Prof. Bryan Caplan

Econ 854


Week 10: Ignorance, Irrationality, and Aggregation: Theory and Evidence

I.             Return to the "Miracle of Aggregation"

A.           The leading explanation for democratic failure is voters’ rational ignorance.

B.           There can be little doubt that voters are highly ignorant.

C.           But as discussed earlier, many assert that voter ignorance is quite compatible with well-functioning democracy!

D.           Why?  If we interpret "ignorance" as "random error," then the Principle of Aggregation kicks in.  If you tabulate millions of random errors and take an average, the aggregate acts "as if" it were fully informed.  That is the Miracle of Aggregation.

E.           So far, so good.  But why should we believe that voters' errors ARE random in the first place?

II.            Ignorance, Irrationality, and Systematic Error

A.           There are two distinct ways economists apply the concept of “rationality”:

1.            Rationality of action

2.            Rationality of belief

B.           In the last thirty years, rational expectations has been a standard technique for modeling economic actors’ beliefs.  Economists often refer to “rationality” and “rational expectations” interchangeably.

C.           Key feature of RE: calibration.  RE requires some connection, albeit imperfect, between agents’ beliefs and the real world.

D.           RE partitions error between:

1.            Irrationality - the systematic component

2.            Ignorance - the random component

E.           RE then rules out the first type of error.  Non-random errors ipso facto become evidence of "irrationality."

F.            But merely defining systematic errors as "irrational" is hardly evidence that they don't exist on a wide scale.  Maybe RE is false, and in that sense, people are not rational.  It is an empirical question.

G.           There are weaker definitions of rationality that allow mere ignorance to co-exist with systematic error. 

1.            Bayesian rationality, for example, merely demands that people update their beliefs in a certain way, but puts no constraints on their priors.  These may be wildly unrealistic.

H.           A still weaker sense of rationality: truth-seeking.  However deluded they are, agents qualify as long as they want to have true beliefs.

I.             These weaker senses of irrationality still have some connection to systematic error.  If you do not update your beliefs conditional on evidence, or if you do not care about truth, you are more likely to have wildly unrealistic views.

J.            If you switch to a non-RE definition, you can save “rationality,” but rationality is no longer enough to make democracy work.

K.           In practice, most economists do equate “rationality” with RE, so I will stick with this definition throughout the lecture.

III.          Rational Ignorance Versus Rational Irrationality

A.           What reason is there to believe that the rational expectations assumption is true?

B.           The main argument is that systematic errors are costly, so people try to:

1.            Avoid them in the first place.

2.            Learn from the systematic mistakes they do happen to make.

C.           Big problem here: Some systematic errors are less costly than others, and some can hardly be called costly at all.

D.           One of my main ideas: Just as economists think of agents weighing the costs and benefits of information, so too can we think of agents weighing the costs and benefits of rationality.  Just as it is sometimes rational to be ignorant (have little information), it may sometimes be rational to be irrational (deviate from full rationality).

1.            Psychological interpretation?

E.           In other words, we can think of irrationality as a normal good.  Why does anyone want this "good"?

1.            Big reason: People derive comfort, security, and sense of identity from their belief structure – and rational thinking is often hard, painful, discouraging work.

2.            Indirect reason: Other people you depend on may treat you differently depending on your beliefs.

3.            For more: See Mosca, The Ruling Class, chapter 7 (on the syllabus)

F.            What is the "price" of irrationality?  It is the material success that you give up in order to retain systematically mistaken beliefs.

G.           Writing down an individual's "demand for irrationality" curve for a given issue is easy.  Just put quantity of irrationality on the x-axis, and the implicit price of irrationality on the y-axis.

1.            Neoclassical demand for irrationality

2.            Near-neoclassical demand for irrationality

H.           When the price of irrationality is high - as it often will be - people consume less.  Perhaps they consume none at all - on at least some issues, they might be fully rational.

I.             When the price of irrationality is low, people consume more.  When irrationality is completely free, people stick with whatever belief makes them most happy, however crazy.

J.            Remember our old friend, the probability of voter decisiveness? 

K.           Immediate implication: The expected price of voter irrationality is essentially zero, so we should not be surprised if voters hold highly irrational beliefs!

L.            Question: How is this different from expressive voting?

M.          Answer: Expressive voting says that people don't really care if policies work.  Rational irrationality says people believe their favored policies do work, but have irrational beliefs about what works!

1.            Ex: The public reaction to WWI.

V.           Systematically Biased Beliefs About Economics

A.           There are many subject matters where irrational beliefs may lead to inefficient policy. 

B.           But one subject matter that seems especially interesting for public choice is economics itself.

1.            Most policy decisions of modern government have significant economic content.

2.            Economists have written about economic misconceptions for hundreds of years - most famously, French economist Frederic Bastiat.

C.           I have done a lot of empirical work on this topic; chapter 3 of The Myth of the Rational Voter summarizes it.

