« Fun With Longevity-Adjusted Benefits | Main | O Unhappy Day »

Ending Exits As We Know Them

Unless you spent Election Night and The Day After ignoring the whole thing, you probably know there was, er, ah, a bit of a problem with the official exit polls conducted by Edison Media Research/Mitofsky International. Basically, the exits showed Kerry winning the popular vote by about the same margin that Bush actually won the popular vote; and also showed Kerry ahead and likely to win in Ohio and Florida, thus clinching the electoral college. This discrepency obviously created a lot of embarrassment (and in the case of Democrats, soon-to-be-dashed victory expectations) for the millions of people who got access, via personal contacts or the internet, to the exits during the afternoon and early evening of election day. And it also helped feed the conspiracy theorists who decided that the actual vote count, not the exits, must be screwed up thanks to the devilish ministrations of GOP election officials or the Diebold Corporation or whatever. To their credit, the people who conducted the exits didn't retreat into a technical haze, and intone, like a parochial school teacher asked by a student to explain the Holy Trinity: "It's a mystery; let's move on." The Edison/Mitofsky combine did an internal study, and have now released a 77-page analysis of What Went Wrong. Turns out, they say, it wasn't a poor selection of sample precincts, or erroneous weighting of results, but mainly a skewed response level to the requests for interviews that led to a slightly higher percentage of exit poll participants among Kerry voters than Bush voters. And that, in turn, was partly caused by access problems in some precincts, and also by the youthfulness of exit poll interviewers, who may have unconsciously over-selected their peers. The study suggests remedial steps to avoid bad exits in the future (mainly uniform access rules for exit poll interviewers and better training), but I have a different proposal: getting rid of the Election Day/Night predictive function of exits altogether. Lest we forget, exit polls are conducted for two very different reasons: (a) to make sure news organizations can "call" states and elections as quickly as possible after the polls close, and (b) to provide empirical data to support interpretations of why voters chose this candidate over that. The civic value of the first function is nil, and perhaps negative. The second does matter, and not only to political scientists or pundits, because without exits, how could we immediately and confidently tell the president he's full of crap when he says the electorate endorsed his Iraq policies? So the obvious thing to consider is to get rid of the first function of exit polls, and preserve the second (the latter goal being easy since they can be adjusted to reflect the actual weighting of actual votes, as they were a few days after this election). A flat ban on distribution of exit poll data until the day after the election should do the trick, since all the leaks invariably come from subscribing news organizations rather than the exit poll firm. Maybe a hefty fine for leakers would help as well. What, specifically, would we lose from ending exits as we know them? News organizations would have to use actual voting data, measured against historic performance and census data, to "call" states and elections. But they already moved in this direction after the 2000 Florida fiasco, using actual votes to "correct" exits; that's why the networks did not erroneously call Ohio, Florida, and the country for Kerry this last time. Sure, America's political junkies would have to abandon the election day game of searching for and then widely disseminating (often distorted) exit poll findings in the middle of the day. But believe me, O ye junkies, it was actually a lot more fun way back in the pre-lapsarian, pre-exits era when you actually had to wait to figure out what had happened. Instead of sitting on your butt and expecting the results to land in your lap at 4:00 p.m., you did your homework, looking for oracular signs of trends in the early votes from long-time bellweathers like Campbell County, Kentucky. With all the resources now available on the web, election day sleuthing could become a lot easier, and a lot more entertaining--without exits.