Predicting Interfirm Human Capital Flow: Leveraging Networks and Heterogeneous Effects
49 Pages Posted: 16 Apr 2018
Date Written: March 30, 2018
Firms compete for human capital which is a key component of the knowledge economy. The ability to predict the flow of human capital between firms affords firms' managers and recruiters, investors, market analysts as well as policy makers actionable insights. The web presence of employees through their LinkedIn profiles establishes a rich data source for analyzing the interactions between firms via their employees’ job changes over time. In particular, we model and study such interactions as a network of employee migrations between firms with the goal of predicting the future interfirm human capital flow. For this purpose, we track the movement of 89,943 individuals over 3,467 public firms from the years 2000 to 2014. The granular nature of this employee-level data allows us to propose new metrics such as the similarity between firms in the space of the skills possessed by its employees. At the same time, we exploit the network representation of the labor market to propose metrics with global cues on human capital flows that are not available through local firm-level variables. In addition to providing accurate predictions, we are also interested in potential explanations that can inform theory and practice. For this reason, we experiment with the causal tree algorithms which were originally designed for estimating heterogeneous treatment effects. We provide a framework for using these new class of algorithms for effective predictive modeling while also using the resulting tree-structure to provide insights on the human capital flow between firms.
Keywords: predictive analytics, human capital, network analysis, machine learning
Suggested Citation: Suggested Citation