Why Policymakers Should Ignore Public Opinion Polls

To view this page ensure that Adobe Flash Player version 9.0.124 or greater is installed.

Please login or register to make a comment!

Policymakers often assume that public opin- ion is a reliable guide to making public policy, but they should not. Public opinion polling mea- sures the wishes and preferences of respondents, neither of which reflect the costs or risks associ- ated with a policy. Public opinion expressed in polls cannot inform policy choice, which requires attention to tradeoffs among values, to second-best possibilities, and to unexpected risks.

Polls are unlikely to be improved enough to help with policy choices. Improvements would make the product (poll results) too expensive or too difficult to obtain from weary respondents. We should not expect to see the day when polling can replace reasoned policy choices by elected representatives of the people.

Despite all the fancy numerology surround- ing modern polling, the extracted advice should not guide public policy. Although public desires for cmore government intervention to help (fill in worthy cause) d are real in that people sincere- ly crave the promised improvement, those cries for government action fail to meet even the most minimal standards of legitimate counsel. This paper shows how little polls tell us about public policy and why we should ignore the proffered guidance to policymakers.

Why Policymakers Should Ignore Public Opinion Polls by ... more. less.

Robert Weissberg _____________________________________________________________________________________________________ Robert Weissberg is a professor of political science at the University of Illinois at Champaign-Urbana. Executive Summary No. 402 May 29, 2001 Introduction: Why Polls Matter Polls are a daily part of our political life, yet their power is a mystery. Polls plainly lack any legal authority. A wise officeholder may prudently consult his or her pollster, but, so far, surveys cannot defeat incumbents or veto legislation. Technically, they can be readily ignored. So, why do polls matter so much, especially to public policy? Keep in mind that the United States did not begin as a direct democracy under majori- ty rule. The Framers of our Constitution hoped to create a constitutional republic, which required constraints on the power of the majority. Such restraints both prevented the tyranny of the majority and promoted the stability of the new regime. The Framers did not doubt that the legitimacy of the American republic lay in the consent of the governed, but they did not ask the people to decide every last detail. They did not expect that the people could or should govern directly. 1 We have come a long way from the Founders 9 balanced, representative democra- cy. Public opinion has achieved a remarkable, though largely unnoticed, ascendancy. The burden of proof is now on those who oppose public opinion. Chalk up a mighty victory for early 20th-century Progressives (and numerous contemporary academics), who argued that the cure for democracy 9s ills is more democracy. Indeed, recent polls suggest that the public has become enamored of its own wisdom: in one 1999 survey, some 80 percent of respondents believed that the nation would be better off if leaders followed public views. 2 If public opinion drives democracy, then what the people think becomes the central question of politics. Polls are powerful because they provide answers to that ques- tion. And their answers are not just opinion but cscience. d Polling methodology has become complex and highly quantitative, an important mark of expertise in an innumer- ate world. Imagine rejoinders to those who doubt the wisdom of polls. The new priestly class of pollsters, like pedants mouthing Latin to befuddle the ignorant, might explain that the data were drawn from a mul- tistage, stratified random sample weighted to capture major SMSA 9s with an oversample of higher SES respondents, or something equally technically abstruse. Who has the self-confidence to question such erudition? That aura of science translates directly into policymaking. As the old saying goes, cIn Washington, good numbers beat bad numbers and bad numbers beat no numbers at all. d As former presidential speechwriter Peggy Noonan artfully explained: cIn every political meeting I have ever been to, if there was a pollster there his work carried the most weight because he was the only one with hard data, with actual numbers on paper. Everyone else had an opinion, the pollster has a fact. d 3 The contest between statistics and hunch is hardly an even battle: the for- mer almost always win. Polls also set boundaries on legitimate policy debates. Each survey result incremen- tally shapes the contemporary Zeitgeist, the ceverybody knows d delineation of cnormal d versus cextreme. d Recall British social scien- tist Walter Bagehot 9s observation from near- ly 150 years ago: cPublic opinion is a perme- ating influence, and it exacts obedience to itself, it requires us to think other men 9s thoughts, and to speak other men 9s words, and to follow other men 9s habits. d 4 As histo- ry shows, today 9s cobvious d remedies may have once been Utopian extremism, and vice versa. Who would have believed a half centu- ry ago that charter schools or privatized Social Security might be creasonable d while state-mandated racial segregation would be unthinkable? A steady barrage of polls, all pointing in a new direction, helped facilitate that shift. Armed with convenient hard data, people who endorse once-lonely causes gain authority and respectability. Against this background, we should not be surprised that Americans look to polls to make tough policy choices, a kind of direct democracy through scientific technique. If 2 If public opinion drives democracy, then what the people think becomes the cen- tral question of politics. Polls are powerful because they provide answers to that question. the vox populi is the voice of wisdom, if not of God, those who hear its words most clear- ly should dominate policymaking. Should we take notice of the polls 9 reports of public demands for more benevolence from Washington? Or, in line with traditional skepticism toward heeding popular outcries, might those demands be dismissible foolish- ness? On the whole, judging by the growing number of commissioned polls and the ris- ing stature of the pollster, one must conclude that reverence for unrestrained majority rule is growing. Moreover the high priests of pub- lic opinion insist that their polls convey legit- imate advice about policies and political strategies. They are wrong. Polls and Public Policy What exactly is a poll supposed to mea- sure? Surprisingly, that question is seldom addressed in the many books on the craft of survey research. An inquiry might assess hopes and aspirations, what Americans want from government or the economy. Or a poll might calibrate political job performance. A survey might also predict future behavior 4 voting intentions or whether parents would send their children to a charter school. Polls might also conveniently reveal hidden behav- ior, for example, campaign donations or riot- ing. A survey can also measure even vague emotions by asking respondents to express likes and dislikes for controversial groups or famous personalities. All of those inquires are easily (and properly) executed via the poll, and none requires especially demanding civic performances by respondents. Queries about policies are another matter. Provided certain modest technical details are satisfied, the door to fantasyland is ajar. The public 9s unbounded cravings can safely be brought to the fore. Provided some client can be found, researchers can literally ask about desiring eternal life; meanwhile the respon- dent is perfectly free to say, cYes, put me down as 8Strongly Agree 9 regarding cheating death. d In other words, accurately assessing popular sentiments down to four decimal places proves nothing about the political ger- maneness of those utterances, despite adroit statistical