On the determinants of subnational tax progressivity in the U.S.

Differences across states in tax progressivity are substantial and have important effects on the equity of state and local finance. This paper presents an empirical model to explain these differences. Incidence choices are assumed to be influenced by the ability to export tax burdens through the deductibility of state and local taxes, the tax behavior of geographic neighbors, the pattern of expenditure redistribution, the political structure, and the level and distribution of income in a state. The model is estimated using effective tax rate data by quintile for 1977, 1985, and 1991, and instrumental variables are used to remove the potential bias in both deductibility and neighbor effects. The most important influences come from tax exporting through deductibility, neighbor state behavior, and political structure. The strong positive effect of deductibility suggests that eliminating or curtailing the deductibility of state and local taxes would substantially reduce the progressivity of subnational tax systems. In contrast to recent literature suggesting the importance of interstate tax competition through tax mimicking, I find a negative neighbor effect, with more progressive states geographically contiguous with more regressive states. This result suggests that, despite the contention that capital has grown more mobile, there remains considerable scope for different distributional choices in a federalist system. Party control by Republicans, as opposed to a divided government, leads to a more regressive tax structure. Measures of expenditure progressivity and tax progressivity are not closely related. For example, a greater taste for welfare is associated with proportional increases in tax burdens for all income quintiles, rather than a change in tax progressivity. There is a small offset to greater inequality of the pre-tax income distribution via increased progressivity of taxes. In future work, it would be useful to investigate more deeply the role of horizontal and vertical tax competition in determining state and local tax incidence, and the complementarity or substitutability of tax and expenditure incidence.

INTRODUCTION

Despite the standard fiscal federalist prescription that state and local governments cannot engage in redistribution, there is substantial variation across states in the distributional incidence of both taxes and spending. While overall state-local tax systems are regressive, in 1991 the most progressive state-local tax system was more than three times as progressive as the least. The standard deviation of progressivity was equal to a third of the mean (Citizens for Tax Justice (CTJ), 1991). In this paper, I estimate an empirical model to determine the role of various economic and political factors in the choice of tax progressivity across states. Factors include tax exportation through federal tax deductibility, interstate tax competition, the distributional incidence of state and local expenditures, the level and distribution of income, and the political party composition of states. The model is estimated using state and local incidence data for 1976, 1985, and 1991.

State and local taxes are important in magnitude--in 1995 they equaled 11.3 percent of personal income. Hence, variations in their incidence will have a significant impact on the overall distributional incidence of the public sector. However, most research on tax incidence has focused on either measurement or normative aspects. There has been relatively little positive analysis of subnational progressivity in federalist systems. Explaining this variation is important to our understanding of fiscal federalism and, in particular, the equity implications of further devolution of fiscal responsibilities from the national to the state level.

The paper is organized as follows. The first section briefly discusses previous literature. The second section presents the empirical model. The third section discusses data and estimation issues. Results are presented in the fourth section, which is followed by a brief conclusion.

LITERATURE REVIEW

The political science literature contains several studies of state and local tax progressivity. Jacobs and Waldman (1983) find that greater income inequality leads to more progressive tax systems, while a higher percentage of blacks is associated with more regressivity. Morgan (1995) finds that a better informed electorate, as measured by newspaper circulation, has a negative effect on tax progressivity. Berch (1995) shows that the number of years of democratic control of the governorship has a positive effect on progressivity, while Lowery (1987) finds that the degree of political competition is positively associated with the degree of tax progressivity. Other studies by political scientists are discussed below.

Economists have focused on tax exporting in explaining variations in time shares of various state taxes in state revenues (Gade and Adkins, 1990; Metcalf, 1993). Their work is relevant to the analysis of tax incidence because of the strong correlation between tax shares and tax incidence. (1) Metcalf (1993) finds that tax exporting through the deductibility of state and local taxes from the federal income tax is important in explaining both the share and level of income taxation, but it does not explain the share and level of the sales tax. In his work, the decisive voter group for the income tax is estimated to be above the 90th percenthe of the income distribution, while the decisive group for the sales tax is closer to the middle. Metcalf's (1993) results suggest that the imposition of relatively progressive state taxes is heavily dependent on the ability to shift burdens via deductibility. Chemick (1992) analyzes incidence directly, but based on only a single year of data. Bahl, Martinez-Vazquez and Wallace (2002) investigate the relationship between progressivity on the tax side, as measured by reliance on the income tax, and the expenditure side, and find the two to be complementary. The Bahl et al. (2002) study is discussed further below.

MODELING STATE AND LOCAL TAX INCIDENCE

Descriptive Data on State and Local Tax Incidence

Table 1 provides summary data on the tax incidence in U.S. states, both overall and by region. As a measure of incidence, I use the ratio of the tax burden, gross of federal deductibility, in the highest quinthe of a state's income distribution to the tax burden the lowest quinthe. This measure is discussed extensively below. The ratio was 0.66 in 1977, 0.69 in 1985 and 1981, and 0.71 in 1995. Thus, in a typical state, the top quinthe of taxpayers paid about 30 percent less in state-local taxes as a fraction of their income than the bottom quinthe. The coefficient of variation in progressivity was substantial, equaling 12 percent in 1976, 29 percent in 1985, 31 percent in 1991, and 25 percent in 1995. Time South has the most regressive tax systems and the least variation across states.

Figure 1 displays the geographic patterns of average progressivity in each state over the period 1977 to 1995. The map shows that regressive states are found in each region of the country, as are relatively progressive states. In some instances neighboring states have similar patterns of incidence, while in other cases highly progressive states border highly regressive states. The empirical model will include a test for the spatial relationship between geographic neighbors.

[FIGURE 1 OMITTED]

Empirical Model

In their choice of state and local taxes, politicians are assumed to choose an incidence pattern that minimizes the political costs of raising a given amount of revenue. (2) In equilibrium, the marginal cost of raising the tax rate on a given income group will be equated to the additional revenue from the tax increase. The additional revenue depends on the size of the tax base and on the elasticity of the base with respect to the tax rate. Variables in the model are interpreted as affecting either the political costs of taxation or the elasticities of the tax base for different income groups.

States vary substantially in the relative importance of state versus local taxes, and the incidence pattern is in part a reflection of this choice. (3) Relatively progressive systems are usually associated with reliance on a broad-based graduated state income tax, which is likely to increase the state share of taxes relative to the local share. I do not model the state versus local decision per se. Instead, the model explains the aggregate incidence pattern for all state and local taxes, though variables influencing the aggregate pattern at the same time affect the state-local choice. (4)

The empirical model, with the expected effect on progressivity in parentheses, can be summarized as follows:

[1] PROGRESSIVITY = F[PCTITEM (+), NEIGHBOR PROGRESSIVITY (+), REPUBLICAN (-), DEMOCRAT (+), TASTEWELF (+), EDINEQUAL (?), INCINEQUAL (+), INCOME (?)]

PCTITEM, the percentage of tax filing units that itemizes, is a measure of the exportability of state and local taxes through federal tax deductibility. Itemization lowers the marginal tax price for deductible taxes from one to one minus the federal marginal tax rate. Because the likelihood of itemizing is strongly correlated with a taxpayer's income, the greater the fraction of taxpayers who itemize, the lower is the marginal tax price for taxpayers with incomes above the median for the state. The reduced tax price reduces the elasticity of the high-income tax base, leading to higher relative tax burdens on high-income taxpayers. (5)