Percentiles of crime impacts due to climate change, organized by level of aggregation, climate scenario, and time period. Crime impacts are estimated by applying future climate data to dose-response functions of property and violent crime rates, aggregated by population levels. The uncertainty represented by these percentiles comes from a combination of statistical uncertainty in the dose-response functions, climate uncertainty across GCMs, and within-month weather realization, as sampled by Monte Carlo runs. The archive contains four folders: county_20a with county-level impacts, state_20yr with state-aggregated impacts, nca_20yr with NCA-region aggregated impacts, and national_20y with nationally aggregated impacts. Aggregation is weighted by crimes within each region. The files are labeled as follow: crime-<TYPE>-<RCP>-<YEAR><SUFFIX>. The TYPE can be property or violent crime. The RCP may be RCP 2.6, RCP 4.5, RCP 6.0, or RCP 8.5; the years are "2020", for average impacts during 2020 - 2039, "2040" for impacts 2040 - 2059, and "2080" for impacts 2080 - 2099. Files with the suffix 'b.csv' contain percent changes; those with the suffix '-absolute.csv' contain level changes in MT. The columns of these files specify the region (using the FIPS code for counties), and the quantiles from 1% ("q0.01") to 99% ("q0.99"). This data is provided for non-commercial research and educational purposes.