Friday 29th November 2024, 09:30–09:45 (Australia/Melbourne), Sutherland Room
In economics and finance, many decisions are based on uncertain outcomes. Published economic data, such as the unemployment rate or GDP growth, are based on estimates that have confidence bands. The uncertainty around the estimates is even larger for forecasts or projections. Most economic data, actual data, and forecasts are published as point estimates that lead people to ignore the uncertainty around the estimates. Ignoring these uncertainties leads to worse decisions.
Our paper explores how to convey probabilistic information in economics and finance to laypersons. We conducted a survey experiment within ANUpoll, a quarterly survey representative of the Australian population. ANUpoll is based on Life in Australia, a probability-based panel run by the Social Research Centre.
We randomly showed respondents one of four plots: a kernel density plot, a histogram, a quantile dot plot with 20 dots, and a quantile dot plot with 50 dots. We then asked each of the treatment groups the same four questions to gauge their understanding of the information conveyed by the plots. Respondents were told what was shown but received no further instruction regarding the interpretation. Each plot summarised the same probability distribution of forecasts of the Australian unemployment rate. We took these forecasts from actual figures published by the Reserve Bank of Australia. When shown the randomly assigned plot for the first time, respondents were asked to provide their best forecast of the unemployment rate one year out. We repeatedly reminded respondents to base their estimates on the shown plot.
Our experiment was included in the online questionnaire for January 2024. A good-sized sample of about 4,000 respondents completed that questionnaire. In addition to the data from the experiment, we were able to access rich data from the current and earlier waves of ANUpoll. This enabled us to examine how well the four plot types performed and whether this varied by sociodemographic and economic variables. We also found that incorporating figures into the online questionnaire was straightforward and worked quite well. Online surveys should consider doing this more often.
Markus Hahn is a research fellow and lecturer at POLIS: The Centre for Social Policy Research, Australian National University. His research interests include survey methodology, the analysis of economic inequality and various aspects of labour economics.