Weather or Not People are Bayesians

A new paper by Tatyana Deryugina finds that people make inferences about global warming from local weather but, given that they use local information, their inferences are mostly consistent with rational updating with some deviations in the very short run. Much more important than local weather, however, are other factors such as education and ideology.

…a Bayesian who is perfectly informed about world weather and science should

not give signicant weight to recent weather in his county when updating his beliefs. However, I

find that some forms of temperature and precipitation abnormalities have an effect over short time

scales of 1-2 days. Average weekly deviations and extreme events such as heat waves or droughts

weeks or months before the survey have no effect on beliefs, suggesting that the short run effects

are temporary and due to psychological heuristics.

Unlike previous studies, I also consider the effects of prolonged periods (1-12 months) of

abnormal weather. I find that abnormally low precipitation and abnormally high temperatures are

signicant predictors of the degree to which people believe the effects of global warming have

already begun to happen. The estimated patterns are consistent with how a Bayesian who only

observes local information would update his beliefs, but I cannot rule out that informed individuals

simply overweight their local weather.

…The marginal effects of education, relative to high school [on "the effects of global warming have already begun to happen"]  is 0.045 for “some college”, 0.101 for “college”, and 0.166 for “graduate school” A day on which precipitation is 2.5 standard deviations above normal would produce a change

in beliefs about the timing of global warming comparable to the estimated correlation between beliefs and “some college”. Precipitation would have to be 8 standard deviations above normal to produce a change in beliefs comparable to the coefcient of “graduate school”…In addition, the [weather] effects are short-lived because the average standard deviation over the past week does not change beliefs.

Hat tip: @jzilinksy via @bryan_caplan.

Addendum: Yes, the title of the post was on purpose!