When Ignorance Isn’t Bliss: How Political Ignorance Threatens Democracy

Inadequate voter knowledge has two major negative implications for democracy. First, it prevents democratic government from reflecting the will of the people in any meaningful sense, undercutting the “intrinsicist” defense of democracy as a government that reflects the voluntary decisions of the populace. Likewise, voter ignorance imperils the instrumental case for democracy as a regime that serves the interests of the majority, since ignorance potentially opens the door for both elite manipulation of the public and gross policy errors caused by politicians’ need to appeal to an ignorant electorate in order to win office.

In this paper I review the overwhelming evidence that the American electorate fails to meet even minimal criteria for adequate voter knowledge. I then examine the implications for American politics. Part I lays out minimal knowledge prerequisites for voter control of public policy, summarizes the massive evidence of voter ignorance that students of the subject have accumulated over the years, and highlights some of the most disturbing implications of those studies. Part II examines more recent evidence of widespread political ignorance. It shows that extensive voter ignorance plagued the 2000 presidential election and apparently continues during the current election cycle. These data are significant because the extremely close and controversial nature of those two elections might have been expected to cause an increase in voter knowledge. In Part III, I review and criticize theories that claim that “information shortcuts” enable voters to control government in spite of pervasive ignorance. Those mechanisms for dealing with voter ignorance are unable to overcome it and sometimes even exacerbate the problem. Part IV restates the argument that ignorance is largely “rational,” rooted in the very low likelihood of a single vote being able to influence electoral outcomes.

Part V advances the claim that the size and scope of the modern state are so great that it is often impossible for voters—even the most knowledgeable among them—to be adequately informed about its operations. This conclusion leads us to the counterintuitive suggestion that a smaller government may actually be more democratic than that which we have now, in so far as voters would stand a greater chance of being able to influence government policy in an informed manner. I briefly examine some evidence from 19th century American history that suggests that the much less educated electorate of that era was able to consider far more complex policy arguments than those that are presented to it today and hypothesize that this difference was largely due to the ability of voters to focus on the relatively small number of issues over which the strictly limited government of that era exercised control.

Finally, Part VI shows that voter ignorance provides an unanticipated argument in favor of decentralized federalism. Decentralization allows citizens to “vote with their feet” by moving out of jurisdictions with policies they dislike and into those that have more favorable ones. Because each person can decide for herself whether or not she will move, there is a much greater incentive to acquire relevant information with “foot voting” than with traditional ballot‐​box voting.

To the extent that the arguments of this study are correct, efforts to increase the stock of knowledge possessed by voters are unlikely to be more than modestly effective. A more promising path for reform would be to reduce the amount of knowledge required for democratic control of public policy.

Finally, the study’s conclusion stresses the tradeoff between big government and democratic government that all advanced industrialized states must face.

What must voters know in order to exercise democratic control over government policy? In their classic work, The American Voter, a University of Michigan Survey Research Center team defined three minimal knowledge prerequisites for voters to be able to exert meaningful influence over a given issue:

Those three conditions have formed the basis for most empirical investigations of individual‐​level voter ignorance since The American Voter. But they are insufficient prerequisites for meaningful control over public policy. In addition to awareness of the existence of relevant issues and of candidate positions on them, informed voters must have at least substantial understanding about which of the available policy options are most likely to advance their goals. Unless the value voters attach to policy in a given area is purely a matter of symbolic “position taking,” they cannot use the ballot to force elected officials to serve their interests without knowing what the likely effects of alternative policy options are.

Just how informed should voters be? Ideally, they should be able to choose between opposing candidates and their platforms on the basis of “the preferences that people would have if their information were perfect.” Although that is impossible, minimally informed voters should at least be aware of basic tradeoffs between alternative policies in cases in which those tradeoffs would be immediately obvious to political elites informed about the issue at hand. Moreover, minimally informed voters should not draw linkages between policies and outcomes that informed observers would consider obviously absurd; for instance, the fact that a majority of American voters with an opinion on the issue believe that the federal government is too large and powerful and simultaneously favor increased spending in almost every major area of federal involvement is a clear case of ignorance of tradeoffs that falls below the threshold of minimally necessary knowledge. On the other hand, where there is clear disagreement among the well informed about the effects of a particular policy, support of or opposition to that policy should not in and of itself count as presumptive evidence of voter ignorance. That standard adds a crucial substantive dimension to the criteria for minimally necessary voter knowledge without judging voters’ competence by how closely their policy preferences match those of the analyst. Moreover, nothing in the definition forecloses the possibility that it may not be necessary for each individual voter to possess the requisite knowledge so long as the electorate in the aggregate possesses sufficient information and signaling capacity to act “as if” voters were individually informed.

As political scientist John Ferejohn has written, “Nothing strikes the student of public opinion and democracy more forcefully than the paucity of information most people possess about politics.” Few people dispute the well‐​established conclusion that most individual voters are abysmally ignorant of even very basic political information. Ever since the seminal research of the 1950s and early 1960s, evidence has accumulated to reinforce this finding.

Nonetheless, the sheer depth of most individual voters’ ignorance is shocking to observers not familiar with the research. Currently, almost 70 percent of Americans do not know that Congress recently adopted a law adding a massive prescription drug benefit to the Medicare program, the largest new federal entitlement in decades, and arguably the most important piece of domestic legislation adopted during the administration of George W. Bush. Equally striking is the fact that more than 60 percent do not realize that a massive increase in domestic spending has made a substantial contribution to the recent explosion in the federal deficit.

A survey taken immediately after the closely contested November 2002 congressional elections found that only about 32 percent of respondents knew that the Republicans controlled the House of Representatives prior to the election. That result is consistent with research showing widespread ignorance of congressional party control in previous elections.

Such widespread ignorance is not of recent origin. As of December 1994, a month after the takeover of Congress by Newt Gingrich’s Republicans, 57 percent of Americans had never even heard of Gingrich, whose campaign strategy and policy stances had received massive publicity in the immediately preceding weeks. In 1964, in the midst of the Cold War, only 38 percent were aware that the Soviet Union was not a member of NATO. Most of the time, only bare majorities know which party has control of the Senate, some 70 percent cannot name either of their state’s senators, and the vast majority cannot name any congressional candidate in their district at the height of a campaign. Overall, close to one‐​third of Americans can be categorized as “know‐​nothings” almost completely ignorant of relevant political information.

