Skip to content

Latest commit

 

History

History
141 lines (140 loc) · 6.52 KB

README.md

File metadata and controls

141 lines (140 loc) · 6.52 KB

ECON 499: Labor Unions and Racial Wage Inequality

PDF: https://aadsouza.github.io/econ499/Labor_Unions_and_Racial_Wage_Inequality.pdf

Contents

  • /code
    • cleaning:
      • clean_nber_morg.do
        • cleans nber extracts of cps morg.
      • do_clean_nber_morg.do (in: morgXX.dta ; out: cleaned_nber_morg.dta)
        • do clean_nber_morg.do
      • pareto_topcoding.do (in: cleaned_nber_morg.dta)
        • gen top-code adjusted wage distribution using pareto parameters following fll2021.
    • pre-cleaning:
      • gen_cpi.ipynb
        • generates cpi using cpiaucsl from fred + other stuff from fred.
      • dind_nind_crosswalk.xlsx
        • crosswalk between cps industry and fll2021 industry categories.
      • state_cps_fips_abb_name.csv
        • adapt crosswalk from dm2020 for state, statefips, stateabb, state.
    • analysis (either in paper/appendix or referenced existence in replication package):
      • dfl1.do
        • heywood parent extended dfl decomposition.
      • linreg.do
        • state-ind unionization rate ols regressions.
      • linreg2.do
        • naive ols wage on union coverage.
      • med_wage_time.do
        • gen median lwage3 time trends by race and sex.
      • nindnocctab.do
        • percentage and union coverage in each ind table.
      • rifdid-plots.ipynb
        • rifdid plots w/o sltt (state linear time trends).
      • rifdid-sltt-bw-plots.ipynb
        • rifdid plots w/ sltt separately by race and sex.
      • rifdid-sltt-bw.do
        • run rifdid separately for B/w w/ state linear time trends
      • rifdid-sltt-plots-wo-ar2w.ipynb
        • rifdid plots w/ sltt w/o always rtw states.
      • rifdid-sltt-plots.ipynb
        • rifdid plots w/ sltt.
      • rifdid-sltt-wo-ar2w.do
        • run rifdid w/ sltt w/o always rtw states.
      • rifdid-sltt.do
        • run rifdid w/ sltt.
      • rifdid.do
        • run rifdid w/o sltt.
      • rifdiddiagnostics.do
        • run rifdid separately for B/w w/o sltt.
      • rifreg2.do
        • run naive rifols wage on union coverage.
      • rifreg2bw.do
        • run naive rifols wage on union coverage separately by race and sex.
      • rifreg2plots-bw.ipynb
        • rifols wage on union coverage plots separately by race and sex.
      • rifreg2plots.ipynb
        • rifols wage on union coverage plots.
      • stag_event.do
        • run event study that doesnt work.
      • stagdid-ptaplot.do
        • gen plots for predicted real log wage trends.
      • stagdid-wo-ar2w.do
        • run did - effect of rtw on B/w wage inequality - w/o always rtw states.
      • stagdid.do
        • run did - effect of rtw on B/w wage inequality.
      • stagsynthblack.do
        • synth - effect of rtw on wages for Black people.
      • stagsynthwhite.do
        • synth - effect of rtw on wages for white people.
      • stagsynthwhitewone.do
        • synth - effect of rtw on wages for white people. - dropped states for lack of common support.
      • sumstats.do
        • gen sumstats table.
      • ucov_race_sex_time.do
        • gen coverage rate time trends by race and sex >= 1983.
      • stagdid-fe-sltt.do
        • run did - effect of rtw on B/w wage inequality w/ year fe and state linear time trends.
      • stagdid-fe-sltt-wo-ar2w.do
        • run did - effect of rtw on B/w wage inequality w/ year fe and state linear time trends - w/o always rtw states.
      • med_centile_time.do
        • gen plot showing median person's position on white wage distribution.
    • miscellaneous stuff + graveyard (attempts and analyses that may not have made it into the paper)
      • restructure_mwage.do (in: mw_state_quarterly.dta ; out: qmwage7919.dta)
        • restructure Zipperer's quarterly state minimum wages.
      • dist_dataprep.do (in: cleaned_nber_morg.dta)
        • prepare data for distribution regressions following fll2021.
      • distreg1.do
        • first (only?) attempt at distribution regression following fll2021.
      • marg_crate.do
        • marginal effects of unionization rates consistent with fll2021.
      • bacondecomp.do
        • intended to run bacondecomp i.e. Goodman-Bacon decomp of stagdid (need to manipulate goodman-bacon ado to accomodate interaction).
      • cit_hisp_linreg.do
        • state-ind unionization rate ols for Hispanic people with citizen variable.
      • kde1.do
        • yearly kernel density estimates for white people, Black people, and Hispanic people.
      • morediagnostics.do
        • "more diagnostics"? for the RIF-DiD analysis of RTW laws (not sure what this ended up being).
      • rifdistributionplots-revised.ipynb
        • draft?
      • rifdistributionplots8919-bootstrap-w-errors.ipynb
        • i guess there are errors here?
      • riffig7.ipynb
        • more rifs.
      • rifreg1.do
        • rif on dfl attempt.
      • rifreg3.do
        • rifols state-ind unionization rate regressions.
      • stagarnr
        • compare alwaysr2w to neverr2w - so treat_st = treat_s (not DiD, post treat and post cont comparison)?
      • staghisp.do
        • run did - effect of rtw on Hispanic-white inequality.
      • dfl-3per.do
        • heywood parent extended dfl decomposition, only 8388 period.
  • /dfl1
    • dfl1.log
  • /distreg
    • note: distribution regression log files.
  • /figures
    • note: figures and plots for analysis and graveyard.
  • /linreg
    • note: log files for most analysis and graveyard.
  • /tabs
    • note: tables for analysis and graveyard.
  • lab_var_wage.txt
    • wage and related variable labels outline
  • 495preliminaryprospectus.bib
    • bibtex file for most/all literature cited in proposal.
  • 499data.bib
    • bibtex file for literature cited in thesis.
  • thesis-defense-slides.pdf
    • beamer slides used for honors thesis defense.
  • ECON_499_003__Honours_Thesis.pdf
    • final draft of thesis submitted to vse.
  • Labor_Unions_and_Racial_Inequality.pdf
    • final pdf.

NBER extracts of CPS MORG

Acknowledgements

I am extremely grateful to Professor Nicole Fortin and Professor Marit Rehavi for their continued guidance and support. I thank Professor Thomas Lemieux for extensive feedback and insights. I am indebted to Professor Nicole Fortin and Neil Lloyd for generously providing Stata codes that expedited the data cleaning process. I also thank Sheldon Birkett, Felipe Grosso, Elisabeth Hatting, Wenxin Ma, Evan Mauro, Javier Cortés Orihuela, and Sarah Kirker Wappel for helpful discussions and comments. All errors are my own.