“Objectivity” and Value
Public Opinion and American Democracy.
by V. O. Key, Jr.
Knopf. 566 pp. $7.50.
How do political systems work? In trying to answer this question, the exponents of classic political theory, from Aristotle to Montesquieu, looked at the global, institutional features of political societies and developed hypotheses accounting both for the differences between systems and for the stability or decay within any given system. But, though concerned with the facts of political life and the concrete causal explanation of recurrent processes, classic theory did not stop there: one of its main preoccupations was with questions of value, and a central problem was that of the “best” government, the “ideal” state. For many classical theorists, this problem appears in relativized form: Montesquieu, for example, no longer asks which government is the “best” but only which is “best” for a society of a given type. But concern for value never disappeared from political studies so long as these were based upon the global, institutional approach.
Departing radically from this approach in trying to account for the working of present political systems, contemporary “political science” eliminates value considerations as well. Both features of the classical approach—its orientation toward global institutions and its preoccupation with value—are rejected for the same reason: that they are incompatible with “scientific objectivity.” If we wish to be “scientific,” we must be concerned only with “facts,” not with “values” (that is, not with those “values which we ourselves hold; other people’s “values,” recorded with detachment, are of course a proper object of “scientific” investigation). And the “facts” must be objectively ascertained. But global features of societies do not lend themselves to such purely controlled observation: only what individual people do and say can be put into an “objective” record of facts. Hence, “political science” today seeks to account for the working of political systems in terms of a “grass roots” approach, using observed facts about individuals as its basic data.
A collection of data, however, is not science: correlations between data have to be established, and such correlations, to be scientific, must be stated in numerical form. Therefore, “political science” tells us, we cannot just look at what people do spontaneously. We have to ascertain how they respond to some identical, standardized stimulus, and then tabulate their responses. Having done this, we proceed to find out how the responses correlate with such sociocultural variables as economic status, occupation, education, age, sex, and the rest. “Political science” investigates significant statistical correlations between variables of this kind on the one hand and such responses to standardized political stimuli as voting in elections or answers given to political attitude questionnaires on the other. This is how we are supposed to get a “scientific” account of the working of our political system.
_____________
“Political science” has held the stage since before the Second World War. By now, it has accumulated an impressive body of voting studies and of statistical survey studies. How far have these labors advanced our knowledge of how the American political system works?
We may turn for an answer to Public Opinion and American Democracy, by V. O. Key, Jr. We find here a masterful exposition and evaluation of the methods used and the findings arrived at in the most important primary research work done in “political science” since early World War II times. Much of the book consists of an analysis in depth of a number of political attitude survey projects run by the Survey Research Center of the University of Michigan during the election campaigns of 1952 and 1956, but other important work (e.g. Berelson’s and Lazarsfeld’s famous voting studies) also receives careful attention.
The research effort reviewed in Professor Key’s volume deals with the political behavior and attitude profile of the entire American population as reflected in representative samples. To derive such profiles, one must study pervasive modes of behavior and pervasive attitude patterns: behavior that all participate in, attitudes that concern issues relevant to all. The main pervasive mode of political behavior is participation in elections; and everyone contributes to the behavior profile by voting for this or that candidate, or by abstaining. A pervasive attitude pattern is based upon questions on which every individual takes a stand; profiles are therefore made up of everyone’s answers to questions of common interest, whether “yes,” “no,” or “don’t know.”
Now the relevance, the cognitive value of voting “profiles” is not problematic. Elections are genuine political stimuli, and every response to them is politically meaningful: the outcome is determined jointly by voting and non-voting. Here “political science” is on firm ground, and the voting studies surveyed in Professor Key’s book afford a wealth of insights into American political behavior and into its sociological background. The case is, however, different with attitude profiles. One can always get a “profile” by addressing a questionnaire to a representative sample and tabulating the answers. But such a “profile” does not necessarily indicate how Americans define their stand on political issues that are significant for them. It shows, in itself, only a distribution of responses to a stimulus which has a well-defined political meaning for the researcher but may have a different meaning, or none at all, for the respondents.
