Democracy demands an informed electorate.ters who lack adequate knowle - PDF document

Routing Democracy demands an informed electorate.ters who lack adequate knowledge about poli-tics will find it difficult to control public policy.Inadequate voter knowledge prevents govern-ment from reflecting the will of the people in anymeaningful way. Such ignorance also raisesdoubts about democracy as a means of servingthe interests of a majority. Voters who lack suffi-cient knowledge may be manipulated by elites.They may also demand policies that contravenetheir own interests.The American electorate does not have ade-uate knowledge for voters to control public pol-icy. Scholars have long documented the limits ofter knowledge about the institutions and poli-cies of the government. That ignorance is not amoral failing. The rational voter has little incen-tive to gain more knowledge about politicsbecause his or her vote is unlikely to affect theoutcome. Since gaining more knowledge offersw benefits and substantial costs, the averageSome scholars have argued that citizens useÒshortcutsÓ to gain enough knowledge to partic-ipate in self-government. The evidence does notsupport the ÒshortcutÓ argument.The size of modern government is often sogreat that it is impossible for votersÑeven themost knowledgeable among themÑto be ade-uately informed about its operations. Smallergovernment may actually be more democraticthan that which we have now: voters would bemore likely to exercise informed control over pol-icy. Voter ignorance also suggests the value ofdecentralized federalism. In a decentralized fed-eral system, citizens may Òvote with their feetÓ bymoving out of jurisdictions with policies theydislike and into those that have more favorableones. Because each person decides whether ornot to move, there is a much greater incentive toacquire relevant information with Òfoot votingÓthan with traditional voting at the polls. When Ignorance IsnÕt BlissHow Political Ignorance Threatens Democracyby Ilya Somin Ilya Somin is an assistant professor of law at George Mason School of Law. His research interests are constitutionallaw, property law, and popular political participation and its implications for constitutional democracy.Executive Summary No. September 22, 2004 An informed electorate is a prerequisiter democracy. If voters do not know what isgoing on in politics, they cannot rationallyexercise control over government policy.Large-scale voter ignorance poses a seriouselection and beyond. It is particularly trou-bling at a time when we face a close wartimeelection with major policy decisions at stake.Inadequate voter knowledge has twomajor negative implications for democracy.First, it prevents democratic governmentfrom reflecting the will of the people in anymeaningful sense, undercutting the Òintrinsi-cistÓ defense of democracy as a governmentthat reflects the voluntary decisions of theLikewise, voter ignorance imperilsthe instrumental case for democracy as aregime that serves the interests of the major-ity, since ignorance potentially opens thedoor for both elite manipulation of the pub-lic and gross policy errors caused by politi-torate in order to win office.In this paper I review the overwhelming evi-meet even minimal criteria for adequate voterknowledge. I then examine the implicationsor American politics. Part I lays out minimalknowledge prerequisites for voter control ofpublic policy, summarizes the massive evi-dence of voter ignorance that students of thesubject have accumulated over the years, andhighlights some of the most disturbing impli-cations of those studies. Part II examines morerance. It shows that extensive voter ignorancetion cycle. These data are significant becausethe extremely close and controversial nature ofthose two elections might have been expectedto cause an increase in voter knowledge. InPart III, I review and criticize theories thatclaim that Òinformation shortcutsÓ enable vot-ers to control government in spite of pervasiveignorance. Those mechanisms for dealingwith voter ignorance are unable to overcome itand sometimes even exacerbate the problem.Part IV restates the argument that ignorance islargely Òrational,Ó rooted in the very low likeli-hood of a single vote being able to influencePart V advances the claim that the size andscope of the modern state are so great that itis often impossibler votersÑeven the mostknowledgeable among themÑto be ade-uately informed about its operations. Thisconclusion leads us to the counterintuitivesuggestion that a smaller government mayhave now, in so far as voters would standa greater chance of being able to influencegovernment policy in an informed manner. Ibriefly examine some evidence from 19th-century American history that suggests thatwas able to consider far more complex policytoday and hypothesize that this differencewas largely due to the ability of voters toocus on the relatively small number of issueser which the strictly limited government ofthat era exercised control. Finally, Part VI shows that voter ignoranceprovides an unanticipated argument in favorof decentralized federalism. Decentralizationallows citizens to Òvote with their feetÓ bymoving out of jurisdictions with policies theydislike and into those that have more favor-able ones. Because each person can decide forherself whether or not she will move, there isa much greater incentive to acquire relevantinformation with Òfoot votingÓ than with tra-ditional ballot-box voting.study are correct, efforts to increase the stockof knowledge possessed by voters are unlike-ly to be more than modestly effective. A morepromising path for reform would be toreduce the amount of knowledge requiredor democratic control of public policy.Finally, the studyÕs conclusion stresses thetradeoff between big government and demo-cratic government that all advanced industri-alized states must face. modern state areoften impossiblefor voters to beadequately I: How Ignorant Are Voters?Knowledge Prerequisites for DemocraticWhat must voters know in order to exer-cise democratic control over government pol-In their classic work, The American Votera University of Michigan Survey ResearchCenter team defined three minimal knowl-edge prerequisites for voters to be able toexert meaningful influence over a given issue:ters must be aware of the issueÕs exis-2. They must have a position on the issue.3. They must know the positions on thegiven election.Those three conditions have formed thebasis for most empirical investigations ofindividual-level voter ignorance since American VoterBut they are insufficient pre-requisites for meaningful control over publicpolicy. In addition to awareness of the exis-positions on them, informed voters musthave at least substantial understandingabout which of the available policy optionsare most likely to advance their goals. Unlessthe value voters attach to policy in a giventakithey cannot use the ballot to forceelected officials to serve their interests with-out knowing what the likely effects of alter-native policy options are.Just how informed should voters be?Ideally, they should be able to choosebetween opposing candidates and their plat-orms on the basis of Òthe preferences thatpeople would have if their information wereperfect.ÓAlthough that is impossible, informed voters should at least beware of basic tradeoffs between alternativepolicies in cases in which those tradeoffselites informed about the issue at hand.Moreover, minimally informed voters shouldnotdraw linkages between policies and out-comes that informed observers would con-sider obviously absurd; for instance, the factthat a majority of American voters with anopinion on the issue believe that the federalgovernment is too large and powerful andsimultaneouslyfavor increased spending inalmost every major area of federal involve-ment is a clear case of ignorance of tradeoffsthat falls below the threshold of minimallynecessary knowledge.On the other hand,ll informed about the effects of a particu-lar policy, support of or opposition to thatpolicy should not in and of itself count aspresumptive evidence of voter ignorance.That standard adds a crucial substantivedimension to the criteria for minimally nec-essary voter knowledge without judging vot-ersÕ competence by how closely their policypreferences match those of the analyst.Moreover, nothing in the definition foreclos-es the possibility that it may not be necessaryter to possess the requi-site knowledge so long as the electorate in theaggregatepossesses sufficient informationand signaling capacity to act Òas ifÓ votersere individually informed.Extent of Ignorance As political scientist John Ferejohn haswritten, ÒNothing strikes the student of pub-lic opinion and democracy more forcefullythan the paucity of information most peoplethe well-established conclusion that mostindividual voters are abysmally ignorant ofeven very basic political information. Eversince the seminal research of the 1950s andreinforce this finding.onetheless, the sheer depth of most indi-vidual votersÕ ignorance is shocking toobservers not familiar with the research.Currently, almost 70 percent of Americans donot know that Congress recently adopted alaw adding a massive prescription drug bene- 3 individual votersare abysmally en very basic the most important piece of domestic legisla-tion adopted during the administration ofGeorge W. Bush.Equally striking is the factthat more than 60 percent do not realize thata massive increase in domestic spending hasmade a substantial contribution to the recentexplosion in the federal deficit.A survey taken immediately after the close-ly contested November 2002 congressionalelections found that only about 32 percent ofrespondents knew that the Republicans con-trolled the House of Representatives prior toresearch showing widespread ignorance ofcongressional party control in previous elec-Such widespread ignorance is not of recentorigin. As of December 1994, a month afterthe takeover of Congress by Newt GingrichÕsepublicans, 57 percent of Americans hadnever even heard of Gingrich, whose campaignstrategy and policy stances had received mas-sive publicity in the immediately precedingIn 1964, in the midst of the Cold War,only 38 percent were aware that the Sovietnota member of NATO.the time, only bare majorities know whichparty has control of the Senate, some 70 per-cent cannot name either of their stateÕs sena-tors, and the vast majority cannot name height of a campaign.Overall, close to one-Òknow-nothingsÓ almost completely ignorantof relevant political information.Three aspects of voter ignorance deserveparticular attention. First, voters are ignorantnot just about specific policy issues but alsoabout the basic structure of government andhow it operates.such basic aspects of the U.S. political systemas who has the power to declare war, therespective functions of the three branches ofgovernment, and who controls monetary pol-icy.That suggests not only that voters can-not choose between specific competing policyprogramsbut also that they cannot easily assigncredit and blame for highly visible policy to the right officeholders. The long-noted tendency of voters to almost automati-one reflection of this problem.The second salient aspect of voter igno-rance is that most voters lack an ÒideologicalÓple issues into a single analytical frameworkderived from a few basic principles; ordinaryconsistent stance on issues that is evident insurveys of political elites.Most scholars fol-low Anthony Downsin emphasizing theusefulness of ideology as a ÒshortcutÓ to pre-dicting the likely policies of opposing partiescompeting for office.at least equally important, is the comparativeinability of nonideological voters to spot inter-ware electorate would not be oblivious to thecontradiction between seeking a reduction ingovernment power and an expansion of near-structure and of ideological interconnectionscreates serious obstacles to democratic controlof government. Without knowledge of theideological interconnections between issues,there is unlikely to be a sufficient preexistingbase of knowledge for the impressions gainedfrom