1968 to 1987, and our selected sample consists of 127 firms totaling 2127 observations classified into the above four industries as described in Table 3.4. The time series data are then deflated by the implicit GNP deflator, and pooled cross-sectionally into the four industries based on the assumption that the firms in the same industry have similar technologies and similar financial structures. Each observation of the data consists of four entries: total assets, total sales, R&D expenditure, and total debt outstanding. As defined in (3.40), the index for any firm's state of technology, T, is constructed as the ratio of the firm's sales and assets normalized by the annual average of the same for the firms in the particular industry in the sample-Keeping in mind the structure of available data, we are now able to appreciate the merit of the formulation of (3.36) and (3.37) as compared with the alternative formulation which may appear first after reading (3.34) and (3.35). Obviously, instead of (3.34) and (3.35), we have the following approximation for (3.33): Dt-D0 « ^qt-qo) + ^-(Tt-T0) + ^t Chapter 3. R&D Investment, Technology and Financial Structure 71 W o * | ( „ - 9 0 ) + § ( r , - r 0 ) + | t which leads to the regression equations: Dt = b0 4 bxqt 4 b2Tt 4 b3t + et, h = c 0 + cxqt 4 c2Tt 4 c3t 4 vt. Since we are to use pooled cross-section time series data, the above equations turn out to be Dit = b0 4 bxqit 4 b2Tit 4- b3t 4 eit, (3.45) lit = c 0 4 cxqit 4 c2Tit 4 c3t 4 vit. (3.46) Although we have implicitly assumed the constancy of dD dl 3D dl dD dl dA' dA' #T' df' ~dt' ~di' (3.45) and (3.46) still pose some difficulty. Since the size of firms in our sample varies widely among firms (for example, in the computer industry, IBM had an average size about $35,000 million in assets while many small firms had only about $10 million), the regression coefficients in (3.45) and (3.46) will largely reflect the differences among firms. It would then be necessary to include firm dummies in the regression equations so as to enable b2 and c2 to capture the effect of changing technology within firms, which is the main focus of our model. However, setting up a large number of firm dummies would tremendously increase computational effort while providing us with little useful information. Therefore, equations (3.41) and (3.42), because they are formulated in terms of first differences, which eliminate the need for firm dummies, are more suitable for our purpose. The results of the regressions from the pooled cross-section time-series data for each industry are reported in Tables 3.5 and 3.6. Also reported in these tables are the results Chapter 3. R&D Investment, Technology and Financial Structure 72 for the four industries pooled together, in which a set of industry dummies are added. Conceivably, the major determinant of the R&D investment level and the debt level is the firm's size as evidenced by the fact that all of the 8[s and j[s are positive and highly significant. However, it is more interesting to look at the 8'2s and 7 2 s because these coefficients reflect the effect of technology, a factor not examined before. Table 3.5 shows that 82 is positive and significant (at the 5% level) for all four industries. Table 3.6 shows that, except for instruments, 72 is negative, and that for computers and machinery, 72 is also significant. For the pooled four industries, 82 is positive, 72 is negative, and both are significant. Almost all of the signs of 82 and ~y2 agree with the pattern listed in (3.30), indicating that firms tend to increase their R&D investment while reducing their debt level when they possess a better than average technology, and vice versa. Despite the fact that all 8'2s are significant, in general, the R2 for R&D equations are less satisfactory than R2 for debt equations. This, however, should not be surprising. R&D investments are project specific. For example, some projects require larger invest-ment than others, different firms have different projects, and any firm may be involved in different projects at different times. Corporate debt, on the other hand, can be con-sidered as basically homogeneous. Hence, an increment in R&D investment has larger idiosyncratic components than a change in financial leverage in response to changing firm size or changing states of technology. As a result, the fit of the debt equation is better than the fit of the R&D equation. It should be mentioned that there are several limitations in the above empirical tests. For example, the instrumental variable, T, defined as sales to assets ratio normalized by industrial average, may not accurately reflect the firm's state of technology. Firstly, by normalizing the firm's sales/assets ratio by industrial average, we hope to extract the impact of fluctuations in general demand condition. In other words, we have implicitly assumed that firms in the same industry with similar technology in the same year face Chapter 3. R&D Investment, Technology and Financial Structure 73 Table 3.5: Regression of R&D • AI = 0O + 0i • AA + f32 • AT Industry 8o ft 02 R-square D-W Name (t-ratio) (t-ratio) (t-ratio) f Statistic Computer 5.75 0.042 21.92 0.38 1.79 (3.74) (16.88) (2.49)* Drugs 5.38 0.019 30.95 0.18 1.31 (10.10) (8.79) (6.08)** Machinery 0.21 0.014 3.68 0.09 2.12 (1.03) (8.26) (2.51)* Instruments 1.97 0.034 10.66 0.13 1.70 (2.78) ( 7.90) (2.07)" Pooled 1.85 0.037 15.79 0.33 1.69 (2.07) (30.17) (5.35)** f only t-ratio of 02 is marked according to significance * significant at 5% level ** significant at 1% level Chapter 3. R&D Investment, Technology and Financial Structure 74 Table 3.6: Regression of Debt AD = 70 + 7i • A A + 72 • A T Industry 7o 7i 72 R-square D-W Name (t-ratio) (t-ratio) (t-ratio) f Statistic Computer -18.61 0.60 -150.5 0.82 1.39 (2.23) (45.0) (3.16)" Drugs -7.05 0.61 -16.1 0.76 1.85 (1.57) (33.9) (0.38) Machinery 1.52 0.64 -38.8 0.78 1.94 (0.94) (48.3) (3.42)" Instruments -3.16 0.65 9.7 0.68 1.10 (0.88) (29.8) (0.38) Pooled -1.22 0.61 -68.7 0.80 1.44 (0.24) (87.5) (4.16)" f only t-ratio of 72 is marked according to significance * significant at 5% level ** significant at 1% level Chapter 3. R&D Investment, Technology and Financial Structure 75 similar market demand. If this is not the case, i.e., some firms find a particular year is a good year whereas some other firms in the same industry with similar technology find the same year is a bad year, then the normalization of firms' sales/assets ratio does not eliminate the random effect of market demand. Secondly, in the short-run, it is possible to see that process R&D may reduce commodity price. If this reduced price leads to a reduced sales/assets ratio, the instrument for the state of technology also suffers from measurement errors. On the other hand, since the size of the firms included in the sample varies widely, the problem of heteroscedasticity is likely to be present. As a result, the Ordinary Least Square estimates may not be efficient. One way of avoiding heteroscedasticity is to transform the terms AD, AI, AA in Equations (3.41) and (3.42) into relative changes, AD AI AA , , , , . , i.e., , ——, and ——. However, the results thus obtained are similar to the results reported in the previous tables, i.e., all of the /32's and 72's have same sign as reported in Tables 3.1 and 3.2 and are mostly statistically significant. This similarity of the results obtained from two alternative measurements of the variables (absolute change and relative change, respectively) assures us that the potential problem of heteroscedasticity may not severely impair our basic findings. Finally, the sample used here consists of firms with at least twelve consecutive obser-vations reported in the Compustat data tape. Presumably, this will exclude some firms which bankrupted during the sampling period, perhaps resulting in a biased sample. The effects of this possible sampling bias on the estimated coefficients in the R&D equation are generally unclear. However, the bankrupted firms should have relatively high debt levels and a poor state of technology. Including these firms in the sample would then be AD likely to push the coefficient of ~^r, i - e - 72, downward. From Table 3.6, 72 is mostly negative, therefore, including the bankrupted firms in the sample would probably enlarge the magnitude of 72. Chapter 3. R&D Investment, Technology and Financial Structure 76 3.5 Conclusion In this essay, a model was constructed which incorporates both R&D investment decisions and financial decisions of a firm. It was assumed that the firm's objective in investing in R&D projects was to seek a better technology which would bring in extra profits when applied in production. Since R&D projects produce random outcomes in the future, such projects are financially risky and are best examined in a two-period model which can take account of the trade-off between today's investment and tomorrow's extra profits. The other dimension of the behaviour of the firm examined in this model concerns its financial structure; specifically, the amount of debt the firm is to issue. There has been a large body of literature discussing optimal financial structure for a firm, but this essay is not primarily concerned with the sole issue of firms' capital structures. We are interested rather in the interaction between a firm's R&D investment decisions and its financial decisions. Therefore, the theories of managerial agency costs were chosen to represent the basic trade-off concerning the financial decisions of a firm. The trade-off in our model is essentially that between the agency costs of equity, which can be reduced by debt financing, and the bankruptcy costs, which take the form of foregone future profits made possible by present investment. The interaction between R&D investment and capital structure turns out to be related to the state of technology of a firm in addition to the firm's size. This yields testable empirical relations between the firm's R&D expenditure, financial