I am looking for studies that compare the accuracy of prediction markets for forecasting elections to forecasts from alternative forecasting methods (e.g., polls, expert judgment, quantitative models).
I am interesting in studies from all countries and all types of elections that analyze forecasts of numerical election outcomes (e.g., vote-shares or seats gained in an election). I am not interested in probability forecasts such as a candidate’s chance to win an election.
The attached document shows the countries/elections for which I have already collected data. I will post the final document here once available.
Thank you very much in advance,
Andreas
Relative accuracy is measured by the limits on the statistical error given by the surveyor. The question is reliability, which depends on how closely the regularity conditions, specified by basic theory, are fulfilled in the survey.
You may have already seen the June issue of Significance, the magazine published by the Royal Statistical Society, authored by Timothy Martyn Hill, which does a fair job of reviewing the forecast errors of betting and forecasters., but his questions about the pre-election day polls doesn't distinguish satisfactorily between polls of themselves never forecast, but only measure with definitive reliability the views of the sample on the day they were taken which is a principal reason that the exit in Britain have an excellent 'forecast'' and why pre-election polls do not if public opinion has changed, even on the margin, from the findings of surveys taken before the day of the election itself.
- Badges
- Science topic
- Similar topics
- Psychology