U.S. states, the Medicaid program, and tax smoothing.

• Author: Kula, Maria Cornachione

1. Introduction

   The result that tax rates over time should be equalized, the theory of "tax smoothing," has been obtained in several settings, including Barro (1979), Kydland and Prescott (1980), and Lucas and Stokey (1983). Empirical tests have focused on Barro's (1979) result that governments smooth tax rates given expected permanent spending, with temporary changes in spending or income resulting in surpluses or deficits. As only new, unpredictable information should cause a change in the tax rate, the tax rate should behave as a random walk.

   Several studies have used U.S. federal tax rates to test the random walk prediction, with conflicting results. Barro (1981) is unable to reject tax smoothing, while Bizer and Durlauf (1990), using spectral analysis techniques, reject tax smoothing. (1) Focusing on the implied behavior of surpluses and deficits, Ghosh (1995) and Huang and Lin (1993) are unable to reject tax smoothing. (2) Other tests of tax smoothing have used U.S. state and local tax rates. Strazicich (1997) uses frequency domain testing and rejects tax smoothing. Strazicich (1996) performs a Levin-Lin (1992) panel unit root test and rejects that state tax rates are unit root processes. (3) Dole (2000) performs Im, Pesaran, and Shin (1997) unit root tests and cannot reject tax smoothing for those states required to balance smaller portions of their budgets. (4)

   This paper tests the tax smoothing theory by focusing on its implication that a change in permanent government spending should result in an equal sized change in the tax rate. The effect of Medicaid, a state administered, federal and state funded medical insurance program for the poor, on state tax rates is investigated. The Medicaid program provides a natural experiment for this test as states are required to cover certain groups in order to receive federal matching money. From the mid 1980s to the early 1990s, a series of federal laws was passed which increased the universe of people covered by Medicaid. To give an idea of the extent of the mandates, in 1984, one out often Americans was enrolled in Medicaid; by 1992, that number had risen to almost one in seven. Consequently, the federal mandates resulted in large, irreversible increases in expenditure for most states. (5) The variation in mandated Medicaid expenditures across states and within states over time is used to identify the effect of changes in Medicaid spending on state tax rates. Two stage least squares is used on a panel of U.S. states (1978-1994) to test whether changes in permanent state Medicaid expenditures resulted in equal sized tax rate changes.

   This work is similar in spirit to Barro (1979, 1986) and Sahasakul (1986), which focus on periods of war as times of temporary spending increases. (6) However, a potential problem with this work is that during wars, the tax elasticity of output (due to changes in the labor supply response to tax rates) cannot be considered to be constant, an underlying assumption of the tax smoothing model. (7) Hess (1993) finds that the tax elasticity of output changes predictably during wars. This may be due to the effect of patriotism on agents' labor supply decisions; Mulligan (1998) finds that changes in workers' budget sets cannot explain the increase in civilian work in the U.S. during World War II. A tax smoothing test focusing on permanent Medicaid expenditures is consistent with the tax smoothing model.

   Tax smoothing as a positive theory of state government behavior is rejected. An additional insight from this analysis relates to the work on the effectiveness of balanced budget requirements, as no behavioral differences are found when comparing states with the most stringent balanced budget requirements to those with the least stringent.

2. The Use of State Data

   The use of state data raises two issues when investigating optimal tax setting policy, the first of which relates to population mobility. Tax smoothing might not be optimal if people change state residence based on differing tax rates. However, interstate movement is not as common as is sometimes assumed. For example, according to the U.S. Census Bureau (1986), in 1985 only 8.7% of Americans lived in different states from their 1980 residences. The percentage moving specifically because of state tax or spending reasons is presumably much smaller. In addition, Barro and Sala-i-Martin (1991) found that an increase of 1% in a state's per capita personal income level raised the state's population growth rate by only 0.026% per year. It is possible that the potential for even greater migration could deter states from pursuing tax smoothing strategies; however, this effect would be most prevalent at the local level. (8)

   Additionally, in order for tax smoothing to be viable, the government must be able to run deficits during "bad times" and surpluses during "good times." Balanced budget rules, which most states have, would seem to preclude the possibility of actual tax smoothing by state governments. However, the effectiveness of these rules is an open question. Bohn and Inman (1996), for example, find that states which have regulations limiting the amount of debt have lower average deficits, while Sorensen, Wu, and Yosha (1998) find that states are able to systematically smooth income shocks. In addition, balanced budget rules would not prevent states from running surpluses and varying these surpluses over time to smooth tax rates. (9) Given this, it is assumed that balanced budget rules do not prohibit states from implementing tax smoothing strategies. The validity of this assumption is tested in the empirical section.