leverage, size, and state of technology. Although it is common to relate the firm's technological progress to R&D investment, the reverse link between the firm's R&D expenditure and the firm's state of technology has not been examined before. In the empirical section of this essay, the implications of the model are tested by industrial data for four U.S. industries. Since state of technology is firm specific and Chapter 3. R&D Investment, Technology and Financial Structure 77 is not easily obtainable, a proxy variable is used instead. This variable, namely, the firm's sales/assets ratio normalized annually by industrial average, may be subject to measurement error, but it captures the basic implications of the technological state. The simple regression model using this proxy variable yields coefficients which have consistent signs and are mostly statistically significant. The general findings are that firms' R&D expenditure is positively correlated with the firms' state of technology, whereas firms' debt is negatively correlated with the state of technology. This implies that when firms possess a better than average technology, they tend to increase their R&D investment and at the same time reduce their debt level, and vice versa. Although the test is rather unsophisticated, the results do encourage further research. Chapter 4 Economic Consequences of Technological Progress 4.1 Introduction It is well established that R&D has greatly contributed to productivity growth in manu-facturing industries (See, for example, Mansfield, 1980, and Griliches, 1986). As a result, manufactured consumption goods appear in much more abundance than ever before. However, due to declines in the relative prices, consumers may actually pay less while enjoying a greater quantity of manufactured commodities. In fact, the value of manufac-tured goods as a share of total GNP has been declining while the value of services has been increasing. For example, in the U.S., the ratio of the price index of goods to the price index of services declined from 1.63 in 1948 to 0.89 in 1986. In 1948, goods accounted for 44.5% of GNP, while services accounted for 33.8%. In 1986, goods accounted for only 27.5%, while services had risen to 60.3%. (CITIBASE: Citibank economic database). Obviously, the decline in the relative price of goods is attributable largely to R&D ac-tivity in the manufacturing industries. However, the fact that the service industries have outgrown manufacturing industries indicates a structural shift in the economy (See, for example, Maddison, 1980, Patton and Reilly, 1987, and Tschetter, 1987). Hence, the overall economic consequences of productivity growth in the manufacturing sector require closer investigation (See Varaiya and Wiseman, 1981, and Gertler, 1986). Traditionally, manufacturing industries are regarded as capital intensive whereas ser-vice industries are labour intensive. Therefore, the relative growth of service industries 78 Chapter 4. Economic Consequences of Technological Progress 79 will decrease the total amount of new capital employed per unit of labour in the economy. Presumably, this will have a negative effect on business investment. In fact, however, the recent growth of the service industries did not depress business investment as might have been expected (See Figure 4.1). This makes one speculate that services are in fact more capital intensive. Service industries may rely heavily on buildings such as business offices, stores, warehouses, etc. Hence, the growth in service output must be accommodated by investment in non-residential construction. While the depreciation of buildings is slow, the original costs of construction are high. Therefore, the service industries may initially require more capital than manufacturing industries. Here, a distinction should be made between cap-ital installation and capital consumption. Specifically, in terms of the stock of factor input mix, services are more capital intensive than manufacturings. However, since the physical depreciation of buildings is slow, in terms of the flow of capital consumption in production, services may still be regarded as labour intensive in the long-run. Two questions then arise. Firstly, can the observed structural shift in the economy be explained by underlying technological progress in the industries? Secondly, given that services are capital intensive in installation but labour intensive in consumption, what predictions can be made about future trends in business investment? Since the structural relationship between technological progress, equilibrium factor demand, and business investment can provide valuable information to direct government policies, these questions are undoubtedly of practical importance. However, trying to answer these questions is by no means easy. One has to consider various complicated mechanisms that are connected with manufacturings, services, busi-ness investment, and technological progress. Combining all these relevant mechanisms together into a general equilibrium model, while maintaining its tractability, is a real challenge. Complexity aside, to be of any use, the model must meet two requirements. Chapter 4. Economic Consequences of Technological Progress 80 Firstly, the model must be able to explain the established stylized facts regarding the structural shift in the economy. Secondly, under the perturbation of technological change, the model must be tractable in order to make predictions about the future course of the economy. This essay makes an attempt to relate the observed structural shift in the economy to productivity growth in manufacturing industries. The analytical framework involves a model of a three-sector economy which is an extension of Jones (1965). Jones studies a two-sector model where one sector is capital intensive and the other is labour intensive. The capital stock in Jones (1965) is treated as exogenous. The model in this essay incorporates another sector, the capital producing industry, to endogenize total capital stock, thereby the impact of technological progress upon business investment can be conveniently analyzed (See, also, Uzawa, 1962, 1963, 1964, and Johansen, 1967). The analysis shows that if the elasticity of substitution between manufactured goods and services on the demand side is less than unity, the value of the manufactured goods may decline even though the quantity of the goods will increase as the result of technologi-cal progress. The impacts upon other industries depend on the nature of the technological progress. In particular, if the technological change improves productivity of both labour and capital without altering the ratio of capital to labour in the input mix, the service industries will expand while manufacturing industries will shrink. Moreover, growth of service industries will stimulate investment in the short-run but reduce investment in the long-run, although it will raise the overall ratio of capital stock to labour in the economy. For other kinds of technological progress which cause a change of the capital/labour ratio in the input mix, the effect is sometimes ambiguous but can be analyzed by simulation methods. It should be mentioned that the noted structural shift in the economy, namely, from producing relatively more goods to relatively more services, is the result of the effects Chapter 4. Economic Consequences of Technological Progress 81 of both supply side and demand side. On the supply side, fast growing productivity enables the manufacturing industries to use less labour and capital while producing more goods. Consequently, resources of production are released from manufacturing industries, providing the basis for the growth of service industries. Furthermore, as national income rises, consumers will enjoy more goods and services. If services demand is more income elastic than manufactured goods, the marginal propensity to consume will be biased toward service consumption. In other words, when consumers in the economy are getting richer, they will demand more and more services relative to goods, dictating the expansion of service industries. While both productivity and income growth contribute to the structural shift in the economy, this essay will concentrate on the analysis of supply side effects, i.e., the economic consequences of technological progress. Since increasing income works in the same direction as does increasing productivity, incorporating demand side effects will not offset the predictions obtained in this analysis. The extent that the structural shift in the economy affects business investment is clearly revealed in an experimental estimation of the investment function. It appears that an expansion of service industries would require more capital investment than man-ufacturing industries. In addition, aggregate output tends to be more and more capital intensive in production. Since the composition of aggregate output has become increas-ingly service intensive, these facts reflect that service industries are likely to be capital intensive. On the other hand, empirical evidence shows that in the past technological progress was mostly labour saving in nature. For this particular kind of technical change, the model predicts a boom of service industries and an expansion of capital investment, which is consistent with observed facts. However, should technological progress turn to-ward capital saving or toward neutral in the future, the model would predict an eventual decline in investment even though service output would continue to increase. The remainder of the essay is organized as follows. Section 2 presents the model, Chapter 4. Economic Consequences of Technological Progress 82 Figure 4.1: Nonresidential Fixed Investment Figure 4.2: Volumes of Goods and Services Chapter 4. Economic Consequences of Technological Progress 84 Figure 4.3: Prices of Goods and Services Chapter 4. Economic Consequences of Technological Progress 85 section 3 analyzes the solution to the model, section 4 reports simulation results for some particular situations, section 5 investigates the implications for the investment function, and section 6 discusses the relevant empirical findings and gives conclusions. 