Podemos, Ciudadanos or the nightmare of polls

​Now that new elections in Spain seem unavoidable fresh polls will start to be released soon. They will be closely followed not only by political parties but also by voters who tend to vote strategically. Before last December 2015 general elections, opinion polls offered substantially discrepant forecasts. And, may be understandably, they were especially wrong about the results of new parties (Ciudadanos and Podemos). So now some voters could ask themselves, among all the options available, which is the opinion poll that does best in tracking voters’ intention? This is virtually impossible to find out as voters may decide their vote depending on the surveys themselves. However, as a rough measure of accuracy, it is possible to evaluate polls by contrasting their forecasts with actual results obtained by parties in last December elections. To do so, the table shows the five most voted parties together with the last forecast made by each poll before the elections. The root mean square error (RMSE), a common measure of forecast accuracy, is also shown. The RMSE penalizes large deviations, so having one big mistake is worse than incurring in a series of small errors. According to this measure, polls were on average 3.2 points above or below actual voting results. El País did particularly well in quantifying the support that new parties would receive and this allowed its poll to have the lowest RMSE. La Razón did better in forecasting the relative strength of traditional parties (PP and PSOE). The large miscalculation of El Periodico de Cataluyna regarding Ciudadanos places the newspaper as the worst forecaster in this occasion. The coming repeat elections are a great opportunity to see how the various poll organizations have learnt from their mistakes.

Opinion polls' forecasts and actual results. Spanish general elections, December 2015 (%)