Has anything changed?

The General Social Survey has a question beginning "Some people think that the government in Washington should do everything possible to improve the standard of living of all poor Americans; they are at point 1 on this card. Other people think it is not the government's responsibility, and that each person should take care of himself; they are at point 5," and asking them to place themselves.  The figure shows the estimated effects of  (log) income and education on opinions for each year that the question was asked. 

Positive numbers mean that the effects are in a conservative direction--"not the government's responsibility."  The effects of income are consistently in a conservative direction, and don't show any trend.  The effects of education are initially in a conservative direction, but go towards zero and negative in 2014, 2016, and 2018.  That is, more educated people are more liberal on this question than people with the same income but less education. 

I've had a number of other posts about changes in the effects of education on opinions about redistribution, most recently this and this.  They all suggest education used to go with opposition to egalitarian redistribution, but no longer does.  (Income still does, and the magnitude of the difference after controlling for education hasn't changed much). 

This point doesn't seem to be widely recognized--most observers seem to assume that "elites" of all kinds are opposed to redistribution.  An example is a paper that appeared in Science by Raymond Fisman, Pamela Jakiela, Shachar Kariv, and Daniel Markovits called "The Distributional Preferences of an Elite."  The elite was Yale Law students, and according to the abstract, they displayed "selfishness" and were "less fair-minded" than the average person.  The paper began by speaking of the "sense of entitlement" of "the American elite," with a footnote that spoke of "the phenomenon of growing elite entitlement."  The research itself was interesting, but the packaging was misleading.  It didn't deal with preferences involving the overall distribution of income and wealth, but with behavior in a low-stakes online exercise where you chose how to divide a sum between yourself and an anonymous "partner."  Basically, they found that the Yale Law students played the game differently than average people--they had more concern with their own gains and less with an even split between participants.  OK, but that's pretty far from issues like whether to raise taxes on people with high incomes.    A recent survey of technology entrepreneurs found that they were about as favorable to redistribution as Democrats in the general population (e. g. 76% favored raising taxes on people who earned over $250,000 per year), and from my experience of universities, I think that if Fisman et al. had asked similar questions of their participants, they would have found a lot of support for redistribution. 

If some elites, especially educational elites, have become more sympathetic to redistribution, why hasn't the change received more attention?  One reason is the natural appeal of the principle that opinions reflect straightforward self-interest.  Another possibility is what I have called an "anti-elitist mood," in which people are reluctant to say anything positive about "elites." This mood seems to be especially strong among people who are part of an elite themselves.   

Twenty-four more years

In a previous post, I mentioned that I had once compiled data on education, occupation, and vote in American presidential elections from 1936 to 1992, using a combination of Gallup polls (1936-68) and the GSS (1968-92).  I showed changes in the relationship between college education and major party presidential vote.  I have finally gotten around to updating this with the GSS data through 2016.  Here is the estimated effect of college education (0=no college; 1=some college; 2=college degree) on Democratic vs. Republican voting among non-blacks, 1936-2016.

There is some variation from one election to the next--the pro-Democratic effect in 2016 election was higher than expected, but the most exceptional election was 1972, which set a record that was not broken until 2016.  However, the dominant thing is a trend that seems to have started at the beginning (the correlation with a time trend is about .85).  I would have guessed that the trend began or at least accelerated in the 1960s, but there's no sign of that.  That is, the relative shift of educated people towards the Democrats is a gradual process that was underway long before people started noticing it, and shows no signs of stopping. 

How big are the effects?  If you fit a linear trend for the education effect to smooth out the short-term variations, in an election where 50% of high school graduates voted Democratic, the predicted Democratic vote among college graduates goes from 34% in 1936 to 61% in 2016. 

The numbers above are not a contrast with all people without college education--they control for the changing effects of primary/secondary education (no formal education, elementary, grades 7-8, attended high school, high school diploma).  Those estimates are shown in the next figure.  Note that they are consistently in a pro-Republican direction, and have become more pro-Republican since the 1960s.  (This might be because the people with less than a high school diploma are increasingly likely to be immigrants, who tend to vote Democratic).

 

So what needs to be explained is apparently not a change in the effects of education, but a change in the effects of higher education.  Why has this happened?  In a paper published in 2002, I offered an idea: "growth in affluence and the division of labor has produced occupational niches for educated people who are critical of the status quo."  Previously, people might have developed "radical" ideas in college, but tended to abandon them after they had to earn a living.  Now, such people can find occupations where their ideas are accepted or even dominant.   I don't know if my explanation is the whole story, or even part of the story, but I was at least looking in the right direction:  the change in the effect of higher education seems to be a gradual shift, not closely tied to specific political events.

