|Authors||S. M. Hohle and K. Teigen|
|Title||Forecasting forecasts: The trend effect|
|Afilliation||Software Engineering, Software Engineering|
|Publication Type||Journal Article|
|Year of Publication||2015|
|Journal||Judgment and Decision Making|
|Publisher||Society for Judgment and Decision Making|
People often make predictions about the future based on trends they have observed in the past. Revised probabilistic forecasts can be perceived by the public as indicative of such a trend. In five studies, we describe experts who make probabilistic forecasts of various natural events (effects of climate changes, landslide and earthquake risks) at two points in time. Prognoses that have been upgraded or downgraded from T1 to T2 were in all studies expected to be updated further, in the same direction, later on (at T3). Thus, two prognoses were in these studies enough to define a trend, forming the basis for future projections. This "trend effect" implies that non-experts interpret recent forecast in light of what the expert said in the past, and think, for instance, that a "moderate" landslide risk will cause more worry if it has previously been low than if it has been high. By transcending the experts' most recent forecasts the receivers are far from conservative, and appear to know more about the experts' next prognoses than the experts themselves.