Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Private governments, public authority : homeowners’ associations and their impact on local public finance Cheung, Ronnie King Gi 2005

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

Item Metadata

Download

Media
831-ubc_2006-129824.pdf [ 9.38MB ]
Metadata
JSON: 831-1.0092783.json
JSON-LD: 831-1.0092783-ld.json
RDF/XML (Pretty): 831-1.0092783-rdf.xml
RDF/JSON: 831-1.0092783-rdf.json
Turtle: 831-1.0092783-turtle.txt
N-Triples: 831-1.0092783-rdf-ntriples.txt
Original Record: 831-1.0092783-source.json
Full Text
831-1.0092783-fulltext.txt
Citation
831-1.0092783.ris

Full Text

PRIVATE G O V E R N M E N T S , PUBLIC AUTHORITY: HOMEOWNERS' ASSOCIATIONS AND THEIR IMPACT ON L O C A L PUBLIC FINANCE by R O N N I E K I N G G l C H E U N G B.Sc. (Hons.), The University of Alberta, 1998 M . A . , The University of British Columbia, 1999 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 M E N T OF 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 OF 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 (Economics) T H E U N I V E R S I T Y OF BRITISH C O L U M B I A November 2005 ©Ronnie King G i Cheung, 2005 Abstract This dissertation examines the impact of homeowners' associations, an increasingly popular innovation in public service provision, on local governments and city residents. A n introductory chapter (Chapter 2) outlines the increasing role of the homeown-ers' association in urban housing development. It overviews the history, motivation and governance of homeowners' associations in American housing markets, and it compares them to other methods of collective decision-making. A s local governments have trans-ferred public authority to these private associations, homeowners' associations can be' considered a form of residential private government. The goal of the three papers of the thesis is to examine the interactions between res-idential private governments, traditional l o c a l governments and city, residents. The first paper (Chapter 3) identifies the impact that homeowners' association membership has on local government expenditures. A key contribution is the construction of a thirty-year panel data set of homeowners' associations in California. Est imation results suggest that local governments have lowered their expenditures in response to the increasing mem-bership in private governments. However, the response differs depending on the public service considered. Local governments download services that are highly substitutable by private providers, such as garbage collection and parks, but they do not download services wi th public good aspects, such as roads and government administration. The second paper (Chapter 4) studies property tax limitations as a motivation for why homeowners' associations have become so popular. The paper is structured in two parts. In the first part, a theoretical model examines how the decision of whether to join a homeowners' association may be altered by the imposition of a property tax l imitation. In the second part, an empirical model tests the theoretical implications by using data on homeowners' associations in the era of California's Proposit ion 13. The third paper (Chapter 5) extends the canonical theoretical model of private gov-ernment by introducing a housing market. Equi l ibr ium is described in terms of the in-teraction between homeowners, the homeowners' association and the local government. The relative elasticities of housing and of public goods play a key role in interpreting the equilibrium conditions. i i Table of Contents Abstract i i Table of Contents i i i List of Tables v i i List of Figures ix Acknowlegements x i 1 Introduction and Overview 1 2 A Brief Survey of Residential Private Governments 6 2.1 Introduction 6 2.2 Homeowners' Associations as Private Governments 7 2.3 The Structure of Homeowners' Associations 9 2.4 The Rise of Homeowners' Associations 11 2.5 The Roles of Homeowners' Associations 15 2.6 Related Research on Homeowners' Associations 17 3 The Interaction Between Local Governments and Residential Private Governments 20 3.1 Introduction 20 i i i 3.2 Theoretical Framework 22 3.3 Empir ica l Specification 27 3.4 Data 31 3.4.1 Public Government Variables 31 3.4.2 Private Government Variables 34 3.4.3 Other Covariates 39 3.4.4 Instrumenting for Endogenous Variables 41 3.5 Analysis 44 3.5.1 Without Private Governments 44 3.5.2 Instrumental Variables Est imation 49 3.5.3 Are Planned Developments Different from Condominiums? . . . . 61 3.6 Further Directions 64 3.7 Conclusion . . . 67 4 The Effect of Property Tax Limitations on Residential Private Gov-ernments 68 4.1 Introduction 68 4.2 Property Tax Limitations and Private Government 70 4.2.1 Expenditure Cuts 71 4.2.2 State A i d • 71 4.2.3 Alternative Sources of Revenue . . . . 71 4.2.4 A Fourth Option: Private Providers . 72 4.3 Theoretical Model 73 4.3.1 The Housing Market 75 iv 4.3.2 Equi l ibr ium . . 76 4.3.3 Intercommunity Equi l ibr ium 77 4.3.4 Internal Equi l ibr ium 78 4.3.5 Computational Solutions 79 4.4 Background on Proposition 13 85 4.4.1 Local Responses to Prop 13 86 4.5 Empir ica l Model 89 4.5.1 Prop 13 in an Event Study Model 90 4.5.2 Publ ic Government Data 92 4.5.3 Private Government Data 94 4.5.4 Other Covariates 96 4.5.5 Analysis 96 4.5.6 Heterogeneity in the Prop 13 Treatment Variable 102 4.6 Conclusion 110 5 Private Government and the Urban Housing Market 111 5.1 Introduction I l l 5.2 Private Governments and Housing Markets 112 5.3 Mode l 112 5.3.1 Players . 113 5.3.2 T iming 115 5.3.3 Stage 2: Clearing of the Housing Market 116 5.3.4 Stage 1: Nash Equi l ibr ium 119 5.4 A Numerical Example 126 v 5.4.1 Stage 2, . 1 2 6 5.4.2 Stage 1 126 5.4.3 Some Numerical Results 130 5.5 Conclusion 133 6 Conclusion 135 Bibliography 137 vi List of Tables 2.1 Share of Homeowners Paying Homeowners' Association Fees . 13 2.2 Share of Homeowners Paying Homeowners' Association or Con-dominium/Cooperative Fees . 14 2.3 Comparison of Two California Planned Developments 15 2.4 Share of Occupied Units in Gated Communities 16 3.1 Summary Statistics of Planned Development Measures 38 3.2 Correlations Between HOA Membership Rate and Selected Co-variates, Selected Types 41 3.3 City Land Area Descriptive Statistics 45 3.4 Summary of Variables . . . 46 3.5 Fixed-Effect Estimation Results. Private Government Measured by Planned Developments 48 3.6 IV Estimation Results - A l l Covariates. Private Government Measured by Planned Developments 50 3.7 IV Estimation Results for Individual Expenditure Categories. Private Government Measured by Planned Developments . . . 54 3.8 Fixed-Effect Estimation Results for Individual Expenditure Cat-egories. Private Government Measured by Planned Develop-ments 55 vii 3.9 IV Estimation Results for Individual Expenditure Categories, Measured Per Non-Planned Development Household. Private Government Measured by Planned Developments 58 3.10 IV Estimation Results for Current Expenditures Versus Capital Outlays. Private Government Measured by Planned Develop-ments 59 3.11 Summary Statistics of Condominium Measures 62 3.12 IV Estimation Results for Individual Expenditure Categories. Private Government Measured by Condominiums 65 3.13 IV Estimation Results for Individual Expenditure Categories. Private Government Measured by Condominiums 66 4.1 Number of Cities in the Survey of Governments 93 4.2 Summary Statistics for Private Government Variables . . . . . . 95 4.3 Fixed-Effects Estimation Results for HO A Membership Rate 98 4.4 Fixed-Effects Estimation Results for HOA Membership Change 99 4.5 Poisson Estimation Results for New HOA Incorporations . . . . 100 4.6 Negative Binomial Estimation Results for New HOA Incorpo-rations 101 4.7 Fixed-Effects Estimation Results with Pre-Limit Interactions. Dependent Variables Are HOA Membership Rate and Change in HOA Rate 108 4.8 Fixed-Effects Estimation Results with Pre-Limit Interactions. Dependent Variables Are New HOA Incorporations 109 5.1 Numerical Model: Public Property Tax, Private Fee Financing. Some Comparative Statics 131 vi i i List of Figures 3.1 Mean Level of Local Expenditures in the 110-City Panel . . . . 33 3.2 M e a n Categorized Expenditure in the 110-City Panel 34 3.3 Growth in H O A s in California - Entire State, A l l Types . . . . 36 3.4 Overall Membership Rates in Planned Developments and Con-dominiums in the 110-City Panel 39 3.5 Rate of Planned Development Membership by County 40 3.6 Capital Outlays Versus Current Operating Expenditures . . . . 60 3.7 Rate of Condominium Membership by County 63 4.1 Computational Model: Equi l ibrium Membership Decisions . . . 83 4.2 Computational Model : Changing the Property Tax Rate . . . . 84 4.3 Histogram of Property Tax Rates in 411 California Cities . . . . 87 4.4 Quintile Plots by Pre-Limit Property Tax Rate Heterogeneity . 105 4.5 Quintile Plots by Pre-Limit Property Tax Dependence Hetero-geneity 106 4.6 Quintile Plots by Pre-Limit Crime Rate Heterogeneity 106 5.1 • Numerical Model : Exogenous Housing Prices. Weibull Distri-bution of Incomes. Median Voter Is a Member 132 ix Numerical Model: Endogenous Housing Prices. Uniform Dis-tribution of Incomes. Median Voter Is Not a Member x Acknowledgement s I gratefully acknowledge Robert Helsley, my thesis supervisor, for his expert guidance and support throughout my time at U B C . Thanks, Bob, for sharing your enthusiasm and talent for urban economics wi th me, and for motivating me in a way that only .a great supervisor can. I also sincerely thank my co-supervisor Patrick Francois and my thesis committee member K e v i n Mi l l igan for their open doors and continual support. They have never failed to astound me with their insights and their patience. I thank the Canadian Social Science and Humanities Research Counci l for their generous support during the first four years of my P h . D . This P h . D . was not a solo project. I have benefited greatly from discussions, ideas and the usually-gentle crit icism of my fellow graduate students. Thank you for putting up wi th my lapses of intelligence and my inglorious moments. Particular thanks go to Aneta Bonikowska, Michael Coell i , L i l i a Karnizova, Stephanie McWhinnie , Kazutaka Takechi, Genevieve Verdier and participants at departmental workshops. A n d to my ever-suffering office mate Jacob Wong: Jake, I owe you. There, we're even. Finally, this thesis is dedicated to M o m , Dad and Luanne. Your love and encourage-ment provided me wi th the wi l l , the desire and the power to accomplish this goal. x i CHAPTER 1 Introduction and Overview In recent years, cities have faced increasing fiscal and polit ical pressures as they try to provide public services for their residents. A s these pressures escalate, the formation of residential private governments is an innovation in local public finance that is blurring the distinction between what is public and what is private. In the residential setting, these private governments take the form of homeowners' associations, and they are found in planned developments, condominiums and cooperatives. Local governments transfer public authority to the boards of private associations by giving them government-like powers in service provision, taxation and enforcement. A s an institution, the homeowners' association has characteristics in common with other forms of collective decision-making. For instance, like a club, individuals who want the services the association provides pay to join. The services that the association provides are usually restricted to members. However, unlike a club, membership in a homeowners' association is tied to housing choice. If a homeowner chooses to buy a house in a development that is governed by a homeowners' association, membership is compulsory and automatic. This implies a similarity to local government, where citizenship ("membership" in the city) is also directly linked to locational choice. The homeowners' association also shares characteristics wi th those of private cor-porations, and so it may be viewed as a manifestation of the privatization movement. Stabile (2000) notes that although homeowners' associations are usually established as nonprofit corporations with an elected governing board, they "are a product made and 1 sold by businesses for profit,..., a corporation, a community, and a lifestyle." The objec-tive of the association is to maximize the welfare of its members, whether by providing excellent public services or by maintaining high property values. O n the other hand, sometimes the objectives of the homeowners and those of the governing board do not match precisely. This may lead to well-publicized instances of neighbour against neigh-bour, where homeowners claim they are harassed by the association's stringent rules and regulations. Finally, the homeowners' association shares many characteristics wi th local govern-ments. It can tax homeowners for membership dues and can enforce payment through civ i l law. The board has the power to impose conditions and regulations not only on allowable architecture and landscaping, but also on acceptable behaviour. The charac-teristic most reminiscent of local government is that the association provides goods and services similar to those provided by a city. These include sanitation, policing, recre-ation facilities and many others. However, these services are meant to supplement the existing provision by the municipal government as most associations are located within city boundaries. This means that rather than considering associations as simply another layer of government, it is appropriate to analyse their interaction wi th the existing local government. The homeowners' association's popularity has soared to the point where an estimated half of all new housing in the United States includes membership in one. The prolif-eration is phenomenal. A few hundred homeowners' associations existed in the United States in the 1960s; their number climbed to 249,000 by 2003. The Community Asso-ciations Institute (2005) states that, as of 2005, these private associations now govern roughly 54.6 mil l ion Americans in 22 mill ion housing units. The creation of private governments arose from a combination of cash-strapped local governments and profit-minded real estate developers. But the rise of these institutions call into question the impact that they have on traditional municipal governance and city residents. There is fierce debate about the benefits of private governments in housing develop-ment. Supporters argue that they fill the gaps left by underfunded and inefficient local governments. They claim that private governments reflect the self-help attitude of their members, who pay for and receive local services that would otherwise not be provided. Proponents also contend that transferring some responsibility to private associations 2 may free up public resources to be used elsewhere, so that everyone, even non-members, can benefit. O n the other hand, critics of private governments claim that homeowners' associa-tions erode support for public institutions. Those who can afford to join can bypass the public system: homeowners who fear crime do not have to vote for tax dollars to attack the root of the problem, they can build a gate to keep the criminals out. Opponents maintain that the erosion of public support, reflected at the ballot box, leads to further deterioration of municipal services, hurting residents who do not have the resources to belong to private governments. Local governments, under pressure to cut expenditures, shift the responsibility of providing public services to private developers. Non-members experience a reduction in public service levels and may be worse off. This dissertation contributes to the understanding of the impact of residential pri-vate governments by examining their effects on local public finance. The dissertation begins with Chapter 2, which presents a brief survey of the private government. The chapter gives a short history of homeowners' associations and overviews their structure, governance and responsibilities. The chapter closes on a discussion of how this method of public service provision differs from other forms of private provision and justifies why it can be called a private government. W i t h the background survey providing a motivation for the dissertation, the next three chapters shed light on the impact of residential private governments by addressing three instances of interaction between the public and private sectors. In Chapter 3, an empirical analysis examines the interaction between private governments and traditional local governments. The overarching motivation for this paper is the theoretical model of private government by Helsley and Strange (1998), who view the public and private governments engaging in mutual provision of public services to residents of a city. The major result is "strategic downloading," where the public government, recognizing that a private government exists to serve member households, cuts back on its own services. Chapter 3 tests whether local governments in California have acted consistently wi th this theoretical hypothesis. Using a novel panel data set of California homeowners' associations, the paper esti-mates a fixed-effects model of public government expenditure. The question of interest is 3 whether cities that have higher rates of homeowners' association membership experience lower public expenditures. The study accounts for the endogeneity between public and private government activity wi th instrumental variables. The results support the view that local governments view private governments as substitute providers and so lower their expenditure accordingly. The study also finds that the types of expenditures pub-lic governments download to private governments depend on the substitutability of the service. For instance, parks and waste collection services are downloaded, but services wi th public-good aspects, such as roads and government administration, no downloading occurs. Interestingly, expenditures per non-private government member also falls, which may indicate that private government members are less wil l ing to support spending on services that do not directly benefit them. This may have important implications for those residents who do not have the desire, or the resources, to join private governments. Chapter 4 examines the impact of exogenous public government shocks on the for-mation and proliferation of private governments. In particular, the paper focuses on state-imposed property tax limitations. The desire to curb the perceived overspending and overtaxing of local governments culminated in the passing of property tax l imi -tations in over half the States in the 1970s and 1980s. Previous literature has shown that property tax limitations have induced local public sector response in many ways, including reductions in expenditures and service quality, increased reliance on charges and higher dependence on state aid. This chapter proposes that the downloading of provision responsibility to private governments was also a public government response. The motivation for the paper is the empirical observation that the years of the prop-erty tax revolt coincided wi th the period of the most dramatic growth in homeowners' association incorporation. The study is organized in two parts. In the first part, a the-oretical model focuses on residents of a city who must choose whether or not to join a private government. Publ ic and private government service levels by determined by the respective government budget constraints and the exogenous tax rates. Conditions for equilibrium are outlined, and a computational model provides the testable comparative static: as the property tax rate is exogenously lowered by the state, the proportion of homeowners who choose to purchase a house in a private government increases, notwith-standing the higher cost of housing within the private government. In the second part of Chapter 4, the California homeowners' association data are 4 again used to test the theoretical implication. In particular, the analysis examines the effect that Proposition 13, passed in 1978, had on homeowners' association membership, growth and creation. The empirical analysis takes the form of an event study. Results indicate that cities have responded to Proposition 13 by exhibiting higher rates of mem-bership and growth in homeowners' associations. The paper then shows mixed evidence that differences in pre-Proposition conditions can help explain why some cities experi-enced higher private government growth than others. In particular, pre-limit crime rate, which may be interpreted as a measure of demand for local public services, is a good pre-dictor of homeowners' association membership post-Proposition 13. This is consistent wi th the view that residents may have feared that Proposition 13 would severely reduce policing effectiveness, and the rise of planned developments and gated communities was a private-sector response. Chapter 5 examines the role of private governments in the housing market. The chapter begins first wi th a brief review on sorting models that combine jurisdictional choice wi th housing markets. This is then integrated wi th the literature on private governments and why housing markets may play an important role. A theoretical model is presented that is based on Helsley and Strange (1998), wi th the addition of a housing market into the interaction between homeowners, the private government and the public government. Equi l ibr ium conditions are characterized, wi th the elasticity of demand for public services playing a key role. A numerical model is presented to show some characteristics of the equilibrium. Finally, Chapter 6 concludes the dissertation. 5 C H A P T E R 2 A Brief Survey of Residential Private Governments 2.1 Introduction This chapter describes the concept of residential private government and addresses its relevance in local public finance. There is an emphasis on the California experience, as much of the subsequent analysis is set in the state, although homeowners' associations are very common in many other parts of the country. 1 The survey begins by defining the homeowners' association and introducing the con-cept of private government. This is followed by an overview of the motivation behind their formation and popularity. Then the survey examines what homeowners' associ-ations do. The chapter concludes by examining theoretical background research that serves as the motivation for the three papers in the dissertation. *It is est imated that F l o r i d a has the largest number of homeowners' associat ions i n the U n i t e d States, owing to i ts fast growth and h igh propor t ion of retirees. 6 2 . 2 Homeowners' Associations as Private Govern-ments A homeowners' association 2 is a body, found in certain housing developments, that manages property owned in common and charges fees for its provision. The association also establishes and enforces covenants and restrictions governing land use. Homeowners who buy property in the housing development must become members in the association. A homeowners' association is an example of a private government, which is any organization of private individuals, bound by geography, empowered by public authority to act as a government in providing public services, taxing members and enacting and enforcing regulations. The survey begins by highlighting five characteristics of a private government and by arguing that homeowners' associations satisfy them. 3 Along the way, it wi l l be possible to distinguish this type of institution from other forms of collective decision-making. First , membership in the private government should be voluntary. This appeals to the free mobility assumption common to multi-community local public good models based in Tiebout (1956). However, "voluntary" refers to the fact that if one does not want to join the private government, one can move away from its jurisdiction. If the public services provided by the private government are consumed, there is an obligation to participate in the private government. 4 In determining membership, geography plays the foremost role, hence its inclusion in the definition of private government. Once the boundaries of a private government are set, the services that it provides are available to all members. This characteristic makes the homeowners' association distinct from a club, which is less dependent on the residence of the club member. 2 T h e s e are also sometimes known as community associations ( C A s ) . Ano the r te rm that encompasses the governing associat ion and the member households is common interest developments (C IDs ) . T h i s d issertat ion uses these terms interchangeably. 3 T h e s e five t ra i ts were formulated by Hels ley and Strange (1998). 4 R e i c h m a n (1976) notes that there is a key legal issue in the regulat ion and operat ion of pr ivate governments. W h i l e munic ipa l i t ies are author ized to enforce payment of proper ty taxes and obedience of by- laws, how can inst i tu t ions of pr ivate ind iv idua ls do the same? T h e most common method of enforcement, Re i chman notes, is that the pr ivate government is ent i t led to place a l ien on the offending property. However, as Ma l l e t t (1993) states, the blurred l ine between pr ivate and publ ic means that pr ivate governments may lack the author i ty to per form these governing act iv i t ies. O n the other hand , th is legal s i tuat ion can be subject to abuses. A p r imary legislat ive goal is thus the effective reconci l ia t ion of pr ivate inst i tu t ions and pub l ic author i ty. 7 The next characteristic, which is closely related, is exclusivity. It must be possible to exclude the consumption of the supplementary services to non-members of the private government. A n extreme example is a residential gated community, which denies even the access to its streets to non-residents. B y their l imited geographic nature, exclusivity is a characteristic exhibited by many local public goods. For instance, homeowners' associations can readily ensure that their swimming pools, extra garbage pickup and security patrols are consumed only by their members. Th i rd , private governments provide services that are supplementary. Indeed, M a l -lett (1993) cites that private governments init ial ly arose because the public sector was perceived to be unresponsive to consumers' demands for recreational and leisure facili-ties. Another popular service provided is enhanced security, which supplements police services provided by the municipal government. Thus a key feature of a private govern-ment is that members remain residents of the municipality, and therefore continue to pay property taxes to the local government. 5 This characteristic differentiates homeowners' associations from simply another layer of local government, as there is no separation of powers between the public and private governments. Fourth, private governments should also be self-financing. This adds to the auton-omy that they enjoy. There are three main ways that private governments raise their revenues. They may require the payment of a fixed fee from each of its members. In many homeowners' associations, each homeowner pays a monthly membership fee, called an assessment. Other methods of financing include a surcharge on the property tax and user fees for the use of certain services. The revenues that homeowners' associations col-lect can be substantial, wi th the average monthly assessment for a US household being $180. ( C A I Research Foundation 1999) The budget of a homeowners' association can rival that of a small municipality, particularly in very large developments. Finally, a private government possesses is scope for strategic interaction with the public sector. It is this interaction that most differentiates the private government from other types of private sector provision. For instance, in most models where the consumer 5 However , not a l l resident ia l pr ivate governments coexist w i th pub l ic governments. T h i s is because they may be located in un incorpora ted terr i tory. W h i l e these represent a minor i t y of ex is t ing R C A s , they can be viewed as a developer 's decision to w i thdraw f rom in teract ing w i t h a t rad i t iona l loca l government. 8 has a choice between consuming a publicly or a privately provided public good 6 , the private provider of the service has no objective of its own. Rather, these models assume that consumers can freely purchase as much or as little of the private alternative as they desire. Indeed, in most models of local public good provision, even the traditional public government is often a passive agent. In each community, the median voter decides on the level of public good that should be provided, and the government's budget constraint and production technology imply automatically the level of taxation. Yet in the study of private governments there are some reasons why governments should be viewed as strategic players in the public good provision game. First , the nature of the services a private government provides prevents members from consuming as much or as little of the public good as they want. Residential asso-ciations sell themselves on their security patrols and their recreational facilities, and the decision of membership in a private government implies an all-or-nothing consumption of the package of services the government offers. Whi le it is conceivable that certain services such as home security or leisure facilities can be procured in variable amounts by individuals, the purpose of the private government literature is to examine what mo-tivates a group of private individuals to purchase services collectively. The second reason for considering strategic governments is to examine the widely-held belief that the ex-istence of a private alternative threatens the traditional public provision. B y modelling the private and the public governments as economic actors, it is possible to see if this shifting of responsibility is indeed motivated and justified. These characteristics imply that homeowners' associations are not identical to other methods of collective decision-making, but there is clear overlap. The following sections survey association structure and history, with the goal of arguing that they have evolved into an institution that may have important impacts on urban development. 2.3 The Structure of Homeowners' Associations Legally, a homeowners' association is found in developments in which homeowners have separate interests in their own unit and an undivided interest in common wi th other owners. The homeowners' association is responsible for managing the common property. 6 S e e , for example, E p p l e and R o m a n o (1996), G l o m m and R a v i k u m a r (1998) and G o u v e i a (1997) 9 There are three main types of housing developments wi th homeowners' associations: planned developments, condominiums and cooperatives. They differ in the structure of the housing and the way that ownership of the housing unit is defined. The type of development that has attracted the most controversy is the planned development. It usually consists of detached or semi-detached houses, although larger developments may also include some multi-family housing. In this case, the homeowner owns the house and the lot under it. The developer designs the entire community, often including streets, parks, security facilities and other common property; the developer then charges homeowners monthly assessments for the provision of the common facilities. Gated communities are planned developments, although not all planned developments have gates. A condominium is a set of units attached to each other in a complex. Homeowners own the units that they occupy. Each homeowner is a tenant in common ownership of the common property, which include common hallways. The homeowners' association manages the common hallways and amenities. In recent years, many apartment buildings have been converted into condominiums because in rising housing markets, it was often more profitable for building owners to sell the units than to rent them out. Conversion entails the creation of a homeowners' association. Finally, in cooperatives, homeowners do not own their unit, but have an exclusive right to occupy it. A corporation holds the title to the units and the common areas. Indeed, many of the first homeowners' associations were cooperatives. The Community Associations Institute (2005) estimates that of the 274,000 homeowners' associations in the United States as of 2005, planned developments account for 55-60% of the total, condominiums 35-40% of the total and cooperatives 5-7% of the total. The remainder consists of homeowners' associations in some mobile home parks, marinas and timeshare communities, but their total numbers are relatively small. How are homeowners' associations governed? A l l allow for self-government; their governing board consists mostly of elected homeowners. Usually, the board of directors of the homeowners' association consists entirely of volunteer homeowners. However, when a development is newly created, local authorities usually allow the developer to have representatives on the board. This is to protect the developer's interests. A s more and more properties in the development are sold, homeowners gradually replace the 10 developer's representatives on the board. Governing boards sometimes contract wi th professional management companies to oversee day-to-day issues, but the ultimate au-thority lies wi th the association. A s well, the vast majority of homeowners' associations are incorporated. 7 This means that the association board members are immune from individual suits from aggrieved homeowners. This last condition has led to one of the deepest criticisms of the homeowners' association: rather than leading to the a greater sense of community, critics contend that the power that the board has over individual homeowners often pits neighbour against neighbour. 2.4 The Rise of Homeowners' Associations The rise of the homeowners' association as a key feature of many present-day housing development can be attributed to several factors. The first is suburbanization. Housing developers took advantage of cheap land on the outskirts of cities and recognized the cost-efficiency and marketability of large-scale planned communities. Homeowners, par-ticularly in middle-income households, were looking for affordable ways to own their own home, and they wanted safe neighbourhoods wi th parks, schools and recreational facili-ties for their families. Roland (1998) states that developers addressed these demands by reducing lot sizes and providing common services, such as swimming pools and tennis courts, through the homeowners' association. A second factor is the growing heterogeneity in the demand for public services. W i t h i n a city, some neighbourhoods have a strong demand for police, while other neigh-bourhoods may demand leisure facilities. A city government that acts in the interest of the median voter is unlikely to please al l residents equally; whereas Bogart (1998) states that "the advantages of private associations are their efficient decision making and the responsiveness of the 'government' to local concerns." Stabile (2000) notes that residents in a homeowners' association may agree on a service level more readily, and because they contract for services directly, they may be able to seek out the most cost-effective ways to provide the services. A final factor in the rise of the homeowners' association is the constraints of local 7 I t is dif f icult to get an accurate number, but C h u l a k (2004) estimates that 99 per cent of Ca l i f o rn ia homeowners' associat ions are incorporated. 