[Data from the Roper Center for Public Opinion Research and the General Social Survey]

Before I forget

I saw a reference to a survey question asked by the Gallup Poll in 1963:  "On the whole, would you say that you are satisfied or dissatisfied with the honesty and standards of behavior of people in this country today?"  34% were satisfied, 58% dissatisfied.  The question was repeated a number of times, most recently in 1999.  I think that I've seen that question before.  What I hadn't known was that it was asked in several other nations.  The percent dissatisfied:

Norway          20%
France            39%
UK                 47%
W. Germany   51%
USA               58%

[Source:  Robert Lane, Political Thinking and Consciousness, p. 216]

Changes, part 2

My last post took changes in the effects of education and income on economic opinions (or at least one important economic opinion) back to 1978.  However, given the upheavals of the 1960s, it seemed possible that there had already been substantial changes even before that starting date, so I wanted to see if I could get anything from the 1950s or before.  Between 1945 and 1947, several survey firms asked "Which of these statements do you most agree with?...The most important job for the government is to make certain that there are good opportunities for each person to get ahead on his own, the most important job for the government is to guarantee every person a decent and steady job and standard of living."  Unfortunately, that question has never been repeated, but there has been a similar one,  "Should the federal government see to it that every person has a job and a good standard of living, or should the government stay out of it and let every person get ahead on their own?" which was asked a few times in the 1970s and 1980s and most recently in 1999.*  I took one of the 1946 surveys and the 1999 surveys and did logistic regressions with education (coded into four categories from not a high school graduate to college graduate) and log income (using the midpoints of the original categories) as independent variables.  The results:

                 1946       1999
Log inc     .155       .388
                (.061)     (.091)

Educ        .565         .102
                (.049)      (.065)

The positive sign means that higher values of the independent variables were associated with the "get ahead on their own" option.  The effect of education was substantially smaller in 1999 than it had been in 1946, and the effect of income was larger (although there's a good deal of uncertainty about changes in the effect of income--the standard error of the difference is .110). 

Putting this together with the question from my last post, I think that it's not just the relative importance of economic and social issues that has changed since the middle of the 20th century--another factor is that education used to have a substantial "conservatising" effect on economic opinions, but no longer does.   


*There was also one that asked people to place themselves on a 7 point scale running from "The government in Washington should see to it that every person has a job and a good standard of living." to "The government should just let each person get ahead on their own," which was included in the ANES in the 1980s and 1990s and revived in a PRRI survey in 2013. 

[Data from the Roper Center for Public Opinion Research]

Changes, part 1

In the 1950s, people with higher "social standing" were more likely to vote Republican.  That was true if you measured social standing by income, education, or some combination of the two factors.  Today, higher income continues to be associated with Republican voting, but college education is associated with a greater chance of voting Democratic.  One potential explanation for this change is that it reflects an increase in the importance of "social issues." More educated people have always been more liberal on social  issues, but the consensus in the middle of the 20th century was that they didn't matter very much--they would sometimes become prominent, but then would fade pretty quickly.  Basically, politics was about economic issues, and on these more educated people were more conservative.  But since the 1960s, "social issues"--e. g., abortion, affirmative action, gun control---have come to be consistently important.  So now income and education are pulling in different directions rather than reinforcing each other. 

In this account, the effects of education and income on opinions are constant--only the relative importance of different types of opinions changes.  What makes this appealing is that there are straightforward arguments for why education should be associated with more liberal opinions on social issues and more conservative opinions on economic issues,  and for why social issues should have  grown in importance.  So it rests on three points that are all clear and easy to understand.  I was reminded of this issue by  a recent column by Thomas Edsall, which points to a paper by Herbert Kitschelt and Phillipp Rehm, who argue that it explains changes in voting patterns through 2016.  They consider the possibility of change in the effects of education on opinions about economic issues, and find no evidence of it, but their data don't cover a long period of time.  In a previous post, I considered a question on redistribution from the General Social Survey that goes back to 1978.  It seemed that educational differences were declining, but I just considered education.  Suppose we take account of both income and education:

 
Positive numbers mean that higher values of income or education go with more conservative opinions on redistribution (the government should stay out of it).  The effects of income have stayed about the same, maybe declining a little, but the effects of education clearly have declined and are now near zero.  That is, it's apparently not just the relative importance of economic and social issues that has changed, but also the relationship between education and opinions on economic issues.  Why?  There's no obvious explanation.  In terms of material interests, if there was any change in the effect of education it should have been in the opposite direction, since the payoff to education has increased--that is, when you compare two people with the same income but different levels of education, the gap in expected lifetime earnings is bigger.