Casper Dahlmann and Niels Bjørn Petersen in the WaPo:

We conducted a survey of 954 Danish local politicians. […] We then divided the politicians into two groups. One group got the data — but without any information as to whether the school was public or private. The schools were just labeled “School A” and “School B.” The other group got the exact same data, but instead of “School A” and “School B,” the schools’ titles were “Public School” and “Private School.”

If politicians are influenced by their ideologies, we would expect that they would be able to interpret the information about “School B” and “School A” correctly. However, the other group would be influenced by their ideological beliefs about private versus public provision of welfare services in ways that might lead them to make mistakes.

This is exactly what we found. Most politicians interpreting data from “School A” and “School B” were perfectly capable of interpreting the information correctly. However, when they were asked to interpret data about a “Public School” and a “Private School” they often misinterpreted it, to make the evidence fit their desired conclusion.