Three aspects of voter ignorance deserve particular attention. First, voters are ignorant not just about specific policy issues but also about the basic structure of government and how it operates. Majorities are ignorant of such basic aspects of the U.S. political system as who has the power to declare war, the respective functions of the three branches of government, and who controls monetary policy. That suggests not only that voters cannot choose between specific competing policy programs but also that they cannot easily assign credit and blame for highly visible policy outcomes to the right officeholders. The long‐​noted tendency of voters to almost automatically attribute “good times” to incumbents is one reflection of this problem.

The second salient aspect of voter ignorance is that most voters lack an “ideological” view of politics capable of integrating multiple issues into a single analytical framework derived from a few basic principles; ordinary voters rarely exhibit the kind of ideologically consistent stance on issues that is evident in surveys of political elites. Most scholars follow Anthony Downs in emphasizing the usefulness of ideology as a “shortcut” to predicting the likely policies of opposing parties competing for office. Generally ignored, but at least equally important, is the comparative inability of non‐​ideological voters to spot interconnections among issues. An ideologically aware electorate would not be oblivious to the contradiction between seeking a reduction in government power and an expansion of nearly all its major programs.

Voter ignorance of basic institutional structure and of ideological interconnections creates serious obstacles to democratic control of government. Without knowledge of the basic “rules of the game” of politics and of ideological interconnections between issues, there is unlikely to be a sufficient preexisting base of knowledge for the impressions gained from campaign information to be accurate; even if they are accurate, they cannot be easily connected with the voter’s policy objectives. It is no surprise, therefore, that the small minority of well‐​informed voters is much better able to process new political information and more resistant to manipulation than is the uninformed mass public.

Finally, it is important to note that the level of political knowledge in the American electorate has increased only very slightly, if at all, since the beginning of mass survey research in the late 1930s. A relatively stable level of extreme ignorance has persisted even in the face of massive increases in educational attainment and an unprecedented expansion in the quantity and quality of information available to the general public at little cost. This striking failure throws doubt on the expectation of political theorists from John Stuart Mill onward that increased availability of formal education can create the informed electorate that the democratic ideal requires.

Individual‐​level voter ignorance seems deeply rooted, perhaps ineradicable. It follows that the ability of voters to meet the conditions of the democratic ideal is dependent on the validity of “shortcut” theories that predict that voters can cast informed votes without themselves possessing even minimal levels of political knowledge.

This part considers recent evidence of widespread political ignorance in the United States. I consider limited data available from the current election cycle, as well as more systematic data from the time of the 2000 election. The analysis of voter ignorance during the 2000 election is based on data from the 2000 National Election Study, an extensive nationwide survey of more than 1,800 respondents that included 31 political knowledge items covering a wide range of subjects.

We do not as yet have a large‐​scale comprehensive data set on political knowledge in the current election cycle. However, Table 1 presents evidence from a number of recent surveys that indicates extensive political ignorance on major issues in the current campaign.

The available data cover a number of basic questions related to widely discussed issues that are currently prominent in both the press and political debate. Perhaps the most disturbing result is that large majorities are unaware of the passage of some of the most important and controversial items on the Bush administration’s domestic policy agenda: almost 70 percent did not know of the passage of the massive Medicare prescription drug benefit, and nearly 65 percent do not know of the recent passage of a ban on partial birth abortion. Similarly, 58 percent admit they have heard “very little” or “nothing” about the USA Patriot Act, the much‐​debated 2001 legislation that increases law enforcement powers for the claimed purpose of fighting terrorism. This survey result probably actually understates the number of respondents who know little or nothing about the act.

The survey evidence also indicates considerable ignorance about various hot‐​button domestic and foreign policy issues. Despite widespread press coverage of large recent job gains,33 the majority of respondents in a June 7 poll mistakenly believed that there had been a net loss of jobs in 2004. With regard to the most important foreign policy issue in the campaign, a majority mistakenly believed that the Bush administration sees a link between Saddam Hussein and the September 11 attacks (despite the administration’s own repeated disclaimers of any such connection), and most do not know even approximately how many American lives have been lost in the Iraq war. Despite the ongoing debate over America’s troubled relationship with Europe and the onset of European unification, 77 percent admit they know “little” or “nothing” about the European Union.

Particularly significant is the fact that, on many issues, the majority is not only ignorant of the truth but actively misinformed. For example, 61 percent believe that there has been a net loss of jobs in 2004, 58 percent believe that the administration sees a link between Saddam Hussein and 9/11, and 57 percent believe that increases in domestic spending have not contributed significantly to the current federal budget deficit. Whether those misconceptions will have an impact on the outcome in November remains to be seen.

The data in Table 1 should not be taken as proof that the public is universally ignorant on every issue. Some basic facts about current public policy are well known. For example, 82 percent know that there is currently a federal budget deficit, and 79 percent know that the deficit has increased during the last four years. Nonetheless the evidence compiled in Table 1 does show that majorities are ignorant of numerous basic facts about some of the most important and most widely debated issues at stake in the present election. That result is particularly striking in view of the extremely close and controversial nature of the contest and the high level of press coverage many of those issues have received.

The limited evidence of widespread ignorance in the current election cycle is powerfully reinforced by much more systematic data from the 2000 election provided by the 2000 National Election Survey. Undertaken during every election year since 1948, the NES is generally considered the most thorough social scientific survey of the U.S. electorate.

The 2000 NES contained a total of 31 political knowledge questions. They are listed in Table 2 along with the percentage of respondents giving correct answers.

Nearly all of the 31 survey items identified in Table 2 are quite basic in nature and would have been well‐​known to political elites and activists at the time. Most addressed issues that were widely debated during the 2000 campaign, including environmental policy, government spending on services, abortion, and policy toward African Americans. Several questions related to factual matters relevant to the record of the Clinton administration, for which presidential candidate Al Gore and the Democratic Party more generally attempted to claim a share of credit. Although the 31 questions do not cover all possible relevant issues and facts, they do include a wide range and are therefore a good representative sampling of Americans’ political knowledge. Moreover, previous studies have found that political knowledge in one area is highly inter‐​correlated with knowledge in others. Thus, we can be reasonably confident that individuals who scored well on the 31 items in the 2000 NES also possess greater political knowledge on other matters than those who scored low.

