As intelligent as smart, Leonel Brizola, during one of his innumerable campaigns, went out into the streets asking each passer-bye if he had already been interviewed for research into electoral opinion. “Neither have I (been)”, said he after listening, almost always, to a “no” from the interlocutor. Last month the Presidential candidate Geraldo Alckmin equally preferred to cast aside the numbers, classifying as a “joke” the results of a poll that was unfavorable to him. “I only speak about serious things, and this poll is a joke. Do you take this thing seriously?, he asked the journalists who had accompanied him. “At each election the controversy concerning the performance of the research institutes during the election campaign repeats itself, when it is questioned whether or not the numbers affect the vote etc. This occurs because of the difference of ethos that moves politicians, journalists and research coordinators concerning them. Candidates, for example, tend to discard them when the result is not good and vice-versa”, explains the political scientist Alberto Almeida, the coordinator for the FGV Opinion Poll and author of the book entitled, Como são feitas as pesquisas eleitorais e de opinião [How electoral research and opinion polls are done] (published by Editora FGV) and who is now publishing the study entitled, [Why Lula?].
In the end, can one have confidence in electoral research? Elementary my Dear Brizola and Alckmin, Sherlock Holmes would have replied, who in the book entitled, The Sign of Four, revealed the curious mechanism that moves electoral soundings. “Although the individual man is an insoluble enigma, a human grouping represents a mathematical certainty. Nothing can be predicted about what a man will do, but it’s possible to forecast the attitudes of a certain number of them. Individuals vary, but the percentages remain constant”, understood the detective away back in the 19th century. “I remember, when I was a child, having attended a Janio Quadros political rally during the presidential election of 1960. It was the last scene of a rural Brazil campaign. Twenty nine years later the country awoke from the military nightmare for the elections of 1989, urbanized and television maniacs. The campaigns moved from the wooden platforms to the television studies, governed by electoral polls, which, in theory, expressed the vagaries of public opinion, a statistical miracle that made one thousand people speak for millions”, observes the journalist Maurício Dias.
“Since the democratization, the research polls have indicated errors and right things in campaign tactics, redirected strategies, provided incentive for denouncements, forced alliances, provided elements for government programs and altered the candidates’ agendas”, analyzes the sociologist Antonio Teixeira Mendes, from the Brazilian Center of Analysis and Planning (Cebrap). “For all of this, they can no longer be seen as extras on the electoral scenario. Candidates suffer the direct impact of ‘the dance of the numbers’: good results in a poll mean political support, contributions, funding from business, perhaps more space in the media. The opposite can represent a loss of alliances, intra-party pressures, apathy, crisis and even desistence.” Or in the somewhat cruel definition from the American ex-President Richard Nixon: “Secret service agents protect the candidates against the psychopaths and the opinion polls protect them against the electorate”. In spite of all of this power, the polls are a mystery for the electorate and even for many journalists who, very often, wrongly interpret the ups and downs of the numbers game. “In truth, one doesn’t believe in the poll because one knows how it works, but simply because it works”, points out Mendes.
And one shouldn’t have any doubts about this. The average correct results of the big four Brazilian research institutions (Ibope,Datafolha, Vox Populi and Sensus) are of international standards: considering the numbers referring to the presidential elections starting in 1989, the average correct result is higher than 90%. Democracy thanks this precision. “To sum up, the polls create political facts, from the point of view that they reduce the cost of information as to what is politically occurring in the country. It has become an important phenomenon for citizens, helping them to decide, to reflect and to judge those that present themselves, recommending, desiring and asking for their vote”, observes the political scientist Wanderley Guilherme dos Santos. “These types of facts created by the polls are welcome, because they are an indispensable ingredient in turning democracy not only a collective benefit, but a benefit as cheap as it can be”, explains the researcher. Indeed there are those who consider the instrument as a “spectacularization” of elections, transformed into ‘horse races’ from which serious political discussion is excluded. “Even if this were true, we have to ponder that this conversion of politics into a sport makes the public pay more attention to the contest and, in this context, ends up being better informed about the candidates’ platforms etc. The electorate gain political information through the release and discussion of the polls by the media”, says Lydia Miljan, from Windsor University in Canada. Information is a precious benefit.
