Abstract
The literature on the tragedy of the commons is voluminous, and application to the fiscal commons is well established. In this article, we extend this application by examining the effects of the distribution of state and local governments’ tax liability on budgetary outcomes. Different tax structures yield vastly different contributions from members of the polity, but all members may influence the draw from the fiscal commons through the political process. We find that when the tax burden is heavier for taxpayers at the top of the income distribution and lighter for taxpayers at the bottom of the income distribution, state and local government expenditures grow, and governments spend more on social welfare. However, we do not find a link between the distribution of tax liability and debt.
Concern and continuing debate over federal budgetary outcomes—from spending, deficits, and debt to the escalating entitlement expenditures that fuel them—have mounted for years and seem to be reaching an all-time high. Because of this concern over the federal budget, there has been a tendency to overlook the budgets of state and local governments. Nevertheless, in the 2012 fiscal year, state and local governments took in and spent $3.1 trillion and were in debt $2.8 trillion (US Government Spending 2012). By comparison, the federal government spent $3.8 trillion and was in debt $15.4 trillion. Obviously, state and local government balance-budget constraints have a sizable impact on the size of the related debt.
The magnitude of spending by state and local governments—almost as large as the federal government’s spending—invites investigation into not only the size but also the determinants of expenditure. And, the laboratory of states provides an opportunity to observe and learn about actions that could be compared with and applied to federal policies.
This article contributes to the literature on public sector revenues and spending by considering a critically important factor that affects budgetary outcomes: the distribution of the tax liability across the population. Put another way, the article addresses the question, Does the ability to shift tax liabilities for state and local government services from lower- to higher-income groups leads to relatively larger or smaller expenditures on those services? Or, does it matter at all? Using data from the Institute on Taxation and Economic Policy (ITEP), we calculate the share of state and local government tax liability for different segments of the income distribution for each state and the District of Columbia. We find that tax systems that shift the burden to high-income taxpayers and away from low-income taxpayers result in higher state and local government spending as a share of total state personal income. The composition of this spending is affected too, in that state and local governments that shift the tax burden in this way spend relatively more on social welfare. In other words, when the tax price for receiving welfare transfers goes down, recipients of these benefits apparently lobby successfully for more payments. When we consider state and local government debt, however, we find no statistically significant linkage to the distribution of the tax liability.
Our article is organized as follows. First, using a fiscal commons metaphor where, given existing tax systems, political agents compete to draw revenues from a common pool, we review the literature on the fiscal commons and institutions that govern it and then briefly present the simple analytics on how the distribution of the tax liability affects fiscal outcomes. We then turn to the data and the empirical models and estimates that test the link between the distribution of tax liability and budgetary outcomes. After offering some concluding thoughts on extensions of this body of research, we briefly consider the policy implications of our results.
The Fiscal Commons: An Institutional Overview
Hardin’s (1968) pathbreaking article in Science identifies the institutional failure that leads to tragedies of the commons: overconsumption due to unrationed access to a common pool resource. The overgrazed pasture, where each shepherd has an incentive to bring one more sheep to the common pasture, is the metaphor for this problem. In a prompt follow-up in the same journal, Crowe (1969) casts strong doubt on the conditions Hardin recommends for avoiding destructive overuse of a commons, specifically commonly held values by resource users, mutual acceptance of restraints on behavior, and effective administration. He warns that if tragedies of the commons of all stripes are to be averted, the natural and social sciences cannot be insulated from one another. Crowe’s warning has been heeded, as shown, for example, by Anderson and Leal (2001) and Ostrom’s (1990) landmark work.
In work traversing into the political sphere, Weingast, Shepsle, and Johnsen (1981) apply the concept of the tragedy of the commons (if not the term) to the size of political projects. In their model, geographic political districts reap concentrated benefits from projects financed by other districts. This tendency for concentrated benefits to trump dispersed costs leads to large, inefficient government projects, a tendency exacerbated by legislators with understandably more regard for serving their constituencies and getting reelected than for overall efficiency. In a similar vein, Bradbury and Stephenson (2009) find empirical support for the “law of 1/n” when they examine each state’s share of federal personal income taxes paid and federal parochial spending in that state. Predictably, the higher the tax share a state’s citizens pay, the lower the per capita parochial spending that state receives.