The average knowledge level in the 2000 NES was roughly similar to that detected in earlier studies and generally low. On average, respondents answered correctly only 14.4 questions out of 31. The data also seem to confirm Stephen Bennett’s findings that about one‐​third of respondents are “know‐​nothings” possessing little or no politically relevant knowledge. About 25 percent of respondents got 8.5 or fewer correct answers. Since 17 of the 31 questions had only three possible answers, two had only two possible answers, one more had two correct answers out of a possible three, and several others could also potentially be guessed with lower probabilities of success, a score of 8.5 is almost exactly equal to the score that could be expected as the result of random guessing. My finding of 25 percent “know‐​nothings” is very similar to Bennett’s finding of 29 percent.

Nonetheless, it is possible to argue that the average knowledge level revealed in the 2000 NES is not too low because the average respondent did achieve correct answers on almost half the questions (46 percent). This claim is flawed for two reasons. First, with minor exceptions, the items in the survey represent very basic political knowledge, without which it is difficult or impossible to place more complex and specific knowledge in useful context. Knowledgeable political activists and even citizens who follow politics reasonably closely would probably be able to answer all but a tiny handful of the questions correctly.

The second reason for pessimism regarding the 2000 NES results is that they probably actually overestimate current American political knowledge. That overestimation is the result of two factors. First, surveys in general somewhat overestimate the amount of political information possessed by the public because of the possibility of guessing by respondents and because more knowledgeable citizens may be over‐​represented among those surveyed. The average respondent in the 2000 NES got only about 6 more correct answers out of 31 than would be expected as a result of random guessing. Although NES respondents had the option of giving “don’t know” answers to questions, past research shows that survey respondents often express opinions about issues they know nothing about to avoid seeming ignorant. Thus, it seems likely that many respondents who did not know the answer to various questions attempted to guess, especially on those items that had only two or three possible answers.

Second, three of the five items with the highest percentage of correct answers are personal information about candidates in the 2000 election that has little or no value for understanding politics more generally. Those three items are the home states of George W. Bush and Al Gore (90 percent and 68 percent correct answers, respectively) and Joe Lieberman’s religion (70 percent). Bill Clinton’s ideology, the second‐​highest scoring item (81 percent correct answers), is an artifact of generous coding on my part, under which both “liberal” and “moderate” answers were deemed correct. Eliminating the three high scoring low‐​value items and two other similar questions, which produced much lower percentages of correct answers, produces an average score of 11.5 correct answers to 26 questions, for a 45 percent average, which is a slightly lower percentage than that observed on the 31‐​point scale. Much more significant, the elimination of the five low‐​value questions (while retaining the Clinton ideology question) increases the proportion of “know‐​nothings” to about 34 percent, a percentage considerably higher than Bennett’s estimate. Table 3 summarizes the aggregate results of three knowledge scales from the 2000 NES.

As Table 3 shows, the already low average knowledge scores on the 2000 NES conceal the existence of a large political knowledge underclass of “know‐​nothings” who possess very little if any basic political knowledge. Depending on which scale is used, this group constitutes from 25 percent to 35 percent of the American public. Overall, considering (1) the very basic nature of the questions asked, (2) the possibility of guessing, and (3) the high percentage of “know‐​nothing” respondents, it is difficult to avoid the conclusion that the 2000 NES, like most research using earlier evidence, reveals a low level of political knowledge.

If voters have generally low knowledge levels, they may be able to make up for it by using information “shortcuts.” Until recently, this was the clearly dominant view among political scientists. This part critically assesses the most important of the various shortcuts to informed voting proposed in the literature on the subject: information from daily life, political parties, cues from opinion leaders, retrospective voting, issue publics, and the so‐​called miracle of aggregation.

Some scholars have taken up Anthony Downs’s suggestion that rational voters will make use of information acquired through ordinary daily‐​life interactions; such information is virtually “free” since the activities that produce it would, by definition, be undertaken even in the absence of any political purpose. Unlike Downs, more recent advocates of this shortcut argue not only that it will be used but that it goes a long way toward meeting voters’ informational needs. For example, voters allegedly can obtain “a good deal of information” about the economy from personal financial transactions such as managing a checking account or seeking employment.61 Morris Fiorina goes so far as to suggest that “[i]n order to ascertain whether the incumbents have performed well or poorly citizens need only calculate the changes in their own welfare.”

Although it would be foolish to deny that some helpful information can be derived from ordinary life, its usefulness to otherwise ill‐​informed voters is greatly overestimated. Three major limitations of such information are particularly important. First, by definition, this approach is of no help in dealing with the many political issues that the vast majority of voters do not encounter in daily life. Second, even if the voter has, following Fiorina’s suggestion, carefully calculated the changes in his welfare and developed a judgment about an incumbent’s policies, he cannot readily determine whether his welfare will be improved by electing the opposing candidate. Even if things have gotten worse under President X, perhaps Challenger Y’s program is even more harmful. That possibility cannot be ruled out without substantive issue knowledge going beyond personal experience.

Most important of all, substantive knowledge is required to determine whether or not a particular personal experience really is the result of public policy and, if so, which political actors are responsible. Ill‐​informed voters attempting to make political judgments on the basis of personal experience may fall into egregious errors. Even with respect to unemployment and inflation, basic economic issues with which most people have substantial personal experience, ill‐​informed voters tend to make spectacular errors. In a survey taken during the 1992 election, during which economic issues were a particular focus of publicity, the vast majority of respondents could not estimate the inflation or unemployment rate within 5 percent of the actual level;63 the electorate’s mean estimates of both rates were approximately twice as high as the real level. Such misperception apparently played a major role in swinging the 1992 election against incumbent president George Herbert Walker Bush. Poorly informed voters are more likely than well‐​informed ones to make sweeping generalizations from personal experience with unemployment but less likely to make accurate connections between experience and policy.

If errors of this magnitude occur in the cases of inflation and unemployment, even more serious mistakes can be expected with other, more remote, issues. And even a correct estimate of unemployment and inflation is only a minimal prerequisite to determining which side’s policy on those issues will better serve the voter’s interests. One still needs to know to what extent incumbents are responsible for current rates and whether or not their opponents are likely to do better. In and of itself, information from daily life is unlikely to be of much help in making such decisions.

The idea that political parties can help voters economize on information costs has a long and venerable lineage, dating back to Democratic Party leader and later President Martin Van Buren, founder of the first modern mass‐​based party. The basic argument claims that voters can infer candidates’ policy stances from their partisan affiliations rather than undertaking the much more difficult task of inquiring into the views of each individual aspirant to office.

