Dueling Tax Truths

With so much of this year’s campaign focusing on taxes and tax policy, it might seem unlikely that prospective voters could still be swayed by learning a basic fact about the U.S. tax system. However, last week’s YouGov survey found that 44% don’t know it—and that they are 6 points less likely to vote for Mitt Romney as a result. What is this magic fact? That almost half of Americans pay no federal income tax.

When a video surfaced of Mitt Romney telling donors that 47% of Americans pay no federal income tax, pundits were quick to declare it “an utter disaster” that “killed Mitt Romney’s campaign for president.” After first saying that his points were “not elegantly stated,” Romney later called them “just completely wrong.”

Much of the commentary on the “47%” flap challenged Romney’s characterization of people who don’t pay federal income tax as “dependent on government” and unwilling to “take personal responsibility and care for their lives.” But almost all of it noted that Romney’s basic premise was essentially correct—according to the latest figures from the Tax Policy Center, 46% of American households paid no federal income tax in 2011.

With all this attention to “the 47%,” one might suppose that everyone now knows this simple fact about the U.S. tax system; but that supposition would be wildly off-base. The YouGov survey reminded 1000 people that “Mitt Romney has claimed that almost half of Americans pay no federal income tax,” than asked, “Do you think that claim is true or false?” Only 55% said true, while 44% said false. Nor were the skeptics simply committed Democrats toeing the party line; people who are generally less knowledgeable about politics were much less likely to accept Romney’s claim regardless of their own partisan views.

Would it matter if these people knew how many Americans don’t pay federal income tax? Comparing the vote intentions of otherwise similar respondents (controlling statistically for party identification and assessments of President Obama’s job performance, Romney’s likeability, and national economic conditions), those who accepted Romney’s claim were 6 points more likely to say they would vote for him than those who didn’t. Was this just a matter of trusting Romney? Apparently not, since they were also 9 points more likely to say they would vote for a Republican House candidate. These estimates suggest that Republicans could gain significant support by convincing even half of those doubters that Romney’s controversial remarks about a dependent class were grounded in a true fact about the U.S. tax system.

Another item in the YouGov survey reminded people that “Mitt Romney has proposed a major reduction in tax rates financed by the elimination of various tax deductions,” then asked, “Do you think Romney’s plan would increase or decrease the proportion of total income tax paid by middle-class taxpayers?” 46% of the respondents said the middle class would pay a larger share, while 53% said its share of the total tax bill would remain unchanged (35%) or even decrease (18%).

As with beliefs about how many people don’t pay taxes, these expectations about the impact of the Romney tax plan seem to have a significant impact on vote intentions. Other things being equal, someone who expected the middle-class tax share to increase was 4 points less likely to support Romney—and 6 points less likely to support a Republican House candidate—than someone who expected the middle-class tax share to decrease or remain unchanged.

A variety of independent analysts have argued that Romney would be unable to reduce tax rates significantly without either escalating the federal deficit or increasing the tax share of the middle-class; there simply are not enough deductions going to high income-earners to offset a major reduction in their tax rate. In the first presidential debate, Barack Obama struggled to explain this apparent inconsistency in Romney’s proposal, but he was stymied by the complexity of the issue and by Romney’s flat insistence (with a nod to “six other studies”) that his plan “will not add to the deficit” and “will not under any circumstances raise taxes on middle-income families.”

Not surprisingly, in light of this complexity and contentiousness, the pattern of beliefs about the impact of Romney’s tax plan looks quite different from the corresponding pattern regarding how many people don’t pay federal income tax. While better-informed Democrats were significantly more likely to say that Romney’s plan would increase the share of taxes paid by the middle class, the impact of political information among Independents was much more muted—and the best-informed Republicans were slightly less likely to agree.