11 governments. The last thirty years have seen a period of fiscal restraint and backlash in local governments. Voters, whether spurred by ballooning deficits or inefficient bu-reaucracy, demanded change. Local governments grappled with drastic cuts in funding from higher levels of government or wi th strict limitations of property tax revenue im-posed by voters. A famous example of fiscal constraints imposed on local governments was Proposition 13 in California, which was approved by two-thirds of voters in 1978 and which slashed property taxes in half and capped further tax increases to two per cent. O n the other coast, Massachusetts voters approved Proposition 2 | in 1980. This initiative imposed an immediate property tax ceiling of 2.5 per cent of assessed value, and it also limited annual nominal increases of property tax to 2.5 per cent. 8 Fiscal burdens often meant reductions in public services. This made housing developments wi th privately-provided public services even more attractive. These factors made the private government a major force in urban governance. Be-cause of their private nature, accurate data on the number and nature of homeowners' associations are limited. However, some estimated figures can be given to motivate their growing relevance in urban settings. To begin, according to the 2002 US Census of Gov-ernments, there are 87,849 units of local governments in the United States. These include county governments, general purpose local governments, special purpose districts and school districts. In comparison, the Community Associations Institute estimates that in 2002 there were 231,000 residential private governments. Yet not one of these private governments is counted by the Census of Governments, while they all provide, to some extent, services that are also provided by the traditional public sector. The growth of homeowners' associations in housing development is driven by new construction. The Community Associations Institute (2003) says that half of all new housing construction in the United States incorporates a private government, while Bar-ton and Silverman (1994) state that this number approached 80 per cent in fast-growing suburban California counties. These figures are estimated and may be overstated, but their fundamental claim is borne out by the 2001 American Housing Survey. Tables 2.1 and 2.2 provide some summary values for this survey. Table 2.1 reports the percentage of owner-occupied households that reported paying a homeowners' association fee in 2001, 8 Cutler, Elraendorf, and Zeckhauser (1999) provides an excellent study of Proposition 2^, its effects on municipal financing and why it was overwhelmingly supported. One reason for its support is that voters perceived their high tax burden as a sign of the government was inefficiently run. 12 Table 2.1: Share of Homeowners Paying Homeowners' Association Fees ( A l l figures are in thousands unless otherwise indicated.) H o u s e h o l d s p a y i n g T o t a l o w n e r - P e r c e n t a g e M e d i a n H O A fees o c c u p i e d u n i t s a n n u a l fee A l l o w n e r -o c c u p i e d u n i t s 6 , 4 4 8 7 2 , 2 6 5 8 . 9 % $ 3 1 2 New vs Old Units N e w (< 4 y rs ) u n i t s 1,313 4 ,690 2 8 . 0 % $372 Region N o r t h e a s t . 462 12,987 3 . 6 % .$504 M i d w e s t 1,120 18,049 6 . 2 % $192 S o u t h 3,207 26 ,715 1 2 . 0 % $300 W e s t 1,659 14,514 1 1 . 4 % $432 Urban/Rural C e n t r a l C i t y 1,190 16,870 7 . 1 % $264 S u b u r b s 4 ,795 39 ,420 1 2 . 2 % $336 N o n - m e t r o 463 .15 ,975 2 . 9 % $264 Source: 2001 American Housing Survey - U S Bureau of the Census while Table 2.2 adds in those homeowners that pay a condominium or cooperative fee. Whi le nine per cent of al l American homeowners belong to a housing association, if only units less than four years old are considered, then 28 per cent of these households are members of a housing association. If condominiums and cooperatives are included in the sample, then 34 per cent of new housing, or 14 per cent of overall housing, belong to residential private governments. Regional differences also underline this proliferation. Residential private governments are most common in the South and the West. A s these regions are also the fastest-growing regions in the United States, it is not surprising that the percentage of new housing in private governments has increased so dramatically. Given the proliferation of homeowners' associations in the housing market, the next section outlines some of their key responsibilities. 13 Table 2 . 2 : Share of Homeowners Paying Homeowners' Association or Condo-minium/Cooperative Fees ( A l l figures are in thousands unless otherwise indicated.) H o u s e h o l d s p a y i n g T o t a l o w n e r - P e r c e n t a g e M e d i a n a n n u a l H O A o r c o n d o fees o c c u p i e d u n i t s H O A / c o n d o fee All owner-occupied units 10,209 72,265 14.1% $312/$2016 New vs Old Units N e w (< 4 y rs ) u n i t s 1,413 4 ,690 3 0 . 1 % $ 3 7 2 / $ 1 4 5 2 Region N o r t h e a s t 1,349 12,987 1 0 . 4 % $ 5 0 4 / $ 2 3 5 2 M i d w e s t 1,894 18,049 1 0 . 5 % $ 1 9 2 / $ 1 6 8 0 S o u t h 4 ,295 26 ,715 1 6 . 1 % $ 3 0 0 / $ 2 0 0 4 W e s t 2 ,669 14,514 1 8 . 4 % $ 4 3 2 / $ 2 1 1 2 Urban/Rural C e n t r a l C i t y 2,361 16,870 1 4 . 0 % $ 2 6 4 / $ 2 2 6 8 S u b u r b s 7,245 39 ,420 1 8 . 4 % $ 3 3 6 / $ 1 9 6 8 N o n - m e t r o 603 15,975 3 . 8 % $ 2 6 4 / $ 1 2 9 6 Source: 2001 American Housing Survey - US Bureau of the Census 14 Table 2.3: Comparison of Two California Planned Developments Rinconada Hills Leisure World City Los Gatos Laguna Woods Types of homes 394 townhouses and 40 single-family houses 12,736 housing units, consisting of single-family houses, townhouses and high-rises Target homeowners Families Adul ts 55 and over Selected amenities Clubhouse, pools, tennis courts, landscaped parks and ponds Six clubhouses, equestrian centre, shuttle buses, fitness and recre-ation classes, community college classes, libraries, al l exterior land-scaping Security services Traffic control service Guarded gate, 24-hour mobile pa-trols Source: Rinconada Hills and L W Laguna Woods web sites, 2004 2.5 The Roles of Homeowners' Associations The primary responsibility of homeowners' associations is to provide public services for members. These vary substantially from development to development, and they can be as simple or as lavish as the developer likes. The most basic is the provision and upkeep of common hallways and thoroughfares. Many also provide landscaping and sanitation, in the form of extra street cleaning or garbage disposal. Security is popular, ranging from patrols to gates. Access may also be restricted through the construction of private streets. To give an example of the variety of planned developments, Table 2.3 compares two sample communities in California. A brief note regarding the gated community is warranted here. The gated community is a form of planned development, where access to the community is restricted through a gate or barrier. These have become highly visible because the motivation for their popularity is rich households' fear of crime. Reich (1991) refers to this as the "secession of the successful," where the wealthy wall themselves in their secure compounds for fear of criminals and ineffective police. Figures from the American Housing Survey, shown in Table 2.4, demonstrate that while gated communities form a significant proportion of 15 Table 2.4: Share of Occupied Units in Gated Communities (Units, owned or rented, that report that "Community access is secured wi th walls or fences." A l l figures are in thousands unless otherwise indicated.) Units wi th Total number of Percentage wi th secured access occupied units secured access A l l occupied units 7,033 106,262 6.6% New vs Old Units New (< 4 yrs) units 821 5,853 14.0% Region Northeast 691 20,321 3.4% . Midwest 579 24,758 2.3% South 2,969 38,068 7.8% West 2,793 23,114 12.1% Urban/Rural Central C i ty 2,689 31,730 8.5% Suburbs 3,986 53,574 7.4% Non-metro 358 20,958 1.7% Source: 2001 American Housing Survey - US Bureau of the Census housing units (over ten per cent of new housing units in the West), most homeowners' associations are not gated. Interestingly, gated units are more common in central cities than in the suburbs. This may be due to the survey, which may regard condominiums with electronic access systems as a restricted access community. O n the other hand, it may be reflective of gated communities being popular in high-crime areas. To pay for the services, each homeowner must pay a monthly assessment. Generally, all homeowners pay the same monthly fee. The 2001 American Housing Survey finds that the median monthly assessment is $26 for all private governments. (American Housing Survey 2001) For condominiums in particular, the median monthly assessment is $168. This is comparable to a condominium owner's property tax l iabi l i ty . 9 Homeowners' associations can therefore have a considerable budget, and their spending may rival that of a local government. 9 C o n d o m i n i u m fees are typ ica l l y higher than p lanned development fees because condomin ium fees often pay for bu i ld ing maintenance and repairs. P lanned developments fees general ly go toward service and ameni ty provis ion. 16 Apar t from providing services, homeowners' associations also enact and enforce land use regulations. These are summarized as the association's covenants, conditions and restrictions ( C C & R s ) Their goal is to l imit the negative externalities neighbours' actions may impose; as Stabile (2000) states, "by joining a [community association], members have implici t ly agreed to a private contract to eliminate the social costs of neighborhood effects in a prescribed way." The C C & R s maintain order in the community by providing homeowners a way to resolve dispute among neighbours and to maintain a high standard of property values. 2.6 Related Research on Homeowners' Associations A n examination of economic research on private government reveals that there has been little theoretical research and hardly any empirical research on private governments. Perhaps this is due to the newness of the concept and the vague line between public and private. Theoretical research in economics have tended to group these governments in with more traditional models of club goods. However, as explained above, private governments have specific characteristics that distinguish them from other forms of pub-lic good provision. Besides, the tremendous growth of homeowners' associations as a method of public service provision means it is appropriate to consider them as interest-ing institutions in their own right. Research on private governments has instead been based largely in the poli t ical science and urban planning literature; as such, it is mostly descriptive and heavily based on case studies. This section gives a brief outline of the relevant economic research that the three papers in the dissertation use as background. The canonical theoretical motivation comes from Helsley and Strange (1998), which identifies the private government as a distinct institution. Its main contribution is highlighting the strategic interaction that takes place between consumers, the private government and the traditional public government. This strategic policy making is attested in the planning literature. Helsley and Strange consider only one urban area, where there is a local government providing a public good. The local government's objective is to maximize the aggregate welfare of the urban population. A private government maximizes the welfare of its members and provides a supplementary level of the public good for members. Helsley and Strange think of the model as a game, in 17 which the three players — residents, the public government and the private government — interact strategically in determining their actions. The principal implication is that the public government scales back its provision, better suiting the desires of nonmembers, in response to the existence of a private government. The existence of this "strategic downloading" demonstrates that viewing either the public or the private government as non-strategic, passive players misses a key element of the public good provision problem. In a followup paper, Helsley and Strange (2000) focuses on the formation of private governments due to cost differentials. Private governments are assumed to have lower costs of service provision than public governments. When this outweighs the social cost of forming the private government, it is optimal to have a private government forming. However, the public sector, seeking to protect its rent-seeking position, may have an incentive to block the formation of the private government. This may result in an inefficient outcome, where the private government ought to be formed but is not allowed, to. The interaction between public and private providers of a public good is not a new area of study. Papers in the study of local public goods, especially those that follow in the footsteps of the multi-community model of Tiebout (1956), have examined how equilibrium outcomes can change when there is a private provider. In Epple and Romer (1991), for example, consumers must choose either a public provider of health care or opt out in favour of private provision. If the consumer chooses private provision, he or she can consume as much or as little as desired. In contrast, private government studies, such as Helsley and Strange (1998) and Helsley and Strange (2000), usually assume that the private government produces a fixed amount of the public good offered to its members. That is, the governments act as separate players in this model, maximizing some objective function through its choice of public provision. There is good reason to think that this is an appropriate way to model private government activity, as they often provide local services that a consumer cannot purchase in variable amounts. Only by joining a private government can consumers enjoy the supplementary services they desire. The previous theoretical research agrees that a key motivation for the private govern-ment literature is the interaction between the private government, the public government and city residents. The following three chapters in the dissertation focus on three sep-18 arate sources of interaction involving private governments. Taken together, they show that the rise of the homeowners' association has substantial impacts on the city and on local public f i n a n c e . 19 C H A P T E R 3 The Interaction Between Local Governments and Residential Private Governments ...Homeowner associations in gated enclaves (and in the country's far more numerous un-gated private communities) are now pushing for more favorable tax treatment of the association fees homeowners pay for trash collection, street maintenance, recreational facilities, and other amenities. Some are lobbying for rebates of local property taxes, arguing that their residents do not use the full complement of local services. Others are seeking to make the fees deductible on state and federal income taxes, accorded essentially the same treatment as local property taxes. If these advocates succeed, gated communities will get a large and undeserved breath of life. - Steven Lagerfeld, http://www.civnet.org/journal/issue5/revlager.htm 3.1 Introduction This paper examines whether local governments alter their levels of public expenditure in response to increased membership in private residential governments. Previous the-oretical research has modelled the interaction between private and public governments, but empirical analysis has been l imited by the lack of data. This paper presents the first direct econometric study of the effects of private government prevalence on local 2 0 public finance. From a novel data set on homeowners' associations, this paper develops a measure of private government membership in California cities from 1970 to 1999 and employs panel data methods to test whether public and private government activities can be regarded strategic substitutes or strategic complements. The motivation for this chapter is that the tremendous growth of private government in new housing construction has engendered fierce debate about their effects on public finance. Supporters argue that private governments are a way for those homeowners who wish supplementary levels of public services to purchase them efficiently. Opponents, on the other hand, argue that the growth of private governments erode the support for local government and encourage the bypass of public provision. This paper sheds light on these controversies by making four contributions. Most importantly, this is the first direct empirical study of the effects of private government on public government behaviour. Previous studies relied on descriptive statistics or isolated case studies. This paper develops an empirical methodology to estimate the interaction effect and takes into account the endogeneity between public and private governments. Second, the paper generalizes the Helsley and Strange (2000) model, the theoretical framework for the empirical analysis. In the Helsley and Strange model, a public and a private government choose public good provision levels simultaneously. The supplementary nature of private provision gives rise to strategic interaction. In the Nash equilibrium, the public government scales back its provision level in response to the private government; that is, public and private government activities are strategic substitutes. This paper generalizes the theoretical model to allow for the possibility of strategic complements, and this provides a consistent framework in which the empirical findings may be applied. Th i rd , the study constructs a panel of California cities over a thirty-year period, merging novel private government data with local government finance data. The panel allows for examination of the geography, variety and growth of private governments. Finally, the empirical analysis shows that local governments in California have reacted to private government activity, consistent wi th strategic substitution. For the most part, this paper focuses on the effects of planned developments, as opposed to condominiums and cooperatives, on local public finance for three reasons. First , planned developments are more likely to provide the wide range of services typical of local governments, and so they are the most likely to be perceived as an alternative to 21 public government. Second, planned developments are the fastest-growing type of private government. Gordon (2004) claims that sales in planned developments represented forty per cent of new home sales in the 1990s. Finally, homeowners in planned developments know that the association has legal powers to force them to pay assessments and comply wi th regulations. Planned developments use these powers consistently, and the lengthy history of court challenges by individual owners has generally favoured the association. These legal precedents reinforce the notion of the planned development as a government. However, the effects of condominiums, the other major type of homeowners' associations, wi l l not be ignored. A s a preview of the results, a 10 per cent increase in the rate of planned development membership in a city wi l l , on average, decrease per capita total expenditures by 1.63 per cent. However, the rate of condominium membership in a city has no effect on local expenditures. The analysis also uncovers differences between various expenditure categories. Higher rates of private government membership decrease public government spending in some categories, such as police and garbage disposal, but not in others, such as roads and government administration. The paper is organized as follows. Section 3.2 discusses relevant theoretical research on private government and outlines an. extension of the Helsley and Strange model to serve as a framework to address the empirics. Section 3.3 explains the empirical methodology, and Section 3.4 the data. Section 3.5 presents the analysis. Section 3.6 gives some further directions for research. Section 3.7 concludes. 3.2 Theoretical Framework This chapter presents an extension of Helsley and Strange (1998) and Helsley and Strange (2000), the principal models of private government. In these models a private government and a public government simultaneously provide local public services to residents of a city.. Helsley and Strange treat public and private government services as perfect substitutes in consumption and find that private and public government spending are strategic substitutes. In other words, a public government, knowing that a private government exists to supplement services to some residents, provides less of the public service in equilibrium. The upshot is that city residents who do not belong to a private 22 government receive less of the public service, while members of the private government receive more. Helsley and Strange refer to this result as "strategic downloading." The strategic downloading hypothesis bolsters popular claims that private govern-ment institutions are usurping authority previously held by local government. Roland (1998) and Johnston and Johnston-Dodds (2002) examine the laws that regulate home-owners' associations in California and conclude that their governance structure resembles both business enterprises and municipal governments. Like directors in a corporation, H O A board directors are not personally liable for any damages the association may incur. But like municipal governments, they can enact enforceable regulations and laws, and the association can be compensated for breach of these regulations. McKenzie (1994) ar-gues that such broad powers mean that "CIDs [common interest developments] currently engage in many activities that would be prohibited if they were viewed by the courts as the equivalent of local governments." The transfer of public authority to private gov-ernments, and the similarity that private governments bear to local institutions, may feed back to a deterioration in the public system, as argued by Roland (1998): "These affluent communities provide not only amenities like tennis courts, but also better public services like garbage collection and security. Members of the C I D communities then vote down taxes needed by local governments to provide services to the broader community." Tax relief and secession are two more ways private governments affect local gov-ernment. Anderson (1996) notes that some homeowners' association lobbies have been particularly vocal in demanding reimbursement for public services that they provide for themselves. If public governments are indeed strategically offloading responsibilities to private associations, this argument may have some merit. Ben Lambert, an attorney for the Community Associations Institute in New Jersey (quoted in Kle in (1995)), argues that "if homeowners do not burden local governments wi th needs for services, then they should hot be paying for those services." Indeed strategic downloading may push locals to the point of leaving the city altogether. In a recent example, a coalition of homeown-ers' associations and businesses headed a campaign to remove the San Fernando Valley, an area of 1.3 mill ion people, from the Ci ty of Los Angeles. The most common complaint was that residents of the Valley paid more taxes to the Ci ty than they were receiving back in services. The debate culminated in a secession referendum in 2002. Whi le the referendum was defeated, it drew attention to the growing power of the homeowners' 23 associations and their influence on local politics. Finally, there is a possibility that private governments may increase public expen-diture. There may be complementarities in the provision of private and public services such that members of private governments, through their participation in the voting process, demand more output from the public government. Gordon (2004) notes some evidence, although weak, that voters who live in planned developments are more po-li t ically active and are more likely to vote in state elections. In such a situation, it may be welfare-enhancing for local governments to welcome the formation of private governments. However, the models of Helsley and Strange do not admit strategic com-plementarity. To address this issue, this section presents a theoretical framework in which the empirical results for interaction may be applied. It is a generalization of Helsley and Strange (2000). Assume there is a public government and a private government in a city. Residents are divided into two types: some are members of the private government, and the rest are non-members. The total population of the city can be normalized to 1, so that the population of members can be represented by a fraction / i , 0 < p < 1. The membership decision is exogenous and is abstracted away by assuming that members have a higher preference for the public service. 1 There is only one local public service, but both governments can provide it. A l l city residents can consume the service the public government provides, but only private government members can consume the service the private government provides. Bo th governments produce their services at a per capita unit cost of c. Assume that residents finance local public service provision by fees. This is consistent wi th empirical findings that many local services are produced wi th a constant returns to scale technology. 2 The assumption is that public and private government services enter in a sub-utility function v(gpub, gprw), where gpub represents public government provision, gprw represents private government provision and v(•) is increasing and strictly concave. This is a depar-1 F o r endogenous membership decisions, see Hels ley and Strange (1998). 2 See Rei ter and Weichenr ieder (1997) for a comprehensive survey of studies of the demand for loca l pub l ic goods. T h e y review studies in the spir i t of Bergs t rom and G o o d m a n (1972) and Borcherd ing and Deacon (1972), and they indicate that the " large major i ty of these studies do not f ind signif icant economies of scale, but conclude that the crowding of pub l i c l y provided services is so h igh that these goods are essential ly l ike pr ivate goods." 24 ture from Helsley and Strange (2000), who assume perfect substitutes: v(gpub + gpriv). The more general formulation admits the possibility of complementarity i n consump-tion. For example, v(-) may represent the uti l i ty from "security," so that the public governments provides police spending while the private government provides a gate and patrols. Then if the gate or private security forces increases the marginal effectiveness of police, public and private governments would be complements in consumption. The uti l i ty of a private government member is UM = v(gpub, gpriv) - cgpub - cgpriv, (3.2.1) arid the uti l i ty of a non-member is UN = 6v(gpu\ 0) - cgpub. (3.2.2) Here, 9 lies strictly between 0 and 1, and so represents the lesser weight that non-members place on consuming the public good. Governments, as the active agents in this model, maximize the aggregate welfare of their constituents. Thus, there are no income effects.3 Each government holds a Nash conjecture about the provision level of the other government. The provision decisions are taken simultaneously. The private government's problem is straightforward: meiXLi\v{gpub,gpriv)-cgpub-cgpriv]. (3.2.3) gpriv ^ Thus the private government's optimal provision level is determined by the first-order condition: V2(gpub,gpriv) = c (3.2.4) The subscripts denote the argument of a partial derivative. Equation (3.2.4) implies that the private government provides a service level up to where the marginal benefit of provision equals the marginal cost. The public government maximizes the aggregate uti l i ty of the entire population, con-sisting of both private government members and non-members. The public government 3 O t h e r government object ives are possible, for instance, max im iz i ng aggregate proper ty values, the ut i l i ty of the median voter or the size of the government budgets. A s this model focuses on the strategic act ions of the two governments, there is no need to in t roduce these elements. 25 solves the following problem: max (JL \v(gpub, gpmv) - cgpub - cgpriv] + (1 - p) \9v(gpuh, 0) - c / u b l . (3.2.5) The first-order condition that determines gpub, taking g p r w as given is misT*, 9prw) + (1 - H)0vi(9pub, 0) - c. (3.2.6) Equations (3.2.4) and (3.2.6) jointly determine the equilibrium levels of gpub and gPrw. These two equations implici t ly define the two governments' reaction functions: Equation (3.2.4) determines gPr™(gPub), while Equation (3.2.6) determines gpub{gpriv). Totally differentiating the private government's reaction function gives: dgpriv v12{gpub, gpriv) dgPub v22(gpub,g-priv) (3.2,7) Similarly, totally differentiating the public government's reaction function gives: dgPub -pv12(gpub,gpriv) dgpriv fxvn(gPub,gpriv) + (1 - Li)9vn(gpub, 0) ' (3.2.8) -dgPriv- • dgpub -sign _ dgPub _ = sign dgP^ _ If v(gpub, gpriv) is strictly concave, then v22(gpub, gpriv), vn(gpub, gpriv) and vn(gpub, 0) are negative, so the following result holds: =szgn[vl2(gpub,gpriv)]. (3.2.9) The cross-partials determine the sign of the reaction function. If the cross-partial vi2 is positive, public and private government spending are strategic complements, while if the cross-partial is negative, they are strategic substitutes. 4 The test for strategic interaction comes down to a test of whether the proliferation of private governments in a city has any influence on the local government's choices. A priori, the strategic effect can either be negative or positive, indicating strategic substitutes or complements, respectively. 4 H e l s l e y and Strange (2000) assume that vi2 = - 1 , and so publ ic and pr ivate spending are always strategic subst i tutes. No te that if bo th react ion funct ions are downward s loping, there exists a unique N a s h equ i l ib r ium. T h i s is not necessari ly t rue if the react ion funct ions are upward s loping. A unique N a s h equ i l ib r ium in this case can be ensured by assuming the appropr ia te cond i t ion on the relat ive magni tudes of the slopes of the react ion funct ions. 26 What k ind of local services could be expected to be strategic substitutes in private government? A n example is local police; Many homeowners' associations feature private security. In condominiums, this usually takes the form of security guards and limited access to the building. In gated communities, barriers restrict access to the roads within the community. The strategic substitutability effect arises if, for example, the local police does not send as many patrols to areas served by homeowners' associations. The marginal uti l i ty of public spending falls as policing resources are redirected away from residents in homeowners' associations. Wha t about strategic complements? They arise from the complementarity in con-sumption of private and public services. This might come from positive spillovers from one government to the other. For instance, police effectiveness may be enhanced if pri-vate associations detain suspicious persons or share information. Alternatively, strategic complementarity could come about from a reallocation of residents' demands. For ex-ample, if a large proportion of voters are very concerned about crime, they may demand more local police spending, even while they live in gated communities. Where strategic complementarity exists, public government may have an incentive to increase spending in response to private government. In this case, local authorities may welcome the for-mation of private governments as a way to increase the overall level of public service consumption. 3.3 Empirical Specification The simple theoretical model outlined above suggests that, depending on the complemen-tarity or substitutability of privately and publicly provided services, public governments may increase or decrease spending in response to private governments. This section lays out the empirical model that is subsequently used to estimate the sign and the magnitude of this effect. The basis of empirical work is the estimation of demand functions for municipal public services. Bergstrom and Goodman (1972) and Borcherding and Deacon (1972) are seminal articles that posit demand for local services depends on traditional economic variables, such as price and income, as well as demographic characteristics of residents. Following these papers, various authors, such as Greene and Nelson (1994), focus on 27 non-demographic, polit ical motivation in explaining public sector spending. Subsequent research recognizes that local decisions do not consist simply of a city responding to the demographic changes in its population; rather the relationship be-tween a local government and the population it serves is tightly intertwined. It is also important to account for the fact that cities respond to the presence of neighbouring cities and higher levels of government. Thus further research integrates strategic action in estimating local government choices. Brueckner (1998) is one of the first to test a strategic model of local government spending. He asks whether cities strategically implement certain growth controls in response to neighbouring cities' actions. The key result is that policy equilibrium can be described by a set of reaction functions, and the key coefficient is the slope of the reaction function, which represents the effect of other cities' growth controls on one's own policy. The sign of the slope may be positive or negative, which indicates that if there is strategic interaction, a rival city's growth controls may be strategic substitutes (negative slope) or complements (positive slope) to a given city's controls. Subsequent studies, surveyed by Brueckner (2003), consider other local choices in similar strategic models. Most agree that local governments can and do act strategically. However, previous studies are virtually all cross-sectional analyses, so there is no way to examine questions such as the pattern of growth controls enactment, the effect of long-standing versus recently-enacted measures and other dynamic issues. Thus, the literature on strategic interaction sti l l lacks a panel analysis on how governments respond to each other. The literature also lacks empirical studies of homeowners' associations. Few stud-ies have considered planned developments as rivals in a local government's provision problem, even wi th their similarity. A key issue is the lack of data. Because homeown-ers' associations are private associations, there is no oversight or regulatory agency to compile statistics. It is even unclear how many homeowners' associations exist in the United States. However, as these associations become much more commonplace and polit ically visible, various authors are beginning to contribute to empirical understand-ing. These, notably McKenzie (1998) and Gordon (2004), deal mainly wi th the effect of homeowners' associations on voter behaviour. However, they do not focus on the impact of homeowners' associations on the spending behaviour of local governments. 