peter-mccrory/high-freq-sah-ui-restat

Data Archive for "Unemployment Effects of Stay-at-Home Orders: Evidence from High Frequency Claims Data" by ChaeWon Baek, Peter B. McCrory, Todd Messer, and Preston Mui

Data Archive for "Unemployment Effects of Stay-at-Home Orders: Evidence from High Frequency Claims Data"

By ChaeWon Baek, Peter B. McCrory, Todd Messer, and Preston Mui

Intro

This repository contains all the data and code required to replicate the analysis in "Unemployment Effects of Stay-at-Home Orders: Evidence from High Frequency Claims Data".

The repository is organized as follows:

• ./data/: Contains the raw data as well as the compiled Stata datasets used in the analysis. A description of each raw data file follows below.
• ./src/: All the code necessary to do the analysis is contained in this folder. There are three sub-directories for each of the languages used. A description of each file is contained below. The empirical analysis is done in Stata, the theoretical analysis is done in Matlab, and some of the figures are created in a Jupyter notebook using a Python 3 kernel.
• ./output/: This folder contains the plots and tables appearing in the paper. Running the code (in the order described below) will reproduce all analysis from the paper (main-text and appendix).

The root directory also contains a .gitignore file for ignoring files for the Github repository (this repository is hosted at: https://github.com/peter-mccrory/high-freq-sah-ui-restat).

All scripts were verified to work with the following software: (i) Stata 14, (ii) Python 3.8.3 installed with Anaconda, and (iii) Matlab 2020b.

Steps for Replicating Analysis in Paper

To replicate the analysis in Baek, McCrory, Messer, Mui (2020) do the following:

• Main Empirics: From within the ./src/stata/ directory, run run_all.do in Stata. This will call on the following do files contained in the ./src/stata/ directory, briefly described below. Each step can be run separately from run_all.do.
  • step1_build_state: This file imports all underlying data files in the data directory needed for the state-level analysis. Intermediate files are saved in ./data/stata/
  • step2_build_county: This file imports all underlying data files in the data directory needed for the county-level analysis. Intermediate files are saved in ./data/stata/
  • step3_merge_state: This file merges together the state-level datasets to produce the main state-build dataset, ./data/stata/state_build.dta
  • step4_merge_county: This file merges together the county-level datasets to produce the main state-build dataset, ./data/stata/state_build.dta
  • step5_analysis_main: This file estimates state-level regressions and saves all regression tables appearing in the main text and online appendix to
  • step6_analysis_county: This file estimates the county-level regressions underlying Tables 3 and 4 in the main text, saving those tables to ./output/tables/.
  • step7_eventstudies: This file estimates event study regressions described in the online appendix, saving the figures in the online appendix to ./output/plots/.
  • step8_early_late_figure: This file creates Figure 3 in the main text, saving it to ./output/plots/
  • step9_rel-implied-agg: This file calculates the relative-implied aggregate employment loss attributable to SAH orders.
• Theoretical Results: From the command line Matlab and while located in the ./src/matlab/ directory, using Dynare, run baseline.mod with the command dynare baseline. This will create and save Figure 5 to ./output/plots/. It will also calculate and report numbers appearing in Table 2 of the main text.
• Additional Figures: To make all remaining figures in the paper (main text and online appendix), run ./src/python/QCEW.ipynb in a Jupyter notebook (Python 3 kernel). This will create and save figures to ./output/plots/. Note: the user will need to install plotly-related packages to their conda distribution; if these are not already installed, an error will be produced prompting them to install the appropriate package (e.g. cufflinks, plotly-orca).

Brief Description of ./data/ Directory Files

---

The data used in the analysis is organized by source. Each source has its own separate folder, each of which is described in brief below:

• bls: This folder contains five data files used to build the state and county-level datasets used in the paper
  • county_qcew_emp.csv: This file reports 2018 county-level employment from the Quarterly Census of Employment and Wages (QCEW). ./src/python/QCEW.ipynb queries the QCEW data and creates this file. These data were downloaded April 10, 2020.
  • county_qcew_subset.dta: This file is created from the 2019 county-level employment from the QCEW. It is created with ./data/bls/build_county_qcew_subset.do. The raw file from which this subset was created is available at https://data.bls.gov/cew/data/files/2019/csv/2019_annual_singlefile.zip. These data were downloaded June 27, 2020
  • lau.csv: This file contains county-level employment and unemployment statistics produced from the BLS Local Area Unemployment Statistics. These data are downloaded and saved with Load-LAU.ipynb. These data were downloaded with this script June 4, 2020.
  • state_bartiks.csv: This file is created in the Jupyter notebook BLS-jobs.ipynb. This represents the Bartik-style control described in the main text. The underlying data were downloaded April 10, 2020.
  • state_emp_ind.csv: This file reports average annual employment from 2018 levels by sector and state. It is created with QCEW.ipynb. These data were downloaded by this script on April 10, 2020.
• cdc: This folder contains the excess deaths associated with Covid-19 as downloaded from the CDC
• census: This folder contains data from U.S. census Bureau, including state demographic data (https://www.census.gov/data/tables/time-series/demo/popest/2010s-state-detail.html; accessed June 16, 2020) and county population estimates for 2020 (https://www2.census.gov/programs-surveys/popest/datasets/2010-2019/counties/totals/; accessed March 28, 2020)
• CZ: This folder contains the USDA commuting zone to county crosswalk created by the USDA (https://www.ers.usda.gov/data-products/commuting-zones-and-labor-market-areas/; accessed July 1, 2020)
• dol: This folder contains state UI replacement rates calculated by the Department of Labor (https://oui.doleta.gov/unemploy/ui_replacement_rates.asp; accessed June 16, 2020)
• google: This folder contains the Google Mobility Report (https://www.google.com/covid19/mobility/; accessed May 21, 2020)
• kong-prinz: This folder contains the business closure dates as reported in Kong and Prinz (2020), hand-coded September 9, 2020)
• nytimes: This folder contains reported county-level SAH order dates as reported by the New York Times. (https://www.nytimes.com/interactive/2020/us/coronavirus-stay-at-home-order.html; accessed various dates using Wayback Machine). This folder also contains presidential vote shares by state (https://www.nytimes.com/elections/2016/results/president; accessed and hand-coded June 17, 2020)
• opportunitylab: This folder contains data from the Opportunity Insights Economic Tracker (https://github.com/OpportunityInsights/EconomicTracker; downloaded September 4, 2020)
• stata: This folder contains intermediate files created when running run_all.do
• usaFacts: This folder contains confirmed daily, county-level COVID-19 cases (https://usafacts.org/visualizations/coronavirus-covid-19-spread-map/; accessed and downloaded June 5, 2020)
• WAH_DingelNeiman: This folder contains the Dingel and Neiman (2020) work at home index, hand-coded April 17, 2020