Marginal tax rates facing low- and moderate-income workers who participate in means-tested transfer programs.

"The possible work disincentives and other efficiencies created by the existence of multiple transfer programs is one of the most important, and one of the oldest, issues in the economics of transfer programs in the U.S.A."

--Keane and Moffitt (1998)

"It's like being persecuted for getting a raise."

--"Karen H." A Wisconsin home health care worker, upon finding out that her food stamps and housing assistance were reduced after she received a legislatively-mandated hourly wage increase (Romich, 2006)

INTRODUCTION

Two trends--the increasing use of the tax system for social policy ends and the decoupling of cash welfare from other means-tested programs--have combined to complicate the budget constraints facing low-income working families. Researchers, practitioners and policymakers concerned with the relatively high marginal tax rates (MTRs) created by the phaseout of the Earned Income Tax Credit (EITC) and other transfer programs embedded in the income tax system should also be aware of the implicit tax rates created by the benefit schedules of common social transfer programs, such as Food Stamps.

Over the last two decades, the Internal Revenue Service and state revenue departments are increasingly serving as delivery systems for social policies aimed at increasing the well-being of low- and moderate-income (LMI) households (Dickert-Conlin, Fitzpatrick, and Hanson, 2003). Over the same time period, eligibility rules governing common means-tested programs have been reformed to include families for whom earnings are a significant source of income. As of the early 1980s, eligibility for common means-tested social programs, notably Medicaid, was restricted to members of households receiving cash welfare (what was then known as Aid to Families with Dependent Children, AFDC). The Family Support Act of 1988 de-linked Medicaid from cash assistance. The Personal Responsibility and Work Opportunity Reconciliation Act of 1996 (PRWORA) enacted further reforms guided by the principle that earnings from employment rather than welfare benefits should raise family income (Haskins, 2002). Reforms aimed at supporting employment and reducing the work disincentives found in the AFDC system included a decentralized combination of time-limited grants with earnings disregards administered by state Temporary Assistance to Needy Families (TANF) programs, and subsidized benefits including TANF child care grants and expansions of public health insurance, notably the loosening of eligibility for Medicaid and the creation of State Children's Health Insurance Programs (SCHIP) (Sawhill and Haskins, 2002). The cumulative effect of tax schedules and the implicit taxation arising from the benefit schedules of other commonly used social transfer programs means that LMI households may routinely face combined MTRs well over 50 percent.

Calculating MTRs

Knowing the size and distribution of such high MTRs can help researchers and policymakers understand potential effects on labor supply, the relationship between earnings and financial well-being for LMI households, and the possible impact of changes to individual programs to the overall system of support. For a household of a given size and program participation status, calculating the combined tax schedule is a straightforward--if cumbersome--process of applying tax provisions and the benefit formulas of the relevant programs. This requires specifying the one or more combinations of the key parameters of family structure, earnings and program participation. For instance, Acs, Coe, Watson, and Lerman (1998) presented effective MTRs for one-parent, two-child households at four different levels of employment (none, part time at minimum wage, full time at the 1997 federal minimum wage of $5.15, and full time with a $9.00 wage). Assuming the household receives food stamps, files state and federal tax returns, and, when applicable, uses TANF cash benefits, they find that the raise in earnings associated with moving from minimum wage to $9.00 per hour is associated with an effective MTR of 65 percent in the median state in their 12-state sample. Similar analyses have calculated MTRs for different time periods, geographical areas, household types or benefit use packages (Wilson and Cline, 1994; Shaviro, 1999; Ellwood and Liebman, 2001; Sammartino, Toder, and Maag, 2002; Wolfe, 2002). These analyses, while informative as to the budget constraints faced by the exemplar families, do not capture the actual distribution of MTRs within a population with varying patterns of participation in means-tested programs and heterogeneous household structures and earning levels. Assuming that households use all available benefits is not realistic and likely leads to overestimates of the MTRs facing LMI workers.

Calculating the actual distribution of MTRs faced within a population requires combining program rules and tax schedules with information on individual households' program and tax participation, family structure, and earnings. Creating such estimates requires household-level tax and program participation data, a data need not fulfilled by widely available data sets (Dickert-Conlin, Fitzpatrick, and Hanson, 2005). Common secondary data sources can be used if tax information and some benefit amounts are imputed. For instance, Dickert, Houser and Scholz (1994) used 1990 Survey of Income and Program Participation (SIPP) data in a microsimulation of the marginal return to increased earnings. This approach requires assuming that self-reported income is the same as reported taxable income and is sensitive to misreporting of program participation, which is a significant and arguably growing problem with the SIPP. (1) Administrative data are more accurate, but they do not include both tax and program information. Internal Revenue Service data do not contain information on program participation data. Using quality control data from the Department of Health and Human Services (for AFDC/ TANF) or the Department of Agriculture (for Food Stamps) requires imputing or relying on self-reports for EITC participation.

To our knowledge, the current study is the first use of merged tax and program administrative data to calculate a population distribution of MTRs. A unique data set consisting of state human services caseload data merged with Unemployment Insurance and Department of Revenue data allows us to examine program participation and the resulting effective tax rates for substantially all Wisconsin LMI households with earned income in 2000. As background, we begin with an overview of program and tax schedules affecting LMI households. After describing the process used to assemble our data, we present descriptive evidence on the patterns of joint participation in the tax credits and transfer programs. By applying statutory benefit formulas and tax schedules, we then calculate actual MTRs. We conclude with a discussion of the implications of our findings of a significant incidence of high MTRs among some LMI households.

MTRs ARISING FROM TAX AND BENEFIT SCHEDULES

In this section, we describe the eligibility formulas and schedules governing the MTRs in the tax and transfer systems. We are looking here at hypothetical maximum statutory tax rates, meaning we assume that each household receives all benefits for which it is eligible. We also assume that readers are familiar with the basic structure of the federal income and payroll tax systems and the common means-tested programs covered here. For those interested in greater detail, Scholz and Levine (2002) summarize recent changes in and the current state of programs for the poor, and Hassett and Moore (2005) review tax policies affecting LMI households. Throughout this exposition we use a single parent with two children as the standard household configuration in examples, both because this household is frequently referenced in anti-poverty policy debates and because its circumstances illustrate the issues well.

Our data are from Wisconsin in 2000, so we describe the relevant features of its state income tax system and use statutory rates and rules for that year. Although it was an early adapter of welfare reform, the set of supports offered in 2000 is largely comparable to that offered by other states (Wolfe, 2002). In terms of income and population dimensions that affect the use of means-tested benefits, Wisconsin is reasonably representative. In 2000 Wisconsin had a slightly higher median income than the US average, lower poverty rates, and less unemployment, suggesting that the state is marginally more economically advantaged. (2) However, the state did have an almost-average portion of the population under age 18, a feature that is relevant since many of the means-tested benefits are most generous to households with children. (3)

Federal Payroll Taxes, Income Tax, and Tax Credits

Although each dollar of earnings is subject to federal income tax, personal exemptions and the standard deduction operate to shield some low-income families from a positive tax liability. In 2000, the tax threshold was $14,850 for a single parent with two children. Through a taxable income (income less personal exemptions and the standard deduction) of $35,150, the tax rate was 15 percent. Taxable income above that level (up to $90,800) was taxed at 28 percent.