# Tax Reform 1986 and marginal welfare changes for labor.

I. Introduction

During the 1980s in the United States, before-tax income shifted away from the poor and toward the upper income groups. Changing the distribution of tax burdens or after-tax income is one way to redress a growing imbalance in before-tax income. Pechman [14], Feldstein [6], and Wallace et al. [17] have all shown that the Tax Reform Act of 1986 (TRA86) increased the progressivity of the combined personal and corporate income taxes, an achievement not accomplished in the last two decades of income tax reform [14].

Tax reform can also increase or decrease the welfare of individuals in an economy through its effect on the allocation of labor supply resources. TRA86 reduced the highest marginal income tax rates from 50 percent to 28 percent for high income persons and removed some lower-income persons from the tax roles. Such large marginal tax rate changes introduce significant potential for increased economic efficiency in labor supply and for welfare gains for the rich and poor alike. These changes should be part of the benefit analysis of tax reform. Indeed, Hausman and Poterba [10] report welfare cost changes that result from TRA86 for the average male and female. Their estimates, however, do not account for welfare cost changes throughout the income distribution. In addition to our use of microdata to estimate the marginal welfare cost of tax reform, we use more precise measures of marginal tax rates than other researchers.

In this paper, we use Browning's partial equilibrium framework to examine the influence of the reduction in marginal tax rates from TRA86 on the distribution of welfare gains from labor supply reallocation for males by population decile and for females by population decile and by marital status. The empirical results illustrate how the tax reform affected low-income female household heads, where poverty is so heavily concentrated, as well as other household heads and spouses. The results indicate that TRA86 reduced welfare costs in the economy, but these welfare gains were distributed in a pro-rich manner or they grew as income increased for males and for females regardless of marital status. In fact, TRA86 resulted in less allocative efficiency and in welfare losses for individuals in the lowest population decile. Thus, in spite of TRA86's progressive changes in tax burdens, the labor supply welfare gains afforded by TRA86 were oriented toward the highest income individuals and away from the lowest income individuals.

In the next section of the paper, we review several methods of estimating welfare changes from tax reform. We examine Browning's method and explain why we use it for our analysis. We then turn to the empirical values for key parameters of the model and discuss the micro data simulation file that we use to estimate the welfare gains. The results are reported in a fourth section of the paper and are followed by the conclusions that we draw from our work.

II. Marginal Welfare Cost

Hausman and Poterba [10] report changes in welfare costs associated with labor supply decisions of men and wives due to the tax reforms of 1981 and 1986. Using Hausman's [9] indirect utility function and the IRS Public Use Tax Return data file, they compute welfare gains per dollar of revenue change associated with the labor supply of the average male wage earner at 0.03 and of the average wife at 0.055.(1)

More general studies by Stuart [16] and Ballard, Shoven and Whalley [2] use general equilibrium analysis to estimate the marginal welfare cost changes associated with labor supply when tax rates change. They examine the marginal effects for hypothetical tax reforms and use aggregate data. We do not use their general equilibrium analyses, because it would require considerable aggregation, thus forcing us to forgo the benefits of our rich micro data set.

Our analysis is performed in a partial equilibrium framework, as developed by Harberger [8] and modified by Browning [4]. Using welfare formulas, elasticity of labor supply elasticities, and changes in the tax system, welfare cost effects are then calculated for each individual by substituting the relevant compensated labor supply elasticities and tax parameters into the appropriate formula. Fullerton [7, 305] reconciles the general equilibrium approaches cited above with Browning's approach, concluding that, although not without limitations, Browning's approach is useful for comparing tax changes (our problem), because his method measures distortions appropriately.

We employ Browning's [4, 12] modification of Harberger's [8] measure of the partial equilibrium welfare cost.(2) Our primary interest is in the marginal change in welfare cost or in the change in welfare cost divided by the change in tax paid for different groups of individual taxpayers.(3) The computation assumes that the additional expenditures financed by the tax changes have only an income effect.(4) Under these assumptions Browning [4] shows that the change in welfare costs (dW) when tax revenues from labor income increase by dR is given by:

dW/dR = [(m + 0.5dm)/(1 - m)]nw[L.sub.2]dm/[w[L.sub.2]dt + wdL(m + dm)] (1) where

m = marginal tax rate before tax reform,

m' = marginal tax rate after tax reform,

dm = (m' - m),

n = compensated labor supply elasticity,

w = net wage before tax reform,

[L.sub.2]= labor supply before tax reform,

dt = change in the average tax rate due to the reform,

dL = total (income and substitution effects) change in labor supply due to tax reform.

The numerator of equation (1) measures the change in the welfare costs associated with labor as a result of TRA86. The denominator of equation (1) measures the change in the tax revenue derived from labor income as a result of TRA86. The change in the tax revenue in the denominator means that the change in labor (dL) in the denominator represents both the income and the substitution effects for labor supply. On the other hand, the compensated elasticity of labor supply is included in the numerator to measure the change in the welfare cost.

Because we do not have data on dL, it is useful to rewrite (1) in terms of the ordinary labor supply elasticity, [E.sub.s.sup..5.]

dW/dR = [(m + 0.5dm)/(1 - m)]ndm/[dt + (dwl(1 - m)w)[E.sub.s](m + dm)]. (2)

Moreover, the expression dw/(2 - m)w can be written as:

[(1 - m')w - (1 - m)w]/(1 - m)w = dm/(1 - m).

Equation (2) then becomes:

dW/dR = [(m + 0.5dm)/(1 - m)]ndm/[dt + (-dm/1 - m)[E.sub.s](m + dm)]. (3)

Equation (3) is used to estimate the marginal welfare effects of the 1986 tax reform.