This claim is not entirely without merit. To this general argument, V. O. Key adds the notion of party identification as a “standing decision” to be reevaluated in the face of experience with the party’s officeholders, and Morris Fiorina presents evidence that voters rationally make use of past experience with the two parties in creating a “running tally” of their relative merits.

Appealing as it is, the argument obscures as much as it reveals. At best, a candidate’s party affiliation is a clue to his policy stances, but it tells the voter little about the likely effects of those policies. Party affiliation helps voters meet the first and third knowledge requirements of the Michigan criteria but provides little guidance on relating that information to the voter’s own goals. In principle, a running tally may help a voter to determine the merits as well as the content of a party’s policies. But it is difficult to do so without substantial underlying substantive knowledge. If conditions are good under the rule of Party X, how does the voter know that this is due to the success of the party’s policies rather than to factors beyond political control, preexisting favorable trends resulting from decisions made by the party’s predecessors in power, personal characteristics of the party’s officeholders that are not representative of the party as a whole and thus might be misleading as predictors of future behavior; or crafty manipulation of public policy by the party’s leadership as a result of which temporary success is achieved at the price of long‐​term harm, the effects of which are felt years or even decades after those leaders leave office? The voter cannot get around this dilemma simply by aggregating large amounts of experience, since it is unlikely that a given voter has been following politics long enough to experience more than two or three governments headed by any one party.

Most dramatic, existing research on parties has largely ignored an important way that the existence of political parties may actually reduce the flow of information to voters relative to a nonpartisan electoral system. If politicians are organized into (relatively) centralized parties, the number of effective political actors in the system is reduced; in most democratic political systems, there are rarely more than four or five major parties at any given time. As in any other competitive situation, the smaller the number of competitors, the greater the chance of successful collusion among them. In any situation in which the number of major parties is small, especially in a two‐​party system such as that in the United States, there is the possibility that the parties may conspire to take an issue of potential interest to the public off the political agenda when doing so benefits political elites at the expense of the rest of the citizenry. Such a cartel also diminishes the flow of information about the issue to the public, since it is no longer discussed by candidates running for office and the media are less likely to cover it.

Empirical examples of such collusion are not difficult to come by. The very first modern‐ style party system was established in part to take the issue of slavery off the political agenda in this way. That cartel lasted for about 25 years. Other cases from a variety of political systems could easily be cited. Interparty collusion can be broken by the entry of new parties into the system, just as the Republican Party eventually emerged to challenge collusion over slavery. But organizing a major new party is extremely costly, and the existing parties can defend their oligopoly position through their ability to manipulate the electoral and campaign finance systems. At the very least, such “policy cartels” can persist for a long time, even on so momentous an issue as slavery. This argument should not be interpreted to imply support for a nonpartisan political system; such systems have their own serious shortcomings. Nonetheless, it is striking that a generation of scholars heavily influenced by economic theories of competition should have maintained a largely uncritical enthusiasm for strong political parties without taking account of a serious objection derived from those same theories. Although real, the informational benefits of parties are almost surely exaggerated by conventional wisdom.

If the vast bulk of the electorate is ignorant, perhaps it can follow the lead of the knowledgeable minority of political activists, or opinion leaders. That line of argument is one of the most common in the literature on shortcuts. Instead of keeping close track of issues themselves, voters can respond to cues issued by political activists whose values are similar to their own. What is important is that there are perhaps 5 percent of voters who are activists and news junkies who do play close attention. If they see that something is seriously wrong in the country, they sound the alarm, and then ordinary people start paying attention.

Unfortunately, the strategy of following cues from opinion leaders creates at least as many difficulties for ignorant voters as it solves. Because of the immense asymmetry of information between leaders and followers and the low incentive of the latter to monitor the leaders’ performance effectively, serious principal‐​agent problems will surely arise. From the perspective of the principal, it is difficult to conceive of a more difficult principal‐ agent relationship than that between ignorant voters and highly knowledgeable, well‐​organized political activists.

Where voter interests and activist interests coincide closely, the difficulties of monitoring need not be so acute. But that state of affairs is far from common. Political activists differ greatly from the general population on a wide range of demographic and socioeconomic characteristics; they also tend to be more extreme in their views. Most important of all, opinion leaders acquire interests that diverge sharply from those of voters simply by virtue of becoming opinion leaders. As political activists, their power, prestige, social status, and opportunities for pecuniary gain will tend to rise with the public’s perception of their issue positions; they thus have strong incentives to exaggerate the importance of political problems and to push for political solutions (or at least solutions with a prominent role for activists) in preference to private‐ sector ones. Even when voters are aware of the incentives for exaggeration and attempt to discount activist claims as a result, they have no way of knowing how much discounting is required.

Even if there exists a subset of opinion leaders whose interests do coincide with those of a given voter, that voter still faces an extraordinarily difficult problem in determining who they are. Since the whole point of relying on opinion leaders is to economize on information costs, the voter is unlikely to invest heavily in researching the leaders’ qualifications. And, unlike in the case with most private‐​sector specialist professionals, the voter cannot simply judge the quality of activists’ performance by the results of the policies they advocate, since it is not usually possible to determine which social outcomes are the result of public policy without considerable substantive knowledge of the issues.

A successful strategy of following cues from opinion leaders requires voters to first decide which leaders’ cues to follow and then monitor those leaders in order to avoid a variety of principal‐​agent problems that are likely to arise. Neither of these is possible without considerable substantive voter knowledge of the issues. Without such knowledge, opinion leaders are as likely to be misleading as they are to be informative.

The retrospective‐​voting hypothesis holds that voters judge politicians by past performance rather than current promises. The argument is advanced that “retrospective voting requires far less of the voter than prospective voting.” But does it?

There are at least three reasons to doubt that the answer is yes. First, as noted above, it is often difficult for ignorant voters to determine which social outcomes are the result of public policy and which aren’t. To take a prominent example in the literature, many models of electoral retrospection are based on “sociotropic” voting, in which voters make their decisions on the basis of the condition of the national economy rather than that of their own personal finances, Yet a person ignorant of economics (sometimes even a trained economist) cannot tell whether economic conditions are the result of (1) the policy of the current government, (2) lagged effects of its predecessors’ policies, or (3) factors completely independent of any government action.

Even if option 1 is the case, the voter may not be able to determine whether current conditions are positive or negative—if, for example, temporary economic sacrifice might be a necessary precondition for future progress. Even where the voter does know that a given outcome is the result of government policy, ignorance of the structure of government may make it difficult for him to decide which elected officials deserve credit or blame. Under an American divided government or a European coalition government, the voter may not even be able to tell which party is responsible. That defect is of particular importance for retrospective‐​voting theory, which emphasizes that the electorate “passes judgment on leaders, not policies” and thus implicitly assumes that voters know which leaders are responsible for what.