Turning from political districts and states to interest groups, Velasco (2000) contrasts the fiscal outcomes of a benevolent planner with those that result when government assets are shared as a commons. He finds that the efficiency maximizing outcomes of a benevolent planner are not obtained under the common property scenario. Instead, the results—sustained deficits and broached debt ceilings—are “logically quite similar to the ‘tragedy of the commons’” (p. 108).
The case for the fiscal commons is perhaps most clearly and plainly articulated by Wagner (1992) and Brubaker (1997). Wagner explains unrestrained federal spending and chronic deficits as the logical outcome of “common property budgeting, where choice is divorced from responsibility for the consequences for those choices” (p. 107). Brubaker argues that fiscal processes—from bureaucratic incentives to the rational ignorance of voters—offer little hope to restrain the looming tragedy.
Jakee and Turner (2002) advance the discussion to the modern welfare state, arguing the problem is twofold: one of practical limitations on free riding and a second of bounded rationality. The welfare state fails on both counts. In the modern welfare state, citizens expect certain entitlements, as the word itself implies, and government agents have little latitude or incentive to effectively monitor the distribution of benefits. In effect, denial of access to the fiscal commons is problematic. 1 These problems are exacerbated when it is difficult to assess the status of the fiscal commons, as is likely the case for politicians with short time horizons, especially when demographic trends, future tax rates, and program changes, among a host of other variables, are difficult to forecast.
The common thread in all these models is an institutional failure to establish a property rights scheme to ration access to the fiscal resources. This failure results in citizens potentially gaining benefits from the commons, while shifting the costs of these benefits onto others. In Wagner’s words, “(o)nce tax discrimination is allowed, choice and liability are . . . separated” (p. 118).
We should point out that Raudla (2010) calls into question literature that equates fiscal tragedies of the commons with their natural counterparts. Referencing the work of Elinor Ostrom, Raudla points out that the problem may be one of coordination, or assurance over what other participants in the budgetary process may do. Further, transactions costs and efforts to curb rent dissipation may limit demands on the budget, and coercive taxes and fiscal investments may replenish the pool. Nevertheless, Raudla concedes that time horizons governed by short election cycles, along with legislature turnover, are unlikely to yield the “stable community of appropriators” (p. 214) necessary to avert fiscal tragedy. This conclusion is strikingly similar to the one recognized by Crowe (1969) more than forty years earlier. Buchanan and Yoon (2004) also recognize this possibility. They argue that while a “membership externality” may limit taxes taken in majoritarian democracy, political reality may differ when taxpayers have different incomes and, as a result, different tax generating capacities.
The Simple Analytics of the Distribution of Tax Liability
While we see the commons model as a useful thought facilitating device when seeking to understand fiscal decisions in a representative democracy, we recognize that collective decisions made on the commons will be affected by the will of a relevant majority somehow determined. We are sympathetic with Becker’s (1983) work that focuses on interest group competition as it affects political outcomes. Becker’s model takes account of the relative costs and effectiveness of competing interest groups in their rent-seeking efforts. Given voting rules and the relative abilities of interest groups to gain political influence by lobbying and ultimately by voting, Becker argues that the final outcome will be efficient in the sense that all costs, including deadweight losses, will be minimized when the final political transfers are made. With this in mind, we call attention to the work of Bueno de Mesquita and Smith (2011) and their selectorate analysis. Their key point is as powerful as it is simple: state representatives, senators, governors, presidents, and even dictators must satisfy key citizen groups if they are to maintain power, and the relative size of these essential groups rises to a maximum in a democracy. Whether the fiscal pasture be an unconstrained commons or not, and no matter what the voting rule, or if voting is allowed or not, political power can always be shifted from one political agent to another one. Dictators can be deposed and elected representatives not returned to office. In short, “voting” within the essential group still matters.
