The experiment found evidence in the fact that the direct manipulation of the policy maker's identity has an effect on the policy outcome. More specifically, they found significant results in policy decisions when there is reservation for women in local government.
Another contribution is that under reservation policy there is more investment in public goods that are linked to women's preferences and less in those linked to men's preferences. The authors also make sure that this differences in the policy outcome are due to the gender of the policymaker and not because of other reasons, such as a greater concern of women about general interests, as the literature suggests.
This experiment might be interesting because reservations for women are increasingly being implemented at various levels of government also in other countries, such as Morocco, and, even at the lowest level of a decentralized government, all mechanisms that affect politician's identities may affect policy decisions. In this paper, thanks to the random assignment of the treatment and control group and the full compliance of the treatment group (and almost full compliance also for the control group), the ATE can be clearly identifiable. Furthermore, the random assignment of the policy status is well implemented and the effects of the treatment on behavior can be measured with reliability \cite{burtless1995}.
Nevertheless, there are some weak points that I would like to mention. First of all, when they explain the results, in Rajasthan we can see in Table 4 that men care more about education than women and the authors do not report any results of the number of educational centers, while in West Bengal yes when maybe this result also could be interesting for the paper. Moreover, the fact that Rajasthan does not increase women participation, however the authors justified this result, plus the fact that this increment does not seem to be persistent, due to they do not check if after that fact that there was a female Pradhan it increment women participation, and also female Pradhan are very unlikely to be elected again, makes the assumption that the policy increases women participation weak.
Secondly, there are also some weaknesses in the model. For instance, the number of complaints might not be such a good proxy for preferences if, for example, some women's complaints are influenced by the opinion of their male relatives. Additionally, the authors assume that everyone can influence in the same way when they complain (everyone has same weights) but that could be not true if maybe the complainant is rich or more influent.
Last, in the paper, they consider as an outcome of the reservation policy the number of policies the Pradhan did when he or she was in charge. Maybe another measure of the outcome such us the quality of the policies (e.g. the durability of the achievement project) could be also interesting to see. In the robustness checks, the authors considered the possibility that maybe the outcomes for women are affected by the fact that women have less experience. They took a subsample of all new Pradhans and checked that the results were not different from the ones obtained with the normal sample. However, they don't report if the durability of the project was longer, for example.