Easy Riding in the Provision of Local Parks and Recreation Services

One of the most important and contentious issues in the theory of public finance concerns the relationship between the structure of local government—that is, the number and size of local jurisdictions—and the level and nature of local government spending on public services. More than thirty years of theoretical work and a number of recent empirical studies suggest that the two are intimately and often surprisingly related. Although no consensus has been reached concerning its precise nature, the relationship between the two is of great policy interest because it is possible that desirable spending outcomes are affected or even prevented by the structure of local governments.

Perhaps the most well-known theory of local government finance is the seminal hypothesis formulated by the economist Charles M. Tiebout. According to the Tiebout hypothesis, individuals would choose to live in a community that provides their preferred level and type of public services if circumstances permit them to move from one community to another. The hypothesis requires that there be fragmented and numerous units of local government within a metropolitan area, each of which offers a different combination of services and taxes among which individuals can choose. When this is the case, Tiebout proposes, individuals would vote with their feet to choose the local community offering a mix of services best matching their preferences, thereby disciplining the spending patterns of local government. In short, the Tiebout hypothesis suggests that spending on public goods would be efficient if local governments were fragmented.

Several hypotheses state the converse of the Tiebout hypothesis—that is, spending outcomes would be inefficient if local governments were not fragmented. Perhaps paramount among these hypotheses is the so-called Leviathan hypothesis, which is articulated most forcefully by economists Geoffrey Brennan and James Buchanan. Brennan and Buchanan argue that, absent the interjurisdictional competition posited by the Tiebout hypothesis, government will act as a Leviathan (a vast bureaucracy) to maximize its share of the economy, and public spending will be inefficiently high. Indeed, a number of empirical studies have found that the level of public spending is greater when local government is less fragmented—that is, when a metropolitan area is represented by relatively few political jurisdictions—than when it is more fragmented. It is worth noting that Tiebout's reasoning could also be invoked to support the opposite of the Leviathan hypothesis—that is, local governments tend to spend too little for local services absent competition to boost spending for such services. Up to the present time, however, the Leviathan hypothesis is the hypothesis that is consistent with empirical studies.

Nonexcludability in the provision of urban parks

The empirical studies referred to above have an important limitation in that they have largely ignored the degree to which different public goods and services are truly "local" goods and services. By definition, local public goods and services are provided to members of a group but are denied to nonmembers. Fire protection, police protection, and education are examples of services for which local authorities can easily deny access to nonresidents of a jurisdiction. To varying degrees, local authorities can ration access to local public goods and services such as libraries and sanitation and community development services and charge a fee to nonresidents of the jurisdiction for the use of such goods and services.

Local authorities have considerably more difficulty excluding nonresidents from other public goods such as environmental quality, roads, and local parks. While access to some recreation goods and services such as swimming pools and sports leagues may be limited to residents of the jurisdiction providing the services, access to most urban parks is not limited. Of course, services like police protection, which we consider a largely excludable service, cannot be denied to nonresidents visiting a community. While nonresidents have to come on site to enjoy the benefits of either local police protection or local parks, they would be much less likely to visit a community in order to enjoy police protection than they would be to visit a community in order to enjoy local recreation facilities. This means that individuals would be more likely to consider parks in neighboring communities to be a substitute for parks in their own communities than they would be to consider police services in neighboring communities to be a substitute for police services in their own communities.

The inability to exclude nonresidents from urban parks and open spaces raises the possibility that the provision of local parks will be inefficiently low. From the perspective of a local community where a park is located, such provision might be efficient when its marginal costs to residents is equal to its marginal benefits, without consideration of benefits accruing to nonresidents. If some of the benefits of expenditures on a park by one community spill over to neighboring communities, each community may have an incentive to consume park services provided by its neighboring communities and to reduce its own park expenditures accordingly. The result of such strategic interaction among communities may be that too few parks are provided from the perspective of the metropolitan area taken as a whole.

This strategic interaction is related to one of the classic problems in public finance—the so-called free rider problem. This problem stems from the insufficient incentive that individual members of a group have to provide goods or services that can be enjoyed by all the members of the group. In the present context, the free rider problem is more accurately labeled the "easy riding" problem because communities would be expected to provide some level of nonexcludable goods, but not as high a level as if they took into consideration the benefits that accrue to everyone within their metropolitan area.

