Click anywhere on the chart, drag the slider, use the navigation buttons, or put a date in the url to view
Intrade maps for the complete 2012 election season. States were colored gray until they reached a volume threshold.
*Note that Intrade had a 100% correct (51/51) prediction for 24 consecutive days before the fallout from Benghazi and the Denver debate. The prediction market only missed the final forecast by one state, Florida. Read analysis.
Most 2012 election forecasts were very accurate, here's why
Starting on September 13th, 2012 the Intrade prediction maintained a 51/51, 100% correct prediction for 24 consecutive days. Fallout from the Benghazi controversy and the first debate in Denver started to show up in polling data in early October and shook up the Intrade prediction market for a few weeks. Then things trended back to where they had been before. The final forecast before the election was 50/51 - missing only the state of Florida.
Although Intrade performed exceptionally well this season, so did a lot of other forecasters. The Real Clear Politics (RCP) poll average virtually mirrored Intrade's performance and statistics whiz Nate Silver (FiveThirtyEight) scored a perfect 51/51, calling Florida for President Obama just before the election. Most forecasting losers this season were partisan pundits predicting massive landslides for their candidate of favor, although there were some credible models based on economics and unemployment that had predicted a Romney victory.
So why were the forecasts so accurate this year? The answer is simple: the polls were accurate and they aggregated well. The pollsters came up with good turnout models and the electorate showed fierce party loyalty. This combination yielded poll results very well suited to aggregation. Averaging a bunch of polls together that are all fundamentally sound will usually produce a result that is of better quality than any one of the polls individually. The idea being that the various biases and errors tend to cancel each other out. They use different methods, but RCP, Intrade, and FiveThirtyEight are all essentially aggregation systems - and they all performed exceptionally well.
And now for a word of caution. There is an understandable level of exuberance over the accuracy of forecasts this year, especially on the heels of another very successful forecast in 2008. But we should be very careful about declaring a Golden Age of Election Forecasting. It's seductive to believe that, with proper analysis, an accurate forecast can be coaxed out of any set of polling data. But that's simply not the case. We don't need to look any further than the state of Indiana in 2008 to see election results that defied nearly every major forecast.
The problem is similar to being handed a bag of diamonds that have been graded by a jeweler. You can look at his equipment, his credentials, and you can look at other bags from the same diamond mine that have been graded by other gemologists. But you can never open the bag and actually look at the stones until the end. There's always a chance that you got a lump of coal. Now realize that we're not talking about diamonds, we're talking about people and political opinions, and the problem becomes even more apparent. An accurate Intrade forecast relies on people making informed bets, and those people are relying on quality data from pollsters. Crunching good numbers yields illumination. Crunching bad numbers give you crunchy bad numbers.
Thanks to all the traders who put their money on the line to bring us a great forecast in 2012! Hopefully the next election will have even higher volume and market efficiency on Intrade.
See you again in 2014 for the midterms
©2011 ElectoralMap.net, all rights reserved
Map Technology by DougX.net