Indeed, if we agree that the ingredient is indispensable, one needs to understand in what manner it is used in the electoral ‘kitchen’. The first mystery that surrounds polls is the sampling. How can a few speak for so many and, in the end, everything be correct? “Size doesn’t mean quality”, replies Alberto Almeida. The mania for polls began in 1916 in the United States when the magazine The Literary Digest decided to send out millions of questionnaires to all of the electors they could manage to list (also with an eye on future subscribers) and got it right, with reasonable precision, the results of the presidential election of that year and onwards. In 1936 they repeated the dose (and they increased: they sent out 20 million questionnaires), with the research concept that consisted in reaching the greatest possible number of people. A year before, more modestly, George Gallup founded the American Institute of Public Opinion, where they developed a more scientific method of research, based on sampling. In order to conquer the newspapers he promised to forecast the winner of the 1936 election in which Franklin Roosevelt and Alf Landon were running. The Digest, which had sent out 20 million questionnaires, said that the latter would win the race. Gallup sent out interviewers all over America who spoke with 3,000 people and bet on Roosevelt, who won the race by a margin of 19 points. What had happened?
The Digest had sent their questionnaires to all car and telephone owners, an important sector but not representative of the country as a whole. “It didn’t help carrying out millions of interviews if those interviewed weren’t representative of the population. It’s better to carry out fewer representative interviews”, explains Almeida. As Gallup had done, based on quota samples, this is still the method adopted until today by the Brazilian research institutes. In it, the population is divided into sub-groups (men and women, blacks and whites, young and old etc.) and the proportional size of each of these groupings is calculated. From this, the total number of interviews to be carried out is defined, which will be divided according to the proportions found for each sub-group. The more variables that were controlled in the definition of the quotas, the better the representation of the research, or that is to say, its degree of similarity with the population. The sample size is independent from the population researched. This system is cheap, fast and works. In Brazil, it has been decided to work with a sample of around two thousand people. But what is that so-called error of margin?
Anyone who thinks that one is dealing with a hoax from the researchers, a manner of ducking out in the case where a forecast is not confirmed in the polling booths, would be wrong. “There always is the chance that those who were included in the sample are not truly representative of the whole population. Thus, the estimates are subject to differences between the data obtained by means of sampling and the researched population. If these were random differences, they’re called sampling errors, and can be calculated and controlled”, clarifies Márcia Cavallari, the director of Ibope. The margin of error states how close the sample statistic falls or is in relation to the parameter of the population. Onto the playing field comes the second ethos of research polls, as observed by Alberto Almeida, that of the media. “For the journalists, there’s nothing more frustrating than an electoral campaign in which the rates of voting intention for the candidates don’t change. Change is news and continuity is not. Often a candidate goes up from 30 % to 32% in two polls with a margin of error of 3 percentage points and by the statistics, there was no alteration. By the journalist’s ethos this is impossible to be believed, and probably to be accepted, and the headline of his newscast will be that candidate X went up 2% in the electoral race. The media prefers to run the risk of remaining with an incorrect hypothesis than to lose a headline”, says Almeida. For the research coordinator, for the third ethos in the scenario, the reverse occurs. The researcher prefers not to discover that there was a change in voting tendency than to accept an erroneous hypothesis that will place his trustworthiness in question.
But it is not enough to know how to choose, one needs to know how to ask. “The questions on a questionnaire are the center of the research project”, explains Márcia, who recommends that they be short, objective, clear and in an vocabulary assessable to any voter. There is even a joke about this. A young monk was severely chastised by his superior when he questioned if he could smoke while praying. Ask the question differently, a friend suggested. Ask if you can pray whilst smoking. “The formulation of the question affects the results of a poll and for this, very often, the results could be fabricated in a subtle manner. It’s common that the media don’t present the questions as they were made up, but only their summary. The citizen needs to be careful with this and always look to know the all the details of the questions”, says Almeida. “Polling is an excellent marketing instrument, assuming one is aware of its power and of its limitations”, emphasizes the director of Ibope. In the end, no matter how good the question is, it will come from a human. And to err is…
“It’s worth nothing to have the perfect sampling and a questionnaire without errors if the interviewers err or fraud the questionnaires. They’re only collecting information and this doesn’t give them the right to expose their own ideas or counter opinions. A well trained and professional field team is a key element in the success of whatever research institute”, evaluates the researcher. There are rules as to how to conduct the interviews that include instructions such as never to interview people accompanied by third parties (the tendency is to respond to what the companion would like to hear), not to interview more than one person from a group of people, no matter whether or not the group is formed by two or more people etc. An interviewer’s life is not easy. “During 1998 we carried out a poll in Rio de Janeiro. In the southern zone, for example, it was necessary to bring in female interviewers with the ‘right face’ so that the passers-bye would not segregate them, as they did with professionals of a more popular phenotype. Then in the shantytown of Rocinha, we were approached by armed men and those interviewed responded to the questions with their eyes on them, frightened”, recalls Almeida. The modality of field work, as a matter of fact, differentiates the Brazilian institutes.
