• October 2014
    M T W T F S S
    « Aug   Nov »
  • Pages

  • Marginalia

  • Accumulations

  • Advertisements

Sensus Con Sensus?


Update: Isto É magazine and Sensus are reiterating their prognostications today, 25 October, two days before the election.


… Although  Dilma had celebrated victory in 2 of 3 Brazilian cities in the first turn of the election, the sampling methods used by the marketing juggernauts, it is reported, were cities favorable to both candidates in equal measure.

After Ibope and Datafolha indicated a technical tie between  Aécio Neves (PSDB) and Dilma Rousseff (PT) in the search for a second-round victory, with a numerical advantage to the Toucan candidate, the Sensus market research firm, published in Isto É on October 11, reported an advantage of 17 points over the PT candidate, gladdening the hearts of Toucan supporters  and shocking Rousseff supporters.

The Ibope and Datafolha currently predict a fairly tight victory for the incumbent candidate.  «Technical tie» is a catchphrase of the day.

In an attempt to explain the discrepancy, watchdog site Olho Neles took the trouble to analyze the basis of the poll, based on the same data on file with the Superior Elections Tribunal (TSE): the municipalities selected for the Sensus poll and how they were collected.

According to information supplied by Sensus to the TSE, the interviews were realized in 136 townships, and the choice of cities to be polled and analyzed could be used to explain the divergence with Ibope and Datafolha.

Olho Neles compared the vote by municipality as  selected by Sensus for its second turn polling and came to the conclusion that these were proportionately supporters of Aécio [by a 2-3 margin].

In the first turn, Dilma won in 3.648 townships, while Aécio finished in first place in 1.821 cidades, while Marina prevailed in 99 townships. That is to say: Dilma won in 65.51% of the townships, while Aécio won in 32.70% and Marina, in 1.77%.

In the Sensus poll, however, of the 136 cities selected, 66 voted for the president in the first turn, while 61% supported the PSDB and 9% the Socialists.  As a result, in the Sensus study, 48.52% of the cities were “Dilmist” in the first round, 44.85% went to Aécio supporters and  6.61% wanted Marina.

The case was widely followed in Minas Gerais, where Aécio has served as governor and senator. In the Estado, Dilma won 640 cities in the first turn, and Aécio received an additional 213. For that reason, Dilma  won in 75.02% of the cities, while Aécio received 24.97%. It might seem natural that the sampling would produce such a scenario. The problem is that among the 15 cities selected by chance for the institute’s second-round polling, 9 were considered “Aécio” (60% of the total) in the first turn, and only six were Dilma supporters (40%).

Though Minas has 853 townshiops and São Paulo has 565, residents from 15 cities in Minas and 23 in São Paulo were heard from. From that perspective, the victory of Aécio appeared much greater.

The available data do not allow pollsters to know the distribution of interview subjects in each city. It is not possible to know how many persons were interviewed in São Paulo … and how many in Minas [first and second largest electoral colleges in Brazil.]

The other side

Questioned about how townships are selected for surveys, Sensus director Ricardo Guedes said the selection is calculated using the PPT — « probability proportional to size» — method. He said that Sensus data reflects a diverse distribution of prospective voters.

Ricardo Guedes also denied that the poll had been sold to  “Isto É” upon its completion. TSE records reflect that Sensus paid for the poll itself, which was later published by the newsweekly.  According to Guedes, the article was a “partnership.”