Following up on Tim Graham's NewsBusters report on a Washington Post article about a study claiming "that supporters of President Bush and other conservatives had stronger self-admitted and implicit biases against blacks than liberals did," I have a few questions I wish the Post story had answered.
Here's three paragraphs from the article, explaining that study. As you read them, ask yourself: Does the conclusion of paragraph three follow from what is said in paragraphs one and two?
For their study, Nosek, Banaji and social psychologist Erik Thompson culled self-acknowledged views about blacks from nearly 130,000 whites, who volunteered online to participate in a widely used test of racial bias that measures the speed of people's associations between black or white faces and positive or negative words. The researchers examined correlations between explicit and implicit attitudes and voting behavior in all 435 congressional districts.
The analysis found that substantial majorities of Americans, liberals and conservatives, found it more difficult to associate black faces with positive concepts than white faces -- evidence of implicit bias. But districts that registered higher levels of bias systematically produced more votes for Bush.
"Obviously, such research does not speak at all to the question of the prejudice level of the president," said Banaji, "but it does show that George W. Bush is appealing as a leader to those Americans who harbor greater anti-black prejudice."
The way I read this (particularly the sentence "But districts that registered higher levels of bias systematically produced more votes for Bush."), they gave people a test (the accuracy of which is not known) to determine if they are biased against blacks, and then correlated this not with the test subjects' own voting pattern, but the voting patterns in the Congressional districts in which the test-takers live.
(Am I reading the article correctly on that point? Maybe others read it differently.)
130,000 people divided by 435 Congressional districts yields an average sample size of 298 people per Congressional district.
Questions that come to mind:
1) Is a sample size of 298 big enough to accurately mirror the entire Congressional district's voting pattern vis-a-vis Bush?
2) Were the 130,000 test-takers more or less evenly distributed over the 435 Congressional districts, or were some districts evaluated on the basis of a few dozen or fewer residents? (For that matter, were registered and/or frequent voters treated the same in the study as non-voters/non-citizens?)
3) The Post article says the 130,000 people who took the test "volunteered online to participate." How were they recruited?
The Post concludes the article with a paragraph containing this quote: "Republicans are supported by whites with prejudice against blacks." As a straight factual statement, as far as it goes, that presumably is true, just as the statement "Democrats are supported by whites with prejudice against blacks" presumably also is true (though not something the Post said). Both statements are too generalized to be meaningful, however.
Just like this Post story.