With “all these studies out there” (as Romney put it in the debate), attention to politics mostly seems to have widened the partisan gap in assessments of how Romney’s tax plan would turn out to affect middle-class taxpayers. That is a big challenge for Obama and others who want to convince uncommitted voters that Romney’s plan is a Trojan horse for reducing taxes on the wealthy. Obama said in Denver, with some exasperation, “it’s math.” But math is hard, especially when it’s about the future. By comparison, convincing uncommitted voters that the existing tax system exempts a lot of people from paying income taxes seems a lot less daunting. If Republicans can spread that word—without getting caught up in the ugly stereotyping that sidetracked Romney last month—the information campaign could have a significant political pay-off in November.


Cross-posted at Model Politics.

12 Responses to Dueling Tax Truths

  1. Andrew Gelman October 15, 2012 at 1:05 pm #


    I have no disagreement with any of your analysis above, but I think it’s misleading to say that “Romney’s basic premise was essentially correct.” Romney said, “[Obama] starts off with a huge number. These are the people who pay no income tax. Forty-seven percent of Americans pay no income tax.” It is not essentially correct to claim that all (or an overwhelming majority of) the people who pay no income tax are Obama voters. The people who pay no income tax include many Republican voters and many many nonvoters.

  2. RobC October 15, 2012 at 1:51 pm #

    The statement that correlation is not causation is by now so trite that it’s painful to repeat it, and yet . . . .

    Professor Bartels says that “44% don’t know it—and that they are 6 points less likely to vote for Mitt Romney as a result.” He also says that “those who accepted Romney’s claim were 6 points more likely to say they would vote for him than those who didn’t” and that “[o]ther things being equal, someone who expected the middle-class tax share to increase was 4 points less likely to support Romney—and 6 points less likely to support a Republican House candidate—than someone who expected the middle-class tax share to decrease or remain unchanged.”

    In the first instance Professor Bartels specifically states causality and in the others he implies it. But wouldn’t it be equally correct to say that someone who supported Romney and/or the Republican House candidate was x% more likely to believe Romney’s claim (or perhaps more precisely, to say he or she does)? If there’s evidence that causality runs in the direction Professor Bartels says, he hasn’t provided it here.

    • Eric October 16, 2012 at 12:05 pm #

      I agree with RobC’s idea, but I would go a little further. It is plausible, perhaps even likely, that the causality is the reverse of what Bartels implies. Conservatives (i.e. Romney supporters) are more likely than Democrats to get their news from outlets that emphasized the true bits of Romney’s statements. In other words, they were Romney supporters first, then they learned that almost half of Americans don’t pay federal income tax.

      I also agree with Andrew that that was not the main thrust of Romney’s comments. Romney’s apparent confusion in thinking that the about half of Americans who don’t pay federal income tax is the same as the about half who receive federal benefits and the about half who support Obama reveals a massive ignorance about the country and about basic logic.

  3. Jeremy Pressman October 15, 2012 at 2:27 pm #

    I am not a tax guy, but many (most?) of those 47% pay many other federal taxes, right? The whole line of argument is HIGHLY misleading. It falsely implies the 47% are a bunch of freeloaders.

    Maybe the polling data he mentioned shows some Americans don’t know which taxes count as income taxes. If you take it out of my paycheck, maybe I assume it is the same as income tax.

    • ben October 16, 2012 at 12:17 pm #

      for the 1 millionth time folks, federal income taxes are not the same thing as payroll (e.g. FICA) taxes… yes our politicians have run away from the Al Gore “lock box” model, but the reality is social security was sold to the American people as a social insurance program. people who are paying payroll taxes do so with the idea that they get something back in a benefit later on. people who pay federal income taxes are not expecting a monthly check from the government later on for their having paid an federal income taxes. Why is this so hard for people to understand?

      • Jeremy Pressman October 16, 2012 at 3:00 pm #

        You can say it for the millionth time but what is the popular perception? If most people don’t know the difference, the fact that they differ is less relevant here.

