PollCheck – classification of pollsters

After the close of the 2024 electoral cycle (elections in Galicia, Euskadi, Catalonia and European elections), the time has come to update the classification of pollsters, our renowned PollCheck.

How it was made

Following the usual trend in the analysis of deviation of the pollsters, after each election we calculate the deviations of each of the last polls published by each pollster prior to the election date (surveys at the closing of the polls are not included, where they are usually done settings).

With each election, we have classified the pollsters into various blocks according to the degree of deviation. Those located in the upper part of the table (smaller deviation) have a plus. The best survey adds +0,75 to your grade, the next block of surveys adds 0.5 and the next block 0.25. The pollsters in the middle part of the classification do not see their grade altered, that is, they neither add nor subtract, so their plus is 0. In the same way, the pollsters in the lower part receive negative marks in blocks, according to its deviation, of -0.25, -0,5 or -0,75. If the deviation is very very large, a plus of -1 is applied.

Keep in mind that The poll deviation represents 35% of the PollCheck total, so the maximum score in this section is 3,5 and the minimum is 0. In cases like the CIS, this section was going to be negative, so it is logically left at 0.

As four electoral processes have been carried out for this classification, the final PollCheck score corresponds to the score from the previous PollCheck (2023) plus the sum of all the bonuses that the pollster has in the four processes.

Finally, we have included three penalties for three pollsters, in line with what was noted during these months, after detecting incoherent data, missing breakdowns for some party, post-hoc modification of the data published by the media/pollution company, etc. These penalties were published and anticipated on our Twitter account at the time of publication of the poll in question.

The ranking: The Panel is crowned, 40db and GAD3 improve, Sigma Two worsens. The CIS sinks

With all this, the result shows a classification in which it is appreciated an important improvement of pollsters like our ElectoPanel (produced by EM Analytics), which overtakes Sociometrica and is placed in first position. They also harvest good data 40db and GAD3.

On the other hand, pollsters like Sigma Dos see their classification worsen after this electoral cycle. And the CIS data is very notable, which sinks to position number 30, with the worst classification in historical deviation of all.