28 This paper addresses these gaps in the literature. The analysis focuses on local public expenditure, determined by the following equation: gftub = Pofr + Ma + di + dt + drt + e«, (3.3.1) where i indexes cities 5 , r indexes regions and t indexes years. pf" f t represents public expenditure per capita. Expenditures, rather than revenues, constitute an ideal depen-dent variable because local politicians make the spending decisions. Also, categorized expenditures make it possible to analyse the effect of private government on individual types of public services. The explanatory variable of interest is gfpv, the measure of private government activ-ity in city i at time t. The analysis explores two different measures of private government in this paper, which Section 3.4 wi l l describe. The null hypothesis is the absence of inter-action of private government on local spending: (3 = 0. If this hypothesis is rejected, then the relevant question is whether (3 is positive, consistent wi th strategic complements, or negative, consistent wi th strategic substitutes. The other explanatory variables, Xu, on the right hand side of (3.3.1) are other variables posited to affect the level of local public spending. Section 3.4 also describes these variables in detail. The rest of the estimating equation consists of dummy variables for various unmea-sured effects, di is the city fixed effect that absorbs permanent heterogeneity at the city level. These might come from geographic amenities, infrastructure or the efficiency of local bureaucracy. They also include poli t ical and historical idiosyncrasies. For example, a city that is the county seat may attract state and county spending. This may arise from the flypaper effect, where tax and grant revenues are capitalized into expenditures on administration. In turn, residents of the city demand complementarily higher levels of local services. The type of city government and the relationship between the executive and legislative branches can also matter. The city fixed effect also picks up the variation coming from cities wi th permanently high or low levels of public expenditure. Next, dt is a time dummy for each year from 1971 to 1999, which picks up the time-varying trend in the error that is the same for all cities in the state. These may 5 A c i ty is an incorporated ent i ty w i t h an autonomous government and self-rule. T h i s is impor tan t be-cause one-quarter of Ca l i fo rn ia 's popu la t ion does not l ive in incorporated cit ies. P l a n n e d developments in un incorporated areas, by def in i t ion, are exc luded f rom the analysis. 29 come from business and investment cycles, election cycles and the effects of statewide budgetary l imitat ion initiatives. Aside from the famous Proposition 13, which strictly l imited local property taxes, California has had a long history of initiatives to curb perceived government overspending. Finally, as many cities are members of an interdependent metropolitan conglomer-ation, it is appropriate to consider region-specific effects as well. drt is a region-year interaction dummy that absorbs region-specific, time-varying heterogeneity. Shocks of this type may arise from large scale natural disasters, regional crime and regional eco-nomic shocks. For instance, one of the most notable regional economic shocks is the technology boom that hit municipalities in the Silicon Valley in the 1990s. The region is defined by the metropolitan statistical area ( M S A ) in which city i is located. It is a natural choice for delineation of regions, as the Census Bureau itself defines M S A s on the basis of economic interdependence of the municipalities in them. There are sixteen M S A s in California, although not all of them are represented by cities in the sample. Thus, wi th all of the dummy variables in the specification, identification is based on changes over time in private government proliferation within a city. The main source of heterogeneity not controlled for is eu, a time-varying city-specific shock. A key assump-tion in a fixed-effects estimation, therefore, is that gP[w, the level of private government proliferation, must be uncorrelated wi th the unobserved error eit. Therefore, a threat to identification would be a city-specific, time-varying shock, to local expenditure that is correlated wi th private government formation. A possible source of such a shock is a common, exogenous factor that drives both public and private government activity together. This leads to an omitted variables problem. A high level of crime in a city, for instance, may spur increased police spending by the public government, while at the same time raise the demand for gated or patrolled private communities. If this common cause problem is not addressed, then the estimation of (3.3.1) may indicate a spurious positive effect of private government on police spending. To give another example, the election of a "pro-development" council might try to attract residential development by lowering development fees, encouraging private government formation, and lowering taxes, necessitating cuts in expenditure. In this case, the estimation of (3.3.1) wi l l result in a spurious negative coefficient on gPrw. Then public governments would not be strate-gically substituting private governments for public provision; rather their provision is 30 correlated wi th private government formation through an underlying polit ical preference for development. Another threat to identification is the simultaneity of public and private government provision levels. The basis for this problem is that the theoretical papers of Helsley and Strange (1998, 2000) model public and private governments as players in a strate-gic game. Instead of an exogenous common factor that drives both public and private governments, the two provision levels are jointly determined in a Nash equilibrium. In-strumental variables estimation is an appropriate way to deal with the omitted variables and the simultaneity issue. Section 3.4.4 discusses this approach in detail. 3.4 Data 3.4.1 Public Government Variables The dependent variable is the level of local public government expenditure, which comes from the Annua l Survey of Governments, administered by the U.S . Bureau of the Census. The cities surveyed generally meet a minimum standard for population, but some smaller cities are included as well. The sample is the 110 cities that report thirty years of expenditure data from 1970 to 1999. This represents one quarter of the state's cities, yet they account for a substantial percentage of the urban population. 6 In 1999, the cities in the sample have a combined population of 16,941,390, which is .63 per cent of the state population in incorporated cities. Because of the sampling methodology, the cities in the sample are larger than the average municipality in California. The mean population for cities in the sample is 129,745 over the period from 1970 to 1999, while it is 44,554 for all 471 cities in the state. Does this pose a serious sample selection problem? There seems to be a tradeoff. If smaller cities are more receptive to the demands of residents and developers and hence more likely to respond strategically to private governments, then this may underestimate the true impact of private governments on local spending. However, if smaller cities in fact respond to a lesser degree because economies of scale make it difficult for them to 6A list of the cities in the sample and the regional metropolitan areas may be obtained from the author. 31 adjust local spending significantly, the strategic downloading effect of private government may be overestimated. A n y significant coefficient on the private government variable should then be taken with caution. However, this latter situation is unlikely to arise. There is little empirical evidence that local public service provision enjoys substantial economies of scale. In addition, the empirical model controls for population size by using per capita values for all expenditure variables. Nine variables measure annual local public expenditure. The principal measure is to-tal direct government expenditure per capita, including current operating expenditures, construction and capital outlays. This variable does not include payments to other gov-ernments. 7 The other eight measures are expenditures on specific services: police, fire, highways and roads, solid waste disposal, parks and recreation, housing and community development, libraries and government administration. 8 Each of these categories is used as the dependent variable in an estimating equation,- but they are auxiliary tests of the main hypothesis. The motivating question for using categorized expenditures is this: if cities do engage in strategic downloading of service responsibilities to private institu-tions, does the effect act more strongly for some services compared to others? Intuition suggests that the substitutability of the privately-provided service for the public one plays a crucial role. The top part of Table 3.4 shows the relative importance of the eight categories of spending and provides sample statistics. The dependent variable is log per capita real public expenditure. For the auxiliary regressions on specific categories, the dependent variable is log(per capita expenditure + 1) as some cities report zero expenditure for certain categories on occasion. 9 Throughout 7 T h i s is a potent ia l source of mismeasurement. T h e on ly substant ia l payment to other governments made by cit ies is to special d ist r ic ts. These distr ic ts provide a substant ia l level of loca l services; how-ever, they ma in ly provide water, sewerage and other pub l ic ut i l i t ies. A s such, they are not readi ly subst i tu table w i t h goods provided by pr ivate governments. To ta l expendi tures also do not inc lude spe-cific amounts for the water supply, electric power and pub l ic mass t rans i t , p resumably as the survey counts these as payments to separate pub l ic u t i l i t y companies. T h e inc lus ion of da ta on special dis-t r ic ts and pub l ic ut i l i t ies wou ld undoubted ly add to a more complete descr ipt ion of publ ic government behaviour , but the dif f iculty in reconci l ing these da ta w i t h mun ic ipa l da ta places it outside the scope of th is analysis. 8 Sev e r a l of these categories require add i t iona l exp lanat ion . "H ighways and roads" encompasses the const ruct ion and maintenance of al l pub l ic streets, inc lud ing sidewalks and bridges. T h i s includes snow removal and street l ight ing. "So l id waste d isposa l " not on ly includes garbage col lect ion but also sani ta t ion services such as street c leaning. "Hous ing and commun i ty development" is a broad category that most ly comprises publ ic housing and urban renewal programs. F ina l l y , "government admin is t ra t ion" is the sum of financial admin is t ra t ion , jud ic iary , pub l ic bui ld ings and the central staff. 9 Supplementa l regressions showed no qual i ta t ive difference in coefficient est imates if the observations 32 Figure 3.1: Mean Level of Local Expenditures in the 110-City Panel 1,200 7 200 i i 1 1 i 1— Year Source: U.S . Census Bureau, Annua l Survey of Governments, 1970-1999 the study, all dollar values are expressed in 1997 dollars using the G D P deflator. Examining the trend of local expenditure from 1970 to 1999 reveals something pos-sibly unexpected. Despite the popular perception of fiscal restraint, real expenditures per capita have actually been on the rise. Figure 3.1 demonstrates this growth. Mean real local expenditure rose from $512 per capita in 1970 to $1,099 per capita in 1999 for cities in the sample. Expenditures did decline from 1979 to 1982, perhaps reflecting the immediate impact of Proposition 13 on local revenues. However, there is ample evidence that cities recovered from the bite of Proposition 13 and subsequent budget-tightening initiatives by reallocating revenue sources. Smith (1991) argues that cities moved to an increased reliance on fees and charges and notes that Proposition 13 had different effects on different expenditure categories. Figure 3.2 shows that a shift in spending priorities d id occur. Publ ic safety services such as fire and police experienced an increase in ex-penditures. There is also a large increase in housing and community development. O n the other hand, parks and highways show little growth in real expenditures, with zero spending are removed. 33 Figure 3.2: M e a n Categorized Expenditure in the 110-City Panel 250 ~a 200 -S 150 100 Y e a r • P o l i c e - H i g h w a y s F i re - P a r k s - S o l i d w a s t e - L i b r a r i e s - H o u s i n g a n d c o m m u n i t y d e v e l o p m e n t - G o v e r n m e n t a d m i n i s t r a t i o n Source: U.S . Census Bureau, Annua l Survey of Governments, 1970-1999 3.4.2 Private Government Variables The key explanatory variable is the level of private government activity. The biggest obstacle to rigorous empirical study is the lack of comprehensive data on these private institutions. A novel source of information about homeowners' associations allows the development of two measures of private government. A database of homeowners' associations in California was obtained from the account-ing firm of Levy and Company in Oakland. It is a unique record of planned developments, condominiums, cooperatives and other homeowners' associations, and it is one of the most complete sources of information about residential private governments available. The database lists every one of the 37,655 incorporated homeowners' associations in the state as of M a y 2003 along with some information of its size, location and function. Figure 3.3 shows the number of associations by year of incorporation and type. The figure clearly demonstrates that homeowners' associations were virtually unknown up to 1970. There is a gradual increase in the number during the 1970s, but in 1978, the number of new incorporations soared and remained at a very high level for about twelve years. There has been a gradual decline in recent years and a return to new incorporation 34 levels resembling those of the mid-1970s. The data also show a distinction between the various types of common interest development and their relative popularity. Planned developments, which generally consist of single-family houses; were the predominant form of common interest development in California unti l the early 1970s. Condominiums began to rise in popularity and took over as the most common form of homeowners' association. For the major part of the analysis, only planned developments (PDs) enter the panel. Assume that the incorporation year of each development is the first year that the pri-vate government is in operation and providing services. Addit ionally, assume that the death rate of homeowners' associations is negligible. 1 0 The panel is then constructed by counting how many associations were in existence in each year. There were 3,199 planned developments in the sample as of 1999. The next step is to impute the total number of P D housing units in each city and year. For confidentiality and data availability reasons, the data set does not report the number of housing units in each association. Instead, it provides a size category variable, which classifies each community association into one of twelve size categories. Each development is assumed to have a number of units equal to the median of the size category it belongs to. Table 3.1 summarizes these imputed values for the two end years of the panel, 1970 and 1999, as well as over the entire thirty-year per iod . 1 1 The first measure of private government is the number of housing units per capita in planned developments in the city. This takes into account the large variation in associ-1 0 A s the typ ica l p lanned development has about 75 single-fami ly housing uni ts , they are unl ike ly to be destroyed and redeveloped. In add i t ion , Ca l i f o rn ia statutes require that each development ma in ta in a resident ial communi ty associat ion to provide services and enforce covenants. Thus , as long as the st ructure is st i l l inhab i ted , the pr ivate government is s t i l l in existence. 1 1 C o m p a r i n g these imputed numbers to those of other surveys helps to check the p laus ib i l i ty of the impu ta t ion . F i r s t , the Ca l i f o rn ia Depar tment of F inance est imates that there are 6,229,316 housing uni ts in the 110 cit ies of the panel in 1999. Thus , as Tab le 3.1 indicates, the impu ta t ion suggests that 6.06 per cent of housing uni ts in the state belong to a p lanned development in 1999. In compar ison, the 1999 A m e r i c a n Hous ing Survey est imates that for the Western states, 6.5 per cent of occupied housing uni ts belong in p lanned developments. T h i s figure represents the number of housing uni ts that report pay ing a regular homeowners' associat ion fee, but wh ich are not condomin iums or cooperat ives. It appears that the imputed housing numbers match the actual s i tuat ion qui te closely. In add i t ion , the L e v y and C o m p a n y da ta set also includes some associat ions for wh ich it is unknown whether they are p lanned developments, condomin iums or some other type. These are assumed to occur at random and are dropped f rom the analysis. Thus , the imputed numbers p robab ly underest imate the t rue number of p lanned development uni ts in the state. However, even w i t h these p laus ib i l i ty checks, measurement error may be a signif icant source of bias in the est imated interact ion effect. 35 Figure 3.3: G r o w t h i n H O A s i n C a l i f o r n i a - E n t i r e S t a t e , A l l T y p e s 2,000 -i : : 1,800 1,600 1,400 1,200 • <S 1,000 E 800 600 400 200 0 4- , B , a , B ° , n , ° Year of Incorporation • Condominiums • Mobile home parks 11 Co-ops H Planned developments Source: Levy and Company, 2003 36 ation size. It is reasonable to believe that a development wi th ten units should not have the same marginal effect on public spending as a development wi th five hundred units, especially if the two are located in relatively small municipalities. The second measure of private government is the membership rate, the number of planned development housing units as a proportion of all the housing units in the c i ty . 1 2 This measure has intuitive appeal. Publ ic governments may be more inclined to act strategically if they realize that a large proportion of their population is paying for similar private services. The 110 cities in the panel exhibit wide variation in private governments. Table 3.1 presents some summary statistics of the private government variables. The mean rate of planned development membership from 1970 through 1999 was 5.3 per cent. In 1999, the mean was about 8.1 per cent, reflecting the upward trend in private government membership. 1 3 Figure 3.4 shows the membership rate from 1970 to 1999, calculated as the sum of al l P D units divided by the sum of al l housing units in the cities of the sample. In addition, planned developments are not distributed equally across municipalities. In 1999, the highest rate of membership, three-fifths of al l housing units, was in Auburn in Placer County. Auburn is a suburb of Sacramento. Indeed, the highest rates of membership are found in smaller suburban cities. To look at a cross-section, Figure 3.5 shows a map of the counties of California shaded wi th the mean rate of planned develop-ment membership in 1999; planned developments are most heavily located in suburban counties surrounding San Francisco, Los Angeles and San Diego. 1 2 T h e denominator of the p lanned development rate combines different sources to ob ta in a reasonable t ime series of to ta l housing uni ts in each city. For 1970, 1980 and 1990, the U .S . Census Bu reau provides housing uni t counts. For 1991 through 1999, the Ca l i f o rn ia Depar tment of F inance 's Demograph ic Research U n i t provides housing uni t est imates. For the other in tervening years, the number of housing uni ts is in terpolated using the Census values as anchors and assuming a constant growth rate. 1 3 O u t of concern for the stat ionar i ty of the t ime series of pub l ic and pr ivate governments, the panel uni t root test given by L e v i n , L i n , and C h u (2002) w i t h an ind iv idua l specif ic mean and t ime t rend was performed. T h e panel uni t test soundly rejects the nu l l hypothesis of non-stat ionar i ty for log real per cap i ta pub l ic expendi tures: the t s tat is t ic is -17.05. T h e pr ivate government measures also reject the nu l l : for per cap i ta p lanned development uni ts, the t s tat is t ic is -1.82, wh ich is signif icant at the 5 per cent level. For p lanned development membership rate, the stat is t ic is -2.95, wh ich is signif icant at the 1 per cent level. These test results indicate that the regression techniques are va l id . 37 Table 3.1: Summary Statistics of Planned Development Measures Sample: 110 cities over 30 years Actua l number of planned developments in 1970, a l l cities: 142 Total imputed P D units in 1970, al l cities: 50,585 Ac tua l number of planned developments in 1999, all cities: 3,199 Total imputed P D units in 1999, all cities: 377,269 Year(s) Mean . Std Dev. M i n . M a x . Total PD units in city 1970 only 382 817 0 4,557 1999 only 3,430 5,332 0 40,556 1970-1999 1,900 3,465 0 40,556 PD units per capita in city 1970 only .007 .057 0 .152 1999 only .030 .091 0 .268 1970-1999 .021 .031 0 .296 PD membership rate in city 1970 only 2.1 % 2.0 % 0 % 38.4 % 1999 only 8.1 % 3 .7% 0 % 60.2 % 1970-1999 5.3 % 7 .4% 0 % 60.2 % Source: Levy and Company 2003, and author's own calculations 38 Figure 3.4: Overall Membership Rates in Planned Developments and Condo-miniums in the 110-City Panel 9% i 3.4.3 Other Covariates The model also includes variables that are commonly encountered in empirical studies of local government spending. The covariates include population density, median household income, proportion wi th a college degree, proportions of blacks, Hispanics and Asians, proportion of foreign-born, proportion reporting welfare income, proportion of persons 17 and under, proportion of persons 65 and over, proportion of dwellings that are owner-occupied and 5-year population growth rate. A l l of these are expected to exert some pressure on the demand for local public expenditures. The data come from census sources. Finally, two sets of dummy variables absorb the yearly effect and the yearly . regional effect. Before moving to the regressions, it is useful to look at some of the correlations between P D membership rate and some of the covariates. Table 3.2 shows some selected correlations between covariates and various types of private government. For the most part, correlations between a city's P D membership rate and the covariates are weak. 39 Figure 3.5: Rate of Planned Development Membership by County Planned Development Membership Rate, 1999 (based on 110-city panel) kHumboldt , Y o l o V ^ ^ ^ El Dorado (Sacra-nento I Alpine, 0% to 4.9% 5% to 9.9% 10% to 14.9% 15% to 19.9% 20% and higher No data (Santa Barbara Angeles Riverside San Diego Imperial Source: Levy and Company, 2003, and author's own calculations 40 Table 3.2: Correlations Between HOA Membership Rate and Selected Covari-ates, Selected Types Sample: 110 cities over 30 years Correlation wi th rate of membership Covariate A l l H O A s Planned Developments Condos Per cent Black -.17 -.15 -.14 Per cent Asian .02 -.05 .11 Per cent Hispanic -.11 -.15 -.04 Per cent wi th 4 years college .44 .27 .53 Per cent 65 and over .28 •17 .33 Per cent 17 and under -.30 -.12 -.45 Median household income .32 .19 .39 Population density -.30 ' -.35 -.10 Per cent owner-occupied dwelling .16 .22 .03 Source: Levy and Company 2003, Census data and author's own calculations The proportions of Blacks and Hispanics are weakly negatively related to membership rate, which may confirm the perception of P D s as a subtle means of exclusion. The proportion of elderly is somewhat positively correlated wi th membership rate, probably reflecting the popularity of retirement communities in some cities. Interestingly, cities wi th high rates of homeowners' associations are not necessarily wealthy cities. The correlation between membership rate and median family income over the sample is 0.32. Whi le the most popular image of homeowners' associations is that of an exclusive, gated community, the data reveal that they are becoming more and more common among middle-income families. However, there is a clear positive correlation between average education levels and membership in a homeowners' association. Finally, population density is negatively correlated wi th membership rate, which is another reflection of the popularity of private governments in newer suburban developments. 3.4.4 Instrumenting for Endogenous Variables The estimating equation of public spending may have an endogeneity bias. Private government activity may not be exogenous to public expenditure. Private government belongs in the public government's reaction function, but theory suggests that public 41 provision can affect private government activity. One demand-driven possibility is that homeowners, frustrated by the deterioration of local services, demand neighbourhoods that provide higher-quality services and pay for the privilege by joining a homeowners' association. There may also be a supply-driven reason. Many authors (Roland (1998), Johnston and Johnston-Dodds (2002)) argue that local authorities, faced with voter-imposed and state-imposed spending cuts, turn to land developers wi th a deal. The government unloads the cost of providing infrastructure and local services on the devel-oper, who may receive additional tax incentives. The developer then passes the costs to homeowners by setting up a private government. The empirical estimation must address these sources of endogeneity. In order to correct for the simultaneity and endogeneity problems, instrumental vari-ables are needed that affect private government formation but not public spending. Two types of variables serve as instruments, lagged private government and city land area. IV: Lagged Values of Private Government The justifying assumption is that past values of private government proliferation in the city do not affect current levels of public spending. Using them as an instrument assumes there is an underlying factor about the city that encourages the growth of private governments. This can be institutional, taste-driven or developer-driven. Lagged private government values of fifteen and twenty years serve as instruments. It is reasonable to think that private development decisions taken fifteen to twenty years ago wi l l influence present public spending only through the creation of more private governments. The instruments are simply the lagged values of per capita planned de-velopment units or planned development membership rate, whichever happens to be the key measure of private government. The use of the lagged endogenous variable in instrumental variables estimation has a long history in econometrics. 1 4 For lagged private government to serve as an appropriate instrument, it is necessary to assume that if there is any persistence in local public expenditures, it is not as long as the lag in planned development measure. A s municipal councils would have turned over at least several times in fifteen years, this'assumption is likely defensible. 1 4See, for instance, Villas-Boas and Winer (1999) and Aronsson, Lundberg, and Wikstrom (2000). 42 IV: Land Area Ci ty land area serves as an instrument for private government prevalence because a pattern of annexations generally indicates residential development. New residential de-velopment drives the growth of private governments. Between 1970 and 1999, almost every city in the panel changed in land area. Most boundary changes result from the annexation of undeveloped land and hence add to the land area of a c i ty . 1 5 Land area changes in turn reflect the feasibility of and taste for increased residential development. First , planned developments tend to require large tracts of land, so cities that border on unincorporated land have more "room to grow." A city should also have the demand for increased residential development in order to undertake the annexation process: while cities may annex for many economic and polit ical reasons, suburban developers initiate a large proportion of annexations. 1 6 Fleischmann (1986) corroborates this view with his-torical evidence that developers use annexation as a way to shift development costs to local governments. Therefore, land area changes should, in part at least, reflect episodes of increased residential development. California reality reflects this argument. Developers often spearhead annexation campaigns by petitioning the city they wish to join. This request must then receive approval from a county-level Local Area Formation Commission ( L A F C O ) . The L A F C O assesses the environmental, economic and polit ical impact of increasing city boundaries on surrounding areas before granting approval; thus, local boundary changes are usually justified by a demand for development. 1 7 Evidence also suggests that increased development activity is strongly responsible for the tremendous growth of private governments. Roland (1998), for instance, notes that developers responded to the rising cost of land in the 1960s by combining smaller lots wi th common open areas and marketing them as a planned development. This also coincided wi th changes in the building industry, where large-scale corporate builders started to dominate. These builders had the capital to construct large planned commu-nities. In a county-level analysis, McKenzie (1998) shows that developer driven action 1 5 E p p l e and Romer (1989) notes that annexat ions make up 98% of a l l boundary changes in their survey. 1 6 S e e A u s t i n (1999) for an empi r ica l example. 1 7 A n explorat ion of minutes f rom L A F C O meetings conf i rms that pet i t ions for annexat ion are most often in i t ia ted by landowners and developers. 43 can explain much of the variance in private government popularity. Land area changes are assumed to be uncorrelated wi th shocks to public spending. This is likely to be the case because land area changes tend to be incrementally small. Such small changes in land area should not result in large increases in per capita public service expenditure. • Table 3.3 gives some descriptive statistics for the land area instrument. 1 8 Cities vary widely in size, from barely two square miles for Fortuna to nearly 470 square miles for Los Angeles. There has been tremendous growth in city sizes, an average increase of 40.5 per cent over the period from 1970 to 1999. However, comparison wi th single-year growth rates demonstrates that area growth is incremental and consistent with developer-driven demand. The mean single-year change is only 0.3 square miles. The mean is pulled down by years in which cities do not change in area. For only those years in which a city experienced a change, the mean change is 1.7 square miles, and the median change is 0.35 square miles. This pattern of heavy but incremental annexation activity suggests that residential development, rather than other polit ical motivation, is the driving force behind land area changes. A s development is correlated with the formation and proliferation of private governments, its use as an instrumental variable is justified. To summarize, Table 3.4 gives a complete list of al l the variables used in the analysis. 3.5 Analysis 3.5.1 Without Private Governments First , consider a model of local public spending using all of the explanatory variables except for measures of private government. This acts as a gauge of the quality of 1 8 Un fo r t una te l y , the land area var iable is subject to measurement error, as it is dif f icult to obta in year ly land area numbers. A combinat ion of sources is used. T h e U . S . Census gives accurate area measurements for 1970, 1980 and 1990. T h e County and City Data Book provides areas for 1975. T h e U . S . Census Bureau 's B o u n d a r y and Annexa t i on Survey provides land area da ta for 1976 through 1979 (U .S . Census Bu reau 1980). T h e California Planner's Book of Lists provides land area for the fo l lowing years: 1981, 1984, 1985 and 1995-1999 inclusive. A s these are self-reported values, however, they may not be rel iable as the Census measurements. If the area for a certa in year is miss ing, it is assumed that area is tha t of the closest past year w i t h data . 44 Table 3.3: City Land Area Descriptive Statistics Sample: 110 cities over 30 years Mean Std. dev. M i n . Max . Median Land area (sq. mi.) 27.5 55.1 1.9 469.3 15.2 Single-year change* in area (sq. mi.) .3 2.2 -17.6 72.3 0 Single-year percent change* in area 1.1% 7.5% -45.6% 128.5% 0% Change in area, 1970-1999 (sq. mi.) 7.4 14.5 -2.1 89.1 2.4 Percentage change in area, 1970-1999 40.5% 67.5% -40.0% 344.0% 16.7% *Single-year changes from 1970-1971 to 1998-1999. Because of data inconsistencies for years in the 1960s, the single-year change from 1969 to 1970 is omitted. Source: U .S . Bureau of the Census and the California Planner's Book of Lists the demographic and public expenditure data. Columns (1) through (3) of Table 3.5 report the estimated regression results. A Hausman test suggests that the appropriate error structure is a fixed-effects rather than a random-effects estimator. The R2 on the regressions suggests a relatively good fit by the explanatory variables, given the cross-sectional data. They also indicate that the year dummies and the metro-year dummies have considerable explanatory power. A s both the fixed-effects and the. I V estimations give similar qualitative results for the demographic covariates, they are discussed here. Population density has a strong negative coefficient, which suggests that there is a spreading of the fixed costs of services among a dense population. Median household income affects per capita spending in a positive manner, consistent wi th local expenditures being a normal good. The income elasticity, evaluated at the mean, is .26 for specifications (1) and (2), which include year dummies, and .15 for specification (3), which does not. This is consistent with other papers, which find an income elasticity less than one. The proportions of Blacks and Hispanics have a positive effect on per capita spend-ing, suggesting a redistributive element in municipal budgets; this supports the results of Bergstrom and Goodman (1972). The proportion of Asians does not have a signif-icant impact on local spending; but as Asians have.similar median family incomes to white non-Hispanics ($61,383 versus $65,342 respectively in 2000), this is not surpris-ing. Interestingly, the proportion of young people in a city is negatively correlated wi th 45 T a b l e 3 .4 : Summary of Variables Sample: 110 cities over 30 years V a r i a b l e S o u r c e M e a n S . D . M i n M a x Dependent Variables - per capita T o t a l g e n e r a l e x p e n d i t u r e (a) 808 5 7 8 . 