The Samaritan’s dilemma, a term coined by James Buchanan in 1978, refers to the idea that if someone expects to be bailed out when a disaster strikes, they will take fewer precautions. If fewer precautions are taken then the disaster will be more severe. The “Samaritan,” i.e., the government, has a dilemma because policy intended to help out in a bad situation might actually make matters worse.

But is the Samaritan’s dilemma really a thing? My coauthor, Tatyana Deryugina, and I have found the first evidence that the Samaritan’s dilemma matters on a broad scale. Although the idea has been around for almost 40 years, until very recently we have had no evidence that the Samaritan’s dilemma even existed. That’s because researchers face two problems when investigating the Samaritan’s dilemma. The first problem is the rarity of disasters; it is hard to get a good sample size for analysis. The second problem is coming up with the right counter factual.

Conceptually, the problem of measuring the Samaritan’s dilemma can be solved by observing parallel universes. In one universe, the government offers a bailout when a distaster strikes. In the other universe, it is impossible to offer a bailout. Otherwise, these two parallel universes are identical. We determine the
effects of the Samaritan’s dilemma by comparing the precautions taken by the same people in each universe. If we find a difference, it must be because of the potential bailout. In reality, however, we cannot observe parallel universes. Instead we need a way to randomize who should expect a bailout and who shouldn’t.

We solve both of these problems in our working paper “Does the Samaritan’s Dilemma Matter? Evidence from U.S. Agriculture.” We solve the first problem by focusing on U.S. agriculture in 1990-2010, when every year, farmers somewhere in the country suffered from extreme weather. In fact, the U.S. government issued disaster payments in every year of our time frame. Over twenty years we observe a lot of agricultural disasters.

To address the second problem, we focus on the political process behind disaster payments. Disaster payments require emergency authorization from Congress. Since congresspeople face frequent elections, incumbents are always trying to curry favor with voters. One strategy they have is to target voters who voted against them in the previous election but can be persuaded to change their vote this time. Third-party voters seem to fit that bill. Our analysis finds that Congress is more likely to send disaster payments to areas with a lot of third-party voters in the previous presidential election. In other words, if two very similar counties are both hit by a disaster, the county with more third-party voters will get more disaster payments. Since third-party voting is unrelated to disasters experienced by local farmers, we use the share of voters that voted third party to design our “experiment.” Loosely speaking, places with a high share of third-party voters are part of the “treatment” group, and places with a low share of third-party voters are the “control” group.

Using this approach, we find that the Samaritan’s dilemma really is a thing. Farmers who are more likely to receive disaster payments simply because of voting patterns purchase less crop insurance than farmers who are otherwise statistically identical. We find that farmers who expect 10% higher disaster payments purchase 2% less crop insurance. They also spend less on farm labor and fertilizer, and have lower revenue from crop sales.

Theory tells us the solution to this problem is to provide free crop insurance, which seems to be the direction farm policy is heading.