In the recent past, especially British election research has increasingly utilised the possibilities of new media. Pre-election surveys have always played a major role in Britain and, particularly, in the USA, but in the recent past, elections and election campaigns have also routinely been observed in surveys based on online panels – often with great success, regarding the interest in these surveys as well as the quality of predictions that can be made on their basis. In the UK, for instance, several better election forecasts could be made with online samples than with traditional survey methods.

In Germany, however, only few attempts have so far been made to describe and predict election behaviour and results by means of data collected in online surveys.

On the occasion of the Bavarian state parliament election in September 2008 we intended to explain the election behaviour in Germany based on an online panel and, thus, to predict the results of the election.

Approximately 1.500 Bavarian members of the YouGov-psychonomics panel were interviewed with respect to their attitudes as well as their intended and actual election behaviour immediately before and after the election. The turnout of voters, which is of central relevance for the 0utcome of state parliament elections, was subject to special consideration: this aspect was analysed with an experimental design (including the random variation of questions phrased and preset scales). In the scope of our contribution we would like to present the experiences we made with this pilot survey in Bavaria as well as the possibilities such online panels (e.g. the integration of experimental designs) have to offer.