campaign information to be accurate;even if they are accurate, they cannot be easilyconnected with the voterÕs policy objectives. Itis no surprise, therefore, that the small minor-ity of well-informed voters is much better ableto process new political information and moreresistant to manipulation than is the unin-Finally, it is important to note that thelevel of political knowledge in the Americanelectorate has increased only very slightly, if atall, since the beginning of mass surveyresearch in the late 1930s.A relatively stablelevel of extreme ignorance has persisted evenin the face of massive increases in educationalattainment and an unprecedented expansionin the quantity and quality of informationThis striking failure throws doubt on the interconnectionscreates seriousobstacles tocontrol of expectation of political theorists from JohnStuart Millonward that increased availabil-ity of formal education can create theinformed electorate that the democratic idealrequires.Individual-level voter ignorance seemsdeeply rooted, perhaps ineradicable. It fol-lows that the ability of voters to meet thedent on the validity of ÒshortcutÓ theoriesthat predict that voters can cast informedtes without themselves possessing evenminimal levels of political knowledge.II: Recent Evidence of This part considers recent evidence of wide-spread political ignorance in the UnitedStates. I consider limited data available fromthe current election cycle, as well as more sys-tematic data from the time of the 2000 elec-tion. The analysis of voter ignorance duringthe 2000 election is based on data from the2000 National Election Study, an extensivenationwide survey of more than 1,800 respon-that included 31 political knowledgeitems covering a wide range of subjects.Cycledo not as yet have a large-scale compre-hensive data set on political knowledge in thecurrent election cycle. However, Table 1 pre-sents evidence from a number of recent surveysthat indicates extensive political ignorance onmajor issues in the current campaign.The available data cover a number of basicthat are currently prominent in both thepress and political debate. Perhaps the mostunaware of the passage of some of the mostimportant and controversial items on theda: almost 70 percent did not know of thepassage of the massive Medicare prescriptiondrug benefit, and nearly 65 percent do notknow of the recent passage of a ban on par-tial birth abortion. Similarly, 58 percentadmit they have heard Òvery littleÓ or Ònoth-ingÓ about the USA Patriot Act, the much-debated 2001 legislation that increases lawenforcement powers for the claimed purposeof fighting terrorism. This survey resultprobably actually understates the number ofrespondents who know little or nothingThe survey evidence also indicates consid-erable ignorance about various hot-buttondomestic and foreign policy issues. Despitewidespread press coverage of large recent job7 poll mistakenly believed that there hadbeen a net loss of jobs in 2004. With regard tothe most important foreign policy issue inthe campaign, a majority mistakenly believedbetween Saddam Hussein and the September11 attacks (despite the administrationÕs ownrepeated disclaimers of any such connec-tion), and most do not know even approxi-mately how many American lives have beenlost in the Iraq war. Despite the ongoingdebate over AmericaÕs troubled relationshipwith Europe and the onset of European uni-ication, 77 percent admit they know ÒlittleÓor ÒnothingÓ about the European Union. Particularly significant is the fact that, onmany issues, the majority is not only ignorantof the truth but actively misinformed. Forexample, 61 percent believe that there hasbeen a net loss of jobs in 2004, 58 percentbelieve that the administration sees a linkbetween Saddam Hussein and 9/11, and 57percent believe that increases in domesticspending have notcontributed significantly tothe current federal budget deficit. Whetherthose misconceptions will have an impact onthe outcome in November remains to be seen.The data in Table 1 should not be taken asproof that the public is universally ignoranton every issue. Some basic facts about currentell known. For example, 82percent know that there is currently a federalbudget deficit,and 79 percent know that thedeficit has increased during the last four are unaware ofimportant andcontroversialitems on the onetheless the evidence compiled inable 1 does show that majorities are ignorantmost important and most widely debatedissues at stake in the present election. Thatresult is particularly striking in view of theextremely close and controversial nature of thecontest and the high level of press coveragemany of those issues have received.Summary of Aggregate Findings of2000 National Election Study 6 Item (date of survey)% Correct Answer% Wrong Answer% Don't KnowKnow that defense spending is one of two largest expenditure 51436areas in federal budget (Mar. 15-May 11, 2004)Know approximate number of U.S. troops killed in Iraq 40 (within 200) 3426(Apr. 23-25, 2004)Know that increased spending on domestic programs has 39574(Feb. 11…16, 2004)Claim to have heard or read at least someŽ information 39 (some,Ž 27; 58 (not much,Ž 28; 3about USAPatriot Act (Apr. 28, 2004)or a lot,Ž 12)or nothing,Ž 30)Know that there has been a net increase in jobs during 200436613Know that Congress has recently passed a bill banning 361748Know that Congress has recently passed a Medicare 311654prescription drug benefit (Apr. 15, 2004)Know that Social Security spending is one of two largest 32626expenditure areas in federal budget (Mar. 15…May 11, 2004)Know that the Bush administration does not believe that 255817Saddam Hussein was involved in 9/11 attacks Know that current unemployment rate is lower than average 226315rate for last 30 years (Mar. 23, 2004)Claim to know at least fair amountŽ about the European 22 (great deal,Ž 3; 77 (very little,Ž 37; 1or fair amount,Ž 19)or nothing,Ž 40)Princeton Survey Research Associates Survey, March 15…May 11, 2004, Roper Center, Accession no. 0454615.NYTimes/CBS survey, April 23…25, 2004, Roper Center, Accession no. 0452110.Pew Research Center Survey, February 11…16, 2004, Roper Center, Accession no. 0448774.NYTimes/CBS survey, April 23…27, 2004, Roper Center, Accession no. 0452116.IPSOS Public Affairs Poll, June 7…9, 2004, Roper Center, Accession no. 0455028.Princeton Survey Research Associates Survey, December 7…9, 2003, Roper Center, Accession no. 0449953.Princeton Survey Research Associates Survey, April 15, 2004, Roper Center, Accession no. 0451671.Princeton Survey Research Associates Survey, April 15, 2004, Roper Center, Accession no. 0451671.NYTimes/CBS survey, December 14…15, 2003, Roper Center, Accession no. 0444433.Fox News/Opinion Dynamics Survey, March 23…24, 2004, Roper Center, Accession no. 0450576.Gallup survey, May 21…23, 2004, Roper Center, Accession no. 0455008. rance in the current election cycle is powerful-ly reinforced by much more systematic datafrom the 2000 election provided by the 2000ational Election Survey. Undertaken duringevery election year since 1948, the NES is gen-entific survey of the U.S. electorate.The 2000 NES contained a total of 31political knowledge questions. They are list-ed in Table 2 along with the percentage ofrespondents giving correct answers.early all of the 31 survey items identifiedin Table 2 are quite basic in nature and wouldhave been well-known to political elites andMost addressed issuesthat were widely debated during the 2000 cam-paign, including environmental policy, govern-ment spending on services, abortion, and poli- 7 Political Knowledge Survey Items from the 2000 NESCorrect Answer Identify Texas as home state of George W. Bush90Know Bill Clinton is moderate or liberal81Know Al Gore favors higher level of government spending on services than George W. Bush73Know Democratic vice presidential candidate Joe Lieberman is Jewish70Identify Tennessee as home state of Al Gore 68Know federal budget deficit decreased, 1992…200058Know Gore is more liberal than Bush57Know Democrats favor higher level of government spending on services than Republicans57Identify attorney general as post held by Janet Reno55Know Republicans controlled House of Representatives before election55Know Gore is more supportive of gun control than Bush51Know Republicans controlled Senate before election50Know Democrats more supportive of government guarantee of jobs/standard of living than Republicans49Know George W. Bush is conservative47 (30 chose moderate)Know Gore is more supportive of abortion rights than Bush46Know Gore is more supportive of government guarantee of jobs/standard of living than Bush46Know Democrats favor higher level of government aid to blacks than Republicans 45Know Gore is more supportive of environmental regulation than Bush 44Know Bush is more likely to favor jobs over environment than Gore41Know presidential candidate Pat Buchanan is conservative40Know Gore favors higher level of government aid to blacks than Bush 40Know Al Gore is liberal Know federal spending on the poor increased, 1992…200037Know crime rate decreased, 1992…200037Identify British prime minister as post held by Tony Blair35Identify Connecticut as home state of Democratic vice presidential candidate Joe Lieberman 30Identify Wyoming as home state of Republican vice presidential candidate Dick Cheney 19Correctly name at least one candidate for House of Representatives in respondents district 15Identify Supreme Court chief justice as post held by William Rehnquist 11Identify Senate Majority Leader as post held by Trent Lott 9Correctly name second candidate for House of Representatives in respondents district4Note: All percentages rounded to whole numbers. N = 1,543 respondents. cy toward African Americans. Several questionsof the Clinton administration, for which presi-dential candidate Al Gore and the DemocraticParty more generally attempted to claim aAlthough the 31 questions donot cover all possible relevant issues and facts,they do include a wide range and are thereforea good representative sampling of AmericansÕpolitical knowledge. Moreover, previous stud-ies have found that political knowledge in onearea is highly intercorrelated with knowledgein others.Thus, we can be reasonably confi-dent that individuals who scored well on the 31political knowledge on other matters thanthose who scored low. A Glass Half Empty or Half Full? HowLow Is the Knowledge Level Revealed inthe NES Data?The average knowledge level in the 2000NES was roughly similar to that detected inearlier studies and generally low. On average,respondents answered correctly only 14.4uestions out of 31.The data also seem toconfirm Stephen BennettÕs findings thatabout one-third of respondents are Òknow-nothingsÓ possessing little or no politicallyrelevant knowledge.bout 25 percent ofrespondents got 8.5 or fewer correctanswers.Since 17 of the 31 questions hadonly three possible answers,two had onlytwo possible answers,one more had twocorrect answers out of a possible three,several others could also potentially beguessed with lower probabilities of success,is almost exactly equal to theMy finding of 25 percentÒknow-nothingsÓ is very similar to BennettÕsinding of 29 percent.netheless, it is possible to argue that theerage knowledge level revealed in the 2000NES is not too low because the average respon-dent did achieve correct answers on almost halfthe questions (46 percent). This claim is flawedor two reasons. First, with minor exceptions,the items in the survey represent very basicpolitical knowledge, without which it is diffi-cult or impossible to place more complex andspecific knowledge in useful context. Know-ledgeable political activists and even citizenswho follow politics reasonably closely wouldprobably be able to answer all but a tiny hand-ful of the questions correctly. The second reason for pessimism regard-bly actually overestimate current Americanpolitical knowledge. That overestimation isthe result of two factors. First, surveys in gen-eral somewhat overestimate the amount ofpolitical information possessed by the publicrespondents and because more