PREDICTING ELECTIONS

In most cases, scholars analyse elections after they have taken place. By doing so scholars know the final result and can argue what the reasons were for the given outcome. However, following the increasing use of pre-election polls, especially in the US, more and more people try to analyse the voters’ mood before an election and with that forecast the voters’ decision at election day.

Generally, one can group forecasting methods into structural models, which use economic and/or political indicators to predict an election outcome, and models using polls. The latter has grown in the latest years given the better data availability resulting from the higher number of pre-election polls. For the 2015 Swiss national elections, a colleague from Brazil, Neale Ahmed El-Dash, and I conducted a forecasting model using Swiss polls. In contrast to many other countries where hundreds of polls are conducted, in Switzerland very few polls are published and even forbidden in the last 10 days before the actual election. Hence, we used Switzerland as a typical “low-information” context to test in how far the statistical model works. In fact, our forecast was pretty close to the actual result.

For more background information see my blog entry on polithink.ch (in German only) and more information about the actual model and data used on www.pollingdata.com.br.

---

E-VOTING

Several countries work on procedures how to vote by internet. One of those countries is Switzerland which started its project “Vote électronique” in the year 2000. So far e-voting is available in several test communities in different Swiss cantons and also for parts of the Swiss living abroad. With a mandate from the canton of Geneva, we conducted a study in which we tested the effect of the introduction of e-voting in several communities. The idea was to see whether the overall participation can be raised by offering this new way of voting and in how far e-voting is used by a specific part of the publication. Often, the procedure of e-voting is linked to the hope of motivating more young people to participate in elections when they are offered with a more modern way of voting which fits their normal habits.

Our study results, though, do not confirm this boosting effect among young participants. Rather we find that people betweem 25-34 are the typical e-voters. Furthermore, men, married and more educated people have a (slightly) higher tendency of using the possibility to vote online. Regarding political ideologies we find that left voters use somewhat more the possibility of e-voting. In general, though, only around 15-20% used the possibility when offered, which did not increase the overall level of participation. A typical characteristic of e-voters is that they vote rather shortly before the actual election day (in Switzerland the huge majority normally votes by post, i.e. several days/weeks before the election). Even more information, also about confidence in the procedure can be found in our report (in French).

Sciarini, P., F. Cappelletti, A. Goldberg, A. Nai and A. Tawfik (2013). Etude du vote par internet dans le canton de Genève. Rapport final à l’intention de la Commission externe d’évaluation des politiques publiques.