4.2 The Model Consider an economy consisting of three industries: a manufacturing industry producing consumption goods, an investment industry producing capital goods, and a service in-dustry providing services to consumers. The outputs of these industries are denoted by M, I, S, respectively. The technologies of the economy are characterized by the following matrix a-ml amk an a-ik where a^ - represents the requirement for factor input j to produce one unit of output i. There are two factors of production, namely, labour and capital. Labour is paid a wage rate wi, capital is paid a rate of return wk plus a replacement cost of depreciation 8pk, where the depreciation rate 8 varies across industries; pk is the current price of capital goods. The following five assumptions are made throughout the analysis, (z) All industries are perfectly competitive Let the prices of manufactured consumption goods and services be pm and pe respec-tively. Then competition implies Pm = CLmlWl + amk(8mPk + Wk) , (4.1) pt = a„iwi + aak(8,pk +Wk), (4.2) Pk = Oiitvi + aik(8ipk + w k ) . (4.3) A = Chapter 4. Economic Consequences of Technological Progress 86 i.e., commodity prices equal costs. (ii) Community taste patterns are nomothetic such that the ratio of the demands for M and S depends only upon the relative commodity price ratio (as in Jones, 1965): £ = ' te) ( 4 4 ) Accordingly, the elasticity of substitution between the two commodities on the demand side, <T, can be defined as dDm dDe (dpm dps\ A7" X = -"\1Z-T)- ( 4 5 ) Rearranging (4.5) yields d(Dmpm) _ d(D„pa) _ _ (djprn _ dps\ DmPm Daps \pm p„ ) Clearly, when a = 1, the consumption shares of M and S will be constant. However, the past four decades have witnessed a secular declining share of goods consumption relative to services consumption along with a declining relative price, — , strongly suggesting that Vs cr < 1 (See Figures 4.2 and 4.3). This is formally stated in the following assumption. (iii) The elasticity of substitution on the demand side is less than unity. This assumption can be expressed as <T<1, (4.6) where a is defined in (4.5). (iv) Industry S is capital intensive relative to industry M in the sense that (1 + 8B)a,k (l + 8m)amk However, the physical depreciation of capital in S is slower than that in M such that Chapter 4. Economic Consequences of Technological Progress 87 The capital intensiveness of industry / is similar to that of industry M. Moreover, the whole capital stock in industry I cannot be reproduced in one period, i.e., a»fc > 1, although Sidik < 1 • (y) The total endowment of labour of the economy, L, the wage rate, wi, and the rate of return on capital, Wk, are exogenous and fixed, while the total capital stock, K, is endogenous. The demand for new capital goods is Di = SmKm + 6aKa + 6iKi + (Km + K„ + K~i — K). (4.7) where K represents the capital stock in last period. If the economy is in a steady state, Km + K. + Ki - K = 0. However, when the economy is expanding, Km + K, + Ki- K >0. It should be noted that the assumption that the factor prices, wi and u>k, are constant is not as stringent as it appears. Since the ratio of the demands for M and S depends on the relative commodity price, which in turn depends on technology and factor prices, a technological change will alter the relative demand by changing the relative commodity price. As a result, the composition of the aggregate consumption of the economy will shift toward either more capital using or more labour using, causing temporary disequi-librium in factor markets. Consequently, the relative factor price must move to invite (or Chapter 4. Economic Consequences of Technological Progress 88 deter) investment. Call the change in factor prices the induced change. Obviously, both the changes in factor requirement (technological change) and in factor prices (induced change) will cause commodity prices to vary. However, compared with the impacts of technological change, the induced change in factor prices is likely to be of secondary importance if the elasticity of factor demand (or supply) is not too small. Suppose, for example, that technological progress reduces the cost of producing com-modity M. The decline in the price of M will alter the ratio of demands for M and 5, which in turn will disturb the factor market equilibrium as resources are switched between industries. Specifically, suppose the result is a shortage of capital, then the price of capital must rise relative to the price of labour in order to invite new investment. This induced change in factor prices will also affect commodity prices. However, if the elasticity of capital supply is not too small, i.e., investment is indeed sensitive to the price of capital (See, for example, Shapiro, 1986), the change in the relative price of capital is likely to be small. Accordingly, the effect on the commodity prices of changing factor prices is not comparable to the effect of the technological progress itself. Other-wise, in general equilibrium, we could observe a technological change that reduces factor input but causes total factor costs to increase. We abstract from induced factor price changes to avoid unnecessary complications in the analysis. The assumption of constant factor prices will greatly simplify subsequent analysis without altering the main results materially. There is also the possibility of exogenous shocks in factor prices (due to government regulation, for example), which may have larger impact on commodity prices. Since this essay focuses on the consequences of technological change, any other exogenous shocks will be ignored. The other possible environment where fixed factor prices are realistic can be found in planned economies, where factor prices are indeed fixed (For example, before recent Chapter 4. Economic Consequences of Technological Progress 89 experiences of high inflation, the wage level in China had been virtually fixed for some twenty years. However, relative commodity prices and composite consumption bundles did vary from time to time). Based on these assumptions, we can carry on our analysis of the economic conse-quences of technological progress. In general equilibrium, all the markets will clear, i.e., M = Dm, (4.8) S = D., (4.9) I = Di, (4.10) L = L. (4.11) In view of assumption (ii), the demands for M and S are not independent. Substi-tuting (4.8), (4.9) into (4.4) yields According to (4.7), the capital market clearance condition (4.10) can be written as I = 8mKm + 6.K. + SiKi + (Km +Ks + Ki-K) . (4.13) Similarly, the labour market clearance condition (4.11) can be written as L = amiM + a.iS + aaI. (4.14) (4.12), (4.13), (4.14) and (4.1), (4.2), (4.3) are six independent equations relating six vari-ables: M, S, I, Pm, ps, and pk- These equations clearly reveal the structure of the model. Firstly, commodity prices, which always equate costs, are determined by production technology. Secondly, consumption of goods and services is determined by commodity prices and the capacity of each industry. Finally, capital investment is determined such that the gap between the desired capacity and the initial endowment of the economy is Chapter 4. Economic Consequences of Technological Progress 90 closed. Hence, this system of equations completely determines the equilibrium state of the economy. Now, suppose the technological progress in manufacturing industry raises the produc-tivity of that industry. Specifically, suppose dami < 0 , and damk < 0 . The task of the following analysis is to investigate the economic consequences of such technological progress. Differentiating (4.1), (4.2), (4.3) yields dpm = widami + (8mpk + wk)damk + amk8mdpk , dp, = aek8sdpk, dph = aikhdpk • Therefore, dpm = wtdaml -f (8mpk+ wk)damk , (4-15) dp,=dpk = 0. (4.16) Equation (4.15) can be rewritten as dpm _ wiami damt ^ (wk + 8mpk)amk damk Pm Pm 0-m.l Pm O-mk damL damk = Cmi r cmk , (4.17) where is the cost share of labour and cmk is the cost share of capital. Clearly, cmi + cmk = 1, The interpretation of (4.17) is clearly that the change in output price is the cost share weighted average of the changes in productivities of the factors. Combining (4.5), (4.8), Chapter 4. Economic Consequences of Technological Progress 91 (4.9) and (4.16) gives dM dS dPm — - — = -<r—. (4.18 M b pm Further, differentiating (4.y) and (4.z) gives dl - 6mdKm - 8adK, - 8{dKi - dKm - dKs - dK{ = 0, (4.19) and amldM + Mdaml + asldS + aitdl = 0. (4.20) Rearrange (4.19), using the coefficients of production, •[1 - (1 + 8i)aik]dl - (1 + 8m)amkdM - (1 + 8,)askdS = (1 + 8m)Mdamk . or 1 (l + 8i)Ki Writing dl (1 + 6m)Km dM (1 + 6.)K. dS = (1 + Sm)Km damk I I M I S I amk ' [ " ' (l+8i)Ki (l+8m)Km (l+8e)Ks I ' ™ I (4.21) can be expressed as (4.22) dl dM dS damk {1 - U i ) — - um—- ut— = wm . (4.23) 1 M b amk Note that tu's are proportional to the shares of additional capital consumption of the expanding industries and need not sum to unity. Similarly, Equation (4.20) can be rewritten as '-^ +4 + 4 = ( 4 24) M b 1 ami where L = ° ^ , *. = ^ , k = °f (4.25) are the shares of labour employed by three industries and sum to unity. Chapter 4. Economic Consequences of Technological Progress 92 Finally, substitute (4.17) into (4.18), dM dS damt damk —T} 7T = -o-Cmi o-cmk . (4-26) M S ami amk Solving (4.23), (4.24), and (4.26) for ^ , and ^ yields 1V1 O 1 dM L + {l° + l i T ^ ) a C m l da '•ml 1 - U J i k- h I t, + k~ crcmk , 1 — U!j V 1 — (VjJ ttflmfe 1 - U > i lm ~\~ Is ~r~ li' S , , , , ,U.+Um ar 1 — UJ{ ' i ^ ( tm + it" 0"C m f c , 1 - u^ i \ 1 - ujj J da>mk 1 — Vi d£ _ (ivm + u s ) l m + (u; m / „ - (vjm)<rcml dami ~(^ m + *«)^m + - (V8lm)<TCmk damk 0~mk (4.27) (4.28) (4.29) ( i -u ,o (z m + z, + Z i ^ - ± ^ ) V 1 - (Vi J Based on (4.27), (4.28), and (4.29), we shall investigate the economic consequences of technological change in industry M in the next section. 