This Stata package implements confidence intervals in shift-share designs (also called Bartik 1991 designs) using the AKM and AKM0 procedures from Adão, Kolesár, and Morales (2019). See the ShiftShareSEMatlab package for Matlab version of this code, and the ShiftShareSE package for an R version.

IV regression using data from Autor, Dorn, and Hanson (2013), including the first-stage results

    use "data/ADH_derived.dta", clear
    local control_varlist t2 l_shind_manuf_cbp reg_encen reg_escen reg_midatl reg_mount reg_pacif reg_satl reg_wncen reg_wscen l_sh_popedu_c l_sh_popfborn l_sh_empl_f l_sh_routine33 l_task_outsource
    ivreg_ss d_sh_empl_mfg, endogenous_var(d_tradeusch_pw) shiftshare_iv(d_tradeotch_pw_lag) control_varlist(`control_varlist') share_varlist(emp_share1-emp_share770) weight_var(weight) alpha(0.05) akmtype(1) firststage(1)

Corresponding reduced-form regression

    reg_ss d_sh_empl, shiftshare_var(d_tradeotch_pw_lag) control_varlist(`control_varlist') share_varlist(emp_share1-emp_share770) weight_var(weight) alpha(0.05) akmtype(1)

The datasets needed to run these commands are stored in the data directory: ADH_derived.dta and sector_derived.dta. The do file ADHapplication.do contains additional examples that use these datasets to illustrate how to use the reg_ss and ivreg_ss to estimate linear regressions in which the regressor of interest has a shift-share structure, and instrumental variables regressions in which the instrument has a shift-share structure.

The released version of reg_ss and ivreg_ss is available at REPEC (see reg_ss and ivreg_ss). You can install these packages from within Stata using the ssc command:

ssc install reg_ss
ssc install ivreg_ss

Alternatively, to download the development version of these packages from GitHub, download the files reg_ss.ado, reg_ss.sthlp, ivreg_ss.ado, ivreg_ss.sthlp from the code directory, and put them into Stata's personal ado directory, typically

• c:\ado\personal on Windows
• ~/Documents/Stata/ado on a Mac
• ~/ado on Linux
• The simple way to find out your local Stata directory is to run the command sysdir in Stata.

For more information on how to use personal ado files, please refer to Stata Official FAQ.

Please open issues and leave feedback, use click on the Issues tab.

• Please make sure you have no missing values in your dataset. We don't handle missing values in our code, and having missing values in the dataset causes the program to shut down.
• Since Stata doesn't support nested preserve and restore, our ivreg_ss code cannot return your original dataset after computing AKM standard errors. reg_ss will return your original dataset if you don't compute clustered standard errors. We suggest reloading your dataset every time after you compute AKM standard errors.

• no observations: You may have missing values in your share variable list / shiftshare variable