Behaviorally designed training leads to more diverse hiring
Data files
Jan 23, 2025 version files 59.50 KB
-
Arslan_et_al._Code.zip
54.20 KB
-
README.md
5.30 KB
Abstract
Many organizations are interested in increasing the diversity of their workforce and spend millions of dollars on diversity training. Yet there is little empirical evidence that such training increases diversity in organizations. We implemented a large-scale field experiment in a global telecommunications and engineering firm (n = 10,433) testing whether behaviorally designed training increases the diversity of who is hired. In particular, the diversity training was timely (delivered immediately before hiring managers shortlisted candidates), tailored to the hiring decision, delivered by senior members of the organization, and made diversity salient. Results show that behaviorally designed diversity training can positively influence the hiring of women and non-national applicants relative to business as usual. Our findings suggest that behaviorally designed diversity training can work to change the diversity of hires but that its success relies on carefully considered design choices and the decision context.
README: Behaviorally designed training leads to more diverse hiring
Overview
The program (Stata do-files) for “Behaviorally designed training leads to diversity hiring” by Arslan, Chang, Chilazi, Bohnet, and Hauser, published in Science (2025).
The program files run all the code to import raw data files (xlsx, csv), clean and generate the data (in dta format), prepare the data for the analysis, run the regression analyses, export output, and thus generate tables presented in the paper. The replicator should expect the code to run for up to 15 minutes.
Data Availability and Sharing
The organizational data used in this manuscript is of a proprietary nature. We, the authors of the manuscript, have legitimate access to and permission to use the data, but we are unable to make the data publicly available due to a strict data use agreement with our field partner (global telecommunications and engineering company).
Interested researchers are encouraged to contact MoreThanNow to have a confidential conversation about data sharing. To request access to the data used, each institution has to submit a data access request form and needs to arrange signed Data Sharing/Use and Non-Disclosure agreements for all researchers to be covered under the agreements. Researchers from the same institution can be named on the same data request form. Once completed, please email the form and all attachments to experiment@morethannow.co.uk.
Upon submitting this request, MoreThanNow will share the request with the field partner. The field partner will review the request and decide whether they are willing to consider sharing their data with the named researchers. If they approve the request, the field partner will send Non-Disclosure and Data Sharing/Use Agreements with the legal contact listed above, which must be executed prior to data being shared. The ultimate decision authority on whether data requests are approved lies with the field partner.
Data Files
Main data files (administrative hiring and intervention tracking) are not available here and were obtained through institutional collaboration: Dataset #1 Treatments Tracking - Final.xlsx, Dataset #2 - Video - Final.xlsx, DatasetHiringMarch2023.xlsx, Dates_reqid_mid.xlsx (this is a template to be filled based on main data files).
Auxiliary data files are shared publicly at Dryad: GenderEquality.xlsx (based on the Global Gender Gap Index 2023) and global_south_nationalities.csv (based on the Organization for Women in Science for the Developing World’s list) for heterogeneity analyses.
GenderEquality.xlsx: This file includes the variable 'req_country' which refers to the country of job requisition raised during our study period, the variable 'gender_equality' which is the gender equality score (varying between 0 and 1) of the countries in the world (as obtained from the Global Gender Gap Index 2023), the variable 'req_georegion' which refers to the geographic region of each country, the variable 'global_north' which takes the value 1 if the country of job requisition is a Global North country (e.g. United States, France, Australia) and 0 if the requisition country in the Global South (e.g. India, Colombia, Kenya).
global_south_nationalities.csv: The variable 'can_nationality' refers to the country of nationality of the job applicant and the variable 'global_south' shows if the country of nationality of the candidate is a Global South country or not. It takes the value 1 if the candidate is a national of a country in the Global South (e.g. India, Colombia, Kenya) and 0 if the candidate is a national of a country in the Global North (e.g. United States, France, Australia).
Software Requirements
-Stata, code was last run with version 18.
-The following packages will be installed in Stata: dropmiss, outreg2, ivreg2, ranktest
-The approximate time needed to reproduce the analyses on a standard desktop machine is under 15 minutes (2024).
Description of programs/code
-DivExp_MasterDoFile.do will run all four do-files below which are necessary for data preparation and analysis.
-DivExp_InitialCleaning.do will import raw files, clean, manipulate, and save the data in dta format.
-DivExp_MergeFiles_Conditional.do will merge different data files (excluding some observations) and prepare the data for the analyses.
-DivExp_MergeFiles.do will merge different data files and prepare the data for the analyses. This is the file generating the dataset for our experimental analyses.
-DivExp_Analysis.do will run the regressions and generate the output (tables in Supplementary Materials).
Instructions
-Place all data files in the ‘data’ folder. There are some auxiliary data files we share, but the main data files will have to be obtained from the field partner.
-Place code (all do-files) in the ‘do-files’ folder.
-Change global ‘path’s in DivExp_MasterDoFile.do
-Run program DivExp_MasterDoFile.do (which will run all the other do-files).
-The provided code reproduces the regression results presented in the article.
-The folder 'output' was left empty on purpose. This is where the regression results (article tables) will be exported when the program code (DivExp_Analysis.do) is run.