Xenophobic violence in Sweden 2009-2022

This dataset contains all police-reported and hate-crime–flagged incidents of xenophobic physical violence recorded in Sweden between 1 January 2009 and 31 December 2022. It is a nationally complete event-level dataset comprising 2,522 unique police reports. The data material includes only cases involving actual or attempted physical violence where a xenophobic motive was either explicitly expressed or clearly inferable from the police narrative. Non-violent offences, harassment, threats, symbolic hate expressions, and other hate-crime categories are not included. The dataset is constructed from structured and unstructured fields in Swedish police reports. Standardized variables (e.g., offence codes, timestamps, and reporting metadata) were extracted programmatically, primarily using Python scripts. Free-text descriptions were manually reviewed to classify motives, locations, and offender/victim characteristics. Several analytic variables were created by combining multiple fields (e.g., collapsed offence categories, derived motive groups, temporal variables). Strict exclusion criteria were applied to remove misclassified cases, unclear motives, incidents involving professional enforcement personnel, and intergroup conflicts not tied to xenophobic targeting. All coding procedures and variable definitions are documented in an accompanying codebook. The dataset consists of 26 variables grouped into six domains: 1. Metadata (e.g., report date, reporting method) 2. Offence characteristics (official offence codes, collapsed offence categories, hate-crime motive) 3. Contextual information (incident date and time, physical setting, coordinates aggregated to 1×1 km grid cells, municipality and region) 4. Victim characteristics (age category, gender) 5. Offender characteristics (number of offenders, gender) 6. Type of violence (type of physical action and weapons/objects used) The dataset is provided as a de-identified, row-level event file. Free-text narratives and personal identifiers have been removed to comply with ethical and legal requirements; geographical information is aggregated to grid level to reduce the risk of re-identification. No additional software beyond standard statistical tools (e.g., R, Python, Stata, SPSS) is required to read the file. Scripts used for automatic extraction and variable derivation can be supplied on request. A codebook describing all variables, coding decisions, and exclusion criteria accompanies the dataset. This material enables researchers to examine long-term patterns, spatial and temporal distributions, and characteristics of xenophobic violence in Sweden using a transparent and replicable administrative-data framework.