A New Approach to Rank Forecasters in an Unbalanced Panel

This study presents a new approach to ranking professional forecasters in an unbalanced panel. Ranking professional forecasters while not accounting for missing forecasts can lead to arbitrary results particularly depending on the forecasted variable and time period chosen. Here, our focus is on a third, very important but neglected, factor—the missing forecast. This paper identifies some serious issues related to the current methodology of some organizations, here we use the Bloomberg Survey as an example although Bloomberg is not alone with this problem. We re-rank top-10 forecasters of nonfarm payrolls using a new approach which accounts for missing forecasts. For many forecasters, the ranking based on our approach is significantly different than those of Bloomberg’s ranking. For instance, Bloomberg declared Credit Agricole as a winner (rank 1) but the new approach assigned 10th position (rank 10) to Credit Agricole. One major reason of different rankings for a firm is that Credit Agricole did not forecast for all 24 months and it was rewarded in the Bloomberg methodology for being absent in certain months. Our methodology does not reward a forecaster for being absent nor does it penalize a forecaster for submitting a forecast and thereby provides a fairer, more rigorous and accurate ranking. In addition, traditional forecast evaluation criteria, such as, MAE, MSE or RMSE are good for a balanced panel but not accurate for an unbalanced panel. Our approach provides a more rigorous and accurate forecasters ranking for any unbalanced panel.