Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Numbers and Decisionmaking
Public Statistics arid Democratic Politics KENNETH PREWITT If, to paraphrase Harold Lasswell, politics has become how much for how many, it is clear that measurement moves toward the center of political life. The result is a politics of numbers-What is to be counted? By whom? Can the numbers be trusted? In which direction is the trend line moving? Who is at fault for the (now numerically defined) failure of a policy or program? The intrusion of numbers into politics is global, as the world's nations now endlessly debate issues couched in numerical estimates and forecasts: weapon counts, oil reserves, trade balances, North-South ine- quities, debt ratios. With reason, then, scholars have focused their attention on how numbers are generated and subsequently used or misused in politics. This important scholarship rests on the assumption that public statistics are not politically neutral. Decisions about what to count are influenced by the dominant political ideologies, and numbers enter the political fray on behalf of social interests. The approach adopted in this essay accepts this assumption but focuses it as follows: public statistics in the United States are generated as part of democratic politics. This invites inquiry into the ways in which this par- ticular nation's "number system" advances or retards democracy, informs or distorts civic discourse, helps or hinders political participation. For just as public statistics are not neutral with respect to the everyday politics of group interests, so they are not neutral with respect to the principles and practices of democracy. Consequently, to study constitutional democracy, as it is today practiced in the United States, requires a perspective on 113
114 KENNETH PREWITT numerical reasoning and the nation's number system. Providing this per- spective is a task for social theory. There are of course unresolved issues in what does, or should, constitute democracy in the United States. We cannot attempt here to sort out the relative emphasis that contending theories of democracy give to such issues as popular participation, economic and social equalities, the protection of property, civil liberties and citizen rights, or democratic procedures. In this chapter we take the simpler route of concentrating on two central issues: accountability-how public leaders are held accountable for their perfor- mance in office; and representation how diverse interests are represented in setting the political agenda. DEMOCRATIC ACCOUNTABILITY The centrality of the concept of accountability in democratic theory de- rives from the observation that democracies no less than other forms of government have public officials with immensely more power than average citizens. Democratic theory does not deny the power advantage enjoyed by those in charge of the government, nor does it optimistically presume that democracies are free of the tendency of power-holders to expand their control. Embedded in a democracy, again no less than in other forms of government, is a structure of bureaucratic and political power. The task of democratic theory is to direct us toward practices that rec- oncile the inclination of power systems toward dominance with the dem- ocratic ideal of popular sovereignty. The basic terms of this reconciliation are to be found in the Constitution, especially in the provision for separating and fragmenting official power so that leaders can check and control each other, and in the companion provision that regular electoral competition will force leaders to contest with each other for the favor of the voters. The general idea of this second provision is summarized in the phrase "theory of electoral accountability" as first adumbrated in The Federalist Papers and subsequently elaborated by Schumpeter and other democratic theorists. There is competition for public office. Leaders present themselves and their records to the electorate. Voters, basing their judgments on the past performance or estimates of future performance of leaders, elect, re- elect, or evict accordingly. Leaders, knowing this, and wanting to gain and retain office, promote policies that will attract public support. This theoretical formulation is a reasonably accurate though partial de- scription of what, in fact, does happen. The empirical evidence has been most compellingly presented by political scientist Morris Fiorina (1981; also Kramer, 1971), who has demonstrated the use voters make of retro- spective evaluations. Voters routinely reject incumbents who governed dur
PUBLIC STATISTICS AND DEMOCRATIC POLITICS 115 ing a period marked by deterioration in social and economic conditions. Another political scientist (Kiewiet, 1983:115) reports that voters "clearly react in an incumbency-oriented fashion to the record of current office- holders, responding positively to success in the economic and other arenas but negatively to perceived failures." Two explanations are available. Citizens vote according to recent changes in their own economic conditions, or, citizens vote according to the im- provement or deterioration of national economic conditions. Under the fast explanation, votes would be influenced by the personal experience of unemployment, the personal loss of purchasing power through inflation, or the need to postpone homeowning because of high interest rates. The citizen experiencing these negative economic conditions votes against the political party in power. Under the second explanation, voters punish or reward politicians de- pending on the performance of the national economy