manipulation. To introduce a modicum of order to cpublic opinion, d consider the following dis- tinctions. When a poll solicits opinions on the evidently unobtainable 4 cfree d health care, a world without pollution, all students exceeding the average 4the results may be called wishes . Nothing commands that those wishes be legal or financially feasible, let alone fulfillable under existing circum- stances. Nevertheless, expressed urges may be exceptionally important politically, regard- less of their imaginary flavor, and a crafty official may use them for political advantage. Unrealistic wants can shape spellbinding rhetoric or lofty programs designed to seduce the unwary. Speechwriters undoubtedly love polls eliciting such aspirations. Nothing (at least technically) forbids asking citizens whether they favor every American boy and girl receiving a world-class education as a pre- lude to campaign promises. We should not, however, conflate shameless pandering with an informed public choice of a deliverable policy. More commonplace in the poll cosmolo- gy are what may be called preferences, wants or desires with some reasonable connection to reality. At a minimum, they are legal, fis- cally doable, and enjoy some leadership sup- port. More mundane prerequisites separat- ing wishes from preferences, for example, would include sufficient technical acumen, properly trained personnel, clear perfor- mance standards, and everything else neces- sary for successful implementation. A poll showing a widespread desire to link federal educational assistance to student test scores reveals a preference. The defining element is that, unlike a wish for universal superior per- formance, preferences could be achieved. In principle, contemporary polls are fully capa- ble of soliciting precise public preferences. Whether preferences are an adequate guide for policy is another story. Consider for a moment the distance 3 The high priests of public opinion insist that their polls convey legit- imate advice about policies and political strategies. They are wrong. between the world of wishes and preferences and the world of a policymaker who must exercise policy choice. A policymaker always must deal with grim reality. Choices are rarely ideal, and second-best outcomes are generally accepted as inevitable. Moreover, every choice must, to the extent feasible, be balanced against every other choice past, pre- sent, and future; policy choice inevitably involves tradeoffs. For example, allocating $100 billion to hire additional teachers may mean defunding other worthy programs, borrowing, or raising taxes. Moreover, people making policy choices must attend to conse- quences, since they are inevitably held accountable by attentive citizens, lobbyists, fellow decisionmakers, and untold others. Policy choices also typically reflect a degree of expertise and are subject to multiple reviews. It is perhaps physically impossible to legislate in complete ignorance, and, if it were attempted (as sometimes does occur), the howls of outrage would be deafening. To be sure, individual legislators might be occa- sionally perplexed, but various institutional mechanisms (e.g., staff, advising bodies) rou- tinely ensure minimal technical know-how. In sum, soliciting the wishes and prefer- ences of the public can be done but is barely relevant to the world of the policymaker. A huge political gap thus separates facile aspira- tions from legitimate policy advice. For lead- ers to conflate wishes with policy choices would be the equivalent of allowing person- alized money printing. If polls are to advise leaders, they must elicit public views about policy choice. Can they do that? The Limits of Polls Contemporary polls tell us almost noth- ing worthwhile about the policy choices fac- ing the nation. Even if we were to believe that America is a nation of philosopher kings, and that every poll is perfectly executed, this heretical judgment still stands. How many polls eliciting public generosity for innumer- able worthy causes actually present respon- dents with a final bill? Every spending choice is independent of any other, a situation at odds with any known political reality. The term ctradeoff d has apparently been ban- ished from the pollsters 9 vocabulary. No Second-Best Choices Policy choices and survey responses inhabit different worlds. Politics, as the old saying goes, is the art of the possible. Seldom are first 4or even second 4choices readily obtainable. Settling for a few cents on the dollar is often the best possible deal. Minimizing losses rather than maximizing gains might even be the superior outcome. Unfortunately, even polls that meet the high- est technical standards forbid respondents to cplay politics d and settle for less than optimal choices. Without such haggling, it is point- less to speak of the public 9s conveying a legit- imate message. At best, polls might uncover collections of first choices, all of which are totally independent of each other. Consider, for example, a philanthropic soul who wishes to spend $10 billion of gov- ernment money to combat AIDS. This is the stark message to the policymaker: one person wants a $10 billion increase. If, however, that cperson d were a legislature with multiple preferences, not an isolated individual, this statement would merely be an opening offer, subject to the negotiation necessary to reach a majority decision. If perchance $10 billion was an excessive amount (or a mean-spirited pittance) according to the philanthropic soul 9s frugal (or philanthropic) colleagues, compromise would be essential. Conceivably, this $10 billion devotee might eventually happily settle for a billion more. Indeed, the initial $10 billion might have served as a clever ploy, an opening move known to be unrealistic and designed to achieve a less ambitious outcome. Government decisionmaking can never be the mechanical aggregation of individual appetites. Decisions require horse-trading and settling for less than the ideal; yet this vital aspect of rational policymaking lies beyond the poll. The political process cannot 4 Contemporary polls tell us almost nothing worthwhile about the policy choices facing the nation. be mimicked by statistically manipulating the data. If the entire sample were physically assembled and told to reach a majority judg- ment on the various survey items, the out- comes would scarcely resemble the first off- the-cuff opinions given in the typical survey. No Tradeoffs Choosing policy requires attention to tradeoffs. Conspicuously absent from most polls is the extractive side of the ledger. The polling format thus differs profoundly from ordering items over the Internet, since adding goodies to one 9s polling shopping cart cannot bump the final bill upward. No Congressional Budget Office annoys benevo- lent interviewees by announcing huge deficits should their gluttony continue. Nor, for that matter, are the lucky respondents in any way obligated to balance their good- heartedness with statutory fiscal limits. The questionnaire thus serves as a credit card with no limits, no interest, no payments until the year 3000; best of all, one 9s credit applica- tion cannot be declined. Choosing policy also requires concrete numbers. Even a superficial glance at typical entitlement questions in polls reveals an indif- ference to concrete figures. That is hardly triv- ial, politically, unless one assumes that any fig- ure is equal to any other. Pollsters typically ask about cmore d or cless d spending and rarely push respondents to offer precise figures. What, for example, might be meant by the wish to spend cmore d money to combat AIDS? Ten million dollars? A billion dollars? A hundred billion? Take the average, even if it includes outlandish sums such as a trillion dollars? Just imagine what would happen if legislators introduced vague bills calling for cmore spending d for highways but cless d for foreign aid. Yet that is what leaders are being cadvised d to do by the oft-repeated polls. The issue is not whether goodhearted citi- zens crave the familiar