Forecasting national team medal totals at the Summer Olympic Games

The paper reports the results of an exercise to forecast national team medal totals at the Beijing Olympic Games, 2008. Forecasts were released to the media before the competitions commenced. The starting point was an established statistical model based on a regression analysis of medal totals in earlier Games, with past performance and GDP among the principal covariates. However, we based our own forecasts on a model with additional regressors, including a measure of public spending on recreation. This adaptation is shown to have improved the forecasting performance. We also made subjective, judgemental adjustments before releasing our final public forecasts, and we demonstrate that this led to a further increase in accuracy. These final forecasts were successful in predicting the principal changes in medal shares relative to the 2004 Games, namely the surge in medals for China and Great Britain and the substantial fall in medals for Russia.