during the incumbents' tenure. For example, citizens who may be secure in their own employment nevertheless vote against leaders whose policies bring about high rates of unemployment. Or, citizens not themselves seriously affected by high in- terest rates nevertheless take into account double-digit interest rates when evaluating the performance of incumbent officials. Somewhat counterintuitively, current research supports the second ex- planation. Voters in the United States give more weight to negative or positive trends in national economic conditions than to changes in their own economic circumstances. The most extensive development of this finding is offered by Roderick Kiewiet, who concludes (1983:1311: "Changing perceptions of the national economy account for a considerably larger pro- portion of the swing in support for the incumbent party from good years to bad than do changes in personal economic conditions." This research finding is important in the present context for what it indicates about the function of national statistics in implementing democratic accountability. If voters punish and reward officeholders less in terms of personal experience than in terms of national economic performance, they can vote responsibly only if they have some reasonably accurate information about that performance. This information of course is often made accessible when it is summarized as statistical trends. Political leaders can be judged by the upward or downward movement of statistical indicators of those socially important issues for which government has assumed responsibility: unemployment, inflation, balance of trade, interest rates, test scores, pov- erty levels, crime rates. When economic and social indicators are moving in politically popular directions, political credit is claimed; when they are moving in unpopular directions, political blame is assigned. Here, then, is a contribution of public statistics to the workings of democracy.
116 KENNETH PREWI1T This application of numbers to the purposes of democratic accountability occurs in a period when many other political developments undermine the conditions necessary for holding public officials to account: the decline of parlor discipline, even of political parties themselves; the increased costs of electioneering and the related packaging of candidates by media experts; the growing political influence of single-issue interest organizations; the comparatively low rates of political participation. These trends occur as the political agenda is ever more crowded with issues difficult for the average citizen to comprehend. A weakened party-electoral system combined with a crowded and complicated issue agenda is not conducive to democratic accountability. Against this background, it is all the more important to understand whether numeric descriptions of major social conditions and trends can improve the reasoning capacity of modern democracies. The hypothesis can be generalized. Just as a particular administration in power can be evaluated by statistical trends, so also can broad social pol- icies. In this generalized version, citizens continually evaluate and reev- aluate broad policy commitments made by previous political generations. In modern nation-states, this retrospective public reflection is facilitated by measures of long-term trends. Descriptive statistics, especially when pre- sented as trend lines, offer voters before-and-after information about the performance of incumbents as well as of general policies. Consequently, these statistics contribute to the procedures that establish accountability in democratic politics. If we could leave matters at this point, the story would be a welcome one for democratic theory. But it is more complicated. Numbers, just as much as words, have the power to distort as well as enhance the reasoning capacity of the public. The greater the importance of numbers to the securing of power, the stronger the incentives to those in power to make certain that numbers present a favorable even if inaccurate picture. Across a broad front democratic politics must contend with ways in which numbers distort and mislead. At this point it is necessary to draw attention to an important subset of any nation's number system, what are called "performance indicators." Performance indicators typically serve two functions: they act as internal signals for the agency, telling it whether its goals are being achieved; and they serve as signals to those outside the agency, including of course those who set policy and control budgets. These two functions subject an agency to conflicting pressures. When an agency designs performance measures in a manner that maximizes internal information, it invites external attention to its failures as well as achievements. It risks sending negative signals that those having power over the agency can use to trim budgets or punish incompetence. It is a familiar complaint that when officials are rewarded or punished
