lionup / mongey_pilossoph_weinberg Goto Github PK

This repository contains the code which replicates "How Many Jobs Can be Done at Home?" by Simon Mongey, Laura Pilossoph, and Alex Weinberg. This replication package downloads the data available online and reproduces the results available in our paper.

The following data on occupation work-from-home ability and physical-proximity may be useful for your research:

• 2 digit Census OCC
• 3 digit Census OCC

We use publicly available data from:

• O*NET

• OES

• ATUS

• CPS

• PSID

• O*NET, OES, and ATUS files

  • Downloaded in the process of running the replication.

• CPS files:

  • 2019 ASEC is downloaded from IPUMS and can be found here clean_CPS2019/input/asec19_raw.dta
  • The basic monthly CPS is downloaded from IPUMS
    • Sample is February and March from 2010 - 2020
    • We keep a subset of variables (see end of this note)
    • The replication code then loads in the files unemployment_figs/input/basic_monthly_CPS_raw_YYYY.dta

• PSID files:

  • We downloaded FAM2017ER from PSID
  • We keep a subset of variables (see end of this note)
  • The replication code then loads in the file clean_PSID/input/htm_psid2017.dta

• The Safegraph data we use in the paper is not publicly available.

1. Download this repository by clicking the green Clone or download button above. Unzip into the directory of your choice.
2. From the terminal, navigate to the folder containing this README.
3. Type make and hit enter to excute the code from top to bottom. The Makefile will execute all the stata and shell scripts.

Note:

• After running make for the first time, any subsequent changes will not run the whole code again from top to bottom, but only the parts of the codes that have changes and their dependencies.
• For example, if after running make for the first time you wish to change sample selection in the CPS and PSID from individuals aged 25-65 to individuals aged 20-65, you could change that line of code in /unemployment_figures/unemp_figs.do, run make again and only the code from this change onwards would be ran.

The best way to replicate the paper is to run make. It is also possible to replicate the paper by running the scripts found in the folders in the following order:

1. download_data/src
2. clean_ONET/src
3. clean_OES/src
4. clean_PSID/src
5. make_LWFH_PP_measures/src
6. clean_CPS2019/src
7. main_figs/src
8. unemployment_figs/src
9. atus_figs/src

The workflow is organized as a series of tasks. Each task folder contains three folders: input, src, output. A task's output is used as an input by one or more downstream tasks.

• Jonathan Dingel's note on the automated task-based organization of this code.
• The replication package for Dingel and Neiman 2020 has helpful notes on running make and stata.

Sample is February & March from 2010 - 2020.

Variables

• age
• occ
• educ
• empstat
• compwt
• month
• year
• marst
• earnweek
• uhrswork1
• bpl
• citizen
• race
• sex
• labforce

From the PSID’s Main Study Family File 2017.

• ER66002
• ER66009
• ER66017
• ER71331
• ER71333
• ER71335
• ER71347
• ER71349
• ER71351
• ER71353
• ER71305
• ER71377
• ER71363
• ER71361
• ER71365
• ER71277
• ER71427
• ER71429
• ER71433
• ER71435
• ER71437
• ER71439
• ER71443
• ER71445
• ER71447
• ER71449
• ER71451
• ER71453
• ER71455
• ER71457
• ER71459
• ER71461
• ER71463
• ER71465
• ER71467
• ER71469
• ER71471
• ER71473
• ER71475
• ER71481
• ER71483
• ER71485
• ER71570
• ER71512
• ER66195
• ER66196
• ER66197