Both the easy riding and Leviathan hypotheses suggest that increased fragmentation of a metropolitan area will lead to reduced provision of public goods. However, the hypotheses differ in their range of application and in their implications. While the Leviathan hypothesis suggests that a less fragmented metropolitan area might provide too much public spending on all public goods—from the viewpoint of economic efficiency—the easy riding hypothesis suggests that a more fragmented area might produce too little public spending on certain types of public goods.

Preliminary analysis of spending by a metropolitan area as a whole

To investigate how local government structure affects spending on nonexcludable public goods, we performed a statistical analysis comparing expenditures for local parks and recreation services—which we characterize as nonexcludable goods—with expenditures on police, fire, and community development services—which we characterize as excludable goods. In the analysis we examined spending in 314 metropolitan statistical areas (MSAs) containing 3,508 municipalities with populations greater than 2,500 in 1982.

As a preliminary exercise, we compared per capita spending aggregated across all levels of local government within an MSA for nonexcludable goods and for our index of excludable goods in a system of two equations. In doing so, we departed from previous studies of the effects of local government structure on local public spending. These studies have measured spending for an undifferentiated bundle of goods, without accounting for the characteristics of different types of goods and for the attribute of excludability, in particular. They have found that this aggregate amount of spending depends negatively on the number of independent political jurisdictions within an MSA—that is, the less fragmented a metropolitan area is, the more it spends on all public goods. This finding is consistent with the Leviathan hypothesis.

Analysis of aggregated data on the overall public spending of an MSA provides little insight into the determinants of the potential strategic interaction among communities with respect to the provision of public goods. Therefore, we used spending patterns by individual municipalities as the basis for most of our research. Specifically, we estimated per capita spending on nonexcludable and excludable goods by individual municipalities with respect to several independent variables, one of which is per capita spending on the same goods by other municipalities within an MSA.

According to the Leviathan hypothesis, an increased level of spending by one jurisdiction signals reduced competition among all the jurisdictions within an MSA. This implies that spending on all types of goods by a given municipality increases as such spending by other municipalities within an MSA increases, and vice versa. In contrast, the easy riding hypothesis predicts that spending for nonexcludable goods by a given municipality decreases as such spending by other municipalities within an MSA increases, and vice versa. Hence, in our estimation of per capita spending on public goods by individual municipalities, the Leviathan and easy riding hypotheses' predictions concerning the effect of other jurisictions' spending for nonexcludable goods on one jurisdiction's spending for such goods are diametrically opposed.

The effect predicted by the Leviathan hypothesis and that predicted by the easy riding hypothesis could both exist. To control for this, we separated the two effects by comparing patterns of spending on nonexcludable goods with patterns of spending on excludable goods within a system of two equations and compared estimated coefficients across the system. In our estimation, coefficient values represent the percentage change in the level of spending by each municipality for the good characterized in each equation when there is a percentage change in each independent variable of interest.

If the easy riding hypothesis is valid, then the coefficient for spending in the equation for nonexcludable goods would be less than the coefficient in the equation for excludable goods. Our analysis showed that spending by each municipality for nonexcludable goods does depend negatively on the spending of other jurisdictions within an MSA and that the coefficient in the equation for nonexcludable goods is less than the coefficient in the equation for excludable goods, as the easy riding hypothesis predicts. A statistical test of the probability that the coefficients in the two equations are equivalent revealed that probability to be less than 15 percent, given our data. This result provides some evidence of the influence of easy riding on the provision of nonexcludable goods. It appears that the spending levels of neighboring jurisdictions affect spending on nonexcludable goods and on excludable goods differently and in the manner suggested by the easy riding hypothesis.

Another variable of interest in our estimation of per capita spending on public goods by individual communities is the geographic size of a jurisdiction. As noted above, the Leviathan hypothesis predicts that spending for all types of goods will increase as the geographic size of a jurisdiction increases. The easy riding hypothesis would seem to predict an increase in spending for nonexcludable goods only. Hence, theory predicts that the coefficient in the equation for nonexcludable goods should be strictly greater than the coefficient in the equation for excludable goods. Our analysis did not find this to be the case.

