The Value of Prediction in Identifying the Worst-Off

TL;DR

This paper evaluates how machine learning predictions can effectively identify the most vulnerable individuals in government programs, comparing their welfare impacts to other policy tools through models and a case study.

Contribution

It introduces analytical frameworks and data-driven tools to assess the effectiveness of prediction in prioritizing the worst-off, informing principled policymaking in welfare systems.

Findings

Prediction effectively identifies the worst-off in welfare programs.

Prediction can outperform expanding bureaucratic capacity in targeting vulnerable groups.

The paper provides practical tools for policymakers to evaluate prediction-based interventions.

Abstract

Machine learning is increasingly used in government programs to identify and support the most vulnerable individuals, prioritizing assistance for those at greatest risk over optimizing aggregate outcomes. This paper examines the welfare impacts of prediction in equity-driven contexts, and how they compare to other policy levers, such as expanding bureaucratic capacity. Through mathematical models and a real-world case study on long-term unemployment amongst German residents, we develop a comprehensive understanding of the relative effectiveness of prediction in surfacing the worst-off. Our findings provide clear analytical frameworks and practical, data-driven tools that empower policymakers to make principled decisions when designing these systems.