4.3 General Predictions of the Model This section analyzes the solution to the model and makes the general predictions upon the economic consequences of technological progress. To facilitate the analysis, we first prove two lemmas which require the following definitions. Chapter 4. Economic Consequences of Technological Progress 93 Definition The economy is said to be in steady state if i) the economy.is in equilibrium as defined by the markets clearing conditions (4.8), (4.9), (4.10), and (4.11) and ii) the industries keep production constant, i.e., there is no switch of labour between the industries and no change in capital employment. The output of industry I will exactly replace the depreciated capital in the economy. Definition Suppose the economy is initially in a steady state. A perturbation of a technological change in industry M will cause the outputs and factor inputs of all the industries to change. Before the economy reaches another steady state, we say that the changes are short-run, while after the economy has reached the new steady state, we say that the changes are long-run. With this terminology, we can prove the following lemmas. Lemma 1 In the long-run, cml < uj I + I-and 1 — UJj I A- I 1 — UJ; Proof By definition, cm; is the cost share of labour in industry M, i.e., wiami Cml = i Chapter 4. Economic Consequences of Technological Progress 94 By (4.1) and (4.2), pm = wiami + (wk + 8mpk)amk wiau + wkaik = wiami + wkamk + O m a m f c — I - 6iaik So, 1 , , c amk au uikamk ( 8maik = 1 + om 1 1 + Cml 0-ml 1 - ^iaik WlO-ml \ 1 ~ ^ife. vm On the other hand, lm is the share of labour employed in industry M, Zj is the I- Wi share of labour employed in industry / producing supporting capital for industry M. According to (4.22) and (4.25), noticing that in the long-run, capital consumption shares become SmKm ^sK6 SiKi ^m — J , Us — j , U)i — —j- , lm aml 1 i 7 W m I c aH — a-mi -r omamk-— Tilt I — J I L K -. r--U>i 1 - biaik I A. I UJm-1 m 11 - Vi _ wkamk (^ ^ 8maik ^ ^ therefore, Cmi lm u>iami \ \ — 8iaiki which demonstrates the first inequality. Using this inequality and the identity Cmk = 1 - Cml will establish the second inequality. Q.E.D. Lemma 2 The following ratio R = lm ~\~ Is ~\~ li' 1 — U)i is larger in the long-run than in the short-run and is always bounded by zero and one. Chapter 4. Economic Consequences of Technological Progress 95 Proof The reciprocal of the ratio can be written as a,i + (1 + 8„)ask-(i) - j , 5 " " " " l - q + gQofl. \RJSR M A M L + 8 M ) A M ** 1 - (1 + bi)aik in the short-run and , K A I L J L ] = I + — 1 - fl.Q.fe M aml + 6 m a m k - ^ -in the long-run. By the assmption (iv) regarding the relative capital intensiveness, (1 + St)ask (1 + ^ )"mt in the short-run, while 8,0,8k < 8mamk in the long-run. Also since 1 - (1 + 8i)aik < 0 and 1 - 8i<Lik > 0 , it can be shown that R)SR > (R)LR ' Furthermore, it is easily seen that 1 ti-J LR Hence, 0 < RSR < RLR < 1 > which proves the lemma. Q.E.D. Chapter 4. Economic Consequences of Technological Progress 96 Before we present our major propositions about the economic consequences of the technological change in industry M, we shall introduce some additional terminology. Definition We say the technological change in industry M is labour saving if . dami < 0, and damk = 0 , we say the technological change is capital saving if dami = 0, and damk < 0 , and we say the technological change is neutral if dami _ damk Q O-ml amk Proposition 1 A capital saving technological change in industry M will increase the output of M. A labour saving technological change will increase the output in the long-run, but it may decrease the output in the short-run. Neutral technological progress will also increase the output; however, the percentage increase in output in the long-run is larger than that in the short-run but less than the percentage increase in productivity. Proof From Equation (4.27), it is obvious that both of the coefficients dM/M , dM/M and - daml /aml - damk j amk will be positive in the long-run; this is expected if M is not a inferior good. However, in the short-run, dM/M —dami/ami Chapter 4. Economic Consequences of Technological Progress 97 may be negative although will still be positive. dM/M -damk/amk When the technological change is neutral, . m k — —— . Hence, (4.27) leads to dM ~M~ Im ~\~ l{ O-mk O-ml Urn 1 — Ui lm + l. + k Ua + U m l - U i [a + (1 - a) R] da. ml O-ml where R is the ratio defined in Lemma 2. From (4.6), 1 - a > 0. da-&ml Hence, by Lemma 2, 'dM* M < (dMy \ M SR \ 1V1 I LR Furthermore, since the ratio R is less than unity, we have dM ~M~ < LR dami O-ml which completes the proof. Q.E.D. It is obvious that any technological progress in industry M will reduce the price of M and stimulate demand for M. However, since the productivity in the industry is improved, even if a larger quantity of M is produced, the overall scales of the industry may not necessarily increase as is shown in the following proposition. Proposition 2 Even though any technological progress in the long-run will increase the output of industry M in physical units, due to the resulting declining in price, the change in the total sales of M is generally ambiguous. However, if the technological Chapter 4. Economic Consequences of Technological Progress 98 change is capital saving, the value of total sales of M will decline in the long-run; while if the technological change is neutral, the value of sales will decline both in the short-run and in the long-run. Proof The change in the sales of M can be expressed as d(PmM) = PmdM + MdPm = pmM (^ + ^ ) . \ M pm J According to (4.18) and (4.27), if the technological change is capital saving, dM | dpm M pm . Vr •l-Ufi (1 - <r)(lm +le+ l i 0 ^ ^ ) + <r(lm + UjZ^)] *nk da. mk lm ~\~ Is ~r h Us + Vm k~ ~ (lm + li- — ) Cmk —<r(le+U- f— ) Cmk _ . 1 — u>i V 1 — u>j J J V 1 — VjJ lm ~\~ Is ~f" li Vs + V m 1 — V{ da. mk amk By Lemma 1, in the long-run liT^-~ (lm + h~IL-)cmk<0, 1 - Vi \ 1 - Vi J therefore, dM dpm ^ Q M Pm in the long-run. On the other hand, if the technological change is neutral, Chapter 4. Economic Consequences of Technological Progress 99 Hence, using Lemma 2 gives d M dp, < 0 M p, both in the short-run and in the long-run. Q.E.D. Now, the attention is turned to industry S. The following proposition reveals that the impacts of productivity growth in industry M upon S are different in the short-run and in the long-run. in industry M upon the output of industry S is ambiguous in the short-run. However, in the long-run, a labour saving technological change in M will increase S. Moreover, if the technological change is neutral, it will increase S both in the long-run and in the short-run; the percentage increase in S in the long-run is larger than that in the short-run but less than the percentage increase in the productivity of industry M. Proof According to (4.28), for a labour saving technological change, By Lemma 1, in the long-run, the right side of the above equation will be positive. Therefore, a labour saving technological change will increase S in the long-run. For neutral technological change, (4.28) gives Proposition 3 The impact of a labour saving or capital saving technological change Since, by Lemma 2, the ratio lm+U m Im Is 4" l{ 'm Chapter 4. Economic Consequences of Technological Progress 100 is larger in the long-run than in the short-run, and since it is always positive but less than unity, the following inequalities must hold. Q.E.D. From the above proposition, it is clear that except, perhaps, for a pure capital saving technological change, the productivity growth in industry M will increase output of S. Since the technology of industry S is fixed, this indicates a physical expansion of the industry, i.e., more labour and capital will be employed to produce S. Then the important question arises: How will business investment respond to the need to accommodate the expansion of industry 5? The following proposition gives the answer. Proposition 4 If the technological change in M is labour saving, industry I will expand both in the short-run and in the long-run. However, if the technological change is capital saving, industry / will shrink both in the short-run and in the long-run. Moreover, if the technological change is neutral, industry / will expand in the short-run but shrink in the long-run. Proof For a labour saving technological change, (4.29) gives while for a capital saving technological change, (4.29) gives dl lm (wm + ujeacmk) + ujmle (1 - acmk) / damk\ 1 ( i _ ( l / i ) ( i m + /. + / i ! ^ ± ^ ) Chapter 4. Economic Consequences of Technological Progress 101 For a neutral technological change, (4.29) becomes dl T (umL - u)elm)(l - a) \ 1 - U; J ddr, = -I I [ — U, cr Since in the short-run (4.30) ( l + * m ) a m f c (l+6t)aBk &ml < 0. while in the long-run Im U, I. Smdr 0-ml sk > 0. Equation (4.30) leads to the conclusion that (dl\ > 0 SR and that for the neutral technological change. dl < 0 LR Q.E.D. No doubt, the most interesting case is neutral technological progress. In this case, the short-run effect and the long-run effect on investment are contrary to each other. In view of this, one should not be too optimistic if the boom in service industries relative to manufacturing industries brings a surge in business investment. In the long-run, the effect may turn out to be negative. Another question one may ask is how national income will change as a result of a technological change in the manufacturing. Unfortunately, the answer to this question is Chapter 4. Economic Consequences of Technological Progress 102 unclear. By definition, gross national income can be expressed as y = PmM +p„S + p k I Accordingly, the change in y can be calculated as dy = pmM d(PmM) PmM or dy = pmMdM p^dS_ p j j d l pmM dp y y M y S y I y p„ The change in y is the weighted average of changes in M , S, I, and p m , the signs of which generally do not agree. Hence, the direction