A simple model of political popularity, as recorded by opinion polls of voting intentions, is proposed. We show that, as a consequence of aggregating heterogeneous poll responses under certain assumptions about the evolution of individual opinions, the time series of poll data should exhibit long memory characteristics. In an analysis of the monthly Gallup data on party support in the UK, we confirm that the series have long memory and further show them to be virtually pure ‘fractional noise’ processes. An explanation of the latter result is offered. We study the role of economic indicators in predicting swings in support, perform event analyses and use our estimates to generate post-sample forecasts to April 1997.