November 18, 2004
Another look at Hispanic voting for Bush

Oddly buried in the This Week community section of today's Chron is this article on Hispanic voting patterns from the 2004 election. This bit in particular piqued my interest:

Nationally, Latinos cast 8 percent a little more than 9 million votes of the total vote in the Nov. 2 election, and there's still some controversy as to whether 44 percent of the Hispanic vote went for Bush, as exit polling indicated, said Tatcho Mindiola Jr., director of the Center for Mexican American Studies and associate professor of sociology at the University of Houston.

"It's a considerable improvement over what he received last time," Mindiola said. "The question is why."

In Texas, Hispanics accounted for 23 percent or 1.6 million votes of the number cast statewide, he said.

In other states, "there is no question," he said, that Hispanics came out to the polls in large numbers, accounting for 21 percent of the vote in California, 32 percent in New Mexico, 10 percent of voters in Nevada and in Hawaii, 9 percent in New York and 3 percent in Ohio.

"South Texas still went for the Democratic Party, whereas the border went for President Bush," Mindiola said.

Twenty-three percent of the total vote coming from Hispanics (which by my calculations would be 1.7 million voters, since 7,389,864 people cast ballots in the Presidential race) is the figure that Big Media Matt cited in demonstrating that 59% Hispanic support for President Bush in Texas was plausible. I disputed his calculation based on my belief that 23% was too high a figure. Now here it is again, which means that maybe I'm full of it. So let's take another look at the numbers and see what they tell us.

In the most heavily Hispanic counties in Texas (see spreadsheet), Bush got 44.5% of the vote. 471,000 votes were cast there; 85% of the population is Hispanic, so if they voted proportionally that's 400,000 voters, of which about 178,000 voted for Bush.

That leaves 1.3 million Hispanic voters in the rest of the state. If Bush got 59% of the Hispanic vote in Texas, that means a smidgeon over 1 million Hispanics total voted for him, or 822,000 of the Hispanics in these other counties. That comes to 63.2% of the Hispanic vote in all the other counties in Texas, including Bexar, Dallas, and Harris, in which Bush got 53.5% of over 2.2 million voters overall. These three counties would account for a huge proportion of those 1.3 million remaining Hispanic voters, and if they were going for Bush to the tune of 63.2% there's no way in hell he gets only 53.5% of the vote overall there. White voters are still the bulk of the electorate in the big urban counties, even as they drop to non-plurality of the total population (as an example, George Strong assumed 68% of the voters in the 2003 Houston mayoral runoff vote would be Anglos), so the only way to make these numbers add up is to assume that white voter support for Bush dropped like a rock in the big cities. Once again I say, no way in hell.

I continue to believe that the 23% figure for the Hispanic portion of the Texas electorate is too high, but there is another factor to consider when evaluating both that claim and the claim that Bush got 59% of their vote, and that factor is the percentage of the Anglo vote that Bush got. Yglesias thought that 72% white support for Bush was pretty darned high. Maybe so, but in Texas, maybe not. Consider the following:

County Pct Anglo Bush vote Pct Bush vote Total vote
Collin 76.1 173,014 71.23 242,889
Denton 76.0 140,541 69.99 200,791
Ellis 71.3 34,577 74.59 46,352
Johnson 83.2 34,739 73.54 47,232
Lubbock 62.5 69,675 75.28 92,548
Montgomery 81.4 104,361 78.10 133,624
Wichita 73.3 32,472 71.29 45,545
Williamson 73.5 83,079 64.97 127,857

Total 672,458 71.77 936,838

The "Pct Anglo" figures are 2000 Census numbers for non-Hispanic whites. Presumably, their share of the electorate in those counties is higher than those Census figures. As you can see, other than those slackers in Williamson County, these heavily Anglo areas came out pretty strongly for Bush. I don't think 72% of the Texas Anglo vote going to Bush is too much - heck, I'm not sure it's not 75%. Put it all together, and I think there's a pretty strong downward pressure on both the Hispanic share of the electorate, and on the percentage of those voters who went for Bush. I continue to feel confident about my original calculations.

UPDATE: Ruy Teixeira performs similar calculations and comes to the same conclusions.

Posted by Charles Kuffner on November 18, 2004 to Election 2004 | TrackBack

I think we've learned a little bit about trusting exit polling.

Posted by: Tim on November 18, 2004 6:13 PM

Good sleuthing; those exit poll Hispanic Bush percentages are thoroughly unbelievable. One can conclude that deliberate faking of the numbers was resorted to. They would have started with a gross excess of minority precincts, causing the overall numbers to move falsely toward Kerry. This had to be fixed in a few hours somewhere. Now they can't easily go back towards the truth, if they deleted Latino Kerry respondents then, and now they would have to reduce the Hispanic participation from 8 to 6%, all from Bush support. On top of that, the black vote stated, would have to be made to drop from 12 to 10 or lower, thus maintaining the Bush/Kerry margin.

Posted by: John S Bolton on November 24, 2004 1:41 AM