Unfortunately, the marginal welfare formula as derived by Browning [4] has the same sign for both an increase and a decrease in welfare.(6) For example, the formula produces a positive sign if marginal tax rates increase, which represents a welfare loss. More specifically, when the marginal tax rate increases then the term (dm) is positive, making the numerator positive. The denominator is simply the extra revenue that is collected when the marginal tax rate increases. Unless the amount of taxes an individual pays decreases when his or her marginal tax rate increases (an unlikely strong labor supply response), the denominator will also be positive when marginal tax rates increase, even though the sign of the denominator is theoretically ambiguous. Similar logic applies to a decrease in the marginal tax rate. In that case, the sign of (dm) and, therefore, the numerator will be negative, and the change in revenue (dR) will also generally be negative when marginal tax rates are reduced.

To solve the sign problem, we note that the sign of the numerator more meaningfully represents the welfare change, while the denominator only standardizes the change in welfare costs per dollar of change in taxes. For most individuals in the data set, the numerator and the denominator have the same sign. In that case we took the absolute value of the denominator, so that the marginal change in welfare cost expression took on the sign of the numerator or the indicator of change in welfare costs. Therefore, a marginal welfare gain has a negative sign, as negative dm means that marginal tax rates decline, and a marginal welfare loss has a positive sign, as a positive dm means that the marginal tax rates increase. It is possible, however, for an individual's marginal tax rate to decline and the change in revenue to increase. In this case, the substitution effect is greater in absolute size than the income effect and the individual chooses to work more as a result of increased wages. These relatively few cases are treated as a marginal welfare gain resulting from the reduction in the marginal tax rate.(7)

After calculating the marginal welfare costs of TRA86 for each taxpayer, the taxpayers are grouped into population deciles. The results of the average change in marginal welfare cost per taxpayer in each decile are presented for males, female spouses and female household heads, in three separate tables.

III. Values of Key Parameters

The above formula for excess burden or marginal welfare cost requires measures of the marginal tax rate before and after tax reform, the change in the average tax rate, the compensated elasticity of labor supply, and the ordinary elasticity of labor supply.(8)

Marginal and Average Tax Rates

We calculate the federal personal marginal tax rate for each individual using both the 1986 (pre-reform) and the 1991 (post-reform) tax structure using the U.S. Treasury Individual Income Tax Model (ITM). The micro data identifies each filer's state of residence and we supplement the federal tax rate with the state income tax rate for each filer. We also include, as part of the marginal tax rate calculation, the combined employee and employer social security payroll tax rate when wage income for the individual is below the social security ceiling and a measure of the average sales and excise tax rates assuming the taxes are borne by factors of production.

To complete the calculation of the marginal tax rate, we account for state and federal deductibility of federal and state income taxes, respectively.(9) While other researchers include state taxes in their marginal tax rate calculations, no one accounts for deductibility of federal income taxes from state income taxes and for the deductibility (and lack thereof post TRA86) of sales taxes in computing marginal and average tax rates.

The average tax rate is the tax paid on labor income divided by pre-reform labor income. The change in the average tax rate (dt) is based on the difference between the average tax rate that would exist under the 1991 tax structure and that in existence before the 1986 reform.(10) In order to focus on the effects of the individual income tax, we hold the social security payroll tax rates at their 1986 levels.

Elasticities of Labor Supply

Equation (3) indicates that the computation of welfare cost changes requires estimates of both compensated and uncompensated labor supply elasticities. In his survey of labor supply behavior for men, Pencavel [15, 3-102] puts the mean value of the compensated elasticity of male labor supply at 0.11. The actual range of compensated male labor supply elasticity estimates in the studies cited by Pencavel is from - 0.19 to 0.84.(11)

Killingsworth [11] and Killingsworth and Heckman [12] review comprehensively the empirical estimates of the elasticity of labor supply for females. We use the elasticity estimates from both in our work. The estimates reported in the two papers have significant overlap. Female household heads have compensated elasticities of between 0.16 and 0.77. The compensated elasticity estimates for wives are between 0.11 to 4.73.(12)

Based on the figures above, there are no definitive estimates of compensated labor supply elasticities. Thus, most excess burden studies calculate welfare costs based on a range of elasticity values. We use 0.11, 0.2, and 0.55 for compensated elasticities of labor supply for males. These elasticities come from Pencavel's [15, 69] reported mean value (.11), and two compensated elasticity point estimates that he reports. For females, we use a broader range of values for the compensated elasticity of labor supply. For spouses, we use elasticities of 0.2, 0.4, 0.8, 1.0, 1.5 and 2.0. For female household heads, we use elasticities of 0.3, 0.5, and 0.8. These figures seem to us to be somewhat on the low side of the accepted range. Hence, our estimation of the change in the excess burden due to tax reform will tend to be conservative.

For ordinary labor supply elasticity, we use Pencavel's [15, 69] reported mean of - 0.12 for males. There is considerably less agreement about the female ordinary labor supply elasticity. Killingsworth and Heckman [12, 195] point out that many studies report ordinary labor supply elasticties for wives in the range of 0.5 to 1.0, with a few studies reporting much higher elasticities. They suggest that female labor supply decisions require further study. Hausman [9] is one of the few studies of U.S. female labor supply that estimates an elasticity for wives as well as an elasticity for female household heads. As his ordinary labor supply elasticity estimates are within the range of those produced by recent studies, we use his estimates of ordinary labor supply elasticities for wives of about 1.0 and for female household heads of about 0.5.

Data and Simulation

We use micro data from the U.S. Treasury Individual Income Tax Model (ITM) to estimate the welfare cost changes for individuals that result from TRA86.(14) The ITM simulates the tax system in 1986 and 1991 (the first year that most provisions are fully phased in), at 1986 levels of income. The Treasury ITM is a microsimulation model, based on the Internal Revenue Service 1985 Statistics of Income (SOI) file. The SOI tax data are supplemented with information from the U.S. Census, Current Population Survey. A series of extrapolation weights are available, which allow the data to be extrapolated beyond 1985 [5].

