Last Updated: 03 April 2019

About this site

Welcome to the Dectech Elections Website.

Currently this website displays the actual results from the 2010 general election. Previously we were making predictions on the outcome of each seat. We used Populus polling data to devise a different way of predicting the election.

Most election predictions use a "uniform swing" model, where the change in national vote share is assumed to be mirrored in every constituency. This model has various problems. For example if there is a national 2% swing away from Labour, a constituency recording a 1% result for Labour in 2005 would in 2010 predict a negative number of votes!

Instead, we attempt to model the flow of voters between the parties (and non-voting) and also include relevant local factors for each constituency.

Populus data allows us to investigate the current voting intentions of past Conservative, Labour, etc. voters. This allows us to estimate how well each party is holding their vote, which parties they are losing voters to, and where they are gaining votes from. In addition, we allow for an incumbency effect, and also take the age and class structure of each constituency into account.

The resulting constituency-tailored transition matrix gives an election prediction for each constituency. Because we investigate past polls, we are able to give each party's chance of winning in each constituency, in addition to predicting which party will win.

*A transition matrix approach was applied by Mebane and Wand to American Elections:

Mebane, Walter R. Jr., and Jonathan Wand. 1997. \Markov Chain Models for Rolling Cross- section Data: How Campaign Events and Political Awareness Affect Vote Intentions and Partisanship in the United States and Canada." Paper presented at the 1997 Annual Meeting of the Midwest Political Science Association, April 10-12, Palmer House Hilton, Chicago, IL.

back