        People who pay any form of tax expect certain benefits. Yes, the nature of the benefit varies. But people have expectations whether it is a government check, the provision of national security, or support for building a new bridge.

        • ben October 16, 2012 at 4:53 pm #

          People who pay any form of tax expect certain benefits. Yes, the nature of the benefit varies. But people have expectations whether it is a government check, the provision of national security, or support for building a new bridge.

          Bingo! That’s the entire point about why people who pay both federal income taxes and federal pay roll taxes are correct in seeing them differently.

          The person who pays federal income taxes is paying toward military, roads, etc. (general costs of government) with the expectation that they will get a benefit and use those things. Yet, the citizen who ONLY pays federal payroll taxes but not income taxes is getting both these generally distributed benefits that others are paying to support via federal income taxes AND the benefits that come in the form of social security benefits later in life because of the payroll taxes (to which he or she contributes).

          The point is that non-retired Americans who only contribute to the payroll tax are getting more in benefits than they use relative to non-retired Americans who pay both the payroll tax and the federal income tax. It’s intellectually dishonest to suggest otherwise. That’s why the makers vs. takers thing has some legitimacy. Now, of course the GOP is hypocritical about this because the reason so few Americans in the lower and middle classes pay an effective income tax rate of ZERO is due to GOP tax cuts over the years. But it doesn’t discount the empirical reality that a lot of folks have less “skin in the game” so to speak.

  4. Mark October 15, 2012 at 2:28 pm #

    I think this is a common problem not just related to math but also to messaging and the amount of time people have to focus on details. For instance, President Obama has continually spoken about raising taxes on “millionaires and billionaires”. I have had numerous conversations with people in the past two years who agree with that sentiment because they understand it to be people who not only have a high annual income but are truly wealthy in terms of having huge assets. But they did not realize, and are horrified to learn, that when he says “millionaires and billionaires” he also means single people making 200K or couples making more than 250K who may have recently reached that level of income and are still trying to build a nest egg. I think it is very effective messaging by the President albeit completely misleading.

  5. Larry Bartels October 15, 2012 at 4:29 pm #

    Andrew: I hope it is clear from the sentence you quoted, and from the post as a whole, that the “basic premise” at issue is about who pays federal income tax, not about Romney’s associated sociological and political musings (which have been much-dissected).

    RobC: Correlation is not causation, but it is a start, and not to be overturned simply by repeating that correlation is not causation. Judgments about causation require, well, judgment. So what considerations are relevant here? As is clear from the first graph, well-informed Democrats are much more likely than less-informed Democrats to accept Romney’s claim; that pattern is inconsistent with the idea that the difference in vote intentions is primarily attributable to “someone who supported Romney” simply being “more likely to believe Romney’s claim.” As I mentioned in the post, the fact that the pattern is stronger for congressional vote intentions than for presidential vote intentions also weighs against the interpretation that this is “just a matter of trusting Romney.” Finally, do the multiple regression analyses do a reasonable job of “controlling” for other factors influencing whether respondents “supported Romney and/or the Republican House candidate”? The adjusted R-squared statistics (excluding the tax fact variable from the regressions) are .91 and .86, respectively. Does all of this guarantee “that causality runs in the direction Professor Bartels says”? Of course not. But it makes that interpretation of the data much more plausible than the alternative.

    Jeremy: There are lots of reasons for people to be confused about this fact, and lots of reasons to resist the conclusions Romney wants to draw from it. My point is simply that the fact itself seems to be politically powerful.

    Mark: I think it’s right to suppose that “messaging” is especially consequential (and potentially misleading) for people who don’t choose to “focus on details” of political debates. But it isn’t just them, in my view. All of us struggle to form coherent preferences in a very complex world with very limited cognitive equipment. Here is a brief statement of my take on the implications for democratic theory: http://www.princeton.edu/~ccameron/KoreaIIE/IIE337/Bartels.Popular.pdf.