Several important considerations complicate the estimation of our equations for spending on nonexcludable goods and excludable goods and thus the interpretation of results reported thus far. One is that access to parks and recreation services provided by one jurisdiction can sometimes be denied to residents of neighboring jurisdictions. With some effort, access to recreational activities such as exercise classes and sports leagues can be controlled through enforcement of residency requirements; with considerably more effort, nonresidents could be denied access to local parks altogether. Another consideration is that current public spending on local parks may not be a good measure of the availability of parks. Better measures of this availability might be the acreage of parks in local jurisdictions or the cumulative past investment by local jurisdictions in recreation facilities; however, nationwide data on both park acreage and past recreation investment are difficult to find. Given that local parks and recreation services are not entirely nonexcludable goods and that current public spending on local parks may not be a good measure of the availability of parks in neighboring jurisdictions, the easy riding effect might be stronger than our imperfect measures indicate.

While we were not able to take the above considerations into account in our analysis, we were able to take into account other factors that complicate analyses of how government structure affects local public spending. One of the most important of these factors is expenditures for local public goods by more than one level of government within an MSA. In many cases, county governments or special districts, which may serve part or all of one or several counties within a metropolitan area, provide a vehicle for spending on certain types of local public goods. If easy riding is a widespread phenomenon, one would expect local governments to attempt to mitigate it by coordinating their spending activities through these governments or districts.

While we can separate the amount of spending by a county government from the amount of spending by individual municipalities within a county, we cannot discern whether county governments are providing services only to unincorporated areas, as opposed to the county as a whole—including incorporated municipalities. It is even more difficult to tell precisely which populations or geographic areas within an MSA are served by special districts. To account for this difficulty, we examined a subset of our data on spending by only those metropolitan areas not served by special recreation districts, and separated spending by county governments from spending by municipalities within those areas. Although one might suspect that the previous results of our analysis of per capita spending on nonexcludable and excludable goods by individual municipalities were biased by the spending role played by special districts, we found them to be robust.

Median income and spending for nonexcludable goods

We found that the most important variable affecting per capita spending on parks and recreation services is the median income of households within a municipality. In an early pretest estimation of such spending using a random selection of 120 municipalities, we discovered that, in low-income communities, per capita spending on parks and recreation increases as income increases and that, in high-income communities, per capita spending on parks and recreation decreases as income increases.

We subsequently divided the municipalities according to median income for each part of the county and estimated equations for spending on nonexcludable goods and for spending on excludable goods for the poorest 50 percent of municipalities and for the richest 50 percent of municipalities. We found the difference between low-income municipalities and high-income municipalities to be striking with respect to easy riding. The easy riding effect appeared to be strong among the poorest 50 percent of municipalities, but it disappeared among the richest 50 percent of municipalities.

An important limitation of empirical studies of this type is that typically they cannot confirm or deny theory. Rather, they can at best demonstrate whether the data appear consistent or inconsistent with theory. Previous empirical studies of local government spending have been careful not to make claims about whether patterns exhibited in local public spending reflect optimal or suboptimal levels of spending, and we emphasize the same.

We find evidence consistent with the easy riding hypothesis. All other factors being equal, the easy riding effect should lead to under-provision of nonexcludable local public goods, from the viewpoint of economic efficiency. However, if the effects predicted by the Leviathan hypothesis or other effects are also influencing local public spending, the net effect is difficult to evaluate.

From our perspective, an important implication of our research is the possibility that easy riding affects expenditures for nonexcludable local public goods and services that are less easy to measure than expenditures for parks and recreation services. Quality of the environment in metropolitan areas is the public good that most directly affects people during the vast majority of their daily lives. The factors that determine environmental quality in any one jurisdiction have effects on the environment in neighboring jurisdictions. We are currently extending our research to assess the influence—if any—of easy riding on the design of various local government policies that affect the quality of the environment.

Our research also highlights the distinction between economic efficiency and optimality. We have uncovered evidence that differences in median income among communities has an important bearing on the tendency to easy ride. From the perspective of individuals residing in a relatively wealthy local political jurisdiction, easy riding may be an unwelcome effect. Non-resident users of public facilities contribute to the congestion of public facilities, and they might also spur the substitution of private goods for public goods, as when private recreation clubs and residential lots take the place of local public parks. From a broader perspective, however, the implied income redistribution that might be accomplished through easy riding may be a positive development. The definition of economic efficiency hinges on the definition of the political jurisdiction for allocating resources, and the optimal level of the provision of local public goods ultimately lies in the political domain.