CHAPTER 5

Wishing and Choosing
THE WORD "EXPERIENCE" CONTAINS AN IMPLICIT EXPEC- tation: the assumption that past successes or failures are yardsticks for the future, as well. Certain materials will continue to wear well, friends will continue to be amicable, committees will continue to squabble, and so on. The moment, however, we put this tacit assumption to the test, we find that many of our "experiences" are based on a very small number of instances, or on only a single one. For instance, the man whose washing machine is faulty and who wants to replace it with a new one, may be said to know infinitely more about the old than about the new machine (the ratio of his knowledge about them is 1 : O), even though his original experience was purely negative. However, this purely negative conclusion may be false for he may not have known how to use the original machine correctly or for the purposes for which it was designed. In any case, he would be quite wrong to generalise from an exceptional case.
This is a difficulty encountered in any one of countless decisions that face us with alternatives. If the decision, i.e. a particular purchase, is important enough and if we have time enough, we may consult someone more experienced, but generally no such person is on the spot, and we are forced to rely on the salesman or some other interested party. Only if the salesman looks particularly unreliable do we generally ignore his advice and rely on our own experience, however meagre.
Freedom of choice
Pseudo-experiences, i.e. experiences based on insufficient evidence, play a very important role in political, and particu- larly in economic life. It is due to them that mankind makes so
164
WISHING AND CHOOSING 165
many false predictions and is so often disappointed even where it would be relatively simple to have genuine experiences and to form correct judgments. A very good example is our ignor- ance of the quality of most of the goods we purchase in shops. While the state often relieves us of worry in that respect, for instance in the case of dangerous drugs, in the assaying of metals, and in seeing that weights and measures are standardised, it generally prefers to allow us to learn by our mistakes. Only in technical colleges and particularly in domestic science colleges are young men and women taught anything at all about the quality of materials, and most men, no matter how educated, will make their way in the world without the slightest inkling that, as far as common purchases are concerned, they are at any swindler's mercy.
Undoubtedly, this not an accidental fault of our educational system, but reflects the democratic wish that human beings must be allowed to choose freely, at whatever risk to themselves. And every choice contains a risk: the risk of choosing wrongly. The less experienced a man, the more his choice becomes like a gamble — he must trust to luck, to price differences, or to the good faith of others.
The wide variety of choices is most certainly one of the greatest attractions of large shops. True the choices are not quite as large as some would have us believe, for consumer research has shown clearly that we shop according to fairly rigid patterns. The vast majority of customers are conservative and buy what they have always bought. This is just another reason why purchasers have very much less experience than they usually believe. Freedom of choice is widest in women's dress shops, but here fashion rules too supreme to leave very much room for individual decisions.
Client motivation
Nevertheless, the range of consumer goods that is subject to sudden changes of demand is large enough for thousands of manufacturers and merchants to ask themselves year in and year out what it is that the public really wants. The correct answer to
166 CHAPTER 5
this question is a matter of life and death for any progressive business, particularly in times of prosperity and growing individual incomes, when large sections of the population buy goods that they could not afford before, and thus face entirely new choices.
During the prosperous twenties, industrialists, and American industrialists in particular, began to take a keen interest in this subject and their interest did not diminish when times became so bad that salesmen had once again to go chasing after customers.
Surprisingly enough, neither manufacturers nor merchants knew very much about the public's real needs. To bridge this gap, a new science — consumer research — ^was bom between the two world wars. Experts began to investigate what goods were in demand, which markets were saturated, which sections of the population were potential customers for which products and what individual wishes they had.
In the course of a few years, an entirely new technique for answering these apparently trivial, but in practice fairly complex, questions was developed. Whole mountains of questionnaires were sent out — some quite straightforward, others less so. Thus, General Motors did not enquire what cars their potential customers would most like to own, but presented an illustrated list of technical innovations and asked the public which of these interested them least.
Written questionnaires were frequently supplemented by personal interviews, and most firms used all their ingenuity to consult the largest possible number of people. However, after some time it emerged that more significant results could be obtained from interviewing a relatively small number of care- fully chosen people rather than a very large unrepresentative mass. Quality was found to be more important than mere quantity — where quality did not of course mean intellect or education, but representative opinion.