PUBLIC STATISTICS AND DEMOCRATIC POLITICS 117 according to statistical evaluations, they are drawn to policies that favor how the agency presents itself to the oversight process rather than policies that improve the conditions for which they have responsibility. The numbers become more important than the progress toward policy goals they pre- sumably index. Khrushchev (1977:71) is said to have lamented: "It has become the tradition to produce not beautiful chandeliers to adorn homes, but the heaviest chandeliers possible. This is because the heavier the chan- deliers produced, the more a factory gets since its output is calculated in tons." Our interest is not in this well-known flaw in command economics, but in the implications for democratic accountability. If the number system is systematically manipulated so that personnel and policies are presented to the public in the most favorable light, we have little warrant for claiming that public statistics enhance democratic procedures. We come here to a point in the discussion where the larger analysis of democratic accountability intersects with a more specific argument about the professional accountability of those who administer the nation's statis- tical system. This accountability is to professional peers who evaluate, against the standards of their disciplines, whether government statistical agencies are maintaining the integrity of the numbers. Professional statis- ticians, in and out of government, hold that proper controls and procedures can protect the public from the abuses associated with fraudulent or mis- leading statistics. In recent congressional testimony, Courtenay Slater, an informed and experienced observer of national statistics, comments (1983:541: "One of the finest things about our statistical system is that our statistics have credibility. They are produced by professionals in the statistical agen- cies and the press release that gives us our economic data comes out of that statistical agency. It is written by professionals. Everybody knows it is objective, and they believe the numbers are honestly presented." There is no question that in well-established statistical agencies, such as the Bureau of Labor Statistics and the Bureau of the Census, the production and reporting of statistics is managed by professionals. The norms of profes- sional control are deeply rooted in the origins of these agencies. Consider, for example, the history of the Bureau of Labor Statistics, which just celebrated its centennial. The bureau was established during the post-Civil War period of intense civil strife, and the statistics it produced were quickly implicated in the arguments of the day about the causes and consequences of industrial conflict. Labor reformers especially felt that if information about prevailing employment conditions could be put "before the legislators and the public, a cry of mingled surprise, shame, and indignation will arise that will demand an entire change in the method of earnings and pay" (Wright, 1870:381.
118 KENNETH PREWllT Although not disputing this assumption, the statisticians and social sci- entists, who by now had organized themselves as the American Social Science Association, wanted to ensure that the new labor statistics not be seen to favor any particular economic interest radical, reformist, or con- servative. The advice given to and heeded by Carroll Wright, founder in 1873 of the first state Bureau of Labor Statistics in Massachusetts and subsequently first commissioner of the Federal Bureau of Labor Statistics, is instructive (Walker, 1877:vii-viii). Your office has only to prove itself superior alike to partisan dictation and to the seductions of theory, in order to command He cordial support of the press and the body of citizens.... I have strong hopes that you will distinctly and decisively disconnect Be [bureau] from politics. This advice is no less heeded today than it was a century ago, and for the same reasons. Perhaps no stronger testimony to the credibility of our major statistical series is needed than the reliance placed on them by both the political process and the marketplace. A member of Congress (Horton, 1983:2) comments, "It would be a public administration catastrophe if we were to find that the statistics we rely on so heavily did not adequately describe the real world of which we are a part and the problems we are trying to solve." In the marketplace substantial funds are routinely trans- ferred on the assumption that national statistical series are trustworthy. The monthly statistical reports of the Crop Reporting Board of the Department of Agriculture, for instance, have such high credibility that hundreds of thousands of dollars change hands through the commodity markets as soon as the data are released. But even if we accept that professional control over national statistics can largely eliminate fraud and greatly lessen bias in the most important of our social and economic indicators, other issues remain. The statistics of even the most professional agencies suffer from measurement problems for which there are no presently available solutions. When these problems lead to errors of serious magnitude and yet the numbers are used by political leaders to set policies and by citizens to evaluate these policies, the ac- countability process is compromised. The aptly labeled "unobserved economy" offers a telling illustration. Two scholars report (Afford and Feige, n.d.:14~: "Recent research suggests that systematic biases associated with a large and growing sector of un- measured economic activity have been introduced into the system of social indicators. The unobserved sector escapes the social measurement apparatus because of accounting convention, nonreporting and underreporting." If the unobserved economy is growing more rapidly than the observed, that is, "counted" economy, but policy is guided by statistics only about the
