Tax policy colloquium, week 11: Julie Cullen's "Political Alignment and Tax Evasion"

This past Monday, Julie Cullen presented her (coauthored) paper, “Political Alignment and Tax Evasion," at the colloquium.  (I had to put off blogging about it due to my travel earlier this week.)  It intriguingly finds evidence (from IRS tax data) suggesting that people may evade income tax more when they are politically opposed to the president.

Pretty much all of the effect comes from income that is reported on Schedules C and E regarding income that is subject to neither withholding nor information reporting.  Since evasion obviously can’t be observed directly (or at least, certainly not through tax  returns), it’s based on the size of the tax gap.

Also, they obviously they can’t link tax evasion to individual returns based on people’s (unknown) political views.  So it’s based on county-level data.  What they find is that, when a county has a strong partisan lean in presidential election voting, the tax gap seems to be bigger when a president of the other party is in office, as opposed to one of the same party as that favored by the voters in that county.  But the paper has a lot of sophisticated controls that push one towards accepting the conclusion that people evade tax more when they oppose the politics and policies of the current president.

Obviously, there is no data in the paper regarding tax returns filed during the tenure of the current president.  From other information in the paper I strongly surmise (although the authors do not say) that its findings are driven mainly by reduced tax compliance from Republicans when there are Democratic presidents, rather than the reverse.  This surmise reflects, for example, the fact that there is an overall Republican lean to people who have the flexibility to cheat more by reason of having a lot of Schedule C and E income as to which there is no third-party reporting backup.

Leaving aside that aspect, one could regard the paper’s main finding as either (a) so obviously true that it is unsurprising, or (b) so counter-intuitive and surprising that one wonders if it can actually be true.

Why unsurprising?  We know from past research that attitudes towards government affect tax compliance, and we know that partisan loyalties affect attitudes towards a particular government, so via transitivity it seemingly had to be true.

Why surprising?  Well, the implication here is that people who are reporting cash income add or drop a few hundred dollars (say) from the amount that they report just because George W. Bush or else Obama happens to be president and they like/dislike this particular individual and his administration.  That’s a striking thought when one ponders it at the ground level, as something that happens (at least in the aggregate) case by case.

A possible payoff for policymakers would be that, purely on grounds of maximizing the revenue yield relative to the cost of audits, Democratic presidents should audit people in “red” counties more, and Republican presidents should audit people in “blue” counties more.  However, this is unlikely to be considered an appealing takeaway for policymakers.  So what it mainly does, within the tax literature on compliance, is less a matter of providing useful policy payoffs than of fleshing out our understanding of taxpayer behavior (i.e., regarding the relevance of “tax morale” motives that are distinct from the Allingham-Sandmo picture of financial optimization subject to risk aversion).

But it also contributes to the political science literature, where the voting paradox suggests that voters’ behavior must reflect consumption / expressive motives rather than involving rational choice relative to the question of who will win the election (since one’s time has value and one can only infinitesimally affect the likely outcome).