Ballot-box zoning, transaction costs and land development

By Samuel R. Staley
Reason Public Policy Institute

Samuel Staley, Ph.D. directs the Urban Futures Program at the Reason Public Policy Institute in Los Angeles. Comments and suggestions should be directed toward Dr. Staley at the Reason Public Policy Institute.

Working papers are intended to make scholarly research on urban development and planning issues more widely available to urban policy analysts and researchers on urban policy. Working papers represent research in progress and are published to invite comment and discussion as preparation for their submission to academic journals and other professional publications. The authors are solely responsible for the content of their research and analysis.

Abstract

This paper uses a transaction-cost approach to urban growth and policy to assess the impacts of public referenda on land development in cities. Public referenda, or ballot-box zoning, increases uncertainty, and hence transaction costs, in the land development process and can discourage investment in land and buildings. An analysis of housing unit growth from 1980 to 1994 in 63 Ohio cities finds consistent and robust evidence that subjecting land use decisions to public referenda created a housing unit "growth penalty" for cities. Moreover, the empirical results were consistently negative irrespective of whether the city rejected or accommodated the proposed zoning, suggesting subjecting land-use decisions to political processes was more important than the outcome of the election.

I. Overview

A community's general policy environment, or "policy infrastructure," can have important implications for land-use and economic development. Local programs and strategies set the "rules of the game" for local land development and provide the institutional environment in which investment and growth takes place. Examples of this policy infrastructure include commitments to infrastructure or local tax policy, both of which impact the relative costs of investing in a community.

Another element of this policy infrastructure is land-use regulation. The generalized nature of the regulatory process means land-use policy impacts every investment in a community, from the construction of a new home to the expansion of a commercial enterprise. Development regulations are pervasive since the effective enforcement of a local zoning code requires subjecting almost all development projects to some form of review and/or approval by a public body or agency. Thus, planning and zoning decisions are particularly relevant to discussions of economic development policy.

The nature of these impacts is not well understood, however. To the extent development regulations improve the quality of land development, the economic and social impacts can be beneficial. If, on the other hand, regulatory processes impose costs on development without improving its quality, the impacts can be negative. Since regulatory processes interfere with economic exchange and may impose higher costs, the economic consequences are likely to reduce economic growth and development.[1] This may be especially true if regulatory processes inject uncertainty into the land development process (Titman 1985; Mayo and Shephard 1991; Evans 1992).

One of the more recent developments in urban planning has been the expanded use of referenda and initiatives to broaden citizen participation in local land use decisions. Referenda have been viewed primarily from a political and sociological perspective as effective ways to enfranchise local citizens (e.g., Caves 1992). In the U.S., the phenomenon has been called "ballot-box zoning." Few researchers and analysts, however, have explored the potential economic consequences of using initiatives and referenda to set (or reverse) economic development policy, particularly in the area of land use regulation. This paper begins to fill this void by exploring the impact of ballot-box zoning has on land development in urban communities.

The next section provides a framework for assessing the way popular referenda influence uncertainty in land development using a transactions-cost paradigm. A transaction-cost approach hypothesizes that public referenda increase the costs of land development and reduces incentives to invest and build in urban areas. Section 3 provides an overview of the empirical model and data used to test this hypothesis. Section 4 reports and discusses the empirical results derived from the empirical model. Section 5 concludes the paper with a discussion of the policy implications.

II. Transaction costs and urban development

Transaction costs are the costs of negotiating, monitoring and enforcing contracts. Coase (1937; 1960) pioneered the concept as theoretical tool, first using transaction cost theory to explain why firms exist and then why market failures occur. In essence, the costs of contracting among parties determines whether production will be internalized into a firm or whether goods and services will be bought and sold through the market using third parties. In the context of market failure, Coase emphasized the importance of well defined and enforceable property rights for establishing an institutional environment favorable to contracting. If property rights were not enforceable or well defined, markets were unlikely to incorporate all costs and benefits of a transaction into an exchange thus creating externalities. In a zero transaction-cost environment, all costs and benefits, including third party impacts, would be included in the exchange and markets would efficiently and equitably allocate resources in a socially beneficial way.