PUBLIC STATISTICS AND DEMOCRATIC POETICS 119 latter, serious errors can hardly be avoided. This in turn of course distorts the process by which fault is assigned, and moves us away from democratic accountability. The technical and conceptual errors associated with measurement are serious, but for important economic and social indicators continuous profes- sional attention and public discussion offer safeguards. The scholarly com- munity here carries a major responsibility. Social scientists and professional statisticians have the technical skill and career incentives to discern discrepancies between what the statistics purport to measure and what they actually measure. These safeguards can operate only when the statistics are indeed public, that is, accessible to professional attention. Such is not the case for critical domains of national security policy, where secrecy prevails. Professional review of the adequacy and integrity of, say, unemployment or inflation measures is orders of magnitude more informed than professional review of numbers purporting to describe, for instance, the comparative weapon systems of the United States and the Soviet Union. In a telling essay, McGeorge Bundy (1984) compares statistics on U.S. nuclear weapons appearing in two publications: the officially produced Defense Department ~4nnucz1 Reportfor the Fiscal Year 1985 and a privately sponsored report of the Natural Resources Defense Council, the Nuclear Weapons Databook. The official publication consistently underestimates American resources in a manner, Bundy argues, designed to make the "Russians look big" and the "Americans small." This is the success indicator issue stood on its head, similar to when police departments inflate crime statistics to justify larger budgets. At issue is not the inevitable tension between the claims of national security and the right of the public to be informed, for we refer here only to those numbers that are presented by the government in public discussion, from the controversial "body counts" in the war of attrition in Vietnam to the equally controversial "missile counts" in the debates about the window of vulnerability. In sharp contrast to the care with which major statistical series affecting social and economic policies are professionally monitored, there has been little serious attention given to how independent professional controls can be applied to military numbers routinely advanced in open forums. A democratic society is preserved when the public has reliable ways of knowing whether policies are having the announced or promised effect- Is inflation being brought under control? Is a war of attrition being won? Are defense expenditures buying national security? Numbers, a part of this publicly available political intelligence, consequently contribute to the ac- countability required of a democracy.
120 KENNETH PREWI17 Flaws in the statistics, whether inadvertently or deliberately introduced, mislead citizens about the performance of their government, thereby di- minishing accountability, but it can be plausibly argued that the wide public availability of reasonably accurate statistics about social conditions for which government is responsible enhances more than it diminishes dem- ocratic accountability. This conclusion, at best an informed guess, rests on assumptions about what is required if civic discourse is to be reasonably informed under the conditions of advanced industrial societies. It also rests on (largely untested) assumptions about the capacity of an electorate to make intelligent use of statistical information. This last point we briefly return to in the concluding section, after reviewing the second of our two major issues connecting national statistics with democratic theory. REPRESENTATION OF DIVERSE INTERESTS As a document in democratic political theory, the Constitution's genius is in its provision for the representation of diverse interests in political decision circles. This commitment to representation involved the founders in political engineering, one aspect of which established the close associ- ation between political representation and the nation's number system. In order that seats in the House of Representatives might be fairly allocated, the Constitution mandated a population count. It further directed that this count distinguish among the free citizens, the slave population, and the untaxed Indian population. This distinction arose because the founders wanted wealth as well as property to be reflected in apportionment count- ing slaves as three-fifths of a person was a way to recognize their property value. Representation had to be apportioned according to politically ac- ceptable criteria. Moreover, the method chosen had to allow for adjustments as the population expanded, redistributing itself among the existing states or spilling over into territories that would later achieve statehood. Thus was established the decennial census, the centerpiece of our statistical system. The limited use of the census to apportion congressional seats did not satisfy James Madison. In early congressional debates Madison (1790:1077) urged that the census "embrace some other objects besides the bare enu- meration of the inhabitants." Madison suggested that the census describe "the several classes into which the community is divided." On this basis, continued Madison, "the Legislature might proceed to make a proper pro- vision for the agrarian, commercial, and manufacturing interests, but with- out it they could never make these provisions in due proportion." We know from The Federalist Papers that Madison viewed society as consisting of multiple and diverse interests. To govern such a society in a