7 5,945 P o l i c e e x p e n d i t u r e (a) 140 100 0 1,053 H i g h w a y s a n d r o a d s e x p e n d i t u r e (a) 95 83 0 1,367 F i r e e x p e n d i t u r e (a) 76 66 0 543 P a r k s a n d r e c r e a t i o n e x p e n d i t u r e (a) 69 70 0 1,844 W a s t e d i s p o s a l e x p e n d i t u r e (a) 30 40 0 342 L i b r a r i e s e x p e n d i t u r e (a) 17 23 0 379 H o u s i n g a n d c o m m u n i t y d e v e l o p m e n t e x p e n d i t u r e (a) 70 126 0 2 ,407 G o v e r n m e n t a d m i n i s t r a t i o n e x p e n d i t u r e (a) 78 71 0 1,066 Explanatory Variables P r o p o r t i o n B l a c k (b) .07 .11 .001 .75 P r o p o r t i o n H i s p a n i c (b) .20 .17 0 .91 P r o p o r t i o n A s i a n (b) .08 .08 .001 .75 P r o p o r t i o n o f a d u l t s w i t h 4 yea rs o f co l lege (b) .09 .06 .006 .32 P r o p o r t i o n 17 a n d u n d e r (b) .29 .07 .13 .48 P r o p o r t i o n 65 a n d ove r (b) .10 .04 .02 .26 P r o p o r t i o n r e p o r t i n g we l f a re i n c o m e •(b) .09 .04 .02 .26 P r o p o r t i o n foreign-born (b) .21 .10 .02 .59 P r o p o r t i o n of h o u s i n g u n i t s o w n e r - o c c u p i e d (b) .55 .12 .19 .92 M e d i a n h o u s e h o l d i n c o m e , i n t h o u s a n d s (b) 36.9 9.69 4.5 71.8 P o p u l a t i o n d e n s i t y ( t h o u s a n d s o f p e r s o n s / s q . m i . ) (b ,d) 5.3 3.1 .4 22.7 5-year p o p u l a t i o n g r o w t h r a t e (b) .12 .19 '• - .13 2.93 M S A - y e a r i n t e r a c t i o n d u m m i e s Y e a r d u m m i e s P e r c a p i t a n u m b e r o f P D u n i t s (c) .020 .031 0 .30 R a t e o f P D m e m b e r s h i p (c) .053 .074 0 .60 Instruments P e r c a p i t a P D u n i t s l a g g e d 15 yea rs (c) .009 .021 0 .26 P e r c a p i t a P D u n i t s l a g g e d 20 yea rs (c) .006 .017 0 .25 P D m e m b e r s h i p r a te l a g g e d 15 yea rs (c) .023 .054 0 .56 P D m e m b e r s h i p r a te l a g g e d 20 yea rs • ( c ) .016 .046 0 .52 C i t y l a n d a r e a i n s q u a r e m i l e s (d) 27.6 55.1 1.9 469 .3 All dollar values are GDP deflated to 1997 values. Sources: a = City public expenditures data from the Annual Survey of Governments, U.S. Bureau of the Census. b = Population and demographic data from the U.S. Census of Population and Housing (1970, 1980, 1990). Yearly population data also from the California Department of Finance, Demographic Research Unit (2003). c = Homeowners' associations data from Levy and Company (2003). d = Land area data,from the U.S. Census of Population and Housing (1970, 1980, 1990), the County and City Data Book (1975), the U.S. Census Boundary and Annexation Survey (1976-1979) and The California Planner's Book of Lists (1981, 1984, 1985, 1995-1999). 4 6 expenditures. As school expenditures are not included in a city's budget, the negative coefficient is probably a reflection of the lower tax base. On the other hand, the propor-tion of senior citizens in the city does not have a significant effect on local expenditures. Perhaps this is due to the nature of local services: they may be no more likely to be demanded by older persons than by younger ones. The coefficient on owner-occupied housing is significantly positive; this contradicts some previous findings that owners, being the most sensitive to property tax, are a persistent force in keeping expenditures low. The remaining explanatory variables raise some questions. Five-year population growth varies in sign depending on the inclusion or exclusion of year dummies. How well-educated a municipality is has either no significant or a positive effect on spending. But as the positive sign comes in only on the specification without year dummies, this might simply be picking up the fact that education levels have risen over years of the sample. Also, the insignificant coefficient on education might reflect collinearity with median income. Finally, the proportion of people who are foreign-born and. the proportion who receive welfare income do not have significant effects on local expenditures, but these are likely noisy measures of the effect of migrants and poverty. 4 7 Table 3.5: Fixed-Effect Estimation Results. Private Government Measured by Planned Developments Sample: 110 cities over 30 years Dependent variable: L o g per cap i ta real publ ic expendi ture (1) (2) (3) (4) (5) (6) (7) (8) (9) Pe r cap i ta P D uni ts -2.149** (.435) -1.227** (.403) 1.933** (.42) P D membersh ip rate - .975** (.171) _ 714** (•161) .625** (.165) % B l a c k 1.058** (.238) 1.272** (.24) 1.315** (.269) 1.08** (.237) 1.279** (.24) 1 3** (.268) 1.105** (.237) 1.291** (.24) 1.294** (.268) % A s i a n .02 (.128) .056 (.129) .037 (.141) .026 (.127) .056 (.129) .026 (.14) .052 (.127) .075 (.128) .012 (.141) % Hispan ic .532** (.135) .739** (.131) .98** (.077) .511** (.135) .712** (.131) .936** (.078) .512** (.134) .701** (.13) .938** (.078) % w i t h 4 years college -.026 (.279) .523* (.266) 1.693** (.189) -.263 (.282) .398 (.268) 1.616** (.189) -.381 (.284) .283 (.27) 1.633** (.189) % 17 and younger -1.125** (.282) -1.304** (.279) -1.446** (.237) -1.242** (.282) -1.345** (.278) -1.39** (.237) -1.341** (.283) -1.418** (.279) -1.378** (.238) % 65 and older -.472 (.403) .222 (.397) .645 (.411) -.479 (.402) .257 (.397) .513 (.411) -.601 (.402) .169 (.396) .597 (.41) % w i t h welfare income -.068 (.565) -1.134* (.52) -.936 (.545) -.149 (.563) -1.253* (.521) -.839 (.544) -.217 (.562) -1.331* (.521) -.836 (.545) % foreign-born .045 (.165) .203 (.162) .067 (.113) -.0002 (.164) .169 (.162) .129 (.113) .009 (.164) .154 (.161) .103 (.113) % owner-occupied homes 1.166** (.158) 1.307** (.155) 1.309** (.169) 1.274** (.159) 1.36** (.155) 1.247** (.169) 1.282** (.159) 1.374** (.155) 1.277** (.169) M e d i a n household income .007** (.001) .007** (.001) .004** (.001) .008** (.001) .007** (.001) .004** (.001) .007** (.001) .007** (.001) .004** (.001) Popu la t i on densi ty - .038** (.007) - .027** (.007) .005 (.007) - 04** ('007) - .027** (.007) .004 (.007) - .04** (.007) - .027** (.007) .003 (.007) 5-year pop. growth rate .048 (.028) .043 (.027) -.086** (.029) .052 (.028) .048 (.027) - .089** (.029) .066* (.028) .059* (.027) - .097** (.029) Cons t . -1.024** (•2) - .947** (.159) -1.297** (.157) - .956** (.199) - .898** (.16) -1.29** (.157) - .884** (•2) - .825** (.161) -1.313** (.157) T i m e dummies Yes Yes No Yes Yes No Yes Yes N o Reg ion- t ime dummies Yes N o No Yes N o N o Yes N o N o N 3,290 3,290 3,290 3,290 3,290 3,290 3,290 3,290 3,290 R2 w i t h i n .647 .604 .497 .650 .605 .501 .651 .606 .500 R2 overal l .100 .147 ,126 .093 .143 .129 .093 .142 .128 * and ** denote significance at the 5 and 1 per cent level, respectively. 3.5.2 Instrumental Variables Estimation This section estimates the public expenditure equation by two-stage least squares. A Hausman test supports the use of a fixed-effect I V model versus a fixed-effect O L S model. The set of instrumental variables is city land area and fifteen- and twenty-year lagged values of the planned development measures. 1 9 Table 3.6 reports the results of the I V estimation of total per capita expenditure regressed on P D membership measures. In al l specifications, the coefficient of the lagged private government variable and land area is positive and highly significant in the first-stage regression. 2 0 A Sargan overidentification test of the instruments does not reject the null hypothesis that the instruments are uncorrelated wi th the error term in the local government equation. This gives some confidence to the overall set of instruments used. The instrumental variables estimation indicates the presence of a significant interac-tion effect. Interestingly, the sign differs depending on the inclusion or exclusion of the year and region-year dummies. Column (6) suggests that P D membership rate positively affects local spending at the ten per cent level, indicating a weak strategic complemen-tarity. However, the private government variable, which is trending up, may simply be capturing the upward trend of local expenditure. The fact that the positive coefficient turns to negative when time effects are included suggests that they are picking up a time effect that is driving both public and private government in the same direction. After controlling for time effects (columns 1, 2, 4 and 5), the coefficient on private government becomes negative. This is consistent wi th the view that public governments see private government activity as a strategic substitute. Elasticities of public expen-diture wi th respect to private government are calculated by mult iplying by the mean of the private government measure. For example, Column (4) of this table suggests that for a city wi th the mean rate of planned development membership of .0533, the 1 9 E s t i m a t i o n was also performed w i t h different combinat ions of lags, ranging f rom ten to twenty years, in the first-stage regressions. A s long as the lags used are suff iciently long, they give qual i ta t ive ly s imi lar results. Tables on ly report the best- f i t t ing combina t ion of fifteen and twenty year lags. 2 0 T h e F s tat is t ic for the jo in t signif icance of the inst rumenta l var iables in the first-stage regression using per cap i ta P D uni ts on the instruments is 62.4. In the first-stage regression us ing P D membership rate, the jo in t F s tat is t ic on the instruments is 49.2. (These last two stat ist ics come f rom the specif i-cat ions w i t h year dummies and interacted y e a r - M S A dummies.) F ina l l y , in the speci f icat ion w i t h t ime and M S A - t i m e dummies, the par t ia l R2s of the excluded inst ruments in the first-stage regression are .039 for per cap i ta P D uni ts and .038 for P D membersh ip rate. 49 Table 3.6: IV Estimation Results — A l l Covariates. Private Government Mea-sured by Planned Developments Dependent var iable: L o g per cap i ta real pub l ic expendi ture Instruments: 15 and 20 year lagged pr ivate government measure and land area of ci ty (1) (2) (3) (4) (5) (6) Per cap P D uni ts -7.74** -4.272** .943 (1.63) ( i . i i ) (.972) P D membership rate -3.063** -1.829** .67 (.726) (.494) (.393) % B lack .665** .908** .84** .742** .941** .831** (.224) (.221) (.245) (.225) (.221) (.245) % A s i a n .075 .087 .022 .137 .121 .003 (.118) (.117) (.126) (.119) (.117) (.127) % Hispan ic .128 .377** .576** .139 .32* .504** (.136) (.128) (.082) (.131) (.126) (.085) % w i th 4 years college -.404 .74* 2.537** -.692 .523 2.579** (.363) (.293) (•2) (.436) (.33) (.201) % 17 and younger -1.26** -1.504** - .985** -1.488** -1.597** - l .o i r* (.307) (.293) (.228) (.314) (.294) (.228) % 65 and older -.687 -.242 .098 -1.045** -.434 .077 (.409) (.396) (.391) (.393) (.382) (.383) % w i th welfare income -.46 -1.677** -.942 -.6 -1.884** -1.046* (.538) (.501) (.506) (.541) (.499). (.504) % foreign-born -.348* -.143 .035 -.273 -.073 .096 (.167) (.161) (.12) (.164) (.159) (.117) % owner-occupied homes 1.41** 1.402** 1.061** 1.384** 1.412** 1.114** (.177) (.158) (.165) (.177) (.156) (.162) M e d i a n household income .006** .002 .0004 .006** .002 -.0003 (.002) (.002) (.002) (.002) (.002) (.002) Popu la t i on densi ty - .025** -.01 .023** - .025** -.01 .023** (.007) (.006) (.007) (.007) (.006) (.007) 5-year pop. growth rate .004 .013 - .137** .043 .034 - .145** (.032) (.03) (.031) (.033) (.031) (.031) Cons t . -1.007** - .956** -1.122** - .875** - .896** -1.118** (.175) (.166) (.149) (.168) (.164) (.147). T i m e dummies Yes Yes N o Yes Yes N o Region- t ime dummies Yes N o N o Yes N o N o N • 3140 3140 3140 3155 3155 3155 R2 w i t h i n .646 .61 .517 .65 .612 .515 R2 overal l .062 .118 .123 .067 .122 .123 * and ** denote significance at the 5 and 1 per cent level, respectively. 50 elasticity on the private government variable (with both lagged values and land area instruments) is -.163. That is, a 10 per cent increase in the rate of planned development membership in a city wi l l , on average, decrease per capita total expenditures by 1.63 per cent. This result is consistent wi th the model prediction of Helsley and Strange (2000), where the slope of the public government's reaction function lies somewhere between -1 and 0, implying that strategic downloading is occurring at a less than one-for-one rate. Different combinations of instruments, year dummies and region-year dummies reveal that estimated elasticities of public government at the mean never exceed .5 in absolute value. Next, compare the I V results with fixed-effects O L S results without instruments. Columns 4 through 9 of Table 3.5 presents the results. Ignoring the endogeneity of pri-vate government decreases the magnitude of the strategic effect greatly, wi th calculated elasticities several times lower than the I V case. For example, an increase in planned development membership of ten per cent wi l l decrease loca l expenditures only .5 per cent on average. IV Regressions on Expenditure Categories Finding public sector responses to private government leads to the next question: is strategic interaction the only story that generates downloading? In addition to strategic substitution, the following alternative mechanisms could generate the empirical down-loading result: • Competing governments: Planned developments may act as a check on the spend-ing of local governments. This has parallels to the Leviathan literature, where a large number pf competing municipalities wi l l better reflect the demands of me-dian voters. In the context of this paper, public governments recognize the growing number of homeowners' associations as competitive, r ival governments. This forces them to adjust expenditures downward. The self-help role of private governments is echoed by papers such as Matsusaka (1995), which demonstrates that localities with direct democracy measures can place limits on excessive government spend-ing. • Political activism: Voters in private governments may be able to better exercise 51 lobbying power by concentrating polit ically active homeowners into association boards. In Lang and Danielsen's (1997) summary of a panel discussion on gated communities, Edward Blakely, a noted researcher on urban planning, states that "it's easier to organize these people in 269 housing units through their community association than it is to go door to door and try to organize people in any other place." In other words, association boards take advantage of economies of scale in poli t ical organization, and they often encourage their members to vote as a bloc on matters that affect members and non-members. Planned community members may then be less likely to favour spending on items that benefit primarily those who live outside the community, such as redistributive spending. • Efficiency: Publ ic governments may recognize that some types of local services are actually private goods, in which case an efficiency argument may justify the private sector's provision of supplementary services. Governments may want to focus their efforts providing public services with more "socially beneficial" characteristics. • Cooperative bargaining: Publ ic and private governments may interact with each other, but not necessarily in a non-cooperative game. Bargaining between the governments often arises in the construction stage of a planned development. M u -nicipal officials may require private developers to install municipal infrastructure in the housing development as a condition for allowing the planned development to be built . In return, the developers can set up a homeowners' association to finance the expenses. Spli t t ing total spending into categories helps to determine which mechanisms are more likely. Different categories of local expenditure have different characteristics, and hence some categories may be more likely to respond to private government than others. This section considers eight expenditure categories as dependent variables. Table 3.7 reports the I V results. For comparison, Table 3.8 reports O L S regressions. A s in the last section, the estimated I V coefficients tend to be larger in magnitude than the fixed-effects coefficients. The IV estimates show that public expenditures on police, fire, parks, waste disposal and housing development are downloaded to private governments. There is no evidence of interaction for highways and libraries. Finally, the O L S results indicate downloading of government administration expenditures; however, when the 52 simultaneity is accounted for, the I V estimations do not find a downloading effect once year and metropolitan region effects are included. No category of expenditure has a significant strategic complementary effect. The estimations show a clear distinction between two classes of the services and the effect that private government has on them. This distinction appears to be based on the substitutability between private and public providers of the service. This is consistent wi th the efficiency mechanism: public governments reduce their spending in services for which private governments provide close substitutes. In terms of the magnitudes, the greatest decreases in local spending occur where substitutability is likely the highest: housing and community development, police and waste disposal. These services have a high degree of substitutability between private and public providers. The strong, nega-tive coefficient on policing suggests that residents of planned communities view privately provided security services as substitutes for city policing. It is interesting that a negative sign appears on both public safety services, police and fire protection. The results run counter to the claim in Gordon (2003) that voters in planned communities might support increased spending on public safety services while rejecting spending on redistributive and duplicative services. Waste disposal and recreation facilities have strong private good characteristics: once a homeowners' association picks up its members' garbage or provides them wi th a swimming pool, there is less of a need for the public government to provide these services to members of the private government. O n the other hand, planned developments have no significant effect on public road and highway spending in any specification. The most plausible explanation is that there is no private substitute for public roadways, especially those outside the homeowners' association. Alternatively, the absence of strategic downloading may be due to the con-siderable spending on durable infrastructure. If expenditures primarily pay for capital and equipment, it might be hard to reallocate spending away from this category even when private government is relieving some of the budget. The next section explores differences between capital and current expenditures to see if they are consistent with this explanation. The I V estimation also shows no strategic effect for library expenditures. This could indicate a lack of substitutability between public and private provision. Planned devel-opments rarely provide library services, and so the absence of a strategic decrease in 53 Table 3.7: IV Estimation Results for Individual Expenditure Categories. Pri-vate Government Measured by Planned Developments Sample: 110 cities over 30 years Instruments = 15 and 20 year lagged private government measure and city land area I Dependent variable Coefficient on private Per capita P D units (1) (2) (3) government variable: P D Membership rate (4) (5) . (6) AIL expenditures -7.741** (1.630) -4.273** (1.110) .943 (.972) -3.064** (.726) -1.830** , (.494) .670 (.393) Police -2.114** (.287) -1.444** (.182) -.356* (.117) -1.029** (.152) -.724** (.083) -.165** (.061) Highways & roads -.082 (.360) .081 (.247) .134 (.196) -.057 (.161) .004 (.110) .037 (.079) Fire -.798** (.157) -.522** (.101) -.250** (.082) -.362* (.071) -.244** (.045) -.097** (.033) Parks & recreation -1.214** (.305) -.747** (.205) -.694** (.165) -.555** (.139) -.370** (.092) -.295** (.066) Solid waste disposal -1.585** (.199) -.955** (.120) -.552** (.092) -.660** (.091) -.405** (.055) -.199** (.037) Libraries -.181 (.100) -.279** (.069) -.246** (.055) -.051 (.045) -.099** (.030) -.082** (.022) Housing & community development -2.092** (.612) -1.472** (.404) -.236 (.325) -.633** (.276) -.689** (.181) -.036 (.131) Government administration -.472 (.300) -.778** (.209) -.559** (.168) -.183 (.134) -.346** (.093) -.221** (.068) Time dummies Region-time dummies Yes Yes Yes No No No Yes Yes Yes No No No N 3,140 3,140 3,140 3,155 3,155 3,155 * and ** denote significance at the 5 and 1 per cent level, respectively. 54 Table 3.8: Fixed-Effect Estimation Results for Individual Expenditure Cate-gories. Private Government Measured by Planned Developments Sample: 110 cities over 30 years I Dependent variable Coefficient on private g Per capita P D units (1) (2) (3) ;overnment variable: P D Membership rate (4) (5) (6) A l l expenditures -2.150** (.435) -1.229** (.403) 1.933** (.420) -.976** (.171) -.715**' (.161) .625** (.165) Police -.285** (.063) -.298** (.057) .161** (.058) -.156** (.024) -.174** (.023) .040 (.023) Highways & roads -.014 (.088) .127 (.081) .180* (.076) -.044 (.035) -.001 (.032) .029 (.030) Fire -.081* (.038) -.075* (.034) .053 (.032) -.061** (.015) -.059** (.013) -.0004 (.013) Parks & recreation -.272** (.076) -.154* (.070) -.076 (.066) -.126** (.030) -.097** (.028) -.057* : (.026) Solid waste disposal -.093* (.040) -.099** (.036) -.042 (.034) -.013 (.016) -.021 (.014) .005 (.013) Libraries -.040 (.024) -.057** (.022) -.045* (.020) -.015 (.009) -.028** (.009) -.022** (.008) Housing & community development -.436** (.148) -.504** (.132) -.051 (.126) -.118* (.058) -.159** (.053) .028 (.049) Government administration -.302** (.073) -.298** . (.067) -.155* (.064) -.139** (.029) -.154** (.027) -.086** (.025) Time dummies Region-time dummies Yes Yes Yes No No No Yes Yes Yes No No No N 3,300 3,300 3,300 3,300 3,300 3,300 * and ** denote significance at the 5 and 1 per cent level, respectively. 55 library expenditures is not surprising. However, an alternate explanation is that about one-third of the cities in the sample do not report any library expenditures at all. In smaller municipalities, the county often provides library service. However, regardless of the level of private government, these cities would not be expected to alter their provision of libraries, and so this may contribute to an absence of a downloading effect. Finally, the effect of private government on government administrative spending is either negative or insignificant. This category may represent association members' perception of government. If homeowners' associations are a way to return local decision-making to the neighbourhood level, then members would support tightening the budgets of bloated, inefficient municipal governments.21 However, the empirical results give mixed evidence to support this view. The prevalence of private governments does not seem to be a way for voters to rein in bureaucratic local governments. In addition, government administration, which includes the local justice system, clearly has public good characteristics that make it difficult for private governments to provide a substitute. Effects on Non-Member Households The different substitutability of the various expenditure categories provide a way to examine the efficiency mechanism further. If the services that the public and the private governments provide are true private goods, then the private government's provision simply displaces the public government's provision. The negative downloading result should therefore not be surprising. However, if the only motivation for public sector downloading is to transfer provision responsibility to efficient providers, then public expenditure per non-member should not be affected by the rate of private government membership. In other words, the coefficient on private government membership should be statistically insignificant if the dependent variable is public expenditure per non-member. The dependent variable is real city expenditure, divided by the number of housing units that are not planned development units. It is not possible to know the population 21Indeed, many critics of private governments point to this as a sign of the deterioration in trust of government: "A study of gated communities, most of which are CAs [community associations]...sees them as 'a dramatic manifestation of a new fortress mentality growing in America.' The exclusivity of CAs may induce their members to evade public service to a larger community. The money needed to pay for such exclusivity can increase the costs of CAs." (Stabile 2000) 56 of the city that does not live in planned developments, so the use of the non-planned development housing units is a justified alternative. Table 3.9 presents the fixed-effects O L S and I V results for total expenditures and the eight specific categories of expenditure. When municipal expenditures are considered per non-member housing unit, the O L S results generally provide no evidence of downloading. Private government membership has no impact on total expenditures per non-member. When Individual categories are examined, downloading is evident only for parks and recreation and housing development spending. Police, libraries and government administration spending per non-member are unaffected by the degree of private government membership. A n d most interesting, the planned development coefficient estimate on highways, fire protection and, to a lesser extent, garbage disposal is positive, indicating complementarity between public and private governments. However, once the instrumental variables control for the endogeneity bias, an in-teresting result brings up additional questions. Publ ic downloading is st i l l present for police, parks, waste disposal and housing development spending, and, in one of the specifications, total spending. Not only is the public government responding to the in-creased private government membership by cutting expenditures, it cuts expenditures per non-member. This indicates that something beyond an efficiency motive drives the strategic downloading. The polit ical activism mechanism may help explain this result. Voters in planned developments may be less inclined to approve funding that they know is destined for non-members. This may have serious welfare effects, as non-members experience a decrease in government expenditure that goes beyond simple displacement of private provision for public provision. Current Expenditures Versus Capital Outlays Not only can cities reallocate spending between expenditure categories, but they can also reallocate wi thin them. Ci ty expenditures can be divided into current operating expenditures and capital outlays. A city can reduce spending by cutting current expen-ditures, such as reducing staff, providing less frequent trash pickup or street cleaning and so on. Alternatively it can lower capital outlays by postponing infrastructure in-vestments and construction projects. Figure 3.6 separates the mean per capita total 57 Table 3.9: IV Estimation Results for Individual Expenditure Categories, Mea-sured Per Non-Planned Development Household. Private Government Mea-sured by Planned Developments Sample: 110 cities over 30 years Instruments = 15 and 20 year lagged private government measure and city land area Coefficient on private government variable: Per capita P D units P D Membership rate J, Dependent variable O L S I V O L S I V A l l expenditures .098 -3.166* .120 -1.021 (.432) (1.576) (.170) (.703) Police .188 -2.102** .080 -1.133** (.133) (.560) (.052) (.261) Highways & roads .543** .850 .201** .386 (.197) (.792) (.077) (.356) Fire .360** -.596 .124** -.271 (.084) (.346) (.033) (.156) Parks & -.266 -2.973** -.125* -1.289** recreation (.159) (.636) (.063) (.289) Solid waste .086 -3.132** .095* -1.270** disposal (.096) (.462) (.038) (.210) Libraries .031 -.187 .026 .045 (.055) (.221) (.022) (.099) Housing & community -.684* -2.923* -.120 . -1.148* development (.316) (1.284)- (.124) (.577) Government -.203 .311 -.090 .194 administration (.156) (.635) (.061) (.285) N 3,290 3,140 3,290 3,155 * and ** denote significance at the 5 and 1 per cent level, respectively. A l l specifications include year dummies and region-year dummies. 58 Table 3.10: IV Estimation Results for Current Expenditures Versus Capital Outlays. Private Government Measured by Planned Developments Sample: 110 cities over 30 years Instruments = 15 and 20 year lagged private government measure and city land area Coefficient on private government variable: Per capita P D units P D Membership rate J. Dependent variable (1) (2) (3) (4) (5) (6) Capi ta l Outlays -5.294 -.232 -1.598 -1.233 .236 .405 (6.498) (4.386) (3.524) (2.909) (1.959) (1.423) R2 wi thin .151 .084 .045 .156 .084 .046 R2 overall .017 .015 .010 .019 .018 .012 N 3,138 3,138 3,138 3,153 3,153 3,153 Current Expenditures -7.380** -4.129** 2.380** -2.929** -1.744** 1.286** (1.326) (.880) (.839) (.590) (.390) (.340) R2 wi thin .758 .747 .628 .761 .750 .625 R2 overall .097 .175 .180 .105 .182 .177 N 3,140 3,140 3,140 3,155 3,155 3,155 Time dummies Yes Yes No Yes Yes No Region-time dummies Yes No No Yes No No * and ** denote significance at the 5 and 1 per cent level, respectively. expenditure into current operating expenditures and capital outlays. Clearly, growth in current expenditures is driving growth in total expenditure. Real spending on capital and construction is flat over the thirty-year sample period. To see if the proliferation of planned developments affects these two types of expen-ditures differently, Table 3.10 reports the results from separate I V regressions of capital and current expenditures. The estimations demonstrate that cities wi th high levels of planned developments have no significant effect on capital outlays, but they cause a sig-nificant decrease in current expenditures. This suggests that current spending is much more manipulable than capital spending, which provides further support for a strategic substitution interpretation for the downloading result. 59 Figure 3.6: Capital Outlays Versus Current Operating Expenditures $1,200 -i 1 $0 - I — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i— I O L O O LT5 O L D r^ . oo co cr> cn CTl CTl CT) CTl CTl CT) Year Source: Census of Governments and Annua l Survey of Governments, 1970-1999 60 3.5.3 Are Planned Developments Different from Condomini-ums? The results of the previous section indicate the presence of a strong, negative effect of planned developments on local public spending, even in a period of increasing expendi-tures. But planned developments make up only one type of residential private govern-ment. Indeed, there are more condominiums than planned developments in California. Does the spread of condominiums have a similar effect on local public expenditures? This question's motivation is that currently, local policy and laws treat planned devel-opments and condominiums equivalently as a "common interest development." Some authors argue this is inappropriate: Rosenberry (1996) states that "legislation often treats these projects as if they have the same problems, and they do not." How different are condominiums from planned developments in practice? Aside from the legal distinction on ownership, there are differences on the membership, size and lo-cation that would suggest different effects on local expenditures. First , condominiums are inhabited by people who place a high value on amenities, and their monthly assess-ments are higher in account of it. Homeowners in condominiums may therefore be more receptive to local spending, which may lead to a dampening of any strategic downloading by public authorities. Second, condominiums are more numerous than planned developments, but they have fewer housing units on average. The median condominium development in the database has 18 housing units, while the median planned development has 76. The greater size of planned developments may confer some economies of scale in poli t ical activism. Therefore, homeowners' associations in planned developments may attract boards of directors that are more wil l ing to get involved in local politics, compared to homeowners' associations in condominiums. Table 3.11 presents some summary statistics of condominium prevalence, and F ig -ure 3.4, presented earlier, graphs over time the overall condominium membership rates alongside P D rates. Like planned developments, condominiums have experienced dra-matic growth. However, even in cities wi th the highest rates of condominium member-ship, condominiums make up less than a thi rd of the housing units in the city. Compare this to some cities that have over half of their housing units in planned developments. 61 Table 3.11: Summary Statistics of Condominium Measures Sample: 110 cities over 30 years Year(s) Mean rate Std. Dev. . M i n . M a x . Number of condo units per capita 1970 only " .003 .008 0 .062 1999 only .029 .029 0 .214 1970-1999 .020 .026 0 .278 Condo rate as a proportion of all housing units 1970 only .8 % 2.6 % 0 % 21.4 % 1999 only 7.5 % 5.6 % 0 % 28.6 % 1970-1999 5.0 % 5.1 % 0 % 30.2 % Source: Levy and Company 2003, and author's own calculations There is also an important geographic difference between the two types of private government. Figure 3.7 shows a county map in which condominium membership rate is shaded. Condominiums are more common than planned developments in central urban counties of Los Angeles, San Diego and San Francisco, reflecting the premium for space. In suburban counties such as San Bernardino, Monterey and Sonoma, planned developments dominate. 