Brexit and Trump

...

Although in several cases the assumption that polls were incorrect is attributed to the way in which those polls were reported in the news, and many media outlets frame their electoral coverage just as a horse race (Searles, Ginn, & Nickens, 2016), in the cases of Brexit, Trump, and Colombia, there has been a consensus between pollsters and analysts that the polls were truly wrong. A mix of factors contributed to it. In addition to problems of validity, there were problems related with sampling. As Fowler states, that means problems of generalization from the sample of respondents to the population. In the case of Trump, for example, there was misrepresentation of the population in the sample, including aspects like the difficulties that represent the rise of cell phones, the “likely-voter” problem in polls, and the context of confirmation bias among pollsters and journalists.

Callegaro and Gasperoni (2008) found similar problems when they analyze the accuracy of 89 published and unpublished pre-election polls in Italy in 2006. The failure of polls to predict an Italian election outcome was attributed to sampling error issues, households having only cell phones or no phones at all, and reluctance of voters to declare their intention to vote, among other aspects.

References

Callegaro, M., & Gasperoni, G. (2008). Accuracy of pre-plection polls for the 2006 Italian parliamentary election: Too close to call. International Journal of Public Opinion Research, 20(2), 148–170. https://doi.org/10.1093/ijpor/edn015

Moffett, M. (2016, October 6). The “spiral of silence”: How pollsters got the Colombia-FARC peace deal vote so wrong. Vox. Retrieved from http://www.vox.com/world/2016/10/6/13175608/polls-colombia-farc-peace-deal-vote-wrong

Searles, K., Ginn, M. H., & Nickens, J. (2016). For whom the poll airs comparing poll results to television poll coverage. Public Opinion Quarterly, 80(4), 943–963. https://doi.org/10.1093/poq/nfw031