Finally, even when the voter knows both that a given outcome is the result of government policy and which leaders are responsible, that is still not quite enough to make an informed choice. The voter will also want to know whether the opposition party is likely to do better. In order to make retrospective voting “work,” the voter must first establish how good past performance has been, and it is not possible to do so without addressing the types of problems discussed here.

The retrospective‐​voting argument does, however, possess a kernel of truth. As Fiorina puts it, retrospective voting can impose a kind of “rough justice” on political leaders who have failed badly. If a policy failure is large, highly visible, and easily attributable to a particular set of leaders, it is certainly likely that they will be voted out of office, as the elections of 1932, 1952, 1968, and 1980 suggest. Moreover, the bigger the failure, the less likely it is that the opposing party’s performance will be worse. The ability of voters to punish large and obvious policy failures by incumbents is one of the major advantages of democracy over dictatorship.

Unfortunately, the preconditions of magnitude, visibility, and easily traceable accountability rarely obtain in real life. Even in the case of a very large policy failure, leaders may escape blame if the full impact of the failure is not felt until after they are out of office. Herbert Hoover and Jimmy Carter paid the price of perceived failure in 1932 and 1980, respectively, but their predecessors (Coolidge, Nixon, and Ford), who arguably had at least as much to do with of the failures in question, did not.

If voters cannot keep track of all the important issues, perhaps they can at least focus on a few that are of particular concern to them. For instance, blacks are more likely than whites to be familiar with civil rights issues. In theory, such “issue publics” can make up for ignorance of more general policy issues within the electorate as a whole.

Attempts to confirm the issue‐​public hypothesis empirically show that it has only very limited validity. Knowledge of different aspects of public policy is highly inter‐​correlated. Even where significant differences in knowledge between groups do exist, they do not necessarily demonstrate that the knowledge of the better‐​informed group is adequate for informed voting; they show only that members of that group know more about an issue than does the rest of the electorate. The difference is crucial, because most studies showing that issue publics are better informed about a particular issue than the rest of the electorate rely on surveys tapping only very basic knowledge.

Even if the voter does have adequate knowledge of the narrow issue of particular concern to him, informed voting with respect to that issue might still be inhibited by ignorance of the “rules of the game” of government policy. A black voter may have sufficient specific knowledge to conclude that current civil rights policy should be changed but not enough general knowledge of the structure of government to determine which elected officials have to be voted out to do it. Even in the land of the blind, the one‐​eyed man cannot be a true king if kingship requires seeing things that can only be discerned with two eyes.

Voter ignorance also undercuts the utility of issue publics in two further, less obvious ways. First, the rationally ignorant voter cannot readily tell which aspects of public policy really are part of the issue of interest. One of the problems in issue‐​public research is the question of how the scope of the relevant “issue” is defined in the first place. If the connection between two or more matters of public policy is not obvious or is ignored by politicians and the media for their own reasons, voters may fail to pick it up. Social Security reform, for instance, is almost never defined as a racial issue, yet the lower life expectancy of blacks combined with the fact that they pay Social Security payroll taxes at the same rate as whites turned Social Security into a major hidden redistribution from black workers to white retirees. The subtlety of the connection, combined with collusive politicians’ lack of incentive to focus on the issue, leads the relevant black issue public to ignore it. Such problems often prevent an issue public from forming in the first place. High knowledge costs combine with collective action problems to ensure that many potential issue publics are numbered among Mancur Olson’s “‘forgotten groups’ who suffer in silence.”

Most fundamentally, voter ignorance of general issues may vitiate the benefits of issue publics even in situations in which the issue publics have fully adequate information about their more specific concerns. If each specific issue area is controlled by a subset of the electorate with a special interest, while these same subsets remain ignorant of generally applicable issues, the outcome may well be a process of mutually destructive rent seeking that leaves each group worse off than it would have been had there been no issue publics to begin with. Within its particular bailiwick, each issue public pushes for policies beneficial to itself without regard to the costs to others—costs of which its members are ignorant even if self‐​interest would not lead them to ignore these costs in any case. A classic “tragedy of the commons” ensues in which the general interest is routinely neglected in favor of the particular. For these reasons, it is by no means clear that an electorate divided into issue publics is in a better position to pursue its policy objectives than one that is uniformly ignorant across the board.

If the rationally ignorant portion of the electorate commits its errors randomly, the power of aggregation might result in those errors canceling each other out. Where mistakes are truly random and the electorate sufficiently large, every “erroneous” vote for Candidate X should be offset by one for opposing Candidate Y. Only the non‐​randomly distributed votes of the relatively informed voters will have a real impact on the outcome; that outcome will thereby be decided “as if” the electorate as a whole were informed.

It is ironic that this line of argument should be put forward by writers committed to developing a defense of “majoritarian democracy” against charges of voter incapacity. Taken seriously, it implies that the votes of the vast majority of the electorate are just “noise” obscuring the “signals” sent by the informed few, as one advocate explicitly states. If the argument were correct, elections would have the same outcome if only the ballots of the well‐​informed minority were counted! Even in the more moderate version of the theory, which allows that some of the ill‐​informed votes nonetheless turn out to be “correctly” cast, a large proportion of the electorate is nonetheless viewed as a source of random errors that fortunately offset each other. Regardless, the “miracle of aggregation” can take place only if (1) the errors really are random and (2) the informed minority that decides electoral outcomes adequately represents the interests of the rest of the population. Overwhelming evidence suggests that neither precondition holds true.

One of the main reasons why errors are non‐​randomly distributed is that voters really do try to use several of the other information shortcuts discussed above. As a result, ill‐​informed voters often draw misleading inferences about economic conditions and other issues. Uncertainty about a candidate’s policy stances itself creates a systematic bias in favor of incumbents whose positions are generally better known.

The random distribution hypothesis fares little better in meeting the second precondition, that of representativeness of the informed. The small minority of well‐​informed voters (no more than one‐​fifth of the total and perhaps a lot less) differs systematically from the rest in gender, income, race, age, religion, ideology, and a host of other politically relevant attributes. It would be remarkable indeed if the interests of this small, unrepresentative subset of the population coincided even roughly with those of the population at large, and there is little reason to believe that they do.

