Prof. Bryan Caplan

Econ 410


Week 6: Basics of Public Opinion and Voter Motivation

I.                     Is the Median Voter Model Correct?

A.                 In order to determine whether or not the median voter model is correct, we must first find out "What voters want."

B.                 Once we know what voters want, we can see whether actual policy conforms to the policy preferences of the median voter.

C.                Probably the most popular account of voter motivation is that voters are essentially self-interested.

D.                Economists typically think this, but so do many political scientists, journalists, and "men in the street."

E.                 Example #1: "Rich people vote Republican, and poor people vote Democratic, because Republicans favor lower taxes and lower spending on redistribution than Democrats."

F.                 Example #2: "Blacks were treated worse under Jim Crow because they weren't allowed to vote.  Politicians didn't worry about losing their votes for racist policies."

G.                Example #3:  "People opposed to conservation laws must own stock in the timber industry."

II.                   Defining the Self-Interested Voter Hypothesis (SIVH)

A.                 There is a danger of tautology here: Is all behavior "self-interested" by definition?  Was Mother Theresa self-interested?

B.                 Throughout this course, I will only use the term "self-interest" in the falsifiable, ordinary language sense of directly valuing only one's own material well-being, health, safety, comfort, and so on.  Two provisos:

1.                  I interpret "people are self-interested" as "on average, people are at least 95% selfish," not "all people are 100% selfish."

2.                  Drawing on evolutionary psychology, I interpret altruism towards blood relatives in proportion to shared genes as self-interest.

C.                The self-interested voter hypothesis (SIVH) can then be defined as the hypothesis that political beliefs and actions of ordinary citizens are self-interested in the preceding sense.

III.                  Empirical Evidence on the SIVH

A.                 There is an enormous literature on the SIVH.  But the most convincing tests are those that rely on data concerning individuals' political beliefs and their personal characteristics (income, education, race, age, etc.) relevant to self-interest.

B.                 Amazing and important conclusion: the SIVH flops.  You can find some sporadic and debatable evidence for self-interested political beliefs, but that is about it.

C.                Consider the case of party identification.  Conventional wisdom tells us that "the poor" are Democrats and "the rich" are Republicans.

D.                In fact, the rich are only slightly more likely to be Republicans than Democrats. 

1.                  Race matters far more than income: Black millionaires much more likely to be Democrats than white minimum wage workers. 

2.                  Gender also dwarfs the effect of income: a man earning $25,000 per year is about as likely to be a Democrat as a women earning $100,000 per year.

E.                 The SIVH fails badly for individual issues as well.

1.                  Unemployment policy - The unemployed not much more in favor of relief measures.

2.                  National health insurance - The rich and people in good health are about as in favor.

3.                  Busing - Childless whites are as opposed as whites with children.

4.                  Crime - Crime victims and residents in dangerous neighborhoods are not much more likely to favor severe anti-crime measures.

5.                  Social Security and Medicare- The elderly are if anything slightly less in favor than the young.

6.                  Abortion - Men are slightly more pro-choice than women.

F.                 The SIVH fails for government spending, but has some moderate support for taxes. 

1.                  People expecting large tax savings from Proposition 13 were more likely to support it.

2.                  But recipients of government services and government employees were about as likely to support Prop. 13 as anyone else.

G.                The SIVH fails for potential death in combat!  Relatives and friends of military personnel in Vietnam were more in favor of the war than the rest of the population.  Similarly, draft-age males support the draft as strongly as other people.

1.                  Marginal evidence for SIVH - exact draft age.

H.                 Best example of a strong self-interest effect: Smoking! 

1.                  Even though smokers and non-smokers are demographically similar, non-smokers are much more in favor of restrictions on smoking. 

2.                  The heavier the smoker, the stronger the opposition. 

3.                  Only 13.9% of people who "never smoked" supported fewer restrictions, compared to 61.5% of "heavy smokers."

I.                     Overall, this body of evidence can only be described as revolutionary.  It is very hard to argue against it, and it means that most of what people think and write about politics is wrong.  Thousands of articles - and millions of conversations - have been a big waste of time because no one bothered to examine the empirical evidence.

J.                  Moreover, the empirical evidence is intuitive plausible.  Are your richer friends really the Republicans, and your poorer friends the Democrats?  Can you find any connection at all?  It isn't easy.

