Differences between the single-event and frequency formats of seasonal-climate-forecast probability

Abstract

Since the late 1980s, Australian forecasters have used the seasonal climate forecast (SCF) statement 'In the next three months, the probability of getting above median rainfall is 30%.' Study one (n=63) established a baseline of whether laypersons interpreted this statement as forecasting wetter or drier conditions than normal. Although the statement is forecasting a greater likelihood of drier conditions than normal, 76% incorrectly interpreted the statement as forecasting a wetter season than normal. Using testing conditions identical to study one, in study two (n = 71), to improve accuracy, we inserted the word only in the statement (i.e. ' ...probability of getting above median rainfall is only 30%'). While the probability itself had not changed, a drop to just 24% errors suggested participants were ambiguous about how to interpret this probability. A body of research by Gigerenzer and others shows peoples' ability to reason with probabilities is enhanced by frequency probabilities rather than single-event probabilities. Study three (n =51) used the frequency (i.e. 3 in 10) rather than single-event (i.e. 30%) format tested in studies one and two. Only 22% made errors showing participants better understood how to interpret the frequency probability. We discuss formats for presenting the SCF and argue the frequency format more effectively conveys the chance nature of forecasts.