knowledge-able citizens may be overrepresented amongthose surveyed.The average respondent inthe 2000 NES got only about 6 more correctanswers out of 31 than would be expected asa result of random guessing.respondents had the option of giving ÒdonÕtknowÓ answers to questions, past researchshows that survey respondents often expressopinions about issues they know nothingabout to avoid seeming ignorant.seems likely that many respondents who didnot know the answer to various questionsattempted to guess, especially on those itemsthat had only two or three possible answers.Second, three of the five items with thehighest percentage of correct answers are per-sonal information about candidates in the2000 election that has little or no value forunderstanding politics more generally.Those three items are the home states ofGeorge W. Bush and Al Gore (90 percent and68 percent correct answers, respectively) andJoe LiebermanÕs religion (70 percent). BillClintonÕs ideology, the second-highest scoringitem (81 percent correct answers), is an artifactof generous coding on my part, under whichboth ÒliberalÓ and ÒmoderateÓ answers werescoring low-value items and two other similaruestions, which produced much lower per-centages of correct answers,erage score of 11.5 correct answers to 26uestions, for a 45 percent average, which is a Surveys in overestimate the amount ofpossessed by the slightly lower percentage than that observedMuch more signifi-cant, the elimination of the five low-valueuestions (while retaining the Clinton ideolo-gy question) increases the proportion ofÒknow-nothingsÓ to about 34 percent, a per-centage considerably higher than BennettÕsresults of three knowledge scales from theAs Table 3 shows, the already low averageknowledge scores on the 2000 NES concealthe existence of a large political knowledgeunderclass of Òknow-nothingsÓ who possesslittle if any basic political knowledge.constitutes from 25 percent to 35 percent ofOverall, considering (1) the very basicnature of the questions asked, (2) the possi-bility of guessing, and (3) the high percentageof Òknow-nothingÓ respondents, it is difficultto avoid the conclusion that the 2000 NES,like most research using earlier evidence,reveals a low level of political knowledge.III: The Shortcomingsof ShortcutsIf voters have generally low knowledge lev-els, they may be able to make up for it byusing information Òshortcuts.Ó Until recent-ly, this was the clearly dominant view amongThis part criticallyassesses the most important of the variousshortcuts to informed voting proposed in theliterature on the subject: information fromdaily life, political parties, cues from opinionleaders, retrospective voting, issue publics,have taken up AnthonyDownsÕssuggestion that rational voters willmake use of information acquired throughordinary daily-life interactions; such informa-tion is virtually ÒfreeÓ since the activities thatproduce it would, by definition, be undertakeneven in the absence of any political purpose.nlike Downs, more recent advocates of thisshortcut argue not only that it will be used butthat it goes a long way toward meeting votersÕinformational needs.or example, votersallegedly can obtain Òa good deal of informa-tionÓ about the economy from personal finan-cial transactions such as managing a checkingaccount or seeking employment.Fiorina goes so far as to suggest that Ò[i]n orderto ascertain whether the incumbents have per-ormed well or poorly citizens need only calcu-late the changes in their own welfare.ÓAlthough it would be foolish to deny thatsome helpful information can be derived knowledge Òknow-nothingsÓconstitutes from25 percent to 35percent of the Aggregate Knowledge Scales from the 2000 NES verage Number of % Know-NothingŽScaleCorrect AnswersRespondents31-Question Scale 14.4 (46%) 25 26-Question Scale 11.7 (45%) 34 25-Question Scale 10.9 (44%) 35Know-nothingsŽ include those who scored 7.5 or fewer correct answers out of 24. The calculation is identical from ordinary life, its usefulness to otherwiseill-informed voters is greatly overestimated.Three major limitations of such informationare particularly important. First, by defini-with the many political issues that the vastmajority of voters do not encounter in dailylife. Second, even if the voter has, followingchanges in his welfare and developed a judg-not readily determine whether his welfarewill be improved by electing the opposingcandidate. Even if things have gotten worseprogram is even more harmful. That possi-bility cannot be ruled out without substan-tive issue knowledge going beyond personalMost important of all, substantive knowl-edge is required to determine whether or nota particular personal experience really is thecal actors are responsible. Ill-informed votersattempting to make political judgments onthe basis of personal experience may fall intoegregious errors. Even with respect to unem-ployment and inflation, basic economicissues with which most people have substan-tial personal experience, ill-informed voterstend to make spectacular errors. In a surveyeconomic issues were a particular focus ofpublicity, the vast majority of respondentscould not estimate the inflation or unem-ployment rate within 5 percent of the actuallevel;the electorateÕs mean estimates of bothrates were approximately twice as high as thereal level.Such misperception apparentlyplayed a major role in swinging the 1992 elec-Herbert Walker Bush.oorly informed vot-ers are more likely than well-informed ones tomake sweeping generalizations from personalexperience with unemployment but less likelyto make accurateconnections between experi-ence and policy.cases of inflation and unemployment, evenmore serious mistakes can be expected withher, more remote, issues. And even a correctestimate of unemployment and inflation isonly a minimal prerequisite to determiningwhich sideÕs policy on those issues will betterserve the voterÕs interests. One still needs toknow to what extent incumbents are responsi-ble for current rates and whether or not theiropponents are likely to do better. In and ofitself, information from daily life is unlikely tobe of much help in making such decisions. litical PartiesThe idea that political parties can helpters economize on information costs has along and venerable lineage, dating back toDemocratic Party leader and later PresidentMartin Van Buren, founder of the first mod-ern mass-based party. The basic argumentclaims that voters can infer candidatesÕ policystances from their partisan affiliations ratherthan undertaking the much more difficultask of inquiring into the views of each indi-vidual aspirant to office.This claim is not entirely without merit.this general argument, V. O. Key adds thenotion of party identification as a Òstandingexperience with the partyÕs officeholders, andMorris Fiorina presents evidence that votersrationally make use of past experience withthe two parties in creating a Òrunning tallyÓof their relative merits.as much as it reveals. At best, a candidateÕsparty affiliation is a clue to his policy stances,but it tells the voter little about the likelyeffectsof those policies. Party affiliation helpsters meet the first and third knowledgerequirements of the Michigan criteria butprovides little guidance on relating thatinformation to the voterÕs own goals. In prin-ciple, a running tally may help a voter todetermine the merits as well as the content ofa partyÕs policies. But it is difficult to do sowithout substantial underlying substantiveknowledge. If conditions are good under therule of Party X, how does the voter know thatthis is due to the success of the partyÕs poli- voters attemptingto make politicalbasis of personalfall into egregiouserrors. cies rather than to factors beyond politicalcontrol, preexisting favorable trends result-ing from decisions made by the partyÕs pre-decessors in power, personal characteristicsof the partyÕs officeholders that are not rep-resentative of the party as a whole and thusbehavior; or crafty manipulation of publicpolicy by the partyÕs leadership as a result ofwhich temporary success is achieved at theprice of long-term harm, the effects of whichare felt years or even decades after those lead-ers leave office? The voter cannot get aroundthis dilemma simply by aggregating largeamounts of experience, since it is unlikelythat a given voter has been following politicslong enough to experience more than two orthree governments headed by any oneparty.Most dramatic, existing research on par-ties has largely ignored an important waythat the existence of political parties maythe flow of information to vot-ers relative to a nonpartisan electoral system.If politicians are organized into (relatively)centralized parties, the number of effectivepolitical actors in the system is reduced; inrarely more than four or five major parties atany given time. As in any other competitivepetitors, the greater the chance of successfulIn any situation inwhich the number of major parties is small,especially in a two-party system such as thatin the United States, there is the possibilitythat the parties may conspire to take an issueof potential interest to the public off thepolitical agenda when doing so benefitscitizenry. Such a cartel also diminishes theof information about the issue to thepublic, since it is no longer discussed by can-didates running for office and the media areless likely to cover it. Empirical examples of such collusion arenot difficult to come by. The very first mod-ern-style party system was established in partto take the issue of slavery off the politicalagenda in this way.That cartel lasted forabout 25 years. Other cases from a variety ofInterparty collusion can be broken by theentry of new parties into the system, just asthe Republican Party eventually emerged tochallenge collusion over slavery. But organiz-ing a major new party is extremely costly, andthe existing parties can defend their oligop-late the electoral and campaign finance sys-tems. At the very least, such Òpolicy cartelsÓcan persist for a long time, even on somomentous an issue as slavery.This argument should not be interpretedto imply support for a nonpartisan politicalsystem; such systems have their own seriousshortcomings.onetheless, it is strikingthat a generation of scholars heavily influ-enced by economic theories of competitionshould have maintained a largely uncriticalenthusiasm for strong political parties with-out taking account of a serious objectionderived from those same theories. Althoughreal, the informational benefits of parties arealmost surely exaggerated by conventionalCues from Opinion Leadersrant, perhaps it can follow the lead of theknowledgeable minority of political activists,shortcuts.Instead of keeping close track ofissues themselves, voters can respond to cuesissued by political activists whose values aresimilar to their own. What is important isthat there are perhaps 5 percent of votersplay close attention. If they see that some-thing is seriously wrong in the country, theysound the alarm, and then ordinary peoplestart paying attention.tunately, the strategy of followingcues from opinion leaders creates at least asmany difficulties for ignorant voters as itsolves. Because of the immense asymmetry ofinformation between leaders and followers litical partiesreduce the flowof information tovoters. and the low incentive of the latter to monitorthe leadersÕ performance effectively, seriousrom the perspective of the principal, it is dif-cult to conceive of a more difficult princi-pal-agent relationship than that betweenignorant voters and highly knowledgeable,Where voter interests and activist interestscoincide closely, the difficulties of monitor-ing need not be so acute. But that state ofaffairs is far from common. Political activistsdiffer greatly from the general population onMost impor-ant of all, opinion leaders acquire intereststhat diverge sharply from those of voters ply by virtue of becoming opinion leadersical activists, their power, prestige, social sta-tus, and opportunities for pecuniary gain willtend to rise with the publicÕs perception oftheir issue positions; they thus have strongincentives to exaggerate the importance ofpolitical problems and to push for politicalnent role for activists) in preference to pri-vate-sector ones. Even when voters are awareof the incentives for exaggeration andattempt to discount activist claims as aresult, they have no way of knowing howmuchdiscounting is required. Even if there exists a subset of opinionthose of a given voter, that voter still faces anextraordinarily difficult problem in deter-on information costs, the voter is unlikely toinvest heavily in researching the leadersÕ qual-ifications. And, unlike in the case with mostprivate-sector specialist professionals, theter cannot simply judge the quality ofactivistsÕ performance by the results of thepolicies they advocate, since it is not usuallypossible to determine which social outcomessiderable substantive knowledge of theA successful strategy of following cuesfrom opinion leaders requires voters to firstwhichleadersÕ cues to follow and thenmonitor those leaders in order to avoid a vari-ty of principal-agent problems that are likelyconsiderable substantive voter knowledge ofthe issues. Without such knowledge, opinionleaders are as likely to be misleading as theyare to be informative.trospective VotingThe retrospective-voting hypothesis holdsthat voters judge politicians by past perfor-argument is advanced that Òretrospective vot-ing requires far less of the voter than prospec-tive voting.