The effect of tax rebates on consumption expenditures: evidence from state tax rebates.

INTRODUCTION

Tax rebate policies have often been discussed as a tool that the government may use to stimulate consumer spending and, so the hope goes, boost the economy in general. However, the efficacy of tax rebates to stimulate consumer spending is still an open question, both theoretically and empirically. This paper estimates the consumption response to a series of state rebates distributed between 1995-2001 using data from the Consumer Expenditure Survey.

At the federal level, two rebates have been distributed in recent years. A provision in the Economic Growth and Tax Relief Reconciliation Act of 2001 (EGTRRA) consisted of a decrease in the lowest tax rate from 15 percent to 10 percent, the benefits of which were mailed out in a rebate check that amounted up to $300 for individuals or $600 for those married tiling jointly. In addition, the Jobs and Growth Tax Relief Reconciliation Act of 2003 (JGTRRA) provided for a $400 increase in the Child Tax Credit, the benefits of which were mailed out to eligible taxpayers in July and August of 2003. A few recent papers (Shapiro and Slemrod (2002, 2003) and Johnson, Parker, and Souleles (2004), for example) have attempted to estimate the effect of these federal rebates on spending, but the evidence on the effect of such rebates is still sparse.

The contribution of this paper, then, is to estimate the expenditure response to state level rebates that were implemented in Connecticut, Minnesota, Oregon, and Wisconsin between 1995 and 2001. Studying these state tax rebates is advantageous for a number of reasons. First, the timing of the rebates varied across states, so that one can identify the effect of the tax rebates off of the difference between the expenditure paths of individuals in a state-quarter in which a rebate was received, and the expenditure path of individuals in states in which rebates were received at some point, but not received in that particular quarter. (1) Second, the magnitude of rebates varied widely across states, even for a given income level, and varied considerably within states. As a result, these rebates contain much more identifying variation than the federal rebates, whose amounts were predominantly $300 for single filers and $600 for married filers. (2) Third, in this study, it is possible to differentiate between the response to the enactment of rebates (when the law is signed) and the implementation of rebates (when the rebate is received). Although previous studies have been able to examine households' response to the receipt of the rebate, since they all examined federal policies, there was no variation in when the law was signed by the president, which could be used to identify the consumption response to these events. In this study, signing of the state level rebates varied across states, and so it is possible to examine whether households' consumption patterns responded to this enactment.

However, these strengths come at a cost, in that respondents to the Consumer Expenditure Survey were not asked specifically about the receipt of these state rebate checks, and so rebate amounts must be imputed using the data available. As a result, any lack of significance in the resulting estimates could simply be due to attenuation bias resulting from a low signal to noise ratio in the independent variable. Despite this weakness, some significant effects of the rebates emerge.

Results from the base specification suggest that receipt of a rebate check has a positive, though insignificant, effect on the expenditures of individuals who receive them in the overall sample. Consistent with Shapiro and Slemrod (2003) and Johnson et al. (2004), these results suggest that approximately one-fifth to one-fourth of the rebate amounts were spent in the quarter of receipt. The estimates also suggest positive responses to the rebates in nondurable spending and spending on apparel in particular, and among households with single respondents. The results vary substantially, however, depending on the sample, regressor used, and component of expenditure examined, with several variables entering insignificantly or with the wrong sign.

When the effect of the announcement of the rebate check amounts is estimated, mixed results are found. The announcement of a rebate is estimated to have had a small and insignificant effect on the amount of spending, but may have shifted the composition of spending toward durable goods in the quarter of announcement.

The paper is organized as follows. The next section reviews the relevant literature, and the tax rebates passed in the states under analysis are described in the third section. The fourth section details the data and empirical strategy used. The fifth section presents estimation results on the estimated response to the receipt of a rebate, and the sixth section presents results on the estimated response to the announcement of the distribution of rebates. The final section concludes.

RELEVANT LITERATURE

Theoretically, the effect that receipt of a tax rebate will have on consumer spending depends on the assumed behavioral model. For example, under a simple intertemporal choice model in which the rebate is anticipated, such a policy would have no effect on spending, since receipt of the rebate would not affect the family's expected lifetime resources. However, in an intertemporal model in which some individuals are credit constrained, such individuals might be induced to increase spending by such a policy, in order to move closer to their unconstrained optimal consumption path (see Browning and Lusardi (1996)). From the behavioral literature, "rule of thumb" or "mental accounting" type models would predict that individuals might spend most or all of their rebate checks. However, this prediction depends crucially on how individuals classify the tax rebate in their "mental accounts," which is inherently unobservable (see Thaler and Loewenstein (1989) and Thaler (1990)).

Empirically, a number of studies have attempted to evaluate the effects of transitory changes in income, both predictable and unpredictable, on consumer spending. Though these papers focus on whether people smooth consumption in line with a lifecycle model, they might also shed light on how individuals' consumption responds to changes in tax parameters.

For example, Wilcox (1989), Parker (1999), and Souleles (2002) each find significant responses to pre-announced changes in income due to various tax changes. Since these changes in income were predictable, these authors argue that their results are evidence against a lifecycle model. Further, several papers (including Souleles (1999), Barrow and McGranahan (2000), and Hsieh (2003)) have estimated (at least in part) consumers' responses to the receipt of tax refunds, and found significant consumption responses to refund receipt.

Several older papers have also attempted to estimate the effect of tax rebate checks per se on consumer spending using, including Blinder (1981) and Poterba (1988). In addition, three recent papers have used individual data to estimate the effect of tax rebates on consumer expenditures. Shapiro and Slemrod (2002, 2003) found that only about a quarter of individuals in a special supplement to the Survey of Consumers reported that the EGTRRA 2001 tax rebates led them to mostly increase spending, regardless of whether they were asked before or after the receipt of their rebate check. Unfortunately, no questions are asked about by how much the rebate induced the individuals to increase their spending, so it is impossible to translate these directly into marginal propensities to consume. However, Shapiro and Slemrod (2003) do some back-of-the-envelope calculations, suggesting that the marginal propensity to consume could range from 0.34 up to 0.5. Finally, Johnson et al. (2004) use variation in the week in which the EGTRRA 2001 rebate checks were mailed to estimate the consumption response. They find that households consumed a large portion of their rebate checks, spending between 20-40 percent of their rebate check on nondurables in the quarter of receipt and an additional third of the amount of the check in the following three-month period.

STATE TAX REBATES, 1995-2001

In the years preceding the passage of EGTRRA, states' projected tax receipts increased substantially from year to year, and actual tax receipts were quite often higher than had been projected. For example, according to the National Association of State Budget Officers (various years), in 2000, tax collections were projected to be approximately 4.3 percent above the previous year's receipts, and came in at 3.9 percent above these projections. This year was not an aberration; in the previous five years, revenues had on average exceeded projections by at least two percent. As a result, throughout this period, the majority of states ran general fund surpluses. (3) States responded to these surpluses in a number of ways, the most prevalent of which were increasing rainy day or budget stabilization funds, investing in capital construction or education, and reducing taxes. (4)

Most pertinent to the study at hand, of the states that reduced taxes, a handful of states distributed rebate checks to some or all of the state's taxpayers. These states were Connecticut, Minnesota, Oregon, and Wisconsin. A brief description of each of these states' rebate programs follows.