2SLS estimations, wi th the same set of instruments, are run with private government measured by condominiums. Table 3.12 reports the results. The difference between con-dominiums and planned developments is striking. As the first row of the table indicates, there is very little evidence of strategic downloading by local governments in the face of condominium activity. Al though Columns (3) and (6) show a strong, significant positive coefficient on condominium prevalence, these are specifications that do not include year dummies. A s condominium growth is strongly positive during the sample period, this suggests that coefficient is simply reflecting this trend. Once the model controls for year effects, the significance disappears. The regressions on individual service categories gives similar interpretations. Only two services show strong negative coefficients on condominium prevalence: fire and waste disposal. Police spending has a significant negative effect in one specification. Rather than reflecting genuine public downloading, these estimates may simply indicate that 62 Figure 3.7: Rate of Condominium Membership by County Condominium Membership Rate. 1999 (based on 110-city panel) iHumboldt San Francisco San Mateo Santa Cruz San Bernardino [Santa Barbara M ^Ventura\ l o s ^ Angeles > Imperial Source: Levy and Company, 2003, and author's own calculations 63 the higher densities of condominiums produce economies of scale in emergency response and garbage pickup, rather than reflecting strategic substitution. In addition, fire codes have become more strict over the period, so a city wi th recent condominium development may not need to spend as much on fire protection services. Next, a question arises as to whether or not land area is a good instrument for condo-minium proliferation. Land area is a suitable instrument for planned developments as it proxies for residential development. However, if condominiums tend to take up vertical space rather than lateral space, land annexations may not reflect any growth in condo-miniums. In addition, many condominiums are the result of apartment conversions and hence would not require additional city land. To remove the possibility of an unsuitable instrument, the condominium regressions are repeated, dropping land area from the list of instruments. Table 3.13 shows the results. They st i l l show very little sign of strategic substitution. Al though condominiums have experienced a much more rapid growth than planned developments, the results of this section find no interaction effect by public governments. Hence, local policy that addresses these two types together as a broad class of common interest development neglects the different impacts that they have. 3.6 Further Directions The idea of private homeowners' associations taking over roles of public authority is a rich topic for further research. This paper demonstrates that public governments do treat planned developments as providers of substitute services, and so they respond ac-cordingly. A natural question for future research is to tie local government decisions to voter choices. This paper finds evidence of strategic downloading, but is voter behaviour consistent with this? Are local governments responding to voter preferences for lower expenditures when more of the population belong to homeowners' associations? The answers depend on the way local government decision making is modelled. This pa-per assumes governments that maximize aggregate welfare of constituents, but models that maximize the median voter's ut i l i ty or the city budget are common as well. A n interesting extension would be to be see if the difference in behaviour between planned development members and non-members can say anything about competing models of 64 Table 3.12: IV Estimation Results for Individual Expenditure Categories. Pri-vate Government Measured by Condominiums Sample: 110 cities over 30 years Instruments = 1 5 and 20 year lagged private government measure and city land area I Dependent variable Coefficient on private Per capita condo units (1) (2) (3) • government variable: Condo membership rate (4) . (5) (6) A l l expenditures -.830 (1.005) .532 (.928) 5.318** (.625) -1.065 (.892) .322 (.743) 3.750** (.704) Police -.397* (.168) -.032 (.147) .948** (.095) -.243 (.148) .010 (.117) .680** (.108) Highways & roads -.359 (.234) -^252 (.212) -.080 (.127) -.430* (.204) -.256 (.171) -.104 (.141) Fire -.300** (.098) -.162 (.085) .057 (.083) -.325** (.092) -.205** (.071) -.011 (.058) Parks & . recreation .010 (.190) -.006 (.173) .029 (.168) .031 (•167) .002 (•138) . -.003 (.116) Solid waste disposal -.565** (.110) -.464** (.098) -.112 (.093). -.409** (.102) -.325** (.081) -.029 (.064) Libraries -.096 (.065) -.002 (.058) .052 (.035) -.099 (.056) -.020 (.046) .008 (.039) Housing & community development -.023 (.387) .069 (.344) .062 (.340) -.051 (.341) .145 (.275) .470* (.235) Government administration -.182 (.193) -.094 (.177) .194 (.173) -.186 (•171). -.130 (.142) .052 (.119) Time dummies Region-time dummies Yes Yes Yes No No No Yes Yes Yes No No No TV 3,140 3,140 3,140 3,155 3,155 3,155 * and ** denote significance at the 5 and 1 per cent level, respectively. 65 Table 3.13: IV Estimation Results for Individual Expenditure Categories. Pri vate Government Measured by Condominiums Sample: 110 cities over 30 years Instruments = 15 and 20 year lagged private government measure J. Dependent variable Coefficient on private Per capita condo units (1) (2) (3) government variable: Condo membership rate (4) (5) (6) A l l expenditures -.039 (1.042) 1.050 (.932). 5.317** (1.030) -.367 (.926) .673 (.743) 3.857** (.739) Police -.036 (.170) .123 (.146) 1.071** (.157) .104 (.150) .113 (.117) .827** (.116) Highways & roads -.316 (.243) -.202 (.213) -.020 (.127) -.395* (.208) -.221 (.171) -.048 (.746) Fire -.189 (.100) -.107 . (.085) .104 (.085) -.228* (.092) -.169* (.071) .031 (.060) Parks & recreation -.053 (.198) .049 (.173) .149 (.172) -.008 (.175) .039 (.138) .123 (.121) Solid waste disposal -.498** (.113) -.440** (.098)' -.095 (.095) -.340** (.104) -.308** (.081) -.008 (.067) Libraries -.044 (.067) .019 (.058) .105 (.057) -.049 (.060) -.005 (.046) .064 (.040) Housing & community development -.378 (.404) .149 (.346) .580* (.346) .316 (.361) .203 (.277) .451 (.246) Government administration -.172 (.201) -.011 (.177) .334 (.176) -.176. (.179) -.075 (.142) .195 (.124) Time dummies Region-time dummies Yes Yes Yes No No No Yes Yes Yes No No No N • 3,140 3,140 3,140 3,155 3,155 3,155 * and ** denote significance at the 5 and 1 per cent level, respectively. 66 local government. However, data on a finer level are required to compare the voting behaviour of planned development residents versus that of non-members. 3.7 C o n c l u s i o n Residential private governments take public authority and transfer it to an associa-tion of private homeowners. A s homeowners' associations become more and more the usual method of building and servicing new housing developments, questions arise as to whether they have an impact on local governments. This paper examined whether or not local public governments reacted to the presence of private governments by altering pro-vision levels. A theoretical model underlines the conditions in which public and private government spending are strategically complements, and when they are strategically substitutes. A panel study of California municipalities from 1970 through 1999 suggests that downloading does occur. This is consistent with public and private activities being strategic substitutes. For a ten per cent increase in planned development membership in a city, local expenditures fall roughly 1.6 per cent. The result occurs even in an era of increasing public expenditure. When local expenditures are split up into spending categories, private government activity produces a downloading effect for those categories expected to have a high degree of substitutability between public and private providers. The empirical analysis suggests that local public governments understand that homeowners' associations are taking over responsibility for providing many public services, and they therefore reduce their expenditures in account of it. Finally, the study stressed the difference between planned developments and condominiums. Whi le both are often treated symmetrically by policy makers, the analysis shows that condominiums do not exert the same impact on local governments as planned developments do. 67 C H A P T E R 4 T h e Effect of Property Tax Limitations on Residential Private Governments 4.1 Introduction In most jurisdictions, property taxes form the primary source of local government rev-enue, and they are often viewed as an unfair and unpopular tax. The tax revolt era in the 1970s and the 1980s in the United States ushered in a series of attempts of voters to l imit the power of local governments to increase property taxes. This chapter asks whether state-imposed property tax limitations on local governments encourage formation of and membership in homeowners' associations. It begins by proposing a theoretical model to examine the effect of a tax l imitat ion on the decision of residents whether or not to join a homeowners' association providing supplementary services. The implications of the model are tested wi th empirical data from California in the years surrounding Proposit ion 13. In 1978, Proposition 13 effectively constrained all residential to a one per cent property tax rate, resulting in drastic reductions in local government revenue. Using a panel of public spending in California cities and a novel data source on homeowners' associations, the empirical model tests whether or not cities that were more property-tax constrained experienced higher rates of private government formation. The study controls for the importance of state aid and user charges in city budgets. The results demonstrate the level and the growth of private government 68 membership were significantly increased by the imposition of Prop 13. Referring back to Figure 3.3 shows the incredible proliferation of homeowners' as-sociations in California over the last thirty years. The graph shows that the surge in association incorporation occurred in the late 1970s and early 1980s. This coincided wi th a period of fundamental change in local public finance, the property tax revolt. Voters across the United States reacted to perceived government inefficiency by passing strict limitations on property tax. California, wi th the approval of Proposit ion 13 in 1978, is often considered the birthplace of the property tax revolt. D i d the stringent conditions of Proposition 13 contribute to the tremendous growth of private governments? A t the same time, however, changes in the construction industry and the booming housing mar-ket also contributed to private government growth. Part of the challenge of this paper is to pul l out the effect of the property tax l imitation. This chapter proposes a theoretical model to study the effect of an exogenous tax rate l imitation on the behaviour of private governments. The model is based on residents who decide whether to join a homeowners' association when choosing a place to live. The comparative static of interest in how the membership margin changes with the imposition of a property tax l imitation. The theoretical model leads to empirical testing. The empirical analysis combines novel data on homeowners' associations in California wi th property tax data to see whether cities that faced stronger property tax constraints are more likely to encourage private government growth. The chapter is organized as follows. Section 4.2 presents an overview of state-imposed property tax limitations and their connection to private governments. Section 4.3 out-lines a theoretical model of property tax l imitat ion and derives results. The empirical evidence begins in Section 4.4 by describing the provisions and effects of Proposit ion 13 in California. Section 4.5 introduces the empirical framework, describes the data and provides the empirical analysis. Section 4.6 concludes. 69 4.2 Property Tax Limitations and Private Govern-ment Property taxes, the main source of revenue for local governments, have never been popular. Voters often describe them as regressive and unfair, and in the late 1970s and 1980s, they unleashed the property tax revolt. This period marked a fundamental desire by voters to l imit the taxing and spending authorities of municipal governments. Shadbegian (1998) notes that in that period, half the states in the country passed some kind of measure l imit ing the degree to which governments can raise or spend money. These measures took many forms: explicit l imits on the property tax rate or level, a l imitat ion on the amount of yearly tax increases, a l imitat ion of the rate at which property value assessments can increase and ceilings on the level and growth rate of expenditures. Together, these measures are known as tax and expenditure limitations ( T E L s ) . T E L s share several similarities regardless of where they are enacted. They are almost always statewide measures that affect most, if not all , local governments in the state. T E L s are generally approved through statewide ballots. 1 Finally, T E L s tend to: be effective: local governments suffered large decreases in property tax revenue, and they remain unable to increase property taxes beyond the mandated l imit without voter approval. 2 The effect of T E L s on local public finance has generated a line of research on the way that local governments deal wi th their imposition. Previous research, both theoret-ical and empirical, has suggested three main options cities had to deal wi th the budget shortfall: cut expenditures, turn to the state and look for alternate sources of revenue. The following gives brief descriptions of these three options, provides some related re-search and proposes another approach: the shifting of public responsibility to private governments. 1 Indeed , V igdo r (2004) argues that T E L s are l ikely to succeed at statewide bal lots rather than loca l bal lots because voters want to have a way to contro l t ax rates i n jur isd ic t ions where they do not l ive. 2 I n some states, such as Ca l i fo rn ia , the T E L has even been made part of the state const i tu t ion. 70 4.2.1 Expenditure Cuts A n obvious response to a revenue shortfall is to cut expenditures. The ostensible goal of most T E L s is to l imit perceived government inefficiency and overspending. Economic research suggests that T E L s do affect spending substantially. Shadbegian (1998) uses a panel data set of local government budgets, aggregated to the state level, to show that limitations have decreased both the level and the growth rate of expenditures. In addition Figlio and O'Sull ivan (2001) provide evidence that local authorities may have an incentive to reduce spending on certain items in order to manipulate voters to override a T E L ' s limits. For instance, a city may cut expenditure greatly on visible "service" inputs such as police officers while reducing administrative expenditures only slightly. Further empirical research on the effects of tax l imitat ion analyses the impact on the quality, as opposed to the level, of local expenditures following the l imitation. Figl io (1998) looks at a tax measure in Oregon and concludes that student-teacher ratios increased significantly as a result, while Bice and Hoyt (2000) predict that a fiscal l imit wi l l increase voluntary contributions to the local public sector. 4.2.2 State A i d Cities can turn to the state for help either through state bailout funds to cover deficits or by shifting functional responsibilities back to the state. Authors such as O'Sull ivan, Sexton, and Sheffrin (1995) have suggested that the blow of T E L s are often cushioned by generous state grants. Shadbegian (1998) provides evidence that increased federal or state funds have acted as a substitute for local expenditure cuts. There is also a view, presented by Joyce and Mul l ins (1991) and Schwadron and Richter (1984), that there has been a gradual shifting of responsibility from local governments to state government. 4.2.3 Alternative Sources of Revenue There is much research to suggest that cities have turned to alternative sources of rev-enues such as charges and fees. T E L s caused municipalities to reduce their dependence 71 on the property tax and turn instead to charges, fees and sales taxes. In particular, many authors have pointed out the dramatic increase in impact fees for new development. Shadbegian (1999) demonstrates in a national analysis that T E L s caused municipalities to increase their fees and charges, which partially offset the decrease in property taxes. 4.2.4 A Fourth Option: Private Providers While empirical research supports the use of alternative revenue generating instruments after a property tax l imitation, there is relatively little research on the supplementary role of private providers of public services. Theoretical research into the area of homeowners' associations suggests that they can be viewed as "private governments." A homeowners' association ( H O A ) is a private association that has the authority to provide public services to residents of a neighbourhood. A n association can levy fees for the provision of these services and has abilities to enact regulations and to enforce them. The association has characteristics of a government because by buying a property within a neighbourhood that is governed by a homeowners' association, a resident is automatically a member and subject to all the fees and regulations imposed. The services provided by private government associations are similar to those pro-vided by a traditional local government. These include garbage collection, security (gates or guards), parks and recreation and even private roads. Munic ipal governments may even download the responsibility of providing infrastructure to private associations. Be-cause homeowners who live in an association pay both association fees and public prop-erty taxes, the private government and the public government supplement each other's service levels. This suggests that T E L s on municipal budgets may encourage the for-mation of private governments as a way to supplement decreased public expenditures. Indeed, to a fiscally constrained municipality, encouraging private governments not only means downloading responsibility of service provision to private providers, but also be-ing able to collect development impact fees and surcharges. This makes the formation and proliferation of private governments even more attractive. The option of private providers of public services as a response to T E L s has not been examined thoroughly in economic literature. Yet it is an interesting question to ask as this option may have serious consequences on the quality of public services 72 residents experience. Sonstelie (1979) suggests that Prop 13 may prompt some parents to switch their children from public schools to private schools. Whi le in the short run this may benefit public school students by spreading expenditures over a smaller number of students, in the long run, this may tend to erode polit ical support for public schools. Chapter 3 of this dissertation provides empirical evidence that public governments may respond to the proliferation of private providers by reducing expenditures, especially for services where private associations provide close substitutes. 3 It also finds evidence that non-private government members may suffer disproportionately from this public response. Thus, the evidence suggests that private associations may have large impacts on areas that formerly belonged solely to the public sector. Given that the motivation exists for studying the effect of a T E L on private govern-ment formation and membership, the next section builds a simple theoretical model fo examine how private government composition can be affected by an exogenous property tax l imitation. This is then followed by an empirical test of the main result, using data from California's Proposition 13. 4.3 T h e o r e t i c a l M o d e l This section presents a model of private government membership and considers the effect of an exogenous property tax decrease on membership behaviour. In this context it is not necessary to model the behaviour of the public and the private governments in order to get a testable implication for the effect of a tax and expenditure l imitat ion on private government membership. A relatively simple housing market is incorporated into a model of private government membership choice. The model combines elements from the private government model of Helsley and Strange (1998) and the multi-community sorting model of Calabrese and Epple (2004). A major difference between this model and Helsley and Strange (1998) is that here, the public and private government tax rates wi l l be taken as exogenous. Also, Helsley and Strange do not incorporate a housing choice. The notable departure from Calabrese and Epple (2004) is that they consider how a tax l imit affects sorting among jurisdictions, whereas this paper examines a l imit 's effect on sorting within a jurisdiction. Here a resident has the option to join a private government 3 A s a reminder: policing, parks and waste disposal, for instance. 73 that supplements the municipality's services. The focus of the model will be the membership decision of the residents of a city. The land area and the population of the city are fixed. All residents have utility function U(G, h, x), where G is a level of public services, which must be obtained from a govern-ment, h is units of housing and x is a numeraire consumption good. The function U is twice differentiable, strictly concave and strictly increasing in, each of its arguments. Public services are provided by two governments: the municipal government (the public government) provides g to all residents of the city, whereas a homeowners' asso-ciation (the private government) provides a supplemental level of service 7 to members only. The g and 7 are publicly-provided private goods, and they are perfect substi-tutes in consumption. This model does not explicitly model the governments' provision decisions; rather, the public good levels are determined via the governments' budget con-straints and the exogenous tax rates.4 The governments finance the provision of their goods differently: the public government levies a property tax rate t on the value all city residents' houses. The private government charges a fixed fee of r to each member, which is presumed to exactly offset the marginal cost of providing the service to the member. Each resident lives in a house, which is measured in units h. The house is either located inside the private government or outside; membership in the private government is thus tied to the housing choice, h may represent the quality of the house, the number of rooms, its size and so on. The housing markets for houses within and outside the private government are separate. A unit of h costs pM if it is inside the private government and pN if it is outside. pM and pN are the annualized rental cost of a unit of housing services. Residents differ in their money income, y, and they are identified by this income level. Income is spent on taxes, housing and the numeraire good, y is distributed according to the density function f(y) with support [y, y]. For simplification, y is normalized to 0 and y is normalized to 1. Thus the problem of a consumer with income y will differ depending on whether the consumer is a private government member or not. A homeowner who 4For papers that explicitly model the governments' provision choice, see Helsley and Strange (1998) and Helsley and Strange (2000). 74 belongs to a private government (hereafter called a member) w i l l solve maxU{g + j,h,x) s.t. pM(l + t)h + x = y - r . (4.3.1) h,x Whereas a homeowner who does not belong to the private government (a non-member) wi l l solve the following: maxU(g,h,x) s.t. pN (1 + t)h + x = y. (4.3.2) h,x The uti l i ty function is posited to be separable in the public good and the private goods: U(G,h,x) = v(G)u(h,x), wi th the additional assumption that u(h,x) is homo-geneous of degree 1. This is consistent with empirical evidence that the income elasticity of housing is approximately 1. Solving the consumers' uti l i ty maximization problems, the linear homogeneity of u(h, x) implies that the corresponding indirect uti l i ty functions are linear in net income: VM(y) = v(g + 7)[(y - r)w(pM(1 + t))] for a member, and (4.3.3) VN(y) = v(g)[yw(pN(l + t))} for a non-member. (4.3.4) where w(p(l + t)) is a decreasing function of the gross-of-tax housing price. For each income level y, the consumer chooses to join the private government if VM(y) > VN(y), not join if VM(y) < VN(y) and is indifferent if they are equal. 4.3.1 The Housing Market The housing market must clear both inside and outside the private government. A s u(h,x) is homogeneous of degree 1, the demand for housing can be written as (y — T)hd(pM(l + t)) for a member, and (4.3.5) yhd(pN(1 +1)) for a non-member. (4.3.6) wi th hd being a function that varies negatively wi th housing price. Housing supply is modelled in a similar way as Calabrese and Epple (2004). Housing is produced by price-taking firms from land and non-land factors with a constant-returns 75 neoclassical production function. The price of non-land factors is fixed and uniform both inside and outside the private government. The land area inside and outside the private government is fixed at LM and LN, respectively. The housing supply functions inside and outside the private government are therefore Hl{pi) = Uhs(jP),j = MovN, where hs is housing per unit of land, depending positively on the net-of-tax price of housing. Housing market equilibrium requires the following two equations hold: Here M is the set of all the income levels that choose to be members, and N is the set of all incomes that choose not to be. 4.3.2 Equilibrium W i t h al l the elements of the model introduced, it is possible to define an equilibrium. Definition 1 Given a distribution of household incomes y, an equilibrium is an alloca-tion of households across communities such that 1. Consumers are maximizing their utility according to their choice of h and x. This implies that both the land inside the private government and the land outside the private government are occupied, and no one wants to move; 2. The housing markets inside and outside the private government clear; and 3. Both the private government's and the public government's budgets are balanced. In keeping with the traditional terminology, conditions 1 is a condition for intercom-munity equilibrium, and conditions 2 and 3 are conditions for internal equilibrium.5 The following sections present conditions that must be satisfied for equilibrium. 5 T h e te rm intercommunity is somewhat imprecise here as bo th land inside and outside the pr ivate government is part of the ci ty; however, th is te rm is a holdover f rom models of sor t ing across mul t ip le jur isd ic t ions. It is impor tan t to remember that the publ ic government continues to serve residents inside the pr ivate government. (4.3.7) (4.3.8) 76 4.3.3 Intercommunity Equi l ibrium The following proposition presents conditions that characterize intercommunity equilib-r ium, which is characterized by residents choosing whether to join the private government or not. Proposition 1 Both land inside and land outside the private government are occupied by incomes in a single interval. There exists a unique household y* who is indifferent between membership and non-membership. If(l — r)v(g+j)w(pM (1+t)) > v(g)w(pN(1 + t)), then those y < y* strictly prefer to be non-members, while those y > y* strictly prefer to be members. Proof. Using Equations (4.3.3) and (4.3.4), the difference in uti l i ty between mem-bership and non-membership for income y is AV(y) = v(g + <y)[(y-T)w(pM(l+t))}-v(g)[yw(pN(l + t))} (4.3.9) = [v(g + 1)w(pM(l + t))-v(g)w(pN(l+t))}y -Tv(g + -f)w(pM(l + t)). (4.3.10) AV(y) is linear in y. A V ( 0 ) = — Tv(g + ~f)w(pM(1 + t)) < 0, so the poorest individual never joins the private government because the private government's fixed fee is too high. O n the other hand A V ( 1 ) = (1 - r)v(g + ~f)w(pM(l + t)) - v(g)w(pN(l + t)). B y assumption this is positive, so the richest individual always prefers to join the private government. This last assumption implies that v(g + ~/)w(pM(l + t)) — v(g)w(pN(l + t)) > 0, which is the slope of the AV(y) function. Using the linearity and continuity of AV(y) and the Intermediate Value Theorem, this implies the existence of a unique y* for whom AV(y*) = 0. For y < y*, AV(y) < 0, so these consumers do not join the private gov-ernment, while for y > y*, AV(y) > 0, so these consumers join the private government. • Indeed the linearity of the indirect uti l i ty functions means it is possible to divide the households into three intervals by income: • For those y G [0, r ] , VM(y) is always negative. The membership fee exceeds their income, and so these households would never join, regardless of the extra level 77 of public service provided by the private government. This is an artifact of the normalization of incomes. • For those y G [r, y*], while they can afford the membership fee, the benefit from the supplemental public service does not cover the costs of membership and the capitalized housing price differential. These individuals do not become members either. • For those y € [y*, 1], the marginal benefit of the extra public service covers the marginal cost of membership, as measured by the membership fee and the capi-talized housing price, so they join. The proposition implies that by identifying y*, who can be called the marginal mem-ber, the population of members and non-members can be determined. The marginal member satisfies AV(y*) = 0 v(g + y)(y* - r)w(pM(l + t)) = v(g)y*w(pN(l + *)). (4.3.11) 4.3.4 Internal Equi l ibrium The other four conditions constitute internal equilibrium. These are the housing market clearing conditions and the government's budget constraint. In all these conditions, the fact that residents sort according to the intercommunity equilibrium condition is used. O n land within the private government, the housing market clears, giving hd(PM(l + t)) f{y-r)f{y)r\y = LMhs{pM). (4.3.12) Jy* O n land outside the private government, the housing market equilibrium condition is hd(pN(l+t)) I" yf(y)dy = LNhs(pN). . (4.3.13) Next, the two governments' budget constraints must bind. The private government's budget constraint is tr ivially satisfied as members pay exactly the marginal cost of providing the private supplement to them. The marginal cost of provision is normalized to 1, so the private government budget constraint is r = 7 . (4.3.14) 78 Finally, the budget constraint of the public government must hold. It is assumed that the marginal cost of providing g is also 1. The public government's budget constraint is therefore tpNhd{pN{l + t)) fV yf(y)cly + tpMhd(pM(l + *)) j \ y - r ) / ( y ) d y = g. (4.3.15) Jy J y* Equations'(4.3.11), (4.3.12), (4.3.13), (4.3.14) and (4.3.15) describe the equilibrium. In this simple model, the private government and the public government tax rates, r and t, are taken as exogenous parameters. In other words, this paper does not model how the taxes are set, as the interest is how the private government membership decision responds to changes in tax rates. 6 The endogenous variables are therefore pM, pN, y*, g and 7 . The comparative static of interest is how the membership margin changes with an exogenous decrease in the property tax rate. In other words, what is the sign of ^ ? Unfortunately, the five equilibrium conditions do not permit the derivation of a signable comparative static. The next section presents a computational model for which an equilibrium and its comparative statics can be computed. 4.3.5 Computational Solutions Residents' incomes y are assumed to be drawn from a uniform distribution over [0,1]. Residents have identical ut i l i ty functions U(G,h,x) = gphaxl-a. (4.3.16) Using Equation (4.3.3), the indirect uti l i ty function of a private government member wi th income y is thus ~aa(l - aY~a~ VM(g + 1,h,x) = (g + jf(y-r) _(pM(l + t))a_ ' (4.3.17) From Equation (4.3.4), the indirect ut i l i ty function of a non-member wi th income y is VN(g,h,x) = (g)Vy V( l+ t )H ' (4.3.18) 6 A l l o w i n g for the endogenous determinat ion of t ax rates wou ld compl icate the equ i l i b r ium condi t ions and the comparat ive stat ics considerably. 79 The next step is to derive the housing price conditions. The housing supply functions are assumed to take the following form: Hi(p>) = Lj(p>Y,j = MoxN. (4.3.19) In the private government, the housing market equilibrium condition is a (V - r)f(y)dy = LM (pMy. p M ( l + t ) J y , While outside the private government, the housing market equilibrium condition is (4.3.20) a yf(y)dy = LN{pNy. (4.3.21) W i t h the assumption of a uniform distribution of incomes over [0,1], it is possible to calculate the net housing prices directly. They are V N V M " a ' i "• 1 lLN\ . 1 + t . ' a ' i l + M 1 lLM. 1 + t 1 1 1+M (y*)1+>1 , and i - (y*)2 r(i-y*) 2 1 + M Next, the membership decision condition takes the form AV(y) = \(g + 7 f l-a aa(l - a) (pM(i+t)Y\ - 9 h , 1 - a aa(l - a ( p N ( l + t ) ) a J y-r(g+jY (4.3.22) (4.3.23) aa{l - a)l-a [pM{i + t)Y\ (4.3.24) This is linear, which implies the existence of a unique marginal member. Now in order for the slope of AV(y) to be positive, it is also necessary that (g + -yf > Now the marginal member y* satisfies A.V(y*) — 0. The value of y* can be explicitly determined: „• = ^ + 7)V"(l + t))° (4 3 2 5 ) Note that in order for y* to be positive, (g + j)0(pN(l + t))Q - g<3(pM(l + t))a must be positive. This can be written g + l\p ( P M ( l + t) " 1 pN(l+t) 80 But this condition is exactly what is needed for the slope of AV(y) to be positive, so that if there is a marginal member in equilibrium, those residents wi th incomes less than y* must strictly prefer not to join the private government, while those wi th incomes greater than y* strictly prefer to join. The final conditions to be satisfied are the governments' budget constraints. The private government budget constraint is t r ivia l : r = 7 . (4.3.26) The public government's budget constraint has the form at T+t rv rv / yf(y)dy-r / f(y)dy (4.3.27) W i t h y distributed uniformly on [0,1], yf(y)dy = | , and J*1, f(y)dy = l—y*. The public government's budget constraint thus reduces to at T+t - - r ( l 2 v (4.3.28) The equilibrium values of pN, pM, y*, 7 and g can be solved from equations (4.3.22), (4.3.23), (4.3.25), (4.3.26) and (4.3.28). To calculate the solution, the following parame-ter values are used: (a = 0.37, p = 0.111, t = 0.28, r = 0.05, LM = 0.1, LN = 0.9, ^ = 3). The parameters are chosen as follows. Following Calabrese and Epple (2004), a and P are chosen such that if G, h and x were privately purchased goods, the consumer would spend one-third of gross income on housing, and ten per cent of gross income on G. p is the ratio of non-land to land inputs in housing, production; Epple and Romer (1991) give evidence that p = 3 is appropriate. A LM/LN ratio of 0.1 is arbitrary but is chosen to represent development in the suburban fringes of the ci ty. 7 The private fee r = 0.05 is chosen arbitrarily, but it represents a substantial cost to join the private government but also a substantial private supplement. This should aid in pinning down a marginal member. Finally, the tax rate t = 0.28 is chosen because t is defined on the annualized 7 T h i s paper makes no at tempt to endogenize the amount of land devoted to pr ivate government. In reality, of course, developers vary LM depending on the ant ic ipated membersh ip , the level of service prov ided and other factors. 81 implicit rental value of housing services, whereas the property tax rate is defined as a proportion of the market value of housing. The conversion formula between the two is given by t = tp/kp, where tp is the conventional property tax rate and kp is the user cost of housing. 8 Evidence from California suggests that prior to Prop 13, the effective property tax rate was roughly tp = 2.5%. Thus, t = 0.025/0.089 = 0.28 is used. Given these parameter values, the equilibrium values o f the five endogenous variables are y* = 0.868, 7 = 0.050, g = 0.040, pN = 0.701, pM = 0.762. A range of starting variable values were used to ensure that.there were no multiple equilibria or convergence problems 9 W i t h the selected parameter values, about 13 per cent of the population chooses to join the private government. Figure 4.3.5 calculates the realized uti l i ty of non-membership and membership for all income levels. Whi le graphically the difference is quite small, the single-crossing condition is satisfied. Those w i t h y < y* prefer not to be members, so that they receive a substantially lower level of public services, but in return for which they purchase housing at a much lower price. Those y > y* strictly prefer to be members, and they consume almost twice as much' G as non-members. The high level of G is capitalized into higher housing prices, which are about 9 per cent higher than house prices outside the private government. Comparative Statics: An Exogenous Property Tax Decrease The next step in the computational exercise is to decrease the exogenous property tax rate t. Figure 4.2 shows the effect of changing t on the equilibrium values of all the endogenous variables except 7 , which is tr ivial ly related to the private government's fee. Panel (i) of Figure 4.2 provides the key comparative static of the empirical model: as t gets exogenously smaller, the income of the marginal member decreases; in other words, 8 T h e fo rmula is der ived fol lowing Po te rba (1992). Le t R be the annual ized rental value of housing services and P be the market value of the house. T h e n R = kpP. N o w kp = (1 — ty)i + (, where ty is the income tax rate, i is the nomina l interest rate and ( is the sum of the risk p rem ium on housing investment, the maintenance rate, the depreciat ion rate and in f la t ion. A s tR = tpP, th is impl ies t = tp/kp. T o calculate kp, Po te rba uses ty = 0.15, i — 0.1286 and ( = —0.02. 