The data are used to estimate the marginal excess burdens of tax reform and the estimates are then grouped into population deciles. We use a household economic income concept, which is a comprehensive income definition that begins with adjusted gross income (AGI) and is then augmented with imputations for capital income and transfer income receipts.(15)

IV. Results

Table I reports the average change in welfare cost per dollar of tax paid for males in each population decile. Tables II and III report the results for female spouses and for female household heads, respectively.(16)

Based on the results reported in Table I, individual income tax reform on average reduced the welfare costs of labor supply for males in the fourth through the tenth population deciles. The size of the marginal welfare cost gains range between - 0.02 and - 3.34. The largest changes in welfare costs per dollar of change in individual taxes are with some exceptions in the population deciles with the highest incomes. The fact is that males in the highest income population deciles experienced the largest reductions in marginal tax rates. Males in the three income population deciles with the lowest income have small welfare cost increases due to small increases in their marginal tax rates.

[TABULAR DATA OMITTED]

The results reported in Table II suggest that, except for those in the first and second population deciles, individual income tax reform reduced the welfare costs associated with female spouse labor supply. Based on the overall average, the average welfare cost reduction for female spouses ranged between - 0.28 and - 2.80. Although the range of compensated elasticities is much larger than that for males, the marginal welfare cost reductions are not larger than that for males when comparisons are made for similar compensated labor supply elasticities. One reason for the relatively small welfare cost reduction is the elimination of the dual wage-earner tax deduction. Its elimination has the effect of raising the average tax rates and, in some cases, the marginal tax rates for wives. The largest average marginal welfare cost reductions are in the eighth and ninth population deciles.

[TABULAR DATA OMITTED]

The results for female household heads reported in Table III show that TRA86 also reduced their welfare costs on average in every population decile, except the first decile, and the marginal welfare cost gains were on average higher than those for males. The marginal welfare cost gains were between - 0.01 and - 5.53 and the largest average gains occurred in the ninth decile.

In general the magnitudes of the marginal welfare costs are large as suggested in other studies of marginal welfare changes by Browning [3], Stuart [16] and Ballard, Shoven and Whalley [2]. However, the results in this paper suggest that TRA86 lead to different marginal welfare cost reductions for wives, female household heads and males. Such diverse results cannot be obtained from aggregate data, and our results demonstrate the benefit of using micro rather than aggregate data to judge the welfare benefits of tax reform.

V. Conclusions

The findings in this paper have important implications for future tax reform. The reduction of tax rates is rarely done across-the-board, and marginal welfare cost changes from tax reform can differ substantially among population deciles. More importantly, the elimination of certain provisions from the tax code, such as the dual wage-earner deduction, can have unintended marginal welfare cost effects on certain wage-earners. In the case of the dual wage-earner deduction, its elimination has significantly reduced the marginal welfare cost gains for wives compared to the marginal welfare cost gains for males. The marginal welfare cost gains of TRA86 are generally distributed in a pro-rich manner, for all types of households, although the distribution of marginal gains is more pro-rich for males and for wives than for female-headed households.

The above findings also point to the importance of examining the marginal welfare cost effects using micro data. Marginal excess burden results based on aggregate data can only give an overall impression of what changes in marginal tax rates would imply for welfare costs. However, tax reform rarely involves so simple a change in tax rates. Eliminating certain provisions in the tax code can increase the average tax rate for certain classes of taxpayers. These unintended consequences cannot be gleaned from the results of aggregate studies. The tax reform debate could be misinformed, if the basis for discussion rests exclusively on the results from aggregate studies, as results from aggregate studies are not intended to, nor should they, guide all policy on tax reform.

(1.) Browning [3] points out the problems with Hausman's specification. (2.) Browning [3, 12] modifies the Harberger measure by valuing the variables in the change in welfare cost formula at after-tax wages and labor supply rather than at before-tax wages and labor supply. (3.) One limitation in following this approach is the assumption of linear compensated supply schedules. For some individuals with large changes in their marginal tax rates the linearity assumption may be less than a precise measure of their marginal welfare gain or loss. The linear compensated supply curve assumption may mean that we have only estimated an approximate welfare change for the relatively few individuals that have had a large reduction in their marginal tax rates. See Browning [4, 17]. (4.) Here we use Browning's [4] equation (9) for measuring marginal welfare costs. Although Browning presents three formulas for measuring marginal welfare costs, each formula makes an altemative assumption about the effects of the incremental government expenditure resulting from the hypothetical tax increase. His other two formulas (equations (10) and (11) in his paper) make polar assumptions about the marginal effects of government spending on individual utility levels. See Browning [3].