Overall, the shortcuts to informed voting discussed in the literature are far less helpful to voters than their advocates suggest. In many instances, they may be actively misleading. There is no real substitute for voters adequately informed at the individual level.

Perhaps the most fundamental cause of ignorance resides in the collective action problem created by the insignificance of any individual vote in determining an electoral outcome. Since one vote is almost certain not to be decisive, even a voter who cares greatly about the outcome has almost no incentive to invest heavily in acquiring sufficient knowledge to make an informed choice. An informed electorate is a “public good” the provision of which is subject to the “collective action” problem that arises when consumers of a good do not have to help pay for its provision in order to enjoy its benefits. Only political professionals and those who value political knowledge for its own sake have an incentive to acquire significant amounts of it. Acquiring significant amounts of political knowledge for the purpose of becoming a more informed voter is, in most situations, simply irrational.

An important extension of this logic is that it applies just as readily to highly altruistic and civic‐​minded citizens as to narrowly self‐​interested ones. Even a 100 percent altruistic person— someone who always chooses to prioritize the welfare of others over her own whenever the two conflict—would not rationally devote much of her time to acquiring political information for the sake of casting an informed vote. No matter how great the benefits to others of a “correct” electoral outcome, our altruist’s ballot has almost no chance of bringing it about since in a large electorate the chance that his vote will be decisive is vanishingly small. The rational altruist would therefore seek to serve others in ways in which a marginal individual contribution has a real chance of making a difference, such as contributing to charity. By spending time and effort on becoming an educated voter, the altruist actually diminishes others’ welfare by depriving them of the services he or she might have conferred on them through alternative uses of the same resources.

The applicability of collective action argument to altruistic voters obviates, at least in this case, one of the standard criticisms of economic models of politics: that they rest on unwarranted assumptions of self‐​interested behavior. The prediction of rational voter ignorance rests on no such assumption. The conclusion that even altruists have little incentive to become informed voters also casts a pall on the proposals of political theorists to improve the functioning of democracy through increasing civic‐​mindedness. Whatever virtues those proposals might have, they seem unlikely to dissipate the rational ignorance that stands as a particularly imposing obstacle to effective democratic control of government.

The collective action problem explanation of political ignorance has other important implications for efforts to realize the democratic ideal of voter control. Proposals to alleviate ignorance by increasing the availability of political information to the public run afoul of the collective action problems. Moreover, most citizens do not seek out information now because they find politics relatively uninteresting.

A notable shortcoming of the collective action explanation of voter ignorance is that it seems inconsistent with the fact that voters vote at all. As has often been pointed out, rational choice models of politics appear to predict that all or most voters shouldn’t even show up at the polls, given the infinitesimal likelihood of affecting the electoral outcome. More precisely, they predict nonvoting unless the “duty‐​based” or “expressive” utility of voting outweighs its costs irrespective of the likelihood of affecting the outcome; if that is the case, however, it is possible that the same sense of duty that leads voters to vote may also lead them to become informed.

A complete discussion of this issue, the “paradox of voting,” would take us far afield, but there are at least three important reasons to believe that the paradox does not invalidate a collective action problem explanation of voter ignorance. First, it is possible that the unexpectedly high incidence of voting is simply the result of people overestimating the potential impact of their vote. Polls show that more than 70 percent of voters believe that their individual votes “really matter.” Such overestimation may in fact be rational to the extent that acquiring an accurate knowledge of the impact of voting may for many be more expensive than the relatively minimal effort required to vote in major elections. If so, it is not implausible to hypothesize that the degree of overestimation is great enough to stimulate voting but far too small to stimulate the much greater investment of time and effort necessary to acquire a substantial amount of political information.

Second, even if—as is surely true at least in part—the critics of rational choice are correct and voting really is explainable by the “expressive utility” of voting or by irrational conceptions of duty, it is still possible (and empirical evidence suggests very likely) that such motives are not powerful enough to induce voters to pay the heavy costs of becoming well informed. This point dovetails with our third argument: that none of the competitors of the rational ignorance theory predict the stability of extreme levels of ignorance in the face of rising education levels. Surely explanations of voter ignorance based on differing levels of commitment to duty and cultural variation and varying access to information would predict greater variation in levels of ignorance across time and place than seems to actually exist.

In sum, the rational ignorance hypothesis is only an imperfect representation of reality. Yet to the extent that it has validity, it sets severe limits to the amount of knowledge ordinary voters are likely to be willing to acquire. Any solution to the problem of voter ignorance will have to work within these constraints rather than try to break them.

The debate over voter ignorance has often focused on how much voters know but rarely on the question of how much government there is for them to know about. Yet it is clear that the greater the size and scope of government, the more voters have to know to control its policies through the ballot. To avoid misunderstandings, it is important to emphasize that the increased democratic control that may be achieved by reducing the size and complexity of government is not the only or even the most important factor that should be considered in determining the proper role of the state in our society. Other considerations may well outweigh it in particular circumstances. However, the knowledge‐ based tradeoff between big government and democratic government is an important consideration that has largely been neglected in the literature to this point.

In most advanced industrial democracies, government spending now consumes at least a third and often more than one‐​half of GDP. But it is not the size of government per se that so greatly increases the likelihood of voter ignorance as the extraordinary scope of government activity. A government that commits enormous resources to a narrow range of readily comprehensible activities is not necessarily much more difficult for voters to keep track of than one that commits only small amounts to them. Yet the growth of government over the last century has been characterized by an immense expansion of the domain of government power as well as by increased activity in areas of traditional state responsibility. That process has reached the point where areas of social life that remain outside the government’s domain are arguably no longer the rule but the exception.

In the United States the executive branch of the federal government alone has 15 cabinet‐ level departments, 54 independent regulatory agencies and government corporations, and 5 “quasi‐​official” agencies. They range in function from the United States Information Agency to the Farm Credit Administration to the National Mediation Board. It is doubtful in the extreme that voters could keep adequate track of all their activities even if they paid far more attention to political information than they do today. Current surveys of political knowledge usually do not even ask about the functions of specific government agencies, instead opting for questions about very basic aspects of government structure and opinions about whether government should do “more” in broadly defined issue areas such as “education” or “helping the poor.” The omission is in part the result of researchers’ reluctance to use questions that may prove “intimidating” to ill‐​informed voters who are ignorant of even basic information. The few questions requiring more detailed issue knowledge that have found their way into surveys unsurprisingly show even greater levels of ignorance than those about more basic information.