9 A l s o , other sensible parameter values were t r ied and gave s imi la r results. 82 Figure 4.1: Computational Model: Equilibrium Membership Decisions N o n - m e m b e r and member utility (t=0.28) u N 15 a> Utility of membership Utility of non-membership Realized member versus non-member uti l i ty when (a = 0.37,0 = 0.111, t = 0.28, r = 0.05, LM = 0.1, LN = 0.9, LI = 3). The vertical dotted line indicates the income of the marginal member. The horizontal axes have been adjusted to illustrate more clearly the crossing point. 83 Figure 4.2: Computational Model: Changing the Property Tax Rate (i) Effect on y» (iii) Effect on p" 0.5 S0.07 = 0.06 (ii) Effect on g (iv) Effect on p" 0.5 The effect of changing t on the equilibrium values of (i) the marginal member, (ii) public government spending, (iii) non-member housing price and (iv) member housing price. 84 the private government membership gets bigger. Note also the relationship between t and y* is concave: at relatively low levels of t, a tax cut results in a higher increase in private government membership. The intuit ion for this is that t is charged to both members and non-members. A low t hampers the public government's ability to provide services and so makes private government provision, which is invariant to the property tax rate, even more attractive. The other panels of the Figure show the other comparative static results. A decrease in property tax leads to a decrease in public government expenditure, which is natural given the public government's budget constraint is binding. A property tax decrease leads to an increase in housing prices inside the private government. This is presumably due to the increased private government membership, which is tied to buying a house in the l imited land available for the private government. Finally, it is interesting to note that there is a non-monotonic relationship between the non-member housing price and the property tax. For relatively low tax rates, a tax cut decreases pN. This seems to be driven by the decreased non-member population. However, if taxes are relatively high, a tax cut increases pN. It seems that the tax cut increases the demand for housing outside the private government, and this is strong enough to overpower the decrease in the non-member population. 4.4 Background on Proposition 13 The remainder of this paper presents an empirical investigation into the effect of statewide property tax limitations on private government activity. The focus is on California mu-nicipalities in the years surrounding the imposition of Proposit ion 13 in 1978. This section provides a brief overview of the Proposition's motivation and consequences be-fore introducing the econometric model. Proposition 13 is often cited as a prime example of the property tax revolt. The origins of the tax revolt movement were soaring land and property prices in the 1970s, which led to skyrocketing assessed values. Whi le housing prices in California climbed, local governments were slow to lower taxes. Many homeowners saw an increase of twenty to thirty per cent in their property tax bills every year. The unpredictable housing market also meant taxpayers were uncertain about the amount of next year's tax b i l l . 85 In 1978 California voters approved Prop 13 by a two-to-one margin, despite ominous predictions of drastic cuts to public services. O'Sull ivan, Sexton, and Sheffrin (1995) estimate that the immediate effect of Proposit ion 13 was a 57 per cent reduction in prop-erty tax revenue in fiscal year 1978-79. Besides the l imitat ion on revenues, Proposition 13 also changed the way local taxes are collected. Before Prop 13, each local agency (county, city, school district and special district) set its own tax rate; the average total effective property tax rate was 2.5 per cent of market value just prior to 1978. Prop 13 constrained total property taxes not to exceed one per cent of market value. A system therefore had to be devised to share the total tax revenues among local agencies. The severity of Prop 13 was exacerbated because it also l imited to two per cent the yearly increase in assessed value for those properties that did not change hands. This cap did not allow for growth in government revenues to match the pressures of increased demand of government services or the increasing cost of providing these services. Figure 4.3 shows how Proposition 13 affected municipalities statewide by substan-tial ly shifting the property tax rate distribution downward. The figure is a histogram of effective property tax rates in fiscal year 1976-77, two years before Proposition 13 took effect, and in fiscal year 1981-82, two years after. The property tax rate is calculated for 411 cities by dividing the property tax revenue by the market value of property in the c i ty . 1 0 In 1976-77, the average effective property tax rate was 0.47 per cent; in 1981-82, the average had decreased to 0.27 per cent. Compared to pre-Prop 13 years, the distri-bution of tax rates shifted to the left, but it also got tighter. This is because Prop 13 transferred much of the power of property tax from the city to the county, which now had to divide up revenues among various local jurisdictions in an equitable manner. 4.4.1 Local Responses to Prop 13 How did cities in California cope with Prop 13? The following summary corresponds wi th the options outlined in Section 4.2: 1 0 T h i s rate is the property tax rate for the c i ty only, and it does not inc lude proper ty taxes pa id to the county, school d ist r ic t or special d ist r ic t . 86 Figure 4.3: Histogram of Property Tax Rates in 411 California Cities Property Tax Rates (411 cities) 11976-77 11981-82 # l£ j£ T - (NJ f\J Q Ln o i!n Q Co I D r— r-*. oo O in o cn on o fr- # # # tn o w> O o T - T - rvj Tax Rate on Market Value The sample is the set of al l cities in California for which the assessed value of property is reported in fiscal year ( F Y ) 1975-76. Source: Annua l Survey of Governments and the Census of Governments, U .S . Census Bureau 87 Expenditure Cuts Most cities in California did reduce their expenditures, although not as dramatically as critics feared. State subventions cushioned cities' need to cut spending, but decreases were felt in most categories. O'Sull ivan, Sexton, and Sheffrin (1995) note that libraries, parks and contribution to enterprises suffered the largest immediate cuts following Prop 13. Infrastructure was also affected, as ninety per cent of cities reported cutting back on capital improvement programs. O n the other hand, building regulation, public safety and public works experienced increases following Prop 13, reflecting a shift in priorities for city budgets. State Aid The passage of Prop 13 was followed by a block grant from the state's surplus to cushion the decrease in property tax revenues. The grants, worth $2 bil l ion, were designed so that no local government would experience more than a ten per cent loss in total revenue for the 1978-79 fiscal year. Subsequently the state eliminated the bailouts and shifted a portion of property tax revenues from school districts to local agencies. This was done so that cities experiencing high growth and development would have a wider source of funds. In turn, the state increased assistance to school districts. This , coupled wi th a contemporaneous ruling on school district equalizat ion 1 1 , effectively transferred control of school funding from the school district to the state. B y removing the power of local governments to change their property tax rates, Prop 13 has concentrated fiscal power back to the state. Alternative Sources of Revenue The largest increase in government revenues after Proposit ion 13 came from increases in fees and charges. Whi le fees sti l l make up a relatively small portion of local government revenues (only 10 per cent of city revenues came from fees in 1978), most jurisdictions found them to be polit ically and economically effective. Cities increased building and developer fees, real estate transfer fees, licence fees, ut i l i ty and sewerage fees, parks and 1 1 Serrano vs. Priest mandated the equal izat ion of per p u p i l spending across school distr icts . 88 recreation fees and many others. Residential growth has depended largely on high fees for new construction. Schwadron and Richter (1984) cite between 1978 and 1981, the fees collected by local governments increased by almost fifty per cent, and "if any one group has been a prime target for fee boosts, it has'been the real estate developers." Private Providers of Public Services There has been little empirical research on private governments as a response to Prop 13. However, the passing of Prop 13 coincided wi th an explosive growth in the incor-poration of homeowners' associations in the state. Roland (1998) notes that part of this popularity occurred because local governments and developers were able to come to mutually beneficial agreements. Developers agreed to take over the responsibility for infrastructure and some public services. Cities often relaxed zoning regulations to allow the developer to sell smaller lots and denser developments. Chapman (1981) notes that even before Prop 13, "local governments were finding it difficult to finance capital infrastructure...Proposition 13's passage accelerated the trend [of shifting improvements to the developer] since it virtually eliminated debt financing." This is the response that the empirical examination seeks to identify. The next section builds the empirical model to test whether Prop 13 is an appropriate channel to explain the extraordinary growth in private governments in California. 4.5 Empirical Model The empirical model takes the form of an event study and tests the proposition that an exogenous property tax decrease raises homeowners' association membership in a city. In the first part of the empirical examination, Prop 13 is assumed to have a similar effect on al l cities in the sample. In the second part of the empirical section, the model allows some differentiation between cities and tests whether or not cities that are more constrained by Prop 13, as defined by ini t ial conditions, experience increased growth in private government. 89 4.5.1 Prop 13 in an Event Study Model In the simplest formulation, the test for the effect of Prop 13 on homeowners' association popularity can be expressed as follows: yit = pPropl3it + 5xit + di + dt + eit. (4.5.1) where yu is the homeowners' association membership rate in city i in year t, and Propl3u is an indicator variable reflecting whether city i was subject to Prop 13 in year t. Other time-varying covariates, xu, control for characteristics likely to have changed over 1976-1982, the years of the study, and which are likely to affect private government mem-bership. Particularly of interest, they include the levels of intergovernmental aid and charges to take into account the possibility that cities substitute for decreased tax rev-enues wi th these alternate sources of income. The di control for unobserved city fixed effects. Finally, the dt control for year effects that are the same for all cities in the sample; to allow for flexibility in modelling the time trend, specifications are run with a set of year dummies, a linear year trend and a quadratic year trend. Equation 4.5.1 is estimated by O L S . A n alternative view of the empirical model is the role of Proposition 13 in the growth rate of H O A . A n alternative dependent variable is the change in H O A membership rate in the city. Focusing on the growth in H O A s captures the way in which Prop 13 may have prompted local governments to encourage the growth of other public service providers; thus the limitation's effect would be to increase the growth rate of private governments. However, it may be more sensible to think of each homeowners' association as an individual agent in the interaction between public and private governments. Because all common interest developments have only a single homeowners' association, regardless of the number of housing units, an appropriate measure of the H O A variable should be the number of new incorporations in a city. Each developer's decision whether to set up an H O A can be treated as a separate decision. Because this variable can only take on non-negative values, and because there are likely many observations for which the number of new incorporations in a city is zero, O L S regression would not make efficient use of the data, whereas count data models would. Count data models are well-known for their ability to accommodate a significant number of zero observations. 90 In a count data context, the dependent variable, yu, is the number of new H O A incorporations in city i in year t. yit is a non-negative integer, and its conditional mean depends on some vector or regressors xit. The key regressor is a measure of the binding effect of Proposition 13. M u c h of the econometric analysis comes from Cameron and Trivedi (1998). The most common way to deal wi th count data is to assume that yi is independently Poisson distributed, given the vector of regressors Xi, wi th density 'f(yi\xi) = ^—^1 i / i = 0 , 1 , 2 , . . . . (4.5.2) wi th mean parameter = e x p ( x ^ ) . (4.5.3) The major obstacle to using the Poisson distribution, however, is that it assumes that the conditional mean equals the conditional variance: = = exp(x-/3). The parameter vector ft is estimated by maximizing the log-likelihood function. Va l id statistical inference requires equidispersion, that is, the conditional mean equals the con-ditional variance. When there is overdispersion (when the conditional variance is greater than the conditional mean), M L estimates can give seriously underestimated standard errors, leading to spurious significance of the parameters. Nevertheless, statistical infer-ence is possible by appropriately adjusting the standard error estimates. This leads to the Poisson pseudo-ML estimator ( P M L E ) , which is based on the correct specification of the conditional mean, while allowing for overdispersion. A n alternative way of handling the overdispersion problem is to specify a distribution that permits more flexible modelling of the variance than the Poisson. The negative b i -nomial distribution assumes that the data are Poisson, but there is a gamma-distributed unobserved individual heterogeneity. Al low the conditional variance to be a function of the mean: V^Xi] = & + atf. (4.5.4) In this case, a is called the dispersion parameter and is estimated alongside (3. This is a parameterization of the negative binomial distribution, which Cameron and Trivedi 91 (1986) call the N B 2 model. The density of the negative binomial distribution is Note that the negative binomial density reduces to the Poisson density if a = 0. The conditional mean is st i l l fii = exp(x'if3). Maximiz ing the log-likelihood function gives the N B 2 M L E , 0NB2,C\NB2)- The N B 2 M L E has several attractive properties. It is robust to distributional misspecification. As long as the conditional mean is correctly satisfied, There is some history of count data models in empirical urban economics. M c M i l l e n and Smith (2003) examine the determinants of the number of employment subcentres in a cross-section of 62 large American urban .areas. Using a Poisson regression, they find that metropolitan population and commuting costs account for nearly 80 per cent of the variation in the number of subcentres. However, M c M i l l e n and Smith do not address overdispersion of the count data. Papke (1991) examines the effect of state and local tax differentials on the location of industry. Using a panel of manufacturing firms, she models the number of firm births in a location as a Poisson count process. Finally, Congdon (1993) compares the Poisson and multinomial models of migration flows. He compares various ways of dealing with overdispersed data. He also addresses count data models when the errors are spatially correlated. 4.5.2 Public Government Data The data for the empirical analysis come from a combination of sources. The dependent variable is a measure of private government, which requires the number of homeowners' associations in California and their membership. The time period for the analysis is fiscal year 1975-76 through fiscal year 1981-82. These are referred to simply as 1976 through 1982. Proposition 13 took effect beginning in fiscal year 1978-79. Thus, the pre-Proposition 13 years in this analysis are 1976, 1977 and 1978, and the post-Proposition 13 years are 1979, 1980, 1981 and 1982. The data on local government revenues come from the U.S . Census Bureau's Annua l Survey of Governments and Census of Governments. Cities, counties, school districts - i (4.5.5) a > 0 , .y = 0 , l , 2 , . . . . the estimated coefficient vector j3 is consistent. 92 Table 4.1: Number of Cities in the Survey of Governments Year Number of cities responding Average population 1976 208 70,689 1977 407 41,004 1978 209 72,659 1979 402 43,051 1980 230 70,603 1981 391 45,936 1982 419 44,686 The horizontal line separates pre-Proposition 13 years from post-Proposition 13 years. Source: Annua l Survey of Governments and the Census of Governments, U.S . Census Bureau and special districts are surveyed on their revenues, expenditures and debt over the previous fiscal year. The cross-sectional unit of analysis is the municipality; however, not all municipalities are surveyed in each fiscal year. The two exceptions are 1977 and 1982. These are Census of Governments years, and every municipality responds to the survey. The analysis includes the two Census years in order to capture as much cross-sectional variation as possible. Table 4.1 shows the number of cities surveyed in each year along wi th the average population of those cities. When only a portion of the state's cities are surveyed, the survey favours larger municipalities. However, the sampling methodology guarantees that smaller cities are also represented. The analysis uses an unbalanced panel of cities. The sample includes only cities that respond to the Survey of Governments for at least six of the seven years of the analysis. Out of 419 cities in the state as of 1982, 199 of them fit this criterion. This seems to be an adequate compromise between keeping as many cities in the sample and not having too many missing years of data. This also guarantees that every city in the sample reports at least two years before Prop 13 and two years after. 1 2 1 2 T o check that there is not a serious sample selection prob lem, the analysis is also performed w i t h a balanced panel , in wh ich on ly cit ies that report to the Survey for a l l seven years are inc luded. Es t ima t i on results are qual i ta t ive ly s imi lar and so are not reported. 93 4.5.3 Private Government Data The dependent variables for the three empirical specifications measure the degree of private government activity in a city. These require the number of homeowners' associa-tions in California and their membership. However, because of their private nature, one obstacle to empirical research into homeowners' associations is the lack of comprehensive data. This paper turns to a novel source of information about homeowners' associations to develop the measures of private government. A database of homeowners' associations in California was obtained from the accounting firm of Levy and Company in Oakland. It is a unique record of condominiums, cooperatives, planned developments and other homeowners' associations, and it is one. of the most complete sources of information about residential private governments available. The database lists each of the 37,655 incorporated homeowners' associations in the state as of M a y 2003. Because the incor-poration date of each association is known, it is possible to back out the associations that existed in each year, so a panel can be constructed. The membership of each association is aggregated to the city level using an imputation method, as the exact membership of each association is unknown. The result is a time series of the homeowners' association membership rates for each city in the sample. 1 3 For one city in the sample, Oakdale, the imputation method yields a homeowners' association rate of greater than 1 for some years; this city is therefore dropped from the sample, reducing the number of cities in the sample to 198. For the count data model, no imputation is necessary as the dependent variable is simply the number of new incorporations. Each association is counted once regardless of size (associations can range from a handful of housing units to thousands of housing units) or type (planned development, condominium or cooperative). Thus an additional appeal to the count data model is that there is no measurement error owing to the imputation of the membership rates. Recall Figure 3.3 at the beginning of the dissertation, which shows the number of associations by year of incorporation and by type. The figure clearly demonstrates that homeowners' associations were virtually unknown up to 1970. There is a gradual increase in the number during the 1970s, but in 1978, the number of new incorporations 13For further information on the data and the imputation method used to calculate the membership rates, consult Chapter 3. 94 Table 4.2: Summary Statistics for Private Government Variables Summary Statistics HOA membership rate: Mean: .079 M i n . : 0 Std. dev.: .097 Max. : .677 % change in membership rate: Mean: .006 M i n . : -.063 Std. dev.: .019 Max.': .299 Number of new incorporations: Mean: 4.8 M i n . : 0 Std. dev.: 19.1 Max. : 372 Frequency Table of New Incorporations 0 incorporations: 465 (33.6% of observ.) 1: 232 ' (16.7%) 2: 169 (12.2%) 3: 119 (8.6%) 4: 87 (6.3%) 5-10: 199 (14.4%) 11-20: 76 (5.5%) 21-372: 39 (2.8%) The sample consists of 198 cities over the 1976-1982 period (1,386 observations). Source: Levy and Company, 2004 soared and remained at a very high level for about twelve years. There has been a gradual decline in recent years and a return to new incorporation levels resembling those of the mid-1970s. The data also show a distinction between the various types of common interest development and their relative popularity. Planned developments, which generally consist of single-family houses, were the predominant form of C I D in California unti l the early 1970s. Condominiums began to rise in popularity and took over as the most common form of homeowners' association. Roland (1998) interprets this as a deliberate decision by the building industry to "market housing wi th increased densities." This was combined with the support of federal and local policies, who made investment in condominiums more attractive by policies such as mortgage insurance. Table 4.2 presents some summary statistics for three private government dependent variables. In the thi rd panel, the frequency table of the number of new incorporations in a given year suggests the need to control for the preponderance of zero observations and the integer nature of the data by using a count data model. 95 4.5.4 Other Covariates The other covariates on the right hand side of the estimating equation are variables that are posited to affect the degree of private government, and which are likely to have changed over the study period. In the previous chapter's empirical examination of private government these included demographic and economic characteristics of the community, such as ethnicity, income and education. However the study period there was a thirty-year period as opposed to a seven-year period here. Therefore, it is unlikely that many of these variables wi l l have substantially changed, and, in practice, many of them are measured at only one or two points in the study period. Thus the effect of these variables wi l l be subsumed into the city fixed effect dt, and the estimation wi l l not explicitly control for them. O n the other hand, there are variables that are likely to have changed substan-tially over the sample period and are observed annually, and these are included in some specifications of the fixed-effects regressions. The most notable are the levels of intergov-ernmental revenue and charges revenue. A s mentioned earlier, some cities substituted for the loss in property tax revenue by heavy increases in fees and intergovernmental aid. In additional, three other time-varying covariates are used to proxy for residen-t ial growth: population, five-year population growth and land area. To the extent that membership rates in homeowners' associations is correlated wi th residential growth, it wi l l be important to control for their effect in order to argue that the variation in private government activity that remains is attributable to Prop 13. 4.5.5 Analysis Table 4.3 shows the fixed-effects estimation results when the dependent variable is the imputed H O A membership rate. The coefficient of interest, the effect of Prop 13, is measured by a dummy variable that takes the value 1 for all observations from 1979 through 1982, and 0 for al l observations from 1976 through 1978. Because of the inclusion of city-specific, time-invariant effects, these specifications identify the effect of Prop 13 using only the within-city levels of H O A membership rate. Specifications (1), (2) and 96 (3) control for year effects through a set of year dummy variables. 1 4 Specification (1) is the simplest specification that does not control for alternate sources of government revenue. The positive coefficient on the Prop 13 dummy sug-gests that the constant in the predicted H O A membership rate was shifted up by .039. To get a sense of the magnitude of the estimate, the mean of the dependent variable is about .08 over the sample period. Hence, Proposition 13 accounts for a substantial shift in the intercept. Specification (2) controls for per capita charges revenue and intergov-ernmental revenue to allow the possibility that property tax l imitat ion causes cities to raise alternative types of revenue. The inclusion of these variables has little effect on the Prop 13 coefficient, although it is interesting that an increase in per capita charges revenue is positively related to an increase in H O A membership. This is consistent wi th the increased use of developer impact charges following Prop 13. However, once Spec-ification (3) controls for additional time-varying covariates (population, land area and population growth) the charges revenue becomes insignificant again. This is perhaps due to the positive correlation between charges revenue and residential growth, which is proxied by these additional covariates. St i l l , the Prop 13 effect remains significantly positive in Specification (3). Specifications (4) through (6) report results from various other specifications of the year trend. Specification (4) uses a linear year trend instead of year dummies, while (5) adds in a quadratic term. In (6), an interaction term, equal to the year multiplied by whether Prop 13 was in effect, allows not only Proposition to shift the intercept in the predicted H O A membership rate, but also the slope. However, in these three specifications the effect of the Prop 13 dummy is insignificant. Table 4.4 shows the fixed-effects estimation results wi th the dependent variable being the percentage point change in the H O A membership rate. Columns (1) through (3) use year dummies to control for the year effects; the Proposition 13 dummy is insignificant. This time, however, using a linear or a quadratic year trend uncovers a positive Prop 13 effect. This suggests that not only did Prop 13 have a positive effect in H O A mem-bership rates over the sample period, it also had a positive effect on the growth rate of membership. 1 4 N o t e that two years have to be dropped to avoid the d u m m y var iable t rap: These are the years 1976 and 1982. 97 Table 4.3: Fixed-Effects Estimation Results for HOA Membership Rate Sample years: 1976 through 1982 Dependent variable = H O A membership rate (1) (2) (3) (4) (5) (6) Prop 13 .039*** .038*** .037*** .002 .003 -.163 (.003) (.003) (.003) (.003) (.003) (.120) Year : o o 6 * * * -.028 .005*** (.001) (.034) (.001) Year 2 .0002 (.0002) Prop 13 x year .002 (.002) Charges revenue .084*** .049* .083*** .083*** .083*** (.016) (.027) (.016) (.016) (.016) IG revenue .016 .023 .009 .012 .014 (.017) (.019) (.017) (.017) (.017) Year dummies Yes Yes Yes Other controls Yes R2 wi thin .255 .275 .274 .253 .273 .274 There are 198 cities in the sample. Mean of dependent variable is .079. A l l specifications control for city fixed effects. "Other controls" refer to population, 5-year population growth and land area. *, ** and *** denote significance at the 10, 5 and 1 per cent level, respectively. To explore the growth of H O A s further, Table 4.5 and 4.6 present count data mod-els where the dependent variable is the number of new incorporations of homeowners' associations of any type. Again , it may be reasonable to think of the number of in-corporations as a relevant dependent variable if Prop 13 encouraged city policies that made H O A formation a more feasible option for a developer. Regardless of the number of housing units, a developer would set up only one homeowners' association. Both the Poisson and the negative binomial regressions indicate there is a significant, positive Prop 13 effect on new H O A incorporations. This suggests that Prop 13 had a hand in raising membership rates in homeowners' associations, but also in the formation and growth of them. 98 Table 4.4: Fixed-Effects Estimation Results for H O A Membership Change Sample years: 1976 through 1982 Dependent variable = Change in H O A membership rate (1) (2) (3) (4) (5) (6) Prop 13 .003 .003 .002 .003* .002 .199** (.002) (.002) (.002) (.002) (.002) (.084) Year -.0001 (.0005) .062*** (.024) .002* (.001) Year 2 -.0004*** (.0001) Prop 13 x year -.003** (.001) Charges revenue .010 .0002 .009 .009 .009 (.011) (.018) (.011) (.011) (.011) IG revenue .005 .006 .008 .003 .003 (.011) (.013) (.012) (.012) (.012) Year dummies Yes Yes Yes Other controls Yes R2 wi thin .014 .015 .019 .009 .015 .013 There are 198 cities in the sample. Mean of dependent variable is .006. A l l specifications control for city fixed effects. "Other controls" refer to population, 5-year population growth and land area. *, ** and *** denote significance at the 10, 5 and 1 per cent level, respectively. 99 Table 4.5: Poisson Estimation Results for New HOA Incorporations Sample years: 1976 through 1982 Dependent variable = New H O A incorporations (1) (2) (3) (4) (5) (6) Prop 13 1.009*** .946*** .906*** .980*** .245*** 54.148*** (.068) (.074) (.079) (.052) (.058) (3.119) Year .001 (.014) 18.395*** (.880) .540*** (.035) Year 2 116*** (.006) Prop 13 x year -.684*** (.040) Charges revenue 1.210* 1.262* -.052 .993 .427 (.652) (.687) (.611) (.642) (.631) I G revenue -.574 -.515 1.417*** -2.289*** -2.612*** (.601) (.620) (.471) (.565) (.583) Year dummies Yes Yes Yes Other controls Yes There are 187 cities in the sample (some cities were dropped because of al l zero outcomes). Mean of dependent variable is 4.8. A l l specifications control for city fixed effects. "Other controls" refer to population, 5-year population growth and land area. *, ** and *** denote significance at the 10, 5 and 1 per cent level, respectively. 100 Table 4.6: Negative Binomial Estimation Results for New HOA Incorporations Sample years: 1976 through 1982 Dependent variable = New H O A incorporations (1) (2) (3) (4) (5) (6) Prop 13 .926*** .919*** 1.197*** .834*** .251*** 40.099*** (.096) (.104) (.093) (.088) (.083) (4.486) Year .015 (.022) 14.088*** (1.146) 429*** (.052) Year 2 -.088*** (.007) Prop 13 x year -.506*** (.058) Charges revenue -.513 2.400*** -1.705* -.136 -1.349 (.882) (.751) (.903) (.840) (.906) I G revenue -.647 2.093 .610 -1.324* -1.040 (.721) (.593) (.628) (.788) (.743) Year dummies Yes Yes Yes Other controls Yes There are 187 cities in the sample (some cities were dropped because of al l zero outcomes). Mean of dependent variable is 4.8. A l l specifications control for city fixed effects. "Other controls" refer to population, 5-year population growth and land area. *, ** and *** denote significance at the 10, 5 and 1 per cent level, respectively. 101 4.5.6 Heterogeneity in the Prop 13 Treatment Variable The previous section generally finds a significant post-Proposition 13 effect on the levels and growth of homeowners' association membership as well as the number of new H O A incorporations in California. However, because Proposit ion 13 affected al l municipalities in the state, the previous estimation assumes that Prop 13 has a similar effect on all cities equally. However, as the histogram of effective property tax rates shows, there is a large range in property tax rates across cities. Presumably, some cities experienced relatively large drops in property tax revenue following Prop 13, while others experienced small drops. Thus, a natural question would be to see if cities that were relatively "more constrained" by Prop 13 experienced relatively larger increases in private government levels and incorporations. 1 5 To the extent that Prop 13 may have affected various municipalities differently, it is useful to devise a measure of the constraint that Prop 13 places on a.city. In other words, Prop 13's bite may have differed across municipalities. This would allow cross-city comparison of the effects of Prop 13 on private government formation. In addition, the inclusion of a heterogeneous Prop 13 treatment variable should help allay the problem of not having a counter-factual state or municipalities. Figl io (1998) includes school districts in Washington, a state that did not experience Oregon's l imitat ion measure. In this study, all the cities in the sample experienced Prop 13, so there is no way to know what the incorporation of new H O A s would have looked like in the absence of Prop 13. However, the heterogeneous treatment variables provide a way to compare cities where the impact of Prop 13 was strong to cities where the impact was minimal. This paper proposes three measures of the degree of constraint that Prop 13 placed on a city: ' • Property tax rate: if the pre-Prop 13 property tax rate in the city is high, then it is likely that Prop 13 has a strong constraining effect on the city's revenues. 1 5 Idea l l y , th is could lead to a test of the second comparat ive stat ic result der ived f rom the com-puta t iona l mode l : If the in i t ia l t ax rate in a c i ty is low, an exogenous proper ty tax decrease raises membership more t han if the in i t ia l t ax rate were h igh, al l other th ings being equal . However, this assumes that the exogenous decrease is the same size for bo th h igh- in i t ia l t ax and low- in i t ia l - tax cit ies. P r o p 13 may not an appropr ia te tax decrease as it affected cit ies differently. 