Óthat the answer is yes. First, as noted above, it isoften difficult for ignorant voters to determinepolicy and which arenÕt. To take a prominentexample in the literature, many models of elec-toral retrospection are based on ÒsociotropicÓting, in which voters make their decisions onomy rather than that of their own personalt a person ignorant of economics(sometimes even a trained economist) cannottell whether economic conditions are the resultof (1) the policy of the current government, (2)lagged effects of its predecessorsÕ policies, or (3)factors completely independent of any govern-Even if option 1 is the case, the voter maynot be able to determine whether current con-ditions are positive or negativeÑif, for exam-ple, temporary economic sacrifice might be anecessary precondition for future progress.Even where the voter does know that a givenoutcome is the result of government policy,ignorance of the structure of government maymake it difficult for him to decide which elect-ed officials deserve credit or blame. Under anAmerican divided government or a Europeancoalition government, the voter may not evenbe able to tell which party is responsible. Thatdefect is of particular importance for retro- Opinion leadersacquire intereststhat divergesharply fromthose of voterssimply by virtue of spective-voting theory, which emphasizes thatnot policiesÓand thus implicitly assumesthat voters know which leaders are responsibleFinally, even when the voter knows boththat a given outcome is the result of govern-ble, that is still not quite enough to make aninformed choice. The voter will also want toknow whether the opposition party is likelyto do better. In order to make retrospectiveting Òwork,Ó the voter must first establishhow good past performance has been, and itis not possible to do so without addressingThe retrospective-voting argument does,however, possess a kernel of truth. As Fiorinaputs it, retrospective voting can impose awho have failed badly.large, highly visible, and easily attributable toa particular set of leaders, it is certainly likelythat they will be voted out of office, as theelections of 1932, 1952, 1968, and 1980 sug-gest. Moreover, the bigger the failure, the lesslikely it is that the opposing partyÕs perfor-mance will be worse. The ability of voters toincumbents is one of the major advantages ofdemocracy over dictatorship.tunately, the preconditions of mag-nitude, visibility, and easily traceable account-ability rarely obtain in real life. Even in the caseof a very large policy failure, leaders mayescape blame if the full impact of the failure isnot felt until after they are out of office.Herbert Hoover and Jimmy Carter paid theprice of perceived failure in 1932 and 1980,respectively, but their predecessors (Coolidge,Nixon, and Ford), who arguably had at least asmuch to do with of the failures in question,did not.If voters cannot keep track of all theimportant issues, perhaps they can at leastocus on a few that are of particular concernto them.or instance, blacks are more like-In theory, such Òissue publicsÓ canmake up for ignorance of more general poli-Attempts to confirm the issue-publicypothesis empirically show that it has onlylimited validity. Knowledge of differentaspects of public policy is highly intercorre-Even where significant differences inknowledge between groups do exist, they donot necessarily demonstrate that the knowl-edge of the better-informed group is ade-uate for informed voting; they show onlythat members of that group know torate. The difference is crucial, because moststudies showing that issue publics are betterinformed about a particular issue than therest of the electorate rely on surveys tappingonly very basic knowledge. Even if the voter does have adequate knowl-edge of the narrow issue of particular concernto him, informed voting with respect to thatissue might still be inhibited by ignorance ofthe Òrules of the gameÓ of government policy.A black voter may have sufficientspecificknowledge to conclude that current civilrights policy should be changed but notgeneralknowledge of the structure ofgovernment to determine which elected offi-cials have to be voted out to do it. Even in theland of the blind, the one-eyed man cannot bea true king if kingship requires seeing thingsthat can only be discerned with two eyes.ter ignorance also undercuts the utilityof issue publics in two further, less obviousways. First, the rationally ignorant voter can-not readily tell which aspects of public policyreally are part of the issue of interest. One ofthe problems in issue-public research is theuestion of how the scope of the relevantÒissueÓ is defined in the first place. If the con-nection between two or more matters of pub-lic policy is not obvious or is ignored by politi-cians and the media for their own reasons, vot-ers may fail to pick it up. Social Securityreform, for instance, is almost never defined asa racial issue, yet the lower life expectancy of The ability of voters to advantages ofdictatorship. blacks combined with the fact that they paySocial Security payroll taxes at the same rate asden redistribution from black workers towhite retirees.The subtlety of the connec-tion, combined with collusive politiciansÕ lackof incentive to focus on the issue, leads the rel-problems often prevent an issue public fromorming in the first place. High knowledgecosts combine with collective action problemsto ensure that many potential issue publics arenumbered among Mancur OlsonÕs ÒÔforgottengroupsÕ who suffer in silence.ÓMost fundamentally, voter ignorance ofgeneral issues may vitiate the benefits of issuepublics even in situations in which the issuepublics have fully adequate informationabout their more specific concerns. If eachspecific issue area is controlled by a subset ofthese same subsets remain ignorant of gener-ally applicable issues, the outcome may wellbe a process of mutually destructive rentseeking that leaves each group worse off thanit would have been had there been no issuepublics to begin with. Within its particularbailiwick, each issue public pushes for poli-cies beneficial to itself without regard to thecosts to othersÑcosts of which its membersare ignorant even if self-interest would notlead them to ignore these costs in any case. Awhich the general interest is routinelyneglected in favor of the particular. For thesereasons, it is by no means clear that an elec-torate divided into issue publics is in a betterposition to pursue its policy objectives thanone that is uniformly ignorant across theIf the rationally ignorant portion of theelectorate commits its errors randomly, thepower of aggregation might result in thoseerrors canceling each other out. Where mis-es are truly random and the electorate suf-ciently large, every ÒerroneousÓ vote forCandidate X should be offset by one foropposing Candidate Y. Only the nonran-domly distributed votes of the relativelyinformed voters will have a real impact on theoutcome; that outcome will thereby be decid-ed Òas ifÓ the electorate as a whole wereinformed.It is ironic that this line of argumentshould be put forward by writers committedto developing a defense of ÒmajoritariandemocracyÓ against charges of voter incapaci-ty.en seriously, it implies that the votesÒnoiseÓ obscuring the ÒsignalsÓ sent by theinformed few, as one advocate explicitlystates.If the argument were correct, electionsould have the same outcome if only the bal-lots of the well-informed minority were count-ed! Even in the more moderate version of thetheory, which allows that some of the ill-informed votes nonetheless turn out to beÒcorrectlyÓ cast, a large proportion of the elec-torate is nonetheless viewed as a source of ran-dom errors that fortunately offset each other.egardless, the Òmiracle of aggregationÓdom and (2) the informed minority thatdecides electoral outcomes adequately repre-tion. Overwhelming evidence suggests thatOne of the main reasons why errors arenonrandomly distributed is that voters reallytry to use several of the other informationshortcuts discussed above. As a result, ill-informed voters often draw misleading infer-ences about economic conditions and otherncertainty about a candidateÕs pol-icy stances itself creates a systematic bias infavor of incumbents whose positions are gen-erally better known.The random distribution hypothesis fareslittle better in meeting the second precondi-tion, that of representativeness of theinformed. The small minority of well-informedters (no more than one-fifth of the total andperhaps a lot less) differs systematically fromthe rest in gender, income, race, age, religion,ideology, and a host of other politically rele-vant attributes.It would be remarkable ncertaintyabout a candidateÕs policy stancesitself creates a systematic bias in favor of better known. indeed if the interests of this small, unrepre-sentative subset of the population coincidedeven roughly with those of the population atlarge, and there is little reason to believe thatOverall, the shortcuts to informed votingto voters than their advocates suggest. Inmany instances, they may be actively mis-leading. There is no real substitute for votersadequately informed at the individual level. IV: The Rationality ofrhaps the most fundamental cause ofignorance resides in the collective action prob-lem created by the insignificance of any indi-vidual vote in determining an electoral out-Since one vote is almost certain not tobe decisive, even a voter who cares greatly aboutthe outcome has almost no incentive to investheavily in acquiring sufficient knowledge tomake an informed choice. An informed elec-torate is a Òpublic goodÓ the provision of whichis subject to the Òcollective actionÓ problemthat arises when consumers of a good do nothave to help pay for its provision in order toenjoy its benefits.Only political professionalsand those who value political knowledge for itsn sake have an incentive to acquire signifi-cant amounts of it. Acquiring significantamounts of political knowledge for the pur-pose of becoming a more informed voter is, inmost situations, simply irrational.An important extension of this logic is thatit applies just as readily to highly altruistic andcivic-minded citizens as to narrowly self-inter-ested ones. Even a 100 percent altruistic per-sonÑsomeone who always chooses to priori-tize the welfare of others over her own when-ever the two conflictÑwould not rationallydevote much of her time to acquiring politicalinformation for the sake of casting aninformed vote. No matter how great the bene-its to others of a ÒcorrectÓ electoral outcome,our altruistÕs ballot has almost no chance ofchance that his vote will be decisive is vanish-The rational altruist