Who Got It Right on Election Day 2007?
by Jonathan Singer, Wed Nov 07, 2007 at 01:33:18 PM EST
Yesterday I predicted that Steve Beshear would beat Ernie Fletcher by a 59 percent to 41 percent margin, which as it turned out was a net six-tenths of a percent off the mark from the final tally. How did I come up with that guestimation? By examining the long-term and short-term averages of all of the public polling on the race.
Looking in-depth into all of those polls, which individual pollster performed the best? Who nailed the race? Though you might not have guessed it if you listened to the Beltway pundits and establishment types who have showered robo-pollsters with little other than derision over the years, the outfit that outshone its competitors in the Kentucky gubernatorial race was SurveyUSA. Take a look at this chart:
So which pollster came closes to making the right prediction in this race, both in terms of margin and final spread? SurveyUSA, Rasmussen Reports and Research 2000 were all about the same amount off in the final spread (with the latter two coming 2.4 points under and the former coming 2.6 points over). But in terms of getting the exact performance of the individual candidates, SurveyUSA blew its competition out of the water, missing by a combined 2.6 points. By comparison, the next closes pollster in this regard, Research 2000, was off by a combined 5.0 points -- not by any means bad (in fact very good as these things come), but not nearly as close as SurveyUSA.
The purpose of this comparison is less about stoking the egos of the folks at SurveyUSA as it is about sending a message to some of the folks in the Beltway who disregard SurveyUSA and other pollsters who use automated rather than live interviewers that this relatively new method does work -- and at a significantly lower cost than the more traditional methods of polling. Roll Call seems to get it, contracting with SurveyUSA to poll the most competitive Senate races around the country this cycle. But to those who are still kneejerkedly skeptical about SurveyUSA and other automated pollsters, it seems that the time is now to start rethinking those sentiments.