The Associated Press had a huge story this weekend accusing American voters of racism, but a racism that has increased in the last four years. When people were asked if they were racist, they said no, but when a survey that measures racial preferences via tiny pictures and Chinese characters was used, the racism was found. The subhead at USA Today is "Overall, the survey found that by virtue of racial prejudice, Obama could lose 5 percentage points off his share of the popular vote on Nov. 6." I'm still working through my thoughts about the story, how it was chosen, what it means that the Associated Press chose to investigate this story, the methodology they used, how they interpreted the results, and so on and so forth. But while digging through the data, I came across something very weird. The question respondents were asked was:

Do you happen to know the religion of each of the following people? If you don’t know, you can mark that too.

Here are the answers they gave for "Barack Obama" for 2010 and 2012:

It's not uncommon to see slight variations in data over two years. For instance, 28 percent gave the answer of Protestant, up from 26 percent two years ago. Five percent believe Obama to be Catholic up from 4 percent in 2010. That all seems reasonable.

But what about that 18 percent thinking Obama is Jewish, up from nobody in 2010? Hunh? Or what about 35 percent believing Obama has no religion, up from 2 percent just two years ago? And we're to believe that 41 percent of people didn't know Obama's religion in 2010 but now only 2 percent report that? And why the shocking increase in those who refused to answer?

I don't really know what this means, but it's exceedingly hard to buy the idea that we'd see this many changes in just two years.

Does it make sense to you?

It also makes it difficult to trust the overall results when this section is so weird.

Fishy image via Shutterstock.