   |
|
|
Back to American Metropolitics |
|
|
Classifying Communities |
|
|
In the course of some 18 metropolitics reports throughout the country it became clear that there were a variety of types of suburbs emerging in US metropolitan regions. A large number of places, inner ring suburbs and older satellite cities for the most part, were beginning to experience rapid social changes, particularly in their school systems, with few local resources available for efforts to deal with the changes. A small subset of these places was more troubled than the central cities they surrounded. Another important group involved a large number of developing communities that were growing very fast without adequate resources for local schools or infrastructure. Finally, there emerged a smaller very affluent group of cities that were gaining all the benefits of a regional economy without assuming the costs. The following section shows the results of an effort to categorize the various types of American communities in the twenty-five largest metropolitan areas on the basis of local resources and needs in a systematic way. The balance between a place's ability to raise revenues and the expenditure costs and needs it faces in providing public services determines how onerous local taxes must be in order to provide any given level of local services. In Chapter 3 we estimated how revenue-raising capacities vary across jurisdictions in the 25 largest metropolitan areas. In Chapter 4 we examined how capacities and needs compare in the 30 large cities in those metropolitan areas. In this chapter we combine the fiscal capacity data with socioeconomic characteristics to develop a typology of suburban places. Since there are more than 4,600 suburban jurisdictions in the 25 metropolitan areas, we cannot compare these measures case by case as we did for the largest cities. Instead we use cluster analysis to group suburban areas according to their characteristics in several dimensions. The intent is to pick dimensions that capture differences in both fiscal settings (revenue capacities and expenditure needs) and socio-political environments. Expenditure needs and the costs of providing services are also determined by several factors. First, how public service responsibilities are divided among the various levels of government (state, regional, county and local) determines the mix of services that a locality must (or can) provide. Second, the characteristics and preferences of local residents and businesses affect the demands that localities face regarding the amount and mix of services to provide. And finally, various characteristics of the service environment faced by localities (such as population density, terrain, and concentrations of high poverty populations) affect the cost of providing services. Ideally, a standardized measure of expenditure needs would be estimated place by place with a model that accounts for all of the factors potentially contributing to demands and costs. This measure could then be compared to the standardized measure of revenue capacity to generate a "need-capacity gap."3 It is not possible to fully model expenditure needs at the local level for his work because of data limitations.4 Instead, a set of local characteristics that research has shown to be important contributors to local costs will be used together with the tax capacity measure to group jurisdictions according to both their capacities and service costs. Cost differences in this context are not determined solely by the prices that local governments face when buying the inputs needed to produce local public goods and services. These kinds of costs are not likely to vary substantially from place to place within a single metropolitan area. Instead costs associated with what is usually called the service environment are most important in this context. The environments in which localities provide local services like police and fire protection, schools, sanitation, streets and transit vary a great deal from place to place within metropolitan areas. Elements of the service environment affect how activities like police patrols, teaching, or construction and maintenance of infrastructure and transit systems translate into the public goods actually consumed by residents - things like freedom from crime, learning, mobility or a healthy living environment. Estimating exactly how characteristics of the local environment impact the costs of providing these consumed goods and services is a difficult enterprise. The most notable attempts either involve only large cities or a single state or metropolitan area.5 However, there is a group of local characteristics that show up most consistently in discussions and empirical analyses of service environment costs. From this group, population density, poverty rate, the age of the housing stock (a proxy for the age of the infrastructure) and population growth were chosen for this analysis.6 The evidence shows that more poverty and older housing stocks result in higher costs per person of providing a wide range of local public goods. For instance, poverty populations are more likely to be victimized by crime and to resort to crime for a livelihood. The costs of limiting crime to a given level are therefore likely to be higher in high poverty environments. Similarly, older housing stocks are likely to be associated with aging, costly to maintain, infrastructure. The relationships