Evaluation of model 2024

With more than 99 % of the vote counted (CERA excluded), results are PP:40, BNG:25; PSDG 9; DOu 1. What we have learnt from this exercise:

Surveys have proved to be the most valuable source of information. Reliance of our model in prior vote biases the model towards prior results. Without additional information, the model is uncapable to produce sufficiently accurate estimations. Reliance on reliable sources only (those providing micro-data) will prove to be an useful decission to enable forensic analysis of our models.
Input from experts has been very valuable to remove computational bias. Contrary to results, our model was consistently predicting representation of Vox in the most populated circumscriptions. We only managed to remove this bias after changing parameters on the basis of expert input. The more experts were in contact with the model, the worst their input helped to correct this bias. Analysis of the case of Vox leads to confirm the hypothesis of dual vote in Galicia and more specifically, a “national” will of the Galician electorate, as they are more capable to isolate the political agenda at the State level. Regional interests have prevailed even when the majority party made an active effort to move the agenda to the Spanish agenda, and even with the most active participation from a Spanish President.
The logic of blocks prevails. We should severely limit voting transfer so the algorithm penalizes transfer between conservative and progressive vote. The only noticeable change has been 1 seat transferred from PP to BNG. Other than that, transfer happened inside blocks: 5 from PSdG to BNG, and 1 from PP to DOu. Unless any major event disrupts the moral valuation of the Galician population about any element of the blocks, inter-block transfer should be penalized, or at least carefully managed.
Towards next Galician elections, three facts should be taken into account. First, the PP has lost the monopoly of populist governmentality. Unless entrance of DOu is channelled, this success paves the way both for the emergence of similar initiatives (maybe Lugo), and for the consolidation of this party in a circumscription which is critical for the PP due to the overrepresentation effect. Second, the sustained growing trend of BNG now adds transfer of 1 seat from PP in addition to the 4 seats transferred from PSdG, representing the first time nationalists manage to enter the antagonic block. Third, without a strong PSdG, no change in government is likely. Unles it gets replaced by some new, disruptive, force in the left, the more likely option would be for the PSdG to retrieve votes from the PP. Anaysis on the capacity of PP to channel alternative populism, the capacity of BNG to continue sustaining this growing trend, and the capacity of PSdG to recover their voters and gain some from PP is likely to help discerning results in next elections.
We need to improve the infrastructure to more quickly incorporate data from reliable surveys, with more ready to run plotting, analysis beyond seats at provincial level (number and percentage of votes, plus trends and mapping are examples). We need to have a survey system ready to systematically and with automatic expert retrieval platform. We did not have the web platform ready, needing to deal with technical issues and glitches that have unnecessarily delayed our work.