“Showing scientifically what works, what does not work”

Clément de Chaisemartin is one of the three economists, excluding the winner, who have been selected by the jury bringing together representatives of the Circle of Economists and the World, for their work relating to applied economics and promoting public debate.

You graduated from HEC in 2008 and are now a researcher in economics, specializing in public policy evaluation methods. This is an unexpected trajectory!

I was indeed thinking of going to the private sector, to be an entrepreneur, which is more in line with my family. But, during an internship in a bank in Spain, I discovered “impact assessment”: the bank randomly divided a list of potential customers into two groups and sent a commercial offer in a white envelope to one group. , and in a green envelope to the other group.

By comparing the response rate of the two groups, we could measure the impact of the color of the envelope. It fascinated me to be able to show scientifically what works and what doesn’t, but by extending it to more interesting fields, such as the evaluation of the effects of public policies. I heard about the methods used by Esther Duflo and I undertook a master’s degree in economics at the Paris School of Economics.

In my dissertation, I measured the effect of a new smoking cessation treatment, which had been widely adopted by certain tobacco centers and much less by other centres. To analyze this kind of “natural experiment”, economists have long used the method known as “regressions with double fixed effect”. [RDEF].

Read also: Article reserved for our subscribers Eric Monnet, winner of the Best Young Economist Award: “History helps bring complexity back to economic theory”

But the purpose of your work is precisely to criticize this method and to propose a new one…

I started, in 2011, with my colleague Xavier D’Haultfoeuille, to show the limits of the RDEF. David Card [Prix Nobel d’économie 2021] is the pioneer of natural experiments in economics. His work generally compares two places, one of which implements a policy, say an increase in the minimum wage, while the other does not. The RDEF then makes it possible to measure the effect of the policy, under the assumption that, in its absence, the variable affected by the policy (for example the employment rate) would have followed the same evolution in the two places.

But most natural experiments are more complex: multiple places set the policy in motion, at different times and with different intensities. In this case, we have shown that the RDEF only makes it possible to measure the effect of the policy if this effect is constant in space and time, which in most cases is implausible. We therefore proposed a new evaluation method.

You have 34.21% of this article left to read. The following is for subscribers only.

source site-30