But what are the criteria whereby a given individual's opinion is judged to be representative of a large group.? Though this question has been discussed time after time and at great length, no one to this day can tell you precisely what the ideal guinea pig must look like. The beloved "man in the street", the "average American", the "average buyer", and the "average
WISHING AND CHOOSING 167
driver", are mere phantoms with no real existence. It is the experts' job to find people who resemble this phantom as closely as possible, i.e. people whose tastes, inclinations, and wishes best reflect those of the public at large. In this way the organisers of questionnaires can save much time and trouble, time being the more important factor of the two. A car manufacturer who is about to launch a new model, cannot afford to waste time on analysing a mass of questionnaires, since, by the time he has tabulated all the results, the new model may well have become outdated.
All such questionnaires are, after all, simply an effort to take stock of public opinion at a given moment, and not at some future date, though the results are, of course, to be used for practical actions in the future. It is here that the main difficulty of consumer research lies, for if the results of questionnaires lead to mere short-term prognoses, the enquiries must be repeated at frequent and regular intervals, which involves a great deal of labour, expense, and delay.
Straw votes
In about 1930, consumer research methods were applied to American politics, as well. Since elections are not normally held every day, not even professional politicians can always tell how the electors feel during the interval. The circulation figures of politically biased newspapers cannot help them here, for the political complexion of a given paper is no true guide to its readers' opinion. Since American political parties only really spring to life during elections and play a small part in public life at other times, little is known about a given candidate's chances before nomination. Candidates and their agents are therefore most anxious to test the mood of the electorate before they rush into very costly campaigns. For instance, they would like to know which seats are safe, and which must be made safer by increased propaganda. But even the electors themselves, and particularly those who are not very interested in politics at normal times, take a sudden interest in their fellow citizens' opinions and in their candidates' chances as the election date approaches.
168 CHAPTER 5
Newspapers and journals have always tried to satisfy their readers' curiosity in this respect, publishing forecasts weeks or even months before the poll. Since the readers were not particu- larly interested in mere editorial opinion on this subject — in democratic countries, an editor's voice counts for no more than that of Tom, Dick, or Harry — the papers concerned had to organise public opinion polls, or "straw votes".
The first "straw vote" was organised by the "Harrisburg Pennsylvanian" before the 1824 Presidential Election.^ At that time, communications were such that it was impossible to get anything other than local opinions in good time, and the "straw votes" were therefore bound to differ widely from the actual election results. The public looked upon them as a kind of parlour game in which they participated with relish though with little faith. Many years later, when the great New Tork Herald organised a similar poll, the predictions were still so wide of the mark that they might just as well have been drawn out of a hat.
But as the numerical gulf between Democrats and Repub- licans began to narrow, and as winning an election was often a matter of only a few votes, papers tried desperately to satisfy their readers' curiosity in more reliable ways. Elections results had become almost a lottery, and more and more people had begun to gamble on them. Wall Street would give the odds every day, and everyone had a glorious time — except, of course, the losers. Things changed again in the twenties when during three successive presidential elections — 1920, 1924, and 1928 — the Republicans emerged so obviously victorious that the out- come seemed a foregone conclusion, and the odds offered ceased to be worth taking. Instead people began to bet on what percentage of the votes a given candidate or party would obtain, and a great deal of money changed hands in those days of general speculation.
The press was keen not to let its readers down, and plied them with well-meant tips. Before the 1928 elections, no less than 85 newspapers and journals made private enquiries, generally by means of questionnaires. Others preferred to simplify the procedure by ringing up every tenth or twentieth
1 Albert B. Blankenship: Consumer and Opinion Research (N.Y.- London 1943), p. 4.
WISHING AND CHOOSING 169
subscriber in the local telephone directory, while yet others sent out special reporters to interview voters at home or in the street. The principle common to all these enquiries was that the more people were interviewed, the better the results would be. In contradistinction to consumer research, opinion research stressed quantity rather than quality — after all, everyone over twenty-one years of age was equally entitled to cast his vote. In fact the questionnaires did anything but reflect the real opinions of the total electorate. Telephone enquiries, for instance, completely ignored the many people who were without telephones, and questionnaires, which were generally sent to regular newspaper subscribers whose addresses were easily obtained, but who usually belonged to the middle classes, told one little about the intentions of, say, the workers. In conse- quence, the electoral chances of the conservative candidates were generally over-estimated — this happened not only in American elections.^ Still, the erroneous belief persisted that the greater the number of questionnaires, the more accurate the results would be. The record was held by the American monthly Literary Digest which sent its readers millions of postcards with short and pointed questions before every election, and received many hundreds of thousands of replies. In fact, in 1932, before Roosevelt's first electoral victory the Literary Digest's forecast was out by only 1%.
