Reinstating the Rational Voter

Pity the economist. Unceremoniously accused of being unacceptably two-handed, he (or she) is now further accused of being two-faced, helplessly confused between teacher and research personas. Brian Caplan’s argument that rational voters are a myth (Caplan 2008) takes a direct swipe at this apparent inconsistency between economists’ rational choice modeling of the homo politicus and the homo sapiens that actually inhabit our world.

On its face, the argument is both plausible and convincing. With voters possessing heavily uninformed views on a multiplicity of basic economic issues, it would seem that our hope for a voting model premised on voters’ rational choices are merely academic exercises with little real-world relevance. What’s more, in this case the magic of (political) market discipline (Wittman 1989) appears to be unavailable to correct systematic misperceptions on the part of voters. Specifically, when mispricing occurs in the market, arbitrageurs restore the competitive market to an efficient outcome; thus, even with a preponderance of mistaken valuations about a given product, all we require for efficient pricing to be restored is one informed agent taking a sufficiently large position in the opposite direction. Since all political agents typically have only one vote, this institutional mechanism would not operate in a political market. Or could it?

Although the “quantity” dimension seems limited in political markets, it is in fact possible for a tendency toward efficiency in the political process. In particular, unlike goods markets—where competitive bidding is generally for a single, homogeneous product—political goods are multidimensional. While this introduces problems of its own (McKelvey 1976), to the extent that observed democratic outcomes are stable, the multidimensional nature of voting means that voters choose on the basis of their marginal concern, rather than every single issue at stake. As a consequence, even if the distribution of voter preferences for a given economic policy are heavily skewed to one (incorrect) side, systematic deviations from an efficient outcome will only occur when the marginal issue for all voters happen to coincide.[1] But as conversations with friends and neighbors can attest, one voter’s marginal issue could well be another’s non-issue. Your being wildly off the mark on, say, free trade would only matter if you are voting mainly on that basis; but it is just as likely (and perhaps more so) that you are voting on another issue for which you are in fact well-informed, such as abortion.[2]

We also have reason to believe that voters can, and often do, vote on the basis of marginal issues for which they are sufficiently well-informed. Voters routinely skip items on a ballot with which they are unfamiliar, and this could be the result of rational abstention (Feddersen & Pesendorfer 1996). If so, this self-selection into the voting pool would further reinforce the likelihood of an efficient outcome. The fundamental problem with Caplan’s argument is that he equates a voter being either uninformed, or incorrect, with him (or her) being irrational. This is an assumption, as being uninformed is neither necessary nor sufficient to invalidate the hypothesis that democracies produce efficient outcomes.[3]


Caplan, Bryan (2008), “What if the Median Voter Were a Failing Student?”, The Economists’ Voice 5(6) (September): Article 1. Available:

McKelvey, Richard D. (1976), “Intransitivities in Multidimensional Voting Models and Some Implications for Agenda Control”, Journal of Economic Theory 12(3) (June): 472–482.

Feddersen, Timothy J. & Wolfgang Pesendorfer (1996), “The Swing Voter’s Curse”, American Economic Review 86(3) (June): 408–424.

Wittman, Donald A. (1995), “Why Democracies Produce Efficient Results”, Journal of Political Economy 97(6) (Dec): 1395–1424.


1. This suggests that a given issue market may, in fact, be quite thin; efficient outcomes will nonetheless result so long as the marginal issues are evenly distributed across the space of all available issues, and there are sufficient voters on each side of the (efficient) median.

2. It may be argued, of course, that economic issues appear to matter to voting, as countless regressions would testify. This does not invalidate the argument, however, since regression coefficients capture the effect of changes at the margin. The significance of economic variables merely point to the fact that these factors matter in the final decision calculus, not that they matter more in any given instance of voting choice.

3. As such, I am not arguing that outcomes in a democracy may not be distorted by other factors, such as special interest lobbying. I am merely resisting the notion that the pervasiveness of uninformed voters must lead to a failure of democracies to achieve efficient outcomes.

This article is unpublished.