wouldtherefore seek to serve others in ways in whichchance of making a difference, such as con-tributing to charity. By spending time andeffort on becoming an educated voter, thethersÕ welfare bydepriving them of the services he or she mighthave conferred on them through alternativeuses of the same resources.The applicability of collective action argu-ment to altruistic voters obviates, at least inthis case, one of the standard criticisms ofon unwarranted assumptions of self-interest-ed behavior.ter ignorance rests on no such assumption.The conclusion that even altruists have littleincentive to become informed voters alsorists to improve the functioning of democra-Whatever virtues those proposals mighthave, they seem unlikely to dissipate therational ignorance that stands as a particu-larly imposing obstacle to effective democra-tic control of government.The collective action problem explanationof political ignorance has other importantimplications for efforts to realize the democ-ratic ideal of voter control. Proposals to alle-viate ignorance by increasing the availabilityof political information to the public runafoul of the collective action problems.Moreover, most citizens do not seek outinformation now because they find politicsrelatively uninteresting.A notable shortcoming of the collectiveaction explanation of voter ignorance is thatit seems inconsistent with the fact that voterste at all. As has often been pointed out,predict that all or most voters shouldnÕt evenshow up at the polls, given the infinitesimallikelihood of affecting the electoral outcome.More precisely, they predict nonvoting unlessthe Òduty-basedÓ or ÒexpressiveÓ utility ofting outweighs its costs irrespective of the Even a voter about the almost no incentive toinvest heavily in sufficient knowledge tomake an likelihood of affecting the outcome;is the case, however, it is possible that thesame sense of duty that leads voters to votemay also lead them to become informed.A complete discussion of this issue, theÒparadox of voting,Ó would take us far afield,but there are at least three important reasonsto believe that the paradox does not invalidatea collective action problem explanation ofunexpectedly high incidence of voting is sim-ply the result of people overestimating thepotential impact of their vote.olls showthat more than 70 percent of voters believethat their individual votes Òreally matter.ÓSuch overestimation may in fact be the extent that acquiring an accurate knowl-edge of the impact of voting may for many bemore expensive than the relatively minimaleffort required to vote in major elections. If so,it is not implausible to hypothesize that thedegree of overestimation is great enough tostimulate voting but far too small to stimulatethe much greater investment of time andeffort necessary to acquire a substantialamount of political information.Second, even ifÑas is surely true at least inpartÑthe critics of rational choice are correctand voting really is explainable by theÒexpressive utilityÓ of voting or by irrationalconceptions of duty,empirical evidence suggests very likely) thatsuch motives are not powerful enough toinduce voters to pay the heavy costs ofbecoming well informed. This point dove-ails with our third argument: that none ofthe competitors of the rational ignorancetheory predict the stability of extreme levelslevels. Surely explanations of voter ignorancebased on differing levels of commitment toduty and cultural variation and varyingaccess to information would predict greatervariation in levels of ignorance across timeIn sum, the rational ignorance hypothesisis only an imperfect representation of reality.t to the extent that it has validity, it setssevere limits to the amount of knowledgeordinary voters are likely to be willing toacquire. Any solution to the problem of voterignorance will have to work within these con-straints rather than try to break them.the Size and Scope of The debate over voter ignorance has oftenocused on how much voters know but rarelyon the question of how much governmentthere is for them to know about. Yet it is clearthat the greater the size and scope of govern-ment, the more voters have to know to con-trol its policies through the ballot. To avoidmisunderstandings, it is important toemphasize that the increased democraticcontrol that may be achieved by reducing thesize and complexity of government is not theonly or even the most important factor thatshould be considered in determining theproper role of the state in our society. Otherconsiderations may well outweigh it in par-ticular circumstances. However, the knowl-edge-based tradeoff between big governmentand democratic government is an importantgovernment spending now consumes at leasta third and often more than one-half of GDP.But it is not the size of government per sethat so greatly increases the likelihood ofter ignorance as the extraordinary scope ofgovernment activity. A government thatcommits enormous resources to a narrowrange of readily comprehensible activities isnot necessarily much more difficult for vot-ers to keep track of than one that commitsonly small amounts to them. Yet the growthof government over the last century has beencharacterized by an immense expansion ofthe domain of government power as well asby increased activity in areas of traditionalstate responsibility.That process hasreached the point where areas of social lifethat remain outside the governmentÕs The greater themore voters haveto know to the exception. In the United States the executive branchof the federal government alone has 15 cabi-net-level departments, 54 independent regu-latory agencies and government corpora-tions, and 5 Òquasi-officialÓ agencies.range in function from the United StatesInformation Agency to the Farm CreditAdministration to the National MediationBoard. It is doubtful in the extreme that vot-ers could keep adequate track of all theiractivities even if they paid far more attentionto political information than they do today.Current surveys of political knowledge usual-ly do not even ask about the functions of spe-cific government agencies, instead opting foruestions about very basic aspects of govern-ment structure and opinions about whethergovernment should do ÒmoreÓ in broadlydefined issue areas such as ÒeducationÓ orÒhelping the poor.Ó The omission is in partthe result of researchersÕ reluctance to useuestions that may prove ÒintimidatingÓ toill-informed voters who are ignorant of evenbasic information. The few questions requir-ing more detailed issue knowledge that haveound their way into surveys unsurprisinglyshow even greater levels of ignorance thanthose about more basic information.The Ignorance of ElitesIf traditional research on voter ignoranceuestions the electoral competence of ordi-nary citizens, focusing on the relationship ofignorance to the size and scope of govern-ment leads us to question that of relativelyll-informed elites as well. It is unlikely thateven professional social scientists have morethan a very superficial knowledge of theactivities of government agencies outsidetheir areas of specialized expertise, if indeedthey have heard of them at all. esearchers tend to assume, usually with-out argument, that the best-informed voterscan make effective use of ideology and opin-ion leaders to guide their choices. Yet it is notat all clear that this optimism about elites isjustified. Let us consider the dilemmas thatthe modern state creates for generally well-informed voters who nonetheless remainignorant of specific issues.The ideologically sophisticated voter whois convinced that a government program canaddress some evil must still determinewhether the remedies proposed by candi-are likely to solve the problem at hand.nless voters hold to an ideological positionthat justifies government activism for its ownsake regardless of the results, they must finda way to ensure that the programs put inplace by legislators really do serve the pur-poses they intend. The difficulty is furtherexacerbated in those very common situationsin which voters cannot readily observe a poli-cyÕs effects.Even the people most favorably disposedtoward government activism cannot deny thepossibility of government programs that failin their stated purposes because of unantici-pated consequences or because they are actu-ally driven by rent-seeking interest groups orattempts to redistribute funds to the rich orsome other group whom voters might notwant to reward. Many programs justifiedrhetorically by the need to aid the poor orprovide some public good for the populationat large actually serve well-positioned Òdistri-butional coalitionsÓ at the expense of the veryAnd it is a commonplaceamong specialists that well-intentionedinterventions such as minimum wages ordrug prohibition often produce side effectsthat may be worse than the original problem.The paucity of research on the control ofmodern-sized governments by well-informed,ideologically sophisticated voters prevents anydefinitive conclusions in this area.In part,the lack of research of this type is itself a con-sequence of the growth of government, sincethat growth has greatly increased the propor-notell informedand has diminished the electoral significanceof those few who are.onetheless, there arestrong reasons to believe that even a well-informed electorate, one with an average level have more than avery superficialknowledge of the of knowledge equal to that of the top 5 to 10percent today, would have great difficultyimposing its will on a government as broad inscope as that which we have today. Informational Advantages of LimitedA government of strictly limited powersmight reduce the problem of public igno-rance by reducing the number of issues to bedecided by government to a level that votersould find more manageable. Obviously,issues taken out of the government sphereould have to be ÒdecidedÓ in other ways,whether through market exchanges or othernonpolitical processes. Whatever their otherdefects, however, those mechanisms are rela-tively free of the collective action problemsthat induce massive voter ignorance in theIs there in fact any evidence that limitedgovernment eases the informational burdenon voters and enables them to exercisegreater control than they do today? A defini-tive answer to this question requires muchonetheless, 19th-century American historysuggests that the answer may be yes. Because of the very limited powers of thenational government, 19th-century nationalpolitics revolved around a small set of rela-tively narrowly defined issues, including thespread of slavery, the disposition of newlyacquired western lands, the tariff, federalsupport for infrastructure spending, bank-ing, and, on a few occasions, warfare with for-eign powers. With the exception of obviouserlaps between issues (e.g., slavery andestern lands), rarely were more than two orthree of these matters in contention at anygiven time. The theory developed here predicts thatthis limitation of government power shouldhave allowed voters to focus on the issuesmuch greater detail than has been possiblesince. Although there have as yet been no sys-evidence supports it. During major 19th-cen-tury controversies such as the debates overmonetary policy in the 1830s and 1890s, theongoing battle over the tariff, and the con-frontation over the expansion of slavery inthe 1850s, politicians presented for massconsumption far more sophisticated argu-ments than prevail in electoral politics today. In the case of slavery, the Lincoln-Douglasdebates over slavery expansion, conductedbefore large audiences of ordinary voters,including a substantial proportion of illiterates,addressed in some detail such questions as theeffect of slavery expansion on free labor,whether or not the Supreme CourtÕs interpreta-tion of the Constitution necessarily takes prece-dence over that of other branches of govern-ment, the moral status of blacks in the liberalideology of the Declaration of Independence,and the true meaning of Òpopular sovereign-ty.