If traditional research on voter ignorance questions the electoral competence of ordinary citizens, focusing on the relationship of ignorance to the size and scope of government leads us to question that of relatively well‐​informed elites as well. It is unlikely that even professional social scientists have more than a very superficial knowledge of the activities of government agencies outside their areas of specialized expertise, if indeed they have heard of them at all.

Researchers tend to assume, usually without argument, that the best‐​informed voters can make effective use of ideology and opinion leaders to guide their choices. Yet it is not at all clear that this optimism about elites is justified. Let us consider the dilemmas that the modern state creates for generally well‐​informed voters who nonetheless remain ignorant of specific issues.

The ideologically sophisticated voter who is convinced that a government program can address some evil must still determine whether the remedies proposed by candidates really are the kinds of programs that are likely to solve the problem at hand. Unless voters hold to an ideological position that justifies government activism for its own sake regardless of the results, they must find a way to ensure that the programs put in place by legislators really do serve the purposes they intend. The difficulty is further exacerbated in those very common situations in which voters cannot readily observe a policy’s effects.

Even the people most favorably disposed toward government activism cannot deny the possibility of government programs that fail in their stated purposes because of unanticipated consequences or because they are actually driven by rent‐​seeking interest groups or attempts to redistribute funds to the rich or some other group whom voters might not want to reward. Many programs justified rhetorically by the need to aid the poor or provide some public good for the population at large actually serve well‐​positioned “distributional coalitions” at the expense of the very people whom the policies are supposedly intended to help. And it is a commonplace among specialists that well‐​intentioned interventions such as minimum wages or drug prohibition often produce side effects that may be worse than the original problem.

The paucity of research on the control of modern‐​sized governments by well‐​informed, ideologically sophisticated voters prevents any definitive conclusions in this area. In part, the lack of research of this type is itself a consequence of the growth of government, since that growth has greatly increased the proportion of the electorate that is not well informed and has diminished the electoral significance of those few who are. Nonetheless, there are strong reasons to believe that even a well informed electorate, one with an average level of knowledge equal to that of the top 5 to 10 percent today, would have great difficulty imposing its will on a government as broad in scope as that which we have today.

A government of strictly limited powers might reduce the problem of public ignorance by reducing the number of issues to be decided by government to a level that voters would find more manageable. Obviously, issues taken out of the government sphere would have to be “decided” in other ways, whether through market exchanges or other nonpolitical processes. Whatever their other defects, however, those mechanisms are relatively free of the collective action problems that induce massive voter ignorance in the political sphere.

Is there in fact any evidence that limited government eases the informational burden on voters and enables them to exercise greater control than they do today? A definitive answer to this question requires much more rigorous analysis than is possible here. Nonetheless, 19th‐​century American history suggests that the answer may be yes.

Because of the very limited powers of the national government, 19th‐​century national politics revolved around a small set of relatively narrowly defined issues, including the spread of slavery, the disposition of newly acquired western lands, the tariff, federal support for infrastructure spending, banking, and, on a few occasions, warfare with foreign powers. With the exception of obvious overlaps between issues (e.g., slavery and western lands), rarely were more than two or three of these matters in contention at any given time.

The theory developed here predicts that this limitation of government power should have allowed voters to focus on the issues that did come onto the electoral agenda in much greater detail than has been possible since. Although there have as yet been no systematic tests of this prediction, considerable evidence supports it. During major 19th‐​century controversies such as the debates over monetary policy in the 1830s and 1890s, the ongoing battle over the tariff, and the confrontation over the expansion of slavery in the 1850s, politicians presented for mass consumption far more sophisticated arguments than prevail in electoral politics today.

In the case of slavery, the Lincoln‐​Douglas debates over slavery expansion, conducted before large audiences of ordinary voters, including a substantial proportion of illiterates, addressed in some detail such questions as the effect of slavery expansion on free labor, whether or not the Supreme Court’s interpretation of the Constitution necessarily takes precedence over that of other branches of government, the moral status of blacks in the liberal ideology of the Declaration of Independence, and the true meaning of “popular sovereignty.” Campaign speeches of this complexity would be unimaginable today.

Popular ideological awareness also played an important role in restraining the growth of government beyond its prescribed bounds. Indeed, this factor was singled out as the most important force limiting the power of the U.S. federal government by leading contemporary foreign observers such as Tocqueville and Lord Bryce. More recent theoretical treatments of the subject also emphasize the crucial role of popular ideological consensus in restraining the growth of 19th‐​century American government in the face of interest group pressure. Such a high degree of ideological awareness is in clear contrast to what we know of modern voters. It suggests that ideological coherence in voting may be more feasible if there are fewer issues to keep track of.

This 19th‐​century evidence by no means amounts to a strong test of my hypothesis and is of course subject to alternative explanations. Even so, it provides a prima facie case for the proposition that voter knowledge and control of government policy will be much greater under a regime of strictly limited government power.

The relationship between voter ignorance and big government leads us to question the adequacy of the information possessed by even the best‐​informed voters. It also leads to the counterintuitive suggestion that the extension of government power to new areas of social life undercuts democratization rather than furthers it. Democratic control of government is increased when there is less government to control.

If rational ignorance suggests that government may be more democratic when it is more strictly limited, it may also counsel in favor of greater decentralization. When information problems are taken into account, voting with your feet in a relatively decentralized federal system may lead to greater majoritarian control of government than ballot‐​box voting in a more centralized state.

As we have seen, one of the main causes of political ignorance is the fact that it is “rational.” Because even an extremely well‐​informed voter has virtually no chance of actually influencing electoral outcomes, he or she has little incentive to become informed in the first place, at least if the only purpose of doing so is to cast a “correct” vote. By contrast, a person “voting with her feet” by choosing the state or locality in which to live is in a wholly different situation than is the ballot‐​box voter. If a “foot voter” can acquire information about superior economic conditions, public policies, and other advantages in another state, he or she can move to that state and take advantage of them even if all other citizens do nothing.

That creates a much stronger incentive for foot voters to acquire relevant information about conditions in different jurisdictions than for ballot‐​box voters to acquire information about public policy. Since states and localities seek to attract new residents and businesses as sources of tax revenue, state and local governments have strong incentives to establish policies that will appeal to potential immigrants and convince current residents to stay. The power of the competitive pressure comes from governments’ constant need to attract additional revenue to finance expenditures that can satisfy key interest groups and increase political leaders’ reelection chances. Interstate and inter‐​locality competition for residents facilitates the creation of public policies that advance the interests of the majority, even in the absence of informed ballot‐​box voting. In this way, voting with your feet becomes a powerful and, in many ways, superior alternative to ballot‐​box voting as a mechanism of majority rule.

