Date of Award
Doctor of Philosophy (PhD)
Altonji , Joseph
The initial job placement of a college graduate plays a crucial role in his future job mobility and career growth. Poor initial placements can lead to lasting impediments by placing college graduates in jobs with limited room for training or promotion. Many college graduates obtain their first jobs through career offices which act as liaisons between students and employers. Job placement mechanisms proposed by career offices often dictate how a college graduate obtains his first job. Therefore, a detailed study of such job placement processes offers a unique window into the mechanisms which drive the sorting patterns in job search and, as a consequence, have implications for a richer understanding of the initial determinants of career trajectories and earnings growth. My dissertation studies the job placement processes of a leading post-secondary educational institution in India and a top U.S. business school. Job placement processes typically include job applications, pre-interview screening tests, job interviews, job offers and job choices. Using detailed information on each stage of the job placement process, I explore the roles played by workers and firms in determining labor market outcomes and assess policy interventions to remedy disparities. By studying structured job placement processes, I also quantify the salience of socio-emotional skills, such as compassion, gregariousness, assertiveness etc., in explaining academic behavior, sorting patterns in the various steps of job search, and eventual placement outcomes. In the first chapter, I study the job placement process of an elite technical college in India, half of which is comprised of students from disadvantaged groups or disadvantaged castes. Disadvantaged castes earn 15\% less than comparable advantaged castes in India's urban labor markets, with the largest disparities concentrated in the private sector. Such disparities remain pronounced despite widespread and effective affirmative action policies in college admissions. Moreover, there is no empirical evidence evaluating compensatory hiring practices for disadvantaged castes among private sector jobs in India. Using novel administrative data on all steps of job search, I evaluate policies to promote hiring diversity. The administrative data includes rich student-level information on all stages of job search, including job applications, pre-interview screening tests, job interviews, job offers, and job choices. The data also includes detailed information on pre-college test scores, previous labor market experience, within-college performance, job characteristics and demographics. I offer the first quantitative decomposition of the earnings drop off (across castes) at each stage of job search. I show that the compositions of job applications and job choices do not explain the earnings gap. Pre-interview screening tests including written aptitude tests (first round) and group discussion based ``soft skills" tests (second round) explain only a small fraction of the drop off in earnings. Therefore, almost all of the earnings drop off occurs between one-on-one interviews (third round) and job offers. These findings suggest that policies which provide information about jobs, modify preferences, or improve performance at university are unlikely to close the earnings gap. Guided by the sequential decomposition of the earnings gap, I build a model of the job placement process. The model, which incorporates both the supply of and demand for jobs, is estimated using the method of simulated likelihood. My estimates show that caste disparities in hiring are driven not by differential caste-preferences over job characteristics but by hiring decisions of firms. Additionally, modelled unobservables play an economically small role in jointly determining observed choices. On average, firms need to be compensated 5\% of average salary to be as likely to make job offers to disadvantaged castes as to observably similar advantaged castes. Finally, I evaluate policies to promote hiring diversity. First, I consider a subsidy in which firms are compensated by the amount that makes them indifferent between hiring an observably identical advantaged or disadvantaged caste. Second, I consider a “pre-college intervention” policy which equalizes the distribution of pre-college test scores across castes. To compare cost-effectiveness, I use the model estimates and calculate the change in test scores required to induce the same employment gains for the disadvantaged caste as the gains from the direct subsidy. The change in test scores is large because the model estimates imply that test scores play only a small role in hiring. Even under extremely conservative assumptions, a back of the envelope calculation based on cost estimates of improving student test scores in India shows that direct subsidies can be twice as cost-effective as the “pre-college intervention” policy. In the second chapter, which is joint work with John Eric Humphries, we study the job placement process of an elite MBA program in the U.S. and quantify the role of socio-emotional skills in determining academic outcomes and eventual job placements. A large literature has documented the crucial role played by socio-emotional skills in determining educational attainments and labor market outcomes. However, little is known regarding the mechanisms through which non-cognitive skills affect such outcomes. Using detailed measures of socio-emotional traits taken before students matriculate, this chapter evaluates how such traits affects three types of outcomes: academic performance and course choices, job search behavior, and industry and compensation of the accepted job offer. These personality measures, such as compassion, gregariousness, assertiveness etc., are assessed through forced-choice methods which provide robust safeguards against ``gaming”. Initial descriptive results show that socio-emotional skills are predictive of behavior in each of these three areas, commonly out-predicting the effects of standardized test scores. Moreover, the relative importance of different skills varies across the three areas of outcomes considered. Based on the descriptive evidence, we estimate a sequential model of the key stages of the MBA program, including academic performance, course specialization, job search, and job placements. Using the model, we then decompose the total effects of socio-emotional skills into direct returns and indirect returns through differences in both academic performance and job search behavior. Preliminary descriptive results suggest that gregariousness and industriousness are important predictors of earnings upon graduation. However, they work through different channels: industriousness affects academic performance while gregariousness affects job search behavior. My dissertation also opens up many avenues for further research. For example, one could ask what the optimal job placement mechanism would be. Theoretical first-best mechanisms may not be well-suited for distributional welfare (or, equity). However, ad-hoc job placement processes might sacrifice substantial aggregate welfare (or, efficiency) for modest improvements in distributional welfare. An ideal job placement process would balance both distributional and aggregate welfare of participants. Another study could involve quantifying the inefficiencies (if, any) due to early matching or unraveling. In one of the job placement processes I study, students cannot participate in job interviews scheduled on the next interview day conditional on receiving job offers from firms interviewing them on the current interview day. However, it is not obvious if such a job placement process is sub-optimal, especially if the job placement process has distributional goals. While theoretical works positing the inefficiencies of unraveling in labor markets have been large, they have been so far accompanied by a very slim body of empirical evidence. In future work, I plan to pursue these and other related avenues of research.
Shukla, Soumitra, "Rookie Market: Unpacking the Black Box of Firm-Worker Matching" (2021). Yale Graduate School of Arts and Sciences Dissertations. 115.