Using LinkedIn to Study the Value of Graduate Degrees

I am engaged in a series of studies of the labor market value of graduate education. The goal is to estimate causal effects of specific graduate degrees, such as an MBA or an MS in Engineering, on labor market outcomes. Moreover, I study how characteristics of colleges, graduate programs, and students influence the payoff to graduate education.

In a new project I will use a different data source—LinkedIn---to measure returns. The Tobin RA will help with two aspects of the project. The first is to study the degree to which individuals work in a different state from the place where they obtained their graduate degree. This part of the project is designed to complement research that I am doing using administrative data from Texas. That data provides rich information about education, family background, and ability indicators. However, the earnings data is limited to those who work in Texas. The fact that earnings data is restricted to Texas raises concerns. For degrees such as a JD, it is possible that the most able students from Texas law schools work in major law centers such as New York, Chicago, and Washington DC. I wish to assess the extent to which this leads to bias in estimating the returns to a JD, and similar issues may arise for other degrees.

The second and main objective is to use LinkedIn to provide a new set of estimates of the value of graduate degrees conditional on institution an undergraduate major. I will use an estimation strategy that I have implemented in prior papers. that involves before and after comparisons of earnings and occupation.

The specific tasks for the Tobin RA are as follows:

1. Learn how to access and use LinkedIn data. Yale researchers can access a version of LinkedIn data assembled by Revilio Labs. It can be accessed at Yale through a server maintained at the University of Pennsylvania. You will need to study the documentation for the data set, learn the definitions of the variables it contains and how they were constructed, and select cases and variables.
2. Provide standardized measures of institution and degree type and state of residence of the institution using the entries in the LinkedIn data. Depending on the quality of the variable standardizations that had been performed by Revilio Labs and LinkedIn, you may need to use improve classifications using AI.
3. For selected graduate degree fields and Institutions, measure the fraction of individuals who obtained graduate degrees from Texas schools but are working in other states 2, 5, 10, 15, and 20 years after obtaining their degrees.
4. Estimate regression models relating earnings to undergraduate and graduate degrees.
5. Help with creation of tables and figures and literature review in connection with the larger project on graduate education.