PUBLIC STATISTICS AND DEMOCRATIC POL]lTICS 121 democratic fashion required complex information about the composition of the public. Thus, for Madison, it was not enough that the census enumerate the population for the sole purpose of apportioning. It should be expanded to include many population characteristics, and thereby become the basis on which the legislators could allocate taxes, benefits, and services ac- cording to the "real situation of our constituents." In anticipating a de- mocracy in which numerical proportionality cuts much deeper than assigning congressional seats, Madison was ahead of his time. Madison's opponents started from a different theory of politics. Reflect- ing eighteenth century theories of the organic society, they "viewed the object of government as the pursuit of an undifferentiated common good; for them, politics was a sphere of virtue, and empirical investigation was irrelevant" (Starr, 1984:371. In the early days of the republic Madison's opponents prevailed. Enumeration was sufficient to serve representation. Contemporary practice, however, is much closer to Madisonian plural- ism, as reflected in the vast expansion of the national statistical system and the policy uses to which it is put. The question before us now is how these developments in the statistical system affect the political representation process. Providing for the representation of diverse interests in political decision circles is at the core of the theoretical formulation known as democratic pluralism, the now dominant interpretation of American democracy. Dem- ocratic pluralism takes as its central problem the conditions that allow for the participation by interested parties in various policy domains. Democracy requires that there be no barriers to the organization and expression of the full array of interests in society. Democratic pluralism is an attractive theory. Since Me early days of the republic it has gradually gained adherents among those who have puzzled over the prospects for democracy in large-scale advanced industrial nations. But the theory also has its critics. In recent decades the effort to formulate a democratic theory has emphasized participation as opposed to pluralism, and in the process generated a critique of conventional pluralist theory. This critique holds that pluralism has not offered a satisfactory account of nonparticipation in democratic politics, too readily attributing low levels of participation to presumed citizen defects such as apathy or ignorance. Since levels of participation covary with social and economic resources, the critics argue, pluralism functions as a justification for the representation of middle and upper class interests in politics rather than a description of how the full array of social interests find a political voice. An alternative explanation of nonparticipation is suggested by E. E. Schattschneider's famous phrase, "mobilization of bias." In explaining why the socially and economically disadvantaged often fail to participate
122 KENNETH PREWI~ in politics, Schattschneider wrote (1960:1051: "Whoever decides what the game is about also decides who gets in the game." This introduces the argument that what is on the political agenda provides a referent point that selectively mobilizes participation across different social groups and inter- ests. Citizens participate not just to put issues on the political agenda but also, and more often, in response to the issues already there. This mobi- lization process, according to Schattschneider, is biased against the interests of the less well off groups in society. It is in this theoretical context that we consider how the analysis and political reporting of social statistics intersects the representation system. Although our emphasis is on contemporary politics, the practice we draw attention to is at least 150 years old. Starting around 1820, writes the historian Patricia Cohen (1982:169), "Many private agencies and volunteer groups with reformist agendas adopted the statistical approach to social facts in order to document the dimensions of the problem they were ded- icated to eradicating." Cohen offers several examples: the use of statistics to describe the miseries of public prisons; the effort by the temperance movement to prove quantitatively that alcohol abuse was a growing problem; and local surveys of pauperism as a basis to challenge poor laws. In deploying their privately collected statistics on behalf of social reform, the early nineteenth century activists anticipated developments surrounding publicly collected statistics that did not come fully into view for another half-cenh~, when the federal Bureau of Labor Statistics was established in the 1880s. The 1820 reformers were signaling to later activists that statistics could mobilize political participation and inform public debate. In the latter part of the twentieth century these possibilities are etched much more deeply in our political life. The nation's number system uncovers social conditions and popularizes them as statistical descriptions: proportion of the population below the poverty line; incidence of child abuse; persis- tence of structural unemployment; addictive behavior and its social costs; the differential in infant mortality between whites and nonwhites; the gap between male and female wages in similar occupations. The transformation of politically unnoticed social conditions into visible statistics puts issues on the political agenda that would otherwise be ignored. These statistical conditions then provide a political referent point for interested groups. This is perhaps one of the most striking aspects of twen- tieth century democratic politics. Resource-poor social interests turn to a statistical description of their plight to generate political pressure and to mobilize adherents to their cause. The history of the civil rights movement is suggestive in this regard. The concept of institutional racism, which held that black poverty was caused not just by racial prejudice but also by structural conditions of the economy,