Transaction costs and land-use policy

The transaction-cost paradigm has several implications for land-use policy. Land development can be considered an extension of the contracting paradigm. Given the pervasive influence of zoning and other development controls, land developers almost always require approval from a public agency to move forward on a project. Regulatory processes that require public agency approval thus establish a bargaining relationship between land developers and public agencies (Staley 1997). Ultimately, this bargaining will result in the local planning board granting development permission or rejecting a site plan or rezoning request submitted by a developer or property owner. Development approval becomes a contract that defines the expectations, responsibilities and obligations of parties in the exchange and thus becomes a mechanism for reducing uncertainty and hence the transactions costs of economic exchange. In this context, well defined contracts can facilitate exchange, and, ultimately, promote economic growth and development (North 1990).

A poorly designed or highly uncertain system of local planning and zoning can discourage investment and economic development, particularly in areas faced with weak demand for property development or if it disrupts the contractual dimension of property development (Lai 1996a; 1996b). The local planning system and procedures established and used by state and local governments provide the governance structure - the rules of the game for bargaining, negotiating, monitoring, and enforcing contracts - for land development within the community.[2] Subjecting economic decisions, including land development transactions, to the legislative processes also can weaken the rule of law and destabilize the system of property rights that supports market activity and transactions. Legislatively-driven transactions are often less stable and more uncertain because of the potential for political manipulation and control and the explicit use of force through the state to achieve objectives.

Risk and uncertainty are key elements of these transaction costs. Uncertainties in the plan application and approval process encourage developers to withhold investments in land until more information becomes available about allowable land uses. Unpredictable planning delays can effect private investment decisions and hence economic development (Allom 1991; Staley 1994). Mayo and Shephard (1991), for example, argue that an increase in the variance surrounding development approvals could adversely effect housing supply. Evans (1992) argued that these types of delays significantly reduced the supply of housing in southern England, and Allom (1991) argued that planning delays significantly add to land development costs in Australia. Altschuler et al. (1993, p. 49-51) note that land developers are probably the most time sensitive parties in the development process because for them, more than any other party, "time is money." North (1990) argues that the development of services was crucial to advanced economies because they reduced transactions costs associated with investment and business development. Thus, while land-use planning may reduce some elements of uncertainty by ensuring approved projects proceed, uncertainty surrounding the application approval process or future zoning decisions could increase uncertainty (or be perceived as increasing the randomness surrounding land development) and discourage land development.

Transaction costs and urban development

Theoretically, the key issue in this study is how ballot-box zoning impacts the transaction costs associated with land development. If public referenda increase uncertainty, create higher transaction costs, or de-stabilize the property rights structure of land development, they could reduce investment in land and discourage economic development. If, on the other hand, public referenda were relatively isolated cases, independent of the approval process (e.g., very site or project specific), their impact might be minimal and have little impact in economic growth and development.[3]

Two hypotheses about the relationship between zoning referenda and economic development are implicit when transaction costs are considered in the context of land development: 1) Higher planning-related transaction costs discourage the level of land development and economic growth, and 2) uncertainty in the planning process drives up transaction costs and reduces economic growth. The following sections develop an empirical model that tests these hypotheses.

III. Model and empirical framework

Generally, land development (B) is driven by external variables (E), non-planning policy variables (G), and planning-related policy variables (P). External variables include factors such as the demand for building units, economic composition of the labor force, employment base, cyclical economic effects, and others. Non-planning policy variables included policies local citizens control, such as local government spending, debt levels, or the amount spent on local infrastructure. Planning-related variables represent locally-determined policies that impact land development such as the structure of the zoning code, uncertainty in the zoning and rezoning process, permit processing, or notification of public hearings. Each of these factors could effect the length, detail, and timeliness of the plan approval process and transaction costs associated with development. Thus, a general model or urban development can be specified such that:

(1) B = f(E,G,P)

where E, G, and P are vectors for independent variables representing external, non-planning policy, and planning-related policy variables.

