Demographic Structure and the James M . Poterba

This article examines the relationship between demographic structure and the level of government spending on K–12 education. Panel data for the states of the United States over the 1960–1990 period suggests that an increase in the fraction of elderly residents in a jurisdiction is associated with a significant reduction in per-child educational spending. This reduction is particularly large when the elderly residents and the school-age population are from different racial groups. Variation in the size of the school-age population does not result in proportionate changes in education spending, thus, students in states with larger school-age populations receive lower per-student spending than those in states with smaller numbers of potential students. These results provide support for models of generational competition in the allocation of public sector resources. They also suggest that the effect of cohort size on governmentmediated transfers must be considered in analyzing how cohort size affects economic well-being. Spending on primary and secondary education is the largest expenditure item in state and local government budgets. Although tax-financed public education may provide some benefits to society and the local community at-large, most of the return to such spending accrues to families with children. The costs of public education nevertheless fall on households with and without school-age children. In particular, older households with owner-occupied homes pay local property taxes as well as state sales and income taxes that ultimately finance K–12 education. These generational differences in the net benefits from publicly provided education can lead to tensions in the political process in which education budgets are set. This article explores the empirical significance of these generational tensions, and presents evidence suggesting that, during the postwar period, a state’s demographic composition has affected the level of per-child education spending. Questions of intergenerational burden-sharing and equity in school finance have received less attention to date than intragenerational issues such as inJournal of Policy Analysis and Management, Vol. 16, No. 1, 48–66 (1997)  1997 by the Association for Public Policy Analysis and Management Published by John Wiley & Sons, Inc. CCC 0276-8739/97/010048-19 Demographic Structure and the Political Economy of Public Education / 49 equality in school district spending levels. The prospective aging of the United States’ population, indicated by the projected growth of the population share aged 65 and over from 12.5 percent in 1990 to 18.7 percent in 2030, may, however, lead to heightened generational tensions and interest in these issues. These issues have already become salient in debates over Social Security, where younger workers are taxed to finance benefits for older retirees. The transfers in the public education system flow in the opposite direction. The analysis presented later in the article relates more generally to the question of how cohort size affects the well-being of cohort members. Traditional analyses of this issue in economic demography focus on the supply and demand for workers, and on the induced rates of return on various assets. Those in a large cohort must supply their labor when aggregate labor supply is high, thus they are predicted to earn lower real wages than those in relatively smaller birth cohorts. Similarly, when those in large cohorts decide to save for retirement, they must compete with many other savers in buying assets, thereby bidding up prices and driving down returns. These two effects combine to reduce the lifetime utility of those in large birth cohorts relative to those in smaller cohorts. This traditional analysis neglects the potentially important role of government transfers in altering the intercohort distribution of resources. Because the level and direction of government-mediated transfers reflect in part the relative political powers of different cohorts, those in small cohorts may receive smaller net transfers than those in larger cohorts. Preston [1984] suggests that such generational competition is part of the explanation for the relative improvement in the economic status of elderly households in the United States, and the decline in the well-being of children, during the 1960–1980 period. This hypothesis is confirmed by at least anecdotal evidence on public support for various government programs. Recent survey results suggest that support for increased federal funding of public schools declines from 77 percent if the respondent is under 30, to 47 percent for those over age 70 [Koretz, 1995]. The theoretical relationship between demographic structure and the agespecific pattern of government spending is complex, in part because the agespecific benefits of various government programs may be difficult to assess. Richman and Stagner [1986], for example, develop an alternative to Preston’s [1984] hypothesis that a rising number of elderly households will lead to greater government transfer flows toward this group. They suggest that rising numbers of dependent elderly may seek to raise the training of younger workers, both to raise the pool of resources from which transfers can be funded and to raise the quality of services they receive. Such alternative hypotheses make the relationship between the age structure of the population, and the level of agespecific transfers, an issue for empirical research. The present article explores the link between cohort size and spending patterns by focusing on the relationship between state and local education spending per child and three demographic variables: the share of the population over age 65; the share of the population of school age (5 through 17 years); and the difference in the racial composition of the elderly and school-age populations. The article is organized as follows: The article begins with a summary of previous work on demographic structure and public spending determination, with particular attention to studies of public education. Next, the article presents the econometric specification that provides the basis for this study, and summarizes the demographic variation across states of the U.S. 50 / Demographic Structure and the Political Economy of Public