From Washington Post Associate Managing Editor Robert Kaiser

2004 has proven to be a Republican year.

But trying to explain exactly how this happened will not be easy, because our best available tool for doing so, the exit polls taken this year by the Edison Media Research and Mitofsky International, is broken. I’m not sure how this can be sorted out in the future, I assume we’ll have a good story about it in Thursday’s paper. Essentially, the problem is that the exit poll’s number don’t add up.

Here is one example of the problem. It’s a little complicated, so bear with me. The exit poll breaks down results by religion. It says that John Kerry did better than Al Gore did in 2000 among Protestants (43 percent for Kerry vs. 34 percent for Gore), Catholics (50 percent vs. 45), non-religious people (69 percent vs. 57) and those belonging to “other” religions (75 vs. 53). Only among Jews, a tiny fraction (three to four percent) of the electorate, did Bush do slightly better this time than last (22 percent of the Jewish vote in 2004, vs. 17 percent in 2000).

Since Protestants, Catholics, “other” and the non-religious constitute more than 96 percent of the population, these numbers suggest Kerry should have beaten Bush handily yesterday. But in fact, Bush won the popular vote by about 3.5 million votes, whereas Gore beat Bush in the popular vote by just .5 million last time. To say it another way, the exit poll claims that Kerry did a lot better than Bush with every major religious group in America, including the non-religious, yet he clearly lost the popular vote to Bush. Go figure.

I could cite a number of other such examples that I have found this morning by comparing the 2000 and 2004 exit polls. This makes me really nervous about some of the most interesting findings in the new exit poll—for example, the conclusion that “moral values” were more important to voters this year than terrorism, the war in Iraq or the state of the economy. Is that really plausible? I just don’t know. Similarly, I have to wonder about the finding that young people did not participate this time in substantially higher numbers than they did in 2000. The number of young voters did go up this year, the exit poll says, but only as much as the total number of voters went up, so their share of the electorate, about 17 percent, was the same this time as last.

I guess the cynic could ask, Is it the exit polls or the real polls that are broken?