Rule by Public Opinion
WRITING in 1893 before the present era of straw votes and public-opinion polls, James Bryce described what he considered to be various stages in the evolution of rule by public opinion. In the first stage, public opinion is passive, merely acquiescing in the exercise of a political authority which it is accustomed to obey. A second stage is reached when conflicts arise between unenlightened rulers and a public awakened to a sense of its real power. When, however, public opinion becomes an active and controlling factor in moulding public policy, exerting its influence through representative bodies and periodical elections, a third stage emerges, characteristic of most civilized states to-day.
Of even greater significance than this threefold classification is Lord Bryce’s allusion to a fourth stage. ‘A fourth stage would be reached,’ he stated, ‘if the will of the majority of the citizens were to become ascertainable at all times, and without the need of its passing through a body of representatives, possibly even without the need of voting machinery at all. . . . To such a condition of things the phrase, “Rule of public opinion,” might be most properly applied, for public opinion would not only reign but govern.’
Bryce had no illusions, however, as to our proximity to this stage: ’The mechanical difficulties, as one may call them, of working such a method of government are obvious. How is the will of the majority to be ascertained except by counting votes? How, without the greatest inconvenience, can votes be frequently taken on all the chief questions that arise? . . . But what I desire to point out, is that even where the machinery for weighing or measuring the popular will from week to week or month to month has not been, and is not likely to be invented, there may nevertheless be a disposition on the part of the rulers, whether ministers or legislators, to act as if it existed; that is to say, to look incessantly for manifestations of current popular opinion, and to shape their course in accordance with their reading of those manifestations.’
In view of the interest which is now displayed in public-opinion polls, the question arises, What progress has been made since Bryce wrote in ascertaining the state of public opinion? Have the mechanical difficulties to which he referred been resolved? Is it now possible for public officials to know from week to week and month to month precisely what public opinion is? In short, are we now entering a stage in the evolution of rule by public opinion in which the will of the majority is virtually ascertainable at all times? If so, what are likely to be the consequences of elevating public opinion to a stage where it ‘would not only reign but govern’?
We certainly live in a stage in the evolution of rule by public opinion when political authorities ‘look incessantly for manifestations of current popular opinion.’ One of these manifestations is the official election. As a device for ascertaining the state of the public mind, the election process, however adequate it is for the purpose of selecting officials, suffers from the fact that at best it is cumbersome, expensive, operates only at stated intervals, and seldom functions so as to give a precise indication of what the public actually thinks on major issues. The outcome of presidential elections, for example, may clearly indicate which of several candidates the public prefers, but give only the vaguest notion regarding the actual state of public opinion on selected issues.
Another indication of public opinion is the press. Every morning President Roosevelt and the heads of the national administrative departments find on their desks a publication called the Press Intelligence Bulletin, possibly the strangest newspaper in the world. Its one hundred and more pages contain an indexed digest of articles from 400 newspapers published in every city in the United States of more than 50,000 population and in the state capitals or most important cities of states having no cities that size. This journal has been called an ‘opinion seismograph,’ for it seeks to register trends of public opinion throughout the United States daily. In fact the press has been so generally employed as a means for ascertaining the state of public opinion that in many cases the word ‘ press ’ is often used as practically synonymous with ‘public opinion.’
The newspapers of a country do undoubtedly reflect the opinion of a large number of persons. But just how precise the correlation is between the editorial opinions of particular newspapers and a given public we do not know. And it is no reflection on the press to say this, for it has never assumed the responsibility for auditing public opinion periodically. Editorially it expresses its own opinion, which may or may not represent the current opinion of the time. Lucy Salmon in her definitive study, The Newspaper and the Historian, has evaluated exhaustively the reliability of the press as an index of public opinion, and it is unnecessary to repeat her analysis. It is only necessary to emphasize that the press, even at its best, is merely a manifestation of public opinion — a manifestation, however, which public officials watch most carefully.
