Recent Marginal Labor Income Tax Rate Changes by Skill and Marital Status

         The economy experienced an unusually deep and prolonged contraction, especially in its labor markets.  Employment and hours worked fell during 2008 and 2009 for many demographic groups, but disproportionately so among less skilled people, and among the unmarried.  As of 2012, labor market activity still remained far below pre-recession levels.  Over the same time frame, many facets of fiscal policy were changed, especially policies related to the distribution of safety net program benefits.

            Fiscal policymakers were of course watching the economy closely, and major safety net legislation was certainly a reaction to economic conditions.  But unless behavior is completely unresponsive to tax and benefit formulas, we cannot have a full understanding of the relationship between fiscal policy and the economy without quantifying marginal tax rates, their changes over time, and their differences across demographic groups.  The purpose of this paper is to help examine the labor market impacts of recent changes in safety net programs by measuring time series of implicit marginal labor income tax rates for the safety net as a whole and recognizing that marginal tax rates and their changes vary by demographic group.  In this regard, the paper is a longitudinal version of prior studies appearing in Tax Policy and the Economy that showed how implicit marginal income tax rates vary with household income and other characteristics under a single year’s tax and benefit rules.[1]

            Analysis of implicit marginal tax rates and their differences across groups might seem to be a specialized topic of poverty research, and only relevant for macroeconomic analysis to the extent that the economy is populated by poor people.  However, this paper explains how people from the middle and above-middle parts of the skill distribution can become eligible overnight for safety net programs such as unemployment insurance and now SNAP (formerly known as food stamps) merely by becoming unemployed for a period of time.  Thus, even skilled people have their incentives to seek and retain work determined in part by safety net program rules.  This paper shows that wide swaths of the skill distribution saw their marginal tax rates increase by more than five percentage points in less than two years.

            I consider the entire safety net, as well as payroll and income taxes, but give most of my attention to three programs spending the largest amounts on non-elderly households, and with significant legislative changes: unemployment insurance, SNAP (formerly known as food stamps), and Medicaid.[2]  I focus on the non-elderly population because the elderly have access to a different set of safety net programs and have a different (and weaker) relationship with the labor market.

My marginal tax rate concept is a comparison of the total amount of subsidies net of taxes received if and when a person were not working to the total amount received (or paid) if and when the same person were working full time, expressed as a fraction of the amount produced when working full time.  This measure is a marginal tax rate on the decision margin of working full time or not at all during a specific time interval.  In this regard, my tax rate concept is reminiscent of the implicit tax rates used by Gruber and Wise (1999) and collaborators in their “tax force” measures of the retirement incentives created by public pension and disability programs around the world.[3] 

            The relationship between subsidies received when not working and the amount that could be earned when working full time varies by demographic group.  A number of subsidies are set as specific dollar amounts (such as the SNAP maximum benefit, or the maximum unemployment insurance benefit), or as a specific bundle of services (as with Medicaid) regardless of how much the beneficiary might earn if he worked full time.  Unemployment insurance benefits below the maximum are, on the other hand, specified as a proportion of the amounts earned in prior employment.

Subsidies received by both employed and non-employed people also depend on the income of others in the household, and therefore can vary significantly by marital status.  Finally, subsidy rules have changed over time as new legislation was passed, and calendar time triggered new provisions in old legislation.  This paper therefore calculates marginal tax rates as a function of earnings potential, marital status, and calendar time.

The paper begins with its conceptual framework for measuring work incentives implicit in the composite of programs known as the safety net, with emphasis on isolating groups-specific changes over time that come from changes in safety net program rules rather than changes in the behavior of the population served under a fixed set of rules.  I then identify the changes in safety net eligibility and benefit rules that were significant for the non-elderly population.  The paper concludes with marginal tax rate series that combine the program-by-program results of earlier sections and a first indication of how changes in these rates were correlated with labor market behavior.

One point of view is that the labor market is slack during a recession, that as a consequence labor supply has nothing to do with labor market outcomes, and that household marginal tax rate calculations are of no help in understanding how people behave during recessions.  Even if this conception of slack markets were accurate, marginal tax rate changes are relevant because they tell us where the labor market is headed after it is no longer slack and supply incentives start to matter again.  More important, both theory and evidence might support the opposite point of view: that subsidies for the unemployed and the poor matter as much or more during a period of significant labor market distortions, in part because those subsidies loom large relative to low offer wages.[4]  The conclusion of this paper also presents a puzzle for the theory that labor supply has recently been irrelevant: that 2007-10 changes in work hours per capita correlate so closely across demographic groups with statutory changes in the incentives to work.

---

[1] See Liebman (1998) and Kotlikoff and Rapson (2007).

[2] Because subsidy program participation is voluntary, a subsidy cannot have a large effect on incentives or behavior unless it redistributes a significant amount of resources.  A tax program, on the other hand, can in principle create large marginal tax rates without redistributing much revenue.

[3] My tax rate is equivalent to a weighted average of (one minus) the local slopes of a worker’s budget constraint (in a graph of disposable income versus earnings from work), where the weights are the size of the income interval over which each local slope applies, because my tax rate is one minus the slope of the straight line connecting the no work point of the budget set to the full-time work point.  In this regard, some might say that my tax rate measures “extensive margin” incentives, which are a weighted average of “intensive margin” incentives.  However, readers should recognize that the concepts of intensive and extensive margins have a time dimension, which varies across studies and safety net programs.  For example, the decision whether to work in the month of August is an extensive margin decision from the point of view of a program that monitors beneficiaries’ labor income on a monthly basis, but an intensive margin decision from the point of view of a program such as the Earned Income Tax Credit that monitors income on a calendar year basis (unless not working in all of the other eleven months of that year).

[4] Mulligan (2012) reviews evidence on the cyclicality of labor market and production effects of marginal changes in labor supply and demand.  He finds that labor supply and labor demand shifts had essentially the same marginal effects on employment in 2008 and 2009 as they did in prior years.