The Gallup Poll
In view of such striking achievements, it seemed rather impertinent for the young American journalist George Gallup to claim that large numbers were irrelevant, and that equally accurate or better predictions could be made with a small but very carefully selected sample of the population and a small team of skilled interviewers. While Gallup's approach may not have been entirely new, there was nevertheless a radical differ- ence between the old method of asking a few hundred people which vacuum cleaners they preferred and predicting the polit- ical opinions of the U.S. electorate by similar methods. ^ H. J. Eysenck: Uses and Abuses in Psychology (London 1955), p. 15S.
170 CHAPTER 5
For while the manufacturer merely takes the public pulse and may or may not act accordingly — it is, after all, an open question whether his actual successes were due to his acting on the advice of opinion polls, or to other factors, e.g. good ideas, good organisation, good marketing, etc. — ^political opinion polls stand or fall by the results, and the organisers must take full blame for their errors.
No wonder, therefore, that George Gallup was met with great reserve when he suggested to a number of American newspapers that they apply his system to the 1936 presidential election. It took him a long time to convince the editors that his system was much cheaper than the customary mass enquiries, and that surprisingly accurate predictions with his method had already been obtained, using a single interviewer and a mere 200 samples.^
In the end, 35 newspapers agreed to subscribe to his American Institute of Public Opinion, with the proviso that if his predic- tions about the presidential election of November 1936 proved less accurate than those obtained by the tried method of the Literary Digest, he would have to refund the entire cost of the investigation.^ Fortunately for Gallup, it never came to that, for although the Literary Digest broke its own record by obtaining two million replies to its electoral postcards, its prediction was out by 19% while Gallup's prediction was out by less than 1%. Moreover, the Literary Digest had predicted that the Republican candidate A. M. Landon would obtain 56% of the total votes cast, when Franklin D. Roosevelt was, in fact, re-elected with an unusually large majority. Gallup was one of the few political prophets who had predicted this result correctly, at a time when most political experts were convinced that Roosevelt's New Deal was so unpopular that the Republicans were bound to oust him.
Because of this striking achievement Gallup's name was suddenly on everyone's lips. Not only was he the prophet of the moment, but it was generally conceded that he had founded a new and most important scientific method of prediction. The American Institute of Public Opinion was suddenly showered
1 George Gallup: A Guide to Public Opinion (Princeton 1944), p. 16 f.
2 Pierre LazarefF: L' Opinion publique: comment la diceler? Les Annales (Paris) January 1956, p. 26
WISHING AND CHOOSING 171
with money and commissions, and the "Gallup Poll" became a generic concept for public opinion polls the world over. That same year, a British Institute of Public Opinion — the British Gallup Poll — was founded under the auspices of the News Chronicle, and France, Australia, Canada and Sweden quickly followed suit. In all these developments, the American mother institute not only took pride of place, but continues to play the leading role.
Red beans and white
What are the secrets of Gallup's success, and within what limits can his system operate efficiently?
George Gallup, who likes to romanticise his own career, tells us that the whole idea came to him while he was a student of psychology at Iowa University. During an experiment with a bag of red and white beans, he was struck by the fact that, twice in succession, he pulled out handfuls of beans whose colours were in the same ratio as those of the entire bag (3 : 7 ) . Beans have therefore played the same role in Gallup's life, as the famous apple is said to have played in Newton's though, strictly speaking, there is a difference, since Newton's discovery was the result of an inevitable physical process, while Gallup's was not. Had he put his hands into the bag a few more times, he would have been most unlikely to have continued to pull out red and white beans in the same ratio. Thus Gallup's original discovery was sheer luck, and luck is precisely what he insists must be obviated in scientific polls.
In fact, it was the older methods of predicting election results which had acted as if the whole enquiry was a giant bag of red and white beans mixed in a fixed ratio, and that, short of counting the entire contents of the bag, it was best to count as much of it as possible. This sounds very logical but is an error in fact, for even if as many as 50% or even 66% of the beans had been counted, little could be said about the rest — the beans may not have been thoroughly "shuffled". The whole process is reminiscent of the old story about the peasant who, having been paid ^100 in cash, counted the first 70 notes and then put all the
172 CHAPTER 5
money in his pocket saying that since the count was correct so far, he had no reason to suspect the remainder.