of change in national income is ambiguous. However, one should be aware that the national income defined above is nominal income. Since productivity growth in industry M brings down the price of M , even if the amount of income declines, the purchasing power of the income may not necessarily decline. The welfare of the economy should, of course, be judged according to the real consumption of the consumers. Proposition 5 A neutral technological change in industry M will increase real consumption both in the long-run and in the short-run. The percentage increase in the long-run is larger than in the short-run but less than the productivity growth. Also, a labour saving technological change will increase real consumption in the long-run. Proof Real consumption is defined as Y = pmM + pBS where Pm , Pa a r e fixed (constant) prices. Hence, dY = PmdM + p.dS — Y 6 — + ( 1 - 6 ) — Chapter 4. Economic Consequences of Technological Progress 103 where 6 PmM pmM +paS is the consumption share of M. For a neutral technological change, dY_ , 1 - Uli \ 1 — U>iJ L + + k +(1-6)(1-<T) Us + Un Im ~\~ k 1 - UJ{ da-mi &ml . da Im ~\~ Ig ~\~ k U 8 + U r , 1 — Ui ~\~ k Ur, l - U i da + (1 - a) u> +u; Im ~\~ Is k Z 1 — Ui > Applying Lemma 2 then gives ml O-ml , > -dami ami , 0 < dY ~Y < SR dY LR Also, < [0* + {l-<r)} >• / LR = [1-^(1-^) ] dami O-ml , dami" O-ml , Then using (4.6) leads to dY' dami < — LR a™1 For the labour saving technological change, Y M ' S Chapter 4. Economic Consequences of Technological Progress 104 lm + CmlO~ By Lemma 1, L + 1. + k Us + U m 1 — UJ: lm — CmlCT I Z T O + Z{ 1 — U>{ L+ls+ h Us +Urr, 1 — U»i mi da-mi O'ml Q.E.D. The intuition behind the above proposition is quite obvious. For example, as stated in Propositions 1 and 3, a neutral technological change will increase output of both M and S. Therefore, no matter what happens to current prices, consumers will enjoy more of both manufactured goods and services. In other words, real consumption increases as a result of technological progress. However, in the case of capital saving technological change, the impact on the output of services is ambiguous. Hence, the change in the level of satisfaction of consuming a new consumption bundle, which may consist of more manufactured goods but less services than before, is ambiguous. Finally, it is of great interest to see how the overall capital/labour ratio in the economy will change in the future. In Proposition 4, it is shown that in the case of neutral technological change, investment will shrink in the long-run. However, this does not imply that the total capital stock in the economy will decline. Indeed, since the depreciation rate of capital varies across industries, the total investment in equilibrium may not be directly related to the total capital stock in the economy. Proposition 6 In the long-run, if the technological change in industry M is labour saving or neutral, the total capital stock in the economy will increase. Chapter 4. Economic Consequences of Technological Progress Proof The total capital stock in the economy is K = Km + Ka + Ki = amkM + askS + aikI. dK = amkdM + a,kdS + aikdl + Mdamk = K ™ + K , % + K " + K ^ . M S f amk Using (4.28) and (4.29), dK = [Km + K. + Ki U m + u, \ dM l-ui ) M v, dami + KiZ —O-Cml 1 - Vi a m i + {Km+ K,<rcmk+ Ki v„ + v„ 1 — Wi 1 - Vi o~cmk da. mk amk In the long-run, 1 — u;, = vm + v, .dM dK K-M v„ da^ vm + ve a m l +KV + {Km + K,o-cmk + Ki vT. + V, Vm + V e Vm + V, o-cmk ' 1 damk • i Omfc For labour saving technological change, .dM v, dK = K— +Ki M Vm + Vs da. o-Cmr ml ami = l K \ l m + ( l . + l i )<TCml v„ Um + V, <rcmi )(-da. ml Oral > ( r c a ^ + X ) ! - — I >0. Vm + UB 0>ml For the neutral technological change, dK = K^ + {Km + Kscrcmk + Ki-Ki Vr, + V, Urn +U„ V m + V, O-Cmk 1} dar O-mk Chapter 4. Economic Consequences of Technological Progress 106 - (1 - <7Cm f c) — - (1 - C T ) — ml ami > K lm. Km K By assumption (iv), + — f — (1 - o-) (/; - ^f] + trcm f c f/i - ^ da. ml <Lml Km M M K - ami^ - amk-g M fK amk = aTnl~j^ \ L ami > 0 Hen ce, K I (K a i l K I T dK > 0. Oik an > 0. Q.E.D. Intuitively, as a result of a labour saving technological change, the requirement of capital for producing each unit of goods does not change. Since the output of both M and S will increase in the long-run (See Propositions 1 and 3), the total capital stock will necessarily increase to accommodate an enlarged scale of production. In the case of neutral technological change, the capital/labour ratio of the manufacturing industry does not change. However, since the service industry has a higher capital/labour ratio, the overall capital/labour ratio of the economy will increase as the service industry expands. This, for a given labour supply, implies a larger capital stock. Chapter 4. Economic Consequences of Technological Progress 107 4 .4 Numerical Simulations The previous section conducted general comparative statics analysis. In some cases, the results are ambiguous; in other cases, the comparative statics show only the direction of marginal change. Because the evolution of the economy from the initial steady state to the new steady state after a perturbation of technological change cannot be fully described by the comparative statics, this section will carry out simulations to obtain a more complete picture of the dynamic process and the magnitude of the changes in each sector of the economy. To begin with, the production coefficient matrix of the model economy is specified below: 0>ml 1 2 a,k 1 2.5 a-ik 1 2 It is noted that the units of output can be chosen such that the production of one unit of each output requires one unit of labour input. So, for simplicity, ami, a„i, an are all set to unity and the unit of capital is determined by the unit of output of industry I. Since data on capital installation in each industry are not available, the numerical values of amk, aak, and a.*, used here are not meant to be approximations to real technologies, but only used as an illustration. The rate of return on capital, Wk, is assumed to be 10% per year, and the depreciation rates of capital for the three industries are set as 8m = 15%, 6. = 5%, Si = 15% In other words, capital installed in manufacturing industries (presumably equipment) has a lifetime of about 7 years, while capital installed in service industries (presumably buildings) has a lifetime of about 20 years. It is easily verified that (1 + <$m)amfc (1 + Se)gek Chapter 4. Economic Consequences of Technological Progress 108 and that This implies that in terms of capital installation, industry S is capital intensive; while in terms of capital consumption, industry S is labour intensive. The wage rate of labour, wi, is chosen such that the cost shares of labour and capital in manufacturing industries are initially 50% each. Accordingly, the cost shares of service industries are 57% for labour and 43% for capital. Finally, the elasticity of substitution between M and S is set as a = 0.4 and the total goods output and service output of U.S. economy in 1986 are taken as the initial levels of M and S. All other variables in the model can be computed from the known parameters and variables. Hence, the specification of the initial state of the model economy has been completed. Now, suppose industry M is undergoing technological progress. Three numerical experiments are conducted according to the nature of the technological progress, namely, for a 10% reduction in labour requirement (—— = —.10, a labour saving technological change), for a 10% reduction in capital requirement (—— = —.10, a capital saving amk technological change), and for a 10% reduction in both factor requirements (a neutral technological change). It takes four to six years for the model economy to reach a static equilibrium after the technological perturbation. The main results are summarized in Table 4.7. It is noticeable that there are large discrepancies between short-run and long-run impacts . For example, a capital saving technological change in industry M will cause the investment industry to shrink by about 25 percent in the short-run while in the long-run the reduction in investment amounts to only 4 percent. Also, the same technological change will cause the output of M and S to increase by 8.0 percent and 5.8 percent in Chapter 4. Economic Consequences of Technological Progress 109 Table 4.7; Simulation Results for Model Economy, I New Model: Services Are Capital Intensive Technology Labour Saving Capital Saving Neutral Horizon S-R* L-Rt S-R L-R S-R L-R M -1.8 3.9 8.0 2.5 6.0 6.6 PmM -6.7 -1.3 2.6 -2.7 -4.6 -4.0 s -3.8 1.8 5.8 0.4 1.6 2.2 I 24.4 3.0 -25.4 -4.0 1.4 -1.0 y 1.6 1.3 -1.9 -1.4 -0.1 -0.2 Y -3.1 2.6 6.6 1:1 3.1 3.7 K 1.7 2.5 -1.7 -2.2 0.2 0.3 N.B. All values are percentage changes * Short-Run change denned as the change in the first year tLong-Run change defined as the total change over six years Chapter 4. Economic Consequences of Technological Progress 110 the short-run but only 2.5 percent and 0.4 percent in the long-run. The common belief that service industries are labour intensive relative to manufactur-ing industries makes no distinction between capital installation and capital consumption. This motivates the following simulation to see what economic consequences such a belief would predict under similar circumstances. In this simulation, the value of ask in (4.31) is replaced by 1.75 and the depreciation rate 8S is changed to 15% while all other pa-rameters are unchanged. The same procedure is used to compute the dynamic process of the economy perturbed by a technological progress. It is observed that it takes three to seven years for the alternative model economy to reach the static equilibrium. The main results of the simulation are summarized in Table 4.8. The predicted paths of M, S, I, and K for the two alternative model economies, after a neutral and a labour saving technological perturbation, are depicted in Figures 4.4 to 4.7 and Figures 4.8 to 4.11 respectively. Comparing Table 