Paying fishers to save sharks killed more sharks

Sometimes when a surprising paper comes out, I don’t immediately read it to understand why its counterintuitive result occurs. So I keep it on the burner in the back of my mind. And when I do get to it, my satisfaction is all the greater for my curiosity having simmered for so long. That’s how it went with the first-ever marine conservation RCT, published by Hollie Booth et al. in 2025. The first ever? Yes: wildlife conservation is not like development economics. Of the many interventions that governments and NGOs perform to conserve wildlife in the ocean, Booth et al. is the first experimental evidence, at least as of last year. What about its surprising result? Paying Indonesian fishers for conservation actions actually made the problem worse. Indonesian fishers who caught hammerhead sharks got paid if they released them alive. But treated fishers actually increased the retention of hammerheads. This backfire is worse than it might sound at first. I could imagine treated fishers catching more hammerheads, in order to release them and receive the payment. This would be perverse. But not only do treated fishers catch more hammerheads, they keep them too, killing greater numbers of these critically endangered animals than if the conservation intervention had not been attempted in the first place. This negative result is important. Pay-to-release programs exist in India, Brazil, and other countries. They exemplify a broader category of conservation interventions, premised on the idea that incentivizing natural resource users to perform conservation actions will cause conservation benefits. The backfiring in Indonesia shows why conservation interventions should be experimentally evaluated before they are scaled up. The institutional context helps explain the backfire. Because these details are specific to the experiment’s setting, they also demonstrate the value of creating a culture of experiments in environmental programs, because effects likely differ by the design of each intervention, the economic environment it operates in, and the animals it targets. Why the payments backfired Based on pilot surveys of fishers, the researchers set the payment for a live release at roughly the average amount fishers could earn by retaining that fish and selling it at market. In addition to hammerhead sharks, fishers could receive payments for live releases of wedgefish, which are also critically endangered and look like a cross between a shark and a ray. I haven’t mentioned wedgefish until now because their effect was statistically insignificant. But they are the key to understanding the authors’ story of why their intervention backfired for hammerheads. The authors think they set the payment rate too high for wedgefish in the villages that drove the results. Their explanation, based on indirect evidence, is that treated fishers responded by fishing harder for wedgefish, raising their catch of both wedgefish and hammerheads, since hammerheads tend to occur in the same places and at the same times. The extra wedgefish could be released alive and paid for. The extra hammerheads mostly could not. Hammerheads must swim constantly to avoid suffocation, so once they are caught and immobilized only about half survive. Wedgefish can pump water over their gills while immobilized, and as many as 90% survive capture. A fisher who hauls up a dead hammerhead has no live-release video to submit and no payment to claim, so the only value left in the fish is at market. A near-empty evidence landscape I once co-authored a conservation RCT to see what it’s like. In publishing our paper, we benefited from a similar ability to make novelty claims as Booth et al. However, I found the logistics of running an RCT to be so difficult, compared to performing causal inference with secondary data, that I’ve never attempted a second RCT. But researchers with a greater appetite for managing field teams face a near-empty evidence landscape. Publication prospects are good, and potential policy impacts are high.