cfree lunch. d Undoubtedly, the csomething for nothing d mentality exists. More important, contempo- rary polling conventions scarcely ever mention taxes. Imagine what would happen if super- markets decided to remove price tags since this disconcerting information ruins an other- wise enjoyable outing. To further enhance the rapture of shopping, the stores would not draw up a bill. Customers would merely be vaguely told that the bill cwould be paid. d Many would accept this open invitation to go wild in the aisles. Most others, however, would surely grow anxious as their shopping carts filled with expensive merchandise. After a point, most would say, cOkay, I 9ve had my fun, but what is this extravagance going to cost? d Such reckoning rarely occurs in the compas- sion-friendly world created by the poll. Moreover, pollsters rarely get it right when they try to attach costs to public bene- fits. As anyone completing a tax return knows, it is one thing to offer $100 for an attractive nostrum, quite another to add $100 to one 9s existing tax bill . Some context is needed to help taxpayers understand the question. Imagine that the pollster explained that the average American taxpayer already pays about $6,000 annually in federal taxes and that every cspend-more d will increase that burden, and then calculated the gener- ous respondent 9s new total tax burden 4pub- lic benevolence might quickly evaporate. Occasional cost probes raise doubts about pollsters 9 grasp of fiscal matters. For exam- ple, a 1992 Gallup Poll asked if the respon- dent might pay an extra $200 to combat air pollution. With 114 million 1992 tax returns, such a taxpayer gift would generate an extra $22.8 billion, or a fourfold increase in the entire Environmental Protection Agency budget. Furthermore, fighting air pollution is only one EPA responsibility. Imagine an EPA official testifying before Congress and insisting that the agency 9s annual budget be quadrupled, with all the extra funding going only to fight air pollution. It would be embarrassing for everyone, to say the least. 5 Lack of Knowledge Choosing policy requires some knowledge about a subject. Problems begin with the sheer size of government: many citizens grow befud- dled by the costs associated with programs. 5 Contemporary polling conven- tions scarcely ever mention taxes. For example, in my own recent research, I sought to solicit public opinion about two of President Clinton 9s legislative proposals: gov- ernment subsides for local education and child care. Within each of those large fields, I zeroed in on two narrower legislative propos- als ostensibly promising government-created progress: assistance to reduce classroom size by hiring more teachers and a multifaceted child-care assistance plan. Those propositions had gained entrance to the legislative arena plus extensive media coverage. The purpose of the survey was to elicit cpolicy choices d from respondents rather than mere wishes or pref- erences. The instrument was expressly tilted toward better-educated respondents. Except for the overrepresentation of the better edu- cated, the sample was a virtual mirror image of the population with regard to sex, race, age, and region. The sample was thus entirely national. Questionnaire data on political pro- clivity 4partisan affiliation, 1996 vote, and ide- ological viewpoint 4also display a close simi- larity to standard accounts. 6 After the survey was completed, I asked interviewers to characterize how respondents wrestled with exceedingly large dollar figures. The interviewers repeatedly commented that confusion over million versus billion was commonplace, and the actual data similarly suggested unfamiliarity with terms like ca hundred billion. d For all practical purposes, with this format, the total requested cnation- al budget d depends largely on the number of enticements. Conceivably, the patient inter- viewee could say cspend more d on hundreds of worthy ventures, all devised by compas- sionate investigators ever-attentive to causes needing assistance. Polls revealing widespread public igno- rance of everything from constitutional rights to elementary geography have become staples. Public opinion texts establish that awareness drops dramatically as probes shift from prominent personalities to the details of pub- lic policy. 7 Even devotees of poll-driven democracy acknowledge those deficiencies. This demonstrated insufficiency is forgot- ten when citizens pronounce on complex social welfare quandaries. What permits poll- sters to confidently assert both that citizens scarcely can navigate Social Security arcana, that most citizens seek greater program funding, and that this mandate is worth heeding? One unsatisfactory solution is to accept citizens 9 assurances that they are, indeed, competent to pronounce. In the esteemed National Elections Studies con- ducted by the University of Michigan 9s Survey Research Center, a modicum of com- petence is imputed to those passing a filter question regarding mere interest in the topic. Less common is simply accepting the respondent 9s assurances. One Social Security question repeated 16 times between 1973 and 1994 by two highly renowned firms was, cIn general, how well informed are you about Social Security 4that is the benefits available, who is and is not covered, and so on? Would you say you are very well informed, fairly well informed, not too well informed, or not at all informed d? In 1994 some 16 percent of respondents claimed to be cvery well informed d and another 44 percent said they were cfairly well informed. d Significantly, no snap quiz ensued to verify that self-classifica- tion, an easily executable task. An especially pernicious problem is public ignorance about existing programs. Let 9s start with dollar figures. When assessing public generosity or cheapness the pollster never says something like, cAt the end of fis- cal 1996 the Old Age and Survivors Insurance trust fund component of Social Security had an asset balance of $514 bil- lion. 8 How many more dollars would you like to add? d Pollsters always assume in these endless cmore/same/less d questions that respondents possess a firm grasp of the fiscal status quo. That is demonstrated nonsense. Most people are clueless about federal out- lays by program, even judged by the most generous standards. To see how this misinformation can play havoc, consider what various (hypothetical) health care desires might have meant in 1996 when the federal government spent $336.6 billion on health care. 9 Keep in mind that the 6 Polls revealing widespread public ignorance of everything from constitutional rights to elemen- tary geography have become staples. poll does not explain existing spending levels, let alone category allocations. Ms. A, who endlessly frets over her well-being, erroneous- ly believes that the federal government dis- penses only $100 billion on medical care and, since she wants that sum tripled, pronounces cincrease d to the pollster. By contrast, Mr. B is terrified by soaring taxes and wrongly believes that Washington is wasting a trillion dollars on quackery. He prefers that the sum be cut in half to $500 billion; so he advises, creduce. d In this plausible example, our cincrease spending d respondent desires a funding reduction and our budget slasher demands a hefty jump. Without precise fig- ures, anything and everything is possible. It is impossible to discern an intelligible message about increases or decreases unless the baseline is known, and polls almost never provide such critical data. Confusion is inevitable about the mean- ing of chealth care, d cassisting education, d chelping the homeless, d and the like. Typical questionnaire items assume universal under- standing of complicated entities. An especial- ly vexing expenditure category is cmilitary spending. d No doubt, B-2 bombers, sub- marines, and the like immediately spring to mind when the cmilitary spending d question arises. Yet, as defenders of the military 9s bud- get will correctly argue, the modern military encompasses far more than