Skip to content

Are all rich people now liberals?

So asks James Ledbetter in Slate.

And the answer is . . . No!

Here’s what happened in 2008:


OK, that’s how people vote. How bout party identification and ideology? Check these out:

(Click on image to see larger version.)

And here it is, sliced a different way:


Of, if you want to see it in map form, check out this article (with Daniel and Yair).

P.S. A skeptic might comment that the above graphs, which are based on national poll data, only break down incomes to the top 5% or so. What about the truly rich. Here are my thoughts on the political attitudes of the super-rich.

P.P.S. Ledbetter actually makes some good points in his article, which is about the campaign contributions of rich Americans. The article relies on a recent book by David Callahan, which seems to echo the work of Tom Ferguson (cited in the above-linked blog entry), who’s tracked campaign contributions by industry over many years.

I think that Ferguson (and Callahan) are on to something important, and I’m glad that Ledbetter is thinking about the implications of these trends. I just think his headline is silly and unhelpful. And the fact that it got out there at all–I assume Ledbetter didn’t write the headline himself–is evidence that there is still a lot of confusion about income and voting in the news media.

P.P.P.S. Some more clarification. Let me emphasize that I’m not trying to get into a fight with Ledbetter; I just think he needs to modify one of his empirical claims.

Ledbetter raises important issues in his article. Campaign financing is hugely important, and I agree that it’s important to look at contributions from different industries.

But I’m bothered by his headline, and even more by the sub-headline that answers “Sort of” to the question “Are all rich people liberals?” The data just don’t bear this one out. Yes, some rich people are liberals. But, as we discussed in Red State, Blue State, this varies a lot by region. It’s no coincidence that your article mentions rich people in California, Connecticut, and New York: These are states where there is essentially zero correlation between income and how people vote. There are a lot of other rich people in the country, though. In Texas you’ll be seeing a different story. And, in aggregate, all the evidence I’ve seen suggests that richer people are more likely to be conservative Republicans.

And this is not just about self-definition. See the third graph above. The top row is self-declared ideology and bottom two rows are economic and social ideology, each computed by averaging a bunch of issue questions from a survey. Unsuprisingly, income is more correlated with conservatism for economic than for social ideology. In either case, it’s not at all the case that rich people are liberal.

I’ll say it again. Ledbetter writes: “David Callahan takes aim at this notion, using reams of data to make two ambitious arguments: first, that America’s rich now tilt politically left in their opinions and campaign contributions, and second, despite the surge in corporate funding of Democrats, the Democrats pursue far more liberal economic and business policies than Republicans.”

I think the second claim is definitely true. But the first claim, “that America’s rich now tilt politically left,” is not supported by the evidence I’ve seen.

To put it another way, I think it’s perfectly reasonable to believe the following propositions:
1. Rich people are an increasingly important source of funding for the Democratic Party
2. Democratic officeholders–even those such as Barack Obama who are from the liberal wing of the party–seem to have an instinctive sympathy for the upper class.
3. The Republican party is even more pro-rich-people than the Democrats are.
4. Richer Americans are more conservative and Republican than middle-income or lower-income Americans. Rich people do _not_ tilt politically left.

I was bothered by Ledbetter’s headline for the same reason I was bothered by Wall Street Journal reporter Robert Frank’s credulity toward a P.R. man’s claim about the opinions of the super-rich: The claim of “billiionaires for Obama” is often the first step in a pseudo-populist argument about partisan support. See, for example, these ludicrous claims from Michael Barone, who should really know better.

If Ledbetter wants to argue that the Democratic Party has been captured by Wall Street–or if his counterpart on the other side of the fence wants to argue that the Republican Party is fighting for middle-class Americans–that’s fine. But I’d rather see these arguments made directly (as you do in the meat of your article) rather than via incorrect or misinterpreted statistics (as in the writings of Frank or Barone).

I’m sympathetic to Ledbetter’s point about Democratic economic policies, and I think he weakens his own argument when he ties it to a statistical misconception that rich people are liberal. In chapter 9 of Red State, Blue State we discuss some of the political implications of different patterns of income, ideology, and voting in different states.


  1. A. Zarkov says:

    These graphs show us voting behavior as a function of income conditioned on ideology. But how does income affect ideology? Surely they are not independent. Could ideology also affect income?

    Then we have IQ. IQ affects income, does it also affect ideology? The IQ variable seems to be left out of most social science studies despite its explanatory power. It's a little like trying to do chemistry without temperature. Imagine trying to make sense out of the physical universe without a temperature variable.

