Share this post
FaceBook  Twitter  

What's New on this Site

This page is currently inactive. For older news, see below.


  • September 9, 2010 The race to forecast the U.S. midterm elections is on! Click here for Polly's latest prediction.
  • September 8, 2010 APSA panel on the U.S. midterm election.The Monkey Cage blog has a summary of the forecasts for the U.S midterm elections presented at a panel of the American Political Science Association held in Washington, D.C., on September 4th. Also included are several of the power point slide shows or papers presented at the conference.
  • February 11, 2008 Polly's new page for the 2008 election is online and will be cared for by Andreas Graefe. He is researcher at the Institute for Technology Assessment and Systems Analysis in Karlsruhe, Germany, currently on leave at the Wharton School. Visit:
  • January 19, 2008.Erikson and Wlezien apply damping to achieve substantial improvements in the accuracy of pre-election polls. “Damped” polls provide more accurate forecasts than the Iowa Electronic Markets for the five U.S. presidential elections from 1988 through 2004. This "E&W poll damping" will be used in the construction of the Pollyvote for the 2008 election.

    Traditionally, raw pre-election polls are interpreted as forecasts of election outcomes. However, these polls register preferences on the day of the poll but do not provide predictions of the election outcome because many things might change prior to the election. In contrast, prediction markets try to account for the changes that might occur. By comparing raw polls with prices of the Iowa Electronic Markets (IEM) for the five US presidential elections from 1988 through 2004, Berg et al. (2007) found that the IEM outperformed polls 74% of the time.

    Erikson and Wlezien provide new evidence on using opinion polls to predict the outcome of elections. Longer-term forecasts involved more uncertainty, and uncertainty calls for more conservative forecasts. Thus, for a two-party vote, forecasts should be damped towards 50-50 -- and the longer the forecast horizon, the more damping needed. Erikson and Wlezien showed that damped polls were much more accurate than the original polls in forecasting the five US presidential elections from 1988 through 2004. Moreover, the damped polls were shown to be more accurate than IEM vote-share market prices and especially those from winner-take-all markets.

    Reviewed by Andreas Graefe, January 19, 2008.

    Erikson R. S. and Wlezien C. (2007): “Are Political Markets Really Superior to Polls as Election Predictors?”, Public Opinion Quarterly, forthcoming.

    Further discussion of this paper and the question whether damped polls outperform prediction markets or vice versa can be found here.