EC742 Part 4: How Efficient is Majority Rule? iii. However, for nearly every model that predicts government failure, there are others that predict either complete success, or at least tolerable levels of success. I. The Logic of Government Failure A. One of the main contributions of the Public Choice literature is the concept of government failure. C. A number of public choice models that predict relatively or perfectly efficient government policies. i. That is to say, the observation that political processes do not necessarily adopt Pareto Efficient or Social Welfare maximizing policies. ii. A very broad swath of the Public Economics and Macroeconomics literatures suggest that public policies can be improved in various ways. i. These include benevolent public interest theories of the state, which rely upon altruism and idealism by public officials, or which simply assume benevolent central planners or dictators. ii. However, government success models also include sophisticated self-interested election and interest group models of government decisionmaking. a. Taxes and subsidies have avoidable excess burdens. b. Macro-economic policies may be pro-cyclical or unpredictable rather than counter-cyclic or predictable. c. Regulations may promote monopoly power rather than competition. d. Environmental policies may be overly burdensome. e. Significant externality problems may go unaddressed. iii. Government failure helps to explain why policy problems exist. D. Lecture VI surveys some of the more common theories of government failure and government success, giving special attention to the electoral literature. I. Rational Ignorance, Fiscal Illusion, and Interest Groups A. An implication of the median voter theorem(s) is that the median voter gets what she/he wants. However, the median voter's ability to pick the policy that is most in her (or his) interest is limited by the information, theories, and time that she (he) has available for analyzing the alternatives. B. Analyzing the relative merits of alternative public policies is just like any other activity--it is engaged in only up to the point that maximizes expected net benefits. a. Externalities may exist because governments fail to act b. Excess tax burdens may exist because government chooses to adopt tax policies that have larger burdens in total than necessary given the feasible array of tax instruments c. Regulations may be set at excessively demanding levels, inefficiently low levels, or created in a manner that impose greater cost on those within the country of interest than necessary (command and control rather than Pigovian taxes and/or effluent markets ). i. In most cases this occurs at the point where the expected marginal benefits of more information and more analysis equals its expected marginal cost. ii. An implication of this (stressed by Downs and Tullock) is that voters will rationally remain ignorant of much useful information. iii. They will use smaller than possible samples of data and ignore types and dimensions of information that are relatively costly to acquire and/or to analyze. iv. [Draw a diagram (and/or write down a few equations) that illustrates the collection of information by an expected net benefit maximizing individual.] B. There are a number of public choice models that predict government failure. i. In at least some cases, even well functioning democratic governments with median voter outcomes may fail to “properly” address fiscal and regulatory problems. ii. In other cases, government failures may result from imperfections in democratic procedures: fiscal illusion, rational ignorance, monopoly power, and the effects of interest groups. 1 EC742 Part 4: How Efficient is Majority Rule? ii. In cases in which the median voter's expected marginal benefit from a public policy is less than the actual benefit or her (or his) expected marginal cost is higher than the actual cost, the result will be an UNDER demand for public services, relative to that which actually maximizes net benefits for the median voter. iii. [The latter case was argued by Anthony Downs and the former by Gordon Tullock. Both are clearly possible and each may be associated with different kinds of services and tax systems.] iv. It bears noting that both governments and interest groups may attempt to induce biased expectations by "subsidizing" (freely providing) information about the benefits of programs and/or "taxing" (withholding) information about the costs of programs. v. [The figure below illustrates the biased choices, as opposed to unsystematic mistakes, that fiscal illusion can gnerage. In the case illustrated G’ > G*, but the reverse is also possible.] C. A bit of rational ignorance is not a problem for democracy as long as it does not induce "biased expectations" about the benefits and/or costs of public policy. i. For example, if the sampling of information done by voters is reasonably complete they will tend to have unbiased estimates--although not perfect ones--and the result will on average advance the interests of the median voter. ii. Indeed, as long as voter expectations are unbiased, the Condorcet jury theorem implies that the outcomes of majority rule can also "aggregate" the information in the minds of voters (by using the median of their estimates when assessing candidates or policies). iii. However, if the information included in the sample is not complete, voter expectations may be biased, and the results of elections will not necessarily advance even the interest of the median voter. iv. (See Congleton 2007 and 2001 for more on this.) E. Fiscal