Disproportionality: Exploring the Nature of Ethnic Disparities in Sentencing through Causal Inference

Empirical research consistently shows that, for similar crimes, offenders from ethnic minority groups tend to receive harsher punishments than their white counterparts. These disparities have been extensively documented across different jurisdictions, types of offences, and sentencing outcomes. However, a crucial question remains: can these disparities be definitively interpreted as evidence of discrimination?
This raise the challenge of determining whether these disparities are due to unobserved case characteristics. The common analytical approach has been to ‘control for’ all relevant case characteristics. But what if these differences are influenced by factors like offender dangerousness, which are difficult to measure and control? Moreover, what if these case characteristics are not defined neutrally and are subject to potential biases? These important methodological questions remain unresolved.
Rather than dismissing ethnic disparities outright due to the limitations of achieving perfect ‘like with like’ comparisons, we suggest reframing the question: How strong would the influence of unobserved relevant case characteristics need to be to account for the ethnic disparities observed in existing research?
For this, we propose using a new sentencing datasets from the Data First project made available by Administrative Data Research UK and the Ministry of Justice, alongside some of the latest sensitivity analysis techniques developed in Epidemiology, to address this methodological challenge.