weaponry. The services have evolved into significant social welfare instruments providing education, day care, health care, retirement income, and other social ministrations normally associat- ed with Health and Human Services and other social welfare agencies. Do foes of mili- tary spending wish to abolish hundreds of Army day care centers? Slash military pen- sions? We cannot know if we simply ask about cmilitary spending. d This example is hardly atypical. Imagine a hardheaded questionnaire writer seeking a policy choice from the public about enhanced federal funding for education. He or she would reject the endless parade of cshould the federal government do more (or less, or the same) to assist education? d Why? Put bluntly, polls here are asking citizens to buy high-priced pigs in murky pokes. The consumer equivalent would be an automo- bile salesperson who offered ca car d without divulging details or even price. If shoppers asked the salesperson for details, the response would be, cJust try to do the best you can. d No business would survive with such generic practices, yet this is what poll- sters offer the public. Pollsters seldom mention public befud- dlement about modern policymaking. Such confusion is hardly a badge of civic shame. How many political scientists, for example, could explain Medicare if suddenly interro- gated? Most would demand a few days for further study (and maybe a grant too), and even ordinary citizens might rise to the occa- sion if patiently tutored. The innate capacity of citizens is not the issue here. Modern polling can give us back only what citizens know the moment the phone rings. No Risks An awareness of the risks associated with policy alternatives should inform policy choice. Obviously, even motherhood nos- trums can turn sour. The phrase curban renewal d (and myriad other anti-poverty fail- ures) should instantly conjure up the right image on this point. If poll results are to guide sensibly, those proffering advice should first be quizzed to see if they grasp the attendant risks. Who wants advice from Pollyanna? Alas, this vital measure is virtual- ly unknown in contemporary surveys. No interviewer prudently inquires, cWould you still support aiding the homeless if assistance made matters worse? d Pollsters, evidently, live in a Shangri-La where runaway entitle- ment programs and similar predictable poli- cy nightmares are unknown. This demonstrated insufficiency reveals nothing about cognitive talent. Each of the Pollyannaish respondents could doubtlessly wax eloquent about the risks of, say, living in Newark, New Jersey, or buying a recondi- tioned Pinto. However, such caution is high- ly unlikely to inform responses to poll ques- tions about social welfare. If no attention is 7 Modern polling can give us back only what citizens know the moment the phone rings. paid to risk, the policy advice of the polled should be treated skeptically. Americans are hardly dunces, but we can- not expect informed advice when people are suddenly confronted with policy choices that frustrate experts. Thousands of Social Security experts cannot agree on whether increased funding is the answer, so why should we expect surprised telephone inter- viewees to be any wiser? It is not a matter of democracy unless one defines this term in the most mechanical, plebiscitory way. The debate centers on who is qualified to render recommendations on exceedingly difficult choices. Does the nation benefit when policy choices follow mere preferences and wishes? Policy choice by polls seems to preclude poli- cies guided by informed choice. Improving the Vox Populi? Current poll methodology is clearly inca- pable of extracting sound policy counsel. That inability does not, of course, end the matter: improvement is always possible. Might there be cause for optimism? Hardly. The plebiscitory future looks bleak, regardless of heartfelt intent or prodigious investment. The conventional poll is inherently unsuited to making policy choices regardless of expert claims to the con- trary. Moreover, all the proffered cnew and improved d possibilities, such as deliberative polling, or untold electronic variants are proba- bly even less adequate. The culprits are not the familiar bugaboos of interviewee honesty, loaded questions, shoddy sampling, and all the other well-examined technical impediments. Even if those obstacles were conquered, the bar- riers would remain formidable. If judicious pol- icymaking is the objective, the representative leg- islative assembly, not the sample, is the appro- priate model. A poll 4regardless of how bril- liantly executed 4can never mimic a legislature. The Economics of Polling To appreciate why surveys cannot trans- form public utterances into sage policy coun- sel, the place to begin is polling industry eco- nomics. All survey organizations (including academic ones) must monitor the bottom line. Getting the public 9s two cents is expen- sive, and going beyond cquick-and-dirty d polling may be prohibitively uneconomical. Though modern technology (especially the telephone) has sharply reduced costs, even the most perfunctory technically acceptable study exceeds $20,000. The price tag for a quality poll, one with lengthy face-to-face interrogations conducted by specially trained interviewers, can easily exceed $100,000. 10 Even at that price most of the questions would be imperfect off-the-shelf items. Developing a fresh survey cosmology of orig- inal questions would probably exceed that sum considerably. For most customers, an extensive and expensive poll offers poor return on the addi- tional outlay. It would be as if General Motors sold a superquality, hand-made $95,000 Chevy to compete with a $20,000 assembly-line version that was nearly as good. Since the mass media can only spend so much per poll, why sell a gold-plated, vast- ly superior product to an indifferent public? Nor will anyone care that the vox populi now speaks brilliantly. The bottom line is, indeed, the bottom line. Given that few polling industry executives express serious reserva- tions about product quality, that nobody sues on behalf of those harmed by defective polls, and that money is always tight, the incentive for peeking beyond crude shadows is virtually nil. This frugality results in a pervasive dumb- ing down of the entire enterprise. The typical telephone solicitation virtually precludes conveying information indispensable to ren- dering an informed judgment. Hugely com- plex issues become catch phrases, so even advice from a philosopher king would be gar- bled. Disputes over scientific research agen- das, hospital construction, tax deductions, drug patent protection, subsidized doctor training, patient rights, and insurance regu- lation are all collapsed into cgovernment assistance for medical care. d Even if vital information was dutifully 8 Policy choice by polls seems to preclude policies guided by informed choice. communicated to respondents, today 9s tele- phone poll is unlikely to engender height- ened sophistication. The telephone format is inherently unsuited to conveying prodigious, unfamiliar detail on subjects boring to most respondents. How many respondents can patiently listen as interviewers drone on about essential technical details? Who can accurately recollect it all after the first few minutes? What if the respondent quite cor- rectly says, cThis is so momentous, I 9d like to think about it for a few days, get some addi- tional information, and discuss it with others more expert than myself d? Surveys that impose heavy information burdens on respondents will surely depress already low participation rates even further, and one might assume that those who did participate would hardly be typical. Getting beyond vacuous inquiries also requires prodigious homework for question- naire writers, another expense that does not necessarily yield a more marketable product. Oscar Wilde once quipped that socialism would never work since it occupied too many evenings. Ditto for intelligent poll queries: they demand too much effort. If pollsters want to improve themselves, they will acquire more technical skills, not investigate the issues they ask about. My own research men- tioned earlier dealt with two specific policies (hiring more teachers and subsidized day care) and required more than 100 hours dig- ging up arcane documentation before the questionnaire was constructed. Policy exper- tise is most likely an unaffordable luxury for today 9s pollsters. Second-Best Choices A series of polling experiments dealing with racial integration, conducted during the 1950s, shows the possibility of ascertaining nonoptimal preferences. 