Lies, Damn Lies, and Statistics

The amnesty crowd is hawking a new fable — that a massive margin among Hispanics saved Harry Reid:

Longtime immigration advocate Rep. Luis Gutierrez (D-Ill.) credited Hispanics Wednesday for keeping Majority Leader Harry Reid (D-Nev.) in the Senate.

"Latino citizens responded to Majority Leader Harry Reid's aggressive pursuit of immigration reform by voting for him in overwhelming numbers," he said in a statement. "They were clearly the difference in his victory."

Latinos made up 12 percent of Nevada voters, with 90 percent of them supporting Reid. Angle got only an 8 percent sliver of their support, according to a Latino Decisions poll.

The implication is that these numbers come from a "Latino Decisions" exit poll — an implication the authors of the poll are not in a hurry to dispel. Ben Smith at Politico, as just one example, was taken in: "Reid got an amazing 90% of the state’s 12% Hispanic voters, according to exit polls; Sharron Angle got just 8%."

Except that the Latino Decisions poll wasn't an exit poll — rather, it was an "election eve" poll, conducted before Election Day, just like all the other polling that has been going on for months. The actual exit poll, over at CNN, shows Angle getting a normal Republican share of the Hispanic vote, 30 percent.

Now, Latino Decisions claims their poll of people who haven’t voted yet is better than a poll of people who have actually, you know, voted. I’m not a statistician, and who knows, they may be right, though that would mean that Hispanics vote even more solidly Democratic than blacks, which is improbable in the extreme. But I’ll ask just this: Which poll of an ethnic group would seem more credible on its face — one conducted by a relatively disinterested news organization, or one conducted by a corporation whose very existence is based on hyping the importance of that same ethnic group, and which is paid for by the National Council of La Raza, the SEIU, and the hard-left open-borders group America's Voice?