  2. Maarten Buis says:

    There are a couple of reasons why IQ is less often used in social science research projects than A. Zarkov expected.

    First, it is very hard to pin down what it is exactly that we are trying to measure with it, e.g. think of all the varients like "emotional IQ" that have been developed.

    Second, even if we agree on what it is that we try to measure, it still very hard to get an actual measure of it.

    Third, the fact that a variable might influence an outcome does not mean we must put it in our model. If it is an intervening variable we definately do not want to enter it in our model. IQ could easily be an intervening variable, since it contains a social component (e.g. the kind of stimulation you received when you were young, the kind of food you got when you were young, whether or not your house was build in the neighbourhood of a toxic waste dump, whether your job exposes you to toxic materials, the amount of alcohol your mother drank while she was pregnant, etc. etc.)

  3. Nick Cox says:

    There is an even simpler reason why IQ is not often used, setting aside the enormous doubts about what IQ means and whether it is worth measuring: Researchers would need to couple their survey with individual intelligence tests of the survey respondents.

  4. nameless says:

    Not to mention traditional IQ tests overcompensate for age, meaning as one gets older their IQ declines — even if one were to answer the exact same percentage correctly each time.

    Then IQ would be inversely correlated with age, while Republicans have a higher proportion of older voters. This naturally implies a chart showing Republicans with statistically lower IQs than Democrats, which I have the impression is the opposite of what A Zarkov wishes to see.

    The point is, IQ is not like temperature, and sociology is not like chemistry.

  5. Gina says:

    All of those reasons apply to the educational attainment variable; but that doesn't stop social science studies from including it.

  6. Gina says:

    Not any more than they currently have to couple their surveys with pay stubs and diplomas.

  7. A. Zarkov says:

    First, it is very hard to pin down what it is exactly that we are trying to measure with it, e.g. think of all the varients like "emotional IQ" that have been developed.

    The term "emotional IQ" is a misnomer and refers to a collection of uncorrelated personality traits unrelated to general intelligence. A better name would be "social competence." On the other hand, we know a general intelligence exists because tests on a variety of cognitively demanding tests are all positively correlated. We know from testing data that the rank of the underlying population correlation matrix is meaning that there is 1; thus there is only one kind of general intelligence. This work goes back to Thurstone in the 1920s, and we now have more than 80 years of testing data to back up the concept.

    Second, even if we agree on what it is that we try to measure, it still very hard to get an actual measure of it.

    It's not hard at all. The military uses the AFOQT (highly correlated to IQ) to evaluate recruits. People with very low scores are rejected. People with high scores are assigned to cognitively demanding programs such as jet fighter pilot training. It's very expensive to train these pilots and the military does not want to waste resources trying to train pilots who will ultimately wash out.

    IQ is real, and we can measure it. We can use the measurements to predict who will succeed. Of course it does not provide a perfect measure. But as George Box famously said, "all models are wrong, some are useful."

  8. A. Zarkov says:

    Not to mention traditional IQ tests overcompensate for age, meaning as one gets older their IQ declines — even if one were to answer the exact same percentage correctly each time.

    IQ test scores are ordinal, not ratio measurements. As such they only provide a ranking and not an absolute measure of intelligence. Someone who weighs 120 pounds has 20% more mass than someone who weighs 100 pounds. But someone with an IQ of 120 is not necessarily "20% smarter" than someone with an IQ of 100. We can only say that he will tend to score higher on cognitively demanding tasks.

    Operationally test scores are ranked and then assigned percentiles, and then replaced by the corresponding z-scores from the Gaussian Distribution. This forces the IQ test scores to be Gaussian. For historical reasons, the reference distribution has a mean of 100 and a standard deviation of 15. I think the standard group for IQ testing was English boys aged 18. By construction they have an average of 100. For older people, we would need to rank within the desired age group. Thus an 80 year old with an IQ of 130 would tend to do better on cognitively demanding tasks than an 80 year old with an IQ of 100. Of course as a group men aged 80 are going to get lower raw scores than than men aged 21. In general we don't give IQ tests to old people except perhaps as a medical diagnostic. We don't have to predict their future behavior, it's already happened.

    while Republicans have a higher proportion of older voters. This naturally implies a chart showing Republicans with statistically lower IQs than Democrats, which I have the impression is the opposite of what A Zarkov wishes to see.