11 The researchers 9 aim was not to uncover each respondent 9s most favored position (the nominal goal of the traditional survey) but to classify views into finely differentiated arrays of coppose, d cindifferent, d or cfavor. d For example, some- body fervently wanting unqualified integra- tion might, nevertheless, also be willing to accept integration of only public facilities and housing. The political relevance of ascer- taining second- or third-best desires should be obvious. It is here 4in the realm of the less than ideal 4that real-world politics typically transpires. One study collected 114 statements drawn from real-world discussions (as opposed to the investigator 9s imagination) of integration. Each study participant then sorted every statement into respondent- defined piles ranging from the most to the least preferred. There were large differences in how each statement was perceived. The responses of strongly pro 3civil rights African- American students and anti-integration whites, seemingly distinct groups, were lumped together into a few categories. Interestingly, black students found few state- ments even minimally acceptable and insist- ed on keeping with their elemental group- ings even when specifically asked to expand their classification. Others saw far more pre- cise gradations. 12 Consider how this approach might apply to other issues. Instead of the global cmore/less/same d federal assistance probe, respondents to questions about health care would sort propositions drawn from ongo- ing debates, many of which were gradients of similar nostrums (for example, modifying tax deductions for prescription drugs by $200 increments). Each person could con- struct a rather personalized wish list, but now those wants would include suboptimal desires. Such a subtle procedure might sug- gest to researchers complex (and more use- ful) conclusions, such as, cWhile a handful of respondents wants to expand the Medicare prescription deductible by $500 and an equal number seek to abolish it altogether, most Americans can live with a $50 to $60 reduc- tion in the deductible. d Similarly nuanced characterizations would apply to other key policy details. The range of publicly caccept- able d options might encompass everything on the legislative table though, to be sure, some picks would be judged superior to oth- 9 Policy expertise is most likely an unaffordable lux- ury for today 9s pollsters. ers. In this way, public opinion might inform a politics that would truly be the art of the possible. Unfortunately, this potentially illuminat- ing technique has disappeared into the attic right next to slide rules and Bomar Brain cal- culators. Here, again, the costs of producing and administering questions would be high. Equally troublesome is the bulkiness of the presentation to both the interviewee and the consumer. The valuable richness afforded respondents and researchers would hardly garner media attention 4no small commer- cial consideration. 13 Tradeoffs Forcing respondents to make the most elementary tradeoffs is even harder than pre- senting second-best choices. What is effort- lessly executed when marching down super- market aisles becomes a nightmare on the telephone. Congress itself navigates tradeoffs across hundreds of policies only with diffi- culty, and it too would run wild save for the constraints imposed by budgets. Yet, without the multiple cbutter versus guns d dilemmas, a poll merely affords welfare gourmands a lavish buffet. The most important thing is to impose opportunity costs instead of proffer- ing an enticement parade. The sheer number of competing choices that must be executed if the final outcome is to be judged realistic is a particularly serious obstacle. Even the simplest budgetary classi- fication entails more than a dozen categories, and one can only fantasize how this exercise could be cogently presented to unsophisti- cated interviewees. Even if this Herculean labor can somehow be accomplished, how are we to interpret the final outcomes? What if, as seems likely, the allocations contravened statutory obliga- tions or radically shifted commitments in nonsensical ways? Again, the public 9s unfa- miliarity with existing policy (including exist- ing legal pledges) rears its ugly head. Making laborious tradeoffs demands a degree of pro- ficiency, but proficiency, by itself, hardly guarantees wise counsel. After all, a child spending $10 at the supermarket will surely be highly constrained, and, in his or her own way, will render hierarchical outcomes. Yet, equally likely, the purchases will be nutrition- ally puny. How do we appraise a public cman- date d for spending, say, $500 billion on health care, $500 billion on education, and $100 million on defense? Must the interview- er intercede with, cYou just can 9t reduce the defense budget to $100 million, so let 9s get real. d Try explaining to those cexperts d that choices routinely lie only at the margin and that certain minimums are currently inescapable. The atomistic poll also escapes the formi- dable aggregation problem implicit in major- ity rule. 14 Conceivably, a cwhat do you want? d poll might find the desire for expanding health care spending scattered about from a few million to hundreds of billions of dollars. That diversity might gratify the pollster as the most exact picture possible, but stan- dards of majority agreement remain totally unsatisfied. How is this to be achieved? In legislatures, the answer is simple (at least in principle): bargaining and horse-trading. Unfortunately, that is physically impossible unless, of course, the telephone interviewer adds, cHere are the other 900 survey partici- pant names; and why don 9t you contact them and see if you can work together at reaching a majority? d Naturally, a majority can be manufactured by the researcher either through the design of the initial instrument (e.g., allowing only two choices) or with postinterview statistical com- putations. In a pinch, the median might be presented as a cpublic mandate. d Though acceptable to today 9s conventions, this tactic is but a deus ex machina. Collective decisions are rarely derived by mechanically aggregating iso- lated individual first choices, especially when deals must be struck across multiple policy domains. Differences in priorities, intensities, negotiating skills, and other pertinent ele- ments could, conceivably, yield a final out- come that was disagreeable to everyone and yet gained a majority. The public mandate is not the sum of its individual parts, and can 10 The public man- date is not the sum of its individ- ual parts. never be, the pollster 9s democratic rhapsodiz- ing not withstanding. The perplexities awaiting those seeking to impose even minimal tradeoff discipline are truly horrendous. During the 1970s a bevy of investigators sought to conquer this predica- ment with an experimental device called a cbudget pie. d The exercise appeared simple enough: Participants received a fixed amount of play money and then were asked to make allocations across sundry government ser- vices. In one particularly realistic exercise, borrowing by going into debt or cutting taxes by returning poker chips to the investigator was possible. Ample opportunities existed to assist befuddled respondents. Experiments entailed small groups, physically assembled, dealing with tangible dollars (albeit of the play variety), all under the researcher 9s watch- ful eye. Policy categories were typically three to five, nobody was rushed, and expenditures were for such humdrum services as fire and police protection. Nevertheless, the budget pie as a method has virtually vanished. Ample simplification, investigator helpfulness, and all the rest proved insufficient to inspire many respon- dents to mimic grocery shoppers. In retro- spect, that is hardly surprising 4even a bril- liant supermarket maestro is probably clue- less when asked about police protection vis- à-vis highway construction vis-à-vis educa- tion. Quite likely, if that consumer joined the city council, he or she would learn the ropes, but, without realistic training (and inescapable institutional constraints), carv- ing up a municipal budget is perplexing. Equally predictable, even modest proficiency was powerfully linked to education and social class. One study reported that only about half the sample of low-income respon- dents could navigate a three-part budget pie dealing with policing. 