On Sunday, I showed why the widely-reported claim by the notoriously-flawed $10 million National Election Pool exit poll that George W. Bush's share of the Hispanic vote leapt from 35% in 2000 to 44% in 2004 didn't match up with the actual votes counted.
Today, I will demonstrate that NEP's Hispanic share estimate—reached after massive data massaging to eliminate the embarrassing fact that it had originally predicted a solid Kerry victory—is internally contradictory.
A seemingly technical question—but actually crucial for Republican strategists. Secretary of State Colin Powell has already been in Mexico to revive talks over opening the borders, just days after the election. Yesterday, the Washington Times's Bill Sammon reported that President Bush has met with Senator John McCain to discuss "jump-starting" his amnesty proposal. ["Bush revives bid to legalize illegal aliens," November 10 2004.]
The press has been scratching its head all week, trying to figure out the reasons behind the NEP's Hispanic share data—really, the biggest surprise in an election of remarkably few surprises. The profile of Bush's 2004 voters is almost identical, regionally and demographically, to his 2000 profile, just three points larger…except, supposedly, for Hispanics.
The Latino figure in the exit poll is a particular shocker because the Republican campaign brain trust, which had spent the first years of the Bush Administration boasting about how it was going to win over lots more Hispanics, had lately given up pushing such claims. Poll after poll showed Bush was headed toward roughly the same performance with Latinos as in 2000.
Indeed, the Nov. 15th issue of Newsweek reveals in a chapter of its instant history of the campaign entitled "Down to the Wire" that Karl Rove's pollster Matthew Dowd wasn't expecting to get even 42% of the Hispanic vote.
In a paragraph devoted to Dowd's reactions during Election Day, Newsweek has Dowd thinking: "Bush seemed to be doing surprisingly well with Hispanics, winning 42 percent of their votes…"
Dowd, of course, oversees a private polling operation that dwarfs most of the public polls in size and accuracy. So his surprise at the 42% share (much less than the 44% number that the NEP ultimately came up with) shows that the Republicans' internal polls must have been pointing toward a Hispanic share down in the 30s.
There's a very simple explanation for this Hispanic-share surprise:
It didn't actually happen.
The only evidence of a disproportionately large Hispanic surge toward Bush is in this one NEP exit poll—which also, it's worth remembering, also predicted that Kerry would win by three points.
Reporters keep looking for actual physical locations where the Hispanic tidal wave carried Bush to victory. They aren't finding much.
Consider, for example, this map showing all the counties in the U.S. that switched from Democrat to Republican (in red) since the 2000 election. Very few counties in heavily Hispanic areas changed sides.
In my Sunday VDARE.com column, I showed how implausible was the exit poll's claim that Bush's share of Hispanics in Texas had skyrocketed from 43% to 59%.
Michelle Malkin has directed me to this Houston Chronicle article by Mike Tolson headlined "Latinos' support for Bush debated: Exit-poll math doesn't add up, one institute says." [Nov. 6, 2004,] The point:
"But if Bush actually did claim almost 60 percent of the Latino vote statewide, his overall margin over Kerry in Texas should have been closer to 70 percent, not the final 61 percent to 38 percent, Gonzalez said."
But you don't even have to compare the NEP poll to reality to see that it's untrustworthy. It's also internally inconsistent.
This is weirdly higher than Bush's share in the Southern states of Florida and Texas, which have the largest Hispanic populations.
I did some digging along the lines that Bolton suggested and quickly hit paydirt.
Here's the background: the NEP exit poll, as reported on CNN and other leading outlets, breaks out Presidential election numbers at three levels: nationally, regionally (East, Midwest, South, and West), and by states.
In each of the regions, not just the South, the sums of the individual states' number of Hispanic votes for Bush add up to less than the exit poll's total regional number of Hispanic votes.
The NEP reports the Hispanic share of the total vote in all states, but it only reports exactly whom Hispanics voted for in those states where there's a statistically significant sample size of Hispanics.
In the South, for example, only four of the fourteen states have enough Latinos for the NEP to break out Bush's and Kerry's shares: Florida, Texas, Georgia, and, last and least, Oklahoma.
By combining the exit poll data with turnout data from the United States Election Project, we can see that the Bush's Hispanic vote totals appear to be systematically inflated.