4.8 with Table 4.7, we see that, in general, the long-run impacts are quite similar except for the total capital stock. Table 4.7 predicts that the total capital stock (or, equivalently, the overall capital/labour ratio) will increase whereas Table 4.8 reveals that it will decrease after a neutral technological change. In the short-run, however, the difference in magnitudes is large although the signs generally agree. The exception again arises in the case of neutral technological change. In this case, Table 4.7 shows that a boom in service industry is accompanied by an expansion of investment, contradicting the opposite result in the Table 4.8 where the investment is negatively correlated to service output even in the short-run (See, also, Figure 4.6). This contradiction has interesting implications for the estimation of investment func-tions. Traditionally, an investment function is the association between aggregate business investment and aggregate output (See, for example, Bischoff, 1971, Jorgenson, 1.971, and Clark, 1979). In the case of neutral technological progress, if service industries are labour Chapter 4. Economic Consequences of Technological Progress 111 Table 4.8: Simulation Results for Model Economy, 77 Naive Model: Services Are Lab our Intensive Technology Labour S aving Capital Saving Neutral Horizon S-R* L-Rt S-R L-R S-R L-R M -3.9 3.9 11.9 2.4 7.0 6.5 pmM -8.8 -1.3 6.3 -2.7 -3.7 -4.1 S -5.9 1.8 9.6 0.4 2.6 2.1 I 21.5 2.5 -26.3 -2.6 -1.3 -0.2 y 1.4 1.2 -1.6 -1.3 -0.1 0.1 Y -5.2 2.5 10.4 1.1 4.1 3.7 K 2.8 2.5 -3.4 -2.6 -0.2 -0.1 N.B. All values are percentage changes * Short-Run change defined as the change in the first year fLong-Run change defined as the total change over seven years F i g u r e 4.4 P r e d i c t e d V a l u e o f O u t p u t M a f t e r N e u t r a l T e c h n i c a l C h a n g e M a : f o r N e w M o d e l , M b : f o r N a i v e M o d e l Chapter 4. Economic Consequences of Technological Progress 113 Figure 4.5: Predicted Services : Neutral Chapter 4. Economic Consequences of Technological Progress 114 Figure 4.6 Predicted Investment : Neutral Chapter 4. Economic Consequences of Technological Progress 115 Figure 4.7: Predicted Capital Stock : Neutral Figure 4.8: Predicted Goods Labour Saving F i g u r e 4 . 9 P r e d i c t e d V a l u e o f O u t p u t S a f t e r L a b o u r S a v i n g T e c h n i c a l C h a n g e S a : f o r N e w M o d e l , S b : f o r N a i v e M o d e l 0 1 2 3 4 5 6 T i m e ( i n Y e a r ) Chapter 4. Economic Consequences of Technological Progress 118 IT) C\| L O •>— L O <N *— O Figure 4.10: Predicted Investment : Labour Saving Figure 4.11: Predicted Capital Stock Labour Saving Chapter 4. Economic Consequences of Technological Progress 120 intensive, as assumed in the model underlying Table 4.8, investment will be negatively correlated with aggregate output. However, if service industries are capital intensive, the correlation will be positive in the short-run but negative in the long-run. 4.5 Accelerator Models of the Investment Function According to the common belief that service industries are labour intensive, the fast growing service sector relative to the manufacturing sector would imply a declining over-all capital/labour ratio and therefore should have a negative impact on fixed investment. Clearly, this prediction is not supported by U.S. data which show that the time path of the total fixed investment actually had a positive trend (Figure 4.1). If service industries are capital intensive, however, investment will increase in the short-run even though in the long-run it might be expected to decline. Hence, the evidence of the increasing in-vestment can be explained by two effects. The first is the size effect: with positive growth of population and labour force, the increasing size of the economy will push investment upward. The second is the effect of a structural shift. When manufacturing industries are undergoing constant technological change, continuous structural shift would admit increasing investment as a sustained short-run effect. While an aggregate investment function naturally takes into account the size effect, it cannot reflect the effect of struc-tural shift. Therefore, when both effects are present, a modification of the formulation of aggregate investment function is desirable. Take the traditional accelerator model as an example. Suppose the technology of an economy is of the leontief type, then the desired capital stock will be a constant multiple of output, i.e., Kd = aY, where Kd is the desired level of capital stock, Y is aggregate output of the economy, Chapter 4. Economic Consequences of Technological Progress 121 and a is a constant. If the capital stock could be adjusted instantaneously at a constant price, the actual capital stock would always equal the desired capital stock. Hence, net investment would be proportional to variations in output, NIt = K t - Kt-r = Kf - K*_x = a (Yt - Yt^) . However, in reality, since a radical capital adjustment is costly, the reaction of the capital stock to output should be spread over a number of time periods. On the other hand, the desired capital stock is based on the expected future output. Because past levels of output are the most important determinants of expectations of future output, net investment should be related to past output changes over a number of periods through a set of distributed lag coefficients: T NIt = Kt -Kt-i = E / 3 T (AYt-T) . (4.32) This is the familiar accelerator model of the investment function (Lucas, 1967, Clark, 1979). It is observed that if the technological coefficient a evolves over time, the coef-ficients 8T in Equation (4.32) will not be constants; rather, they will depend on time, For a two sector economy, namely, an economy consisting of a manufacturing industry and a service industry, even if the technologies of both industries are kept unchanged, a shift in output shares will be equivalent to a change in the technology of producing aggregate output. Specifically, let the desired levels of capital stock for the two industries be related to the levels of output as Km=amkM, K* = askS, then the relation between the aggregate capital stock and aggregate output is K = Km + K' = amkM + atkS = (jM^amk + j ^ ^ k ) (M + S). (4.33) Chapter 4. Economic Consequences of Technological Progress 122 Denoting Y as M + S, the relation K = oY implies that M S oc = a m f c — — — + a„k _ -M + S M + S Obviously, a is the (output share) weighted average of capital requirement coefficients of the two industries: amfc and ask- Hence a shift of the output share will change a even if both amk and a„k are constants. The above arguments reveal that a should evolve over time, which motivates the following formulation of the investment function: T NIt = Kt- Kt-r = £ / 3 T (AY;_ T) + e-t(AYt). (4.34) T = 0 Note that for a constant growing economy Equation (4.34) leads to NIt = (J2f3r + e-t)AYt. T = 0 Hence, e can be interpreted as the trend of coefficient change, which includes both the effects of changing technological coefficients, amk and a„k, and the changing shares of output M and S. Another alternative formulation of the investment function may be similarly derived from Equation (4.33), i.e., T T NIt = NI™ + NTt = E br (AM t _ T ) + em • t (AMt) + £ cT (A5 (_ T) + e* • (AS,) , (4.35) T=0 T=0 where em and e" account for the technological changes in M and S industries, respectively. Equations (4.32), (4.34) and (4.35) are estimated using U.S. national accounts data (annually from 1948 to 1986). M is taken as the real output of goods producing industries and S is the real output of service producing industries. Y is the sum of M and S and Chapter 4. Economic Consequences of Technological Progress 123 t is defined as year — 1982. Net investment, NI, by definition, is the difference between total fixed investment and replacement investment. Since the latter is not available in the data set, the Capital Cost Allowance (CCA), deflated by its implicit price deflator, is used instead. Thus NI = 1- CCA. The equations are estimated by ordinary least square regression. A constant term is added into each equation to account for non-output effects (e.g., price, interest rates, etc.) and the lagged variables are fitted using third-degree Almon Polynomials without end-point constraints. The results of the regression are listed in Table 4.9. Comparing these regression equations yields the following observations. (i) The addition of the term e • (AF) raises the R2 from 0.31 for Equation (4.32) to 0.50 for Equation (4.34), which is a substantial contribution. Also, the estimated coefficient e is positive and significant, indicating that the aggregate technological coefficient a indeed evolves over time. Since the sum of the estimates of /3r is 1.25, it can be calculated that _C/3T + e • t varies from 0.47 in 1948 to 1.34 in 1986, indicating that the production of the aggregate output requires more capital today than it did before. However, from the results of Equation (4.34), it is not possible to tell to what extent this increasing usage of capital is caused by the structural shift which puts larger weight on service production or by the fundamental change in technology. (zi).When output Y is disaggregated into M and S, the results of Equation (4.35) show that £ bT < £ c r , which implies that expansion of service industries requires more capital investment than the expansion of manufacturing industries. This result is clearly in line with the assump-tion that service industries are more capital intensive. Chapter 4. Economic Consequences of Technological Progress 124 Table 4.9: Estimated Equations of Investment Function Aggregate output Aggregate output Disaggregate output without trend* with trend** with trend Coefficient Estimates S.E.t Estimates S.E. Estimates S.E. a 54.48 20.22 60.73 17.68 36.29 23.78 e — 0.023 0.007 — em — — 0.006 0.020 e3 — — 0.018 0.023 3o 0.18 0.11 0.30 0.10 — ft 0.25 0.09 0.23 0.08 — ft 0.23 0.07 0.20 