Moreover, unlike a ballot‐​box voter, a foot voter need not connect his judgment of conditions to specific elected officials and their policies. It is enough for her to know that conditions are better in one state than another and then be able to act on that by moving. So long as public officials themselves know that their policies can affect social conditions in ways that attract foot voters, they will have an incentive to implement better policies in order to appeal to potential migrants. Not only does foot voting create a stronger incentive to acquire knowledge than ballot‐​box voting, it also requires less knowledge to implement effectively.

To my knowledge, there has not yet been a study that precisely measures the informational advantages of voting with your feet over ballot‐​box voting. Nevertheless, there is reason to believe that the advantages are quite large. Anecdotal evidence and ordinary life experience suggest that most citizens put far more effort into deciding where to live than into acquiring political information.

Perhaps more telling is historical evidence of the power of voting with your feet even under extremely adverse circumstances. In the Jim Crow–era South of the late 19th and early 20th centuries, African‐​American southerners, most of them poorly educated and many illiterate, were still able to learn enough information about the existence of relatively better conditions in other states to set off a massive migration. Southern black workers relied on information provided by relatives in other jurisdictions and by agents of businesses seeking to recruit African‐​American workers. The resulting migration not only benefited the migrants themselves but also forced racist southern state governments to “grant … African‐​Americans greater educational opportunities and greater protection in their property and person” in an effort to get them to stay and continue to provide labor for southern white‐​owned farms and businesses. In a related dramatic example, inter‐​jurisidictional competition for the labor of migrating black coal miners led to successful lobbying by coal companies for a reduction in school segregation in West Virginia in the early 1900s.

Obviously, the ability of southern blacks to vote with their feet did not come close to fully mitigating the baneful effects of Jim Crow. It did, however, provide important informational benefits and political empowerment to a widely despised and poorly educated minority. Although exact comparisons are difficult, it seems likely that potential southern black migrants of the Jim Crow era were able to learn considerably more about relative conditions in different jurisdictions than most modern voters have learned about the basics of our political system. At the very least, large numbers of poor and ill‐​educated southern blacks learned enough to understand that relatively more favorable employment opportunities and public policies awaited them in other jurisdictions, a realization that contrasts with the inability of most modern citizens to acquire sufficient knowledge to engage in effective retrospective voting. Obviously, if voting with your feet could provide powerful informational advantages in the exceptionally adverse conditions of the Jim Crow‐​era South, there is strong reason to expect that it is more effective in modern times, when education levels are much higher, information costs are lower, and no large group is as thoroughly oppressed as were poor southern blacks a century ago.

Democracy and political ignorance are far from the only issues that must be considered in determining the degree of decentralization that a society should have. A variety of other considerations may in some situations outweigh the advantages of foot voting. The argument advanced here is not intended to be a comprehensive theory of federalism, or even close to it. It does, however, highlight an important consideration that is too often ignored.

I have argued that voter ignorance is widespread, that it is a rational result of collective action problems, that it cannot be circumvented through various “information shortcuts,” and that it is greatly exacerbated by the size and scope of modern government. I have also argued that voter ignorance highlights some unrecognized advantages of decentralized federalism.

Concern for the consequences of voter ignorance is not a new idea in Western political thought. Plato, Aristotle, and John Stuart Mill, among others, all regarded it as one of the most important problems of democracy. Modern democratic thought, however, has tended to sweep the problem under the rug even as the growth of government has made it increasingly severe. Yet ultimately, liberal democrats cannot avoid the inherent tension between big government and democracy.

To say that the tension must be faced is not to say that it must necessarily be resolved in favor of democracy. Democracy is not the only criterion for good government, and perhaps not even the most important one. Nonetheless, the dilemma is particularly acute for those who see democratic self‐​government as an important value in its own right. Even for those who see democracy as a purely instrumental value—for instance, for achieving social justice—the danger of elite dominance over most policy issues in a democracy with largely unconstrained government power should raise troubling questions. Are our elites really so benevolent and so knowledgeable that they will serve the interests of the poor and downtrodden in the absence of effective voter control?

The concern here is not the traditional fear that elites can deviously manipulate public opinion to suit their own advantage, although that possibility cannot be ruled out. It is that elites end up in control of policy simply by default. What the voters don’t know about, they can’t meaningfully control.

A possible response to the argument of this paper is simply to scale down the ambitions of democracy from effective voter control of public policy to a mere ability of voters to change rulers when they so choose. That is indeed an important advantage of democracy in situations in which incumbent officials have committed a massive and highly visible policy error. But while this claim is a point in favor of unrestricted majoritarianism as opposed to dictatorship or oligarchy, it says nothing about the merits of such a system compared with a democratic government with strictly limited powers. Moreover, the ability to remove a failed leader is of little use in that vast majority of cases in which the electorate cannot readily tell whether or not a given leader has failed.

The argument presented in this paper does not give any definitive answer about how limited the powers of government should be, even from the narrow standpoint of ensuring meaningful voter control over public policy. Nonetheless, the depth and apparent intractability of current levels of ignorance strongly suggest that democratic control requires a government more limited in scope than what we have today.

A second limitation of my analysis is the lack of comparison with knowledge problems encountered by consumers in private‐​sector markets. If the latter were as great as those that exist in the case of government, the argument for limiting government power would be weakened. A full analysis of this subject is beyond the scope of this paper, yet there are powerful theoretical reasons to believe that “market ignorance” is both a less prevalent and a less serious problem than voter ignorance. Among the more important of those reasons are the absence of collective action–induced rational ignorance in most product markets, the relative ease of connecting product quality to product performance compared with the great difficulty of linking social outcomes to public policy, the greater ease of finding disinterested advisers on market transactions (e.g., financial advisers, consumer guides), and the existence of enforceable contract and tort remedies against deception by product manufacturers of a sort that is unavailable against deceptive politicians.

The political realm also lacks a conveyor of information as efficient in providing feedback and diminishing information costs as the market’s price system. Yet even if the situation is closer to parity than I suggest, the monopolistic nature of the state and the great powers concentrated in its hands should still lead us to place a particularly high value on democratic control over government. At a bare minimum, the problem of voter ignorance and its connection to activist government should be taken more seriously in political thought than it has been so far.

References that appear in the original policy analysis have been removed.