We point out that some current fiscal outcomes reflect the results of past political decisions. In effect, path dependency matters. An example of particular relevance to our work is that most US states adopted income taxes years ago, but some did not. Fiscal policies, such as tax structures, can be and are changed over time, whether with major overhauls or minor tweaks, but even these changes depend on the starting position.
Because we do not delve into the details of the political process, our approach has limitations. Unlike Meltzer and Richard (1981), who model the political process and for whom taxation and redistribution are endogenous, we take the rules governing the supply of revenue that accrues to the commons as given and then observe how revenues are distributed across taxpayer groups. Put another way, we accept differences in tax structures as being exogenously determined and then observe fiscal outcomes across these different tax structures.
Given this background, we consider a polity in which fiscal rules, the tax structure in particular, have been determined. We envision N citizens divided into M evenly sized groups, n
1, n
2, n
3,…, n
M. If the tax rate for group i is given by ti
and the income per capita of each member of group i is given by Yi
, the total tax take, T, is given by
In effect, for a given value of tax revenues, there are many possible vectors of tax rates and incomes among the M groups. A vector of tax rates or incomes or both that concentrates the tax liability on a minority of taxpayers would cause the majority to favor increased government spending, thereby raising the level of T/N, since the cost of higher levels of government benefits would be disproportionately borne by a minority of taxpayers. 2
In reality, citizen’s incomes differ, and across polities, the effects of different tax instruments are significant. As argued by Davis et al. (2009, 4–11), state and local governments’ mix of income, property, and sales and excise taxes determines the progressivity of tax policy, with income taxes generally the most progressive, property taxes mildly regressive, and sales and excise taxes highly regressive. In turn, the tax structure determines the share of total taxes paid by different income groups.
We note that this simple model ignores many important sources of spendable funds for state and local governments. Transfers from the federal government, liability settlements, other taxes, and user charges, among other sources, provide significant revenues. Once these revenues are in hand, state and local governments may use them to substitute for tax revenues obtained from internal sources, thereby breaking the fiscal constraint that otherwise would affect them.
The size of these transfers can be impressive. For example, under the Disproportionate Share Hospital Allotment program, states receive large amounts of federal funds to augment Medicaid services (Coughlin and Liska 1997, 1). This program distributed $11.2 billion in the 2011 fiscal year (Henry Kaiser Family Foundation 2012). Education supplements to offset the cost of educating military dependents are another source of transfer funding, providing states with approximately $900 million in 2010–2011 (Schweers 2011). Tobacco settlement money provides yet another example of nontax revenues that may be used for general fund purposes. Under the 1998 multistate tobacco settlement negotiated by the state attorneys general, some $246 billion has been distributed (Campaign for Tobacco Free Kids 2012). 3
In earlier research (Lipford and Yandle 2012), we examine the effects of changes in the distribution of the federal tax liability on total spending, social welfare spending, and debt. We find that entitlement spending and debt, as a share of gross domestic product (GDP), have risen as the share of taxes paid has shifted more toward the upper end of the income distribution across time. In this article, we test the effects of the distribution of the tax liability by income group on state and local government spending, the composition of this spending, and state and local government debt.
Data, Tests, and Results
Before analyzing the empirical results, we briefly discuss the ITEP data we use to calculate measures of the distribution of tax liability.
The ITEP Data
The ITEP has published three studies, each titled Who Pays? A Distributional Analysis of the Tax Systems in All 50 States, that measure the distribution of the tax liability across the fifty states and the District of Columbia. Unfortunately, these studies, published in 1995, 2002, and 2007, are not compatible across time. 4 For these reasons, we confine our analysis to 2007 and the fifty-one observations available.
The ITEP data have a second limitation in that they exclude elderly taxpayers. Because states and localities provide a host of complex and inconsistent tax breaks to citizens aged sixty-five and older, the ITEP found it necessary to calculate the distribution of taxes for only the nonelderly population.
