What Tax Researchers Can Learn from Psychology

A splendid new paper sums up some of the interesting developments in the field known as behavioral public finance (I wrote about this in my AFR column recently). The abstract is below, and some choice snippets are over the fold.


Behavioral Economics and Tax Policy (gated stable link, ungated unstable link)
William Congdon, Jeffrey R. Kling, Sendhil Mullainathan

Abstract: Behavioral economics is changing our understanding of how economic policy operates, including tax policy. In this paper, we consider some implications of behavioral economics for tax policy, such as how it changes our understanding of the welfare consequences of taxation, the relative desirability of using the tax system as a platform for policy implementation, and the role of taxes as an element of policy design. We do so by reviewing the logic of specific features of tax policy in light of recent findings in areas such as tax salience, program take-up, and fiscal stimulus.

We argue that the implications of behavioral economics—the integration of economics and the psychology of preference formation and choice—for public policy, including tax policy, have yet to be systematically explored, and that this oversight leads to both mistaken policy and missed opportunity. Behavioral economists have now accumulated several decades of findings indicating that the standard economic assumptions about individual behavior are not accurate, that people do not act rationally, that they are not perfectly self-interested, and that they hold inconsistent preferences. Moreover, and especially in recent years, policy economists have increasingly come to see that these deviations from the standard assumptions about behavior matter for economic policy. …

While the traditional case for tax simplicity is indirect, related to achieving broad tax bases, or administrative, the behavioral approach suggests that the degree of simplicity directly enters the optimal tax calculation. There are at least three elements of this added nuance. First, complexity can affect behavioral responses to otherwise economically equivalent taxes. Second, complexity can be used as a screening mechanism to promote efficiency. Third, complexity may aid attainment of social goals. …

When policymakers can choose to keep some taxes hidden from consumers, this will keep the elasticity low, which, other things equal, is desirable for efficiency purpose. But with a binding budget constraint, spending too much on the good with a hidden tax will leave less income for subsequent purchases—distorting individual consumption and decreasing welfare. Resolving the net welfare consequences of tax salience is thus an important future line of research. …

Complexity, screening and welfare. A related but separate consideration appears allowing for varying degrees of sophistication about tax complexity. Take the following example: social returns to education are high, and as a result a tax credit for education is desirable. Assume, however, that variation in the social return to education across individuals is correlated with their being savvy about tax complexity. To make the example simple, assume two types: one type has social return and is savvy, and the other type has only private return and is not savvy. With a simple tax credit, then everyone participates and the government subsidizes the type that only generates private return as well as the social return type. With a complicated credit only the savvy respond, and the subsidy is provided only to those who generate a social return, improving the targeting efficiency of the subsidy.

Complexity, social goals, and welfare. Finally, complexity and imperfectly rational behavior can combine to create new opportunities for tax policy to achieve social goals. One situation where complexity can create such an opportunity is when the complexity of a tax leads individuals to respond to a tax in error, but that error leads individuals to respond in a way that fits with policymakers’—and possibly society’s—preferences. For example, the result that the EITC strongly encourages work may be due in part the fact that the EITC follows a complicated schedule, to which recipients can bring only limited attention and limited computational resources.

What evidence is available does suggest that EITC eligible individuals do possess only imperfect knowledge of how the credit operates (Smeeding, Ross Phillips and O’Connor 2000; Phillips, 2001; Chetty and Saez, 2009). The evidence of the impact of the EITC on labor force participation and hours worked is consistent with a model in which individuals only understand the work incentives of the EITC in some approximate sense. For instance, an increase in labor force participation with little effect on hours worked could be a result of workers understanding the average effects of the EITC on earnings but not the marginal effects—a variant of “schmeduling”, in the terminology of Liebman and Zeckhauser (2004).

The key is to note that, while this is an error on the individual level, from society’s perspective, this may be thought of as desirable. It may even be possible to combine schmeduling with hidden taxes so that individuals choose in ways that incorporate income effects (from schmeduling) while reducing distortions from substitution effects (from hidden taxes). …

The attractiveness of the automaticity of the tax system comes from the fact that accumulating evidence suggests that barriers to taking up programs, even minor barriers such as application costs or waiting times, can discourage program participation out of proportion to the magnitude of the costs they impose (Bertrand, Mullainathan, and Shafir 2006). That human frailties—procrastinating filing a form, or being put off by the tediousness or hassle of completing it, or failing to understand program rules—can lead qualifying individuals to forgo benefits.

Moreover, there is a concern in such an environment that program non-participants are then not those who value the program the least but instead those who understood the rules the least or faced the biggest procrastination problem. Such barriers to take up may induce an unattractive selection in program participants. In some cases, such as transfer programs, those most put off by complex or burdensome program application requirements might be the very target population. … Estimates suggest that only about 67 percent of eligible individuals receive food stamps, while the EITC may reach as many as 94 percent of eligible households …

Finally, the use of taxes to achieve fiscal policy ends is a good, and timely, example of using taxes as a tool of policy. A goal of fiscal policy is to achieve macro stabilization—in the current crisis the task has been to stimulate the economy in the face of slack aggregate demand. … 

Delivering tax cuts. Another lesson is that how tax cuts are delivered can affect whether and how they are spent. The most recent stimulus bill, for example, the American Recovery and Reinvestment Act of 2009, rather than provide a lump sum tax cut (either as a rebate or a bonus) delivered its primary tax cut in the form of reduced withholding. The idea being that tax cuts that reduce withholding might be more effective than lump sum rebates. This was based in part on past experiences that suggest a greater willingness to spend out of reduced withholding, such as shown in Shapiro and Slemrod (1995), which found a propensity to spend out of a reduction in withholding rates even in the absence of a change in the tax rate. And also on a recent laboratory experiment done by Chambers and Spencer, which found subjects were more likely to plan to spend a hypothetical tax cut delivered as many small payments rather than one delivered as a lump sum (2008).

4 thoughts on “What Tax Researchers Can Learn from Psychology”

  1. Come on, if you’re introducing new words into the language we need more than a cross-reference. What does “schmeduling”, in the terminology of Liebman and Zeckhause, mean?As is, it sounds like a Yiddish gesture of disdain: “Scheduling, schmeduling, as long as there’s time to ring his mother,” or some such ethnic joke.


  2. The screening complexity argument has in fact a long history in the design of welfare payments. The idea is that if you make people jump a lot of hoops to get the payment then they’ll self-select: only those who really neeed the payment will jump.

    I actually don’t like this idea much because I think unwillingness or inability to jump is only loosely correlated with need and it is closely correlated with other things (illiteracy, fear of officialdom, etc) that in fact are correlates of need.

    The point is, though, that this sort of insight into the screening effects of complexity is not new.


  3. ChrisB, Andrew is only quoting the paper with the reference to the schmeduling paper – he didn’t use the term himself. But the Liebman and Zeckhauser paper is anyway well known among welfare economists. And 10 seconds of googling locates a publicly accessible copy here.


Comments are closed.