illusion causes problems within a median voter model, because if voters are unable to accurately assess their own marginal costs and benefits they may vote for the wrong candidate and favor the “wrong” policies. (See Congleton 2001, 2007.) D. When voters have biased expectations about the benefits and/or cost of public programs the are said to exhibit Fiscal Illusion. i. In cases in which the median voter's expected marginal benefit from a public policy is greater than the actual benefit or her (or his) expected marginal cost is lower than the actual cost, the result will be an OVER demand for public services, relative to that which actually maximizes net benefits for the median voter. (Illustrated below) i. If voters make systematic mistakes, the policies chosen through elections may not advance even median voter interests. a. Too little of services with "hidden" benefits will be provided. b. And, too much of services with "hidden costs" will be produced. F. Note that a good deal of what interests groups do is informational in nature (Congleton 1986, 1991, 2001). Fiscal Illusion MC (true) $/unit i. They sponsor research and testimony of researchers in Congress and before regulatory commissions. ii. They sponsor political advertising of various kinds. iii. To the extent that these informational strategies affect voter expectations about their costs or benefits, they may induce fiscal illusion. MC (estimated) MB (estimated) MB (true) G* G' G. Voter ignorance also allows candidates for office to trade favors for campaign resources. G 2 EC742 Part 4: How Efficient is Majority Rule? i. If voters knew everything, campaign resources would not matter. ii. However, voters typically know relatively little about candidate positions, and have to be informed about them in various ways. iii. Fiscal illusion weakens the constraints of electoral competition, allowing secret deals among elected officials, bureaucrats, and interest groups. I. Another Neglected Assumption i. The strong form of the median voter theorem, by neglecting information problems, also ignores possible "agency problems." a. Candidates may say one thing to get elected and do something else once in office. b. Moreover, elected representatives may not be able to fully control the bureaucracy. c. However, candidates that are known to have cheated and one poorly at overseeing the bureaucracy will be more likely to lose the next election than those that have not since the median voter will not have gotten what he or she wants. So this is not a crazy assumption. a. Voters typically will not know very much about specialized issues, like the tax treatment of insurance companies, the specific details of farm subsidies, the manner in which public services are produced (by whom, how and where). b. On the other hand, interest groups may care deeply about such issues and be willing to pay for policies that advance their interests--either with campaign contributions or with other sorts of support during elections. ii. It also assumes that the real issue space can be mapped into a single dimension and that voters rank order candidates according to their policy positions. iv. This allows interest groups to have direct effects on policy that is disproportional to the number of votes that they cast in an election. v. (It is also one reason why campaign reform law tends to be discussed and eventually promoted. Voters know something goes one, even if they do not know what, and they demand some method of control over these deals. Politicians respond by passing legislation.) a. There is evidence that supports this assessment, especially the very detailed analysis of voting in the US Congress by Poole and Rosenthal (1991,1996, 2001). b. However, as the number of relevant dimensions of public policy increase, the probability that a median voter exists shrinks. Very strong symmetry assumptions are required for the distribution of voter preferences in multidimensional issue spaces (Plott, 1967). H. It bears noting that a good deal of campaigning and lobbying is ideological rather than economic in nature. Evidently, one cannot gather majority support simply by arguing that “you” will be better off under “my” policy than others currently being considered. iii. Informational assumptions about candidates and voters, turnout, and electoral institutions also tend to affect the character of an electoral equilibrium. iv. None the less, the median voter model is widely used and widely found to be a useful first approximation of public policy formation in democracies. i. There are, partly for this reason, also “non rational” and “non instrumental” theories of voting. ii. For example, the importance of "expressive voting" in voter behavior is not well understood at this point, but is attracting considerable attention (see for example Brennan and Hamlin 1998, 2000) or Caplan (2001, 2002). iii. Morality, Altruism, and Ideological dimensions of voter choices are also attracting new attention from economists and rational-choice based political science. II. Is there a Median Voter? A. As one moves from one dimensional choices to multidimensional ones, the probability of a median voter declines. i. A multidimensional median requires complete symmetry around a single point (the median). In the absence of such a point, no median exists and so the median voter theorems do not apply. 