Of particular importance to this article, the ITEP data record taxes paid only by state residents. Since citizens may pay taxes to a state or a locality other than the one in which they reside (e.g., a vacationer paying sales tax in a neighboring state), the ITEP data are ideal for assessing the actual tax burden faced by state residents who can vote. 5
For each state and the District of Columbia, the ITEP divides the population into income-based quintiles. (For the top quintile, further breakdowns are provided.) The ITEP reports the average income and percentage of income paid in income, property, and sales and excise taxes for each income quintile. 6 Using these data, we calculate the average tax paid by the average citizen in each income quintile. Summing these figures yields the total taxes paid by five average citizens, one from each income quintile. By dividing the taxes paid by each average citizen in each income quintile by the total taxes paid by all (average) citizens, we calculate the share of total taxes paid by income quintile. With these figures, we calculated a Herfindahl Index for the concentration of taxes paid. We also utilize the ITEP’s measure of tax inequality as another composite measure of the distribution of tax liability. In this measure, high values indicate a progressive tax system and low values indicate a regressive tax system. 7 Table 1 provides descriptive statistics for these variables as well as all variables used in the empirical tests described in the following.
Descriptive Statistics.
To illustrate the variation in tax liabilities across the fifty states, we provide two outline US maps. Figure 1 shows the share of taxes paid by the lowest income quintile across the fifty states. Wide variance across region and income is evident. For example, southern states South Carolina and Kentucky, two lower-income states, are included with three northern, higher-income states, Connecticut, Massachusetts, and New York, where the lowest income quintile pays a small share of state and local taxes.
Share of 2007 state and local taxes paid by the average citizen in the lowest income quintile.
Figure 2 gives a breakdown on the share of taxes paid by the top income quintile. Again, variance by region and income is great. High-income taxpayers pay less in states as different as Mississippi, the Dakotas, and Washington, and pay more than half of the taxes in the South Atlantic states of the Carolinas and Virginia and along the western coast states of California and Oregon.
Share of 2007 state and local taxes paid by the average citizen in the top income quintile.
Tests and Results
The share of total taxes paid by each income quintile and the composite measures derived from these shares (e.g., the Herfindahl Index and the ITEP Tax Inequality Index) provide evidence on the concentration of a state’s tax burden. In effect, these distribution indexes measure how much each income quintile of a state’s population contributes to the fiscal commons. They allow us to test whether and to what extent the fiscal commons is affected by the distribution of the tax burden on citizens. We examine and report the effects of the distribution of the tax liability on total state and local government spending, the composition of state and local government spending, and state and local government debt.
Effects on Total Spending
In order to test the effect of the distribution of the tax liability on total state and local government spending, we regress direct general expenditures of state and local governments as a percentage of total state personal income against the various measures of the distribution of the tax liability. 8 These measures are our variables of key interest. We also consider a number of control variables, including the share of general revenues from all sources other than income, property, and sales and excise taxes; the state unemployment rate; the share of state population aged sixty-five and above; the area of the state; the ratio of state population to number of local governments, and the square of this ratio.
Before examining regressions in detail, we consider the control variables and their expected signs. As mentioned earlier, federal transfers, liability settlements, and other taxes and user fees are significant sources of state and local government revenues. We predict these revenues, measured as a share of general revenues, should be positively correlated with total spending. 9 That is, regardless of the distribution of the tax liability, additional revenues translate into additional spending. As a measure of states’ overall economic condition, we include the state unemployment rate. A high unemployment rate indicates a poorly performing state economy and greater demands on state revenues, particularly those that fund the alleviation of poverty. The correlation here too should be positive. We include the share of state population aged sixty-five and older for two reasons: one, these citizens are omitted from the ITEP data, and two, the elderly may place greater demands on total spending. Again, we expect a positive correlation.
To test the effects of physical constraints on spending, we include state area (square miles) and the ratio of population to the number of local governments, as well as the square of this term. If geographically large states are more costly to administer, spending should be positively correlated with area. But, scale economies may matter too. We predict that state and local spending should rise with the ratio of population to local governments, but that this increase should be at a diminishing rate. The appendix provides sources for these data.