K.                 Thus, tests of the Median Voter Hypothesis that assume voters are self-interested are almost bound to fail.  Why?  If voters are not self-interested, then the failure of policy and the median voter's self-interest to "match" proves nothing.

IV.               Sociotropic and Ideological Voting

A.                 One major alternative to the SIVH, popular among many political scientists, is called "sociotropic voting."

B.                 Sociotropic voting means voting for policies that maximize "social welfare" or something along those lines.

C.                While the sociotropic voting hypothesis is over-simplified, it provides considerable insight.

D.                Ex: Good economic conditions make politicians more popular.  But what matters is mostly overall economic conditions, not those of the individual respondent.

E.                 Sociotropic voting explains quite a bit, but it does little to explain voter disagreement.  If everyone wants to maximize "social welfare," why don't they all vote the same way?

1.                  In contrast, the SIVH has a ready explanation for disagreement.

F.                 Leading answer: Ideological voting.  Most people think their policies are "public-interested," but there is ideological controversy about what policies "work" and what counts as "working."

G.                Ex: Affirmative action.  Conservatives and liberals argue about whether it works (are blacks better-off because of it?), but also disagree about what it means to "work" (a "level playing field" versus a "fair outcome"?).

H.                 So even though almost everyone favors "truth, justice, and the American way" (to quote Superman), ideological differences can explain why strident disagreement persists in spite of the SIVH's failure.

V.                 Education, Ideology, Income, and Opinion

A.                 For specific opinions (as opposed to party identification), income empirically often seems to make a large difference.

1.                  Ex: High income people seem much more in favor of immigration than low income people.

B.                 But the effect of income almost always disappears once you control for education.  Ph.D.s who drive cabs think like other Ph.D.s, not other cab drivers.

C.                Education and ideology are close to unrelated when you look at a random sample of Americans (as opposed to university faculty!).

D.                Yet both education and ideology exert large effects on people's politically-relevant opinions.

E.                 Example #1: Immigration. 

1.                  Opposition to immigration shrinks as education rises.

2.                  Opposition to immigration grows as conservatism rises. 

F.                 Example #2:  "Excessive profits."

1.                  Blaming "excessive profits" for economic difficulties falls as education rises.

2.                  Putting blame falls as conservativism rises.

G.                Fun fact: Education and ideology interact in an interesting way.  As education rises, ideology becomes a more powerful predictor of belief.  For grade-school drop-outs, there is little connection between their political beliefs and their stated ideological label.  For Ph.D.'s, this connection is very strong.

VI.               Group-Interested Voting

A.                 While the SIVH fails badly, there is some stronger evidence for group-interested voting.

B.                 What's the difference?  If a policy hurts you but helps your "group," how do you vote and think?  If you go with the group, your voting is group-interested, not self-interested.

C.                Ex: The black millionaire.  Democrats favor higher and more progressive taxes (which hurts the millionaire a lot), but also care more about the plight of blacks (which does virtually nothing for the millionaire; no one will discriminate against him).  If self-interested, he would vote Republican; if group-interested, he would vote Democratic.

D.                Much of the superficially plausible evidence for self-interested voting turns out to be group-interested when you look more deeply.

E.                 Ex:  Jewish support for Israel.

VII.              Case Study: The Determinants of Party Identification

A.                 What happens if you use basic econometrics to try to sort out the determinants of party identification?

B.                 Linear probability model: Predict the probability of being a Democrat or being a Republican conditional on your personal characteristics.

C.                What if you ignore ideology, and try to predict party identification using only income, job security, income growth, education, gender, age, and race?

D.                [Handout Table 2a] 

1.                  Income.  There is some slight evidence that income matters in the expected direction. 

2.                  Job security and income growth.  People with more job security are a bit more Republican and perhaps a little more optimistic about their future income.

3.                  Education.  Irrelevant.

4.                  Gender.  Males are markedly less likely to be Democrats (6.4 percentage points).

5.                  Age.  Older people are a little more likely to be Democrats.

6.                  Race.  Blacks and "other races" (essentially Hispanics), but not Asians, are massively more likely to be Democrats and less likely to be Republicans.

E.                 Question:  What if you also control for stated ideology (very liberal, liberal, moderate, conservative, very conservative)?

