Counting Counts for Something

So after exit polling went 0 for 3 on Election Day, news groups plan to "delay distribution of data for several hours on future election days." The article is slightly unclear, but seems to be saying that data will be collected over the whole day before being released. (But if that's done, it wouldn't constitute a "delay" of the results, but a wait for complete results, right?)

The article also mentions that there will be a query into whether flawed methodology might have caused inaccurate results and that the news organizations will try harder to stop leaks.

Humorously enough, though, an official involved in looking into the problem points his finger at ... the blogosphere! "There were an awful lot of people on the Internet talking about things they don't understand ... First wave exit poll data is not terribly accurate." Said Sheldon Gawiser, head of the news organizations' steering committee, talking about something he doesn't understand: how the open exchange of information in the blogosphere quickly leads to the debunking of flawed theories and of bad information. For my part, I recall how I learned of the early exit polling data that showed Kerry winning in a landslide via the internet. Soon after, the same blog expressed misgivings over the quality of the data!

Another flaw in the article lies in its focus on just the prediction of the winner of the election: "Those first numbers reflect only interviews with people who have voted in the morning. By delaying release of numbers until between 4 p.m. and 6 p.m., pollsters will have talked to more people and, it is presumed, have a more accurate snapshot of the electorate ..." But what about the other mistakes that were uncovered in the exit polls, namely the high Hispanic vote for Bush and the bad questioning methods that led to the myth that "moral values voters" tipped the election for Bush? Without serious changes in their methodology aside from waiting for "more complete results," the exit pollers won't be any more accurate next time.

The actual conduct of Old Media on Election Day seems to contradict this display of a passion for accuracy. I still remember the call of Florida for Gore before the polls in the Bush-leaning panhandle closed on Election Day 2000. And I remember waiting up for Ohio to be called for Bush long after similar-sized Pennsylvania had been called for Kerry (and by a smaller margin than Bush had in Ohio) in 2004. I can perhaps be forgiven for being cynical when I see an article like this. It looks too much like the Old Media are making a big deal over the appearance of objectivity so they can be in a better position to try to affect the outcome of the next election with phoney results. But I suspect that the blogosphere will prove not only nimble in the next race, but possessed of a long memory. The exit polls will likely carry less weight next time.

Certainly, there can be lessons gleaned from the last election that can make exit polling a more accurate predictor of the results of the election, but it will always be just a statistical sample and subject to the discretion -- and therefore the errors or prejudices -- of the pollster. This is the same reason the idea of "correcting" the 2000 Census by statistical methods was rejected.

What is it that Democrats and their allies have against counting? I think I just answered my own question. Regardless of how well exit polling can predict an election, the election itself is the only poll that really counts.

-- CAV