This paper investigates strategic motives of macroeconomic forecasters and the effect of their professional affiliations. The ‘wishful expectations hypothesis’ suggests that a forecaster predicts what his employer wishes. The ‘publicity hypothesis’ argues that forecasters are evaluated by both accuracy and ability to generate publicity, and that forecasters in industries that emphasize publicity most will make most extreme and least accurate predictions. The ‘signaling hypothesis’ asserts that an extreme forecast signals confidence in own ability, because incompetent forecasters would mimic others to avoid public notice. Empirical evidence from a 26-year panel of annual GDP forecasts is con-sistent with the publicity hypothesis. This indicates that conventional tests of rationality are biased toward rejecting the rational expectations hypothesis. Copyright ? 2008 John Wiley & Sons, Ltd.