of costs with density and population growth are more complicated. With these characteristics, higher costs are likely to be associated with both very low values (very low density or population decline) and very high values (very high density or exploding population). Very low densities can increase per person costs for public goods with a transportation component (such as schools or public safety) as well as for infrastructure (such as roads and sewers). Very high densities on the other hand generate congestion. Population declines tend to increase per person costs of long-lived public goods like infrastructure (streets and sewers) because, in the short run the number of users declines while the supply is fixed. Very large increases in population tend to increase per person costs of the same goods because, for a variety of reasons, the costs of new infrastructure tend to fall disproportionately on current residents (versus future residents). A second advantage of this group of characteristics is that, in addition to the service cost environment, it captures a good cross-section of the socio-economic characteristics that often define political character or logical political coalitions. Poverty, density, housing age, growth characteristics and wealth are the kinds of characteristics that people look to when deciding if another place is like "us". The final important element of this tableau in the United States, of course, is race. In MARC's national study, 4,606 incorporated municipalities and 135 unincorporated areas in the twenty-five metropolitan areas were grouped according to their characteristics in the two broad dimensions - tax capacity and costs. Two tax capacity measures were included - 1998 tax capacity per household and growth in tax capacity from 1993 to 1998. Five characteristics of the service environment were used for the cost measures - percentage of elementary students eligible for free or reduced lunch in 1997, 1998 population density, population growth from 1993 to 1998 and the age of the housing stock in 1990. The proportion of elementary students that were non-Asian minorities in 1997 was also included. This was chosen not as a cost variable per se but rather to reflect a range of factors that work against high-minority communities. Primary among these is the fact these communities tend to be viewed differently from largely white communities in the housing market, depressing growth prospects and housing prices. The variables were computed for each municipality as a percentage of the average value for the municipality's metropolitan area. All of the suburban communities were then divided into groups using cluster analysis, a procedure that divides the observations in a data set (municipalities in this case) into relatively homogeneous groups based on specified characteristics (the two capacity, four cost and one race variables in this case). Central cities were put into their own cluster and not included in the cluster analysis.7 MARC has also used this technique to categorize communities within a single state. These include state studies in California, Ohio, New Jersey, and Pennsylvania. 3 See Ladd and Yinger (1989), America's Ailing Cities, for a full description of the need-capacity gap approach. In the context of that discussion, the tax capacity measure employed here is a variant of the Representative Tax System (RTS) developed by the Advisory Commission on Intergovernmental Relations (ACIR). The primary difference is that this work uses metropolitan-level average tax rates rather than the national averages used by ACIR in their inter-state comparisons. This procedure avoids the pitfalls of applying the RTS to places with access to different mixes of tax bases (discussed by Ladd and Yinger) but limits one's ability to make inter-metropolitan comparisons. 4 Primary among these is the lack of current data on local socioeconomic conditions. When 2000 Census estimates become available it will be possible to model expenditure needs more rigorously. 5 See Ladd and Yinger (1989) for estimates for large cities and Ladd, et al. (1991) for a study of a single state. 6 Other variables often mentioned in this context are total population and job densities (or jobs per capita). Population was not used in this work because it shows such wide variations within metropolitan areas that it would dominate the cluster analysis overwhelming any effects of the other variables. In addition six of the seven variables included in the analysis (all except tax capacity) are significantly correlated with population at the 95% confidence level. Job density was not included in the analysis because adequate data is not available at the local level in most of the 25 metropolitan areas. 7 The procedure used for the grouping was the K-means clustering procedure in SPSS. The K-means procedure is designed to handle large numbers of observations but requires that the number of clusters be specified. Alternative groupings for five, six and seven clusters were compared and the six-cluster run was chosen because it generated groups that could most easily be identified by geographic location within metropolitan areas. In particular the step from five to six clusters is the step that separated low tax capacity outer ring suburbs from low capacity inner ring suburbs (on the basis of the density measure). |
|
|
Back to American Metropolitics |
|
|
©2002-2008 Ameregis Corporation
|
|