Attitude survey studies are often bedeviled by such difficulties; as Professor Key points out, we don’t know too well what surveys actually “measure.” Identical responses need not denote identical or even similar attitudes, since respondents may misunderstand or fail to understand the question. It is also difficult to assess the significance of “don’t know” answers, a problem incisively dealt with by Professor Key. Apart from all this, however, a fundamental systematic distortion seems to me to be involved in the very design of many political attitude survey studies.
Political attitude surveys are typically designed to “measure” how people divide into “liberal” and “conservative” attitude groups, or how they distribute themselves along a scale running from the “liberal” to the “conservative” pole. To “political scientists” in general, the main pervasive attitude pattern running through American society is that defined by the liberal-conservative dichotomy. All studies investigating this pattern, however, follow the rule that identical answers to questions relating to liberalism-conservatism must be given the same score regardless of the respondent’s social position and involvement in the issue on which the score is being established. Thus, a union member’s pro-union answers are rated as indices of “liberalism” on a par with those of businessmen or intellectuals.
This is a fundamental error. If a man favoring the underdog is by definition a “liberal,” the underdog who agrees with him need not be so. Pro- and anti-union answers give us only a profile of pervasive American attitudes toward labor unions: this profile cannot then be taken as reflecting a pervasive American attitude pattern concerning liberalism and conservatism. How such a pattern can be established at all is, indeed, difficult to see. If we directly question our samples on “liberalism” and “conservatism,” we get no usable profile, because these terms mean different things to different people. We can approach the problem only indirectly, by asking questions about issues supposed to provide criteria of liberalism or conservatism, and this is what “political scientists” do. But differential involvement in all such issues (whether they concern labor unions, medical aid, discrimination, or whatever else) tends to make responses non-additive on a “liberalism” scale: issue questions cannot be regarded as “standardized stimuli” that affect everybody in the same way and produce responses that can be added up in the same dimension, with respect to “liberalism” and “conservatism.”
It seems that “political scientists” who, being liberals themselves, are genuinely interested in liberalism vs. conservatism as the fundamental political issue coloring all others, uncritically project this concern upon American society at large. But this shows that “political science” has discarded the supposed pitfalls of classic theory only in intention, not in fact. It also attributes directly unobservable global features to society; it also incorporates the value attitudes of its practitioners. Apparently one cannot avoid doing these things when studying politics and society. The only question is how well one does them, but it is clear that people who are unaware of doing a thing cannot do it very well.
All this does not mean to say that attitude survey studies are generally worthless. On the contrary, many such studies (no less than the voting studies) have derived valuable knowledge about the social background of American politics by the use of excellent, and sometimes truly brilliant, techniques of investigation, deduction, and analysis. Still, the general picture that emerges is both distorted and incomplete. Supposed measures of liberalism often measure nothing, and the mine of information worked by questionnaire techniques is quickly exhausted.
Professor Key indicates that “political science” has found no satisfactory way to bring into focus such phenomena as political power, influence, and leadership. Indeed, as the survey studies show, “political science” seeks to isolate the “influential” members of the political community by finding out who votes most regularly, reads most about politics, does most of the party work, and so on. This is how heroically “political science,” wedded to the grass roots approach, strives to avoid its own subject matter. True, political issues are of concern to all, and in a democracy everyone is supposed to participate in politics. But many things besides pervasive attitudes and modes of behavior that everyone engages in on an equal footing are found among the essential determinants of the political process. These include, on the one hand, the hierarchical features of politics (differentiation between leaders and led, rulers and ruled) and on the other, its institutional features (the “rules of the game,” the established modes of acquiring and exercising authority). Classic theory, which concentrated on these, was on the right track; “political science,” which systematically blocks them out, incapacitates itself thereby. Of course we are aware today of the need for more rigorous and extensive fact-finding techniques than those available to the classics, and it is “political science” that has given us this awareness. But the methods we need cannot be based upon the concept of “objectivity” propounded by “political science.” We cannot do without interpretation and judgment, and the objectivity which is possible (and imperative) in these fields necessarily has an admixture of controlled subjectivity. This is inescapable. The “pure” objectivity postulated by “political science” leads both to loss of subject matter and to an intrusion of uncontrolled subjectivity.
_____________