15 Willingness to Pay The problem of tradeoffs has another dimension. Consider the supermarket again. By definition, consumers are willing to pay for what they take to the cash register. Their shopping carts are filled after a series of hard choices, and their selections are valuable for society precisely because they reflect trade- offs. The willingness to pay for groceries reveals a real choice, rather than a simple wish. Similarly, a willingness to pay for a pol- icy bespeaks serious advice about policy. Can polls find out if citizens are willing to pay for their chosen policies? There are many obsta- cles to that goal. First, ascertaining true policy costs is dif- ficult. Entitlements are notoriously suscepti- ble to vagaries in demography, technology, immigration, economic circumstances, and untold other uncontrollable factors. Few entitlements turn out to be cheaper than originally forecast. How do we interpret polls indicating a sincere willingness to pay for underpriced benefits? Equally well-known are nostrums whose true costs are conscious- ly underestimated to garner legislative sup- port. President Clinton 9s education plan, it will be recalled, ignored essential construc- tion outlays and teacher-hiring incentives. Pricing entitlements is part of the political conflict, and it may be unwise to expect poll- sters to navigate this quandary. A more philosophical reservation con- cerns exactly what is meant by respondent generosity. Looming over all discussions of willingness to pay (or WTP, as it is common- ly abbreviated in the economics literature) is the assumption that fiscal earnestness is valid on its face, that is, agreeing to pay means agreeing to pay. But not every charita- ble pledge is honored, and interviewers do not collect taxes. Stripped of consequentiali- ty, endless cspend more d responses may rep- resent only some vague cdo something about the problem d sentiment. The most impressive attempt to measure citizen willingness to fund a desired policy goes by the name of contingent valuation method (CVM). The common feature of CVM projects is attention to a single project 9s minute details prior to presenting the bill. Respondents may receive information about prevailing outlays, various funding strata- gems, and available substitutes and similar 11 How do we inter- pret polls indicat- ing a sincere will- ingness to pay for underpriced ben- efits? itemized data needed for a well-formed choice. Respondents may also learn the risks of a policy (including increased costs as a result of forgoing expenditures entirely) and might grade the status quo in terms of its minimal acceptability. One especially detailed investigation of cleaning up water- ways for recreational use even painstakingly reviewed how this legislative goal was to be accomplished. 16 The CVM technique tends to be exceed- ingly demanding. Participants in the water quality study, for example, not only were made aware of current dollar outlays; the costs of possible improvements were also pre- sented in terms of their personal added tax burden. Even more remarkable, this individ- ualized approach was also applied to other government functions. Now everyone knew how a water improvement project would affect spending for police protection or high- way construction (among other services). Work sheets offered incremental expenditure combinations, each yielding a unique bene- fit-to-cost ratio. The final WTP figure was arrived at slowly through a method akin to completing a complicated tax return. Choices could be altered as new information arose, and the cno tax increase d response was always available. 17 The greatest strength of CVM 4its atten- tiveness to realistic details and rigor 4is also its greatest deficiency. CVM is exceedingly uninviting to contemporary, cost-conscious practitioners. Like budget pies, the apparatus translates poorly into the telephone format. CVM would drive down survey participation rates (already low), and those who did partic- ipate would not be representative Americans. How many erstwhile citizen-advisers have the patience required to follow all the laborious instructions, let alone repetitive paper-and- pencil computational exercises? Moreover, the technique is executed one issue at a time, so each issue requires a fresh survey. 18 Educating Respondents Experience with adding information to polls is not encouraging. Citizens are gener- ally uninterested in being better informed, boredom quickly replaces initial enthusiasm, and attracting fresh voices (e.g., the poor, excluded minorities) is a severe challenge. 19 That nearly all poll enhancement efforts have vanished despite sponsor enthusiasm and ample resources speaks loudly about the futility of this crusade. One might also note the difficulty of boiling down immense issues to digestible snippets. A five-minute talk on government-subsidized medical care might quadruple public wisdom, but would that improvement be consequential, given the issue 9s true complexity? Furthermore, who will guarantee the fairness of the presen- tation, assuming that it is possible to con- struct a balanced presentation Moreover, as any policy expert will attest, debate often cen- ters on the information itself or on expert projections. Who is to say that cexpert d analyses are as neutral as claimed? Far more consequential is the public 9s reluc- tance to grasp policy intricacies. The possibili- ty of teaching the public is always assumed but never demonstrated. Even if all the necessary information could be transmitted to respon- dents willing to be tutored, that would only set the stage for wise counsel. It cannot be presup- posed that fresh knowledge can be fashioned into an intelligent judgment. Consider, for example, the ubiquitous matter of government health care assistance. Obviously, the concrete choices are not between government aid and no aid across dozens of social welfare policies. Intervention is a foregone conclusion. The debate transpires at the extreme margin; for example, how tax deductions for prescription drugs or government reimbursements for exot- ic medical procedures are to be treated. Documenting cravings for cmore d help is irrel- evant, given that every advanced nostrum, technically, uses government authority to improve health care. This immensely compli- cated subject would have to be communicated in an upgraded survey. Can citizens ever navigate these abstruse issues, even if the communication obstacle is overcome and patient respondents receive expert guidance? Probably not, sad to say. 12 Experience with adding informa- tion to polls is not encouraging. Pollsters are unlikely to succeed where schools (even colleges) routinely fall short. Why should people bewildered by endlessly repeated elementary civics lessons suddenly master the connection between patent law protection and research funding? Can ordi- nary citizens appreciate the nonobvious fact that huge tax credits assist the rich, not the destitute, even when fully explained? To expect sudden curiosity and attentiveness to public affairs is unrealistic. Ironically, those academics so casually optimistic