0.06 — ft 0.20 0.10 0.22 0.09 — ft 0.25 0.13 0.31 0.11 — bo — — 0.46 0.47 — 0.33 0.33 b2 — — 0.10 0.27 b3 — — -0.17 0.28 bA — — -0.41 0.29 c 0 — — -0.20 0.62 Cl — — 0.05 0.50 C2 — — 0.37 0.39 c 3 — — 0.77 0.48 c 4 1.25 0.54 Eft 1.11 1.25 — — 0.31 E cT 2.25 R-Square 0.31 0.50 0.73 R. A.J 0.85 0.62 0.68 * See Equation (4.32) ** See Equation (4.34) *** See Equation (4.35) t Standard Error % Residual Autocorrelation Chapter 4. Economic Consequences of Technological Progress 125 (iii) From Equation (4.35), the estimated trend in technological change for industry M, em, is 0.006 while the estimated trend for S industry, e", is 0.018. It should be noted that the small value of em does not necessarily imply that there has been little technological change in manufacturing industries. Rather, it can mean that the technological change may have been labour saving in nature. (iv) Comparing the results of Equation (4.34) with those of Equation (4.35), we see that the weighted sum of the trends, em and e", in Equation (4.35) on average is mAMt ASt which is only about 60% of the aggregate trend, e, in Equation (4.34). Thus, it suggests that about 40% of the apparent technological change is actually caused by the structural shift in the economy, i.e., the change in output composition, rather than the change in technology. (v) While decomposing aggregate output Y into M and S enhances the explanatory power of the model as evidenced by an increase of R2 from 0.50 in Equation (4.34) to 0.73 in Equation (4.35), it suffers from higher standard errors of coefficients. The standard errors in Equation (4.35) are several times larger than those in Equation (4.34). This is likely caused by the collinearity between output levels of M and S. Although the share of S in total output has markedly outgrown the share of M, in terms of absolute volume, both M and S have been growing steadly. This problem of collinearity between the levels of components of aggregate output seriously impairs the empirical application of the disaggregate output model. Nevertheless, the model does provide useful information and reveals the influences of the structural shift in the economy. Chapter 4. Economic Consequences of Technological Progress 126 4.6 Conclusion The discussions in this essay clearly show that the economic consequences of technologi-cal progress depend on the nature of technological change, and that the long-run effects and short-run effects of technological change can be quite different. For example, in the long-run, technological progress in manufacturing industries generally brings about more production of manufactured goods; however, labour saving technological progress in the short-run may temporarily depress goods production. Intuitively, labour saving technological progress releases labour from the manufacturing sector. To clear labour markets, more capital must be produced, which requires a capital flow from the con-sumption goods industry to the capital goods industry. As a result, the production of consumption goods may temporarily be depressed. On the other hand, the total sales of manufactured goods will generally decline as a result of technological progress, yet a capital saving technological change may instead expand the total sales of goods in the short-run. This can be explained as follows: Since capital saving technological change reduces the required amount of capital for given output, the capital industry will shrink in the short-run, providing resources of labour and capital which are used to produce consumption goods and services. Consequently, increased production may raise the total sales of goods even though the price of goods is reduced. The examples above reveal that the economic consequences of technological progress are in general very complicated. Hence, any useful prediction regarding the consequences of technological change must be made on an empirical basis. In this aspect, an exami-nation of the historical data on U.S. national accounts proves helpful in providing back-ground information. From 1948 to 1986, the growth rate of productivity measured by the ratio of output to labour input in manufacturing differed substantially from that in ser-vice industry (See Barras, 1986, and Kendrick, 1987). In 1948, the output/labour-input Chapter 4. Economic Consequences of Technological Progress 127 ratio in goods producing industries was 21.3 thousand dollars per employee (measured in 1982 constant dollar); in 1986, this ratio was 45.9 thousand dollars per employee, more than double. However, the output/labour-input ratio of service industries was 24.4 thou-sand dollars per employee in 1948 and was 39.5 thousand dollars per employee in 1986. The increase was only about 62%. Moreover, during the same period, the relative price index of pm/ps, i.e., the ratio of the price index of manufactured goods to the price index of services declined by more than 46%. Therefore, based on these facts, we may conclude that the manufacturing industries have been undergoing active technological progress in the past four decades. On the other hand, even though the real-output/labour-input ratio has doubled itself, the cost share of labour, defined as the ratio of labour cost—wages and salaries—to the value of output, in the manufacturing industries was somewhat stable. In 1948, this cost share of labour was 50%. It varied from 49% to 54% thereafter and in 1986 it was 52%. Considering that the wage rate of labour increased about 730% while the price of capital (measured by implicit price deflator of capital investment) increased by 370%, we may reasonably conclude that the technological change in manufacturing industries has been relatively labour saving (See also Barras, 1986, and Watanabe, 1986). However, this sort of technical change cannot last forever — full automation clearly sets a limit. Hence, future technological progress should eventually shift toward capital saving or neutral. In Section 4, the simulations based on alternative assumptions regarding the charac-teristics of technology in service industries show mixed predictions concerning the accu-mulation of capital stock. Specifically, for the case of a neutral technological change in manufacturing, if service industries are relatively capital intensive, the total capital stock will increase; while if service industries are relatively labour intensive, the total capital stock will decline (Computation is based on the assumption of constant labour supply, Chapter 4. Economic Consequences of Technological Progress 128 in fact, incorporating expanding labour supply in the models would lead to similar pre-dictions with respect to the overall capital stock/labour supply ratio). Unfortunately, data on disaggregate capital stock in each industry are not available, hence, the relative capital intensiveness of manufacturing industries versus service industries is not directly verifiable. However, the data on fixed business investment provide indirect evidence supporting the hypothesis that service industries are indeed more capital intensive than manufacturing industries. The conventional accelerator model of investment can be easily modified to incor-porate the effects of structural shift, i.e., the changing composition of aggregate output. Regression results of the augmented accelerator model reveal that a simple decomposition of aggregate output increased the explanatory power (measured in the value of R2) of the model by almost 50%. The sum of the coefficients of distributed lags of service out-put is larger than that of goods output, implying that an expansion of service industries may require more capital investment than an expansion of goods producing industries. Furthermore, it is found that the composite output has shifted toward more use of cap-ital in production. Since only part of this tendency is explained by the estimated trend in technological change, the remainder must be attributed to the changing composition of the aggregate output. The fact that the aggregate output has become more service intensive necessarily lends support to the basic assumption that service industries are relatively capital intensive. On the whole, empirical evidence appears to show that technological progress in manufacturing industries has been significant; that the type of technological change has been principally labour saving; and that service industries are indeed capital intensive. Based on these facts, the model presented in this essay predicts a boom of service output along with an expansion of investment, which is consistent with historical observations. However, technological change cannot be labour saving forever, and the model would Chapter 4. Economic Consequences of Technological Progress 129 predict a decline in investment for other types of technical change. Whether or not a decline in investment would eventually occur when technological progress becomes either capital saving or neutral remains a question for the future. Bibliography [1] Barras, R. 1986. A Comparison of Embodied Technical Change in Services and Manufacturing Industry, Applied Economics, 18:941-958. [2] Bhattacharya, S. and D. Mookherjee. 1986. Portfolio Choice in R&D, Rand Journal of Economics, 17:594-605. [3] BischofT, C. 1971. Business Investment in 1970s: A Comparison of Models, Brookings Papers on Economic Activity, (1):13—58. [4] Bound, J., Cummins, C , Griliches, Z., Hall, B., and A. Jaffe 1984. Who Does R&D and Who Patents? In Z. Griliches (eds.) R&D, Patents, and Productivity, Chicago: Univ. of Chicago Press. [5] Brainard, W.C., Shoven, J.B. and L. Weiss. 1980. The Financial Valuation of the Return to Capital, Brookings Papers on Economic Activity, (2):453-502. [6] Brander, J. A. and T. R. Lewis. 1986. Oligopoly and Financial Structure: the Lim-ited Liability Effect, American Economic Review, 76:956-970. [7] Brander, J. A. and T. R. Lewis. 1988. Bankruptcy Costs and the Theory of Oligopoly, Canadian Journal of Economics, 21:221-243. [8] Brander, J. A. and B. J. Spencer. 