Dear Stata users,
A new command ssaggregate is available through SSC. It helps you implement shift-share (or "Bartik") research designs, in which the instrument averages a set of shocks with unit-specific weights measuring shock exposure. For example, a regional instrument is constructed from some industry shocks averaged using local employment shares, as in Bartik (1991) and Autor, Dorn, and Hanson (2013).
Based on Borusyak, Hull, and Jaravel (2018), our command exploits an equivalence result. In the example of regions and locations, the regional shift-share IV coefficient can be identically obtained from a different IV regression estimated in the sample of industries. In this regression the outcome and treatment are first averaged with exposure weights to obtain industry-level aggregates; industry shocks then instrument for aggregate treatment. ssaggregate produces those industry-level aggregates.
The equivalent industry-level representation is useful when identification relies on as-good-as-random assignment of industry shocks (even if local shares are endogenous)---a quasi-experimental framework that may be plausible in many applications where the number of industries is large. The industry-level regression then helps visualize the identifying variation, produce corrected standard errors, test identifying assumptions, optimally combine multiple industry-level shocks, and more (see Borusyak, Hull, and Jaravel 2018).
For an example of usage, please refer to our replication archive, which applies these suggestions in the setting of Autor, Dorn, and Hanson (2013). We hope the command will be useful in your own shift-share designs, and please let us know if you have any questions!
Kirill
References:
Autor, D. H., D. Dorn, and G. H. Hanson (2013): The China Syndrome: Local Labor Market Impacts of Import Competition in the United States, American Economic Review, 103, 2121-2168.
Bartik, T. J. (1991): Who Benefits from State and Local Economic Development Policies? W.E. Upjohn Institute for Employment Research.
Borusyak, K., X. Jaravel, and P. Hull (2018): Quasi-Experimental Shift-Share Designs. NBER Working Paper 24977 (replication archive: https://github.com/borusyak/shift-share).
A new command ssaggregate is available through SSC. It helps you implement shift-share (or "Bartik") research designs, in which the instrument averages a set of shocks with unit-specific weights measuring shock exposure. For example, a regional instrument is constructed from some industry shocks averaged using local employment shares, as in Bartik (1991) and Autor, Dorn, and Hanson (2013).
Based on Borusyak, Hull, and Jaravel (2018), our command exploits an equivalence result. In the example of regions and locations, the regional shift-share IV coefficient can be identically obtained from a different IV regression estimated in the sample of industries. In this regression the outcome and treatment are first averaged with exposure weights to obtain industry-level aggregates; industry shocks then instrument for aggregate treatment. ssaggregate produces those industry-level aggregates.
The equivalent industry-level representation is useful when identification relies on as-good-as-random assignment of industry shocks (even if local shares are endogenous)---a quasi-experimental framework that may be plausible in many applications where the number of industries is large. The industry-level regression then helps visualize the identifying variation, produce corrected standard errors, test identifying assumptions, optimally combine multiple industry-level shocks, and more (see Borusyak, Hull, and Jaravel 2018).
For an example of usage, please refer to our replication archive, which applies these suggestions in the setting of Autor, Dorn, and Hanson (2013). We hope the command will be useful in your own shift-share designs, and please let us know if you have any questions!
Kirill
References:
Autor, D. H., D. Dorn, and G. H. Hanson (2013): The China Syndrome: Local Labor Market Impacts of Import Competition in the United States, American Economic Review, 103, 2121-2168.
Bartik, T. J. (1991): Who Benefits from State and Local Economic Development Policies? W.E. Upjohn Institute for Employment Research.
Borusyak, K., X. Jaravel, and P. Hull (2018): Quasi-Experimental Shift-Share Designs. NBER Working Paper 24977 (replication archive: https://github.com/borusyak/shift-share).
Comment