F.                 [Handout Table 2b]

G.                Answer:  Ideology matters even more than race.  Moreover, the slight change in the other coefficients shows that ideology is far from a "mere proxy for self-interest."

H.                 Consider two examples.

1.                  Ex. #1:  Black female millionaire Ph.D., with maximal job security and income growth, 30 years old.

2.                  Ex. #2:  White male minimum wage worker, minimal job security and income growth, 30 years old, high school education, very conservative ideology.

I.                     Ex. #1:  [Since we don't know ideology, use Table 2a]  Estimated probability of being a Democrat: 58.8%; estimated probability of being a Republican: 16.9%.

J.                  Ex. #2:  [Using Table 2b]  Estimated probability of being a Democrat: 10.9%; estimated probability of being a Republican: 49%.

K.                 What does all this show? 

1.                  Very strong evidence for ideological voting.

2.                  Strong evidence for group-interested voting, with race being the main group of interest.  (How could this be distinguished from self-interested voting?)

3.                  Self-interest plays a marginal role at most.

VIII.            The Dimensionality of U.S. Political Opinion

A.                 There are many different ways to analyze political beliefs.

1.                  Libertarian-statist spectrum.

2.                  Christian-secular spectrum

B.                 But empirically, U.S. political opinion "fits" quite well on the liberal-conservative or left-right spectrum.

C.                On a deep level, this spectrum may not make a great deal of sense.  Libertarians, for example, often argue that there are really two dimensions - personal freedom and economic freedom:

1.                  Libertarians - pro-personal, pro-economic

2.                  Populists - anti-personal, anti-economic

3.                  Liberals - pro-personal, anti-economic

4.                  Conservatives - anti-personal, pro-economic

D.                But again empirically, most people line up on the diagonal, and the other two boxes are sparsely inhabited. 

E.                 Political beliefs are admittedly fairly random; but once you know someone's place on the liberal-conservative spectrum, you often have as good a prediction as you can hope for.

F.                 Key implication: In spite of the multitude of issues, real-world politics fits the single-issue median voter model well because a single liberal-conservative dimension underlies issue attitudes.

G.                Complication #1: Ideological labels give better predictions for the well-educated.  Education "stretches" the liberal-conservative spectrum.

H.                 Complication #2: Returning to the two-dimensional diagram, education probably shifts the diagonal up and to the right.  More educated people tend to be both more tolerant and more appreciative of free markets.

I.                     This fact suggests that politicians might really compete over two dimensions rather than one.   

J.                  In practice, however, the liberal-conservative dimension appears to be far more electorally salient.  Education affects issue beliefs, but appears to be independent of party identification.  Real-world politicians focus on the left-right dimension far more than the education dimension.

IX.               The SIVH Versus the Logic of Collective Action

A.                 How is all this ideological, unselfish voting possible?  It seems to conflict with the logic of collective action - people sacrifice their own political interests without hope of compensation.

B.                 But this impression is misleading.  Why?  Precisely because one vote is extraordinarily unlikely to change an electoral outcome, it is very safe to vote against your own interests!

C.                Ex:  When Barbara Streisand votes for a candidate that will charge her $2 M more in taxes, is that equivalent to giving $2 M to charity?

D.                Of course not.  Her vote won't change the election's outcome.  If the Democrat wins, she has to pay, but he would have won - and she would have to pay - anyway!  So the MC of voting Democratic is not $2 M, but $2 M times the probability that she casts the decisive vote.  Even if that were a high 1-in-2 M, her expected cost of voting Democratic would only be $1.00.

E.                 The logic of collective action cuts two ways.  It makes people unwilling to contribute serious effort for political change.  But it also makes people unafraid of voting contrary to their own interests.


Table 2a: Conditional Probability of Being a Democrat/Republican






Independent Variables















Job security





Recent Income Growth





Expected Income Growth



































Other race





Mean Dep. Variable



SD Dep. Variable









*=p<.05    **=p<.01    ***=p<.001




Table 2b: Conditional Probability of Being a Democrat/Republican






Independent Variables















Other Ideology










Job security





Recent Income Growth





Expected Income Growth



































Other race





Mean Dependent Var



SD Dep. Variable









*=p<.05    **=p<.01    ***=p<.001