(5.) Browning's formulation of the marginal change in welfare costs does not assume that the tax reform affects labor income only. In particular, the term (dL) in the denominator would in principle include both the substitution effect and all changes in individual labor supply due to income effects from tax reform. We do not, however, have data for w and for dL for the denominator, and, therefore, cannot account for all changes in dL due to tax reform. In changing Browning's equation and using the labor supply elasticity in equation (2) of the text, we are assuming that tax reform's change in nonlabor income did not affect labor supply. While tax reform did affect non-labor income, most of the effect of tax reform on non-labor income was in the highest income brackets. We believe that changes in the taxation of capital income at high income levels do not affect labor supply decisions, and offer it as a partial defense of our use of the modified equation (2). In our calculation of the change in marginal welfare costs due to TRA86, however, we use the change in the marginal and average tax rates that result when all of the changes in capital and labor income are taken into account in the tax base. In addition, tax reform affected the marriage tax. We account for changes in tax revenue due to the higher marginal tax rate for the second wage earner. But we do not capture the income effect from the change in the second wage earner's marginal tax rate on the labor supply of the primary wage earner. (6.) Browning's problem was more straightforward as he deals with a positive change in aggregate revenue as a result of his hypothetical tax reform. (7.) Our data reveal only a few cases where the decline in the marginal tax rate led to a positive change in taxes. While we agree with one reviewer that there are interesting policy implications here, confidentiality of the data prevents us from revealing the exact magnitude of these effects. (8.) The use of micro-level data to calculate the excess burden for each individual eliminates the calculation of the average weighted marginal tax rate for labor necessitated in aggregate studies. One reviewer requested that we report wages and marginal tax rates by population decile to facilitate research using general equilibrium models. Unfortunately, the Department of Treasury will not permit us to report those figures calculated from the ITM model. (9.) In 1986, the following 12 states allowed full deductibility of federal income taxes from their state income tax: Alabama, Arizona, Colorado, Iowa, Kansas, Kentucky, Louisiana, Minnesota, Missouri, North Dakota, Oklahoma and Utah. Importantly, these states allow the deduction of federal tax liability even when taxpayers do not itemize deductions for state tax purposes. Thus, all taxpayers in these states deduct the federal tax liability from their state income, and taxpayers in these 12 states have their marginal federal income tax rate reduced by one minus the state income tax rate. In addition, Oregon taxpayers can deduct their federal tax liability from state income but the deduction is limited to $7,000 [1]. Some states also allow a deduction from state income taxes for social security payroll taxes. We take this deduction into account and also account for TRA86's changes in the ruies governing the deductibility of sales taxes in the computation of marginal tax rates. Due to a lack of data on transfer programs, we do not account for marginal tax rates of transfer programs, but we do account for the marginal tax rates from the Earned Income Tax Credit. (10.) To incorporate the 1991 fully phased-in tax laws into the analysis, we extrapolate the data to 1991 using the ITM to calculate the marginal and average tax rates. Note, however, that we have not incorporated behavioral effects associated with the change in the marginal tax rate on capital gains. Theory suggests that an increase in the tax rate on gains leads to a decrease in realizations. This would lead to a lower tax payment for a given tax rate. Since this behavior is not incorporated, returns that realize gains may have a higher level of tax than they would if realizations actually fell. This may bias the change in average tax, which would be greater for individuals whose level of realizations is affected by changes in the tax rate. The complications added by such behavior are not modeled here. (11.) Killingsworth [11] suggests that the range of positive compensated elasticity of male labor supply estimates is between 0.08 and 1.O. (12.) The availability of labor supply elasticity estimates constrains the extent to which we can estimate marginal welfare costs for racial and age subgroups of males and females. For example, the two studies that split the sample between black wives and white wives do not find a consistent pattern of differences in the compensated elasticities between black and white wives, and, therefore, we do not attempt to estimate excess burdens by race. A few studies report estimates for females in different age groups and for female household heads, for black female household heads and for wives. The studies do not, however, estimate separate elasticities for household heads by age, race and household group. Generally speaking, most data sets would not have enough observations in most cells to estimate labor supply elasticities for age groups by household headship by race. (13.) Households with two wage earners may make labor supply decisions simultaneously. We do not account for the simultaneity in this analysis. (14.) We consider the effects of changes in individual taxes only, and do not attempt to measure the welfare effects of corporate income tax changes. (15.) For a more complete description of economic income, see Cilke and Wyscarver [5] and Nelson [3]. (16.) The population deciles and the respective income ranges are defined using all income tax returns and economic income. Therefore, the income range is the same for any given population decile in the three tables. The implication, of course, is that 10 percent of the male filers are not necessarily in the first population decile of Table I, but that 10 percent of all filers are in the first population decile. A similar statement applies to the deciles for wives and female household heads in Tables II and III.

References

[1.] Advisory Commission on Intergovernmental Relations. Significant Features of Fiscal Federalism 1987 Edition. Washington, D.C.: 1987, p. 80. [2.] Ballard, Charles L., John B. Shoven, and John Whalley, "General Equilibrium Computations of the Marginal Welfare Costs of Taxes in the United States." American Economic Review, March 1985, 128-38. [3.] Browning, Edgar, "A Critical Appraisal of Hausman's Welfare Cost Estimates." Journal of Political Economy, October 1985, 1025-34. [4.] _____, "On the Marginal Cost of Taxation." American Economic Review, March 1987, 11-23. [5.] Cilke, James and Roy A. Wyscarver. The Treasury Individual Income Tax Simulation Model. Washington, D.C.: Office Of Tax Analysis Department of Treasury, 1990. [6.] Feldstein, Martin, "Imputing Corporate Tax Liabilities to Individual Taxpayers." National Tax Journal, March 1988, 37-60. [7.] Fullerton, Don, "Reconciling Estimates of the Marginal Welfare Cost of Taxation." American Economic Review, March 1990, 302-308. [8.] Harberger, Arnold. "Taxation, Resource Allocation, and Welfare," in Taxation and Welfare, edited by Arnold Harberger. Chicago: University of Chicago Press, 1974, pp. 25-62. [9.] Hausman, Jerry A. "Labor Supply," in How Taxes Affect Economic Behavior, edited by Henry Aaron. Washington, D.C.: Brookings Institution, 1981, pp. 27-83. [10.] _____ and James M. Poterba, "Household Behavior and the Tax Reform Act of 1986." Journal of Economic Perspectives, Summer 1987, 101-19. [11.] Killingsworth, Mark R. Labor Supply. Cambridge: Cambridge University Press, 1985, pp. 130-206. [12.] _____ and James J. Heckman. "Female Labor Supply: A Survey," in Handbook of Labor Economics, Volume 1, edited by Orley Ashenfelter and Richard Layard. New York: North-Holland, 1986, pp. 103-204. [13.] Nelson, Susan. "Family Economic Income and Other Income Concepts Used in Analyzing Tax Reform," in Compendium of Tax Research 1987, edited by C. Eugene Steuerle and Thomas Neubig. Washington, D.C.: Office Of Tax Analysis Department of Treasury, 1987, pp. 77-100. [14.] Pechman, Joseph, "The Future of the Income Tax." American Economic Review, March 1990, 1-20. [15.] Pencavel, John, "Labor Supply of Men: A Survey," in Handbook of Labor Economics, Volume 1, edited by Orley Ashenfelter and Richard Layard. New York: North-Holland, 1986, pp. 3-102. [16.] Stuart, Charles, "Welfare Costs per Dollar of Additional Revenue in the United States." American Economic Review, June 1984, 352-62. [17.] Wallace, Sally, Michael Wasylenko and David Weiner, "The Distributional Implications of the 1986 Tax Reform." National Tax Journal, June 1991, 181-198.