3 EC742 Part 4: How Efficient is Majority Rule? ii. However, if there is a mapping from an N-dimensional issue space into a single dimensional issue space (as with a left right ideological spectrum), the median voter models can be applied. iii. This is one explanation for the models reasonably good success in explaining properties of US elections. ii. Thus, all policies that distribute tax burdens or the benefits of public services are inherently plagued by lack of majoritarian equilibria when voters are narrowly self interested and no other institutional constraints exist. iii. (Civic norms, civil law, and takings law may help solve this problem as suggested below.) B. Illustration of a 2-dimensional issue space with three voters (approximately spatial voters). D. There is a very large literature on democratic cycles. i. There are several lines of research within this area: G2 a. attempts to resolve the Arrow Impossibility Theorem (Arrow 1951, Bernholz 1973) b. the rationality of majority decision making (Salant and Goodstein 1990) c. implication of agenda control chaotic (McKelvey 1975, Saari 2001) d. studies of coalitional stability (cooperative game theory) (Xue (1998), Kaminski 2001) X AL Y Bob Cat ii. There is also a literature that investigates possible solutions to cycles, instability, and intransitivity. Z a. This literature includes institutional solutions (Shpesle and Weingast 1981) b. It also includes pieces on the extent to which cycles may make democracy work better or worse (Buchanan 1954, Buchanan and Congleton 1998) G1 E. There are also several other literatures on cycles that may be associated with elections. i. Note that Y is majority preferred to X (Cat and Al vs Bob) ii. Z is majority preferred to Y (Cat and Bob versus Al) iii. But X is majority preferred to Z (Al and Bob vs Cat) iv. Majority rule in two dimensions is not always transitive, as it is in a single dimension (with spatial voters). i. The timing of elections varies among organizations and governments. Judges are elected for life terms, Senators for 6 year terms, presidents for 4 year terms, and Congressmen for 2 year terms. ii. In cases in which elections are held on a regular basis, there are election cycles that affect political and non-political decisions (and outcomes). C. Similar problems are associated with “divide the pie” games. i. Every possible division can be majority dominated by another. Consider a pie of size three to be divided between Al, Bob, and Cathy (A, B, C). The following is one of an infinite number of cycles. v v v v a. This means that all the considerations that go into voter and candidate decisions repeat themselves during the electoral cycle. There are obvious electoral stages of choosing to run, choosing a platform, conducting the campaign etc. (1,1,1) (1.5, 1.5, 0) (.5, 2, .5 (1,1,1) 4 EC742 Part 4: How Efficient is Majority Rule? Office holders also vote on policies, many of which bring home the bacon and reveal their competence and true platform positions. indicator of the relative merits of the candidates running for office, voters will tend to favor incumbents that increase their income at higher than average rates. b. One can imagine electoral institutions in which elections take place within different periods. a. In this case, incumbents may have incentives to increase the money supply or increase the deficit a year or two before their election as a method of increasing their vote share. The early models assumed that voters were not very informed about incumbent policies and used their own circumstances, and perhaps that of their neighbors, as proxies for whether the president (or other incumbent) had done a good job while in office. The political business cycle literature was also of interest to macro economists insofar as they provide an explanation for the macroeconomic policy “failures” of many democratic governments. (That is to say, it explained why “good” economic advice was not often taken.) v For example elections could be once a week, once a month, once a year, once every two years, three years etc. v Elections could be for life (kings and judges). v Elections could also be performed at random (unanticipated) intervals, to reduce or eliminate their cyclic nature. c. In each case, the necessity of winning an election to hold office has consequences. v For example, candidates have to be determined, votes cast, and counted. v In most cases, campaigns for office take place in order to inform the voters (who every they are) of the qualifications and positions of those seeking office. iii. Several lines of research in political economy study the economic and political consequences of election cycles. b. Kramer’s (APSR, 1971) study of US House of Representative elections is one of the first careful studies of political business cycles, although there were several precursors to his work (which Kramer reviews in the first part of his article). He find statistically significant relationships between income growth and electoral support (positive), between unemployment and electoral support (negative), and between inflation rates and electoral support (negative). More or less similar results are found by Nordhaus (1975) and Hibbs (1977). v These do not study problems with democratic instabilities, but rather the manner in which elections per se may affect public policies and thereby the real economy. F. The first aspect of electoral cycles to capture the imagination of economists and political scientists was the possibility of using macroeconomic policies to influence electoral outcomes, e.g. the political business cycle. i. Under Keynesian or Monetarist assumptions it is possible for government to manipulate the macro economy by judiciously increasing expenditures or the money supply. v (This is not possible in most rational