FEJIV: Module for Fixed Effect Jackknife IV (FEJIV) Estimation

[Tymon Sloczynski]

I would like to announce that fejiv, a Stata module for fixed effect jackknife IV (FEJIV) estimation, is now available on SSC.

fejiv implements the fixed effect jackknife IV (FEJIV) estimator of Chao, Swanson, and Woutersen (2023), which enables consistent IV estimation with many (possibly weak) instruments, cluster fixed effects, heteroskedastic errors, and possibly many exogenous explanatory variables.

Consistency of the FEJIV estimator requires that instrument strength satisfies a key growth condition: the concentration parameter must grow faster than the square root of the number of instruments. Mikusheva and Sun (2022) show that this condition is necessary for the existence of a consistent estimator and also propose a test of it, implemented in the Stata command manyweakivpretest, available on Liyang Sun's GitHub.

One of my recent papers ("When Should We (Not) Interpret Linear IV Estimands as LATE?"; available at https://arxiv.org/abs/2011.06695) recommends the FEJIV estimator as an alternative to two-stage least squares (2SLS) when estimating the fully interacted specification of Angrist and Imbens (1995). Within the local average treatment effect (LATE) framework, when strong monotonicity is doubtful but weak monotonicity is plausible, the fully interacted specification eliminates the problem of "negative weights."

[Joao Santos Silva]

Dear Tymon Sloczynski ,

Many thanks for this very useful command right in saying that clustered standard errors are not supported?

Best wishes,

Joao

[Tymon Sloczynski]

Dear Joao Santos Silva:

Thank you very much for your interest and kind words! The command implements only the formula for standard errors presented in the paper by Chao et al. (2023). An interesting question is whether the cluster-sample setup in that paper doesn't imply that the standard errors are already "clustered" in the usual sense. I'm not saying this is the case, but I'd need to think it over more to rule it out. Any thoughts on this topic would be very much appreciated.

Let me also note that fejiv isn't currently as fast as I'd like it to be. I have some ideas on how to speed it up, and we'll be testing them in the near future. I'll post again here if there are any significant improvements.

Best wishes,

Tymon

[Tymon Sloczynski]

I also failed to mention in my initial post (though you'll naturally find this information in the SSC distribution, help file, etc.) that the command is coauthored with Qihui Lei, a fantastic PhD student at the University of Wisconsin.

[Joao Santos Silva]

Thank you very much, Tymon Sloczynski. And my apologies for the fact that some words were missing from my initial post