During the 1980s in the United States, before-tax income shifted away from the poor and toward the upper income groups. Changing the distribution of tax burdens or after-tax income is one way to redress a growing imbalance in before-tax income. Pechman [14], Feldstein [6], and Wallace et al. [17] have all shown that the Tax Reform Act of 1986 (TRA86) increased the progressivity of the combined personal and corporate income taxes, an achievement not accomplished in the last two decades of income tax reform [14].

Tax reform can also increase or decrease the welfare of individuals in an economy through its effect on the allocation of labor supply resources. TRA86 reduced the highest marginal income tax rates from 50 percent to 28 percent for high income persons and removed some lower-income persons from the tax roles. Such large marginal tax rate changes introduce significant potential for increased economic efficiency in labor supply and for welfare gains for the rich and poor alike. These changes should be part of the benefit analysis of tax reform. Indeed, Hausman and Poterba [10] report welfare cost changes that result from TRA86 for the average male and female. Their estimates, however, do not account for welfare cost changes throughout the income distribution. In addition to our use of microdata to estimate the marginal welfare cost of tax reform, we use more precise measures of marginal tax rates than other researchers.

In this paper, we use Browning's partial equilibrium framework to examine the influence of the reduction in marginal tax rates from TRA86 on the distribution of welfare gains from labor supply reallocation for males by population decile and for females by population decile and by marital status. The empirical results illustrate how the tax reform affected low-income female household heads, where poverty is so heavily concentrated, as well as other household heads and spouses. The results indicate that TRA86 reduced welfare costs in the economy, but these welfare gains were distributed in a pro-rich manner or they grew as income increased for males and for females regardless of marital status. In fact, TRA86 resulted in less allocative efficiency and in welfare losses for individuals in the lowest population decile. Thus, in spite of TRA86's progressive changes in tax burdens, the labor supply welfare gains afforded by TRA86 were oriented toward the highest income individuals and away from the lowest income individuals.

In the next section of the paper, we review several methods of estimating welfare changes from tax reform. We examine Browning's method and explain why we use it for our analysis. We then turn to the empirical values for key parameters of the model and discuss the micro data simulation file that we use to estimate the welfare gains. The results are reported in a fourth section of the paper and are followed by the conclusions that we draw from our work.

II. Marginal Welfare Cost

Hausman and Poterba [10] report changes in welfare costs associated with labor supply decisions of men and wives due to the tax reforms of 1981 and 1986. Using Hausman's [9] indirect utility function and the IRS Public Use Tax Return data file, they compute welfare gains per dollar of revenue change associated with the labor supply of the average male wage earner at 0.03 and of the average wife at 0.055.(1)

More general studies by Stuart [16] and Ballard, Shoven and Whalley [2] use general equilibrium analysis to estimate the marginal welfare cost changes associated with labor supply when tax rates change. They examine the marginal effects for hypothetical tax reforms and use aggregate data. We do not use their general equilibrium analyses, because it would require considerable aggregation, thus forcing us to forgo the benefits of our rich micro data set.

Our analysis is performed in a partial equilibrium framework, as developed by Harberger [8] and modified by Browning [4]. Using welfare formulas, elasticity of labor supply elasticities, and changes in the tax system, welfare cost effects are then calculated for each individual by substituting the relevant compensated labor supply elasticities and tax parameters into the appropriate formula. Fullerton [7, 305] reconciles the general equilibrium approaches cited above with Browning's approach, concluding that, although not without limitations, Browning's approach is useful for comparing tax changes (our problem), because his method measures distortions appropriately.

We employ Browning's [4, 12] modification of Harberger's [8] measure of the partial equilibrium welfare cost.(2) Our primary interest is in the marginal change in welfare cost or in the change in welfare cost divided by the change in tax paid for different groups of individual taxpayers.(3) The computation assumes that the additional expenditures financed by the tax changes have only an income effect.(4) Under these assumptions Browning [4] shows that the change in welfare costs (dW) when tax revenues from labor income increase by dR is given by:

dW/dR = [(m + 0.5dm)/(1 - m)]nw[L.sub.2]dm/[w[L.sub.2]dt + wdL(m + dm)] (1) where

m = marginal tax rate before tax reform,

m' = marginal tax rate after tax reform,

dm = (m' - m),

n = compensated labor supply elasticity,

w = net wage before tax reform,

[L.sub.2]= labor supply before tax reform,

dt = change in the average tax rate due to the reform,

dL = total (income and substitution effects) change in labor supply due to tax reform.

The numerator of equation (1) measures the change in the welfare costs associated with labor as a result of TRA86. The denominator of equation (1) measures the change in the tax revenue derived from labor income as a result of TRA86. The change in the tax revenue in the denominator means that the change in labor (dL) in the denominator represents both the income and the substitution effects for labor supply. On the other hand, the compensated elasticity of labor supply is included in the numerator to measure the change in the welfare cost.

Because we do not have data on dL, it is useful to rewrite (1) in terms of the ordinary labor supply elasticity, [E.sub.s.sup..5.]

dW/dR = [(m + 0.5dm)/(1 - m)]ndm/[dt + (dwl(1 - m)w)[E.sub.s](m + dm)]. (2)

Moreover, the expression dw/(2 - m)w can be written as:

[(1 - m')w - (1 - m)w]/(1 - m)w = dm/(1 - m).

Equation (2) then becomes:

dW/dR = [(m + 0.5dm)/(1 - m)]ndm/[dt + (-dm/1 - m)[E.sub.s](m + dm)]. (3)

Equation (3) is used to estimate the marginal welfare effects of the 1986 tax reform.

Unfortunately, the marginal welfare formula as derived by Browning [4] has the same sign for both an increase and a decrease in welfare.(6) For example, the formula produces a positive sign if marginal tax rates increase, which represents a welfare loss. More specifically, when the marginal tax rate increases then the term (dm) is positive, making the numerator positive. The denominator is simply the extra revenue that is collected when the marginal tax rate increases. Unless the amount of taxes an individual pays decreases when his or her marginal tax rate increases (an unlikely strong labor supply response), the denominator will also be positive when marginal tax rates increase, even though the sign of the denominator is theoretically ambiguous. Similar logic applies to a decrease in the marginal tax rate. In that case, the sign of (dm) and, therefore, the numerator will be negative, and the change in revenue (dR) will also generally be negative when marginal tax rates are reduced.

