Here are some observations on the 2004 exit poll data, based on the latest version of the data available. There is much to be explained and understood about these data and certainly legitimate questions can be raised about some of the findings. But the first task is simply to clarify what the poll actually says and does not say, because there is considerable confusion about this.

The figures used here are not the final figures, but, based on my experience in previous election cycles, they probably vary only slightly from the final numbers available when the exit poll authorities (in this case, the National Election Pool (NEP)) release a cleaned-up dataset in a couple of months with final weights.

All 2004 figures discussed here refer to the NEP exit poll not the Los Angeles Times exit poll, since the NEP poll is both substantially larger and far and away the most widely-used and cited. All 2000 figures refer to the 2000 exit poll by VNS, the NEP's predecessor.

1. Gender. According to the NEP poll, Bush carried men by 11 points (55-44), exactly the same margin he had in 2000 when he carried men by 53-42. Among women, however, Kerry's margin was only 3 points (51-48), down from the 11 point margin Gore had in 2000 (54-43). No matter how you measure the gender gap (add the margins and divide by 2 or simply subtract Democratic male support from Democratic female support), this means a substantial compression of the gender gap (from 11 or 12 points, down to 7) and it is entirely due to the Democrats' reduced margin among women.

2. Race. According to the NEP poll, 23 percent of voters this year were minorites, up from 19 percent in 2000, indicating the continued rapid expansion of the minority electorate.

The NEP poll says, however, that Bush widened his margin among white voters--still 77 percent of voters--to 17 points (58-41), up from a 12 point margin (54-42) in 2000. And among hispanics, now 8 percent of voters, the poll indicates a Kerry margin of only 53-44, a dramatic compression from Gore's 62-35 margin among the same group in 2000.

However, there is some dispute about whether the compression of the Democratic margin was as severe as indicated by this poll. An exit poll of Hispanics only by the William C. Velásquez Institute of San Antonio, which sampled 54 counties in the 14 states with the largest number of Latino registered voters, had 68 percent voting for Kerry and only 31 percent voting for Bush.

To further sow confusion, the NEP data on hispanics are now being reported in two different ways--as above, at 8 percent of voters and 53-44 Democratic support and at 6 percent of voters and 56-43 Democratic support in Sunday's New York Times. How did hispanics suddenly get demoted to 6 percent of voters? The answer is complicated, but here it is: the NYT for purposes of their historical chart uses a single race question to capture hispanics, as opposed to a race question plus another question on whether the respondent is of hispanic descent or not, which was included on both the 2000 and 2004 exit polls and is now used by CNN and practically everyone else. The NYT's reasoning for not using this new (and better) two question measure of hispanic respondents is that the historical hispanic data in their chart will at least all be measured in the same way.

I certainly see the point in apples-to-apples comparisons. On the other hand, since the new measure is undoubtedly a better one and we don't really believe the hispanic proportion of voters in 2000 was only 4 percent (as the single question hispanic series also indicates) and just 6 percent this year, it would be better, I think--as well as less confusing--for the NYT to go with the data that is the best and simply acknowledge a discontinuity in the exit poll times time series on hispanics between 1996 and 2000.

People will, after all, play the closest attention to the figures--both support rates and proportion of voters--from this year and, secondarily from 2000. Given that the figures from the two question hispanic measure for those two years are (a) better and (b) comparable with one another, it strikes me as a good idea to feature those data rather than the misleading single question hispanic data. Again, the discontinuity can then be footnoted for those that get into the data that far, but the average NYT reader will be provided with the most accurate measure of hispanic turnout and presidential support.

The data on blacks are much more straightforward. Among blacks, Kerry had an 88-11 margin, down only slightly from 2000's 90-9 margin for Gore. In fact, except for 2000 and Mondale's 1984 campaign, Kerry's margin among blacks is the highest obtained by a Democratic candidate since the exit polls started in 1976.

More exit poll fun tomorrow!