[pr_nik2]

Legacy papers are very popular these days in econ, history, and political science. I don't have the time to dive deep into this one, but spatial identification of long-term causal effects is prone to errors. Specifically, the error terms in regression specifications capture unexplained variance that can be very persistent over long periods of time. These errors can also be spatially correlated across treated observations: https://voxeu.org/article/standard-errors-persistence

[thelamest]

I can't dive deep either, just a quick note that the OP article reference list includes the (important) review you linked.

[pr_nik2]

That's great. Not all methodological insights get adopted by applied studies. Thanks for pointing this out!