To solve the sign problem, we note that the sign of the numerator more meaningfully represents the welfare change, while the denominator only standardizes the change in welfare costs per dollar of change in taxes. For most individuals in the data set, the numerator and the denominator have the same sign. In that case we took the absolute value of the denominator, so that the marginal change in welfare cost expression took on the sign of the numerator or the indicator of change in welfare costs. Therefore, a marginal welfare gain has a negative sign, as negative dm means that marginal tax rates decline, and a marginal welfare loss has a positive sign, as a positive dm means that the marginal tax rates increase. It is possible, however, for an individual's marginal tax rate to decline and the change in revenue to increase. In this case, the substitution effect is greater in absolute size than the income effect and the individual chooses to work more as a result of increased wages. These relatively few cases are treated as a marginal welfare gain resulting from the reduction in the marginal tax rate.(7)

After calculating the marginal welfare costs of TRA86 for each taxpayer, the taxpayers are grouped into population deciles. The results of the average change in marginal welfare cost per taxpayer in each decile are presented for males, female spouses and female household heads, in three separate tables.

III. Values of Key Parameters

The above formula for excess burden or marginal welfare cost requires measures of the marginal tax rate before and after tax reform, the change in the average tax rate, the compensated elasticity of labor supply, and the ordinary elasticity of labor supply.(8)

Marginal and Average Tax Rates

We calculate the federal personal marginal tax rate for each individual using both the 1986 (pre-reform) and the 1991 (post-reform) tax structure using the U.S. Treasury Individual Income Tax Model (ITM). The micro data identifies each filer's state of residence and we supplement the federal tax rate with the state income tax rate for each filer. We also include, as part of the marginal tax rate calculation, the combined employee and employer social security payroll tax rate when wage income for the individual is below the social security ceiling and a measure of the average sales and excise tax rates assuming the taxes are borne by factors of production.

To complete the calculation of the marginal tax rate, we account for state and federal deductibility of federal and state income taxes, respectively.(9) While other researchers include state taxes in their marginal tax rate calculations, no one accounts for deductibility of federal income taxes from state income taxes and for the deductibility (and lack thereof post TRA86) of sales taxes in computing marginal and average tax rates.

The average tax rate is the tax paid on labor income divided by pre-reform labor income. The change in the average tax rate (dt) is based on the difference between the average tax rate that would exist under the 1991 tax structure and that in existence before the 1986 reform.(10) In order to focus on the effects of the individual income tax, we hold the social security payroll tax rates at their 1986 levels.

Elasticities of Labor Supply

Equation (3) indicates that the computation of welfare cost changes requires estimates of both compensated and uncompensated labor supply elasticities. In his survey of labor supply behavior for men, Pencavel [15, 3-102] puts the mean value of the compensated elasticity of male labor supply at 0.11. The actual range of compensated male labor supply elasticity estimates in the studies cited by Pencavel is from - 0.19 to 0.84.(11)

Killingsworth [11] and Killingsworth and Heckman [12] review comprehensively the empirical estimates of the elasticity of labor supply for females. We use the elasticity estimates from both in our work. The estimates reported in the two papers have significant overlap. Female household heads have compensated elasticities of between 0.16 and 0.77. The compensated elasticity estimates for wives are between 0.11 to 4.73.(12)

Based on the figures above, there are no definitive estimates of compensated labor supply elasticities. Thus, most excess burden studies calculate welfare costs based on a range of elasticity values. We use 0.11, 0.2, and 0.55 for compensated elasticities of labor supply for males. These elasticities come from Pencavel's [15, 69] reported mean value (.11), and two compensated elasticity point estimates that he reports. For females, we use a broader range of values for the compensated elasticity of labor supply. For spouses, we use elasticities of 0.2, 0.4, 0.8, 1.0, 1.5 and 2.0. For female household heads, we use elasticities of 0.3, 0.5, and 0.8. These figures seem to us to be somewhat on the low side of the accepted range. Hence, our estimation of the change in the excess burden due to tax reform will tend to be conservative.

For ordinary labor supply elasticity, we use Pencavel's [15, 69] reported mean of - 0.12 for males. There is considerably less agreement about the female ordinary labor supply elasticity. Killingsworth and Heckman [12, 195] point out that many studies report ordinary labor supply elasticties for wives in the range of 0.5 to 1.0, with a few studies reporting much higher elasticities. They suggest that female labor supply decisions require further study. Hausman [9] is one of the few studies of U.S. female labor supply that estimates an elasticity for wives as well as an elasticity for female household heads. As his ordinary labor supply elasticity estimates are within the range of those produced by recent studies, we use his estimates of ordinary labor supply elasticities for wives of about 1.0 and for female household heads of about 0.5.

Data and Simulation

We use micro data from the U.S. Treasury Individual Income Tax Model (ITM) to estimate the welfare cost changes for individuals that result from TRA86.(14) The ITM simulates the tax system in 1986 and 1991 (the first year that most provisions are fully phased in), at 1986 levels of income. The Treasury ITM is a microsimulation model, based on the Internal Revenue Service 1985 Statistics of Income (SOI) file. The SOI tax data are supplemented with information from the U.S. Census, Current Population Survey. A series of extrapolation weights are available, which allow the data to be extrapolated beyond 1985 [5].

The data are used to estimate the marginal excess burdens of tax reform and the estimates are then grouped into population deciles. We use a household economic income concept, which is a comprehensive income definition that begins with adjusted gross income (AGI) and is then augmented with imputations for capital income and transfer income receipts.(15)

IV. Results

Table I reports the average change in welfare cost per dollar of tax paid for males in each population decile. Tables II and III report the results for female spouses and for female household heads, respectively.(16)

Based on the results reported in Table I, individual income tax reform on average reduced the welfare costs of labor supply for males in the fourth through the tenth population deciles. The size of the marginal welfare cost gains range between - 0.02 and - 3.34. The largest changes in welfare costs per dollar of change in individual taxes are with some exceptions in the population deciles with the highest incomes. The fact is that males in the highest income population deciles experienced the largest reductions in marginal tax rates. Males in the three income population deciles with the lowest income have small welfare cost increases due to small increases in their marginal tax rates.

