Current Projects

The Workforce Science Project is currently sponsoring the following projects:

Criminal Background and Job Performance

(With funding from the Joyce Foundation)

Job applicants with criminal records are much less likely than others to obtain legitimate employment. This research presents the first evidence using civilian data on how ex-offenders behave on the job if hired. Our data indicate that individuals with criminal records have a much longer tenure and are less likely to quit their jobs voluntarily than other workers. Some results, however, differ by job. Customer service employees with a criminal record are no more likely than others to be discharged involuntarily and are probably no more likely to leave for reasons of misconduct. Sales people with records, however, may have a higher rate of involuntary discharge and do have a higher rate of misconduct discharge. This complex pattern suggests the need for expanding the public policy menu beyond Ban the Box to include incentives to encourage employers to look more closely at their workforces to identify where the true risk groups are.

The Value of Psychological Tests in Predicting Work Outcomes

Through a partnership with the University of Texas Medical Branch, WSP researchers will have the unprecedented opportunity to develop a psychometric test battery for use in pre-employment hiring. These tests will measure personality, cultural fit, and ability, and then evaluate which elements of the instrument are most predictive of a wide range of performance measures.

Toxic Employees

(In partnership with Evolv)

Everyone has worked with a “toxic employee” whose very presence in the workplace detracts from everyone else’s performance and may even pose a legal risk to their employer. We study termination reason codes (e.g., sexual harassment, drug/alcohol policy violations) to determine whether opaque questions at the point of application can offer up an accurate signal about an applicant's honesty and integrity.

Employer-Subsidized Credentials: Improving the New Training Paradigm

(With Baylor Scott and White Health and Burning Glass Technologies)

Some employers have begun to address the skills gap with a new model of employee training in which employers collaborate with educational institutions to provide portable educational credentials. This study will examine whether employee participation can be increased by improving the information available to employees about available training and its likely return in their future career.

What Privacy Rights Do Employees Value?

As with any use of Big Data, talent analytics raises the challenge of protecting individual privacy rights. With diminishing practical constraints on data collection and analysis, privacy advocates have increasingly turned to legislative and regulatory restrictions. This push for governmental intervention is based on guesses and assumptions rather than an empirical understanding of what data employers use, how employers use that data, or even of what monitoring most concerns employees. The current project attempts to shed light on these issues through a series of surveys of employee privacy attitudes. Using a list of current and proposed monitoring practices gleaned from interviews with talent analytics professionals, these surveys seek to determine what types of monitoring and data usage raise the greatest degree of privacy concerns.

The Legal Regulation of the Employment Relationship

A small but significant economic literature examines the effect of employment laws on a variety of outcomes such as employment levels. Variations in state law and changes in the law over time are used to untangle the causal effect of laws themselves from other factors that might affect the measures of interest. Although this literature represents an important beginning, it has several limitations. First, the literature focuses overwhelmingly on a few slices of the legal regime governing the employment relationship, ignoring much of a large and complex system. Second, almost all studies code laws in a very simple Yes/No format that ignores the complex differences among state laws. Third, almost all of the literature examines one employment law in isolation (e.g. maximum hours) without considering the effect of other aspects of the employment law system. We are addressing these issues by coding a large set of employment laws into a more nuanced form that can be used for statistical analysis. We will then use this novel data in a series of studies that exploit variations in state laws to investigate the extent to which the employment laws effect economic activity.

Recessions and Productivity

(In partnership with Evolv)

In the last three U.S. recessions, productivity (output per hour) ceased to move with output but instead actually started to increase before the recessions officially ended.  Utilizing millions of data points on worker productivity from a variety of different industries, we test two explanations for this change: (1) Firms selectively fired poor performers or hired better performers; or (2) Employees worked harder and became more productive in order to retain their jobs.


Completed Projects

Data Analysis of a Job Board

We consider the problem of modeling time-dependent textual data taking given endogenous and exogenous processes into consideration. Such text documents arise in real world applications, including job advertisements and financial news articles, which are influenced by the fluctuations of the general economy. We propose a hierarchical Bayesian topic model which imposes a dynamic hierarchical structure on the evolution of topics incorporating the effects of exogenous processes, and show that this model can be estimated from Markov chain Monte Carlo sampling methods. We further demonstrate that this model captures the intrinsic relationships between the topic prevalence and the time-dependent factors, and compare its performance with latent Dirichlet allocation (LDA) and the structural topic model (STM). The model is applied to two collections of documents to illustrate its empirical performance: online job advertisements from DirectEmployers Association and journalists’ postings on BusinessInsider.com.