   Finally, states vary in the responsibilities assigned to subgovernments. To ensure compatibility across states, the unit of analysis used in this paper is the combined state plus local sectors. For brevity, "state" is used to denote what is actually "state plus local." State revenue and spending do not include any transfers from the federal government.

3. Medicaid

   This section describes the basic Medicaid program, the federal mandates which greatly increased recipients and expenditures, and the inability of states to control their rising health care costs. It will also show how the states' historical positions on Medicaid, their views on the mandates, and their actions with regard to Aid to Families with Dependent Children (AFDC) combine to illustrate that mandatory Medicaid expenditures are exogenous from the states' perspectives.

   Basic Program Facts

   Medicaid originated as a jointly funded (10) federal-state medical insurance program for those receiving cash assistance--single parents with children receiving AFDC, and the aged, blind, and disabled receiving Supplemental Security Income (SSI). While the federal government specifies eligibility and cash benefits for SSI, thus ensuring uniformity across the country, each state sets its own financial eligibility conditions and benefit levels for AFDC. Thus the states traditionally had control over the main avenue for entry into Medicaid.

   Different state AFDC programs meant wide variation in state Medicaid recipients and expenditures. (11) In 1984, for example, total (federal plus state and local) Medicaid expenditures per capita ranged from a high of $382.87 in New York to a low of $52.05 in Wyoming (Table 1), a gap far greater than can be explained by differences in regional cost of living. California, with more similar demographics to New York, had expenditures per capita roughly three times smaller than New York. From 1984 to 1994, the average total (federal plus state and local) Medicaid expenditure per capita and the average state plus local Medicaid expenditure per capita measured in constant dollars roughly doubled (Table 1). Overall, Medicaid spending is a significant part of state budgets. Aggregating over all states, (12) from 1980 to 1990 its share of state budgets grew from 9% to 14%. Table 2 lists the growth rates of aggregate state and local Medicaid spending and state and local spending on everything else between 1978 and 1994. In every year the Medicaid growth rates dwarf those for other spending.

   Federal Mandates of 1980s and 1990s

   Beginning in 1984, Congress passed a series of mandates (described in Appendix A) which broke Medicaid's tie to cash assistance eligibility. Medicaid coverage was expanded to non-AFDC low income pregnant women and children, and low income Medicare beneficiaries who were not eligible for cash assistance programs but needed help with Medicare cost sharing. These mandates are important not only because they expanded the pool of Medicaid eligibility, but because this expansion was achieved partly by defining this eligibility in terms of the federal poverty line.

   The Medicaid mandates had substantial eligibility implications. Currie and Gruber (1996a) find that Medicaid eligibility for children increased by 100% between 1984 and 1992, with almost one third of all children in America eligible for Medicaid in 1992. Currie and Gruber (1996b) find that almost 45% of women were eligible for the expenses of pregnancy in 1992, an increase of 250% from 1979. Cutler and Gruber (1996) report that two-thirds of those made eligible for Medicaid via the expansions were already covered by private insurance. (13) Federal mandates that increased the pool of eligibles was not the only way costs were imposed on states. Federal legislation, for example, increased the amounts of income and assets spouses of nursing home residents did not have to put towards nursing home care. (14)

   States' Responses to Medicaid

   Even before the mandates of the 1980s and 1990s, Medicaid control and financing issues were a persistent source of tension between the federal government and the states (Hansan, Racine, and Vignola 1979). Many characterized it as the proverbial "800 pound gorilla" in terms of federal interventions which negatively impact the fiscal flexibility of the states (Nathan 1996).

   States...