A third indication of public opinion which serves to guide the actions and thoughts of public officials is the pronouncements of pressure groups. It would be a courageous public official who ignored, for example, the programmes and policies of such groups as the American Farm Bureau Federation, the American Federation of Labor, the United States Chamber of Commerce, the American Legion, to mention only a few. In spite of claims to the contrary, however, such groups cannot speak for the generality of the electorate. Most public officials realize this and seek to find out just how representative a particular group is. Like newspapers, pressure groups reflect in their opinions an indefinable approximation to public opinion in the sense of the opinion of the entire voting population.
These pressure-group manifestations of public opinion are, however, highly significant. Taken together, they may reveal a rather close approximation to the opinion of the electorate as a whole. In fact, as A. F. Bentley pointed out in his classic work, The Process of Government, an understanding of group activities may be the best key we have to an understanding of political processes generally. And if we assemble and analyze pressure-group programmes, even though we do not obtain a precise index of public opinion at the time, we may be anticipating by weeks, perhaps months, public opinion as it finally crystallizes in action.
There are, of course, other manifestations of public opinion. Congressmen as well as scenario artists have their fan mail. Politicians have their scouts tiptoeing in out-of-the-way places, interviewing, observing, cogitating. The expression, or lack of it, on the faces of those attending public meetings; the amount of applause which greets the announcement of a new tax programme; the response to appeals for campaign funds; the reactions of audiences to news reels; the compiled returns from precinct captains and ward bosses, all constitute factors of importance to the official ear.
For those with long experience in the political arena, these manifestations may have an esoteric value, not so much as indices of public opinion on vital issues, but as signs pointing the way to political success or defeat. Nevertheless progress in the art or science of identifying public opinion has, so far as public officials are concerned, remained practically stationary. Substantially the same methods are now employed as were employed by Alexander Hamilton and Andrew Jackson, and with about the same results.
Within the past two decades the possibilities of devising techniques for accurately ascertaining the state of public opinion have been explored by various institutions and private agencies. The motives prompting this activity are varied. In the first place, an announcement of a convincing statement of what public opinion really is has news value. As early as 1900, newspapers began to take preëlection polls on candidates, and in some cases attempted to ascertain opinion on particular issues. As newspapers generally grasped the possibilities of this practice, the results of such polls came to have an established place in the news columns of the press.
In the second place, convincing statements of what public opinion is have a propaganda value. Psychologists have acquainted us with the influence exerted upon wavering opinions by the so-called ‘impression of universality,’ and political campaigners as well as promoters of all kinds have been quick to seize upon the opportunities offered by straw polls to sway the ‘band-wagon’ vote. We have no way of measuring precisely the effect which is exerted upon public opinion by announcements that everybody is doing it, or that 65.33 per cent of the people are opposed to the sales tax. It is sufficient to know that party leaders act as if such statements had a propaganda effect.
In the third place, such information has a commercial value. Knowledge is not only power, but profit. Radio broadcasting companies, motion picture producers, advertisers, manufacturers, and producers generally have to conform more or less to the demands of the public. Hence the rise of market research agencies specializing in opinion census taking; the result, elaborate tables of listening habits, programme preferences, theatre attendance, and classified opinions for different publics regarding soap, automobiles, or the relative desirability of glass and paper milk containers.
And finally, knowledge of what public opinion is has an academic value. Considerable ingenuity has been displayed by academicians in devising questionnaires of varying degrees of complication. Usually the public to which such questionnaires can be and are submitted is small and in some cases unique, but the results, even though they do not give an index of a public opinion which is socially and politically significant, have a certain curiosity value, and may, if properly tabulated and diagramed, satisfy the requirements for a degree. This is not to imply that all academic researches relative to questionnaire technique are of slight value. Professor Thurstone’s discovery of a method for actually measuring degrees of favor and disfavor toward candidates and issues may pave the way to a better understanding of public opinion and the forces that are influential in its formation. But for the present these academic researches have not supplied us with techniques for continuous public-opinion auditing, such as would characterize Bryce’s fourth stage in the evolution of rule by public opinion.
