Every general election that mobilizes millions of voters to consecrate new authorities is usually surrounded by media rituals when the votes are counted. One of them is the debate on the definition of the popular mandate type chosen by the majority, whether for continuity or change, for profound changes or only minor adjustments. Another healthy ritual is usually the exaltation of the democratic method as the most transparent and civilized way of anointing authorities, at least in the face of its historical alternatives in our region.

More recently, a new ritual has emerged that basically rejoices in highlighting the size of the error between the predictions made by voting intention polls and the results finally obtained by different candidacies: “Once again, the polls failed”. We saw it in the recent general primary election in Argentina; in the middle of the year when the president was elected in Guatemala; last year in the presidential vote in Brazil and in the estimates on the fate of the Chilean constitutional plebiscite twelve months ago. The examples are proliferating and even impacting some entities defending the democratic process. This is the case of the OAS observers to verify the Guatemalan election, who did not hesitate to put equal emphasis on the irregularities of the justice system in vetoing candidacies, on the violence that surrounded the elections with murders of candidates and voters, and on the distance between results and polls’ forecasts: they equated the three factors as if they were homogeneously responsible as distortions contrary to the proper functioning of democratic institutions.

However, it would help to understand that the polls have strayed much less than they are accused of if three principles were made clear. One of them is that polls taken in our countries, where voting is compulsory, seek to represent the entire population in a position to vote and, thus, project intentions of the total number of voters without considering that on election day many will abstain from going to the polls. The recent abstentions in Brazil and Argentina, which vary between 20% and 30%, omit an enormous volume of votes that are counted in the surveys but which, by voter choice, end up avoiding the ballot boxes.

This electoral abstention is not distributed homogeneously among supporters of the different contenders. The classes with less education and financial resources, as well as the generational extremes (the youngest and the oldest), tend to exhibit an absenteeism rate that is double that of the higher classes or more educated sectors or of adults at the peak of their life cycle. To the extent that public opinion polls systematically indicated that Lula’s electoral base was strongly concentrated in the lower classes, this higher abstentionism on voting day significantly reduced his vote share. In fact, the polls did not deviate almost at all from the total votes obtained by the then-president-candidate Bolsonaro, but they did calculate more votes than Lula finally got. In the case of an open primary such as in Argentina, this difference between pollsters and voters is even greater.

The second principle is that surveys ask about intentions. We all have a friend or relative who smoked and promised to stop smoking but could not keep his or her promise. Behavior is only observable when it is executed. Before that, we only have projections anchored in admitted preferences. It was always like that. Only in the golden age of electoral polls, from the 1980s to the 2000s, when they were accurate in their projections, the influences that could skew the path between intention and behavior were much less present than they are today. The sources of such deviations today are of a different nature, although the decentralization of the media thanks to the consumption of social networks is the main suspect. In the past, rumors, fake news, and even genuine breaking news against a candidate took a long time to spread among the population. Today they circulate massively in seconds.

On the other hand, the stabilizing forces of preferences had a considerable weight. Political parties organized people’s political identities, as well as their communities of belonging, such as churches, neighborhoods, ethnic groups, unions, or social classes, which gave them a lasting identifying anchor, often transmitted through the family. Today, the generalized institutional distrust punishes parties and unions more severely.

The main effect of this is the voluminous electoral indecision until the last minute. Two weeks before the Argentine primaries, polls indicated that 2 out of 10 voters did not know who to vote for. In the Brazilian case, a little more than a week before the first round election, there were still between 6% and 9% undecided voters in a context where the final difference between the candidates was 5%.

The last principle is that polls reflect how society absorbs the offer coming from the political world: in that sense it is a portrait produced and potentially volatile to the sway of the stimuli present until the last minute. Polling professionals do not resign themselves to these possible changes and seek to triangulate and validate intentions by asking about the firmness of electoral preference, the probability of never voting for any of the candidates, the perception of who will win the election, the degree of alignment of their values with those of the candidates. All tools help to determine how stable the manifest intention is, but they also serve to estimate which way the undecided may move toward.

Each of these measuring instruments is based on a hypothesis. Asking about who will win the election assumes that, among the undecided voters, there may be a strategic final vote: simply to jump on the bandwagon of the victor. Asking about rejection assumes that there is an emotional-affective component in the vote that places personal sympathies or antipathies at the center of the decision, regardless of the programmatic proposals of the candidates. Asking about the priorities of society or the most important values assumes that there is a rational component, where the type of expected future is decisive in the choice of candidate. All these hypotheses served to project preferences in behaviors and simulate reactions of the undecided, but as long as there are no disinformative campaigns or the manipulation of emotions overrides the persuasiveness of rational, strategic, or emotional arguments. By demonizing Lula and the Worker’s Party and multiplying the fear of a possible communist wave, Bolsonaro’s supporters managed to get a critical minority of the electorate to sacrifice their programmatic proximity to the center-left or even their personal dislike for Bolsonaro.

It would be difficult to abandon the ritual of criticizing the polls and their deviations from the final results in the short term. But precisely this phenomenon reaffirms their centrality to contemporary democracy, not only because they continue to be increasingly sought after by all those who seek to influence the public sphere, but also because of their ongoing contribution to a more informed, transparent, and representative society.

_{*Translated by Janaína Ruviaro da Silva from the original in Spanish.}