To test this model, data were collected for Ohio cities for the period 1980 to 1994.[4] While restricting the analysis to Ohio may reduce its external validity, the interpretation of the results is enhanced because political institutions and regional variations in the economy are held constant. Thus, while the differences in political, legal and cultural institutions between Hong Kong, Houston, and Cleveland make inter-city comparisons of planning processes problematic, these sources of variation are minimized by restricting the analysis to one state where political, cultural, economic, and legal institutions are held constant. Thus, the model should more adequately capture differences in implementing planning rules independent of broader institutional factors.

Data limitations[5] also required limiting the sample to cities. Smaller towns and unincorporated areas were excluded, truncating the sample. The omission of small cities and townships is potentially important since townships and small cities often have faster processing times for rezoning applications and small cities are also important areas of new growth. Thus, statistical biases may occur if the traditional assumptions underlying Ordinary Least Squares (OLS) regression analysis are violated.[6] To compensate for data deficiencies, the models were estimated using Generalized Least Squares (GLS) to correct for heteroskedasticity.[7]

To further facilitate data gathering and inter-urban comparability, the sample was limited to central-city counties. Restricting the analysis to central city counties allowed for an analysis of cities within a competitive land market with similar regulatory systems for land use. In rural Ohio counties, development regulation is often a county-level responsibility, and, when local planning exists, township governments often have more flexible zoning codes and systems without full-time planning staff. Five major metropolitan areas were selected: Akron, Cincinnati, Cleveland, Columbus, and Dayton. Toledo and Youngstown were excluded because, as border towns, land development patterns and trends may be effected by different institutional environments operating in neighboring states (Indiana, Michigan and Pennsylvania).[8]

A second statistical problem was multicollinearity. Preserving degrees of freedom was important given the small sample size. Retaining variables was an important goal because dropping one variable could alter the estimates of the model (Moestellor and Tukey 1977). Variables that did not have theoretical plausibility, or explained little economic growth, were dropped.[9]

A multivariate empirical model was specified such that:

(2) H = a + [beta]₁MEDHSHDY + [beta]₂POPGROWTH +[beta]₃TRANSPORT

- [beta]₄DEBT + [beta]₅CLEVELAND +[beta]₆ SOUTH - [beta]₇REFERENDA + [epsilon]

where the variables and their expected signs are described in table 1 and [epsilon] represents error in the model. Descriptive statistics for each of the continuous variables are provided in table 2.

The dependent variable, H, represents housing-unit growth in each municipal jurisdiction and is used as a measure of land development in each jurisdiction.[10] While land development and urban economic growth are not necessarily the same, the focus of this paper is on the impact of uncertainty on land development. The link between housing development and economic growth is fairly straightforward: more housing expands a community's capacity to integrate larger populations and encourages more diverse investments in land (e.g., expansion of commercial and retail property).[11] Housing-unit growth was measured over three different periods to test the robustness of the estimates: 1980 to 1990, 1980 to 1994, and 1984 to 1994.

Source: Data for median household income and population growth from U.S. Bureau of the Census, 1990. Data for transportation spending and debt from City Government Finances in Ohio (Columbus, OH: Ohio Public Expenditure Council) 1994). Data on referenda and initiatives from Ohio Secretary of State.

Median household income is expected to exert a positive influence on building permits since higher incomes signal a positive amenity value as new housing is attracted to higher income communities. Population growth should also have a positive influence on building activity since it reflects the demand for land in the community. Transportation spending should positively impact building activity since it reflects an community's investment in physical infrastructure -- roads and sewers -- critical to property development. Debt, in contrast, should reduce building activity for two reasons: a) the community may be unwilling to take on new projects to facilitate land development because it already services debt and b) higher debt may imply higher future taxes and a less attractive community.

Regional factors that could impact growth in a particular city are controlled for using a dummy variable for whether the city was in the Cuyahoga County (CLEVELAND) or in Southwestern Ohio (SOUTH).

Measuring the effects of ballot-box zoning

A dummy variable (REFERENDA) rather than a continuous variable was used to capture the impact of citizen initiatives on uncertainty in the zoning process for several reasons. First, a relatively small number of cities experienced very high levels (ten or more referenda) of citizen activity on zoning issues. Thus, a small number of very active cities could distort the impacts of ballot box zoning on economic growth for most of the sample.

