From reading this discussion, there is one sentence
Standard errors should be clustered at the city level, i.e. the level
of aggregation at which the treatment occurs
In Dasgupta, 2019‘s paper, the aggregation level is country (laws have impact on all firms in a country).
I am quite new to cluster standard error but I saw the paper of Dasgupta, 2019, in some cases, he did cluster for the country (table 11), in some cases, he did the cluster for industry (table 8), in some cases, he did cluster for firm (table 12).
I am quite curious about this way of clustering. It is because from this discussion , it seems that the old "rules of them" is 30-40 groups. However, in Dasgupta’s paper case, he has 64 countries and hundreds of thousands of firms, so why he need to cluster for firms and countries?