We now turn our attention to the econometric results. Table 2 provides results for the ordinary least square estimates of total state and local government expenditures as a percentage of total state personal income for different measures of the distribution of the tax liability. We note at the outset that all regressions are statistically significant and have goodness of fit exceeding 0.6.
Regression Results for State and Local Government Total Expenditure Estimates.
Examining the control variables first, we find that as expected, revenue from sources other than income, property, and sales and excise taxes is positively and significantly correlated with total spending. A state’s general economic health is also important, as shown by the positive and significant correlation between a state’s unemployment rate and total spending. Contrary to our expectations, we find no evidence that the share of a state’s population that is elderly influences spending.
A state’s physical features also matter at a statistically significant level. Geographically large states are evidently more costly to administer, though we find some weak evidence that economies of scale may matter (at the 10 percent level for the two-tailed and one-tailed tests, depending on the regression and variable considered). As the ratio of population to number of local governments rises, total spending increases, albeit at a decreasing rate, indicating that the average cost of administering government services falls with larger local governments.
Turning to the measures of the distribution of tax liability, we first consider the composite measures, the Herfindahl Index and the ITEP Tax Inequality Index, as shown in regressions (1) and (2). These variables are positive and statistically significant at the 5 and 2 percent level, respectively, for a two-tailed test. Concentrating the tax liability at the top end of the income distribution is associated with higher state and local government spending. Regressions (3) and (4) focus on the share of taxes paid at the top end of the income distribution. These findings reinforce the estimates of the composite measures of the distribution of the tax liability: as relatively few citizens pay a higher share into the fiscal commons, total state and local government spending as a share of total state personal income rises.
Regressions (5) and (6) provide additional insight. When the lower quintiles of the income distribution pay a higher share of taxes, total state and local government spending falls. In effect, when contributions to the fiscal commons are spread more evenly across citizens, more citizens have a stake not only in the benefits of government programs but also in their costs. The evidence here indicates that consideration of these costs (taxes) restrains government spending. As the theory of the commons makes plain, institutions matter, and the institutions that govern which citizens contribute and how much they contribute are important.
Finally, we note that the magnitudes of the coefficients are not trivial. Focusing on the shares paid for ease of interpretation, we find that a ten percentage point increase in the share of taxes paid by the top 5 or 20 percent of the income distribution raises total state and local government spending as a share of total state personal income by over one percentage point. When the bottom 20 percent of the income distribution pays ten percentage points more of the tax burden, the estimates indicate that total state and local government spending as a share of total state personal income falls by over seven percentage points; when those in the bottom 40 percent of the income distribution pay ten percentage points more of the tax burden, government’s share of total state personal income falls by three percentage points. 10
Effects on the Composition of Spending
A second question relates to the composition of spending. Will citizens who pay relatively less in taxes influence the political process to increase spending on government programs that disproportionately benefit them? To test this hypothesis, we regress social welfare expenditures, defined as the sum of expenditures on public welfare, hospitals, health, social insurance, and veterans’ services, as a share of total state personal income, against our same measures of the distribution of the tax liability.
We again include the share of general revenues from all sources other than income, property, and sales and excise taxes, the state unemployment rate, and the share of state population aged sixty-five and older as control variables. However, we omit the variables related to state geographic size and the number of governments. The rationale for this omission is that most social welfare expenditures are administered at the state level instead of the local level, so that physical factors are less important. For the 2006–2007 fiscal year, the state share of total direct general expenditures was 43 percent, but the state share of social welfare expenditures was a comparatively higher 72 percent. We add the poverty rate for each state to the model, expecting that social welfare expenditures are positively correlated with poverty.
Table 3 shows the results of these estimates. We note that the explanatory power and goodness of fit of these estimates are inferior to the estimates for total spending. Nevertheless, the results yield interesting conclusions.
