Yesterday, Lucy Martin of the UNC Political Science Department presented the above paper, discussing survey evidence regarding how Americans think about tax progressivity. The paper addresses an apparent puzzle: the fact that rising high-end inequality, plus stagnant real income growth for everyone outside the top 0.1 percent, has not yielded greater high-end tax progressivity.
Is this lack of an offsetting tax policy response to rising high-end inequality actually surprising? I’m not all that surprised by it, but it clearly is surprising if one’s baseline assumptions reflect (1) the median voter hypothesis, which holds that the median voter’s policy preferences tend to be enacted, plus (2) a view of voting as being based on narrowly defined economic self-interest. Under such a view, the median voter should want higher tax rates at the top under these circumstances, and should be expected to succeed in getting it – wholly independently of the question of whether or not this would be a good policy choice – unless she is sufficiently concerned about the efficiency costs on her of the high-end rate increase (or believes that it wouldn’t actually raise revenue).
The two main explanations for the “puzzle” that political scientists have offered are (1) the rich have too much political power for the sentiments of the median voter to carry the day, and/or (2) voters actually aren’t strongly supportive of increasing high-end redistribution – perhaps because they don’t just focus on narrowly defined economic self-interest.
The paper mainly comes out in favor of explanation #2. It concludes from survey evidence (gathered on yougov, with sampling to replicate characteristics of the general voting population) that, while there is widespread support for mildly progressive tax rates, these sentiments are relatively tepid, and support a progressive rate structure rather like the one that we currently have – rather than one with much higher marginal rates at the top.
However, explanation #1 is not refuted, except insofar as one interprets it as involving affirmative elite override of strongly held popular sentiments. The fact that the public does not care intensely can plausibly be viewed as leaving the elite free to decide. The evidence in the paper might come closer to supporting a strong version of explanation #1 if the 2016 election were to lead to the election of a Republican president, along with continued Republican Congressional control in both houses, and this in turn led to the enactment of a flatter rate structure – promised by all leading Republican candidates – that actually would be at variance with the paper’s findings regarding voter sentiment, including that among Republican voters.
Among the paper’s features that I particularly like are its (1) disaggregating between high-end and low-end inequality issues, (2) disaggregating between the issues under study and those of views on government spending and/or the “size of government,” and (3) offering a gauge on voter sentiments’ intensity and elasticity (defined as the rate of change, for one’s preferences regarding tax rates, as the income level that is under consideration changes).
In preparing for the session, I divided my thoughts into two main topics: (1) political science issues, focusing on survey design and one’s theory of voting, and (2) the tax policy takeaways one might glean from the evidence in the article.
1. VOTING & VOTER PREFERENCES
A) Research design – Here I see four main issues:
(i) Tax base – Marginal tax rates don’t mean much until we know to what they apply. Given the limitations on how much one can lay on the plate of survey respondents, we don’t know what (if anything) they had in mind if they liked, say, a 35% or 40% top rate. How might this relate to – and what would they think about – say, the use of tax shelters, the capital gains rate, or the general non-taxability of unrealized appreciation. These are techie issues, but surely they might have some actual or potential state of mind regarding the relevance (and high likelihood) of significant divergence between taxpayers’ taxable income and their economic income.
For that matter, what about average rate versus marginal rate? For a top bracket of, say, 35%, were they assuming that this was also the average rate that taxpayers with income in that bracket actually faced as to their taxable income as a whole?
(ii) Taxes only, without direct regard to the use of the funds – In principle, one should always think about tax changes in a long-term balanced budget sense. However, since money is fungible and there are many different possible uses of say, increased high-end revenue, the survey design did not offer any indication regarding how marginal revenues might be used (or ceased to be used). Instead, to avoid encouraging complete disregard of budgetary considerations, the survey informed respondents when particular choices would affect or greatly affect net revenue levels.
This probably was a better survey design than proposing particular uses of the funds – especially since, if one used too many alternatives, one would both be degrading the results’ statistical significance and requiring respondents to slog through more. But it did mean that respondents weren’t offered the possibility of taxing the rich more in order to fund something they might like. Obviously, a savvy politician might make earmarking claims of this kind (like NYC Mayor Di Blasio campaigning for higher taxes on the rich that he said could fund universal pre-K). It also may have caused revenue-raising higher tax rates at the top to look as if they merely would have imposed burdens on one group without conveying benefits to any other group.
(iii) $375,000 of income and above as the top group – In order to keep the rate brackets similar to those under actual U.S. income tax law at the time that the survey was being designed – which had advantages in terms of figuring out net revenue effects – the study’s top group was people with income of $375,000 or more. Insofar as actual voter concern today focuses on plutocracy and/or the super-rich – say, the top 0.1%, as opposed to just the top 1% - this meant that one did not learn what the respondents thought about tax rates for people earning at least, say, $1 million or $10 million a year.
(iv) Anchoring – What should we make about the fact that the tax rate structure getting the most support looked rather like what we actually have right now? Does this mean public sentiment is being honored? Alternatively, might the causal arrow run the other way, with people taking cues from what is actually on the books? What would happen if we could run the same test with U.S. voters 60-odd years ago, when the top rate was over 90%? And if we found that people liked that top rate then, once again we’d have to ponder the direction of the causal arrow, i.e., do the rates reflect preferences or anchor them? More feasible, of course, would be doing a similar survey in, say, an EU country with higher marginal rates for individuals at the top of the distribution.