Second, the results of a statistical analysis using a simple count of the number of referenda and initiatives in each city was difficult to interpret. Strong negative relationships were found between referenda and building-permit activity regardless of whether citizens passed or rejected the initiatives. For example, a GLS model for housing-unit growth from 1980 to 1990 found the referenda variable significant at the 99 percent level of significance, regardless of whether referenda were approved ([beta] = -264.06; t statistic = -4.859) or rejected ([beta] = -279.35; t statistic = -5.338).

At first glance, this result appears paradoxical: high pass rates on referenda that accommodate development would be expected to neutralize the negative impact of ballot-box zoning on economic growth or have a positive impact. This paradox is resolved using a transaction-cost approach to urban planning and development. Developers could be reacting to the uncertainty (and higher transaction costs) surrounding potential delays for their projects in more highly politicized environments, not the likelihood a community will approve their zoning request once it is on the ballot. The mere fact citizens are willing to challenge zoning decisions at the ballot box creates uncertainty for property owners and developers. Thus, even though voters may eventually vote to accommodate growth, the project is delayed through the referendum process. The historical presence of citizen referenda or initiatives appears to signal to developers a higher likelihood of delay and level of uncertainty compared to cities without a history of ballot-box zoning.

The use of the dummy variable implies that the politicized nature of the local planning environment (rather than pass or rejection rates) has an impact on the level of economic activity in a city. A continuous variable (e.g., the number of referenda) would imply that marginal unit changes in the number of referenda impacts the level of economic development activity.

Overview of zoning referenda

Data on zoning referenda and initiatives between 1984 and 1994 were obtained from the Ohio Secretary of State for 63 cities in five major urban counties in Ohio.[12] Twenty-two of the cities (34.9%) in the sample reported zoning-related referenda between 1984 and 1994 (table 3). About one quarter of the cities passed or rejected zoning-related referenda. Thirteen cites (60 percent of the cities with referenda on local ballots) reported more than one public initiative. Of those cities with one or more zoning-related issues on the ballot, the pass rate averaged 54.8 percent. Thus, when given a choice, voters tended to approve zoning changes. Eight cities experienced multiple referenda where some passed and some failed. Seven cities reported more than ten zoning referenda during this period. These were typically cities with statutory requirements for voter approval on zoning decisions.

An important limitation of the model is specification bias. While attempts were made to test the validity of the variables, some independent variables were not incorporated because of high levels of multicollinearity or the lack of data. For example, a potentially important determinant of building permit activity might be development patterns within a community. An attempt to gather these data from local communities did not generate useable information for a large enough number of cities to be incorporated into this analysis. Similarly, general community attitudes toward economic growth, particularly city council and planning board members, help determine the overall climate for growth. These data were also unavailable. (In this case, however, data on zoning referenda provide a measure of local concern over growth.) Other factors that could drive local economic growth might include annexation activity, the quality of local education, and the pace of suburbanization within the metropolitan area. This model, however, is not intended to specify a general model of urban development. Rather, the focus is on the importance of planning-related variables on development activity within cities, controlling for as many external factors as practicable.[13]

IV. Empirical results

The results of the GLS estimation[14] of the impact of zoning-related referenda (using the dummy variable) are reported in table 4. The model captures 80 percent of the variation in the growth in housing units among the cities in the sample and the results are robust across time periods. Median household income and transportation spending had substantial impacts on the growth of housing units. For the equation estimating housing unit growth from 1980 to 1990, a $1,000 increase in median household income was associated with an increase of 6.7 housing units per 1,000 population. An increase in $1 per capita in transportation spending increased housing unit growth by about 11.7 units per 1,000 population. Whether the city was located in the Cuyahoga County did not influence the level of housing unit growth during any of the time periods. A city's location in either of the two southern counties (Hamilton or Montgomery), however, appeared to negatively impact growth in housing units.[15]

The presence of zoning-related referenda on the ballot had a consistent, negative impact on housing unite growth in the cities examined. Communities subjecting zoning-related issues to public vote appear to experienced an annual "growth penalty" of 19.4 to 28.7 housing permits for every 1,000. If the penalty were adjusted for the size of the city, the impact of referenda is relatively large. The average city in the sample had a population of 50,000 and annual housing growth of about 321 units from 1980 to 1994, or 89.09 units per 1,000 population. Using the regression estimates for this 14-year period, the model estimated a city of 50,000 would suffer a penalty of -271 housing units over the period, or 19.4 units each year. (Alternatively, a city that began to subject its land-use decisions to public referenda would expect to experience a growth penalty of the same magnitude.) This is more than six times the estimated increase of 2.93 units each year, and three times the actual average increase of 6.36 units each year.

