Accuracy of Pre-Election Polls for the 2006 Italian Parliamentary Election: Too Close to Call

Abstract

In this article, we analyze the accuracy of over 70 published and 19 unpublished pre-election polls reporting results pertaining to the April 2006 election for the Chamber of Deputies in Italy and apply a new measure of predictive accuracy, A, recently introduced in the U.S. context. The center-left coalition won the election with a narrow margin of 0.1 percent of the votes, which came as a surprise: all published pre-election polls had attributed a 3 to 4 percentage point lead for the center-left. The failure of virtually all pollsters to predict the outcome is due to various factors: sampling error issues; a 15-day pre-election embargo on publishing results; high coverage error due to households having only mobile phones or no phone at all; a possible reluctance of Italian voters to declare their intention to vote for the center-right. The new measure of poll accuracy was easy to adapt and apply to the Italian context, due to a recent election reform. Information given by pollsters remains, however, too inadequate to engage in comprehensive evaluations of their polls.