expectations models, although even in those models price inertia and informative aspects of electoral outcome my have effects.) v In general the political business cycle models reflect “mainstream” assumptions about voting behavior and macroeconomic theory at the time the papers were written. c. President Clinton's famous poster for members of his campaign: "its the economy stupid" was evidence that many politicians believe that “pocketbook” issues are very important in national elections. d. In general, however, the results from this literature were mixed, with many studies supporting the existence of political business cycles and many others finding little evidence of such cycles. (See for ii. Insofar as the policy effects generated real positive increases in voter income and voters use their personal economic circumstances as an 5 EC742 Part 4: How Efficient is Majority Rule? example N. Beck, AJPS 1987, re the US Federal Reserve’s monetary policies.) See Paldam 1997 for a nice survey of the political business cycle literature. See also Alesina, Roubini, and Cohen (1999) for a book length treatment of political business cycles, which includes a fairly complete survey of the literature. Brief surveys of the literature appear at the beginnings of most pieces on political business cycles. [The political business cycle literature goes from voters to office holders to public policies to economic effects (which reveal information about the candidate’s policies) to a new round of voter choice, and so forth.] iii. The “rational expectations” revolution in the 1980s and 1990s undermined most of the early macro-political models and reduced interest in Political Business Cycle models for about a decade (until the late 1980s) when a real political business cycle literature emerged. Clearly, the economic expectations of voters matter, and there is significant disagreement about the nature of the economic theories and expectations of voters. Rational expectations models assume that voters/consumers/firms know all that can be known, and, consequently, have unbiased expectations about future events, including the behavior of politicians and the economic consequences of their actions. Clearly, both the particular macroeconomic assumptions about how or whether an economy responds to increases in the money supply or government spending and voter behavior (expectations) determine the feasibility of political business cycles. iv. In one strand, of the rational expectation based literature on electoral cycles focuses on the obvious, but previously neglected, fact that the party that wins an election can not be known before the election with certainty. Elections reveal which public policy is actually going to be used, which has real effects, because economic agents use those anticipated policies to make investment decisions and form other plans. Elections will thus have small real affects that tend to be concentrated shortly after the time of elections. If the policies of these two parties differ, real macro variables will change at the time one party or the other wins the election, partly because once the winner is known investers and consumers will take that information into account and change their behavior. Alberto Alesina and Nouriel Roubini pioneered this approach in their 1990 RE Stud piece on “Political Cycles in OECD Economies.” III. Possible escapes from the Majoritarian Cycling Problem A. Majoritarian Equilibria i. There are a number of possible escapes from the implication that majoritarian public policies are doomed to instability. ii. First preferences or ideology may be such that mappings to a single dimension are possible and so the real multidimensional space that public policies are chosen in can be collapsed to a single dimension without loss. iii. Pool and Rosenthal’s analysis of voting in the US Congress is consistent with this perspective. iv. Another possibility is that institutions have been chosen or have evolved within successful majoritarian political systems that prevent the cycling problem one way or another. B. Shepsle and Weingast (1981) provide an overview of various procedural rules which might generate “institutionally induced” majoritarian equilibria in cases where voter preferences are not fundamentally single dimensioned or arrayed with great symmetry. 6 EC742 Part 4: How Efficient is Majority Rule? i. For the most part, the procedures operate by controlling the options that can be voted on (agenda control). S&W (p. 503) "focus on the manner in which institutions transform pure majority rule into a different legislative game." v Add a project v Substitute one project for another v This process generates a stable equilibrium when the d cheapest projects are funded and d is the minimum # required for a winning coalition. ii. Many institutional arrangements in the US Congress appear to be stability enhancing (although not stability assuring): a. One example of a stability increasing institution is to allow only relatively large policy changes to be considered. If alternatives have to be far enough from the status quo (original position) then all feasible alternatives may be beyond the win set of the status quo. If the win set of Xo [ denoted W(Xo) ] is entirely within distance d of Xo then Xo is stable if only proposals further than d from Xo can be put on the agenda. a. Voting against the status quo after all amendments have been adopted. (Policy changes have to be in the win set of the status quo.) b. Having to be majority approved by a series of committees. (Policy changes have to be in the intersection of