PUBLIC STATISTICS AND DEMOCRATIC POLITICS 123 polity, and society, made its political appearance through statistics on res- idential segregation, black-white income differentials, unequal educational opportunities, inequities in access to health care, and so forth. Civil rights leaders first used the numbers to emphasize the scope of institutional dis- crimination. They then used them to gain political support for new social policies such as Headstart, job training, and afDumative action. Other groups have reached the conclusion that to be "measured" is to be politically noticed, and to be noticed is to have a claim on the nation's resources. Thus the physically handicapped in New York initially resisted being counted, for fear that this would lead to further stigmatizing them, but then reversed their position when they realized that political visibility closely followed on statistical visibility. Data presented in Michael Harrington's The Other America helped initiate the War on Poverty by identifying the poor as a target group for government action. The consumer protection movement, starting with Ralph Nader's Unsafe at Any Speed, has made heavy use of statistical arguments, as have the environmentalists. Describing public-interest citizen groups, one com- mentator (Henderson, 1981:441) writes that "the quality and quantity of information and the way it is structured, presented and amplified" shapes their political choices and strategies. Harold Wilensky (1967:19) generalizes these observations when he writes that "facts and figures" assist those political interest organizations "weak in grass-roots political resources." Information "may give an advantage to the weak, whose case, if strong and technical, can count for something." This is not a trivial observation when examined in the context of the effort through He history of democracy to establish equal civil and political rights in the face of inequalities in resources that different social interests bring to the political arena. In democratic theory as well as actual practice, organization is most often promoted as the corrective when economic inequalities are repro- duced as differential opportunities for political participation. The less wealthy but more numerous social interests combine and increase their political strength through working-class parties, social movements, and interest groups. Consequently, a resource that helps to organize the re- source-poor will help to correct political imbalances and promote broader democratic participation. This observation leads us to consider whether statistical programs can actually help establish group identity and lead to the formation of interest organizations. In a careful account of the interplay between ethnicity and the census, William Petersen (1983:27) writes: "Few things facilitate a category's coalescence into a group so readily as its designation by an official body," and cites the importance of "questions put to them by
124 KENNETH PREWl7 immigration officials and census schedules" for helping to solidify group identification . Hispanic-Americans are particularly important in this regard. More than any group in American political history, Hispanic-Americans have turned to the national statistical system as an instrument for advancing their political and economic interests, by making visible the magnitude of the social and economic problems they face. In the processes by which groups are formed and diverse interests are represented in democratic politics, public statistics are not an unmixed blessing. Just as some groups can establish a political identity by being enumerated, other groups cannot escape the way they are socially clas- sified because of this same enumeration system. For example, for two centuries we have had a statistical practice of racial classification, which undoubtedly has contributed to the continuing salience of race in Amer- ican society. Policies now being implemented could easily result in the Hispanic-Americans becoming a permanent racial minority in the statis- tical system, with what long-term effects it is difficult to foresee. More- over, the statistical system is not sufficiently robust to withstand the distortions accompanying severe political pressures. When political cri- teria are transparently used to determine what should be technical issues, such as the best way to count a population group, statistics lose their credibility. Racially sensitive measurement policies are not likely to be reversed soon, now that so many government services are allocated according to race and ethnicity. The brief period during which it was thought wrong to identify race, gender, or national origins on employment or school appli- cations was swept away by the emergence of affirmative action and statis- tical parity in the 1970s. The nation has entered a period in which "proportionate allocation" is carried to ever greater extremes. There is a contagion effect: Once statistical proportionality is elevated to a principle of government, there is great pressure from various racial and ethnic groups to be fully counted. From the perspective of democratic theory these developments are trou- bling in at least three respects. First, to assign to the statistical system responsibility for group classification and resource allocation is to transform the thing being measured segregation, hunger, poverty into its statistical indicator. Always in tension with the judgmental in politics is an insistent search for objective rules to reduce the element of arbitrariness in subjective judgment. The legal code is one such set of objective rules, formalized bureaucratic procedures another, and now statistical formulas. This search does not eliminate politics; it simply pushes them back one step, to disputes about methods. Arguments about numerical quotas, availability pools, and
