Dataset produced and used in the Cano & Hahn (2023). The data is produced and used by code in the causalCRS github repo: zipcode.fema.fsf.acs.rainfall.v2.csv: Dataset compiled from the FEMA Flood Insurance Mitigation Administration NFIP Redacted Claims data with additional information on community characteristics from: the First Street Foundation data set, US Census Bureau American Community Survey, and PRISM climate group data sets. In total, the data set has 74,288 entries from 14,729. See Cano & Hahn (2023) for details. typology_*.npy: Conditional Average Treatment Effect (CATE) and accompanying measurements for the 27 community typologies defined in the paper. typology_cate.npy: Measured CATE for the 27 community typologies typology_sig_cate.npy: Statistical uncertainty on CATE measurements. typology_y_treat.npy: Claim per policy in USD (outcome) of the community typologies for treated sample typology_y_control.npy: Claim per policy in USD (outcome) of the community typologies for control sample typology_covar.npy: Median covariate values for communities that fall within the typology. This is for reference only. fema_crs.*.npy: Compiled dataset for training/validation/testing the q_phi(Y|X) of the treated/control samples used in CausalFlow. qphi_control.tar.gz / qphi_treat.tar.gz: Compressed tar file that contains all the trained ensemble of flows for the control/treated samples used in CausalFlow. qphi_support.v2.log.control.pt / qphi_support.v2.log.treat.pt: NDEs estimating the distribution of the covariates for the control/treated samples. fema_cid.almost_only_c300.npy: An array of communities that participated in the CRS but only executed the c300 activity, which corresponds to the outreach effort. This dataset is used to estimate the overall impact of outreach.