One of the most elaborate experiments in polling public opinion is the Literary Digest poll, a venture which has done much to stimulate interest in the problem of opinion identification. The Literary Digest began its straw polls and forecasting activities in 1916. Since then it has conducted nine nation-wide polls, three on the question of prohibition, one on the nature of a presidential primary, two on the New Deal, and three on general presidential elections. In all of its polls the Literary Digest submits to its voters a postcard ballot containing a very limited number of direct questions to be answered yes or no. The selection and phrasing of the questions are the work of the editorial staff. These ballots are then sent to a carefully prepared list of persons, numbering, in the recent New Deal poll, more than 10,000,000. In making up this list the Literary Digest uses telephone directories and automobile registrations, taking infinite care to avoid duplications. These polls have established a surprising record for suggestiveness in predicting the outcome of elections, although, as Professor Robinson and others have pointed out, the polls have been affected at times by party bias. As a device for continuous public-opinion auditing, however, the Digest polls have distinct limitations. The undertaking is so tremendously expensive and so cumbersome that it cannot be resorted to frequently. And however advantageous to the periodical from the promotional point of view, its frequent use soon reaches the point of diminishing returns.
During the past year two experiments have been launched to test the possibility of applying to the practice of public-opinion census taking what is known as the proportional method of sampling. In principle the method of sampling is simple. Instead of attempting to secure a complete enumeration of opinions of those who constitute the voting public, an effort is made to secure a cross section of this public which will adequately and fairly represent the whole.
In July 1935, the magazine Fortune undertook to apply to ’factual journalism the technique of the commercial survey — a sampling of public opinion by methods long familiar to the industrialist in the sampling of ore or cotton.’ Every three months the editors select a list of questions which are carefully edited to remove ambiguities and indefiniteness. These questions are then submitted by special interviewers to a group of 3000 persons presumed to constitute a representative sample of the entire voting population of the country.
How is this group of persons selected? Having decided that the primary and determinative factors in the opinionforming process are (1) the geographical section of the country in which persons live, (2) whether they live in urban or rural communities, (3) their income, (4) their sex, and (5) their age, the editors proceed as follows: ‘The number of persons interviewed in each classification bears as nearly as possible the same ratio to the total sample as the members of that class bear to the total population.’ However, because of the commercial interests of the magazine, they do not always ’follow the census as literally as this in balancing the sample by economic levels, because Fortune is interested in opinions and buying habits.’ For example, in seeking to obtain a representative cross section in terms of the income factor, four economic groups are arbitrarily distinguished, — the prosperous, the upper middle class, the lower middle class, and the poor, — and the proportions included in each sample are arbitrarily fixed as follows: 10 per cent for the first group, 30 per cent for the second, 40 per cent for the third, and 20 per cent for the fourth. As the editors state, ‘The proportions are arbitrary from the very nature of the divisions desired, but they make sense when compared with available figures.’ Although in most cases the five factors mentioned above are made the basis for their samples, in some cases, and for particular questions, other factors such as race and religion are used.
It is obvious that the success of the Fortune polls depends upon the validity of the sampling procedure, for, as the editors state, ‘The balancing of the sample according to these variables is the most important element in the accuracy of the results.’
In October 1935, the American Institute of Public Opinion began the publication of the results of its weekly polls on national issues. Every Sunday in seventy-two newspapers throughout the country a full-page display of its findings appears, carefully tabulated, with comments by persons specially informed on the issues involved. The Institute itself is a commercial enterprise, supporting itself from the proceeds of subscriptions to its syndicated service. As in the case of the Fortune polls, emphasis is placed upon sampling, the attempt to find a group of persons fairly representative of the nation’s voters in matters of opinion. The Institute hopes, if successful, to lessen the expense and at the same time to secure a frequent and reliable census of public opinion throughout the country.
The research staff of the Institute, after consultation with the editors in whose papers the results of the polls appear, prepares a questionnaire containing a list of questions pertaining to one or more major issues of the day. Care is taken to frame the questions so that they will be definite, pertinent, and easily understandable. This process of framing and editing is one of importance, and may begin weeks before the ballots are finally printed. The questionnaire is printed on postcard ballots, to be mailed directly to the voters, and on blanks to be used by the Institute’s selected list of 250 interviewers throughout the country.