ÓCampaign speeches of this complexityould be unimaginable today. opular ideological awareness also playedan important role in restraining the growth ofgovernment beyond its prescribed bounds.important force limiting the power of the U.S.ederal government by leading contemporaryoreign observers such as Tocqueville andLord Bryce.More recent theoretical treat-ments of the subject also emphasize the cru-restraining the growth of 19th-centuryAmerican government in the face of interestSuch a high degree of ideo-logical awareness is in clear contrast to whatknow of modern voters. It suggests thatideological coherence in voting may be moreeasible if there are fewer issues to keep trackThis 19th-century evidence by no meansamounts to a strong test of my hypothesisand is of course subject to alternative expla-nations. Even so, it provides a prima faciecase for the proposition that voter knowledgeand control of government policy will beed government power. The relationship between voter ignoranceand big government leads us to question the A government ofstrictly limitedpowers mightproblem of publicnumber of issuesto be decided bygovernment to alevel that voterswould find more adequacy of the information possessed byeven the best-informed voters. It also leads tothe counterintuitive suggestion that theextension of government power to new areasof social life undercuts democratizationrather than furthers it. Democratic control ofgovernment is increased when there is lessgovernment to control. VI: Political Ignorance Federalism: The Informational Benefits of ting with Your FeetÒFoot VotingÓ vs. Ballot Box Voting as anIf rational ignorance suggests that govern-ment may be more democratic when it ismore strictly limited, it may also counsel infavor of greater decentralization. When infor-mation problems are taken into account, vot-ing with your feet in a relatively decentralizedederal system may lead to greater majoritar-ian control of government than ballot-boxting in a more centralized state.As we have seen, one of the main causes ofal.Ó Because even an extremely well-informedter has virtually no chance of actually influ-incentive to become informed in the firstplace, at least if the only purpose of doing so isto cast a ÒcorrectÓ vote. By contrast, a personÒvoting with her feetÓ by choosing the state orlocality in which to live is in a wholly differentsituation than is the ballot-box voter. If a ÒfootterÓ can acquire information about superioreconomic conditions, public policies, andther advantages in another state, he or shecan move to that state and take advantage ofthem even if all other citizens do nothing. That creates a much stronger incentive forters to acquire relevant informationabout conditions in different jurisdictionsthan for ballot-box voters to acquire infor-mation about public policy. Since states andbusinesses as sources of tax revenue, stateand local governments have strong incentivesto establish policies that will appeal to poten-tial immigrants and convince current resi-dents to stay.The power of the competitivepressure comes from governmentsÕ constantneed to attract additional revenue to financeexpenditures that can satisfy key interesttion chances. Interstate and interlocalitycompetition for residents facilitates the cre-interests of the majority, even in the absenceof informed ballot-box voting. In this way,ting with your feet becomes a powerfuland, in many ways, superior alternative toballot-box voting as a mechanism of majorityMoreover, unlike a ballot-box voter, a footter need not connect his judgment of con-ditions to specific elected officials and theirpolicies. It is enough for her to know thatconditions are better in one state than anoth-er and then be able to act on that by moving.So long as public officials themselves knowthat their policies can affect social conditionsin ways that attract foot voters, they will havean incentive to implement better policies inorder to appeal to potential migrants. Notonly does foot voting create a stronger incen-tive to acquire knowledge than ballot-boxting, it also requires less knowledge toimplement effectively.The Power of Foot Voting under Adverseknowledge, there has not yet been astudy that precisely measures the informa-tional advantages of voting with your feeter ballot-box voting. Nevertheless, there isreason to believe that the advantages areite large. Anecdotal evidence and ordinarylife experience suggest that most citizens putfar more effort into deciding where to livethan into acquiring political information. of the power of voting with your feet evenunder extremely adverse circumstances. In the Votiyour feet in a federal systemmay lead togreater majoritarian control of Jim CrowÐera South of the late 19th and early20th centuries, African-American southerners,most of them poorly educated and many illit-erate, were still able to learn enough informa-tion about the existence of relatively betterconditions in other states to set off a massiveSouthern black workers reliedon information provided by relatives in otherjurisdictions and by agents of businesses seek-ing to recruit African-American workers.The resulting migration not only benefitedthe migrants themselves but also forced racistsouthern state governments to Ògrant . . .African-Americans greater educational oppor-tunities and greater protection in their proper-ty and personÓ in an effort to get them to stayand continue to provide labor for southernwhite-owned farms and businesses.related dramatic example, interjurisidictionalcompetition for the labor of migrating blackcoal miners led to successful lobbying by coalcompanies for a reduction in school segrega-tion in West Virginia in the early 1900s.Obviously, the ability of southern blacksto vote with their feet did not come close tofully mitigating the baneful effects of JimCrow.It did, however, provide importantinformational benefits and political empow-cated minority. Although exact comparisonsare difficult, it seems likely that potentialsouthern black migrants of the Jim Crow erarelative conditions in different jurisdictionsthan most modern voters have learned aboutthe basics of our political system. At the verystand that relatively more favorable employ-ment opportunities and public policieswaited them in other jurisdictions, a realiza-modern citizens to acquire sufficient knowl-edge to engage in effective retrospective vot-Obviously, if voting with your feetcould provide powerful informational advan-ages in the exceptionally adverse conditionsof the Jim Crow-era South, there is strongreason to expect that it is more effective inmodern times, when education levels aremuch higher, information costs are lower,oppressed as were poor southern blacks acentury ago.from the only issues that must be consideredin determining the degree of decentralizationthat a society should have. A variety of otherconsiderations may in some situations out-eigh the advantages of foot voting. Theargument advanced here is not intended tobe a comprehensive theory of federalism, oreven close to it. It does, however, highlight animportant consideration that is too oftenI have argued that voter ignorance is wide-spread, that it is a rational result of collectiveaction problems, that it cannot be circum-ented through various Òinformation short-cuts,Ó and that it is greatly exacerbated by thesize and scope of modern government. I havealso argued that voter ignorance highlightssome unrecognized advantages of decentral-ized federalism.Concern for the consequences of voterignorance is not a new idea in Western politi-cal thought. Plato, Aristotle, and John StuartMill, among others, all regarded it as one ofthe most important problems of democra-cy.Modern democratic thought, however,has tended to sweep the problem under therug even as the growth of government hasmade it increasingly severe. Yet ultimately, lib-eral democrats cannot avoid the inherent ten-sion between big government and democracy. say that the tension must be faced isnot to say that it must necessarily be resolvedin favor of democracy. Democracy is not theonly criterion for good government, and per-haps not even the most important one.onetheless, the dilemma is particularlyacute for those who see democratic self-gov-ernment as an important value in its ownEven for those who see democracy as advantages of a purely instrumental valueÑfor instance, fordominance over most policy issues in ademocracy with largely unconstrained gov-ernment power should raise troubling ques-tions. Are our elites really so benevolent andso knowledgeable that they will serve theinterests of the poor and downtrodden in theabsence of effective voter control? The concern here is not the traditional fearopinion to suit their own advantage,although that possibility cannot be ruled out.ply by default. What the voters donÕt knowthis paper is simply to scale down the ambi-tions of democracy from effective voter con-trol of public policy to a mere ability of votersis indeed an important advantage of democ-racy in situations in which incumbent offi-cials have committed a massive and highlyvisible policy error. But while this claim is apoint in favor of unrestricted majoritarian-ism as opposed to dictatorship or oligarchy,it says nothing about the merits of such a sys-tem compared with a democratic govern-ment with strictly limited powers. Moreover,the ability to remove a failed leader is of littleelectorate cannot readily tell whether or not agiven leader has failed.does not give any definitive answer abouthow limited the powers of governmentshould be, even from the narrow standpointof ensuring meaningful voter control overpublic policy. Nonetheless, the depth andapparent intractability of current levels ofcontrol requires a government more limitedin scope than what we have today.A second limitation of my analysis is thelack of comparison with knowledge prob-lems encountered by consumers in private-sector markets. If the latter were as great asthose that exist in the case of government,the argument for limiting government powerould be weakened. A full analysis of thissubject is beyond the scope of this paper, yetthere are powerful theoretical reasons tobelieve that Òmarket ignoranceÓ is both a lessprevalent and a less serious problem thanter ignorance. Among the more importantof those reasons are the absence of collectiveproduct markets, the relative ease of connect-ing product quality to product performancecompared with the great difficulty of linkingsocial outcomes to public policy, the greaterease of finding disinterested advisers on mar-t transactions (e.g., financial advisers, con-sumer guides), and the existence of enforce-able contract and tort remedies againstdeception by product manufacturers of asort that is unavailable against deceptiveThe political realm also lacks a conveyorof information as efficient in providing feed-back and diminishing information costs asthe marketÕs price system.Yet even if themonopolistic nature of the state and thegreat powers concentrated in its handsshould still lead us to place a particularlyhigh value on democratic control over gov-government should be taken more seriouslyin political thought than it has been so far.NotesSome of the material in this paper is adapted fromtwo earlier articles: Ilya Somin, ÒVoter Ignoranceand the Democratic Ideal,Ó Critical Review413Ð58; and Ilya Somin, ÒPolitical Ignorance andthe Countermajoritarian Difficulty,Ó Iowa Law89 (2004): 1287Ð1371.See, e.g., Carole Pateman, Democratic Theory(Cambridge: Cambridgeiversity Press, 1970).2. Anthony Downs, An Economic Theory ofDemocracy(New York: Harper & Row, 1957), chap.3. The discussion in this section is based on a Current levels that democraticcontrol requires agovernment morelimited in scopethan what wehave today. ÒcommonsenseÓ notion of democratic control.Obviously, there are many different theories ofdemocratic control of government, not all ofwhich have the same knowledge prerequisites. Fora detailed discussion of the leading theories show-ing that even the least demanding theoriesimpose substantial knowledge burdens on voters,see Somin, ÒPolitical Ignorance and theCountermajoritarian Difficulty,Ó 1296Ð1304. 