The gap between the model's predicted net growth (-3.74 permits per 1,000 population) and the actual net growth (2.69 permits per 1,000 population) undermines the credibility of the model, however. The model should predict, with a fair level of precision, outcomes derived from the data it manipulates.

Notes: Model transformed by assigning transportation spending to the P Matrix. Significance levels (using two-tail tests): *** for 98%, ** for 95%, and * for 10%. F values for model F(7,55).

Two potential explanations exist for the divergence between the predicted and actual values of the analysis. First, multicollinearity may prevent a precise interpretation of the regression coefficients. This problem could be mitigated by expanding the dataset to include more cities or transforming the model in other ways. As previous sections explained, the GLS estimating technique was adopted to minimize the effects of heteroskedasticity, and models using other transformations (e.g., logarithmic models) did not perform as well as the GLS models. Second, The sample means may not accurately represent the cities in the sample. In this case, outliers may skew the average so that the sample mean does not accurately represent the cities in the sample population. Given the large standard deviations (see table 2), this is likely.

This latter problem can be investigated further by using cases from the dataset. Estimates were calculated for three cities: Cleveland Heights, Worthington, and Kettering. Cleveland Heights represents a city near the statistical mean in the region with the largest number of cities (Cuyahoga County). Worthington is a city smaller than the statistical mean in a growing region (Franklin County) and was selected as an example of a city with a history of zoning-related referenda (neither Cleveland Heights nor Kettering had a history of referenda on land-use issues). Kettering is a city larger than the statistical mean outside Cuyahoga County (Cleveland) and Franklin County (Columbus). The results of the comparison are in table 5.

The standard deviation is a common statistical measure of dispersion, and we would expect the results from the model to be within one standard deviation of the mean. The standard deviation for annual growth in housing units per 1,000 population over this period was 9.3. Predicted annual housing unit growth was within one standard deviation for the sample mean, Cleveland Heights, and Kettering. The predicted growth in housing deviated by more than one standard deviation for Worthington, and in this case the model predicted higher economic growth. The estimates for the ballot-box "growth penalty," then, do not appear to impose a downward bias in the model's ability to predict overall economic growth in this community.

V. Policy implications

This paper tested two hypotheses about planning and urban development derived from a transaction-cost approach: 1) Higher planning-related transaction costs discourage the level of land development and economic growth, and 2) uncertainty in the planning process can drive up transaction costs and reduce economic growth.

Note: * indicates statistically significant variable in comprehensive multivariate model.

An analysis of 63 Ohio cities found that cities that subjected land-use decisions to public referenda were likely to experience lower levels of building activity. This supports the hypothesis that ballot-box zoning injects an element of uncertainty into the development approval process, increases the transaction costs associated with land development, and reduces the level of land development in the community. These results were robust over several time periods.

Importantly, the estimated impacts of referenda on housing unit growth were significant and negative regardless of whether the community voted in favor of the development or against it. This supports the transaction cost interpretation of the impact of ballot-box zoning on land development decisions. The mere fact citizens are willing to place land use decisions on the ballot appears to be a signal to developers that their projects will be subject to higher levels of uncertainty and delay compared to cities that resolve land-use decisions through a legislative or administrative process.

The results have implications for communities that adopt widespread citizen participation as part of a broader political agenda. To the extent referenda reflect community values (e.g., a more open planning process), the higher transaction costs associated with the process will likely translate into reduced economic growth. This is evident in cities where rezoning cases are subject to automatic referenda through local ordinances. Thus, communities that value principles other than efficiency may experience a growth penalty.

