Essays on the Labor Market for Internships

Abstract

In chapter one we first describe the demand for interns. By processing the text of all available ads on a popular internship website, we are able to characterize the occupation that best matches the advertised internship. We also match each internship ad to its local labor market. We find that internships in loose labor markets are more likely to be unpaid. Paid internships are in occupations that have higher wages, require less on the job training, and are more quantitative. We then conduct an audit study with more than 11,500 resumes, where we randomly assign student characteristics and apply to internships. We find that employers are more likely to respond to an application when they are looking for an unpaid intern. We find little effect of major field of study, volunteer experience, and college work experience on employer callbacks and some evidence that a higher GPA and previous internship experience increases positive employer responses. Black applicants receive fewer positive responses than white applicants, but this is entirely driven by greater discrimination against black-named applicants living far away from employers.

Chapter two continues on this work by describing the demand for skills for virtual internsships. Using the available database from all available ads used in chapter one, I focus on a small subset containing only virtual internships. Following Deming and Kahn (2016), I process the text of the advertisements using key words aligned with specific skill sets. From this data set I assess the skills associated with paid, unpaid, part-time, and full time internships. In addition, I link each internship ad to its local labor market to assess the impact of unemployment. I find an increase in the unemployment rate decreases the probability that an internship is paid. Finally, I assess the association between different skill sets and downstream wages. Using the Occupation Employment Statistics data set, I link the ad's to downstream occupations using a machine learning algorithm. I find that cognitive and financial skill sets are statistically significant and positively associated with the downstream wage. Most surprisingly, words associated with positive character traits are negatively associated with downstream wage.