4. Angus Campbell et al., The American VoterArbor: University of Michigan Press, 1960), chap. 8.5. See Michael X. Delli Carpini and Scott Keeter,What Americans Know about Politics and Why It(New Haven, CT: Yale University Press,1996) for a survey of the field. For a more recentbut less comprehensive analysis, see ScottCollective Preferences in Democratic Politics(New York: Cambridge University Press, 2003). 6. See David Mayhew, Congress: The Electoral(New Haven, CT: Yale University Press,Jane J. Mansbridge, Beyond Adversary Democracy(Chicago: University of Chicago Press, 1980), p. 25.8. Linda Bennett and Stephen Bennett, Living withLeviathan: Americans Coming to Terms with BigGovernment(Lawrence: University Press of Kansas,1990), chaps. 2, 4.9. For the most influential work making thisclaim, see Benjamin Page and Robert Shapiro, Rational Public(Chicago: University of ChicagoPress, 1992); see also other works cited in Somin,ter Ignorance and the Democratic Ideal,Ó pp.429Ð31. 10. John Ferejohn, ÒInformation and the ElectoralInformation and Democratic Processesed. John Ferejohn and James Kuklinski (Urbana:iversity of Illinois Press, 1990), p. 3.11. For the classic studies of this era, see, e.g.,Campbell et al.; Bernard Berelson et al., (Chicago: University of Chicago Press, 1954); andespecially Philip Converse, ÒThe Nature of Beliefed. David Apter (New York: Free Press 1964).12. Princeton Survey Research Associates Survey,April 15, 2004, Roper Center, Accession no.0451671.13. Pew Research Center Survey, February 11Ð16,2004, Roper Center, Accession no. 0448774.14. Data calculated from the 2002 NationalElection Study variable 025083. Data from the2002 NES are available from the author or fromthe NES website, http://www.umich.edu/~nes/.15. See, e.g., W. Russell Neumann, The Paradox ofMass Politics(Cambridge, MA: Harvard UniversityPress, 1986), pp. 15Ð16; and Stephen E. Bennettand Linda Bennett, ÒOut of Sight Out of Mind:AmericansÕ Knowledge of Party Control of theHouse of Representatives, 1960Ð1984,Ó esearch Quarterly35 (1992): 67Ð81.16. Somin, ÒVoter Ignorance and the DemocraticIdeal,Ó p. 417.17. Page and Shapiro, p. 10. 18. Delli Carpini and Keeter, What Americans Knowabout Politics, p. 94; and Neumann, p. 15.19. See, e.g., Althaus, Collective PreferencesDelli Carpini and Keeter, What Americans Knowabout Politics20. Ibid., pp. 70Ð71.21. Converse, ÒThe Nature of Belief Systems in22. See, e.g., works cited in Somin, ÒVoterIgnorance and the Democratic Ideal,Ó pp. 417Ð18.23. Downs, chap. 7.24. See, e.g., Samuel Popkin, The Reasoning Voter(Chicago: University of Chicago Press, 1991), p.25. The term is borrowed from Delli Carpini andKeeter, What Americans Know about Politics26. Ibid., pp. 234Ð35; Neumann; and DonaldKinder and Lynn Sanders, ÒMimicking PoliticalDebate with Survey Questions: The Case of WhiteOpinion on Affirmative Action for Blacks,Ó 27. For studies showing no change over time, seeDelli Carpini and Keeter, What Americans Knowabout Politics, pp. 62Ð134; Eric R. A. N. Smith, hanging American Voter(Berkeley: University ofCalifornia Press, 1989); Stephen E. Bennett,ÒÕKnow-NothingsÕ Revisited: The Meaning oflitical Ignorance Today,Ó Social Science Quarterly69 (1988): 476Ð92; Stephen E. Bennett, ÒKnow-things Revisited Again,Ó litical Behavior(1996): 219Ð31; Stephen E. Bennett, ÒTrends inAmericansÕ Political Information, 1967Ð87,ÓAmerican Politics Quarterly17 (1989): 422Ð35; andMichael X. Delli Carpini and Scott Keeter,ÒStability and Change in the U.S. PublicÕsKnowledge of Politics,Ó Public Opinion Quarterly(1991): 583Ð96. For a recent exception, see Althaus,p. 215, which shows a very small increase in knowl- 22 edge between the 1980Ð88 period and 1990Ð98.The increase shown in AlthausÕs study is extremelylow (from an average of 52 percent correct answersin the earlier period to 54 percent in the later one)and may be an artifact of the particular questions28. See John Stuart Mill, Considerations onpresentative Government(1861; Indianapolis:29. A total of 1,543 respondents had completedata on answers to all 31 knowledge items.30. Data from the 2000 NES are available fordownloading from the University of MichiganInteruniversity Consortium on Political and Socialsearch, http://www.icpsr.umich.edu. The 2000NES is data set number 3131. A modified versionof the data set with recoded variables for purposesof the present study is available from the author. 31. The surveys used in this table have beenarchived by the Roper Center for Public Opinion.They are available online at Roper Center for PublicOpinion Research, http://web.lexis-nexis.com/unirse/form/academic/s_roper.html. Individualsurveys are identified by accession number. Insome cases, the ÒDonÕt KnowÓ figure includessome respondents who refused to answer. Researchsuggests that it is very rare for respondents whoknow the correct answer to a question to refuse togive it. 32. Many survey respondents will refuse to admitignorance. Indeed, large numbers will express opin-ions about nonexistent legislation rather thanadmit they have never heard of it. For a notoriousexample, see Stanley PayneÕs famous finding that70 percent of respondents expressed opinionsregarding the nonexistent Metallic Metals Act.Stanley Payne, The Art of Asking Questions (Princeton,NJ: Princeton University Press, 1951), p. 18.33. Indeed, the ll Street Journalrecently reportedthat Òdiscussions of a jobless recovery evaporat-edÓ after the Labor Department reported, onApril 2, a gain of 308,000 jobs during the monthof March alone. Aaron Luchetti, ÒBond RallyEnds As Economy, Job Market Spur Fed toMove,Ó ll Street Journal, July 1, 2004, p. C12.34. Pew Research Center Survey, February 11Ð16,2004, Roper Center Accession no. 0448772.35. Pew Research Center Survey, February 11Ð16,2004, Roper Center Accession no. 0448773.36. Exact wording of questions is available fromthe author or can be found in the 2000 NES code-book available for downloading from the ICPSR,http://www.icpsr.umich.edu/index-medium.html. A list of coding changes is available fromthe author upon request. In general, the codingmethodology followed is similar to that used byDelli Carpini and Keeter, What Americans Knowabout Politics, in their analysis of 1988 NES politi-cal knowledge data.37. The possible exceptions are the home states ofLieberman (Wyoming and Connecticut, respec-tively) and possibly being able to name a secondHouse candidate in oneÕs congressional district,especially in cases in which the House race was notclose. Eliminating those three items would notchange any of the results analyzed in this papersignificantly. Moreover, the first two were repeat-edly mentioned in the press during the campaignand likely would have been picked up by anyonewho followed the campaign at all closely.38. These items were the reduction in the federaldeficit in the Clinton years, the reduction incrime, and increased government spending tohelp the poor. 39. Delli Carpini and Keeter, What Americans Knowabout Politics, pp. 138Ð52.40. Data calculated from questions listed inable 1.41. Bennett, ÒÕKnow-NothingsÕ Revisited,Ó p. 483.42. Data calculated from answers to items listedin Table 1.43. These 17 were the 12 items comparing BushÕsand GoreÕs or Democratic and Republican issuepositions, the 3 questions asking about changes inthe crime rate, the deficit, and spending on thepoor in the 1992Ð2000 period, and the 2 questionsregarding identification of BushÕs and PatBuchananÕs ideology (moderate, liberal, or conser-vative). Although some of the questions had morethan three options on the original survey, I col-lapsed them into three for recoding purposes, withuld have had a one-in-three chance of gettingthe correct answer. For the questions regardingBushÕs and GoreÕs ideologies, I gave half credit torespondents who picked Òmoderate,Ó even though,arguably, most knowledgeable observers would notagree with that answer. Al Gore ran an explicitly lib-eral campaign emphasizing the theme of Òthe peo-ple vs. the powerful.Ó See, e.g., John F. Harris andCeci Connolly, ÒShaking Off the Clinton Strategy,oo; With Populist Push, Gore Looks toward aDifferent Group of Swing Voters,Ó Washington Postugust 24, 2000, p. A1. Bush famously describedhimself as a Òcompassionate conservativeÓ andprominently proposed a number of strongly con-servative policies, including a large income tax cut 23 and the privatization of Social Security.44. These were the two questions regarding partycontrol of the House of Representatives andSenate prior to the election.45. I decided to code as correct answers bothÒmoderateÓ and ÒliberalÓ on the question askingthe respondent to identify Bill ClintonÕs ideology. 46. Guessing, albeit with low probabilities of suc-cess, was possible on the questions asking foridentification of the four candidatesÕ home states,Joe LiebermanÕs religion, and the positions heldby Lott, Blair, Reno, and Rehnquist.47. Half points were possible because I allowedhalf credit for certain answers to two questions. 48. I assume that a respondent guessing random-ly would have gotten right, on average, 5.67 of the17 questions with three possible answers, one ofthe two binary questions, 0.66 point on the ques-tion regarding ClintonÕs ideology, and one moreestion from the remaining 10, for a total scoreof 8.33. 49. Bennett, ÒÔKnow-NothingsÕ Revisited,Ó p. 483 50. See Scott L. Althaus, ÒInformation Effects inCollective Preferences,Ó American Political ScienceRev92 (1998): 545Ð46 (noting that moreknowledgeable respondents may be overrepre-sented among those giving opinions on surveys). 51. See calculation above.52. See, e.g., PayneÕs discussion of classic researchon this issue.53. This finding replicates similar results fromearlier research presenting evidence that the mostwidely known facts about politicians are personaltidbits with little real information value. DelliCarpini and Keeter, What Americans Know about, p. 10. For example, the most widely knownfacts about the first President Bush were his dis-ste for broccoli and that he had a dog named54. Those were the home states of Lieberman (30percent correct) and Republican vice presidentialnominee Dick Cheney (19 percent). Ironically,those two items probably had greater informa-tional value than at least two of the three similaruestions that many more respondents answeredcorrectly. LiebermanÕs issue positions were rea-sonably representative of moderately liberalConnecticut, and CheneyÕs conservatism was cer-inly representative of majority political opinionin Wyoming.55. If we also eliminate the Clinton ideology ques-tion, we are left with a slightly lower average scoreof 10.5 out of 24 (44 percent correct).56. Figure calculated using the methodology out-lined in notes above. I continue to assume thatrespondents guessing randomly would get about5.67 correct answers to the 17 questions with threeoptions, 1 point from the two binary questions,and 0.66 point from the Clinton ideology question.However, because the number of other items hasbeen reduced from 10 to 5, I have assumed thatthey would get only 0.5 correct answer from theseitems by guessing rather than 1.0 as in the modelr the 31-item scale. Thus, a total of just undereight predicted correct answers. A total of 35 per-cent of respondents scored eight correct answers orer on the 25-point scale.57. The tide has turned in recent years, however.As a recent literature review points out, there areÒsigns of an emerging consensusÓ that Òthere is alevel of basic knowledge below which the abilityto make a full range of reasoned civic judgmentsis impaired.Ó William A. Galston, ÒPoliticalKnowledge, Political Engagement, and CivicAnnual Review of Political Science(2001): 221. 58. See, e.g., Popkin; and Donald Wittman, Myth of Democratic Failure(Chicago: University ofChicago Press, 1995), chap. 1.59. Downs, chap. 12. 60. Popkin, pp. 23Ð24.61. Ibid., p. 23.62. Morris Fiorina, trospective Voting in American(New Haven, CT: Yale Univer-sity Press, 1981), p. 5. Fiorina has pointed out tome that he would modify this sentence if he couldrewrite it today (personal communication). Ite it nonetheless, not to criticize him, butbecause it represents a widespread view amongscholars and politically minded intellectualsmore generally.63. Thomas Holbrook and James Garand.