PUBLIC STATISTICS AND DEMOCRATIC POLITICS 125 demographic imbalance become a substitute for democratic discussion of the principles of equity and justice. Second, if statistical identification facilitates political consciousness among some resource-poor groups, these same statistics make invisible to the policy process other groups at the margins of social and economic life, where measurement often fails the undocumented workers, the illegal aliens, and the vagrant, homeless populations. In many government programs, persons not counted are not there. Another difficulty stems from the inertia of statistical systems. For technical as well as bureaucratic reasons, statistics lag behind the dynamic patterns of group formation and change resulting from immigration, internal migration, transformation in the occupational structure, and new levels of social consciousness. Insofar as politics is organized by the numbers, there will be a tendency to overlook more recently established social conditions in favor of those already reflected through the statistical system. The third and most troubling danger is the shift away from a system of representation and public policy based on the individual citizen toward one based on the representation of demographic aggregates: ethnic, racial, in- come, gender, etc. This shift invites, even mandates, the allocation of benefits and rights according to group membership rather than individual accomplishment or need. To many observers this tilt toward group representation undermines the fundamental premise of liberal democracy. Nathan Glazer (1975:220) la- ments the drift toward numbering and dividing up the population into racial and ethnic groups: "This has meant that we abandon the first principle of liberal society, that the individual and individual's interests and good and welfare are the test of a good society, for we now attach benefits and penalties to individuals simply on the basis of their race, color, and national origin." Glazer, of course, does not attribute the rise of quota politics and group-based representation to the availability of statistical information. But if statistical information has not caused, certainly it has abetted the emer- gence of demographically defined groups as a category in public policy. The formal system of political representation itself has not escaped the insistent pressure for demographically defined proportionality. Abigail Thernstrom (1983) artfully traces how the 1965 Voting Rights Act was transformed in two decades from a law to protect black voting rights to one that appears to require the "correct" number of minority seats in legislative bodies. Demands for proportional representation, in which the legislature is to mirror the characteristics of the population from which it is selected, are not new. Until recently, however, group politics intersecting with the electoral process was the preferred avenue for achieving this end. Legal remedies were, appropriately, limited to ensuring fair procedures,
126 KENNETH PREWI17 not particular outcomes. Now, buttressed by statistics, laws have begun to affect the very composition of legislative bodies. As was the case in our discussion of accountability, we see in this discussion of representation that countertendencies are at work. On the one hand, statistical description can bring social conditions to public attention, mobilize disadvantaged groups, and broaden the political agenda in ways that lessen the bias inherent in an electoral representation system based largely on the resources of wealth and political organization. On the other hand, these statistics introduce practices and policies inconsistent with our traditional understanding of democracy: the objectification of politics, the assumption that that which is not counted is not there; the temptation to substitute group membership for individual merit or need as the basis for public policy; the allocation of legislative seats according to designated racial or ethnic criteria. We are far from having the evidence that would allow us to sort out the relative strength of these countertendencies and again must resort to an informed guess. With respect to democratic accountability I suggested that the benefits of statistical descriptions outweighed the hanns. With respect to the representation of diverse interests I am less sanguine. The distortions of the representational process seem to me every bit as strong as the im- provements. Moreover, the negative tendencies are not of the sort that can be corrected with greater professional scrutiny of statistical information. They are much more political than technical in nature and in fact become stronger as statistics become more precise and reliable. CONCLUSIONS I conclude by emphasizing the theme that connects this essay with the efforts of Ogburn and colleagues. The present inquiry has emphasized the importance of close attention to the nation's number system by professional statisticians and social scientists. Assuring the integrity of numbers involves continuous improvements in measurement, revisions in concepts as social conditions change, and the highest standards of statistical interpretation, analysis, and reporting. Moreover, protecting statistical quality and integrity will add little to democracy unless joined to the educational task of ensuring that numeracy takes its place alongside literacy as a skill indispensable to democratic citizenship. In the absence of public understanding of statistical argumentation, the numbers will more likely aid political demagoguery than democratic discourse. Because of, and notwithstanding, the various problems and risks iden- tified in this essay, those who care about democracy have a large task before them: analysis of the political role of numbers, as well as a commitment