The results, however, are not definitive. Further research should focus on testing for the effects of public referenda on larger numbers of cities. In particular, cross-jurisdictional effects could be explored by analyzing housing unit growth rates in neighboring jurisdictions. Regulatory uncertainty in one jurisdiction, for example, could push development into the hinterlands, encouraging urban sprawl. An analysis of cities in different states may yield different results based on alternative institutional arrangements and state and local land use laws. Finally, future research should investigate whether referenda have differential impacts in different types of communities (e.g., rural vs. suburbs vs. central city).

References

Allom, Jamieson. 1991. Delays in Development Approvals Mean Higher Costs.

Australian Property News 3 October: 16.

Altschuler, Alan A., Jose A. Gomez-Ibanez and Arnold M. Howitt. 1993. Regulating for Revenue: The Political Economy of Land use Exactions. Washington, D.C.: Brookings Institution and Lincoln Institute for Land Policy.

Blair, John P. and Timothy K. Kinsella. 1991. Regional Patterns in Manufacturing Development Policy for Ohio Metropolitan Areas. In Manufacturing Development Policy: Economic Restructuring in Ohio, ed. John P. Blair and Keith Ewald, 283-95. Dayton, Ohio: Wright State University Press.

Caves, Roger W. 1992. Land-Use Planning: The Ballot Box Revolution. Newbury Park: Sage publications.

Coase, Ronald H. 1937. The Nature of the Firm. Economica, new series 4: 386-405.

Coase, Ronald. 1960. The Problem of Social Cost. Journal of Law and Economics 3, no. 2 (October): 1-40.

Easton, Stephen T. and Michael A. Walker. 1997. Income Growth and Economic Freedom. American Economic Review 87, no. 2 (May): 328-32.

Evans, Alan W. 1992. Town Planning and the Supply of Housing. In The State of the Economy: 1992, ed. Giles Keating, Peter Warburton, et al., 81-93. London: Institute for Economic Affairs.

Gwartney, James, Robert Lawson and Walter Block. 1996. Economic Freedom of the World: 1975-1995. Vancouver, BC: Fraser Institute.

Hu, Teh-wei. 1982. Econometrics: An Introductory Analysis, 2nd Ed. Baltimore: University Park Press.

Lai, Lawrence Wai-chung. 1996a. Zoning and Property Rights: A Hong Kong Case Study. Hong Kong: Hong Kong University Press.

Lai, Lawrence Wai-chung. 1996b. Rejoinder to Bristow. Planning and Development 12, no. 1: 62-3.

Lai, Lawrence Wai-chung. 1997. Property Rights Justifications for Planning and a Theory of Zoning. In Progress in Planning, forthcoming.

Maddala, G.S. 1983. Limited-Dependent and Qualitative Variables in Econometrics. New York: Cambridge University Press.

Mayo, Stephen and Stephen Shephard. 1991. Housing Supply and the Effects of Stochastic Development Control. Oberlin College Discussion Paper in Economics, 20 August.

North, Douglass C. 1990. Institutions, Institutional Change and Economic Performance. New York: Cambridge University Press.

Staley, Samuel R. 1994. Planning Rules and Urban Economic Performance: The Case of Hong Kong. Hong Kong: Chinese University Press.

Staley, Samuel R. 1997. Urban Planning and Economic Development: A Transaction Cost Approach. Ann Arbor, MI: UMI Dissertation Services.

Titman, Sheridan. 1985. Urban Land Prices Under Uncertainty. American Economic Review 75, no. 3 (June): 505-14.

End Notes [1]Generally, economies that have facilitated market exchange have grown faster than those that have superseded market transactions through economic planning (Easton and Walker 1997; Gwartney, Lawson and Block 1996). The critical dimension of this process is voluntary exchange secured through contract. Thus, North argues that the security of contract is critical to the successful development of an economy. More importantly, from a public policy perspective, government regulation that facilitates contracts will stimulate economic growth, while those that destabilize market transactions will discourage economic growth.

[2]Because most local planning codes in the U.S. presume development and rezoning applications are refused, the governance structure imposes high transaction costs since developers must convince a public body or agency that their development project benefits the community. Often, this requires public hearings at at least two levels (city council and planning board) and sometimes three (city council, planning board, and neighborhood association). Given the comprehensive impact planning has on land development, public policy is thus critical in the land development process.

