In defense of the polls

Abstract

Public opinion polls contain very valuable features and show a strong track record, which implies that they are a valid and reliable research tool. Nevertheless, polls are always confronted with two accusations: their lack of predictive power and the influence they have on their very object of research once their results are published in the media. Furthermore, these attacks often seem to be aimed at the technique itself, leaving the other actors in this process —experts and mass media— untouched. Specialists, on the other hand, have barely shown any creativity in their defense of this technique, and they have done very little to establish whether there are any grounds for these criticisms. Furthermore, practically no studies have been carried out on the reasons why predictions (very seldom) fail. All these charges are discussed in this article.