Two weeks before the general election, the most likely result estimated by Ipsos gives an absolute majority to the sum of the two main right-wing parties, PP and Vox. With a turnout of 70%, lower than the average of all general elections held so far, the people will win 145 seats with 35.3% of the vote. This means 14.5 points and 54 more deputies than those achieved in the 2019 elections. For his part, Fox will get 36 deputies with 12.6% of the vote: 16 seats and 2.5 points less than four years ago. Between the two they would add 181 seats in the House of Representatives, five above an absolute majority. This is the image revealing of the pre-election Ipsos poll data for vanguard set before the start of the election campaign. One certainty: the People’s Party will definitely be the party that gets the most votes and seats in 23J. From there, we enter a zone of uncertainty. Some are sharper than others. Today, they would do it with some ease. But we must not forget that some of the disputed seats are a zero sum: what is included in one block must be subtracted from the other, making the difference estimated now less than it might seem at first. And if we take in the worst ends of the two parties on the right, that adds up to 170 MPs, six short of an absolute majority. Adding the two maximums would add up to 186 seats. But it must be taken into account that there are disputed seats between them, so the two parties cannot happen simultaneously. In any event, the possibility of PSOE and Sumar alone reaching 176 seats was not considered. The best data for the Socialists with the best data for that of Yolanda Díaz would result in 152 seats, 24 short of that absolute majority.
This is the beginning of Ipsos pre-election polling report for vanguard which was published on Sunday 9 July with data collected before the start of the election campaign. After reading it, can we confirm that the polls were wrong?
This scenario provided for an unequal 6-point share in favor of the right versus the left. The hypothesis we used, based on these data, is that the right will vote as in the elections of April 28, 2019 (when it received the support of 11.5 million voters) while the left will do so as in the re-election on November 10 of the same year (when it received the support of 10.5 million).
How does he explain that participation was not ultimately unequal between right and left?
However, the 23-J results confirmed only part of the hypothesis. In fact, last Sunday the Left (PSOE + Sumar) voted almost as much as in 10N (PSOE + UP + Compromís + MP): 10.7 million votes. But the right (PP + Vox + UPN) stayed at an intermediate point: 11.1 million ballots. How does he explain that participation was not ultimately unequal between right and left? In the absence of the first post-election studies, we have a hunch: 400,000 center-right voters were laid off last week. Perhaps because of the expectation of an expected victory of the right, perhaps because of the post-election PP pacts with Vox in city councils and autonomous communities, perhaps because of the expectation of Vox entering the government … Anyway, in April four years ago, the People’s Party with Ciudadanos achieved 8.5 million votes. Last Sunday, the popular figures remained at 8.1 million.
What happened to bring about these changes? Well, neither more nor less than an election campaign that has caused an electoral realignment, especially over the past week. Could we have appreciated the 23J vote better? Yes, but only in the case of PSOE. And we had the data to do that. Polls have failed, in no way (and only in part), the electoral kitchen, that is, the decision on the correct model of vote estimation.
Before this election, at least three explanations can be considered based on the same raw data from the survey. Three models: (1) An estimate using voting memory from the last municipal elections on May 28, 2023 as the main variable. (ii) an estimate using estimates from the last general election on November 10, 2019; (3) Disproportionate layoff adjustment estimate for estimated participation (71%), with more weight given to the last election (28m). This last option was the one we used to make the estimate published in this paper. The average total error for all games was less than 1 point. Its weakness: underestimation of the Socialist Workers’ Party (28.2% estimated vs. 31.7% received, i.e. 3.5 points lower). This was the only difference that kept alive the possibility of a Whig majority. Today we know that a more faithful model of (1) would have been closer to the final result because we could have estimated the PSOE to be between 31% and 32% (victory, using the same raw data from the survey), which would have ruled out the possibility of a PP and Vox majority. This was an analytical decision attributed only to the pollsters, not to the pollster. In short, another lesson on the importance of electoral cuisine.