Predicting Layoff among Fragile Families

Area of Science:

• Social Sciences
• Economics
• Psychology

Background:

• Job loss represents a significant disruption to social and economic roles.
• Layoffs have documented long-term negative economic and psychological consequences for individuals and families.
• Accurate prediction of job loss is crucial for understanding its causal effects using observational data.

Purpose of the Study:

• To compare the predictive performance of an empirical social science model against data science models for predicting job layoff.
• To assess the impact of layoff prediction accuracy on causal inference in social science research.

Main Methods:

• Utilized the Fragile Families data for analyses.
• Developed a model grounded in existing empirical social science research on job layoff.
• Compared the performance of this model against top-performing data science models within the Fragile Families Challenge framework.

Main Results:

• The social science-based model's predictive performance was comparable to the best data science models.
• Layoff was identified as a relatively exogenous shock, influencing predictive model performance.
• Minor improvements in layoff prediction models were observed across different approaches.

Conclusions:

• Empirical social science models can be effective in predicting social processes like job layoff.
• The exogenous nature of layoff may limit the gains from highly complex predictive models.
• Future research should investigate whether enhanced prediction models improve causal inference validity.