Unlike these earlier prophets, Gallup based his investigations on sociological rather than purely arithmetical considerations. To him, the electorate was not an amorphous mass of Repub- licans and Democrats, as the election figures might have led one to believe. In reality the electorate was made up of different social strata with different political trends. Thus farmers vote differently from industrial workers, the North votes quite unlike the South, Negro preferences differ from those of the Whites, employers have other interests than employees, and so on. Then there are differences of age and sex, for it appears that older people and most women usually vote for conservative parties.
We could continue this list of biological, economic and social differences which affect election results adinfinitum. For practical purposes, however, some of the finer distinctions must be ignored, for though the census figures show the composition of the electorate according to age, sex, vocation and to some extent according to income, there are other factors for which we have no up-to-date statistical data. To carry out comprehensive private investigations of all these factors before every election would take up far too much time and money.
It is therefore quite impossible to construct a true mirror image of the entire electorate in the opinion poll laboratory, but even a slightly blurred image is always better than no image at all. Gallup therefore samples his subjects mainly according to six factors: state, size of community, age, sex, income, and political affiliation. Other factors may also be taken into account from time to time, e.g. the national origins of electors during the last war.
Only when the structural composition of the electorate has been adequately reconstructed, can the purely arithmetical question of how many people in every state, in every income group, etc. must be interviewed, be solved. Once this is done, probability laws take over, and the more people are interviewed the more exact the estimates will be. However, above a certain maximum, the accuracy increases by no more than a fractional percentage. Whenever errors of 1-2% are permissible, a few thousand questionnaires will accurately reflect the opinions of the total U.S. electorate.
WISHING AND CHOOSING 173
Predicting elections today
Gallup maintains that, with correct sampling, the predictions made on the basis of only 100 questionnaires would not be out by more than 15%, those made on the basis of 900 question- naires would reflect electoral opinion to within 5%, while 10,000 questionnaires would involve a margin of error of only 1-5%.^ Beyond that figure, the margin of error decreases so insignificantly that, in election forecasts, it may be safely neglected. Before the 1936 elections, when Gallup was still on probation, he interviewed no more than 3000 out of 80 million eligible and 45 million actual voters. In other words, though only one in every 27,000 eligible voters (or one in every 15,000 actual voters) was consulted, the results were excep- tionally accurate.
True, not all Gallup's predictions were as accurate as that, for of the 114 election forecasts made by the Institute between 1936 and 1944, only 19 were out by less than 1%, S9 were out by between 2 and 3%, nearly half were out by more than 3%, and six were out by as much as 10-15%.
The Gallup system had its greatest setback during the 1948 presidential election in which Truman and not Gallup's favourite, Dewey, was elected. Actually, the error was not all that great, for though Truman was 9% ahead of Dewey, two outsiders had obtained 10% of the votes so that the victor did not gain an absolute majority over his opponent — a rare event in U.S. politics. Before Eisenhower's election in 1952 and re-election in 1956 there was little doubt about the outcome, and Gallup did in fact make fairly accurate forecasts.
The American results were excelled by those obtained by the British Institute of Public Opinion which, using the same methods, managed to forecast a number of Parliamentary elec- tions within a margin of error of only 0-5%. In France, too, the "Institut Fran^ais d' Opinion Publique" managed to forecast parliamentary results with a margin of error that rarely exceeded 2%.
Having had more than twenty years experience of Gallup polls, we may say that Gallup's method of sampling the electorate has proved most successful. Despite a few failures, the results go far ^ G. Gallup, op. cit., p. 17.
174 CHAPTER 5
beyond any previous expectations. Before Gallup, political predictions were no more than shots in the dark, and it is thanks to his achievement that we can nowadays speak of scientific fore- casts in this difficult field. This is a significant advance, indeed.
The art of prediction
No doubt, current methods of taking the public pulse are capable of vast improvement. While Gallup and his disciples have written a great deal about the basis and workings of his methods, they have kept the details of their sampling procedure a strict trade secret. But even if this secret were ever wrested from them, they have little to fear from unfair competitors, for election forecasting is an art rather than a strict science.