In selecting its sample of 100,000 persons to be questioned, the Institute endeavors to include a proper proportion of persons in different age groups; of those in different income categories, from millionaires to people on relief; of those living in cities and in the country; of Republicans and Democrats; and a proper proportion of men and women. It seeks to guard against errors by ensuring that so far as possible the distribution of returns, and not merely the distribution of ballots sent, shall be representative. This is important, for experience shows that as a rule not more than 20 per cent of the ballots sent are returned, and those so returned may not fairly represent the several classes polled unless care is taken to see that they do so.
In view of the importance attached to the method of sampling, it is interesting to note in more detail the procedure followed. In the first place, in order to secure a proper distribution of Republicans and Democrats, ballots are submitted to voters in selected counties in each state most nearly representative of the state as a whole in the vote for presidential candidates in 1932. In order to secure a proper distribution of ballots from urban as well as rural dwellers, at least two counties are selected from each state, one representative of the rural vote of the state in 1932 and one representative of the urban vote. In other words, counties are so selected in each state that they will represent fairly the party complexion and urban-rural distribution of the state as a whole. The Institute is now experimenting with other factors in an effort to improve further the representative character of its sampling process.
Allowances having been made for the above-mentioned factors in selecting the counties to be polled, the next step is to obtain in each county a cross section that will adequately take account of the income factor. To this end a fixed percentage of the ballots sent to a particular county will go to persons selected at random from lists of automobile owners, who, on the basis of experience, have been found to represent what may be called the middle income group. In order to obtain an adequate representation of persons in the higher income brackets, a definite number of ballots are sent to persons selected at random from lists of telephone subscribers. And the remaining ballots are distributed among those on relief or among those whose names appear on registered voters’ lists and have not been included heretofore. Finally, in order to assure a proper representation of ‘new voters,’ — young people, who must always be considered in the construction of a fair sample, — special care is taken to see that at least 15 per cent of all ballots returned come from this group.
To what extent have the experiments of the magazine Fortune and the American Institute of Public Opinion demonstrated that the mechanical difficulties to which Bryce alluded have been overcome? Have they at last discovered a technique whereby, without too great expense, frequent, periodic, and reliable polls of electoral opinion throughout the United States can be taken? The answers to these questions depend in large part upon the validity of their sampling techniques. How may we test the validity of the samples employed? No amount of statistical refinements in tabulating figures can compensate for an invalid sample.
In the first place, any attempt to determine arbitrarily what factors in the opinion-forming process are important enough to be taken into account in constituting a sample can at best be only presumptive and not conclusive. How can we be sure, for example, that age, income, residence, party affiliation, and sex are all the factors that should be considered, or even the most important? What about race, religion, occupation, and many other factors which in some instances may be of primary importance? In taking over into the field of opinion research the sampling technique used with so much success in the field of marketing, it should be remembered that on many social, political, and even economic questions there may well be factors to be considered which would be of slight consequence in attempting to sample a public’s reaction to commercial products and practices. Both the Institute and the magazine Fortune are aware of these difficulties and are seeking in an empirical fashion to discover what factors are of primary importance in the case of particular issues. Experience with commercial surveys and candidate polls has furnished many interesting hypotheses, but experience with noneconomic questions has thus far been too limited to justify imperative answers.
In other words, the ultimate test of the validity of a sample is not a logical one in the sense that sponsors of polls on particular issues can determine in advance the precise factors which should be taken into consideration in constructing a valid cross section of the voting population for polling purposes. But there are tests of another character which may be employed to advantage. In the case of market surveys the results can be checked against sales, and in the case of candidate polls, and some issues, against official election returns. If on repeated occasions there is a high degree of correlation, there is a strong presumption that the sample used was valid.
But what about issues where checks of this kind do not exist? There are two tests that may be applied. If, by using the same cross section, as the size of the sample is increased the distribution of returns tends to remain stable, or become more stable, there is a strong likelihood that the sample is valid. In the second place, if by taking a succession of small samples, using the same cross-sectional basis, the distribution of returns in each case is substantially the same, there is also a strong probability of validity. These so-called pragmatic tests are in the final analysis about the only tests of the validity of sampling procedures. The size of the sample is not the important thing, necessarily. The cross section used is also not the significant thing unless the final returns closely approximate what the returns would be if a complete enumeration were made. And the only way to test this approximation, in the absence of a complete enumeration, is to increase the sample and note whether the returns become stabilized, or take a succession of samples and note the degree of correlation among them.
