Past research has revealed a trend effect when people are faced with a revised probabilistic forecast: A forecasted event that has become more (vs. less) certain is taken to signal a trend towards even stronger (weaker) certainty in future revisions of the forecast. The present paper expands this finding by exploring the boundary conditions of the trend effect and how it affects judgments of the forecaster. In Study 1, the trend effect was shown to persist when receivers process the forecast more deliberately, by considering reasons for the revision. In Study 2, trend continuation was predicted even when the two forecasts were made by different experts at different points in time. Study 3 demonstrated that the effect disappears when receivers are given an earlier forecast disrupting the linearity of the trend (e.g., a 60%-70% sequence preceded by a 70% forecast). In Study 4, two forecasters were perceived as more in agreement when revising divergent probabilities in the same rather than in opposite directions. If the event occurs, a forecast with downgraded probability (e.g. from 50% to 40%) was judged to be less accurate than an equally uncertain single forecast (40%). These results demonstrate the robustness of the trend effect based on two forecasts, affecting not only receivers' expectations of what comes next, but also their perceptions of the forecaster and of forecast accuracy. The findings have implications for how people communicate and understand risks and other uncertain events in areas such as climate science, weather prediction, political science and medicine.
When probabilities change: Perceptions and implications of trends in uncertain climate forecasts