about upgrading ordinary citizen awareness often despair at their students 9 (often at elite schools) underwhelming ability to grasp sophisticated policy analysis. Risks Conveying risks to ordinary citizens via a poll is another huge stumbling block. Superficially, the conventional poll appears adequate to this assignment 4one might, for example, offer odds with every policy choice. A question about expanding Medicare might now include a warning that there is a one-in- three chance that this generosity might be more costly than anticipated, or a one-in-five chance that fraud will burgeon. This is hard- ly inconsequential since the mere mention of risk inevitably shapes results. For example, a 1999 ABC News poll 20 asked about sending troops to Kosovo if the air campaign failed. Though 57 percent of respondents endorsed this proposal as stated, endorsements fell to 44 percent when the risk of csome casualties d was introduced. When the risk rose to ca thousand casualties, d endorsement plum- meted to 26 percent. Though seemingly effortless, this cadd-a- risk d element is arduous in the customary sur- vey. Two obstacles immediately come to mind. The first is technical: establishing the precise odds for any outcome. This is a political mine- field, to say the least, and using any one set of plausible figures instead of another might dra- matically alter public preferences. Just imagine respondents 9 being asked, cWould you still support government prescription drug price controls even if there was a reasonable 4say a one-in-three 4chance that this would reduce new drug development? d Who can say if this assessment is accurate, or what creasonable d should signify? No doubt, proponents of price controls would find this wording and risk esti- mate objectionable and dismiss the results as rubbish. The vox populi would soon degener- ate into a Tower of Babble as each interest group sponsored polls with its own creason- able d risk assessment. A more vexing problem is the public 9s ability to comprehend risk that exceedingly rare, though highly momentous, events may occur. Innumerable policy choices 4trans- portation safety, pollution toxicity, medical risk, crime victimization, and even gun acci- dent probabilities 4exhibit this trait. After all, pursuing perfection in safety or cleanliness entails nothing more than moving from cextremely unlikely d to chighly improba- ble. d 21 Given that most people have trouble grasping large numbers, the task of address- ing risk in polls appears hopeless. Can polls be improved so that they provide a reliable guide for policymakers? The answer is no. The economics of polling would not support the extensive measures that would be needed to make polls worthwhile for policy choice. If the money were available, partici- pants would not be: the demands on respon- dents would be immense. Consequently, polling is not likely to provide useful informa- tion to policymakers any time soon. Conclusion This analysis suggests that contemporary polls are seducing respondents, not offering them hard choices of the type faced by legis- latures or policy analysts. Given the typical survey 9s inattention to costs, indifference to risk, and other shortcomings, it is a miracle that polls do not find unanimous support for more social spending. Polls do not pro- vide worthwhile advice about policy; they measure only wishes for a world of benefits with no costs. Polling has crossed the line between 13 The economics of polling would not support the extensive mea- sures that would be needed to make polls worthwhile for policy choice. mechanically recording popular sentiments and becoming a political player. The ques- tion, cWhom do you admire most? d is absolutely harmless; by contrast, cShould the federal government spend more on the homeless? d can be highly mischievous. Assuming that two-thirds of the public will endorse this benevolent outreach, a fresh cfact d is created 4the public desires some- thing, and since democracy means heeding the vox populi, let 9s act! Should this cfact d be regularly publicized, the pollster has brought into being ca consensus d that will surely attract opportunistic office seekers, and those who caution restraint will now be on the defensive. 22 If contemporary polls are poor guides to policymaking and we have no reason to believe they can be improved, where do we go from here? Must we surrender to those poll- sters ever willing to seduce the public with appealing nostrums that quickly become cprograms d to opportunist office seekers? A successful battle against facile entreat- ments must address the way polls are used, not the surveys themselves. Absolutely noth- ing can impede the issuance of unreflective cravings, but this analysis challenges their standing as cwise democratic counsel. d 23 Abstract cravings for public largesse should be treated as cinteresting curiosities d; under no circumstances should they inform policy- making or determine policy choices. Notes 1. James Madison thus counterposes pure democ- racy and representative government: cThe two great points of difference between a Democracy and a Republic are, first, the delegation of the Government, in the latter, to a small number of citi- zens elected by the rest: secondly, the greater number of citizens, and greater sphere of country, over which the latter may be extended. d Madison believed the new American republic was a representative govern- ment. He believed representative government was less likely than direct democracy to decline into fac- tions and civil war. See James Madison, Federalist no. 10 in The Founders 9Constitution, ed. Philip B. Kurland and Ralph Lerner (Chicago: University of Chicago Press, 1988), vol. 1, p. 130. 2. See Steven Kull, Expecting More Say: The American Public on Its Role in Government (Washington: Center on Policy Attitudes, 1999). 3. Peggy Noonan, What I Saw at the Revolution: A Political Life in the Reagan Era (New York: Random House, 1990), p. 249. 4. Walter Bagehot, National Review, July 1856, from Bartlett 9s Familiar Quotations (Boston: Little Brown, 1982), p. 597. 5. Such inattentiveness to creality d is not rare. One careful study of 51 questions dealing with the Panama Canal Treaty found that 10 of the questions contained factual errors such as wrong- ly characterized treaty provisions or misstated historical details. Unfortunately, results from the flawed questions found their way into congres- sional debates over ratification. See Ted J. Smith III and J. Michael Hogan, cPublic Opinion and the Panama Canal Treaties of 1977, d Public Opinion Quarterly 51, no. 1 (Spring 1987): 5 330. 6. The tilt toward education in the sample is clear: 14 percent of respondents had postgradu- ate degrees compared to 7 percent of the general population. By contrast, a mere 5 percent of those interviewed had some high school or less versus 18 percent of the general population. All inter- views were conducted by Angus Reid Associates of Winnipeg, Manitoba, Canada. For more details about this study, see Robert Weissberg, cVoracious Appetites: Public Opinion and Big Government, d University of Illinois, 2000. 7. A useful compilation of such ignorance is Michael X. Delli Carpini and Scott Keeter, What Americans Know about Politics and Why It Matters (New Haven, Conn.: Yale University Press, 1996). Chapter 2 in par- ticular shows widespread unfamiliarity with the fed- eral government 9s fiscal commitments, legislative initiatives, and elementary demographic informa- tion. To be sure, on some items (e.g., the passage of a new minimum wage or the federal savings and loan bailout), awareness was surprisingly abundant. But rooting through mounds of poll data for examples is not the point. What is essential is the alignment of poll queries with appropriate information levels. A citizen ignorant of 99 of 100 things might still offer wise counsel if the question pertained to his one area of knowledge. 8. U.S. Department of Commerce, Bureau of the Census, Statistical Abstract of the United States: 1998 (Washington: Government Printing Office, 1998), Table 381. 