...rain on the previous election day lowers current voter turnout.
Economists and political scientists have observed that a citizen who votes today is more likely to vote in the future, but determining whether that is the result of unobserved individual attributes, or the effect of voting per se, is difficult. In Estimating Habit Formation in Voting (NBER Working Paper No. 19721), Thomas Fujiwara, Kyle Meng, and Tom Vogl conclude that voters are creatures of habit. They estimate that a 1 percentage point decrease in past voter turnout (measured as a share of eligible voters) lowers current turnout by 0.7 to 0.9 percentage points.
To investigate whether voting is habit-forming, the authors match daily weather data to U.S. county-level presidential election returns and population data for all elections between 1952 and 2012. They find that rain on the previous election day lowers current voter turnout. Precipitation has no effect on turnout if it occurs two weeks before or after election day. When it rains on election day the depressing effect on turnout is larger if it also rained on the previous election day. The findings can be illustrated by considering the effect of 1 inch of precipitation on election day- an outcome that is in the 98th percentile of the election day precipitation distribution. If the previous election day was free of precipitation, then 1 inch of precipitation on the current election day reduces turnout by 1.4 percentage points. If there was a similar amount of precipitation on the previous election day, then the reduction in current turnout averages 1.6 percentage points.
The effect of precipitation on turnout is roughly linear, with 1 millimeter of rain (0.039 inch) decreasing voter turnout by 0.07 percent. Although election day precipitation has roughly the same effect on voters from different parties, it depresses turnout more in less densely populated areas.
The authors conclude that their findings are best explained by the view that getting into the habit of voting depends more upon people's perceived rewards from the act of voting itself than on any benefit people expect to obtain from influencing election outcomes. Changes in voters' costs of voting and their beliefs that they might cast a deciding vote do not seem to explain the results.