fernando vilelaThe Datafolha Institute, established in 1983 as a research department of the Folha da Manhã newspaper group to carry out soundings of opinion with “technical rigor and agility”, prefers to research in the streets. “In our DNA we have the spirit of journalism. Besides, today it’s difficult to access luxurious condominiums and one can only enter into shantytowns after an agreement with the drug dealer”, observes Mauro Paulino, for 20 years the director general of Datafolha, which, starting from 1995, was transformed into a business unit. The Vox Populi, for its part, only carries out home interviews. “This increases the possibility of the control of flows in the field. The ideal research is that which gives the same chance of a member of a determined universe to be included in the sample”, believes Marcos Coimbra, the institute’s director. In spite of the conviction in the instrument of his work, the researcher has reservations about the real limit of opinion polls. “A voting intention is only an intention to vote. As yet we don’t have a large number of elections in order to understand the standard of behavior of the Brazilian elector. In its 506 years of history, this is the fifth consecutive direct election in Brazil”, he points out. Hence the ‘youth’ of the research institutes.
The pioneer of them, Ibope, was established in 1942 by Auricélio Penteado, the controversial owner of radio Kosmos, of São Paulo, who, said the apocryphal polls of that era, was the broadcasting station least listened to. Using the techniques developed by Gallup, Penteado quickly discovered that the future was not on the radio waves (effectively the Kosmos was a disaster), but in the growth of research. The institute’s first test was one of the most unusual: Ary Barroso, a Rio de Janeiro counselor, had disputed with Carlos Lacerda the best place to build the future Maracanã football stadium. The composer Barroso defended land at the Derby Club and Lacerda on the sand bar of Jacarepaguá. The composer of Aquarela do Brasil (Barroso) suggested that Ibope should talk to the voice of the people and he won the race with 88% of the citizens on his side. “Barroso’s action placed in suspension the idea of parliamentary representation, indicating, although indirectly, the fact that representatives of the population had no interest in knowing the desire of those who had elected them. The grand majority of the representatives, summoned in speeches maintain themselves restricted to the role of elector/consumer, indebted to choosing the political products elaborated by the political producers”, notes the historian Aureo Busetto, from the São Paulo State University (Unesp). National democracy grew with the opinion polls.
In 1953 came a greatest test for Ibope, capable of forecasting the victory, in Sao Paulo, of mayor Janio Quadros. Through the elected mayor, the success of the opinion poll would turn Brazil upside down. “Although unperceived by the majority of the electorate, the compromise between organs and professions of the press was known by political groups and by the intellectual elites. Thus, the electoral forecasts given by the press, before Ibope, were mere pieces of party or candidate propaganda”, says Busetto. “The prognostic precision of Ibope was an important step for the redemption and qualification of public opinion polls in the areas of study of political-electoral.” The reaction came with the Brazilian ‘jeitinho’ (the way Brazilians get overcome difficulties , and bend the laws). If before the newspapers had flung open doubtful ‘certainties’ and had contradicted themselves on their front pages (‘X has so many percent of the electorate and will win very easily from Y’), after the coming of Ibope, it was necessary to have greater care in order to please the reader. They invented dummy institutes, such as the Epil or the Ipê, which ‘guaranteed’ the scientific precision of the results, although the method continued to be the same: in the lead was the one who could pay the most or whose victory interested the newspaper.
“We carried out two basic pieces of research with the finality of establishing the ‘areas of knowledge and ignorance’ of the researched fact (the elections) and the study of public opinion in the face of the problems that were in need of a solution and the men who would be capable of resolving them.” This was the ‘method’ adopted by the Ipê, in the words of its director. This was only the beginning of problems for Ibope. During 1954, in the São Paulo state governor election, there was polarization between Janio Quadros and Adhemar de Barros, and everyone wanted to get the outcome of the contest right. Even Radio Record risked a guess: JQ. And, by an unhappy coincidence was the only one to get it right. Ibope had given the victory to Adhemar (who lost by only 1%) and fell into the jaws of the newspapers. “If say it simply, we would say Ibope is sloppy”, wrote the Diário da Noite newspaper, belonging to Assis Chateaubriand, who feared the scientific competition of Auricélio’s institute. He even went on to print a series of articles with a clairvoyant Armenian, Sana-Khan, who prophesized things such as ‘there are three Jupiter guided stars in Janio’s destiny’, a form of inciting Ibope and putting in question electoral research polling by sampling on saying that the supernatural was much better equipped than it.
