Forecasting elections by polling the polls

Democratic institutions are filled with individuals choosen on free elections. Although forecasting behaviour of individuals may not be the first and single purpose of social science, this blog will start with an entry about the uses of Bayesian tools to pool knowledge from different sources and provide a forecast for the Spanish legislative elections that will be held next november 20th, 2011.

I have adapted Simon’s Jackman model on “polling the polls” in order to account for a multiple party system, with many different poll houses which publish relatively few polls. This contrasts with the usual environments of two-party systems, where there are less poll houses but they take lots of snapshots at different points in time.

Less than three weeks before the elections the forecast still predicts a bipartisan system. Conservative (PP) and Socialists (PSOE, in power) account for 78 of the support, with the former being ahead by 14 percentage points (46% vs 32%). IU (Leftists) and UPyD with 5.9% and 3.2%. The rest of the parties are territorially centered (CiU with 3.3% and ERC with 0.9% in Catalonia; PNV with 1.2% in the Basque Country; BNG with 0.9% in Galicia and CC in the Canary Islands with 0.7%).

Prediction for Spanish elections at 111101

Consult all the results at Forecasting Spanish elections 2011 and subscribe to the blog via RSS in order to get the latest updates.

This entry was posted in Applied models, Bayes, Elections and tagged , , . Bookmark the permalink.

Leave a Reply

Your email address will not be published. Required fields are marked *