B) Theory of what drives voter preferences
The paper identifies three main explanations for voters’ tax policy (and other) preferences: (mainly economic) self-interest, fairness norms, and partisan identity. Herewith some thoughts on each:
(i) Economic self-interest – Assertions that voters should be expected to vote in their self-interest, generally defined economically, have long struck me as hard to reconcile with the voting paradox. Here I mean not the range of Condorcet, etc., phenomena, but rather the fact that voting itself is not an economically way of promoting narrowly self-interested outcomes.
Everyone knows the basic issue here. I once heard (perhaps apocryphally) about a prior-generation professor who apparently told his students that he never voted, even though he cared about political outcomes, because the chance that his vote would alter the outcome was effectively zero. Therefore, the benefit certainly wasn’t worth the cost, defined in terms of the time he would have to spend on voting. “What if everyone thought that way?” he was asked. “Well, then I’d certainly vote,” he replied.
While awareness of the voting paradox surely does reduce turnout, obviously millions of people vote anyway. But the paradox is still, to my mind, of primary importance in understanding and explaining voter behavior. Since voting is so irrational, if defined as seeking to realize the dollar value one places on a desired election outcome, divided by the likelihood that it will actually change the outcome, obviously something else is going on. Consumption? Self-expression? Sense of obligation? Cooperating rather than defecting with regard to like-minded voters, who face a prisoner’s dilemma insofar as each would rather not bother to vote but they’ll only win if enough of them vote?
Once one is voting based on any of those motives, it is no longer irrational not to vote based on economic self-interest, even if one otherwise acts pursuant to it. Suppose you just like voting for the candidate with whom you’d hypothetically rather have a beer. It’s not personally irrational to vote for this person, even if he or she would be bad for one’s economic interests, if this feeling sufficiently flavors one’s enjoyment of the voting act.
More important still, it is now affirmatively irrational – and people damn well know it – to invest significant effort in figuring out which candidate would best serve one’s interests, unless one happens to enjoy the investigative process. How much time would you spend figuring out what car you ought to buy, if the decision wasn’t up to you but instead would be made by a multi-million person electorate?
Given this point, it verges on being paradoxical that people pay as much heed as they do to the question of which candidates would favor their interests and those of people like them. I would presume that this reflects feelings of affinity that in turn reflect our having evolved to internalize sincere belief in arguments in favor of our own interests.
But one still doesn’t get a strong prediction that the median voter will respond to high-end inequality in the manner presumed by standard political science models.
Fairness norms – The paper notes that people who favor higher tax rates at the top tend to believe that the most economically successful were mainly lucky, while those who oppose such rates tend to believe that merit and hard work play larger roles. Likewise – though the issues are somewhat independent – those who favor high tax rates tend to have a lower valuation of the likely efficiency costs than those who oppose such rates.
While this does not contradict the paper, which looks at correlations without proposing specific causal theories, I tend to wonder which way the causal arrow runs. I would suspect that there is a widespread tendency to take the “progressive” view on both issues as a consequence of one’s (for other reasons) favoring higher rates, and the “conservative” view on both issues as a consequence of one’s (for other reasons) opposing such rates. In other words, I think people often start with biases and then develop the needed rationalizations, although certainly one ought to aspire to the reverse.
Perhaps more intriguing is the paper’s finding that those who favor high rates at the top tend to lean towards being “reciprocators,” as tested separately within the survey, while those on the anti side tend to lean more towards being “free riders.” But given this point , along with the voting paradox, it is interesting that conservatives, whom the survey suggests lean towards being on the “free rider” side, have nonetheless done a better job of countering their own internal group prisoner’s dilemma, by maintaining higher voter turnout levels even in non-presidential election years.
Partisan identification – Not surprisingly, this proves to have the strongest predictive value, arguably supporting the observation that following politics is a lot like picking sports teams to root for.
2. TAX POLICY IMPLICATIONS
Given the structure of the U.S. political system – in particular multiple branches, status quo bias, and partisan entrenchment, frequently yielding gridlock – it’s no surprise that the system hasn’t grown significantly more progressive at the top. This would require (a) Republicans to lose their veto power at all levels, plus (b) Democrats to support as a bloc enacting a more progressive rate structure. (The 2013 budget deal did a bit of this, but only by restoring the Clinton-era top rate.)
So it’s not clear what extent we need to look to voter preferences to explain the “puzzle” of limited change to high-end rates. But nonetheless, because Knowledge is Good (in the words of Emil Faber, but I actually mean it), one does not need policy or outcome relevance in order to find the paper’s analysis interesting.
In any event, however, I see the following three main takeaways for people who favor greater high-end progressivity:
(a) Suppose one favors raising tax rates at the top – even if not to 1960s levels, then at least to something approaching Diamond-Saez-advocated levels on the order of 70%. The information provided is not encouraging, and suggests that the indicated change would most likely have to reflect intra-elite opinion movements, rather than the empowerment of widespread public sentiment. While it’s not clear why the elite should be expected to favor higher tax rates, to some extent on itself, note that the conservative movement, unlike its counterpart on the left, has spent the last 4 decades building a powerful policy advocacy infrastructure in Washington.
(b) Efforts to increase high-end progressivity can try to take advantage of earmarking the net revenues for widely supported functions, although the public is far from being wholly naïve regarding the relevance of earmarking given that money is fungible.