A scalable Bayesian double machine learning framework for high dimensional causal estimation, with application to racial disproportionality assessment

TL;DR

This paper introduces a scalable Bayesian double machine learning framework for high-dimensional causal inference, applied to assess racial disproportionality in Stop and Search practices and its impact on expressive crimes in London.

Contribution

It develops a novel semi-parametric Bayesian method combining empirical likelihood and double machine learning for high-dimensional causal estimation.

Findings

Racial disproportionality can be mitigated by considering Black population proportions.

The proposed method provides valid posterior coverage in high-dimensional settings.

Application reveals potential policy implications for reducing racial disparities.

Abstract

Racial disproportionality in Stop and Search practices elicits substantial concerns about its societal and behavioral impacts. This paper aims to investigate the effect of disproportionality, particularly on the black community, on expressive crimes in London using data from January 2019 to December 2023. We focus on a semi-parametric partially linear structural regression method and introduce a scalable Bayesian empirical likelihood procedure combined with double machine learning techniques to control for high-dimensional confounding and to accommodate strong prior assumptions. In addition, we show that the proposed procedure yields a valid posterior in terms of coverage. Applying this approach to the Stop and Search dataset, we find that racial disproportionality aimed at the Black community may be alleviated by taking into account the proportion of the Black population when focusing on expressive crimes.