102 • Property tax dependence: if, before Prop 13, a city was highly dependent on property tax revenue and lacks alternative sources of revenue, it may turn to private government development to finance municipal infrastructure when faced with property tax cuts. • Crime: a city wi th a high pre-Prop 13 crime rate may meet wi th less resistance from allowing planned developments and gated communities to form. These mechanisms suggest that cities can be differentiated by interacting the Prop 13 dummy wi th these heterogeneous variables. Figl io (1998) suggests that the interaction of the policy dummy should be wi th a pre-limit variable only; this removes the possibility of property tax dependence or crime rate to be endogenous and to depend on the number of H O A incorporations. The equation estimated is as follows: yit = pPropl3it + CjProplSuZij + Sxit + dt + dt + eit. (4.5.6) where yit is the dependent variable, which, as before, wi l l be the H O A membership rate in levels, the one-year percentage point change in H O A membership rate and the number of new H O A incorporations. Propl3u indicates whether Prop 13 is in place, Ztj is an interaction variable representing a pre-Prop 13 condition and Xu are time-varying covariates that affect private government activity. A s in the previous section, the city fixed effect is di and the year effects are dt, which may take the form of a set of year dummies, a linear time trend or a quadratic time trend. The three pre-limit conditions are measured at the year 1976 and come from a variety of sources. The property tax rate is the property tax revenue divided by the assessed value of property. Property tax dependence is defined as the city's property tax revenue divided by total revenue. Tax revenue figures from the Annua l Survey of Governments while the assessed values come from the State Board of Equalization. Crime rates are calculated by taking the FBI ' s crime index (which aggregates violent and property crimes) and normalizing by population. A s the goal is to identify differential effects of Prop 13 from cities wi th different pre-limit conditions, it is important to verify that the private government trends are similar 103 across all municipalities prior to Prop 13. One way to do this is to group cities into groups wi th similar pre-limit conditions and compare their private government variables. If the expectation is that Prop 13 affected some cities more than others, then the various groups of cities should have similar private government trends before Prop 13 but should diverge afterward. The results of this verification exercise are summarized in Figures 4.4, 4.5 and 4.6. The methodology is as follows. Cities are ranked according to the pre-limit condition of interest and grouped into quintiles. The figures plot the mean of the dependent variable of interest for cities in the first and in the fifth quintile. The left panel of Figure 4.4 shows that cities with different property tax rates before Prop 13 react differently in the trend in the private government membership rate after Prop 13. Prior to 1979, low property tax cities (1st quintile) and high property tax cities (5th quintile) exhibited similar trends in homeowners' association membership. 1 6 After 1979, high-tax cities seem to show slightly faster growth in the H O A membership rate. The H O A membership rate of low-tax cities seem to be growing at a decreasing rate. This is consistent wi th the theory that cities wi th high property tax rates before Prop 13 are more likely to feel the proposition's bite. Thus, high tax rate cities may be more likely to turn to private government. However, this difference between the first and fifth quintiles seems small and may indicate that property tax rate may not be the appropriate pre-limit condition to group cities by. The right panel of Figure 4.4 plots the mean number of new H O A incorporations for the first and fifth quintile of the 1976 property tax rate. The difference in growth between the first and fifth quintiles is evident if the number of new incorporations is the measure of private government activity. Cities wi th high tax rates have higher growth in H O A incorporations, consistent wi th the view that these cities experienced a larger bite from Prop 13. However, an important caveat is that the divergence seems to occur a year before Prop 13 took effect. This could indicate that something besides Prop 13 was driving high property-tax rate cities to turn to private government. O n the other hand, as the stipulations of Prop 13 were well-known a year before the referendum, developers and municipal officials in high tax cities may have anticipated the loss in revenue and 1 6 K e e p in m ind that it is the t rend that is of interest, as ident i f icat ion of the P r o p 13 effect comes about f rom changes in pr ivate government w i t h i n a city. 104 Figure 4.4: Q u i n t i l e P l o t s b y P r e - L i m i t P r o p e r t y T a x R a t e H e t e r o g e n e i t y Left: Mean H O A membership rate by quintiles of property tax rate in 1976 Right: Mean new H O A incorporations by quintiles of property tax rate in 1976 turned to private governments even before the proposition was passed. In Table 4.7, the O L S estimation results are presented when the dependent vari-able is either the H O A membership rate or the percentage point change in the H O A membership rate. Table 4.8 reports the results wi th the Poisson and negative binomial count data models, using the number of new incorporations as the dependent variable. The only reported coefficient is that on the Prop 13 dummy, corresponding to j3 in Equation (4.5.6), and on the Prop 13 interaction term, corresponding to £. The interaction terms give mixed evidence to the importance of pre-limit conditions on the degree of homeowners' association membership. The coefficients on the pre-limit conditions are significantly different from zero for only certain specifications. In the specifications in which both the uninteracted Prop 13 coefficient and the interaction term are different from zero, a mean effect of Prop 13 can be calculated. In general the mean effect is positive, which supports the effect that Prop 13 has on H O A membership and H O A membership growth. The strongest evidence for pre-limit heterogeneity is the use of crime rates. Cities wi th high pre-Prop 13 crime rates exhibit higher rates of homeowners' association mem-bership and also more new H O A incorporations. This suggests that pre-limit crime rate 105 Figure 4.5: Q u i n t i l e P l o t s b y P r e - L i m i t P r o p e r t y T a x D e p e n d e n c e H e t e r o -gene i ty Left: Mean H O A membership rate by quintiles of property tax dependence in 1976 Right: Mean new H O A incorporations by quintiles of property tax dependence in 1976 Figure 4.6: Q u i n t i l e P l o t s b y P r e - L i m i t C r i m e R a t e H e t e r o g e n e i t y Left: Mean H O A membership rate by quintiles of crime rate in 1976 Right: Mean new H O A incorporations by quintiles of crime rate in 1976 106 may be acting as a proxy for local demand for public services. Residents of high crime cities may have recognized that the passing of Prop 13 may entail drastic cuts in policing and public safety 1 7, and this may have increased their demand for private provision of public safety. In other specifications the sign of the interaction term is mixed. For example, the upper left corner of Table 4.7 shows when the pre-limit condition is the 1976 property tax, the estimated coefficients suggest that in cities with a high effective property tax rate in 1976, the effect of Prop 13 is lower, although the effect for the city wi th the average tax rate is st i l l positive. W h y would cities that have higher property tax rates, and thus for whom Prop 13 would presumably pose the greatest constraints, experience lower homeowners' association memberships? One possibility is that cities with high property tax rates are slower-growing central cities, and that instead of responding to Prop 13 by encouraging the growth of private governments, they are forced to cut government expenditures. O n the other hand, Table 4.8 shows that cities wi th high pre-13 tax rates lead to more incorporations of H O A s . This would support the fact that these high-tax cities are slower-growing central cities if most of these incorporations took the form of condominiums or condo conversions instead of planned developments. Condominiums tend to have fewer housing units than planned developments, and so it would be possible for membership rates to rise slowly (or even fall relatively to increased residential growth in non-condominium housing) while the number of incorporations grew much more rapidly. Using the property tax as a measure of city constraint gives a mixed result; while this is unexpected, it does suggest that cities in California responded differently to Prop 13, and not all cities may have taken the route of private government development. Finally, the use of pre-limit heterogeneity seems to have had no effect on the per-centage change in H O A membership rate. Surprisingly, the coefficient on property tax dependence is negative, which suggests that cities that were highly dependent on prop-erty tax revenues had slower rates of H O A membership and formation after Proposition 13 was passed. This may indicate that property tax dependence is not an effective way to measure the revenue constraint that Prop 13 had on a city and that cities were better able at adapting to the loss in revenue that otherwise feared. 1 7 A n d indeed dur ing the campa ign per iod leading up to the vote, opponents of P r o p 13 used the threat of pol ice cuts to t r y to convince voters to vote no. 107 Table 4.7: Fixed-Effects Estimation Results with Pre-Limit Interactions. De-pendent Variables Are HOA Membership Rate and Change in HOA Rate Sample years: 1976 t h r o u g h 1982 Speci f icat ion: Dep . var. = H O A membersh ip rate (1) (2) (3) Dep. var. = Membersh ip rate change (4) (5) (6) Interacted with 1976 pro P r o p 13 P r o p 13 x 1976 prop, tax rate perty tax rate 057*** Q^7*#* 018*** (.006) (.006) (.006) -3.424*** -3.430*** -3.428*** (1.008) (1.007) (1.007) .002 .005 .003 (.003) (.003) (.004) .178 .179 .176 (.594) (.595) (.593) Interacted with 1976 pro P r o p 13 P r o p 13 x 1976 prop, tax dependence perty tax dependence .055*** .015** .016** (.006) (.006) (.006) -.063*** -.063** -.063** (.025) (.025) (.025) .003 .007 .004 (.004) (.004) (.004) -.005 -.006 -.005 (.014) (.015) (.014) Interacted with 1976 cri P r o p 13 P r o p 13 x 1976 cr ime rate me rate 019*** -019*** -017*** (.005) (.005) (.005) 281*** 279*** 281*** (.057) (.057) (.057) -.001 .001 -.0004 (.003) (.003) (.003) .045 .048 .045 (.032) (.032) (.032) M e t h o d of year contro l Dummies L inear Quadra t i c t rend t rend Dummies L inear Quadra t i c t rend t rend *, ** and-*** denote significance at the 10, 5 and 1 per cent level, respectively. A l l specifications control for city fixed effects and intergovernmental and charges revenue. For the property tax and property tax dependence specifications, there are 199 cities in the sample. For the crime rate specification, there are 168 cities. 108 Table 4.8: Fixed-Effects Estimation Results with Pre-Limit Interactions. De-pendent Variables Are New HOA Incorporations Dep. var. = New H O A Incorporations Sample years: 1976 through 1982 Speci f icat ion: Po isson (1) (2) . (3) Negat ive b inomia l (4) . (5) (6) Interacted with 1976 pro P r o p 13 P rop .13 x 1976 prop, tax rate perty tax rate . 7 5 0 * * * . 7 7 1 * * * .017 (.091) (.077) (.080) 40 .922* * * 41 .439* * * 44 .889* * * (11.189) (11.288) (11.024) . 7 8 8 * * * . 6 8 5 * * * .082 (.130) (.128) (.127) 32.929* 33.561 33.437 (19.551) (21.156) (20.923) Interacted with 1976 pro P r o p 13 P r o p 13 x 1976 prop, tax dependence perty tax dependence 1.242*** 1.288*** . 5 3 6 * * * (.114) (.102) (.105) - 1 . 6 5 1 * * * - 1 . 6 6 3 * * * - 1 . 5 9 3 * * * (.476) (.477) (.476) 1.076*** . 9 4 1 * * * .368** (.151) (.151) (.145) -.834 -.544 -.734 (.584) (.624) (.591) Interacted with 1976 cri P r o p 13 P r o p 13 x 1976 cr ime rate me rate . 5 7 0 * * * . 574* * * -.154 (.145) (.137) (.139) 4 .814* * * 5 .209** * 4 . 9 0 2 * * * (1.570) (1.570) (1.570) . 5 9 7 * * * . 4 4 8 * * * -.118 (.188) (.197) (.187) 4 . 3 9 3 * * * 5 .226** 4 .372** (2.042) (2:216) (2.060) M e t h o d of year contro l Dummies L inear Quadra t i c t rend t rend Dummies L inear Quadra t i c t rend t rend *, ** and *** denote significance at the 10, 5 and 1 per cent level, respectively. A l l specifications control for city fixed effects and intergovernmental and charges revenue. For the property tax and property tax dependence specifications, there are 187 cities in the sample. For the crime rate specification, there are 166 cities. 109 4.6 C o n c l u s i o n This chapter proposed an avenue by which local governments can' respond to state-imposed tax and expenditure limitations. In addition to increasing fees and intergov-ernmental revenues, a city can download the responsibility of providing public services to private governments, in the form of homeowners' associations. The paper presented a simple theoretical model to study the effect of an exogenous tax rate l imitat ion on the behaviour of private governments. The model is based on residents who decide whether to join a homeowners' association when choosing a place to live. The comparative static of interest in how the membership margin changes wi th the imposition of a property tax limitation. The theoretical model leads to an empirical examination of the pattern of California homeowners' association growth in the years surrounding Proposition 13. The empirical analysis combines novel data on homeowners' associations in California wi th property tax data to see whether cities that faced stronger property tax constraints are more likely to encourage private government growth. In most specifications, Proposit ion 13 had a positive and significant effect of the level and the growth of homeowners' association membership, as well as on the number of new incorporations. Interaction terms try to estimate the effect of pre-Prop 13 conditions on the size of the effect of Prop 13. Cities wi th a high pre-Prop 13 crime rate, proxying a demand for public policing, are more likely to have higher rates and growth of homeowners' associations. This suggests that Prop 13 affected cities in California in different ways. Of course, the ideal way to identify the effect of Prop 13 would be to find a group of cities in another state to serve as a control group. Bu t barring better data on homeown-ers' associations in other states, this is not immediately feasible. Therefore, a further direction research may take is to develop a better measure of heterogeneity among cities and a better way to measure the degree of constraint that a city might have faced. 110 C H A P T E R 5 Private Government and the Urban Housing Market 5.1 Introduction This chapter extends the theoretical model of private government introduced in Chap-ter 4 and examines how, in the presence of a housing market and distortionary property taxes, a public government interacts wi th a private government. The motivation of this chapter is that the incorporation of housing markets adds to the understanding of why residents choose homes governed by private governments. The chapter shows that the presence of the property tax distorts two choices: the consumers' housing consump-tion decisions and the optimal provision of public services. A modified Samuelson rule characterizes equilibrium provision of public goods with elasticities playing a key role. The chapter is organized as follows. Section 5.2 reviews previous research on private government as it relates to the urban housing market. This is done wi th the goal of integrating two strands of the local public finance literature into one view of private government. Section 5.3 outlines the model, which largely works off Helsley and Strange (1998), but which also incorporates elements of multicommunity general equilibrium models of local public good provision, after Tiebout (1956). After highlighting the.basic model and its implications, Section 5.4 gives an example of equilibrium determination by performing some numerical simulations on a solved functional form. Section 5.5 concludes. I l l 5.2 Private Governments and Housing Markets The first step is to situate the housing market wi thin the private government literature and to motivate its integration. W h y does incorporating a housing market add to the understanding of the interaction between private and public governments? One reason is that it aids in integrating the newer private government models wi th the Tiebout models of local public good provision. A s authors such as Rose-Ackerman (1979), Epple, Fi l imon, and Romer (1984) and Hansen and Kessler (2001) have noted, the addition of housing complicates these models, but they act as a stabilizing force that promotes the existence of equilibrium. Another reason for looking in the housing market is to understand another motivation consumers have for joining a private government. Popular opinion, especially among members of homeowners' associations, is that private governments protect and increase property values. Reichman (1976) states that "The sole purpose of the general scheme is to preserve the community as planned and so to secure property values and desired living conditions." Indeed, a 1999 survey of homeowners' association members echoes this sentiment ( C A I Research Foundation 1999). It reports that 85 per cent of members believe their property values are rising, and also that a primary reason for purchasing a home in a homeowners' association is that it represents a good investment. Thus, in a theoretical model it would be reasonable to expect that the value of the extra services provided by the private government be reflected in housing prices. However, in the presence of distortionary property taxes, this capitalization may not be perfect. 5.3 Model This section outlines the private government model wi th housing markets and endoge-nous choice of public service levels. After the elements of the model are introduced, equilibrium conditions are characterized. 112 5.3.1 Players The model consists.of one urban area, whose land area is fixed. There are three main players: consumers who dwell in the city, the public government and the private gov-ernment. The following examines each in turn. Consumers The population of the city is fixed and normalized to 1. Each consumer has the iden-tical util i ty function U(x,h,G), where h represents housing services, G is the level of government services in the form of a local public good and £ is a numeraire composite good. The value of h can vary continuously and can be thought of as the quantity or quality of housing services. U(x,h,G) is increasing, strictly quasiconcave and twice differentiable in all its elements. To address the issue of housing rents, make the ini t ia l usual assumption that all consumers in the city are renters, and their rents accrue to absentee landlords residing outside the city. The consumers differ in their exogenous income yl. This exogeneity and the absentee landlord assumption imply that income does not depend on housing market conditions. The distribution of incomes is denoted f(y) wi th support [y,y], so that /(y^dy1 = 1. Each consumer can be referred to by income, yl. In this model, consumers must do two things: first they choose whether or not to join the private government, and then they choose uti l i ty-maximizing levels of x and h. Regardless of whether a consumer chooses to join or not to join the private government, each wi l l nevertheless receive some level of G from the public government. A consumer can only consume a higher level of G by joining the private government. Two assumptions about U(x, h, G) are needed for the analysis, and it is useful to mention them now. Assumption 1 U(x, h, G) is homothetic. Assumption 2 M(f,G,y) = ^ \ v = ^ is increasing in y. Assumption 1 is made for convenience in deriving the equilibrium conditions of the model. Whi le it is not innocuous, it has been commonly assumed in other models of 113 local public good provision (see, for example, Hansen and Kessler (2001)) Assumption 2 is the important single-crossing condition (as used by Epple, Fi l imon, and Romer (1983) and others). This assumption states that the slope of a consumer's indifference curve in (G,f) space increases as income increases. Here f represents the gross-of-tax price of housing. These assumptions are useful because they imply the desired sorting result - that there exists a marginal consumer who is indifferent between membership and nonmembership in the private government. This consumer partitions the population into a set of members (those wi th income higher than the marginal consumer) and a set of nonmembers (those wi th income lower than the marginal consumer). The Public Government The public government is obliged to provide a level of the public good to all inhabitants of the city. Denote g the level of the public good that is provided by the public government. The average cost of production for each unit of g is c. In order to finance provision, the public government levies an ad valorem property tax t on all housing services consumed in the city. A s the public government's budget must balance, choosing g automatically determines the tax rate t. Remember that even private government members consume g; this model rules out instances of secession. Assume for now that the objective of the public government is to maximize the aggregate welfare of all the residents in the city. 1 The Private Government There exists a private government that determines the level of the public good that wi l l be available only to members of the private government. 2 This level, denoted 7, is the "private supplement." Hence, each member of the private government wi l l consume G = g + 7 in total, while each nonmember wi l l only consume G = g.3 There is no cost. 1 I t is also possible to assume other governmental object ives, such as median voter u t i l i t y m a x i m i z a -t i on and property value max im iza t ion . For the numer ica l models later in the chapter, these al ternat ive object ives may be invoked. 2 A s s u m e here that the pr ivate government exists for granted; th is is not a mode l of pr ivate govern-ment format ion. See Hels ley and Strange (2000) for such a model . 3 T h i s model assumes that the pub l ic ly - and pr ivate ly-prov ided pub l ic goods are pure subst i tutes. T h i s follows Hels ley and Strange's (1998) formulat ion of the publ ic good. In th is sense, the te rm "p r i va te supplement" is apt. However, it is p lausib le to th ink of examples of pub l ic ly -prov ided and pr ivate ly-prov ided local services that are not subst i tutes. For example, pub l ic po l ic ing and pr ivate secur i ty services may be v iewed as complementary pub l ic goods. It wou ld then be more sensible to 114 advantage to the private government, which also produces 7 at the constant average cost of c. Instead of financing wi th a property tax, however, the private government simply charges each private government member the average cost c per unit of 7 pro-duced. This is in line wi th the usual way in which homeowner associations are financed, through a fixed assessment that does not differ across households. One question that can be examined wi th this model, then, is how these two different methods of financing differ. For now, assume that the objective of the private government is to maximize the aggregate welfare of its members. 5.3.2 T iming The model consists of two stages: Stage 1: Suppose the pre-existence of the three players, consumers, the public and the private governments. 4 The values of three variables are determined simultaneously. First, the consumers sort themselves into members or nonmembers. This amounts to determining y*, the income of the member who is indifferent between joining and not joining the private government. This individual may be called the marginal member. The determination of y* is done taking g and 7 as given. Second, the public government chooses g, taking y* and 7 as given. Th i rd , the private government chooses 7, taking the values of y* and g as given. Thus the equilibrium values of y*, g and 7 are resolved as a Nash equilibrium triplet. Stage 2: In the second stage, the consumers, having decided on membership, now maximize their uti l i ty by choosing bundles of x, the composite good, and h, housing services. Finally, the housing market clears and resolves housing prices. think in terms of a general public good function G(g,/y) that could exploit this complementarity. 4 This is in contrast to Helsley and Strange (1998), who add a stage before this one where the public government decides whether or not to allow the private government to form. This chapter chooses to focus only on the provision problem of the two governments as it is interested in how the private government affects the housing market. In addition, in practice public governments often do not have such unilateral authority in the creation of private governments. While their formation is promulgated by law, it is often under state regulation, not local. As well, after the law is passed, creation of private governments tends to be initiated by private individuals and not the public government. 115 5.3.3 Stage 2: Clearing of the Housing Market To characterize the solution of the model, begin wi th the last stage. W i t h i n each stage, it is useful to examine first the case in which there is only a public government maximizing total welfare. This Base Case w i l l be helpful in comparing the equilibrium conditions of this case with a private government. Stage 2 Without Private Government The second stage of the model is quite simple for the base case without a private gov-ernment. There is only one nonexcludable public good financed by a property tax on housing services. Housing services have the same price throughout the city, so that the problem of each consumer y% is simply max U(xl,h\ g) s.t. yl = xl + fhl where g is fixed, and f = (1 + t)r. (5.3.1) xl, hl Note that in the base case, g is the level of public services, solely provided by the pub-lic government and financed by the property tax t. Only one housing price r prevails in the market. Setting up the constrained maximization problem for each household is straightforward and clearly gives rise to the usual " M R S equals price ratio condi-tion," namely jf = f. Each consumer wi l l have a housing demand function h(r,g,yl). However, because the uti l i ty function is homothetic, the housing demand function is multiplicatively separable in prices and income. Therefore it is possible write consumer yl,s housing demand in the form h(f,g,yl) = (f)(yl)h(f,g), where <j)' > 0. Note that the h function is common to everybody; only the arguments are different. This is the unit housing function. Now the indirect ut i l i ty function of each consumer is V(r,g,yl) = U(y>-myi)Hr,g)A(yi)h(?,g),g). • (5.3.2) Finally, clear the housing market, resolving the net-of-tax housing price. In housing equilibrium, aggregate housing demand must equal aggregate supply, characterized by the following equation: [V (p(yi)h((l + t)r,g)f(yi)dyi = HS(r)- (5-3.3) y 116 Hs represents a generic upward-sloping housing supply function that depends on net-of-tax house prices. Now use the separability of the housing demand function to rewrite the equilibrium housing price equation: h{{l + t)r,g)$ = Hs{r) where $ = P' 0(?/) j'{ylW. (5.3.4) A t this point, introduce the shorthand $ to represent an aggregate income measure of the population that depends only on the income distribution and its support. Equation (5.3.4) is enough to determine the equilibrium r, which depends on the variables t and g. These are chosen in the preceding stage. Stage 2 With Private Government When a private government exists alongside a public government, the final stage of the model is slightly more complicated. B y this time consumers have already decided whether they wish to join the private government. A s mentioned earlier, this is captured by the existence of y*, the marginal consumer who is indifferent between membership and nonmembership. In this last stage, assume that y* exists, so that the set of nonmembers is [y, y*) and the set of members is [y*,y]. (For completeness, assume that the marginal member does join the private government.) The problem of each consumer y% is stated as the following: max U(x\ h\ g) s.t. yl = x1 + (1 + t)rNhl if a nonmember, and (5.3.5) maxU(xi,ti,g + 'y) s.t. y* = x* + (1+ t)rMti + vy if a member. (5.3.6) Recall that g is the level of public good provision by the public government, and 7 is the level of the private government supplement. The public government's tax rate is t, while rN is the net-of-tax price of housing services.for nonmembers, and rM is the net-of-tax price of housing services for members. The notation f\i = TV or M indicates the gross-of-tax price of housing for nonmembers and members, respectively. Solving (5.3.5) and (5.3.6) for each consumer gives a common housing demand func-t ion h(r\G,y'1). Aga in using the homotheticity of the uti l i ty function, consumer yl,s 117 housing demand can be written in the form ~ . I • <£>(yl)h(fN, q) if a nonmember, and h{f\g,j,yl) = <j>(yl)h(rl,G) = \ *Ky ' V : - J [ 4>(yl — cj)h(rM, g + 7) if a member. where (j)' > 0. Note that the h function is common to members and nonmembers; only the arguments are different. Hence the indirect uti l i ty function of each nonmember is V(fN, g, y<) = U{y* - r^tfMr"; g), <t>(yl)h(rN, g), g). (5.3.7) Whi le the indirect ut i l i ty function of each member is V(fM, 5 ) 7 , y') = U(yi - c 7 - f M 0(y* - cy)h(fM, g + 7), W - C1)h{fM, g + 1),g + 1). (5.3.8) Finally, in this stage, the housing markets clear, and this resolves the net-of-tax housing prices. The two equations that characterize the housing prices are as follows: [V (j)(yi)h((l+t)rN,g)f(yi)dyi = Hs(rN), and (5.3.9) [ ^-C1)h{{l+t)rM,g + 1)fWW = Hs(rM). . (5.3.10) Jy* The separability of the housing demand function implies it is possible to pul l out the housing demands from the integrals in both equations above. Thus the equilibrium housing price equations are h((l + t)rN,g)$N = Hs{rN), and (5.3.11) h{(l + t)rM,g + ~()<S>M . = Hs(rM); (5.3.12) /y* (j)(yl)f(yz)dyl, and and $ M = I" ttf-cymyW.-Jv* 'yNote that <&N depends only on y* while $ M depends only on y* and 7. These represent aggregate income measures of the nonmembers and the members respectively. Equations (5.3.11) and (5.3.12) determine the equilibrium rN and rM respectively. rN depends on y*, t and g, while rM depends on these parameters and c and 7 as well. 118 5.3.4 Stage 1: Nash Equilibrium Stage 1 Equilibrium Without Private Government Stage 1 without a private government is quite simple. There is no need to determine a marginal member, and of course there is no private government provision; Therefore this stage simply consists of the public government's choice of g. Now the objective of the public government is to choose provision and taxes to maximize the aggregate welfare of its members, subject to the budget being balanced. The problem is thus: max f U [yl - (1 + t y ^ h ^ l + t)r, g), ^ )h{{\ + t)r, g), g] f{ylW s.t. trh((l + t)r,g)$-cg>0. (5.3.13) The constraint simply says that the government's total property tax revenue cannot exceed the total cost of provision. Recall that for both the private and the public governments, each unit of the public good costs c of the numeraire to produce. The government's first order conditions (FOCs) for this problem are: F O C w.r.t. g: f Ux [-(1 + t ) ^ ) ^ ] f(yi)dyi+ Jy [V Uh [4>(yl)h2] f&W + I" UGf(y*W + K [trh2§ - c] = 0. (5.3.14) Jy Jy F O C w.r.t. t: I" Ux [-rfty^h - (1 + t)r7W)hi] f(y%W+ Jy f Uh [Mrh] fiy'W + K [rh$ + tr2h^\ = 0. (5.3.15). Jy where K is the Lagrange multiplier on the constraint, and I have assumed that the budget constraint binds. For clarity, suppress the arguments of the housing function and its partial derivatives. That is, h = h(f,g), hi = | | and h2 .= Now, moving the terms involving K to the other side and dividing (5.3.14) by (5.3.15) gives the the following equation: trh2$ - c fh2Ux - h2Uh - UG rh$ + tr2h^ [rh + (1 + £)r 2 / i i ] Ux - rhxUh' 119 (5.3.16) where Ux = U^iyWW, Uh = J ^ c ^ / G / W and UG = f» UaftfW-These are summary measures that define the aggregate marginal utilities of x, h and G respectively. Because their values differ over the range of incomes, they must remain inside the integrals. To further simplify equation (5.3.16), recall that the Stage 2 problem always implies that 77k = f. After substituting out the Uh terms, the following Samuelson condition describes the level of public provision: c = $ ter 1 + t UG UX + $ t r | | . (5.3.17) Here e r = the own-price elasticity of housing. This equation, combined wi th the private government's balanced budget constraint trh(r,g)$-cg = Q. (5.3.18) characterize the public government's optimal choices of g and t, when there is no private government. Equation (5.3.17) can be examined more carefully. This is the modified Samuelson rule for public good provision in the presence of a distortionary property tax. The left-hand side is the marginal rate of transformation between G and x — recall it takes c units of the numeraire to make one unit of the public good. The first term on the right-hand side comprises distortionary factor multiplied by a term clearly representing the sum of the marginal rates of substitution. The second term measures the effect that an extra unit of g provision has on total tax revenue v ia the housing market. This second term exists because of the complementarity between the public good and housing services. Wha t can be said about the size of the terms in equation (5.3.17)? The term UG/UX by itself represents the first-best level of provision, which would occur if the public good were financed by a non-distortionary lump-sum tax. Now 1 ter ^ tr dh(f, G) 1 + t h dr But. the Slutsky equation says dh(f,G) udh(f,G) 7TZ — Ohh — n n : or oy 120 where Shh < 0 represents the derivative of the compensated demand for housing. Now because uti l i ty is homothetic, d h Q y C ^ = 0 . Therefore . ter tr _ 1 - — = 1 + T S h h < 1. l + t h Hence the "true" benefit of the public good is scaled down from the first-best amount. So wi th property taxation, this term says there is underprovision of g relative to the first-best. The magnitude of the scaling-down factor depends not only on the size of the property tax and housing prices, but also on the compensated demand elasticity. If demand for housing is quite inelastic, the scaling-down factor is small. This is because although the property tax raises the price of housing for consumers, it does not cause a great substitution away from housing. Since property tax revenue is relatively un-changed, an amount of g close to the first-best can be provided. (Indeed, in the l imit ing case, if housing is perfectly inelastically demanded, this term is 1; since the tax revenue term is positive, the second-best level of g actually exceeds the first-best level. This w i l l also happen if housing is very inelastically demanded and the tax revenue term is large enough.) O n the other hand, if housing is relatively elastically demanded, the first term's effect wi l l tend to dominate the second, leading to a lower level of public good provided. What about the tax revenue term? If housing and the public good are complements, then provision of g w i l l increase tax revenue. This wi l l make the second term positive, which means that this distortion term tends to overprovide g relative to the first-best. Clearly, when both distortions are present, it is ambiguous whether the public govern-ment with property tax underprovides or overprovides g relative to the Samuelson rule. The substitutability and elasticity of housing demand are the keys to solving the puzzle. Stage 1 Equilibrium With Private Government After having examined the implications for a distortionary property tax without a private government, this section turns to the situation with a private government. There are simultaneous choices by the consumers, public government and private government. This stage wi l l involve the derivation of the three players' reaction functions. The three reaction functions wi l l characterize the three Nash equilibrium choices of y*, g and 7, if 121 an equilibrium exists. (The public property tax t w i l l fall out of the public government's budget constraint.) It is important to state that in this stage, all the players take the values of rN and rM as given. That is, the marginal private government member y*, the private government provision 7 and the public government provision g are all determined on the assumption that these choices wi l l not influence housing prices. 