RATIONAL IGNORANCE (Public Choice)

1. Theory

Information is a good like any other. The primary benefit of information is that it reduces the probability of acting on false beliefs; the primary cost is that acquiring information requires time. Basic microeconomics predicts that (ignoring risk-aversion) individuals acquire information up to the point where the expected marginal benefits equal the expected marginal costs (Stigler, 1961). Beyond that point, acquiring information becomes selfishly counterproductive; while you will avoid more mistakes, it is on balance cheaper to commit them.

A corollary is that if the marginal benefits of information are always zero, the rational economic decision is to be ignorant. If the market pays nothing for knowledge of ancient Egypt, there is no reason to spend time learning about it. In general terms, then, "rational ignorance" refers to any situation where individuals know little or nothing because the expected benefits of knowledge are negligible.

However, it has long been recognized that rational ignorance has far more empirical relevance in public choice than in other branches of economics. Why? Suppose that spending one more hour learning about politicians’ voting records allows you to shift your vote to a candidate whose policies, if adopted, would be $100 better for you. The expected marginal benefit of an hour of study is emphatically not $100, but $100 multiplied by the probability that you cast the decisive vote, tipping an otherwise deadlocked outcome. In virtually any real-world election, that probability will be essentially zero, implying an expected marginal benefit of zero as well (Olson, 1965; Downs, 1957).

The upshot is that imperfect information matters far more in politics than in markets. Consumers are not omniscient, but they have clear incentives to roughly figure out whether a piece of merchandise is worth the asking price. They are ignorant of details of the marketplace, not its basics. In contrast, voters’ have no more incentive to study the basics of politics than they have to study the minutiae! Even if a person voted in a completely random manner, he could still enjoy personal comfort and security. A person who consumed in a completely random manner could not.

2. Evidence

At any rate, this is what the economics of information tells us. But do these predictions hold up empirically? Obviously they are not literally true. Everyone knows something about politics. Nevertheless, the empirical evidence on political knowledge reveals that citizens are ignorant to a shocking degree. Consider the following table showing the percentage of adult Americans aware of various elementary political facts.

More comprehensive works (e.g. Delli Carpini and Keeter, 1996) are quite consistent with this outline; in fact, they show that knowledge of foreign affairs is even more limited. It is particularly striking that such a small percentage knows unchanging characteristics of the U.S. Constitutional structure like the number of senators each state has or the length of House members’ terms. Without this knowledge, it is hard to see how voters could hold the politicians who represent them accountable for anything. Politicians’ low level of name recognition is less surprising, but similarly disheartening: if voters are unable even to name their representatives, it is wishful thinking to imagine that they keep track of their voting records and reward or punish them accordingly in the next election. Politicians’ party affiliations obviously simplifies this problem to some degree, but voter knowledge of the party to which second-tier politicians belong is also quite limited.

3. Consequences

One of the main themes of public choice is that democracy is paradoxically run for the benefit of "rent-seeking"special interests rather than the general public. Rational ignorance is at the root of most efforts to model this outcome. If voters have no idea what politicians are doing, then politicians can safely cater to the special interests who make it their business to closely follow whatever aspects of policy they depend on. This pattern often goes by the name of "concentrated benefits, diffuse costs." Each piece of special interest legislation takes only a few cents from the average voter, but provides millions for the special interests who back it. No voter, therefore, will bother to even learn of the existence of the legislation, but its beneficiaries may well employ a full-time staff of lobbyists to protect their livelihood.

A more complex account of the connection between rational ignorance and the predominance of special interests focuses on campaign contributions. Special interests can use their money to buy politicians’ support. These politicians then use donations to pay for misleading political advertising. Precisely because voters are so uninformed, this misleading advertising is generally an effective way to get votes. In equilibrium, of course, politicians have to strike a balance between special interests and voter interests, but the less informed the voters are, the less their interests count (Grossman and Helpman, 1994).

