Abstract Empirical evaluation of econometric market share models has shown that they are useful as descriptive tools. This use of the models has led to a number of generalizations about the effectiveness and relative importance of advertising, price, and other elements of the marketing mix. However, little is known about whether these models are useful for prediction. This paper examines the published empirical evidence about the predictive performance of econometric market share models and uses data for fifteen brands from three markets to examine the predictive ability of the models in more detail. Three conclusions are reached: (1) Econometric market share models were not consistently more accurate than simple extrapolation (time series) methods for short-term forecasting. (2) Market share models did not usually capture enough of the important features of the market to be used by themselves as ‘stand alone’ forecasting instruments. (3) Apart from differences between markets there did not appear to be circumstances which indicated where econometric market share models are likely to be more accurate at short-term forecasting. Surprisingly the face validity of the estimated models did not appear to be a good indicator of forecasting accuracy. The final section of the paper poses a number of research questions which need to be resolved before any final judgment can be made about the usefulness of econometric models for prediction.