Article ID: | iaor200969442 |
Country: | United Kingdom |
Volume: | 28 |
Issue: | 1 |
Start Page Number: | 1 |
End Page Number: | 18 |
Publication Date: | Jan 2009 |
Journal: | Journal of Forecasting |
Authors: | Sjberg Lennart |
Keywords: | forecasting: applications |
Four groups made forecasts of the outcome of the Swedish Parliamentary election in the fall of 2006. They consisted of members of the public, political scientists, journalists writing about domestic politics in Swedish daily newspapers, and journalists who were editing sections of readers' letters in daily newspapers. They estimated, using a 12-step category scale, which percentage of the votes that they believed seven parties would get in the election. Data were then obtained on the outcome of the election, and on the two opinions polls closest in time to it. When median forecasts were compared across groups, it was found that the group from the public was most successful in forecasting the outcome of the election. This was in spite of the fact that the median error made by individual members of that group was about 50% larger than the median error made by members of other groups. The two polls were less efficient than the group from the public and overestimated the span between the incumbent government and the opposition by a factor of 2. The members of the public and journalists showed some wishful thinking in their forecasts. There were large and consistent individual differences in forecasting ability. Men performed better than women, as did those who expressed more interest and knowledge in politics.