the win sets of each successive vote.) iii. The committee process: since proposed alternatives must win approval within committees and by the Congress as a whole (or at least in each chamber) committees have an incentive to take account of the preferences of the entire Congress so that their preferred policies can ultimately be adopted. (Consequently, committees provide some screening of policy options for the Congress as a whole.) iv. Bicameralism and presidential systems also constrain the range of policies that can be adopted, because there are multiple veto players. With sincere voting (which is usually assumed) a policy reform must make all veto players better off relative to the status quo. Often the result will be a continuation of the status quo rather than reform. (Note that the effects of multiple veto players are in many respects similar to those of super majority rules, as noted in Buchanan and Tullock [1962] and Tsebelis [2002].) v. Institutional arrangements that constrain the process of collective choice have the effect of reducing the domain of reform and protecting the status quo, but they also reduce the instability of the majority rule method of collective decision making. b. One can also delegate complete agenda control to one person. v In this case, the agenda setter will devise a series of votes which will approach his own personal ideal point, and not allow alternatives to be voted on which would be his ideal point's win set. v Note, however, that an agenda setter has enormous control over the outcome, when a series of votes can be arranged. [illustrate]. c. Policy changes may be voted on only one dimension at time. v This yields median voter outcomes in each dimension even in cases where no overall median position exists, when preferences are single peaked, as is usually implied by the standard neo-classical assumptions about tastes and is always implied by spatial voting. v A rough example of this type of rule is that Congress is supposed to pass thirteen separate appropriation bills each year rather than a single budget bill. (See Congleton and Sweetser, Public Choice 1992.) d. One can restrict the number of alternatives that can be voted on: e.g. let each person add a single project to the status quo or have a time limit for revisions. v Both these methods limit the number of votes undertaken, and assure that a decision will be made. v However, it bears noting that the order of the votes will often affect that decision. e. Restrict voting to a narrow range of discrete alternatives. For example, Allow only votes of the following form (Fiorina 1980): v Strike a project 7 EC742 Part 4: How Efficient is Majority Rule? ii. If voters economize on information costs by collecting no data about a subset of policy-relevant parameters, voter information is incomplete in another sense. IV. Moderating the Problems of Voter Ignorance A. One possible escape from the problem of rational ignorance is gounded in Condorcet’s jury theorem (1785). v Voters will be ignorant, rather than uncertain about relevant policy parameters and implications. v Confusion on this point exists because the Downs and Tullock discussions of rational ignorance are unclear about what is being “sampled” by voters and exactly what is meant by rational ignorance. i. It suggests that the problem of uninformed voters is not a major problem for democracies (Lupia and McCubbins 1998). ii. The jury theorem and the statistical properties of medians suggests that voters may not need very much information for democracy to yield “correct” policy outcomes. iii. When information about some relevant parameters is not available, not gathered, or not analyzed, biased expectations are very likely. v (See Levy and Peart [2002] for an overview of Galton’s analysis of median estimates.) v Lupia and McCubbins argue that party labels have enough informational content that voters rarely mistakenly vote for the “wrong” candidate. v Wittman (1995) argues that information problems are no worse in political markets than they are in private markets. v Congleton (2001) demonstrates that ignorance implies that unbiased estimates cannot be formed by even the most conscientious and rational voter. iv. Thus, the Condorcet Jury theorem moderates but does not eliminate the effects of voter ignorance. iii. These papers and others that more directly explore the implications of Condorcet’s jury theorem suggest that the policy choices made by majority rule tend to be far more accurate than one would expect based on survey evidence on voter knowledge. iv. Elementary statistics implies that even a small sample is sufficient to allow individuals to form unbiased (although imprecise) estimates of any phenomena. So, the source of fiscal illusion cannot be of small samples that result in search models (Stigler 1961). v. The information-aggregating effect of elections yields accurate policy assessments or candidate choices, insofar as the median error term in a large sample asymptotically approaches that of the underlying uncertainty. C. Two tables of simulated election results are provided below. Both are from Congleton 2007. Table 1 Electoral Outcomes as Estimates of Candidate Quality v This would be the case if median estimates are both unbiased and very accurate (Levy and Peart 2002). B. However, there are other ways than sample size in which a voter may economize on data collection. i. Congleton (2001) suggests that fiscal illusion and other forms of biased expectations are not based on small samples, but are rather consequences of complete ignorance of policy-relevant information. Electorate Size 11 Average Challenger Quality in Sample -4.114 Median Estimated Challenger Quality -4.068 Standard Error of Median Est. Candidate Quality 2.68 Number of Electoral Mistakes 13 101 -3.808 -3.975 2.77 14 501 -3.884 -3.827 2.63 14 1001 -3.964 -3.979 2.34 13 2001 -4.07 -4.093 2.31 10 4001 -3.787 -3.746 2.43 13 100 elections are simulated for each community of voters, electoral outcomes are tabulated, and sample statistics computed. The incumbents, reference candidates, and challengers are drawn from the same distribution and vary in each election. 