How did the European polls do?

The polls were once again very useful in the European elections last Sunday. They were exceptionally accurate. The estimate of 40dB., published by Morning Express and Cadena SER, deviated by less than one point per game, which is a remarkable precision. Our polling average reduced that error to 0.69 points.

The average of polls almost nailed the results of PP and PSOE. They also greatly adjusted the results of Vox, Ahora Repúblicas, Podemos, Junts, and even those of Sumar, whom they overestimated by 1.3 points.

Furthermore, the pollsters once again achieved something complicated: they anticipated the emergence of a new formation, Se Acabó la Fiesta (SALF), led by Alvise Pérez. In the polls a week ago they were close to 3% of votes and rising. Our prediction said that they would have representation with 85% probability, and that they could reach three seats, which are the ones they obtained. The polls underestimated his vote by 1.9 points, but anticipating his entry is commendable.

In 2024, demoscopy chains four elections with good results (European, Catalonia, the Basque Country and Galicia), after having done worse last year in the general elections and, above all, in the Valencian elections.

But which big pollsters are being more accurate? The following table shows the success of eight organizations since 2015, taking the main quotes of each year. There you can check the good performance of 40dB. and the average of Morning Express.

The 40dB surveys. They have done better than the average (green color) in all the events this year. And our average, which aggregates the surveys under a certain methodology, has achieved the best overall accuracy. He did better than the average in 14 of 15 elections since 2015. Of course, it is not so much our merit as it is the pollsters’: the average only works thanks to the polls from which it is fed.

The difference between PP and the PSOE

On Sunday, the Popular Party won by four points over the Socialists, which is what the average had anticipated. Last week, the pollsters largely agreed: Sigma Dos gave the PP a two-point lead; which were three for 40dB., five for GAD3 and six for Sociometrica.

However, a few weeks earlier the estimates were different. In March, GAD3 gave the PP a 10-point advantage, and Sociometrica gave them 12 in April, while 40dB. He spoke only of 5.5 points in his general election polls. The difference is striking due to its magnitude. Although, however, later the pollsters saw the same trend: the PP fell and the PSOE rose.

All the estimates agreed in placing the PP in first position, with one exception: the CIS said that the PSOE would prevail by between 1 and 5 points, although it later lost by four.

The CIS was again imprecise

The public body’s latest estimate for the European elections made an error of 1.6 points on average, more than double our average. It deviated by almost five points with the PP and by 2.6 with the PSOE. His estimate was the worst of the 10 analyzed.

The poor performance of the CIS is not new. As the previous graph showed, when we compare the main pollsters, those of the CIS appear as the worst. This has been the case since the arrival of José Félix Tezanos to the direction of the center. Yes 40dB. deviates 0.14 points less per party than the typical survey, the CIS deviates 0.56 points more.

Furthermore, this Sunday the CIS once again exhibited its worst problem: the bias to the left of its estimates. It gave around 42.4% of votes to the sum of PSOE, Sumar and Podemos, which remained at 38.1%, four points below. Nobody can be surprised. As we have been counting for years, this pattern is repeated over and over again: with Tezanos, the CIS has overestimated the left in 41 of the 42 elections that we have analyzed since 2018.

As we have also repeated many times, the worst thing is that years go by without anyone bothering to correct this obvious bias.

_