Partners and I recently started testing software that allows people to answer questions via a smartphone app. In conducting this trial we enlisted a small group of 125 to download the app and respond to a handful of queries. Since the timing of the test coincided with the first presidential debate, we took our group from those that were interested in politics and asked them political questions. The app users, while small, was geographically diverse and responses to most our questions were in line with results from large surveys-- sixty percent of our users expected Obama to win the debate, 65% said Romney actually won it, while 21% went with Obama--all in line with results from traditional polling sources. But when we asked our group if they would switch their vote in light of the debate results, the response diverged with mainstream polls. Even though our group judged Romney the resounding winner there was no net switching with 5% going to Romney and 7% to Obama. Meanwhile traditional polls showed Romney gaining anywhere from 4 to 12 points in national surveys.
Though our user group was far too small to be considered representative, it is worth considering if their response, that there was no real switching, was the more accurate reflection of voter's response to the debate.
Historically there is no statistical relationship between a debate win and an increase in subsequent polls. But the Romney win was unusually large so perhaps this time the debate did actually change the race.
But the election betting markets strongly believe that Romney hasn't drawn very close. The two best known, the Iowa Election Market and Intrade still have Obama comfortably ahead. A week after the debate you can buy a Romney victory at 37 cents (and receive a dollar if he wins). While this is up from the mid 20s before the debate, it's still a far cry from him actually leading. In fact it means bettors think he has a 37% chance of winning. Anyone who has bet on the NFL and marveled at the ability of betting markets to predict the final point differential in a football game will not need convincing of the wisdom of betting crowds. To paraphrase Samuel Johnson, nothing focuses the mind like a $100 bet. But for those that need more convincing, academic studies agree that the betting markets are more effective than polls at predicting election outcomes, besting traditional polls 74% of the time since their founding in 1988.
In the week that we have been asking our test questions we have learned some interesting if not necessarily actionable information. What can we imply from the fact that 64% believe Romney meant what he said about 47 percent of Americans considering themselves to be victims, while only 43% agree with Romney's assertion? Or how could it affect our foreign policy that 60% think military action should be off the table in dealing with Iran's nuclear program but only 8% think the threat is an exaggerated by the Military industrial complex? But the information on lack of voter switching, taken with an enormous grain of salt given the small sample size, but combined with the election market data, and I think I know what to do if someone offers me Obama at even money.