4. Responses

The empirical evidence on voter ignorance is widely, though not universally, accepted. A minority of scholars maintain that "political IQ" tests understate the true competence of the individual voter (Popkin, 1991; Lupia and McCubbins, 1998). The most effective critiques of the standard analysis, however, concede voters’ severe ignorance, but maintain that even so, democracy works remarkably well.

The leading version of this argument is sometimes called "the miracle of aggregation" (Page and Shapiro, 1992; Wittman, 1995; Hoffman, 1998). Suppose that only 10% of voters are well-informed; they vote for the candidate who puts more weight on voter welfare. The remaining 90% are completely uninformed and vote by flipping a coin. Common sense says that this situation is hopeless. Basic statistics, however, implies that as the size of the electorate increases, the randomness of individual voters matters less and less. In percentage terms, errors tend to cancel each other out. In fact, for a reasonably large electorate, we can practically guarantee that in a two-party race, each candidate gets half of the uninformed vote. No matter what a candidate does, he gets at least 45% of the votes. To win, though, a candidate needs more; and the only way to get more is to court the informed voters. What fraction of the well-informed does a politician need to win? 51%. In other words, whichever candidate the majority of well-informed voters prefers wins the election, even though by assumption the well-informed are out-numbered 9:1.

A parallel argument holds for ignorance about policy. Suppose that most voters are ignorant about international economics. As long as this ignorance leads equal fractions of voters to over- or under-estimate the benefits of free trade, the equilibrium policy remains the same. Voter ignorance by itself therefore does not imply that protectionism prevails. As long as errors are random rather than systematic, and the electorate is large, an ignorant electorate acts "as if" it were fully informed.

This sort of reasoning is especially persuasive to economists because it is an obvious extension of the familiar rational expectations assumption. Most formal economic models routinely represent imperfect information as measurement error; they assume that individuals observe the "true value plus noise." In Akerlof’s (1970) famous article on asymmetric information, for example, ignorant car buyers know the average value of used cars, but nothing about any particular car. From a rational expectations standpoint, systematic errors are by definition "irrational," a condition that most economists are unwilling to invoke.

Aside from the miracle of aggregation, another source of optimism in the face of voter ignorance emerges from the economics of crime (Becker, 1968). This literature suggests that politicians can be kept in check even if voter ignorance occasionally rises up to 100%. All that is necessary is that politicians have a strictly positive probability of being caught if they abuse the public trust. If so, voters have a simple way to deter abuses: over-punish whoever they catch. If a politician has a 1% chance of being caught taking a $1 bribe, bribery has to be punished with far more than a $1 fine. Voters could penalize wayward politicians for even minor infractions by throwing them out of office, demonizing them, or putting them in jail. This is not idle theorizing, either; a number of politicians have ruined their careers by saying the wrong word at the wrong time, or by misusing a trivial amount of official resources.

The "concentrated benefits, diffuse costs" story has also been specifically challenged. Perhaps if there were only a single program in the federal budget that fit this description, that program would survive indefinitely. But most economists concerned about programs with concentrated benefits and diffuse costs believe that they number in the thousands. If so, there is a simple way to get rid of them: set up some sort of omnibus repeal bill (Wittman, 1995). Bundle thousands of tiny affronts against the average taxpayer into one bill that pulls the plug on all of them. In this way, politicians could overcome the dilemma that no one program is big enough to attract voters’ attention. Interestingly, after economists raised this theoretical possibility, U.S. politics provided a clear empirical illustration with the post-Cold War base closings bill.

Finally, even if one believes that voter ignorance remains a serious problem for democracy, it is important to carefully work through the direction of the effect. Rational ignorance has generally been thought to make government inefficiently large. But standard asymmetric information models actually predict the opposite (Breton and Wintrobe, 1982). In the Akerlof lemons model, to take the canonical example, informed sellers and uninformed buyers leads the used car market to shrink. The reason is that buyers realize that they are unable to judge product quality, and therefore become more reluctant to buy. In a parallel manner, if political insiders know more about program quality than voters, voters’ rational response is, in effect, to say "When in doubt, vote no." In equilibrium, then, asymmetric political information tends to make government smaller, not bigger.