The study conducted by the research group in Elections and Public Opinion of the Universitat de València (GIPEyOP) affirms that the deviation from the polls is caused by the amount of voters with the same political profile who decline to participate. This concept is defined by the research group as differential participation.  Due to this fact, the pollster has to find new people who are willing to take part in the poll in order to get a sample important enough.

‘Knowing the characteristics of the people who decline to participate in the exit polls is essential when it comes to implement the strategies that reduce the margin of error,’ underlines José Manuel Pavía, full university professor of Quantitative Methods in the Faculty of Economics of the UV.

In order to correct this deviation, GIPEyOP suggests improving the implementation of the multipliers known as correction factors. These numbers have the purpose of applying the necessary sociological correctives to the gross data in order to make a more accurate estimation. These conclusions are included in the article ‘Sampling error or nonresponse bias?’ published in Revista internacional de Sociología and written by Jose M. Pavía, Elena Badal and Belén García-Cárceles.

Several important errors such as the ones that have taken place in other exit polls come from the mistake related to the correction factors. Pavía explains that the most significant case was last general elections in 2016. ‘It was discontent among the sector and the image shown to society threatened to affect the previous success.’

‘The final deviation from the polls prediction is not significantly affected by the people who participate and decide to hide their vote by a lie but by those who do not participate’ explains José Manuel Pavía. Furthermore, in polarised political climates, that is to say, when expressing certain opinion in public can be dangerous, it is not considered acceptable or it can go against the majority public opinion, the differential participation increases and, in turn, the uncertainty in the exit polls.

The study has been developed within the Spanish context and with some concern due to the mistake made in the electoral polls. The research is based on seven exit polls in which 165,655 voters were surveyed. These polls correspond to the Cortes Valencianes elections in 2003, 2007 and 2011 and in the 2012 regional elections in Andalusia, Catalonia, Basque country and Galicia.

In order to carry out the study, the research group in Elections and Public Opinion has gained access to poll’s microdata. However, the research has not determined the reasons why a fraction of the voters do not take part in the exit polls.

GIPEyOP

The research group in Elections and Public Opinion (GIPEyOP) is formed by more than ten researchers and aims at analysing, studying and finding solutions to all issues and questions related to electoral processes and the measurement and monitoring of public opinion. The most relevant research fields of the group include: the generation of electoral predictions, the inference of individual voting behaviour, the analysis of polls and surveys, or the semantic analysis of opinions and the monitoring of the internet sentiment.

Article:

Pavía, J. M., E. Badal and B. García-Cárceles. 2016. “Spanish exit polls. Sampling error or nonresponse bias?”. Revista Internacional de Sociología 74(3):e043.

doi: http://dx.doi.org/10.3989/ris.2016.74.3.043