Causal and computational methods for the study of labor market frictions
Abstract/Contents
- Abstract
- This thesis studies frictions in the labor market. We take a primarily computational lens to the problems described. The first chapter is co-authored with Guillaume Saint-Jacques, Sinan Aral, Iavor Bojinov and Erik Brynjolfsson and is centered on the role of social networks in labor markets. Although the Strength of Weak Ties is one of the most influential social-scientific theories of the past century, two empirical challenges have prevented a definitive causal test of this theory. First, a lack of large-scale data linking social networks to job transmission has prevented empirical verification of their association. Second, network ties and labor market outcomes are endogenous, making causal evidence of a link between weak ties and jobs elusive. We address these challenges using data from multiple randomized experiments conducted on LinkedIn's People You May Know (PYMK) algorithm, which recommends new connections to LinkedIn users. The PYMK experiments we analyzed randomly varied the prevalence of weak ties in the networks of over 17 million people, creating approximately 2 billion new ties and 600,000 new jobs. Using the exogenous variation in weak ties created by these experiments, we tested the extent to which weak ties increase job applications and job transmissions. Contrary to recent correlational evidence, which suggests job transmissions correlate with strong ties, we found weak ties increase job transmission, but only to a point, after which there are diminishing marginal returns to tie weakness. These results help resolve the apparent ``paradox of weak ties" and provide causal evidence for the weak tie theory. They also suggest the need to revise the theory by incorporating the non-linear effects of weak ties on job transmission. Co-authored with Susan Athey, Lilia Chang and Lisa Simon, the second chapter examines labor market job changes in the broader US economy. Workers will need to transition occupations at an increasing rate in the future, as automation and other trends change which occupations are in demand and which skills are required. This work aims to accurately predict fine-grained labor market state transitions of individual workers, given their observable characteristics and their labor market histories. We estimate a discrete choice model of occupational transitions with thousands of choices, using recent new methods developed in Ruiz et al. (2020) and Athey et al. (2018) that employ approaches from machine learning for estimating models with a large number of latent variables and use matrix factorization to reduce the dimensionality in the rich state space. While the model is descriptive, we can compare workers who have the same occupation, education and individual characteristics, but differ in terms of urban or rural location. Holding individual characteristics and the start state fixed, we find that urban workers in declining occupations earn 6.3 percent more a year than rural workers. We are then able to perform a decomposition that find that a large majority of this difference can be accounted for by increased opportunities in urban areas, as opposed to a wage premium in the same state in urban areas per se. In the last chapter, co-authored with Evan Rosenman, Romain Gauriot, and Robert Slonim, we study volunteer labor markets. First, we make a methodological contribution to the regression discontinuity designs literature. A crucial assumption this design rely on is that the forcing variable on which treatment is assigned cannot be manipulated, one that is frequently violated in practice thus jeopardizing point identification. In this paper, we introduce a novel method that provides partial identification bounds on the causal parameter of interest. The method first estimates the number of manipulators in the sample using a log-concavity assumption on the un-manipulated density of the forcing variable. We then derive best- and worst-case bounds when we delete that number of points from the data, along with fast computational methods to obtain them. We use the new method on a large dataset of blood donations from the Abu Dhabi blood bank to obtain the causal effect of deferral on future volunteering behavior. We find that, despite significant manipulation in the data, it has power to detect causal effects where traditional methods, such as donut-hole RDDs, fail.
Description
| Type of resource | text |
|---|---|
| Form | electronic resource; remote; computer; online resource |
| Extent | 1 online resource. |
| Place | California |
| Place | [Stanford, California] |
| Publisher | [Stanford University] |
| Copyright date | 2021; ©2021 |
| Publication date | 2021; 2021 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Author | Rajkumar, Karthik |
|---|---|
| Degree supervisor | Athey, Susan |
| Degree supervisor | Imbens, Guido |
| Thesis advisor | Athey, Susan |
| Thesis advisor | Imbens, Guido |
| Thesis advisor | Wager, Stefan |
| Degree committee member | Wager, Stefan |
| Associated with | Stanford University, Department of Economics |
Subjects
| Genre | Theses |
|---|---|
| Genre | Text |
Bibliographic information
| Statement of responsibility | Karthik Rajkumar. |
|---|---|
| Note | Submitted to the Department of Economics. |
| Thesis | Thesis Ph.D. Stanford University 2021. |
| Location | https://purl.stanford.edu/jg633hg0671 |
Access conditions
- Copyright
- © 2021 by Karthik Rajkumar
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...