It was a week of surveys about voter intentions and assessments of incumbent administrations. Readers don’t need to have any doubts: We are in an electoral season, and the polls are here to stay. It is wise, however, to be cautious.
Folha published its own during the week. The main one was on Sunday, when the newspaper’s headline was: “Serra way out in front in So Paulo.” (It referred to Jos Serra, the centrist Brazilian Social Democracy Party, or PSDB, candidate for mayor). There was also an assessment of the administration of incumbent Mayor Marta Suplicy (of the left-leaning Workers Party, known as PT). Both polls were conducted by Datafolha.
On Monday, a Datafolha assessment of the administration of So Paulo state Gov. Geraldo Alckmin (PSDB) came out. On Tuesday, all the newspapers published Ibope surveys of voter intentions (Folha had: “Ibope showing Serra leads in SP”) and in Rio (“Maia ahead in Rio with 18-point lead, Ibope says”).
On Wednesday, Ibope had to explain. The survey that it conducted between June 15 and 18, commissioned by the PT, had a statistical tie among the three main candidates in So Paulo: Serra, with 22%; Suplicy, with 21%; and Paulo Maluf (a conservative former So Paulo mayor and governor), with 18%. The survey published on Tuesday was carried out between June 25 and 27, 10 days after the first one. It was commissioned by the Globo TV network and gave a different result: Serra was ahead with 30%, Maluf had 21% and Suplicy had 16%.
It is strange that such a big change could occur in such a short time. According to the institute, the big difference between the two surveys was the result of the influence of electoral programming on radio and TV. The PT survey was carried out after the party’s electoral program (Brazilian law gives all parties free air time), and the Globo survey was taken after free ads for Serra and Maluf.
On Thursday, O Estado de So Paulo published a survey by the GPP/InformEstado Institute, carried out at practically the same time as Ibope with a result that was little different: Serra had 33.1%, Maluf was 26.8% and Suplicy had 15.2%. The Datafolha survey, published by Folha on Sunday, also during the same period (July 24 and 25), showed other results for the PT candidate: Serra had 30%; Maluf, 24%; Suplicy, 20%.
There are numeric differences to which the institutes don’t give much importance because they consider that they are within the margin of error. The margin of error is determined by the number of voters surveyed. Datafolha polled 1,083 voters (3 percentage point margin of error); Ibope contacted 1,001 voters (3.1 percentage point margin of error); and GPP/InformEstado talked to 606 voters (4 percentage point margin of error). Besides this, it explained, the three surveys have different methodologies.
The series of surveys continued during the week. On Wednesday, newspapers published a CNI/Ibope poll which evaluated the administration of President Luiz Incio Lula da Silva. Here are headlines on the biggest-circulation dailies: “Lula’s approval rate falls from 34% to 29%” (Folha); “Ibope shows Lula administration in decline for 15 months” (Estado); and “Ibope: Lula’s popularity and level of confidence fall” (the Rio daily O Globo).
On Thursday, Estado published, along with a survey of voter intentions, an evaluation of the Suplicy administration. New discrepancies arose outside the margin of error. Datafolha and GPP/InformEstado made their surveys, as I said, with practically the same data: June 24 and 25 (Datafolha) and June 26 and 27 (GPP). In Datafolha, 29aid Suplicy was doing excellent/good, 37aid acceptable and 33aid bad/terrible. In GPP, 18.6aid she was excellent/good (10 percentage points less), 39.6aid acceptable and 41.5aid bad/terrible.
It is true that the surveys can’t be read as absolute truth: They are only snapshots at a certain moment. But the differences become obvious and are uncomfortable.
The institutes have methodological explanations for almost everything, but there are problems that can’t be ignored. Doing a survey shortly after the commercials ran by the party sponsoring the survey, at the very least, needs to be questioned.
This should be great for the parties because they give the impression of an advantage that we know is artificial and will not be maintained. But, for the institutes and, mainly, for the news media, it calls the credibility into question because they will be burdened with legitimate suspicions of manipulation. Newspapers should be concerned about providing their readers with all the essential information to understand the surveys and question them.
——————————————————————————–
Reading the polls
The best way to deal with the polls is to know how to read them. The Brazilian Association of Polling Companies (Abep) created a guide for this election that helps to understand how they are done. It has a recommendation to news organizations that they publish as much information as possible about their methodology and production. Respect for these norms does not eliminate the possibility of fraud or manipulation, but it diminishes the chances. The guide can be read and copied at Abep’s website: www.abep.org.
The most important recommendations are listed below. Folha has also released the obligatory registry number of the survey with the Regional Electoral Court.
“The public should have access to all necessary information to understand, interpret and evaluate the data published in a suitable way.
“The following are recommended concerning the publication and release of electoral polls:
a) Name of the polling company responsible for carrying out the work;
b) Name of the sponsor/commissioning party for the poll;
c) Population represented;
d) Size of sample and geographic coverage;
e) Dates survey was carried out;
f) Sampling error/margin of error (this information does not replace the necessity to specify the size of the sample);
g) Sampling method;
h) Method of collecting information;
i) Criteria for deliberation if they exist;
j) Whenever the results say something about a segment of a sample and not their entirety, this should be explicit;
k) Whenever the release of the results of a certain segment (for example, age range, sex, region, etc.) indicate the number of interviews which formed the results (base);
l) Explain the rate of “no response,” in other words, the percentage of responses who “don’t know,” whenever they can significantly affect the interpretation of the results;
m) Information about the number of cities, locations or points shown;
“The items listed above should be available in any survey released. Situations in which it is not possible to supply all the data (radio or TV, for example) should anticipate and indicate the available space to give minimum details (for example, the website of the polling company or company that is releasing it). Items a) through e) are obligatory in any type of release, whether in the print media or electronic.”
Translation by John Wright