    • Dave Ely October 15, 2012 at 9:17 pm #

      I think that you are making two significant and related mistakes here. One is interpreting answers to poll questions as if they accurately reflect knowledge of reality, and the other is that you probably have the causation reversed. If your interpretation was correct, your first table would imply that low-information Republicans have better knowledge of income tax distribution than even the highest information Democrats and independents. A much more likely explanation is that the way respondents answer the question is related to how the question is interpreted. The key phrase is “Federal Income Tax”. I believe that a very high percentage of respondents would “hear” tax and not hear the “Federal Income” part, and would correctly say that the claim was false based on their hearing. Lower information voters would be more likely to make this mistake, but it would probably be common among high information voters as well. This would explain why the slopes are as they are, and also would explain why you see 3 distinct lines.

      It is entirely believable that there is a strong correlation between likelihood of supporting Romney or other Republicans and the belief that nearly half of Americans pay no taxes at all. The causation in regards to that correlation is more likely to be that both are affected by a common cause, or that a conservative political outlook causes the belief about who pays taxes. Your second chart is consistent with your first with this explanation, but I don’t see how it makes sense at all in light of your explanation of the first.

  6. Brett October 15, 2012 at 8:33 pm #

    No, Romney’s premise is just about 100% wrong. Let me direct you to this:


    While federal income tax is progressive, the other two big tax levies are highly regressive: the payroll tax (please note that 28.3% of those filing tax returns pay no income tax but do pay payroll taxes. They’re working but not making enough to trigger an income tax liability), and state and local taxes, where the highest rates actually fall on the poorest quintile, while the highest income brackets actually pay the lowest rates:


    Another interesting conclusion from decisionsonevidence comes from the age graph, which shows a sharp divergence between those who pay no income tax but do pay the payroll tax: the 16-21 age cohort: entry level workers and college students (whom Romney characterizes as refusing to take personal responsibility for their lives):


    Roughly speaking, and this is really back-of-envelope-calculation here, 53% work & pay INCOME tax, 28.3% work but pay only payroll tax totaling 81.3% of all taxpayers. Who makes up the rest? The total workforce (both employed & unemployed) is 240 million (http://www.bls.gov/news.release/empsit.a.htm), while 40 million taxpayers are retired (http://www.census.gov/prod/cen2010/briefs/c2010br-09.pdf), some of whom do and some who do not pay taxes, all of whom we can assume file tax returns. Which, just for the sake of argument, may run as high as 17% of total tax base right there. All told, that may account for as much as 98.3% of all tax payers accounted for without finding the mythical welfare queen Romney was trying to push on us. The 1.7% remainder is a far cry from Romney’s 47%.

    So who’s in that remaining 1.7%?

    Well 29.5 million Americans (enough to comprise over 10% of the total workforce, please note) suffered injury due to accidents in the US serious enough to merit hospital treatment: (http://www.cdc.gov/nchs/data/ahcd/nhamcs_emergency/2009_ed_web_tables.pdf). I wonder how many of those people were impaired enough to affect their earning enough to lose their tax liability? Or how many had to temporarily quit work to act as caregivers?

    I won’t even mention serious illnesses, but you can check the CDC for those numbers if you like, the point being I’m not seeing a huge problem with welfare queens kicking back and enjoying the high-life at taxpayer expense in these numbers.

    Unless the author means those enjoying historically low capital gains taxes, benefiting from tax breaks on carried interest or expensing their “working” vacations to their S corporation or sole proprietorship.

    Accept these numbers and the conclusions I draw here or not. The point being that anyone can paint just as convincing a picture with the data I’ve presented here as Romney tried pull on those people he was begging money from. By all means, lets get the real data and then we can go from there, but I really don’t feel like I and people like me deserve the insult Romney tried to deliver.

    Footnote: that 28.3% comes from the same source as that 47% Romney cited: the Tax Policy Center. You can look it up.