It was difficult to recover the credibility of electoral statistics with the Brazilians because, in 1984, it suffered another calamity, this time with the Gallup Institute. Janio was once again on the scene, this time confronting Fernando Henrique Cardoso. For the Gallup there was a technical draw. “The political desire of the media was for a victory by FHC and they lost the bet. They made the electorate go mad”, observes Marcus Figueiredo, from Iuperj. The research institutes perceived that it was necessary to take care with the bearer and even more with the press, closing itself more in professionalism in order to guarantee competitiveness via competence. No matter how they were, the polls had arrived to stay. “Since the start of the decade of the 1980s, the political polls in Latin America have built themselves into a central event in the process of re-democratization”, says the political scientist Fabián Echegaray, from Market Analysis Brasil. They were, he observed, a key vehicle for checking the cleanliness of the vote and of discouraging electoral fraud, a practice common up until then in various Latin American countries. In cases such as those of Argentina and Venezuela, the release of the polls even prevented attempts at military coupes; they also demystified the dictatorial discourse about the exaggerated price to pay for the return of an elected and civil regime; and provided space for the people, at last, to express their desires and see them respected”, he analyzes.
It doesn’t always happen. It was based on opinion polls whose supposed data indicated the popular desire to break with the Constitution that in 1992 president Fujimori carried out a state coup in Peru. During the Reagan administration there was, says Echegaray, also a wave of fraudulent electoral research polls, sponsored by the American government in order to consolidate the interests of the United States in various regions. “One direct consequence of this bad use of research polls was the incorrect representation of the true political agenda and the poor knowledge about what were the priority questions.” Hence the inevitable question: do the polls affect the vote? “Yes. It’s rational that the citizen considers the option choice with the greatest chance of victory as well as the possibility of a second round. The curiosity is that the division of rounds was done exactly to provide incentive for people to express their real preferences, at least in the first round. In Brazil, at times, the elector anticipates the useful vote for this phase, without a doubt fed by the appreciation of the research polls”, believes Alessandra Aldé, from Iuperj.
There is even jargon for indicating this tendency: the bandwagon effect (literally speaking the opening wagon of a circus and indicative of the candidate with the greatest chance of winning, and, right away, attractive to the majority who prefer to follow the general movement and bet on the winner) and the underdog (the “pariah”, the C candidate, the ‘lantern carrier at the back of the line’ for whom one votes in order to jump start B, whom you want, and to force the definition of an election in the second round with A, whom you don’t want). The polls, therefore, will push the elector from one side to the other and must, for this reason, be submitted to controls, such as prohibiting their release during a period of time before voting day (in various countries they are totally prohibited or must be interrupted various weeks before an election). “An elector is submitted to a large volume of information about the candidates, denouncements, deeds etc., and will form an image of each one of them. All of this information will be filtered in accordance with the preferences, visions of the world and sympathies of the electors. One piece of information that could (or not) arrive to the elector is the poll. Notice: it is one among a vast universe”, evaluates Almeida. “The direct influence of polls is not something to be supposed, but proven. In the end, the elector can know the polls well and have doubts about their truth or simply ignore them. It’s not obvious that polls influence the vote and up until today there has been no scientific research that has proven this hypothesis. One could say that the well placed candidates in the polls have more resources, animate their party supporters etc., but only this.”
For Almeida, the same is true for the supposed manipulation of polls carried out by serious and well known institutes. “When this question is taken up, it’s necessary to generate public data that proves this. In the opposite case, this is mere suspicion, without any scientific base.” Professor Teixeira Mendes agrees with the researcher. “The evaluation of the public and the confrontation with the final results have turned manipulation into marketing suicide.” Equally, he does not see any possibility in the precise determination of the degree to which a poll affects a voter’s decision, although he recognizes that it has an impact on an election. “It’s enough for a candidate to go up in the polls to be then offered a bombarding from the others, which thus ends up with him falling in the voters intentions.” In the good sense of the word, the price of democratic liberty is eternal vigilance: understand the ‘dance of the numbers’ but don’t drown in them.