1983. Strategic Commitment with R&D: the Symmetric Case, Bell Journal of Economics, 14:225-235. [9] Brander, J. A. and B. J. Spencer. 1989. Moral Hazard and Limited Liability: Impli-cations for the Theory of the Firm, International Economic Review, forthcoming. 130 Bibliography 131 [10] CITIBASE: Citibank Economic Database. 1946—, New York, Citibank, N.A. 1978. [11] Clark, P. 1979. Investment in the 1970s: Theory, Performance, and Prediction, Brookings Papers on Economic Activity, (1):73—113. [12] Cohen, W.M., Levin, R.C., and D.C. Mowery. 1987. Firm Size and R&D Intensity: A Re-examination, Journal of Industrial Economics, 35:543-565. [13] Comanor, W.S. 1967. Market Structure, Product Differentiation, and Industrial Re-search, Quarterly Journal of Economics, 81:639-657. [14] Dasgupta, P. and J. Stiglitz. 1980a. Uncertainty, Industrial Structure, and the Speed of R&D, Bell Journal of Economics, 11:1-28. [15] Dasgupta, P. and J. Stiglitz. 1980b. Industrial Structure and the Nature of Innovative Activity, Economic Journal, 90:266-293. [16] Dukes, R.E., Dyckman, T.R., and J.A. Elliott. 1980. Accounting for Research and Development Costs: The Impact on Research and Development Expenditure, Jour-nal of Accounting Research, 18(Supplement):l-26. [17] Evenson, R.E. 1984. International Invention: Implications for Technology Market Analysis, In Z. Griliches (eds.) R&D, Patents, and Productivity, Chicago: Univ. of Chicago Press. [18] Gertler, M.S. 1986. Regional Dynamics of Manufacturing and Non-Manufacturing Investment in Canada, Regional Studies, 20:523-534. [19] Gort, G., and R.A. Wall. 1986. The Evolution of Technologies and Investment in Innovation, Economic Journal, 96:741-757. Bibliography 132 [20] Griliches, Z. 1984. Market Value, R&D, and Patents. In Z. Griliches (eds.) R&D, Patents, and Productivity, Chicago: Univ. of Chicago Press. [21] Griliches, Z. 1986. Productivity, R&D, and Basic Research at the Firm Level in the 1970's, American Economic Review, 76:141-154. [22] Grossman, G. M. and C. Shapiro. 1986. Optimal Dynamic R&D Programs, Rand Journal of Economics, 17:581-593. [23] Hamberg, D. 1964. Size of Firm, Oligopoly, and Research: The Evidence, Canadian Journal of Economics, 30:62-75. [24] Horwitz, B.N., and R. Kolodny. 1980. The Economic Effects of Involuntary Uni-formity in the Financial Reporting of R&D Expenditures, Journal of Accounting Research, 18(suppliment):38-74. [25] Hughes, K.S. 1986. Exports and Innovation, European Economic Review, 30:383-399. [26] Jaffe, A.B. 1986 Technological Opportunity and Spillovers of R&D: Evidence from Firm's Patents, Profits, and Market Value, American Economic Review, 76:984-.1001. [27] Jensen, M. C. and W. H. Meckling. 1976. Theory of the Firm: Managerial Behaviour, Agency Costs and Ownership Structure, Journal of Financial Economics, 3:305-360. [28] Johansen, L. 1967. Some Theoretical Properties of a Two-Sector Model of Optimal Growth, Review of Economic Studies, 34:125-141. [29] Jones, R. 1965. The Structure of Simple General Equilibrium Models, Journal of Political Economy, 73:557-572. Bibliography 133 [30] Jorgenson, D.W. 1971. Econometric Studies of Investment Behavior: A Survey, Journal of Economic Literature, 9:1111-1147. [31] Kamien, M. and N. Schwartz. 1975 Market Structure and Innovation: A Survey, Journal of Economic Literature, 13:1-37. [32] Kendrick, J.W. 1987. Service Sector Productivity, Business Economics, 22(2): 18-24. [33] Kim, M. and V. Maksimovic. 1987. Technology, Debt, and the 'Underinvestment Problem', Working Paper, University of British Columbia. [34] Klette, T. and D. de Meza. 1986. Is the Market Biased against Bisky R&D? Rand Journal of Economics, 17:133-139. [35] Kraus, A. and R. Litzenberger. 1973. A State-Preference Model of Optimal Financial Leverage, Journal of Finance, 28:911-922. [36] Kuznets, S. 1930. Secular Movements in Production and Prices, Boston and New York: Houghton Mifflin. [37] Lee, T. and L. L. Wilde. 1980. Market Structure and Innovation: A Reformulation, Quarterly Journal of Economics, 94:429-436. [38] Levin, R.C., Cohen, W.M., and D.C. Mowery. 1985. R&D Appropriability, Oppor-tunity, and Market Structure: New Evidence on Some Schumpeterian Hypotheses, American Economic Review, 75:20-24. [39] Lindenberg, E.B. and S.A. Ross. 1981. Tobin's Q Rates and Industrial Organization, Journal of Business, 54(January): 1-32. [40] Link, A.N. and J.L. Neufeld. 1986. Innovation versus Imitation: Investigating Al-ternative R&D Strategies, Applied Economics, 18:1359-1363. Bibliography 134 [41] Long, M.S. and I.B. Malitz. 1985. Investment Patterns and Financial Lever-age, in B.M. Friedman (eds.) Corporate Capital Structure in the United States, Chicago:Univ. of Chicago Press. [42] Loury, G. C. 1979. Market Structure and Innovation, Quarterly Journal of Eco-nomics, 93:395-410. [43] Lucas, R.E. Jr. 1967. Optimal Investment Policy and the Flexible Accelerator, In-ternational Economic Review, 8:78-85. [44] Maddison, A. 1980. Economic Growth and Structural Change in the Advanced Countries, in I. Leveson and J.W. Wheeler (eds.) Western Economies in Transi-tion: Structural Change and Adjustment Policies in Industrial Countries, Colorado: Westview Press, 41-60. [45] Mansfield, E. 1980. Basic Research and productivity Increase in Manufacturing, American Economic Review, 70:863-873. [46] Mansfield, E. 1981. Composition of R&D Expenditures: Relationship to Size of Firm, Concentration, and Innovative Output, Review of Economics and Statistics, 63:610-615. [47] Mansfield, E. 1984. R&D and Innovation: Some Empirical Findings, in Z. Griliches (eds.) R&D, Patents, and Productivity, Chicago: Univ. of Chicago Press. [48] Mansfield, E., Romeo, A. and S. Wagner. 1979. Foreign Trade and U.S. Research and Development, Review of Economics and Statistics, 61:49-57. [49] Mansfield, E., Schwartz, M. and S. Wagner. 1981. Imitation Costs and Patents: An Empirical Study, Economic Journal, 91:907-918. Bibliography 135 [50] Mansfield, E., Teece, D. and A. Romeo, 1979. Overseas Research and Development by U.S. Firms, Economica, 46:187-196. [51] Modigliani, F. and M. Miller. 1958. The Cost of Capital, Corporation Finance, and the Theory of Investment, American Economic Review, 48:261-297. [52] Modigliani, F. and M. Miller. 1963. Corporate Income Taxes and the Cost of Capital: a Correction, American Economic Review, 53:433-443. [53] Mohnen, P.A., Nadiri, M.I. and LR. Prucha. 1986. R&D, Production Structure and Rates of Return in the U.S., Japanese and German Manufacturing Sectors, European Economic Review, 30:749-771. [54] Myers, S. 1977. Determinants of Corporate Borrowing, Journal of Financial Eco-nomics, 5:147-175. [55] Pakes, A. 1984. Patents, R&D, and the Stock Market Rate of Return: A Summary of Some Empirical Results, in Z. Griliches (eds.) R&D, Patents, and Productivity, Chicago: Univ. of Chicago Press. [56] Pakes, A. 1985. On Patents, R&D, and the Stock Market Rate of Return, Journal of Political Economy, 93:390-409. [57] Pakes, A. and Z. Griliches. 1984. Patents and R&D at the Firm Level: A First Look, in Z. Griliches (eds.) R&D, Patents, and Productivity, Chicago: Univ. of Chicago Press. [58] Pakes, A. and M. Schankerman. 1984. The Rate of Obsolescence of Patents, Research Gestation Lags, and Private Rate of Return to Research Resources, in Z. Griliches (eds.) R&D, Patents, and Productivity, Chicago: Univ. of Chicago Press. Bibliography 136 [59] Patton, S.G. and B.J. Reilly. 1987. The Role of Private Services in the American Economy, Business Economics, 22(2):7—11. [60] Pavitt, K., Robson, M. and J. Townsend. 1987. The Size Distribution of Innovating Firms in the U.K.: 1945-1983, Journal of Industrial Economics, 35:297-316. [61] Scherer, F.M. 1965. Size of Firm, Oligopoly, and Research: A Comment, Canadian Journal of Economics, 31:256-266. [62] Scherer, F.M. 1967. Market Structure and the Employment of Scientists and Engi-neers, American Economic Review, 57:524-531. [63] Scherer, F.M. 1980. Industrial Market Structure and Economic Performance, 2nd ed. Chicago: Rand-McMally. [64] Scherer, F.M. 1982. Demand-pull and Technological Invention: Schmookler Revised, Journal of Industrial Economics, 30:225-237. [65] Schmookler, J. 1966. Invention and Economic Growth, Cambridge, Mass.: Harvard Univ. Press. [66] Scott, J.T. 1984. Firm versus Industry Variability in R&D Intensity, in Z. Griliches (eds.) R&D, Patents, and Productivity, Chicago: Univ. of Chicago Press. [67] Shapiro, M.D. 1986. The Dynamic Demand for Capital and Labor, Quarterly Journal of Economics, 101:513-542. [68] Titman, S. 1984. The Effect of Capital Structure on a Firm's Liquidataion Decision, Journal of Financial Economics, 13:137-151. [69] Titman, S. and R. Wessels, 1988. The Determinants of Capital Structure Choice, Journal of Finance, 43:1-19. Bibliography 137 [70] Tobin, J. and W.C. Brainard. 1977. Asset Market and the Cost of Capital, in B. Balassa and R. Nelson (eds.) Economic Progress, Private Values, and Public Policy: Essays in Honor of William Fellner. North-Holland. [71] Tschetter, J. 1987. Producer Service Industries: Why Are They Growing So Rapidly? Monthly Labour Review, 110:31-40. [72] Uzawa, H. 1962. On a Two-Sector Model of Economic Growth, Review of Economic Studies, 29:40-47. [73] Uzawa, H. 1963. On a Two-Sector Model of Economic Growth II, Review of Eco-nomic Studies, 30:105-127. [74] Uzawa, H. 1964. Optimal Growth in a Two-Sector Model of Capital Accumulation, Review of Economic Studies, 31:1-24. [75] Varaiya, P. and M. Wiseman. 1981. Investment in Manufacturing in U.S. Metropoli-tan Areas 1960-1976, Regional Science and Urban Economics, 11:431-469. [76] Watanabe, S. 1986. Labour-Saving versus Work-Amplifying Effects of Micro-electronics, International Labour Review, 125:243-259. [77] Williams, J. 1987. Perquisites, Risk, and Capital Structure, Journal of Finance, 42:29-48. 

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:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0098182/manifest

Comment

Related Items