PUBLIC STATISTICS AND DEMOCRATIC POllTICS 127 to making the numbers perform according to the responsibilities that a democracy places upon ~em. ACKNOWLEDGMENTS To fully acknowledge collegial contributions to this chapter would require this note to be nearly as long as the paper itself. It was prepared while the author was a Fellow at the Center for Advanced Study in the Behavioral Sciences. The working conditions at the center are much appreciated; col- leagues at the center who provided useful commentary include Joel Aber- bach, Paul Brest, Roger Noll, and Barbara Rosenkrantz. I am grateful for the support of the Guggenheim Foundation, and also for the financial support provided through the center by the National Science Foundation, under grant BNS 76-22943, and by the Exxon Educational Foundation. I appre- ciate as well critical comment from participants at the symposium on which this volume is based and especially from anonymous commentators com- missioned to review the very preliminary draft presented at We symposium. In addition, this chapter benefited from We Social Science Research Coun- cil's Conference on We Political Economy of National Statistics. A slightly modified version is included in a volume based on that conference, which gave me access as well to the editorial advice and sharp editorial pens of William Alonso and Paul Starr. REFERENCES Alford, Robert R., and Feige, Edgar L. N.d. Information Distortions in Social Systems: The Unobserved Economy and Other Ob- server-Subject-Policy Feedbacks. Unpublished paper. Bundy, McGeorge 1984 Deception, self-deception and nuclear arms. The New York Times Book Review, March 11:3,15. Cohen, Patricia Cline 1982 A Calculating People: The Spread of Numeracy in Early America. Chicago: University of Chicago Press. Piorina, Morris 1981 Retrospective Voting in American National Elections. New Haven: Yale University Press. Glazer, Nathan 1975 Affirmative Discrimination: Ethnic inequality and Public Policy. New York: Basic Books. Henderson, Hazel 1981 Information and the new movements for citizen participation. Pp. 434-48 in Thomas J. Kuehn and Alan L. Porter, eds., Science, Technology, and National Policy. Ithaca, N.Y.: Cornell University Press. Horton, Frank 1983 Federal Government Statistics and Statistical Policy. Hearing before the Legislation
128 KENNETH PREWRAP and National Security Subcommittee of the Committee on Government Operations, House of Representatives, June 3. Washington, D.C.: U.S. Government Printing Office. Khrushchev, N. 1977 Cited in Charles E. Lindblom, Politics and Markets. New York: Basic Books. Kiewiet, D. Roderick 1983 Macroeconomics & Micropolitics: The Electoral Effects of Economic Issues. Chicago: University of Chicago Press. Kramer, Gerald 1971 Short-tenn fluctuations in U.S. voting behavior, 1896-1964. American Political Sci- ence Review 65:31-43. Madison, James 1790 Annals of Congress, First Session, House of Representatives. Washington: Gales and Seaton, 1834. Cited in Steven Kelman, The politics of statistical policymaking: jus tification for public information-collection and theories of the role of government, Conference on the Political Economy of National Statistics, October 14-15, 1983:14. Social Science Research Council, New York. Revised version to be published in William Alonso and Paul Starr, eds., Politics of Numbers. Petersen, William 1983 Politics and the measurement of ethnicity. Conference on the Political Economy of National Statistics, October 1~15. Social Science Research Council, New York. Revised version to be published in William Alonso and Paul Starr, eds., Politics of Numbers. Schattschneider, E.E. 1960 The Semi-Sovereign People: A Realist's View of Democracy in America. New York: Holt, Rinehart and Winston. Slater, Courtenay 1983 Federal Government Statistics and Statistical Policy. Hearing before the Legislation and National Security Subcommittee of the Committee on Government Operations, House of Representatives, June 3. Washington, D.C.: U.S. Government Printing Office. Start, Paul 1984 Measure for measure. The New Republic February 13. Thernstrom, Abigail 1983 Trigger of reform, trigger of mischief: the use of statistics in voting rights and redis- tricting. Conference on the Political Economy of National Statistics, October 14-15. Social Science Research Council, New York. Revised version to be published in William Alonso and Paul Starr, eds., Politics of Numbers. Walker, Francis A. 1874 Massachusetts Bureau of Statistics of Labor, Fifth Annual Report. Cited in James Leiby, Carroll Wright and Labor Reform: The Origin of Labor Statistics. Cambridge: Harvard University Press, 1960:63. Wilensky, Harold 1967 Organizational Intelligence. New York: Basic Books. Wright, Carroll 1870 Massachusetts Bureau of Statistics of Labor, First Annual Report. Cited in James Leiby, Carroll Wright and Labor Reform: The Origin of Labor Statistics. Cambridge: Harvard University Press, 1960:55. 1877 Massachusetts Bureau of Statistics of Labor, Eighth Annual Report. Cited in James Leiby, Carroll Wright and Labor Reform: The Origin of Labor Statistics. Cambndge: Harvard University Press, 1960:68.