[3] In some cases, cities require all rezoning decisions and/or "upzonings" (i.e., rezoning commercial to residential, multi-family residential to single family residential, or industrial to residential) to public votes or referenda. Citizens can also petition city councils to place zoning and planning decisions on the ballot by initiative. In practice, referenda are more likely than initiatives. While important legal distinctions exist between initiatives and referenda, the terms are used interchangeably in this analysis. Legally, referenda prevent a legislative action from being implemented. Initiatives install or repeal existing ordinances.

[4] A direct indicator of uncertainty would be a measure of the probability plan or zoning applications would rejected (holding the quality of projects constant). These data, however, are only available at the city level and require extensive, detailed knowledge of plan applications and deliberations over an extended period of time. A survey was sent to city planning departments to gather these data, but response rates were sufficiently low on this portion of the survey they could not be used in the empirical analysis. Moreover, the nature of the plan approval process suggests application rejections may not be the most appropriate measure of uncertainty. The attitudes of planning boards and local citizens, and the nature of objections during the public-hearing process, were also important factors that determined how quickly and decisively a local planning board (and city council) might act.

[5]Cities provide data on building permit activity while villages and townships often allow counties to track this information. County data is not consistently broken out by political subdivision.

[6]In almost every case, the regression models suffered from heteroskedasticity and multicollinearity. Heteroskedasticity is particularly important because its presence increases the standard error (Griffiths, Hill and Judge 1993; Kmenta 1988; Hu 1982 ), thus reducing the likelihood a variable would be statistically significant in the regression. Maddala (1983, 178-9) warns that estimates from truncated regression models with heteroskedasticity present tend to be both inefficient and inconsistent. Since the regression models test the usefulness of a paradigm in explaining development activity, the analytical emphasis is on structural interpretation rather than forecasting and statistical significance is important.

[7] Each model was transformed to correct for heteroskedasticity by assigning an independent variable to the P Matrix for the transformation (Griffiths, Hill and Judge 1993, 502-3). Several models were specified to determine the most robust and consistent estimates. The variable most strongly associated with heteroskedasticity was assigned to the P Matrix to transform the models.

[8] Cincinnati is also a border city, but it is separated from Kentucky by the Ohio River, a formidable natural barrier in the context of land development.

[9]Correlation matrices were used to identify the independent variables most highly correlated with the dependent variable, but least correlated with other independent variables. Multicollinearity was also diagnosed by examining the performance of the models during regression runs. Typically, if two highly correlated independent variables were used, their performance within the model was poor. In general, independent variables highly correlated with other independent variables were dropped from the model to preserve degrees of freedom unless strong theoretical reasons justified their continued use in the model.

[10]Building permit data were available only for a four-year period, 1990 through 1994. Given the availability of data, the longer time horizon was chosen as a more suitable test of the impact of referenda on economic growth.

[11] An alternative measure of economic growth might include per capita income although the link to land use policy would be more tenuous theoretically. Moreover, the focus of this paper is on land development as one indicator of economic growth, not identifying the determinants of urban development more generally. Data limitations would have also restricted the empirical test to one decade. Housing unit growth, on the other hand, allowed for a test over several time periods.

[12]Referenda and initiatives were classified according to whether local citizens approved a change in land-use or whether they overturned an approval by the local city council. A simple count of the number of zoning decisions challenged would be inappropriate. Referenda could address either a decision to approve or disapprove a zoning or planning-related decision by the local city council. Zoning and planning related referenda and initiatives that did not involve a rezoning of property (i.e., updates to zoning codes or passage of a zoning ordinance) were excluded from the analysis.

[13] The legislative nature of the approval process creates transaction costs regardless of community concerns. By establishing a bargaining relationship between local officials and developers, the process creates uncertainty and a bias toward higher costs. The case studies revealed, for example, that staff typically add conditions to development applications and developers attempt to meet concerns voiced by a relatively small number of citizens at public hearings. Both tendencies drive up costs.

[14]Preliminary OLS estimations of the model revealed heteroskedasticity.

[15]In a model not presented here, a dummy variable for Columbus was also included but was not Regional factors that could impact growth in a particular city are controlled for using a dummy variable for whether the city was in the Cuyahoga County (CLEVELAND) or in Southwestern Ohio (SOUTH).