[TABULAR DATA OMITTED]

The results reported in Table II suggest that, except for those in the first and second population deciles, individual income tax reform reduced the welfare costs associated with female spouse labor supply. Based on the overall average, the average welfare cost reduction for female spouses ranged between - 0.28 and - 2.80. Although the range of compensated elasticities is much larger than that for males, the marginal welfare cost reductions are not larger than that for males when comparisons are made for similar compensated labor supply elasticities. One reason for the relatively small welfare cost reduction is the elimination of the dual wage-earner tax deduction. Its elimination has the effect of raising the average tax rates and, in some cases, the marginal tax rates for wives. The largest average marginal welfare cost reductions are in the eighth and ninth population deciles.

[TABULAR DATA OMITTED]

The results for female household heads reported in Table III show that TRA86 also reduced their welfare costs on average in every population decile, except the first decile, and the marginal welfare cost gains were on average higher than those for males. The marginal welfare cost gains were between - 0.01 and - 5.53 and the largest average gains occurred in the ninth decile.

In general the magnitudes of the marginal welfare costs are large as suggested in other studies of marginal welfare changes by Browning [3], Stuart [16] and Ballard, Shoven and Whalley [2]. However, the results in this paper suggest that TRA86 lead to different marginal welfare cost reductions for wives, female household heads and males. Such diverse results cannot be obtained from aggregate data, and our results demonstrate the benefit of using micro rather than aggregate data to judge the welfare benefits of tax reform.

V. Conclusions

The findings in this paper have important implications for future tax reform. The reduction of tax rates is rarely done across-the-board, and marginal welfare cost changes from tax reform can differ substantially among population deciles. More importantly, the elimination of certain provisions from the tax code, such as the dual wage-earner deduction, can have unintended marginal welfare cost effects on certain wage-earners. In the case of the dual wage-earner deduction, its elimination has significantly reduced the marginal welfare cost gains for wives compared to the marginal welfare cost gains for males. The marginal welfare cost gains of TRA86 are generally distributed in a pro-rich manner, for all types of households, although the distribution of marginal gains is more pro-rich for males and for wives than for female-headed households.

The above findings also point to the importance of examining the marginal welfare cost effects using micro data. Marginal excess burden results based on aggregate data can only give an overall impression of what changes in marginal tax rates would imply for welfare costs. However, tax reform rarely involves so simple a change in tax rates. Eliminating certain provisions in the tax code can increase the average tax rate for certain classes of taxpayers. These unintended consequences cannot be gleaned from the results of aggregate studies. The tax reform debate could be misinformed, if the basis for discussion rests exclusively on the results from aggregate studies, as results from aggregate studies are not intended to, nor should they, guide all policy on tax reform.

(1.) Browning [3] points out the problems with Hausman's specification. (2.) Browning [3, 12] modifies the Harberger measure by valuing the variables in the change in welfare cost formula at after-tax wages and labor supply rather than at before-tax wages and labor supply. (3.) One limitation in following this approach is the assumption of linear compensated supply schedules. For some individuals with large changes in their marginal tax rates the linearity assumption may be less than a precise measure of their marginal welfare gain or loss. The linear compensated supply curve assumption may mean that we have only estimated an approximate welfare change for the relatively few individuals that have had a large reduction in their marginal tax rates. See Browning [4, 17]. (4.) Here we use Browning's [4] equation (9) for measuring marginal welfare costs. Although Browning presents three formulas for measuring marginal welfare costs, each formula makes an altemative assumption about the effects of the incremental government expenditure resulting from the hypothetical tax increase. His other two formulas (equations (10) and (11) in his paper) make polar assumptions about the marginal effects of government spending on individual utility levels. See Browning [3].

