Neighborhood Labeling and Criminal Justice Decisions

In 2015, the Swedish National Police began publishing an internal list of “troubled neighbourhoods,” shaping public discourse and increasing negative media coverage despite no accompanying policy changes. Using a regression discontinuity design, we estimate the causal effect of neighbourhood listing on criminal justice outcomes by comparing defendants awaiting trial at the time of release with those sentenced just prior. Sentence lengths increased by over 60% (approximately 260 days) for defendants sentenced after listing updates. The effect is immediate, fades with media attention, and is strongest in courts where fewer judges have lived in these neighbourhoods. Affected defendants also experience a 16 percentage point decline in employment over the three years following sentencing. Using data on the full universe of court cases, we apply machine learning to analyse judicial reasoning and find that judges place less emphasis on legal detail and evidentiary complexity for affected defendants. These findings are consistent with limited-attention models in which salient public narratives influence decisions.