1.            Data: the Survey of Americans and Economists on the Economy (SAEE).  1510 members of the general public, 250 Ph.D. economists.

D.           Standard method of testing for irrationality: Look for differences in mean beliefs of laypeople and experts.

E.           Complication: Critics of economists claim that it is the economists who are biased rather than the public!

1.            Self-serving bias

2.            Ideological bias

F.            In my empirical work, however, I am able to show that large systematic belief differences persist controlling for self-serving and ideological bias.  [Tables]

G.           What main clusters of systematic belief differences emerge?

1.            Anti-market bias

2.            Anti-foreign bias

3.            Make-work bias

4.            Pessimistic bias

H.           What kinds of inefficient policies could each of these four categories explain?

VI.          Group Differences in Economic Beliefs

A.           If you buy my evidence on systematically biased beliefs, the distribution of bias becomes a pressing issue for democracy: Is bias uniformly distributed, or concentrated in specific parts of the population?

B.           While biases appear in all major segments of the public, the following factors reduce bias in the SAEE:

1.            Education

2.            Being male

3.            Job security

4.            Income growth

C.           What does NOT make people think like economists?

1.            Income level

2.            Conservatism

D.           What does this mean?  Presumably, when median economic literacy falls (whether due to franchise rules or personal choice), policy gets worse, because the median voter’s biases are more severe and politicians have to cater to them. 

1.            Ex: Policy is probably better than it would otherwise be due to the higher turnout of the well-educated.

E.           In subsequent work with Stephen Miller (“Intelligence Makes People Think Like Economists”), we found that much of the apparent effect of education is actually an effect of IQ.  IQ is the strongest overall predictor of “thinking like an economist.” 

F.            Controlling for IQ, education is the second strongest predictor of economistic thinking – and remains the strongest antidote to anti-foreign bias.

VII.        Systematically Biased Beliefs About Other Subjects?

A.           Foreign policy?

B.           A large literature in political science documents a “rally-round-the-flag effect.”  FDR’s approval jumped 12 percentage-points after Pearl Harbor.  Bush’s approval rose 35 percentage-points after 9/11!  This seems hard for a rational voter model to explain.  Why should failure make leaders more popular?  Even if you think you have an answer, why would this extra popularity predictably erode?

C.            Other misconceptions about international affairs also seem to have a strong effect on voter preferences.  Ex: “According to an October 21, 2004 Harris Poll, 52 percent of those who preferred Bush thought that Saddam had helped plan and support the hijackers who attacked the U.S. on September 11 (it was 23 percent for those who preferred Kerry) and 58% of those who preferred Bush thought that Iraq had weapons of mass destruction when the U.S. invaded (it was 16% for those who preferred Kerry). Neither of these assertions is true.” (Donald Wittman, “Reply to Caplan” in EJW)  Original survey at:

D.           Misconceptions are not marginal.  Consider Hitler’s argument for conquest: Germany won’t be able feed itself with its current land area, and trade is not a viable solution.  It was a key motive for World War II, yet after the war, Germany grew rich following the strategy that Hitler dismissed.

E.           Global warming?  A strong consensus of climate scientists does agree that it is a real and serious problem.  (See Figures 28-30, 

F.            If laymen accepted the expert consensus, would they favor different policies?

G.           Other areas?

VIII.       The Enlightened Preference Approach

A.           In political science, there is a large literature on “Enlightened Preferences” that also seriously undermines the Miracle of Aggregation.  Best summary of the literature: Scott Althaus’ Collective Preferences in Democratic Politics.

B.           Basic idea: Regress policy preferences on standard variables and a measure of objective political knowledge.  Then simulate the distribution of Enlightened Preferences – i.e. what preferences people would have if everyone had the highest level of objective political knowledge. 

C.           Complication: You can allow the coefficient on political knowledge to vary by sub-group.  Ex: Maybe well-informed people with high income are less supportive of progressive taxation than poorly-informed people with high income, but well-informed people with low income are more supportive of progressive taxation than poorly-informed people with low income.

D.           According to the Miracle of Aggregation, Enlightened Preferences will have the same mean as actual preferences.

E.           Key finding #1: The Miracle of Aggregation fails badly again.  Enlightened preferences are almost always noticeably different from actual preferences.

F.            Key finding #2: Knowledge usually works in the same direction for diverse sub-groups.  In fact, the absolute magnitude of the coefficient is often larger for the groups that normally oppose a given policy.  Example: Preferences for free markets vs. government.

G.           Key finding #3: Enlightened preferences are more economically conservative and socially libel than actual preferences. (summary table, p.129)

H.           Examples (pp.109, 111, 115, 116)

I.             Closing thought: Enlightened Preference results are based on questions that are easy in absolute terms.  So the “maximum” level of knowledge in the simulations is still fairly low.  What would estimated Enlightened Preferences be if the questions were much more demanding?