Regression Results for State and Local Social Expenditures Estimates.
As expected, we find that revenues unrelated to income, property, and sales and excise taxes raise social welfare spending as a share of total state personal income, 11 as does a higher unemployment rate. The poverty rate is also positively correlated to social welfare spending (though usually at the relatively weak 10 percent level for a two-tailed test). Again, the percentage of population aged sixty-five and older is not statistically significant.
The measures of the distribution of the tax liability again conform to our expectations. The composite measures, the Herfindahl Index and the ITEP Tax Inequality Index, have positive signs and are significant at the 10 and 5 percent levels for two-tailed tests. The shares of taxes paid by the top 5 and 20 percent of the income distribution are also positively correlated with more social welfare spending (though statistical significance is somewhat weak at the 10 percent level for a two-tailed test). At the bottom of the income distribution, the signs are again negative and significant: when those at the bottom of the income distribution pay a higher share of taxes, they demand less social welfare spending.
While the coefficient values are smaller than those from the estimates for total spending, they are still of material magnitude. Again focusing on the estimates of tax shares, we find that when the top 5 and 20 percent of the income distribution pay ten percentage points more, social welfare spending as a share of total state personal income rises by 0.5 percentage points. When the bottom 20 and 40 percent of the income distribution pay ten percentage points more, social welfare spending as a share to total state personal income falls by three and one percentage points. 12
Effects on Debt
When we make comparable estimates for state and local government debt as a percentage of total state personal income, we find no statistically significant relation with any of the measures of the distribution of the tax liability. We note this finding is contrary to that of Lipford and Yandle (2012), where we find a strong correlation between the distribution of tax liability and the gross debt–GDP ratio at the federal level. The purposes of debt may provide an explanation. Whereas most federal debt is used to finance current expenditures, most state and local debt is used to finance capital expenditures. Debt to finance capital expenditures that may replenish the common pool is not likely to be of interest to citizens at the lower end of the income distribution who are concerned with immediate payoffs from the fiscal commons.
Concluding Thoughts
This article has expanded the empirical research on the fiscal commons by examining the effects of the distribution of the tax liability on state and local government spending, the composition of this spending, and debt. For total spending and spending on social welfare programs, we find that as the distribution of the tax system becomes more concentrated, spending rises. As theorists in the natural and social sciences have long recognized, unrestrained access to a common pool leads to overconsumption. In particular, as this research has shown, the tendency toward tragedy is magnified when there are sharp differences between what citizens can take from the commons and what they give to replenish it.
While this article’s focus on state and local government spending in the United States contributes to our understanding of the fiscal commons, we submit that additional research would prove beneficial. In particular, cross-country studies or studies of different time periods would provide new evidence, sharpen empirical estimates, and provide further guidance to policy makers struggling with increasingly depleted fiscal resources.
The implications of this research are clear. When the share of taxes paid is concentrated on one group of citizens while others pay relatively little, total spending and social welfare spending increase, since all citizens, including those who pay little into the fiscal pool, have a voice in the expenditure side of fiscal decisions. This finding is consistent with the arguments of Jakee and Turner (2002), who emphasize that in democracies, entitlements have grown in scope and number, and citizens have developed an attitude of expectation toward these benefits. In the same vein, Wagner (1992), writing in context of the US federal budget, places the blame squarely on the “essentially unlimited democracy that characterizes contemporary America” (p. 119).
A broader tax base—or tax structure that evens the shares paid across income groups—is likely a necessary condition for any tax reform intended to restrain government spending. In a reversal of the cause of the American Revolution learned by every school child, the United States now has a system of “representation without taxation” for a significant portion of the population. Unless this system is dramatically reformed, strains on the fiscal commons at every level of government will almost surely worsen.
Footnotes
Appendix
Acknowledgments
The authors thank Eric Daniels, Jerry Slice, Russell Sobel, participants at a session of the 2013 Public Choice Society Meeting, two anonymous referees, and the editor of this journal for comments on an earlier draft. Any remaining errors are our own.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.