    If Republicans are really a bunch of old codgers as compared to Democrats, than they certainly would get lower raw scores on IQ tests. But if the scores were properly group normed then the results could be different. But this is my point exactly. In social science we should have variables like age, sex, race, religion, ideology, geographical location etc. But why leave out IQ all the time? In those cases were such data were available, why not use it? The following example illustrates my point.

    The New York City Fire Department has been sued because because blacks and Hispanics score much lower on the FDNY admissions than everyone else taking the test. No one can find a test question that's biased. If you look at the questions they actually give the answers in the question intro. Thus the test is really a reading comprehension test. I was able to calculate the ratio of test score using only group IQs. In other words, we can predict the test score statistics almost perfectly using IQ. The test scores are exactly what one would expect given the racial mix of the people taking the test. There is no mystery.

    BTW one can do the same kind of IQ analysis for other tests such as the bar exam and get similar results.

    Lesson learned : If you want to understand the world than use all the variables you have.

  9. Andrew Gelman says:

    Ummmm . . . that's all fine. I have no problem with people studying how voting relates to education, occupation, etc.–or even IQ or personality type, if you happen to have surveys that include measurements of these.

    In the meantime, sometimes people go around saying things like "Are all rich people now liberals?" And I can answer this sort of question. Descriptive statistics can be useful in its own right.

  10. A. Zarkov says:


    My question about using IQ was tangential to my main interest. All your curves condition on political ideology. What causes people to hold a particular ideology? Marx would say income. Steven Pinker would say ideology might be an inborn personality trait. Midge Decter, author of Liberal Parents Radical Children and others claim your parents largely determine your political orientation. What does data tell us?

  11. Andrew Gelman says:

    In regard to the question of what contributes to political ideology, I follow the standard political science model in which your continuous ideology is the sum of many factors, including the ideologies of your parents, friends, etc., what you get from the news media, your own personality, economic conditions during your politically formative years, economic conditions today, etc.

    Given a basic knowledge of history and politics, it's no surprise that richer people tend to be more conservative (on various dimensions) and tend to vote for Republican candidates. I wouldn't bother wasting the pixels to point this out at all . . . except that so many people who write about politics don't seem to realize it!

    In some ways, this makes my job easier: even if all I do is describe what's out there, without any explanation, I can be making a contribution to political understanding.

  12. I guess I don't see why the only remarkable finding here might be if the number of rich people who vote Democratic well exceeds those who vote Republican.

    Isn't the remarkable and important point — well illustrated in these graphs — that rich people are generally as likely to vote Democratic as Republican?

    Wasn't there a time when rich people voted against Democrats precisely because they saw them as against being pitted against their interests?

    What does it say about modern day American politics and society, in which the income disparity between the middle and working class and the rich has risen to breathtaking levels, that those occupying the privileged positions at the top see in the Democrats nothing that they need fear? How much has the Democratic Party sold out its commitment to the welfare of the common person to enable that sort of disconnect?

    Why shouldn't the middle and working class simply abandon the Democratic Party as simply another corrupt conglomeration of pols which no longer represents their interests? (And no, it doesn't suffice as an answer to say that the Republican Party is worse. There exists no inherent reason we must forever choose between Evil and Eviler.)

  13. Andrew Gelman says:


    You'll have to take a look at Red State, Blue State. The short story is that richer voters have been voting Republican for decades: Except for the decade starting in 1952, voters in the upper third of income have been about 10-20% more Republican than voters in the lower third.

    To put it another way, no, it's not true that the rich used to vote for Republicans and now vote for Democrats.

    The prevalence of this misconception is one reason we wrote Red State, Blue State, and it's also the reason I write blog entries such as this. I'm sure James Ledbetter and his headline writer mean well, but I think they'd be doing a better job if they were aware of the facts that we know from 70 years of scientific polling.

    See here for more along the same lines. For some reason, it's hard for people to keep it in their heads that the rich-poor difference in voting preference is about the same now as it was in 1940.

  14. A. Zarkov says:

    liberal biorealist writes,

    Why shouldn't the middle and working class simply abandon the Democratic Party as simply another corrupt conglomeration of pols which no longer represents their interests?

    Exactly what working class interest is the Democratic Party not serving? We have a highly progressive income tax with an earned income credit so that currently 47% of workers pay no income tax at all. Surely that covers the working class. The Democrats have passed a new large health insurance program, extended unemployment benefits etc. They did bail out the big banks, but the working class is not going to pay for that, it's the upper and professional classes that will have to pay for the deficits. They have adopted a to-bottom strategy pioneered by Lindsey Administration in the early 1970s. The very rich do well along with the poor and working class. The middle and upper middle class suffers the most as I found out as a resident of New York City in the early 1970s.