(5.) Browning's formulation of the marginal change in welfare costs does not assume that the tax reform affects labor income only. In particular, the term (dL) in the denominator would in principle include both the substitution effect and all changes in individual labor supply due to income effects from tax reform. We do not, however, have data for w and for dL for the denominator, and, therefore, cannot account for all changes in dL due to tax reform. In changing Browning's equation and using the labor supply elasticity in equation (2) of the text, we are assuming that tax reform's change in nonlabor income did not affect labor supply. While tax reform did affect non-labor income, most of the effect of tax reform on non-labor income was in the highest income brackets. We believe that changes in the taxation of capital income at high income levels do not affect labor supply decisions, and offer it as a partial defense of our use of the modified equation (2). In our calculation of the change in marginal welfare costs due to TRA86, however, we use the change in the marginal and average tax rates that result when all of the changes in capital and labor income are taken into account in the tax base. In addition, tax reform affected the marriage tax. We account for changes in tax revenue due to the higher marginal tax rate for the second wage earner. But we do not capture the income effect from the change in the second wage earner's marginal tax rate on the labor supply of the primary wage earner. (6.) Browning's problem was more straightforward as he deals with a positive change in aggregate revenue as a result of his hypothetical tax reform. (7.) Our data reveal only a few cases where the decline in the marginal tax rate led to a positive change in taxes. While we agree with one reviewer that there are interesting policy implications here, confidentiality of the data prevents us from revealing the exact magnitude of these effects. (8.) The use of micro-level data to calculate the excess burden for each individual eliminates the calculation of the average weighted marginal tax rate for labor necessitated in aggregate studies. One reviewer requested that we report wages and marginal tax rates by population decile to facilitate research using general equilibrium models. Unfortunately, the Department of Treasury will not permit us to report those figures calculated from the ITM model. (9.) In 1986, the following 12 states allowed full deductibility of federal income taxes from their state income tax: Alabama, Arizona, Colorado, Iowa, Kansas, Kentucky, Louisiana, Minnesota, Missouri, North Dakota, Oklahoma and Utah. Importantly, these states allow the deduction of federal tax liability even when taxpayers do not itemize deductions for state tax purposes. Thus, all taxpayers in these states deduct the federal tax liability from their state income, and taxpayers in these 12 states have their marginal federal income tax rate reduced by one minus the state income tax rate. In addition, Oregon taxpayers can deduct their federal tax liability from state income but the deduction is limited to $7,000 [1]. Some states also allow a deduction from state income taxes for social security payroll taxes. We take this deduction into account and also account for TRA86's changes in the ruies governing the deductibility of sales taxes in the computation of marginal tax rates. Due to a lack of data on transfer programs, we do not account for marginal tax rates of transfer programs, but we do account for the marginal tax rates from the Earned Income Tax Credit. (10.) To incorporate the 1991 fully phased-in tax laws into the analysis, we extrapolate the data to 1991 using the ITM to calculate the marginal and average tax rates. Note, however, that we have not incorporated behavioral effects associated with the change in the marginal tax rate on capital gains. Theory suggests that an increase in the tax rate on gains leads to a decrease in realizations. This would lead to a lower tax payment for a given tax rate. Since this behavior is not incorporated, returns that realize gains may have a higher level of tax than they would if realizations actually fell. This may bias the change in average tax, which would be greater for individuals whose level of realizations is affected by changes in the tax rate. The complications added by such behavior are not modeled here. (11.) Killingsworth [11] suggests that the range of positive compensated elasticity of male labor supply estimates is between 0.08 and 1.O. (12.) The availability of labor supply elasticity estimates constrains the extent to which we can estimate marginal welfare costs for racial and age subgroups of males and females. For example, the two studies that split the sample between black wives and white wives do not find a consistent pattern of differences in the compensated elasticities between black and white wives, and, therefore, we do not attempt to estimate excess burdens by race. A few studies report estimates for females in different age groups and for female household heads, for black female household heads and for wives. The studies do not, however, estimate separate elasticities for household heads by age, race and household group. Generally speaking, most data sets would not have enough observations in most cells to estimate labor supply elasticities for age groups by household headship by race. (13.) Households with two wage earners may make labor supply decisions simultaneously. We do not account for the simultaneity in this analysis. (14.) We consider the effects of changes in individual taxes only, and do not attempt to measure the welfare effects of corporate income tax changes. (15.) For a more complete description of economic income, see Cilke and Wyscarver [5] and Nelson [3]. (16.) The population deciles and the respective income ranges are defined using all income tax returns and economic income. Therefore, the income range is the same for any given population decile in the three tables. The implication, of course, is that 10 percent of the male filers are not necessarily in the first population decile of Table I, but that 10 percent of all filers are in the first population decile. A similar statement applies to the deciles for wives and female household heads in Tables II and III.

References

[1.] Advisory Commission on Intergovernmental Relations. Significant Features of Fiscal Federalism 1987 Edition. Washington, D.C.: 1987, p. 80. [2.] Ballard, Charles L., John B. Shoven, and John Whalley, "General Equilibrium Computations of the Marginal Welfare Costs of Taxes in the United States." American Economic Review, March 1985, 128-38. [3.] Browning, Edgar, "A Critical Appraisal of Hausman's Welfare Cost Estimates." Journal of Political Economy, October 1985, 1025-34. [4.] _____, "On the Marginal Cost of Taxation." American Economic Review, March 1987, 11-23. [5.] Cilke, James and Roy A. Wyscarver. The Treasury Individual Income Tax Simulation Model. Washington, D.C.: Office Of Tax Analysis Department of Treasury, 1990. [6.] Feldstein, Martin, "Imputing Corporate Tax Liabilities to Individual Taxpayers." National Tax Journal, March 1988, 37-60. [7.] Fullerton, Don, "Reconciling Estimates of the Marginal Welfare Cost of Taxation." American Economic Review, March 1990, 302-308. [8.] Harberger, Arnold. "Taxation, Resource Allocation, and Welfare," in Taxation and Welfare, edited by Arnold Harberger. Chicago: University of Chicago Press, 1974, pp. 25-62. [9.] Hausman, Jerry A. "Labor Supply," in How Taxes Affect Economic Behavior, edited by Henry Aaron. Washington, D.C.: Brookings Institution, 1981, pp. 27-83. [10.] _____ and James M. Poterba, "Household Behavior and the Tax Reform Act of 1986." Journal of Economic Perspectives, Summer 1987, 101-19. [11.] Killingsworth, Mark R. Labor Supply. Cambridge: Cambridge University Press, 1985, pp. 130-206. [12.] _____ and James J. Heckman. "Female Labor Supply: A Survey," in Handbook of Labor Economics, Volume 1, edited by Orley Ashenfelter and Richard Layard. New York: North-Holland, 1986, pp. 103-204. [13.] Nelson, Susan. "Family Economic Income and Other Income Concepts Used in Analyzing Tax Reform," in Compendium of Tax Research 1987, edited by C. Eugene Steuerle and Thomas Neubig. Washington, D.C.: Office Of Tax Analysis Department of Treasury, 1987, pp. 77-100. [14.] Pechman, Joseph, "The Future of the Income Tax." American Economic Review, March 1990, 1-20. [15.] Pencavel, John, "Labor Supply of Men: A Survey," in Handbook of Labor Economics, Volume 1, edited by Orley Ashenfelter and Richard Layard. New York: North-Holland, 1986, pp. 3-102. [16.] Stuart, Charles, "Welfare Costs per Dollar of Additional Revenue in the United States." American Economic Review, June 1984, 352-62. [17.] Wallace, Sally, Michael Wasylenko and David Weiner, "The Distributional Implications of the 1986 Tax Reform." National Tax Journal, June 1991, 181-198.

Printer friendly Cite/link Email Feedback | |

Title Annotation: | Tax Reform Act of 1986 |
---|---|

Author: | Wasylenko, Michael |

Publication: | Southern Economic Journal |

Date: | Jul 1, 1992 |

Words: | 4736 |

Previous Article: | Motor carrier deregulation and highway safety: an empirical analysis. |

Next Article: | Price discrimination with correlated demands. |

Topics: |