9. Ibid., Table 167. 10. During the early stages of my own research I 14 Abstract cravings for public largesse should be treated as cinteresting curiosities d; under no circum- stances should they inform poli- cymaking or determine policy choices. solicited cost estimates from one prestigious sur- vey organization for 500 face-to-face interviews in a single metropolitan area. Part of the expense entailed developing new ways to probe complex social welfare views. The proffered cost estimate was $160,000, a sum well beyond my budget. 11. These are summarized in Carolyn W. Sherif, Muzafer Sherif, and Roger E. Nebergall, Attitude and Attitude Change: The Social Judgment-Involvement Approach (Philadelphia: W.B. Sanders, 1965). 12. These are summarized in ibid., chap. 4. 13. This technique is also notable for its incom- patibility with modern statistical analyses. If sim- ple descriptive portrayals are the only require- ment, only clutter poses a problem. But, if more sophisticated multivariate techniques are to be applied, the computational challenge is sizable. This drawback looms large in the contemporary academy where elaborate computation exercises drive professional prestige. Imagine statistically analyzing items where scale interval and number varied across respondent. In other words, some respondents employed a 5-part metric, others used 10 categories, and so on and on. And, to boot, rather than the researcher 9s grouping simi- lar items, respondents themselves made those cat- egorizations; many might be idiosyncratic or even nonsensical. The Procrustean bed of convention may violate reality, but it is practical. Again, the methodological tail wags the substantive dog. 14. Translation of autonomous individual preference into a collective, majority-based outcome is, of course, a theoretical problem of the first order that has pro- vided full employment to untold public choice practi- tioners. A reasonably succinct and nontechnical analy- sis of this dilemma can be found in William H. Riker, Liberalism against Populism (Prospect Heights, Ill.: Waveland, 1988), especially chap. 10. 15. John P. McIvor and Elinor Ostrum, cUsing Budget Pies to Reveal Preferences: Validation of a Survey Instrument, d Policy and Politics 41 (1976): 87 3110. Significantly, well-educated, affluent respondents displayed similar troubles when the exercise was presented via a telephone poll. Here about a quarter of the respondents could not complete a five-item budget involving basic city services. That performance level is roughly the same as what our research uncovered. See Peter J. May, cA Technique for Measuring Preferences for Spending Reductions, d Social Indicators Research 10, no. 10 (1981): 389 3405. 16. Robert Cameron Mitchell and Richard T. Carson, Using Surveys to Value Public Goods: The Contingent Valuation Method (Washington: Resources for the Future, 1989), chap. 1. 17. CVM studies have also sought realism by con- sciously selecting respondents who have had actual policy encounters. Again, a far cry from today 9s polls permitting ordinary folk to cacquire d billion-dollar anti-missile defense systems. Research on WTP for wilderness recreation, for example, had hunters and hikers from Maine navigate these arduous choices and financial allocations. Dollar figures were kept plausi- ble 4a dollar or two a day to enjoy a local lakeside mari- na 4as opposed to, say, unspecified immense expendi- tures for cthe environment. d At least according to the technique 9s proponents, when all the interviewee traits and experiences are compared with their expressed dollar commitments, the assessment process appears economically rational. Mitchell and Carson, chap. 1. 18. Other practical obstacles are substantial. CVM demands an enormous researcher effort in customizing questionnaire items and, frequently, adjusting dollar figures to individual respon- dents. Polling firms would assuredly have to hire policy experts and economists galore to deter- mine, for example, what it personally costs each of 1,000 people to hire five new teachers in their home school districts. Collecting precise personal information of a type well beyond simple demo- graphics likewise imposes heavy research bur- dens. Those requisite data may also be murky or even unknowable (e.g., accurate costs of projects years down the road). And the ideological pres- sures to underestimate costs (including risks) may be inescapable. 19. Upgrade schemes number in the hundreds, and more surface daily, yet each one repeats the shortcomings of its predecessors. A useful overview of the deliberative poll is offered by Daniel M. Merkle, cThe National Issue Convention Deliberative Poll, d Public Opinion Quarterly 60 (Winter 1996): 588 3619. 20. Gary Langer and Ben Fitzpatrick, cTempered by Risk, d ABCNEWS Poll, April 9, 1999, electron- ic version. 21. Making such colossal numbers meaningful to ordinary people is difficult. One survey asked about the humdrum matter of bus safety, namely about reducing mortality from 8 per 100,000 to 4 per 100,000, and, ultimately, to 1 per 100,000. See M. W. Jones-Lee, M. Hammerton, and R. R. Phillips, cThe Value of Safety: Results from a National Survey, d Economic Journal 95 (March 1985): 49 372. 22. Libertarian readers might at this point sug- gest a poll version of fighting fire with fire. Why not launch counter-polls to subvert the Washington colossus? To wit, when respondents are informed of gargantuan costs, dangerous risks, and the necessity of settling for less than 15 Utopia, devotion to government largesse cools. Now the survey might inquire, cDo you think the homeless problem should be left to local commu- nities if Washington 9s assistance will only make it worse? d Surely the results would generally encourage those opposed to federalizing every societal disorder. Conceivably, a swarm of such items might eventually undermine the alleged welfare state consensus. Though tempting, this cmy-poll-beats-the-hell-out-of-your-poll d strategy is impractical and, more important, disingenu- ous, despite its conformity to today 9s polling standards. Though the commercial pollster will happily pose those questions, people who hope to defeat today 9s pro-welfare consensus face a stacked deck. For one, industry stalwarts, includ- ing the vital academic wing, will remain quite comfortable with statist vox populi outpourings. While their opponents must scamper to fund their episodic poll salvos (at $20,000 or more a shot), those at the controls (particularly in uni- versity settings) can easily repeat the standard, pro-statist item as hallowed convention. Moreover, government itself supports the polling enterprise. As Johns Hopkins University political scientist Benjamin Ginsberg so forcefully argued, our current welfare colossus could not exist apart from widespread popular endorsement, and those who profit from it are deeply motivated to publicize this fact. It is a bureaucrat 9s dream to champion what the public fervently demands. See Benjamin Ginsberg, The Captive Public: How Mass Opinion Promotes State Power (New York: Basic Books, 1986). 23. The reader may falsely see me as a foe of democracy. The burden of this paper has been to show why polls should not be used to make poli- cy choices. For that reason, I should be counted a critic of direct democracy, which in this case num- bers the pollster among its friends. I do count myself a friend of representative democracy, which is antithetical to the rule of the pollster. A more general statement of my doubts about direct democracy through public opinion must await another occasion. 16 Published by the Cato Institute, Policy Analysis is a regular series evaluating government policies and offer- ing proposals for reform. Nothing in Policy Analysis should be construed as necessarily reflecting the views of the Cato Institute or as an attempt to aid or hinder the passage of any bill before congress. Contact the Cato Institute for reprint permission. Additional copies of Policy Analysis are $6.00 each ($3.00 each for five or more). To order, or fo

less