5 The chapter makes this admittedly myopic assumption primarily for tractability. If agents take into account the effects their collective actions have on net house prices, the model's complexity would increase substantially. In addition, for the case of private government membership decisions of individuals, it is reasonable to assume that agents do not believe their decisions are great enough to affect house prices. Of course, it is more difficult to justify the same for the public and the private governments. One would think that the private and the public governments' provision choice would be made with the knowledge that they wi l l influence net housing prices. Determination of the Marginal Member The consumers' choice in Stage 1 is simple. They decide whether or not to join the private government. They take the level of provision of g and 7 as given, as well as the property tax rates and house prices. A consumer of income y% w i l l therefore join the private government if and only if the uti l i ty of membership exceeds the uti l i ty of nonmembership. In other words, yl w i l l be a member if and only if U [y* - c 7 - f *W - n)hM, - c7)hM, g + y]>u[yi- f W ^ " . <f>(yl)hN\ g]. (5.3.19) where hM = h(fM, g + 7) and hN = h(rN,g), the income-independent housing demand functions. The goal is to find the marginal income y* for whom (5.3.19) holds wi th equality. It can be shown that the indifference curves of all individuals are increasing, continuous and concave in the (G, f) plane. The single-crossing assumption then implies that y* w i l l partition the population into two convex sets: all those y E [y,y*) strictly prefer not to join the private government, and al l those y e (y*,y] strictly prefer to join 5 Although residents and governments are not myopic in the sense that residents will adjust their housing consumption in response to changes in government provision or taxes. 122 the private government. Determination of Private Provision Now move on to the levels of government provision. It is simpler to begin wi th the private government's choice. The objective of the private government is to choose provision to maximize the aggregate welfare of its members, subject to the budget being balanced. Because the private government simply charges its members the unit cost of provision, the budget is automatically balanced. So the private government's problem reduces to the following: max [V U & - c 7 - (1 + t)rM^(yl - cry)h((l + t)rM, g + j), 7,r Jy. ^ - c7)M(l.+ t)rM, g + 7), S + 7] fiy'W- (5.3.20) The first order condition ( F O C ) for the private government is: fV U? [-c + f ^W - n)chM - f * W - c 7 ) < ] / (i/W Jy* + [V [-c0'(y* - n)hM + M ~ cry)h?] f(y*W + f UQ f(yi)dyi = 0. (5.3.21) Jy* • Jy* For clarity, suppress the arguments of the housing function and its partial derivatives. That is, hM = h{rM,g + 7), hf = § ^ and = Now, in a similar manner to the base case, simplify the first order condition by using the envelope condition from the Stage 2 uti l i ty-maximization problem: jf = fM for all members. The Samuelson condition results and reduces to the following: = Jy* G J \u ) t> ^ ug_ / 5 g The equations shows the absence of a property tax distortion. The private government sets an efficient level of the supplement, corresponding to an aggregate marginal rate of substitution. Note the private government's optimality condition is derived as if it believed that it was the only government. This is because the private government cares only about the effect of providing 7 on its members; it does not account for the effect 123 of additional government services on housing decisions outside its gates. A n d since the public good is financed by a nondistortionary fee rather than a property tax, a first-best level of the private supplement is chosen. Determination of Public Provision The objective of the public government is to maximize the aggregate welfare of the entire population. This differs from most previous multi-community models of L P G provision because in this model the public government cares about the welfare of members of the private government. The public government cannot deny private government members the consumption of the public good. The problem of the public government is max 9,t s.t. U [y* - (1 + t)rNcf>(y>)h((l + t)rN, g), + t)rN, g),g] f^dy* + f U [yl - cry - (1 + t)rM</>(</ - C1)h({\ + t ) r M , g + 7), Jy* 4>(y% - cry)h((l + t)rM, g + j),g + ^ f(yi)dyi trNh((l + t)rN, g)$N + trMh((l + t)rM, g + 7 ) $ " _ c g > 0 . (5.3.33) Again, find the first order conditions for the public government's optimization problem, and after some algebra, and using the fact that the Stage 2 optimization by consumers implies that up f and EL UP r , the following equation results: c = 1 + t r N h N $ N + 1 -1 + t r M h M $ M M G hNrNTjN + jjM j.M JJ M (i) dG dG (ii) (5.3.24) where in an analogous fashion to the previous section, er hN the own-price elasticity of housing for nonmembers. Also defined are three summary measures of aggregate marginal utility, this time pertaining to nonmembers: ry* uN = y Jy Jy 124 Equation (5.3.24) is the public government's modified Samuelson rule. This equation, combined wi th the public government's budget constraint, trNh{{l + t)rN, g)$N + trMh{{\ + t)rM, g + 7 ) $ M - eg = 0, (5.3.25) characterize the public government's optimal choices of g and t. Examine the Samuelson condition (5.3.24). First , it is important to see that the public government takes the private government's members into its provision decision. This is unsurprising as members contribute.to aggregate city welfare just as nonmembers do. Yet because members have an alternative source of public services, they are weighted differently than nonmembers. Starting from the right, term (iii) is similar to the tax revenue effect present in the optimality condition of the private government. Unlike (5.3.22), however, the public government takes into consideration the effects of public provision on the housing demands of both members and nonmembers. Again , if housing and government services are complementary enough, these tax revenue distortions imply overprovision of the public service relative to the first-best. A less clear-cut interpretation exists for terms (i) and (ii). It is obviously a weighted M R S , but because members and nonmembers are evaluating the partial derivatives of the uti l i ty function at different points, their M R S s are not summable. A s a result the terms do not simplify. Rather it looks like a combination of the M R S of members and nonmembers. Resolution of the Nash Equilibrium The analysis above characterizes the equilibrium choices of the three players in Stage 1. Consumers choose their marginal member given the levels of provision by the two governments. The private government and the public government each choose provi-sion, given the marginal member and the provision of the other government. The Nash equilibrium in Stage 1 wi l l be a 4-tuple {y* ,7 , g , t} such that equations (5.3.19 with equality), (5.3.22), (5.3.24) and (5.3.25) are al l satisfied simultaneously. The subsequent part of the analysis gives a numerical example, derives the equilibrium conditions and demonstrates some of its properties. 125 5.4 A Numerical Example A s the analysis in the previous section is general, it may be difficult to see the properties of the Nash equilibrium without solving the model explicitly. Therefore this section works through an example that is solved numerically. In this section, assume consumers share the following Cobb-Douglas uti l i ty function: U(x\ti,G) = {xi)a(ti)p{G)s, where [3 > 8 to ensure stability. To get tractable results, it is necessary to alter the objectives of the public and the private government. The numerical example assumes majority voting in Stage 1 to determine the levels of g and 7. This assumption simplifies the Samuelson conditions because it no longer involves aggregating the marginal utilities over groups of consumers. 5.4.1 Stage 2 In Stage 2 the levels of g, t and 7 are known and taken as given. The marginal member y* is also exogenous. Standard uti l i ty maximization implies that nonmembers' demands for housing and the numeraire are a + p{l+t)rN ' a + pV: Members' demands for these goods are h a + p(l + t)rM ' X a + p[V l h 5.4.2 Stage 1 Stage 1 now involves solving for the Nash equilibrium values of g, t, 7 and y*. Begin wi th the marginal member condition. The marginal member y* satisfies the following: aa/30 ^ { y * - c i ) a + 0 { { l + t)rM)-^g + 1f / „.a Q0 \ . (y*y+P((l+t)rNrPgs. (5.4.1) aapp 126 Isolating y*, the following condition determines the marginal private government mem-ber: V — ~JT ~J~J 1~- ^ o . ^ . z ; (y-W) a + / 3 (gr - f - /y) Q + / 3 _ (T~-M) a + / 3 (g) Q + / 3 Next derive the Samuelson rule for the level of private government provision. The level of 7 is assumed to be chosen in pairwise elections by members of the private government. As such, the median voter theorem can be invoked because the single-crossing condition implies that members' ideal levels of g + 7 rises with income. Since g is taken as given by all members of the private government, the level of 7 that prevails is that which is preferred by the median voter in the interval [y*\ y]. C a l l this consumer the median voter-member, designated y. Note that the identity of y depends on y*. The private government's problem then amounts to maximizing the median voter-member's utility. It is useful to first do this in for a general ut i l i ty function. The median voter-member's problem is m a x ^ ^ - c 7 - ( l + t ) r M ^ - C 7 ) / i ( ( l + t ) r M , ^ + 7 ) , ^ - C 7 ) / i ( ( l + t ) r M , ^ + 7) ,p + 7]. 7 The first-order condition is C/f [ - c - (1 + t)rMch<p\y - cry) - (1 + t ) r M cj>(y - C1)h2} + U^[-ch<l/(j-cry) + <l>(i-cry)h2] + U^ = 0. (5.4.3) Next, use the envelope condition from Stage 2, which states that for all members, rjM = ( 1 + ^ ) r M • This results in the following first-order condition: U, If™ x (5.4.4) Like the analysis in the previous section, the private government sets a non-distortionary level of 7. This time, the supplement is that which is preferred by the median voter-member. Back to the Cobb-Douglas example, the private government provision condi-tion simplifies to the following: . c(a + p + o) \a + p + dj 127 This is written in the form of a reaction function for 7, and presents a key implication. Note that the negative sign implies that g and 7 are strategic substitutes. As public government spending goes up, the equilibrium private government supplement decreases. However, due to the presence of the distortionary property tax, this decrease is not at a one-for-one rate. The last problem to solve for in Stage 1 is the public government's provision decision. Here it is necessary to consider two cases depending on whether the median voter overall, y, is or is not a member of the private government. Case A: y < y* First consider the case where the median voter overall is not a member of the private government. She chooses g and t to maximize uti l i ty subject to the community budget constraint and taking rN, rM, y* and 7 as given: P ct max a In ( -y I + 8 In g,t \a + 8 ' y s.t. tr N P y y a + p(l + t)r a + p(l+t)r N tr M \V P y C7 £1 i\ J i ' f(y)dy =cg. a + p(l+t)rM-Sing (5.4.6) (5.4.7) It is possible to simplify the budget constraint by the fact that C9 = tr" _Pof, y^Mf(yW + trM f J, a T^JtfW eg eg jy_ a + p(l+t)r t \ ( p r a + p(l + t)r l+tj \a + p t \ ( P ylf{ylW - c 7 / fiy^dy y l + tj \a + p [E[y] - cjM] (5.4.8) where E[y] is the mean income over the whole population, and M is the population of the private government. After this simplification, the two first order conditions for the 128 public government's choices of g and t are the following: t = - , and (5.4.9) 0 = da + W + ^ W-W- (5-4.10) Note that these choices of g and t imply that there is no strategic interaction between the public and the private government. This arises as the private government's provision does not affect the median voter's objective function. Case B: y > y* The next case determines the public government's provision in the case where the median voter overall is a member of the private government. This is clearly the more interesting case, as this case is the one in which the public and private governments truly interact. This case is also increasingly relevant in many municipalities, where over half of the population belongs to a homeowners' association. This time, the median voter chooses g and t to maximize the following uti l i ty function subject to the community budget constraint and taking rN, rM, y* and 7 as given: f a \ ( j3 (y - erf) \ m , f a l n \—p{i> ~ ^ J + P h l [a + f)(1 + t ) r » + ^ C ^ h f k $ i f < * w = c > - ( 5 ' 4 ' n ) The F O C s for this problem reduce to the following: ' = clalWlJ)-1' ^ ( 5 ' 4 ' 1 2 ) = (a+V+t)™-^- (5Ai3) Note that equation (5.4.12) is expressed as a reaction function of g wi th respect to 7. The slope of the reaction function is clearly d# -SM d 7 . (a+ /?)(!+<) 1 < - 1 . 129 That is, there is a more than one-for-one reduction in the public government's pro-vision as the private government provides another unit. In other words, the strategic downloading problem is greater than one-for-one. Also, note that in both of the F O C s , the income of the median voter y has disappeared, and instead the relevant income is those of the mean individual. It is interesting that the median voter takes M, the population of the private government, into account whether he is or is not a member. 5.4.3 Some Numerical Results Exogenous Housing Prices This section presents numerical results from the above model, shows that an equilibrium exists and discusses some of the implications. The following parameter values for the uti l i ty function are used: {a = 0.3, j3 = 0.5,5 — 0.2}. A s most of the interesting action happens in Stage 1, the first numerical exercise is to present results where the housing prices are taken as given and to demonstrate that a Stage 1 equilibrium exists. Assume that the housing prices are taken as given: rN — l,rM — 2. The cost of provision is c = 1 for both governments. Finally, incomes are assumed to follow a Weibul l distribution, which has a probability density function that is right-skewed (that is, the median is less than the mean). Incomes from this distribution range from zero to infinity. The Weibul l distribution has a scale parameter T) and a shape parameter b. The example assumes 77 = 10, b = 3. This results in a median income of 8.85. The mean income is given by while the population of the private government, which is clearly dependent on y*, is [WJ J The parameter values above constitute the benchmark case. These are varied in order to examine the stability and comparative statics properties of the equilibrium. Finally, the median voter member has income 130 Table 5.1: Numerical Model: Public Property Tax, Private Fee Financing. Some Comparative Statics (Incomes are distributed in a Weibul l distribution wi th parameters 77 = 10, /3 = 3.) 2 1 1.2 1.4 1.6 1.8 2.2 2.4 y* 3.0036 2.6427 2.7481 2.8347 2.9044 2.9598 3.0383 3.0658 7 1.7859 1.7684 1.7731 1.7772 1.7807 1.7836 1.7878 1.7894 ' 9 0.0086 0.0214 0.0180 0.0150 0.0124 0.0103 0.0072 0.0061 t 0.0019 0.0048 0.0040 0.0033 0.0028 0.0023 0.0016 0.0014 c 1 1.1 1.2 1.3 y* 3.00364346869 3.00364346863 3.00364346863 3.00364346859 7 1.7859 1.6236 1.4883 1.3738 9 0.0086 0.0078 0.0071 0.0066 t 0.00191029641761 0.00191029641762 0.00191029641772 0.00191029641777 Table 5.1 and Figure 5.4.3 present Stage 1 equilibrium results from the assumption of exogenous housing prices. It is assumed for these calculations that the overall median voter (the one who chooses g) is a member of the private government, that is, y > y*. This is the more interesting case as it highlights the interaction between the public and the private government choices. The first column of each panel represents the benchmark parameters. The figure shows two comparative static exercises: first, the price of housing within the private government increases. Second, the per unit cost of service provision changes. The calculations show that Stage 1 equilibria do exist, and that the equilibrium values behave as theory and intuit ion suggests. A s housing becomes more expensive within the private government, the marginal member increases; that is, fewer people join the private government. As well, voters substitute away from 7 and towards g, which raises the property tax rate. This provides an interesting comparison of the two methods of financing. A s the private government housing price increases, the substitution away from g happens because it reduces the property taxes paid by members. This is replaced by consumption of non-distortionary 7. When c is varied, on the other hand the equilibrium marginal member hardly changes, and in fact the change is non-monotonic. However, other variables behave as expected. 131 Figure 5.1: N u m e r i c a l M o d e l : E x o g e n o u s H o u s i n g P r i c e s . W e i b u l l D i s t r i b u -t i o n o f I ncomes . M e d i a n V o t e r Is a M e m b e r Exogenous housing prices: Effect of increasing m on y* 1 1.5 2 2.5 3 3.5 4 Housing price within priv govt Exogenous housing prices: Weibull distribution, Majority voting, Median Voter > y* I I 1 1 I I Equil ibrium Va lues : Marginal member = 3.004 Private government provision = 1.79 Publ ic government provision = 0.008 ^ ^ 0 * ~ ~ 0 ~ ^ ' - Property tax rate = 0.0019 _ ^ - - - * - ^ ^ \ - - -i i um — un _ _ — — —~ _ _ - - Parameters : c = 1 i i i i i i m = 2; n = 1 1 0 1 2 3 4 5 6 7 8 9 10 Income 132 Both g and 7 decrease, but interestingly the property tax rate increases slightly. Endogenous Housing Prices The next numerical exercise is to endogenize the housing prices to see whether or not the value of public services would be capitalized into housing prices, and how that would affect Stage 1 equilibrium choices. For this exercise the incomes are assumed to be uniformly distributed on [0,1]. The same uti l i ty function parameters are used: a = 0.3, P = 0.5,7 = 0-2- However, c is assumed to be 0.05. To derive the equilibrium, it is assumed that the governments choose the level of provision v ia the median voter. For this exercise, assume that the median voter chooses not to join the private government: y* > 0.5. Figure 5.4.3 gives the equilibrium values of the five endogenous variables. 6 The figure also demonstrates that the residents satisfy the single-crossing condition. The marginal member has an income of 0.8, and al l those wi th higher income choose to join the private government. The equilibrium private government provision is roughly twice that of public government provision, which represents a very high level of service that members enjoy. The equilibrium shows that indeed the housing price wi thin the private government is higher than that outside, reflecting a capitalization of the increased level of public good. 5.5 Conclusion A s private governments take over more of the roles previously reserved for public local governments, the strategic interaction between private and public providers of public goods becomes more relevant an area of study. This paper has presented an analysis of a private government in the context of the urban housing market. B y incorporat-ing a housing market, another channel is opened for the interaction between private governments and the traditional public government. Numerical analysis showed that public and private government provision are strategic substitutes, and this can result in suboptimal provision of public goods. 6 T h e der ivat ion of the equ i l ib r ium first order condi t ions may be obta ined f rom the author. 133 Figure 5.2: N u m e r i c a l M o d e l : E n d o g e n o u s H o u s i n g P r i c e s . U n i f o r m D i s t r i -b u t i o n o f I n c o m e s . M e d i a n V o t e r Is N o t a M e m b e r Endogenous Housing Prices Uniform income distribution, Majority voting, Median voter < y* 0.7 0.6 0.5 0.4 0.3 d) N (0 <D 0.2 CC 0.1 -0.1 Equilibrium Values: Housing price within private government = 0.908 Housing price outside private government = 0.839 Marginal member = 0.800 Private government provision = 2.165 Public government provision = 1.840 Property tax rate = 0.444 134 CHAPTER 6 Conclusion The objective of this dissertation was to examine the impact that homeowners' associa-tions have had on local public finance. The growing popularity of these associations as a dominant method of local service delivery and their roles as residential private gov-ernments underline the importance of understanding their effects on traditional public finance. The thesis was composed of four parts. Chapter 2 presented a brief survey of home-owners' associations. It argued that these private associations share characteristics wi th municipal governments, and so the term private government is apt. It outlined the structure, popularity and roles of homeowners' associations and reviewed evidence that homeowners' associations have important impact on local public spending and city com-position. The chapter closed on the idea that as homeowners' associations become more and more common in residential development, it is increasingly important to under-stand the interactions between private governments, public governments and homeown-ers. This idea forms the motivation for the three following chapters in the dissertation. Chapter 3 examined the impact that homeowners' associations have had on local government expenditure. The study uses a self-compiled panel data set of planned de-velopment membership rates and public government expenditures from 1970 to 1999. The results are consistent wi th the view that local public governments have responded to the increased membership in private governments by reducing expenditures. However, local governments did not treat all expenditures equally. Results suggest that download-ing to private governments exists for those services that are relatively substitutable, and 135 does not exist for services wi th more public good aspects. The chapters concluded by showing that condominiums did not exert the same impact that planned developments did; this also served to verify the robustness of the empirical analysis. Chapter 4 reversed the question and asked if public government policy may have contributed to the substantial growth in private governments. The chapter examined the role of a state-imposed property tax and expenditure l imitat ion in encouraging the membership in and the formation of private governments. A theoretical model of pri-vate government, incorporating a simple housing market, is proposed. This model gen-erated a comparative static, which is tested using Californian data around the period of Proposition 13. Empir ica l estimates are consistent with the view that Proposition 13 contributed to the rise of the homeowners' association. In addition, there is some evidence that cities that anticipated a high levels of constraint due to Proposition 13 had faster growth in private government wi th cities wi th low levels of constraint. Finally, Chapter 5 extended the theoretical private government model of Helsley and Strange (1998) by incorporating a housing market. Equi l ibr ium conditions between a public government, a private government and homeowners are characterized. A s the model is analytically untractable, a numerical model is presented and worked out, and several elements of the model are demonstrated. 136 Bibliography A M E R I C A N H O U S I N G S U R V E Y (2001): "Table 2-13. Selected Housing Costs-Occupied Units," h t t p : //www. census .gov /hhes /www/hous ing /ahs /ahs01/ tab213 .h tml . A N D E R S O N , B . L . (1996): "Common Interest Developments: C I D Homeowners' Fees and the Issue of Double Taxation," Discussion Paper 5, Publ ic Law Research Institute, University of California Hastings College of the Law. A R O N S S O N , T . , J . L U N D B E R G , A N D M . W I K S T R O M (2000): "The Impact of Regional Public Expenditures on the Local Decision to Spend," Regional Science and Urban Economics, 30(2), 185-202. A U S T I N , D . A . (1999): "Politics vs Economics: Evidence from Munic ipa l Annexation," Journal of Urban Economics, 45, 501-532. B A R T O N , S. E . , A N D C . J . S I L V E R M A N (1994): "History and Structure of Common Interest Communities," in Common Interest Communities: Private Government and the Public Interest. Institute of Governmental Studies. B E R G S T R O M , T . , A N D R . G O O D M A N (1972): "Private Demands for Publ ic Goods," American Economic Review, 63(3), 280-296. B I C E , D . C . , A N D W . H . H O Y T (2000): "The Impact of Mandates and Tax Limi ts on Voluntary Contributions to Local Publ ic Services: A n Appl ica t ion to Fire-Protection Services," National Tax Journal, 53(1), 79-104. B O G A R T , W . T . (1998): The Economics of Cities and Suburbs. Prentice-Hall, Upper Saddle River, New Jersey. B O R C H E R D I N G , T . E . , A N D R . T . D E A C O N (1972): "The Demand for the Services of Non-Federal Governments," American Economic Review, .62, 891-901. B R U E C K N E R , J . (1998): "Testing for Strategic Interaction Among Local Governments: The Case of Growth Controls," Journal of Urban Economics, 44, 438-467. (2003): "Strategic Interaction Among Governments: A n Overview of Empir ica l Studies," International Regioanl Science Review, 26(2), 175-188. C A I R E S E A R C H F O U N D A T I O N (1999): National Survey of Community Association Homeowner Satisfaction. Community Associations Institute. C A L A B R E S E , S., A N D D . E P P L E (2004): "On the Poli t ical Economy of Tax Limits ," mimeo. 137 C A M E R O N , A . C , A N D P . K . T R I V E D I (1986): "Econometric Models Based on Count Data: Comparisons and Applications of Some Estimators," Journal of Applied Econo-metrics, 1, 29-53. (1998): Regression Analysis of Count Data, vol. 30 of Econometric Society Monographs. Cambridge University Press, Cambridge. C H A P M A N , J . I . (1981): Proposition 13 and Land Use: A Case Study of Fiscal Limits in California. Lexington Books, D . C . Heath and Company, Lexington, Massachusetts. C H U L A K , M . (2004): " H O A Q & A , " h t tp : / /www.hoaqanda .com/Bylaws.h tm. C O M M U N I T Y A S S O C I A T I O N S I N S T I T U T E (2003): "Facts About Community Associa-tions," h t t p : / / w w w . c a i o n l i n e . o r g / a b o u t / f a c t s . c f m . (2005): "Data on U.S . Community Associations," h t t p : / / w w w . c a i o n l i n e . o r g / a b o u t / f a c t s . c f m . C O N G D O N , P . (1993): "Approaches to Model l ing Overdispersion in the Analysis of Migrat ion," Environment and Planning A, 25(10), 1481-1510. C U T L E R , D . M . , D . W . E L M E N D O R F , AND R . J . Z E C K H A U S E R (1999): "Restrain-ing the Leviathan: Property Tax Limita t ion in Massachusetts," Journal of Public Economics, 71(3), 313-334. E P P L E , D . , R . F I L I M O N , A N D T . R O M E R (1983): "Housing, Vot ing and Moving: Equi l ibr ium in a Mode l of Local Publ ic Goods with Mul t ip le Jurisdictions," Research in Urban Economics, 3, 59-90. (1984): "Equi l ibr ium Among Local Jurisdictions: Toward an Integrated Treat-ment of Vot ing and Residential Choice," Journal of Public Economics, 24(3), 281^308. E P P L E , D . , A N D R . R O M A N O (1996): "Ends Against the Middle: Determining Pub-lic Service Provision When There Are Private Alternatives," Journal of Public Eco-nomics, 62(3), 297-325. E P P L E , D . , A N D T . R O M E R (1989): "Flexibi l i ty of Munic ipa l Boundaries," Journal of Urban Economics, 26, 307-319. — (1991): "Mobi l i ty and Redistribution," Journal of Political Economy, 99(4), • 828-858. F I G L I O , D . N . (1998): "Short-Term Effects of a 1990s-Era Property Tax Limi t : Panel Evidence on Oregon's Measure 5," National Tax Journal, 51(1), 55-70. F I G L I O , D . N . , A N D A . O ' S U L L I V A N (2001): "The Loca l Response to Tax Limita t ion Measures: Do Local Governments Manipulate Voters to Increase Revenues?," Journal of Law and Economics, 44, 233-257. F L E I S C H M A N N , A . (1986): "The Politics of Annexation: A Preliminary Assessment of Competing Paradigms," Social Science Quarterly, 67, 128-142. G L O M M , G . , A N D B . R A V I K U M A R (1998): "Opting Out of Publ ic ly Provided Services: A Majori ty Vot ing Result," Soial Choice and Welfare, 15(2), 187-199. 138 G O R D O N , T . M . (2003): "Crowd Out or Crowd In?: The Effects of Common Inter-est Developments on Poli t ical Participation in California," The Annals of Regional Science, 37, 203-233. (2004): "Planned Developments in California: Private Communities and Publ ic Life," Discussion paper, Publ ic Policy Institute of California. G O U V E I A , M . (1997): "Majority Rule and the Publ ic Provision of a Private Good," Public Choice, 93(3-4), 221-244. G R E E N E , K . V . , A N D P . J . N E L S O N (1994): "Legislative Majorities and Alternative Theories of the Size of Government," Public Finance, 49(1), 42-56. H A N S E N , N . , A N D A . K E S S L E R (2001): "(Non-)Existence of Equi l ibr ia in Mul t icom-munity Models," Journal of Urban Economics, 50(3), 418-435. H E L S L E Y , R . , AND W . S T R A N G E (1998): "Private Government," Journal of Public Economics, 69(2), 281-304. (2000): "Potential Competi t ion and Publ ic Sector Performance," Regional Science and Urban Economics, 30(4), 405-428. J O H N S T O N , J . L . , A N D K . J O H N S T O N - D O D D S (2002): "Common Interest Develop-ments: Housing at Risk?," Discussion Paper CRB-02-012, California Research B u -reau, California State Library. J O Y C E , P . G . , A N D D . R . M U L L I N S (1991): "The Changing Fiscal Structure of the State and Local Publ ic Sector: The Impact of Tax and Expenditure Limitations," Public Administration Review, 51(3), 240-253. K L E I N , K . E . (1995): "Owners Complain of Double Tax," Los Angeles Times. L A N G , R . E . , A N D K . A . D A N I E L S E N (1997): "Gated Communities in America: Wall ing Out the World?," Housing Policy Debate, 8(4), 867-899. L E V I N , A . , C . - F . L I N , A N D C . - S . J . C H U (2002): "Unit Root Tests in Panel Data: Asymptot ic and Finite-Sample Properties," Journal of Econometrics, .108, 1-24. M A L L E T T , W . J . (1993): "Private Government Formation in the D C Metropoli tan Area," Growth and Change, 24(Summer), 385-415. M A T S U S A K A , J . G . (1995): "Fiscal, Effects of the Voter Initiative: Evidence from the Last 30 Years," Journal of Political Economy, 103(3), 587-623. M C K E N Z I E , E . (1994): Privatopia : Homeowner Associations and the Rise of Residen-tial Private Government. Yale University Press, New Haven. (1998): "Homeowner Associations and California Politics: A n Exploratory Analysis," Urban Affairs Review, 34(1), 52-75. M C M I L L E N , D . P . , A N D S. C . S M I T H (2003): "The Number of Subcenters in Large Urban Areas," Journal of Urban Economics, 53(3), 321-338. O ' S U L L I V A N , A . , T . A . S E X T O N , A N D S. M . S H E F F R I N (1995): Property Taxes and Tax Revolts. Cambridge University Press, Cambridge, New York. 139 P A P K E , L . E . (1991): "Interstate Business Tax Differentials and New F i r m Location: Evidence from Panel Data," Journal of Public Economics, 45(1), 47-68. P O T E R B A , J . M . (1992): "Taxation and Housing: O l d Questions, New Answers," Amer-ican Economic Review Papers and Proceedings, 82(2), 237-242. R E I C H , R . (1991): "Secession of the Successful," The New York Times, 16-17, 42-45. R E I C H M A N , U . (1976): "Residential Private Governments: A n Introductory Survey," The University of Chicago Law Review, 43, 253-306. R E I T E R , M . , A N D A . W E I C H E N R I E D E R (1997): "Are Publ ic Goods Public? A Cr i t i ca l Survey of the Demand Estimates for Local Publ ic Services," Finanzarchiv, 54, 374-408. R O L A N D , H . E . (1998): "Residential Common Interest Developments: A n Overview," Discussion Paper CRB-98-006, California Research Bureau, California State Library. R O S E - A C K E R M A N , S. (1979): "Market Models of Local Government: Ex i t , Voting, and the Land Market," Journal of Urban Economics, 6(3), 319-337. R O S E N B E R R Y , K . (1996): "Common Interst Development Issues after Nahrstedt, Sum-mary Report," Discussion paper, Senate Housing and Land Use Committee, Sacra-mento. S C H W A D R O N , T . , A N D P . R I C H T E R (1984): California and the American Tax Revolt. University of California Press, Berkeley, California. S H A D B E G I A N , R . J . (1998): "Do Tax and Expenditure Limitat ions Affect Local Govern-ment Budgets? Evidence from Panel Data," Public Finance Review, 26(2), 118-136. — (1999): "The Effect of Tax and Expenditure Limitat ions on the Revenue Structure of Local Government, 1962-87," National Tax Journal, 52(2), 221-237. S M I T H , R . T . (1991): " L o c a l F i s c a l Arrangements, Home Rule, and California's Fiscal Consti tution after Proposition 13," in Proposition 13: A Ten-Year Retrospective, pp. 51-112. The Lincoln Insitute of Land Policy. S O N S T E L I E , J . (1979): "Public School Quali ty and Private School Enrollments," Na-tional Tax Journal, 32, 343-354. S T A B I L E , D . (2000): Community Associations: The Emergence and Acceptance of a Quiet Innovation in Housing. Greenwood Press. T I E B O U T , C . (1956): " A Pure Theory of Loca l Expenditures," Journal of Political Economy, 64(5) , 416-424. U . S . C E N S U S B U R E A U (1980): "Boundary and Annexation Survey: 1970-1979," Report GE-30-4, U.S . Census Bureau, Washington, D . C . V l G D O R , J . L . (2004): "Other People's Taxes: Nonresident Voters and Statewide L i m -itation of Local Government," Journal of Law and Economics, 47(2), 453-476. V I L L A S - B O A S , J . M . , A N D R . S. W I N E R (1999): "Endogeneity in Brand Choice M o d -els," Management Science, 45(10), 1324-1338. 140 

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.831.1-0092783/manifest

Comment

Related Items