    I really don't understand your objection. What more would you have them do for the working class?

  15. Andrew, you seem to have missed my point.

    To begin with, you say that the "rich" (i.e., in the top 1/3 of the income distribution) have favored Republicans in the past by 10-20%. Now I should think that a number of us don't consider the upper third per se as being "rich". I'd be far more interested in how, say, the upper 10% or 5% or 1% vote in present decades vs past decades. My expectation (though I'll admit I certainly don't know for sure) is that those in those upper ranges now vote far more for Democrats than they have in the past (by your graphs, again, about 50% today, perhaps even slightly more). I'm not saying here that those upper ranges may not vote in fairly significant numbers Republican as well — again, my point is that a good number of these genuinely affluent people choose to vote Democratic, which, in other eras, would seem to be voting against their economic interests.

    My point is not the usual point one sees in, say, What's the Matter With Kansas? There, the author argues that nowadays middle class voters in Kansas vote against their economic interests by voting Republican, which, supposedly, they did not do in the past. If you think the data doesn't support that view, that's one point.

    My point is rather that it would SEEM at first blush (though, again, it may be that the data contradicts this) that the rich nowadays vote against THEIR economic interests — IF one were to assume that the Democrats were actually against their interests.

    What I claim, though, is that, in fact, unlike in the past, the Democrats are NOT in any real way pitted against the interests of the rich. THAT is why so many of the rich feel so completely comfortable voting for them. As I believe Warren Buffet himself may have said, our society has engaged in class warfare, and the rich have won — and the Democrats are as much on board with this victory as any Republican.

    Which brings me back to my original point: why should middle class voters continue to support the Democratic Party when they have done nothing — truly nothing — to ameliorate the increasing problem in income inequality? For example, CEOs of Fortune 500 corporations today typically make 300-500 times as much as the average worker. Back in the 70s, they made only 30-40 times as much — a jump of a full order of magnitude. Does any rational person really believe that today's corporations are more successful, or more competitive with their foreign counterparts, than those of the 70s? Have Democrats done ANYTHING to remedy this? Certainly nothing effective that I've been aware of.

    In general, Andrew, I think much of your work against the claims of What's The Matter With Kansas misses the larger picture here. You argue that the middle and working class in Kansas really isn't voting terribly differently from the way it did decades before — presumably supporting the notion that Democrats are still mostly preferred by the lower middle and working classes, and so that they aren't actually going Republican and thus, supposedly, voting against their economic interests. But that assumes that the Democrats really are allied with the working class and middle class. It may easily be the case that most of those lower middle class voters only vote Democratic because Republicans are actually worse for their interests than Democrats, not that the Democrats are really good for their interests.

    But the situation is not symmetric with respect to the rich and whom they vote for. I don't think anyone would claim that the Republicans aren't fully allied with the interests of the rich. Any rich voter could easily plunk down fro a Republican and see their economic interests exceedingly well served. Yet half of all the rich seem to vote Democratic nowadays. I don't see how that might possibly be true unless the Democrats are themselves very protective of their economic interests as well.

  16. Andrew Gelman says:


    You can take a look at Red State, Blue State or at some of the graphs that have appeared on this blog over the years. We do look at the top 5%. There have been high-income Democratic voters for many decades.

    Regarding your discussion of the economic policies of the Democrats and Republicans, you're right that we're only comparing the two parties. Our summaries reveal how people are voting; we're not attempting to study whether the Democrats really are allied with the working class, or whether Republican policies are good for ordinary Kansans, or whatever. These are all important questions, they're just not the questions we're answering.

    Regarding the asymmetry between the support of the two parties: I think this is an important point, and we discuss it in chapter 9 of Red State, Blue State.

  17. Nick Cox says:

    I disagree. People mostly know what qualifications they have and what income band they are in. I don't think that is true for IQ. Your assertion would seem to imply that you could just ask people what their IQ is.

  18. nameless says:

    You make an excellent point, and your anecdote is an perfect example of why IQ as a single measure of intelligence has gone out of style.

    IQ was always an arbitrary measurement, influenced in varying degrees by literacy, logical thinking, spatial thinking, and mathematics. And who is to say that's all there is?

    So if you wish to include every variable in every study, then IQ is not the answer. You would need to measure every participant's every ability, at which point your model of the world is no longer a model; it is the world.

Where can you find the best CBD products? CBD gummies made with vegan ingredients and CBD oils that are lab tested and 100% organic? Click here.