The conclusions which we may draw are clear. As onlookers and in some cases subjects of experimentation we need to have a lively skepticism of statistical precision in these matters. We, along with the sponsors of these polls, must recognize the difficulties involved. At the same time we may properly color our skepticism with a certain degree of enthusiasm and encouragement, especially when we perceive that serious attempts are being made, not to claim for such researches and experimentation more than should be read into them, but to extend the frontiers of precise knowledge in this important field of public-opinion research.
Notwithstanding the difficulties yet to be overcome, these attempts to devise practical techniques for accurate and frequent public-opinion auditing are significant. In the first place, they demonstrate and emphasize the shortcomings of current methods for ascertaining the state of the public mind, particularly on issues of national importance. In these days of rapid social change, when new and important problems are arising continually, the need for frequent public-opinion auditing is indicated. Official elections, however, because of their infrequency and the vague and elusive qualities of party platforms, have come more and more to be devices for selecting public officials, rather than instruments for bringing to light the precise state of public opinion on specific issues.
Apparent also are the limitations of such devices as the press, representations of pressure groups, mass meetings, and esoteric schemes of canvassing and poll-taking whereby the public mind reveals itself spasmodically, and often in such a way as to place a heavy burden on the imagination of the legislator and administrator. Whether or not the new school of experts in opinion identification ultimately perfect procedures for accurate and continuous publicopinion polls, their efforts presage a more careful reappraisal of our traditional methods of making public opinion articulate.
In the second place, these newer experiments will probably evoke a more critical attitude on the part of public officials as well as the people generally toward the claims of groups and individuals purporting to speak the voice of the people. A public which reads discussions of the technical accuracy of a Literary Digest poll or an American Institute poll, and begins to think in terms of the adequacy of a sample, probable errors, and coefficients of correlation, is certain to become more and more skeptical of statements by individuals and interested minorities that public opinion is thus and so. In time it may be bold enough to ask, ‘What public are you talking about? Did you make a complete enumeration? If not, what method did you employ in sampling your public?’ Pressure groups of all colors and persuasions seek to identify their objectives with the public interest, striving to create the impression that they speak for the whole of the nation, or at least the intelligent portion of it. As the public becomes more statistically-minded the opinions of such groups will stand revealed for what they are, the opinions of those doing the talking. This wall be a gain.
In the third place, these efforts to improve the existing procedure of opinion identification may encourage a more widespread discussion of important public questions. Those who receive the weekly ballots issued by the American Institute of Public Opinion, whether they mark and return them or not, have many problems brought to their attention the importance of which they may have only dimly perceived. Even if they have perceived them as problems, they probably have not grasped the specific issues involved. How difficult it is to grasp these issues no one realizes better than the sponsors of these polls themselves. Political leaders and group leaders generally are often quite as interested in confusing as in clarifying issues. Whatever the reaction of the voter may be to the questionnaires he receives, he has before him as a rule a list of questions which not only may focus his attention upon important public problems, stimulate perhaps a modicum of thought, but also may serve to clarify the issues involved and to that extent further an intelligent decision.
Progress in the art of opinion identification may sooner or later raise questions not only difficult to answer, but at the very centre of the problem of democratic government. In the first place, what of the effect of such polls on public opinion itself? Will public opinion’s new rise to power sober and chasten it, or will the knowledge of its new rôle make it more susceptible than ever to emotional appeals, thereby accentuating the tyranny of mere numbers, and leading to even more violent swings of the pendulum than heretofore?
Again, how may the dangers of fraud, manipulation, and corruption in the use of these new polling devices be obviated? Their value as instruments of propaganda is generally recognized. To the extent that individuals are influenced in their thinking by what others think, to that extent pronouncements that 75.9 per cent of the people in the United States favor or oppose a given programme of action exercise their influence — hence a strong incentive to make such pronouncements, when it is to the advantage of a particular group to do so, regardless of the accuracy of the data upon which they are based.
