Looking forward to reading this! FWIW, I worry so much about error in legal variables because some common sources of error seem likely to be correlated with characteristics of jurisdictions, outcomes of interest, etc. https://twitter.com/scottburrisphlr/status/1306203068808204290">https://twitter.com/scottburr...

It& #39;s not just about noise - it& #39;s neither reasonable nor useful to expect *anything* to be measured without noise. Lawyers also have a non-zero error rate - no matter how much we are trained to pay attention to detail through rigorous blue-booking exercises :-)

But, some sources of error would seem to have a much more insidious impact. So, for example, think about a common problem in legal measurement - operationalizing a variable on the state level when much of the policymaking actually occurred on the local level.

Since it is common that policy innovations originate at the local level and then filter up to the state level (see, e.g., https://tinyurl.com/y4ecbxxv ),">https://tinyurl.com/y4ecbxxv&... I worry that this will systematically underestimate the proportion of the population exposed to the treatment (depending on context)

Moreover, policy diffusion literature suggests that policy transmission from local to state government depends on the characteristic of the state (e.g., https://tinyurl.com/y62jazzv )">https://tinyurl.com/y62jazzv&... - as does the passage of preemption laws preventing local policymaking ( https://tinyurl.com/y3ncr25x )">https://tinyurl.com/y3ncr25x&...

It therefore seems likely that the difference between the legal variable constructed and the policy as it exists in practice would likely be correlated with characteristics of the jurisdiction - some of which (like legislative professionalism) are often not accounted for.