Article written by José Alberto León Alonso in El Día on 03/29/15.
I wrote a few weeks ago about the usefulness of surveys. Polls and electoral barometers are a source of first-hand information about the political pulse of a nation. The decision of whether or not to bring forward the calling of elections sometimes depends on them, an always risky bet that requires good information, and there is no better one than a survey that faithfully portrays the electoral sentiment of the moment. Politicians, very aware of this, take or leave surveys depending on whether the results are favorable or not, resorting if not to the recurring phrase of “the only survey that is valid is the one at the polls.”
In the case of the 2015 Andalusian elections, the reality is that the polls achieve a reasonably good level of accuracy. Despite the current uncertainty, with a panorama of great volatility, the March polls have got the result right, both in terms of the winner and in the relative position of the different parties and their approximate seats. All parties have obtained results within the margin of error without major surprises. Thus, it seems that the electoral advance has paid off thanks to the great precision that, this time, the much-maligned polls have had. In the attached table I show the electoral predictions by seats published during the month of March. The actual seats won are shown shaded in the penultimate column, and the last column details our own estimates, made with data up to the first week of March. The TNS poll was conducted at the exit poll on election day itself.
As can be seen, the pollsters with the highest degree of accuracy (they got the number of seats for three of the five parties right) have been NC-Report and…ourselves (C5-SAO). In our case, we appropriately predict the number of seats for PP, Ciudadanos and IU, while we underestimate those of the PSOE (we predict three less than the real ones) and we overestimate those of Podemos (we predict three more). The reason may be in the concealment of the vote. In fact, the TNS exit polls awarded between 41-44 seats to the PSOE and 19-22 to Podemos, and since these are surveys carried out on voters on the same day of the elections, there are no biases in them regarding the decision to vote or abstain and the sample should be representative. Everything suggests that part of the PSOE voters decided to hide their vote from the pollsters, instead proclaiming their support for Podemos. It is likely that the EREs scandal has a lot to do with this behavior, but for next elections we will have to count, at least in Andalusia, on this concealment of the vote (traditional in the PP, but not in the PSOE) to refine our "kitchen" and optimize the level of success.
In any case, in most cases the differences fall within the inevitable margins of error, in addition to the influence of the distance between the time of the survey and the day of the elections, since the volatility of electoral sentiment is greater than ever in our democratic history, as demonstrated by the fact that a party like Ciudadanos, which did not enter any Andalusian survey just two months before the elections, obtained almost 10% of the votes. Voting intention changes over time, while electoral polls that are published no later than a week before the elections are based on surveys carried out several days, or even a week before their publication, and in that period events may occur that affect the decided vote or the final decision of those who were undecided.
On the other hand, the eight constituencies and the application of the D'Hondt formula when assigning seats introduce additional estimation problems. Thus, the PSOE won a seat in Cádiz to the detriment of Ciudadanos by just 720 votes (0.1% of the votes), and another in Huelva to the detriment of Podemos by only 1,217 votes (0.5%). That is, the allocation of seats is sometimes “played” for a tiny number of votes, impossible to estimate through surveys.
Concluding: The polls were right. At least this time the only valid poll has not been at the polls.
U!
