Voting Intention Polls in the 2018 Italian Parliamentary Election
Giancarlo Gasperoni (University of Bologna)
Keywords: Electoral polls in established demcoracies
AbstractFive years have passed since the last Italian parliamentary election in 2013, in which the most important novelty was the presence of the populist Five-Star Movement (5SM), which garnered approximately one-fourth of all votes and led to the 2013 pre-election polls’ displaying the lowest degree of predictive ability among the four national elections held since the beginning of the century. The 2013 election signaled the end of long-standing bi-polar competition in the Italian political system; in fact, the 2018 election (which will be held on March 4) is expected to be strongly tri-polar in nature, with competition focusing on a centre-left coalition (incumbent), a centre-right one, and the 5SM.
Italian pollsters’ are facing the same issues that plague other established democracies. Moreover, the election will be held under new rules (enacted last November, after two preceding voting regimes were deemed unconstitutional), which further challenge polling institutes: 37% of seats are assigned in single-candidate districts on a first-past-the post basis; 61% are assigned on the basis of proportional representation criteria in multiple-candidate districts; 2% are allotted to voters residing abroad. The new rules comprise the *fifth* voting system since the beginning of the century (two of which were never actually used).
The authors intends, firstly, to monitor the characteristics of published polls carried out during the 2018 election campaign (the number of polls should be between 50 and 100), in terms of administration mode, sample size, focus (single- versus multiple-candidate districts), etc.; this first goal is also geared to evaluating polls’ methodological quality. Secondly, the performance of polls with regard to predictive accuracy will be assessed (via Traugott et al.’s “A” and/or Arzheimer and Evans’s “B” measures); characteristics of better-performing polls will be identified; if possible, polls’ performance will also be gauged in comparison to other forecasting modes (analyses of social media, prediction markets, etc.).