8 EC742 Part 4: How Efficient is Majority Rule? i. Table 1 examines the case in which all voters are “slightly informed” in the sense that they have a bit of information and understand the determinants of candidate quality: Qi = a + bEi + ui ii. Note that relatively few errors are made and those that are made reflect the unavoidable effects of an observable candidate characteristic. The more important these unobservable charactiericis (ui), the more errors are made. i. The upper table assumes that 2/5s of the voter are “weakly” informed as in the above table. They take full account of the model but have very small samples. The others are rationally ignorant about various structural properties of the candidate quality function. Voters in column 2, assume that the challenger is of average quality (a) and ignores their observable talent and experience. Voters in column 3 focus only on the observable characteristics of candidates (Ei). The third column of voters are totally uninformed and votes more or less randomly. ii. The top half of table 2 demonstrates that majority rule can also ameliorate problems associated with voter ignorance (bias) in cases in which a sufficient number of voters have unbiased estimates (complete small samples). D. Table 2 examines the effects of rational ignorance on voter accuracy in assessing their own net benefits or the quality of candidates on the ballot. Table 2 Electoral Mistakes in Populations of Poorly Informed and Rationally Ignorant Voters Simulations: 100 elections, with 202 Poorly Informed Voters, and 101 of Each Type of Rationally Ignorant Voter Range of u Poorly Informed Electorate Rationally Rationally Rationally Ignorant of Ignorant of Uninforme B Electorate H d Electorate Electorate +/- 0.1 0 27 35 40 0 +/- 1.0 5 28 34 55 5 +/- 2.0 13 17 33 48 14 +/- 4.0 28 31 37 45 23 a. In circumstances under which poorly informed voters are a large group of voters, although not necessarily a majority, electoral outcomes are only slightly more error prone than in cases in which only poorly informed voters exist. b. Indeed, in the case in which only a small unobservable stochastic factor exists, majority rule yields essentially error-free results, even though very substantial error rates occur in each of the rationally ignorant sub-populations. c. The errors and biases of the rationally ignorant groups roughly offset each other, and the electoral outcomes are largely determined by the unbiased poorly informed group of voters, even in relatively difficult circumstances. Overall Electoral Mistakes iii. The lower table assumes that that there are equal numbers of each type of voter. This reduces the accuracy of the overall result, without affecting that of any of the voter subgroups. iv. The bottom half of table 2 demonstrates that the “jury theorem” effect is greatly reduced as ignorance becomes relatively more commonplace among voters. Simulations: 100 elections, with 51 Poorly Informed Voters, and 153 of Each Type of Rationally Ignorant Voter +/- 0.1 1 25 54 49 28 +/- 1.0 5 19 34 55 24 +/- 2.0 15 27 49 43 38 +/- 4.0 21 24 36 51 32 v In this case, the error rates for the electorate as a whole are systematically higher than those of the poorly informed, because unbiased voters no longer determine the electoral outcome. 9 EC742 Part 4: How Efficient is Majority Rule? v Instead, idiosyncratic features of experience and ignorance (the observable and unobservable characteristics of the reference politician focused on) determine the accuracy of majoritarian decisionmaking. v Indeed, the effects of rational ignorance appear to be more important than the difficulty of the estimation problem faced! v (There is no systematic increase in mistake rates for the electorate as a whole as the range of u increases, although there is a systematic increase in mistakes as the proportion of rationally ignorant voters increases.) v. Nonetheless, even toward the bottom of table two, where the range of u and the electorate composition are the least favorable to democratic decisionmaking, majority rule still ameliorates the problem of rational ignorance to some extent. vi. The final outcome is less error prone than two of the three rationally ignorant subpopulations, although far more error prone than is the group that collects a small amount of comprehensive information. vii. Rational ignorance is not fully countered by the jury theorem. E. The relevance of rational ignorance for public policy is likely to vary among policy areas because information costs do. i. In areas in which most voters have a bit of independent reliable information, the results of electoral competition should yield outcomes that are unbiased estimates of median voter interests. ii. In areas in which such information does not exist, electoral pressures will not necessarily achieve this result. iii. Interest groups that conduct public informational campaigns may push the equilibrium in such cases in directions that advance their interests, whether ideological, profits, or public. 10