Homework 3

Introduction

Here is a guide to the experimental data of Bertrand & Mullanaithan (2001)

Each row of this dataset contains the variables related to one job application. Here is a summary of the variables you will need:

  • call: a binary variable equal to 1 if the application receives a callback
  • race: a string variable equal to “w” if white sounding name, “b” if black sounding
  • education: an integer variable between 0 and 4 that indicates level of education reported on the applicant’s resume
  • yearsexp: a variable with the applicant’s years of experience
  • h: a dummy variable indicating that the resume was of a high quality

Question 1

Use a regression framework to replicate the paper’s evidence that (1) white sounding names receive more callbacks than black sounding names; and (2) white sounding names receive significantly more (at a significance of 10%) callbacks when the resume quality is “high”, while black sounding names do not.

Question 2:

Re-run the same analysis to document racial bias in callbacks (finding 1) using a regression framework, this time controlling for education and years of experience in the regression.

Question 3:

Use a regression framework to test the null hypothesis that racial differences in callbacks are the same regardless of the applicant’s years of experience.

Question 4

Consider the model:

\[ \mathbb{E}[white|eduction] = \beta_{0} + \beta_{1}education \]

where white is a dummy variable equal to 1 if the applicant has a white sounding name.

What does the random assignment of race imply about \(\beta_{1}\) in this linear model? Run a regression to test this implication.