No one, I believe, can fairly question the honesty of purpose and painstaking care that accompany the polling efforts of such agencies as the American Institute of Public Opinion and the Literary Digest. They have taken every effort to avoid the danger of ballot stuffing and manipulation. They desire to find the truth, if for no other reason than a commercial one. But they are dealing with instruments of power. And in other hands our reverence for statistics might be used against us. The dangers are not merely those of crass fraud. There are an infinite number of ways, some of them very subtle, for using these techniques for private rather than public advantage. The selection of questions, their phrasing, the timing of the polls, as well as the statistical treatment of the results, are phases of the problem that present difficulties. In the case of polls on candidates, the results can be checked against official returns, and frequent errors may of themselves tend to discredit the enterprise. But in the case of polls on issues there exists no such obvious check, not even the check of competition. For it would be only the most extraordinary coincidence that would enable the public validly to check one poll against another. This is a serious problem, and will become more serious as the techniques are perfected, or appear to be perfected. It is of course possible that agencies now engaged in the task of opinion auditing will discover some method to protect themselves against the unscrupulous. Otherwise further progress will be retarded, if not altogether obstructed, by the loss of prestige that follows repeated charges of corruption and manipulation.
Recently there was introduced in Congress a bill designed to prohibit such polls of opinion as are now being conducted by the American Institute of Public Opinion and the Literary Digest. Like many other efforts to mitigate abuses, this attempt to outlaw experiments in opinion identification would simply destroy the wheat with the chaff. When used merely for propaganda purposes, such polls are no more and no less reprehensible than other means that may be employed to further a given cause. Conceived, however, as experiments designed to test the validity of Bryce’s assertion that continuous public-opinion auditing is desirable but impossible, they justify a more intelligent kind of treatment.
Others assert that public-opinion polls are so vested with the public interest that the government itself should conduct them. Would it not be desirable, for example, for the Bureau of the Census to include among its activities frequent and continuous censuses of public opinion? This might be done in connection with its present extensive census-taking activities, and, as sampling techniques are improved, they might be used to supplement more extensive enumerations. Periodic reports on the state of the public mind may be quite as significant and valuable for public purposes as monthly and yearly reports on population density or the number of radios per family. Such polls would assist legislators in formulating general legislative policies, and also serve to guide administrative officials in the exercise of their discretionary powers, as, for example, in the conduct of foreign affairs. There seems to be no valid reason why the government should not undertake such experiments, provided care was taken to protect the public interest as over against the interests of minority groups seeking to exploit such services for private ends. Protection might be afforded by establishing an omni-partisan advisory board composed of legislative and administrative leaders to define explicitly the policy to be followed in selecting questions, conducting the polls, and utilizing the results. Inasmuch as such polls would probably have a prestige greater than that possessed by private undertakings of a similar character, care should be taken to indicate precisely the limits within which such polls could properly be regarded as accurate.
What is likely to be the effect of more accurate and continuous public-opinion auditing upon our representative system of government, upon elections, and upon public policy? In the fourth stage which Lord Bryce envisaged, the need, he said, for electoral machinery and representative bodies would disappear. Perhaps we do not need to concern ourselves with such remote possibilities. Certainly we are far from realizing those conditions which Bryce stipulated as prerequisites for such a fourth stage. Nevertheless, we may not ignore the fact that the results of such polls may influence public officials, especially if they rely on their accuracy. But how will such results influence them? Will they tend to undermine their sense of responsibility for leading public opinion? Will they tend, perhaps, to encourage public officials to follow courses they think unwise? Or will they tend to encourage more effective leadership, more strenuous efforts on the part of those in authority to inform the public and persuade it to follow intelligent courses of action?
If we are indeed entering this fourth stage we should carefully consider just how far and to what extent we really want public opinion to govern. After all, the fourth stage in the evolution of public opinion may prove to be, not a stage in which public opinion governs as well as reigns, but a stage in which it exercises its power more effectively through the wise choice of representatives and leaders.