Education Table 1. Relative per capita social spending by age group, 1980. Relative spending (age 0–14 5 100) Population share (%) Country Ages 15–64 Ages 651 aged 651 Canada 72 265 9.5 France 51 263 14.0 Germany 60 316 15.5 Italy 110 380 13.5 Japan 44 235 9.1 Sweden 43 234 16.3 United Kingdom 54 213 14.9 United States 67 381 11.3 Sources: Columns one and two are drawn from OECD [1988a, Table 18]. Column three is from OECD [1988b, Table 3.1]. Then there is a discussion of the regression evidence on the association between demographic structure and education spending, along with ‘‘control equations’’ that relate state and local spending for activities other than K–12 education to the same set of demographic variables. A brief conclusion suggests several limitations of the current analysis, as well as directions for future work. GOVERNMENT-MEDIATED REDISTRIBUTION ACROSS GENERATIONS Government-provided retirement pensions, tax-financed national health care systems that transfer resources from those in the labor force to older individuals, and publicly provided education are the most prominent examples of government’s substantial role in intergenerational redistribution. The relative levels of these various transfers differ substantially across countries. Table 1 presents a cross-national comparison of per capita social spending on children, middle-aged individuals, and the elderly in several large Organization for Economic Cooperation and Development (OECD) nations. Per-child spending has been normalized to 100 for each nation. Table 1 shows that spending on the elderly exceeds that on children in all of the countries, but that the ratio of spending per elderly individual to spending per child varies from 2.3 in Japan and Sweden to 3.8 in Italy and the United States. The cross-national data do not suggest an obvious relationship between the share of the elderly in the population and the share of government spending devoted to the elderly or to children. Simple statistics such as those in Table 1 are virtually impossible to interpret, however, in the absence of other information on the relative incomes of different cohorts, the costs of providing services in different countries, and the nature of transfers that are not mediated by the government. Nevertheless, the data suggest substantial differences across nations in the government transfer component of the bargain that is struck between generations. 1 Further discussion of international comparisons may be found in O’Higgins [1988] and Smeeding, Torrey, and Rein [1988]. For a historical perspective on the growth of social spending in several OECD nations, with attention to demographic factors, see Lindert [1996]. Demographic Structure and the Political Economy of Public Education / 51 Searching for a link between demographic variables and spending outcomes raises an obvious question: How should potential demographic effects be modeled? Because there is no universally accepted model of political equilibrium, it is difficult to invoke a maintained model of generational bargaining as a basis for developing econometric specifications for public spending on education or other directed expenditure programs that accrue to particular age groups. Different models can yield different answers to questions of the form ‘‘How will government spending on children change if the share of older individuals in the population increases?’’ In a median-voter framework, the answer to this question depends on whether the demographic shift changes the identity of the median voter. It is possible to shift the population age structure without changing the identity of the median voter, and therefore without affecting the political equilibrium or the age-specific structure of government transfer payments. In other models of political equilibrium, the relative size of different voting groups may affect spending outcomes. In this setting, a natural role for empirical work is to search for robust relationships between various demographic variables and the level of education spending. The political dynamic between different generations reduces to a simple transfer game when there are no cross-generational effects of directed spending. Whether a larger generation will succeed in transferring resources from smaller generations to itself depends, in this case, on the structure of the political process, and the degree to which protection against such transfers has been afforded to minority groups. This simple framework ignores cross-generational effects, however, which may be significant and which can make clear predictions very difficult. If individuals are altruistically linked to their parents and their children, then spending that raises the utility of the old will also raise the utility of the young, and vice versa. Even without such altruistic links, there may be cross-cohort direct utility effects associated with government spending. Spending on public education, for example, may reduce the incidence of crime and thereby raise the utility of the old, while also raising the utility of the school-age children who receive this education, as well as their parents who are not required to arrange for private education. Such cross-effects make it difficult to build a satisfactory model of the political process that determines school spending. In addition to generational affiliation, many other individual characteristics may determine the value of goods that benefit particular population subgroups. Cutler, Elmendorf, and Zeckhauser [1993], who label such goods ‘‘directed 2 The notion that different groups within society compete for resources is the centerpiece of many models of distributive politics, such as that developed by Weingast, Shepsle, and Johnsen [1981]. Inman [1987] summarizes the many possible models that can be considered in analyzing the political process that determines policy outcomes. The special case of intergenerational bargaining involves particularly pronounced differences in the objectives of different participants in the political process, with age the primary determinant of interest group affiliation. Inman [1978] explicitly models demographic variables as shifting the median position and hence the behavior of the public sector. 3 Logan and Spitze [1995] present evidence of substantial intergenerational altruism with respect