++Addition++Wikipedia's election page has some errors in voter totals (as reported three days after the election; it may have been subsequently edited yet again by the time of this reading). Kentucky's total was overstated by more than 500,000 votes, and a few states were presented as final results even though the totals displayed did not represent 100% of precincts reporting. As Kentucky is a white state, the table presented in the body of this post slightly inflates the white national total. I've kept my original table in the body (I'm only into revisionism when it serves a constructive purpose other than covering my behind), but also present this more accurate one constructed from state voter totals provided by CNN:
The shifts are very minor. The white percentage drops by .2%, black by .2%. The Hispanic percentage rises by .2%, Asian by .1%, and other by .1%.
Steve Sailer has already predicted exit polling data overstated the Hispanic turnout this election (as it did in the 2004 Presidential election):
Hysterical pundits will announce that the Hispanic tidal wave accounted for 8 or 9 or even 10 percent of the vote!It's not even necessary to wait that long to undercut the 9% figure now deemed the official national total by Edison Media Research and the major media outlets that partner with it.
Then, a year from now, the Census Bureau will quietly announce the results of its huge post-election survey of voting, the gold standard of ethnic voting shares. It will show that the Hispanic share of the vote, which was 5.4 percent in 2000 and 6.0 percent in 2004 actually was only 6.9 percent in 2008, or whatever.
On election night, after exit polling data were in for all 50 states and DC, CNN's national exit poll showed Hispanics constituting 8% of the national total. Steve astutely captured and posted this before it was altered.
An easy (albeit tedious) way to verify the national figure is to simply look at exit polling data from each state. Taking the voting percentages by race in each state and multiplying them by the total number of votes cast in the state allows for the number of actual votes cast by race to be determined. The numbers are rounded to the nearest percentage, but there is no reason they should systematically differ across states (7.54% as 8% here should be balanced out by 12.46% as 12% there) to an extent that they misrepresent the national total, and anyway the national exit polls round to the nearest whole percentage point as well. Presumably EMR creates the national exit poll by aggregating each of the individual state exit polls so this does not offer an explanation for the abrupt post-election editing.
What do the exit polls actually say when aggregated together to form a national exit poll? The table shows the total number of voters by race (in millions), what the states actually reveal in percentages of the total by race, and what the major media are now running with:
That's not carelessness on my part--the media total really does come to 101% even though exit polls at the state level come to exactly 100% (and match what CNN originally reported, with the exception of the "Other" category, which varied by one point). Funny how to 'balance' things out, EMR seems to have arbitrarily taken 1.2% of the total respondent base that was white and converted it to Hispanic, er, "Latino", in the national exit poll.
Even that 7.8% will likely prove to be an overestimate, since smaller groups tend to be oversampled as a way of ensuring they're not undercounted, since it would be more problematic in determining voter tendencies to undercount Asians or Hispanics than it would be to undercount whites.
Data, via Swivel, are here.