Politics Interferes with Reasoning

Americans are really bad at math. Various scientists and educational reforms are always arguing that if people were better educated at math and at scientific thinking more generally, they would be better able to understand issues like climate change. I have never believed this myself; consider how many mathematically and scientifically astute people are libertarians.

Now we have some real data showing that people's politics interfere with whatever limited math skills they do have. (Good summary here.) Yale law professor Dan Kahan and his colleagues gave a bunch of people political questionnaires, to determine their ideologies, a math test, to determine their "numeracy," and then a single math problem that was the real meat of the experiment. The math problem had two versions. The numbers were the same in both cases, but in one the problem was described as a study of whether a new hand cream made rashes go away faster. In the other, the problem was described as a test of whether bans on carrying concealed handguns influenced the crime rate. Both problems were given to half the people with the data reversed; that is, for half the participants the hand cream worked, and for half it did not. Here is one version of the skin cream problem:

Not surprisingly, when the problem was about hand cream people of all political persuasions got it right or wrong at about the same rate, based on their numeracy, and whether the outcome was positive or negative did not matter. Sadly, 59% got the problem wrong, but never mind that. When the problem was about gun control, the result mattered. Both liberals and conservatives were much more likely to get the problem right when the answer supported their ideologies, and much more likely to get it wrong when it did not. This was true especially of the most numerate people in the study; the more numerate people were, the bigger the discrepancy between the two versions. Even for people above the 90th percentile in numeracy, only 57% got the problem right. So while the results on the hand cream problem were entirely predictable just based on the numeracy scores, for the guns and crime problem "the impact of numeracy is minimal."