ÒHomo Economicus? Economic Information andEconomic Voting,Ó litical Research Quarterly(1996): 361.64. Ibid., p. 360.65. Ibid.66. Diana Mutz, ÒDirect and Indirect Routes toliticizing Personal Experience: Does KnowledgeMake a Difference?Ó Public Opinion Quarterly 67. Downs, chaps. 7Ð8; and John H. Aldrich, Why(Chicago: University of Chicago Press,1995), pp. 47Ð49.68. For evidence in support, see, e.g., ibid., pp.69. V. O. Key, The Responsible ElectorateMA: Harvard University Press, 1966); and MorrisFiorina ÒAn Outline for a Model of Party Choice,ÓAmerican Journal of Political Science70. For example, an American voter who had beena mature adult for 20 years as of 1998 would haveexperienced two Democratic and two Republicanpresidential administrations since achievingpolitical awareness.71. See generally Mancur Olson, Collective Action(Cambridge, MA: Harvardiversity Press, 1965).72. Somin, ÒVoter Ignorance and the DemocraticIdeal,Ó p. 423.73. Anti-slavery activists were well aware of the car-tel and routinely denounced the alliance betweenhern and Southern political elites that broughtit aboutÑÒthe lords of the lash and the lords of theloomÓ in Charles SumnerÕs phrase quoted in LouisFiller, The Antislavery Crusade 1830Ð1860(New York:Harper & Row, 1960), p. 244.74. V. O. Key, Southern Politics in State and Nation(New York: Knopf, 1949), chap. 14.75. For representative citations to the extensive lit-erature defending this theory, see Somin, ÒVoterIgnorance and the Democratic Ideal,Ó p. 424.76. W. Russell Neuman, quoted in Popkin, p. 47.77. See the massive analysis in Sidney Verba et al.,ice and Equality(Cambridge, MA: Harvardiversity Press, 1995).78. See works cited in Somin, ÒVoter Ignoranceand the Democratic Ideal,Ó p. 425; the classicanalysis is Converse, ÒThe Nature of BeliefSystems in Mass Publics.Ó79. I have analyzed this problem in greater detail inIlya Somin, ÒResolving the Democratic Dilemma?Óle Journal on Regulation16 (1999): 404Ð11.80. Fiorina, Retrospective Voting, p. 10; see also Key,The Responsible Electorate, pp. 60Ð61.81. See, e.g., D. Roderick Kiewet, (Chicago: University of ChicagoPress, 1983).82. Fiorina, Retrospective Voting, p. 11.83. Ibid., p. 4.84. Converse, ÒThe Nature of Belief Systems in85. Shanto Iyengar, ÒÒShortcuts to PoliticalKnowledge: The Role of Selective Attention andAccessibility,Ó in Information and Democratic, pp. 160Ð85.86. Delli Carpini and Keeter, What Americans Knowabout Politics, pp. 138Ð52.87. Somin, ÒVoter Ignorance and the DemocraticIdeal,Ó p. 429.88. Olson, The Logic of Collective Action, p. 165.89. For arguments in favor of the aggregation theo-, see, e.g., Wittman; Converse, ÒPopular Represen-ationÓ; Page and Shapiro; James Stimson, ÒAMacro Theory of Information Flow,Ó in and Democratic Processes, pp. 345Ð69; and BernardGrofman and Julie Withers, ÒInformation-PoolingModels of Electoral Politics,Ó in Par, ed. Bernard Grofman (AnnArbor: University of Michigan Press, 1993). 90. Page and Shapiro, chap. 10.91. Converse, ÒPopular Representation and theDistribution of Information,Ó in Democratic Processes.92. The phrase is taken from ibid., p. 383.93. Holbrook and Garand; and Mutz.94. R. Michael Alvarez, (Ann Arbor: University of Michigan Press, 1997).95. On gender and racial differences in politicalknowledge, see Somin, ÒPolitical Ignorance and theCountermajoritarian Difficulty,Ó pp. 1354Ð63 andsources cited therein. For other differences, see, e.g.,Delli Carpini and Keeter, What Americans Knowabout PoliticsCollective Preferences96. Downs, chap. 13.97. Olson, The Logic of Collective ActionHardin, Collective Action(Chicago: University ofChicago Press, 1982).98. William H. Riker and Peter Ordeshook, ÒATheory of the Calculus of Voting,Ó litical Science Review 25 99. Empirical evidence provides some support forthis conjecture. Americans spend some 3 percent oftheir income on charity. See Richard B. McKenzie,as It a Decade of Greed?Ó 91Ð96. Judging by survey evidence, that is very like-ly a far greater sum than individual voters spend onacquiring political information. 100. See, e.g., Brian Barry, Democracy, 2d ed. (Chicago: University of ChicagoPress, 1978).101. See, e.g., Michael Sandel, DemocracyÕs Discontent(Cambridge, MA: Harvard University Press, 1996);and Amy Gutmann and Dennis Thompson,Democracy and DisagreementHarvard University Press, 1996).102. For a recent work advocating such proposals,see, e.g., Bruce Ackerman and James L. Fishkin,Deliberation Day(New Haven, CT: Yale UniversityPress, 2004). For citations to earlier literature ofthis type, see Somin, ÒVoter Ignorance and theDemocratic Ideal,Ó p. 437.103. Riker and Ordeshook.104. For a more thorough discussion of the para-dox of voting and its relationship to rationalchoice, see Aldrich, ÒRational Choice andAmerican Journal of Political Science(1973): 246Ð78. AldrichÕs argument that voterturnout is a poor test of collective action theorybecause the expected costs and benefits on eachside are so small is a useful complement to thearguments I develop in the text. In a work largelyignored by political scientists and economists,philosopher Derek Parfit shows how voting may berational even for a fully informed, completely ratio-nal citizen so long as (1) he perceives a substantialdifference between the opposing candidates and(2) he places at least a very small value on the wel-fare of other citizens and not just on his own. SeeDerek Parfit, asons and Persons(Oxford,Clarendon Press), pp. 73Ð75.105. Terry M. Moe, The Organization of Interests(Chicago: University of Chicago Press, 1980), pp.31Ð32. For the explanation to work, it is not nec-essary that all voting be explained by it, merely asubstantial part. 106. Ruy Teixeira, The Disappearing American Voter(Washington: Brookings Institution, 1992), p. 56.107. Irrational, I remind the reader, only in thesense that the individual voterÕs fulfillment of hisduty does not in fact succeed in helping his coun-trymen.108. See, e.g., Robert Higgs, Critical Episodes in the Growth of American Government(New York: Oxford University Press, 1987).109. Figures derived from Department ofCommerce, S. Government Manual 2003Ð2004(Washington: Government Printing Office, 2003).110. Delli Carpini and Keeter, What AmericansKnow about Politics, pp. 91Ð93.111. See Mayhew.112. Mancur Olson, The Rise and Decline of Nations(New Haven, CT: Yale University Press, 1982).113. There is, of course, no shortage of theoreticalmodels of government with perfectly informedters. But they are of limited applicability to sit-uations in which voters are poorly informedabout specific issues but have a strong generalideological awareness.114. This point does not contradict my earlierstatement that the level of ignorance has notdiminished over time. The studies I cited show thatthe amount of knowledge possessed by the averageer has increased only modestly, if at all. A con-stant level of knowledge combined with an increas-ing amount of material that one has to learn to betruly knowledgeable (as a result of the growth ofgovernment) leads to a net increase in ignorancerelative to what one needs to know to vote in aninformed manner. For example, if in 1940,informed voting required a knowledge level of 10,while in 1990 it required a level of 20, then a voterpossessing 5 units of information at both points intime would be relativelymore ignorant in 1990.115. Paul Angle, ed., The Complete Lincoln-Douglas, 2d ed. (Chicago: University of ChicagoPress, 1991); and Harry V. Jaffa, Crisis of the HouseDivided: An Interpretation of the Issues in the Lincoln-(New York: Doubleday, 1959).116. Alexis de Tocqueville, Democracy in Americatrans. George Lawrence (New York: Doubleday,1969), vol. 2, pp. 164Ð65, 262Ð63; and James BryceThe American Commonwealth, 3d ed. (London:Macmillan, 1895), vol. 2, pp. 536Ð37.117. See works cited in Somin, ÒVoter Ignoranceand the Democratic Ideal,Ó p. 435.118. See Thomas R. Dye, American Federalism:Competition among Governments(New York: JohnWiley, 1990), pp. 1Ð33; and Ilya Somin, ÒClosingthe PandoraÕs Box of Federalism: The Case forJudicial Restriction of Federal Subsidies to StateGovernments,Ó Georgetown Law Journal468Ð71.119. See William Cohen, At FreedomÕs Edge: BlackMobility and the Southern White Quest for Racial 26 (Baton Rouge: Louisiana Stateiversity Press, 1991); Florette Henri, BlackMigration: Movement North, 1900Ð1920(New York:Anchor Press, 1975); and David E. Bernstein, ÒTheLaw and Economics of PostÐCivil War Restrictionson Interstate Migration by African-Americans,ÓTexas Law Review120. For a detailed account of these ÒemigrantagentsÓ and their role in providing information tosouthern blacks, see Bernstein, pp. 782Ð83,121. Ibid., p. 784; Henri, p. 75; and Robert Higgs,Competition and Coercion: Blacks in the AmericanEconomy 1865Ð1914(New York: Cambridgeiversity Press, 1977), pp. 29Ð32, 59, 119Ð20,122. Price V. Fishback, ÒCan Competition amongEmployers Reduce Governmental Discrimination?Coal Companies and Segregated Schools in WestVirginia in the Early 1900s,Ó Journal of Law and32 (1989): 324Ð41. For a general discus-sion of the ability of migration to reduce discrimi-nation in education, see Robert A. Margo,ÒSegregated Schools and the Mobility Hypothesis:A Model of Local Government Discrimination,ÓQuarterly Journal of Economics123. It should, however, be noted that its failureto do so was partly attributable to southern stategovernmentsÕ partially successful efforts toreduce black mobility. See Cohen; and Bernstein,pp. 810Ð27.124. See discussion of retrospective voting above.See also Somin, ÒPolitical Ignorance and theCountermajoritarian Difficulty,Ó pp. 1298Ð1300,125. For citations, see Somin, ÒVoter Ignoranceand the Democratic Ideal,Ó p. 444. 126. See, e.g., Pateman; and Benjamin Barber,trong Democracy(Berkeley: University ofCalifornia Press, 1984).127. See, e.g., Benjamin Ginsberg, The Captive Public(New York: Basic Books, 1986).128. For the classic statement of this view, seeJoseph Schumpeter, Capitalism, Socialism, andDemocracy, 3d ed. (New York: Harper & Row,1950), pp. 250Ð302.129. For the classic analysis, see F. A. Hayek, ÒTheUse of Knowledge in Society,Ó 27 OTHER STUDIES IN THE POLICY ANALYSIS SERIESThree Myths about Voter Turnout in the United Statesby John Samples (September 14, 2004)How to Reduce the Cost of Federal Pension Insuranceby Richard A. Ippolito (August 24, 2004)Budget Reforms to Solve New York CityÕs High-Tax Crisisby Raymond J. Keating (August 17, 2004)521.Drug Reimportation: The Free Market Solutionby Roger Pilon (August 4, nderstanding PrivacyÑAnd the Real Threats to Itby Jim Harper (August 4, 2004)clear Deterrence, Preventive War, and Counterproliferationby Jeffrey ecord (July 8, 2004)A Lesson in Waste: Where Does All the Federal Education Money Go?by Neal McCluskey (July 7, 2004)517.Deficits, Interest Rates, and Taxes: Myths and Realities by Alan Reynolds (June 29, 2004)European Union Defense Policy: An American Perspectiveby Leslie S. Lebl (June 24, 2004)Downsizing the Federal Governmentby Chris Edwards (June 2, 2004)Can Tort Reform and Federalism Coexist?by Michael I. Krauss and RobertLevy (April 14, 2004)South AfricaÕs War against Malaria: Lessons for the Developing Worldby Richard Tren and Roger Bate (March 25, 2004)The Syria Accountability Act: Taking the Wrong Road to DamascusClaude Salhani (March 18, 2004)511.Education and Indoctrination in the Muslim World: Is There a Problem? What Can We Do about It? by Andrew Coulson (March 11, 2004)storing the U.S. House of Representatives: A Skeptical Look at Currentby Ronald Keith Gaddie (February 17, 2004)Mrs. Clinton Has Entered the Race: The 2004 Democratic Presidential CandidatesÕ Proposals to Reform Health Insuranceby Michael F. Cannon Published by the Cato Institute, Policy Analysis is a regularfor reform. Nothing in Policy Analysis should be construed asAdditional copies of Policy Analysis are $6.00 each ($3.00each for five or more). To order, or for a complete listing ofve., N.W., Washington, D.C. 20001 or call tolltime). Fax (202) 842-3490 € www.cato.org