And Gallup is an excellent teacher of that art. He has per- sonally trained a large staff of assistants and can throw an army of interviewers into any strategic point, if speedy analyses are needed. In the United States alone, there are about 1,000 inter- viewers, mainly part-time, who must not only know how to ask questions but must also be very familiar with local conditions, in order to select their samples. The Institute merely indicates how many persons in every category have to be interviewed, and the final choice is left to the interviewers' sole discretion, pro- vided that they do not question the same voter more than once a year.
Since the personal qualities of interviewers are of the utmost importance, there is always the danger that, no matter how care- ful the choice of personnel, a subjective note may be introduced into the results. Governments try to avoid this pitfall by prescribing to Census officials exactly whom they are to inter- view, but this merely shifts the problem to a higher level.
Moreover, official sources generally prefer "area sampling" to Gallup's "group sampling" and the choice of area is, of course, a matter of personal judgment. Area sampling is, in any case, the costlier and more laborious of the two methods, and for that reason alone avoided by most Private Opinion Poll Institutes.
It is difficult to say which of the two is generally preferable since, apart from election forecasts, the accuracy of predictions
WISHING AND CHOOSING 176
can be evaluated in only the rarest of cases. In commercial enquiries the predictions can be checked to some extent at least against sale results, but in social enquiries no such criteria exist. Despite this fact — or precisely because of it — Americans have tried to apply Gallup's technique to every possible sphere. Thus polls are taken not only of public opinion about proposed legislation and about the international situation, but also about belief in the immortality of the soul and the Day of Judgment. There is no means of making sure how representative the replies are of public opinion as a whole, since it is unlikely that any government will hold referenda on these topics.
Is democracy endangered^
Many opinion polls are simple investigations rather than fore- casts, let alone methods of propagating ideas. Nevertheless, they may have a persuasive effect and considerable repercussion on the future. When Gallup avers that 75% of the American people believe in life after death, this may go a long way towards persuading some of the sceptics, and when Kinsey in his famous investigation of the sex life of the American adult (which was based on the Gallup system )i concludes that 50% of American women have pre-marital relations with men, the other 50% may be inclined to become less chaste. Such effects are un- avoidable, particularly in a country as conformist as the United States.
"A whole town cannot lie", Sartre has an American Senator say in one of his plays — least of all, we might add, if its opinions are confirmed by a Gallup Poll. We need, therefore, not be sur- prised to learn that most of the people's elected representatives pay great heed to opinion polls, though some are too proud to bow to the popular will. Still, if they will be pig-headed, they must not be surprised if they are thrown out of office.
It is for this very reason that Gallup Polls have been decried as exerting an improper influence on American politics. Gallup himself has hotly contested this imputation. Polls, far from
^ A. C. Kinsey et ah Sexual Behaviour in the Human Male ( Philadelphia 1948); Sexual Behaviour in the Human Female (Philadelphia 1953).
176 CHAPTER 5
endangering democracy, may well be a democratic method of enabling people to have their say even outside the voting booths. Such a system, has, in effect, been used successfully in Switzerland, where all important legislation is decided by popular vote.
Meanwhile, public opinion polls are increasingly being applied to non-political events, even outside the United States. Thus the church of St. Sulpice, one of the largest in Paris, called on the Institut Dourdin (which uses Gallup's methods) to organise a poll on "What non-believers think about us". The result of the poll was published in the church magazine and was made the subject of three sermons by the priest, who assured his readers that he had derived great spiritual benefit from the replies. ^ The Institut Dourdin, however, also plays a more secular role, and has organised discreet polls about the popularity of a number of stars of the stage who were worried about their next public appearance.
But no matter what the subject of the enquiry, all such polls are based on the same principle: correct sampling, and the assumption that public opinion will not change radically until, say, the next election, referendum, or other specific event. Only in very rare cases is it assumed that public opinion will remain unchanged for a longer period.
Opinion polls therefore predominantly aim at short-term prognoses, and generally make no attempt to predict public attitudes in, say, two or three years' time. Even so, periodic questionnaires on specific subjects may well provide the key to trends of opinion, and thus lead to successful long-range fore- casts. In this way public opinion polls would cease being a static and become a dynamic technique instead. ^ Les Annales (Paris), January 1956, p. 28.