If we add up what the exit polls say was the total of Bush Hispanic votes from the broken-out states in each region, you repeatedly find that he would have had to have won an absurdly high share, often over 100% (!), in the other, unreported states in the region for the regional total to be accurate.
Let's start with the South. The exit poll claims Bush won a jaw-dropping 64% of the Hispanic vote there, up 14 points from 2000.
The South has the most Hispanic voters of any region, according to the exit poll—35% of the national total. So, if Bush's Hispanic share is exaggerated in the South, that would have a sizable effect on the national number.
Traditionally, Florida has the most Republican-voting Hispanics in the country due to its middle-class, anti-Communist Cuban population. The NEP poll reported that 56% of Florida's Hispanics voted for Bush. By contrast, a Florida exit poll conducted by the New Democrat Network claimed that only 46% voted for Bush, but no matter. Either number is still below that 64% Bush share the national exit poll claimed to find in the South overall.
Something is strange if Florida's Hispanics are less Republican than the regional average.
Less plausibly, the NEP exit poll alleged that Bush's share in Texas zoomed up to 59 percent. But even if we take that as gospel, that's still less than the 64% claimed for the South overall.
Florida and Texas between them have over 4/5ths of the South's Hispanics. So to get the overall Southern regional Hispanic share to 64%, Hispanics in the remaining Southern states would have had to be incredibly pro-Bush. Yet the NEP reports that the third largest concentration of Hispanics in the South, in Georgia, only gave 56% of their votes to Bush.
So, where are these hyper-Republican Southern Latinos hiding?
In Oklahoma, the exit poll claims Bush won a staggering 74% of the Latino vote, higher even than Bush's non-Hispanic white share.
This seems awfully unlikely.
But if you add up the Hispanic votes from these four states with broken-out shares for Bush, you still see Bush supposedly winning 1.730 million out of 2.981 million Southern Hispanic voters—only 58%. So what had to happen in the other ten states to get him to 64% for the South as a whole?
There are two ways to estimate this. The first is to simply subtract the four broken-out states total from the South's 14-state total and assume the remainder is the result in the other 10 states.
So, if Hispanics made up 9% of the 38.382 million voters in all 14 states of the South, then there must be 0.474 million Latino voters in the other ten states. And if Bush really carried 64% of Hispanics overall in the South, then he must have won 0.480 million Hispanic votes in those other ten states.
That means he won 101% of these states' Hispanic vote.
That seems a little … unlikely, even for Karl Rove.
Yet when you use the second and more reliable method for estimating how many Hispanics voted in the other ten states, the results get even more absurd.
Because the NEP reports Hispanic share of turnout for each state, you can estimate how many Hispanics voted in the other 10 Southern states: only 0.253 million. Bush still needs to have garnered 0.480 million votes to make his regional total. So his share of the Hispanic vote in the 10 hidden states was 190%.
I suspect this exceeds even Mr. Bush's expectations. [Numbers fans should click here to see the Southern exit poll results.]
Similarly, the exit poll claims that in the West region, Bush took 39% of the Hispanic vote. But in the eight broken-out states, which account for something like 97% of all Hispanic voters in the West, Bush only garnered 34%.
So for the unspecified states (Alaska, Wyoming, Idaho, Montana, and Utah) to raise Bush's regional share from 34% to 39%, their Hispanics would have had to cast about 167% of their votes for Bush.
In the Midwest, the exit poll purports that Bush won 0.489 million votes from 1.527 million Hispanics (32%). But in the four broken-out states, he won only 0.216 million out of 0.735 million (29%). So Bush would have had to capture 0.273 million in the unspecified states. The exit poll reports that there were just 0.222 million Hispanic voters in those other states. So Bush must have won a 123% share of them.
In the East, the situation isn't quite so preposterous. The exit poll reports that Bush won 28% in the whole region, and that's what he won in the reported states. However, to make his supposed regional total of votes would still require him to win 95% of the Hispanics in the unreported Eastern states.
Let's make two assumptions that are more realistic
That would put his Hispanic share at 38% to 39%, up 3 or 4 points from 2000, compared to his white share of 58%, which was up 4 points.
Historically, the gap between the white share and the Hispanic share stays relatively stable—and 2004 does not look like too much of an exception.
My conclusion: Bush scored at the high end of the GOP range for Hispanics—but he's not really broken the mold.
Moral for Congressional Republicans: don't ease your skepticism toward Bush's immigration proposals because you think you'll get Hispanic votes.