Reallocating U.S. Election Results from Precincts to Census Geographies

Voting precincts are the most granular spatial units for reporting election outcomes, whereas census geographies, such as block groups, census tracts, and ZIP Code Tabulation Areas (ZCTAs), are commonly used for publishing demographic, economic, health, and environmental data. This dataset bridges the two by reallocating precinct-level votes to standard census geographies through a systematic and replicable framework. The reallocation assumes that votes within each precinct are distributed proportionally to the household population. Household population counts from census block groups—the smallest census unit with regularly updated population estimates—are used to allocate votes to fractions created by the intersection of precinct and census boundaries. This process is implemented using three allocation strategies: areal weighting, impervious surface weighting, and Regionalized Land Cover Regression (RLCR). Results from all three methods are provided. Among these, the RLCR method demonstrates the highest accuracy based on validation against voter-level ground truth data and is recommended as the primary version for analysis. The alternative methods may serve as robustness checks or sensitivity tests. The dataset currently includes the 2016 and 2020 U.S. general elections and is designed for seamless integration with other datasets, such as the American Community Survey (ACS), CDC PLACES, or IRS Statistics of Income (SOI), via the GEOID field.投票区（voting precincts）是报告选举结果的最精细空间单元，而普查地理单元如普查街区组（block groups）、普查片区（census tracts）与邮政编码制表区域（ZIP Code Tabulation Areas，ZCTAs），常被用于发布人口、经济、健康与环境类数据。本数据集通过一套系统且可复现的分析框架，将投票区级别的选票重新分配至标准普查地理单元，打通了两类数据载体的应用壁垒。该重分配机制假设：每个投票区内的选票将按照家庭人口比例进行分配。我们采用普查街区组（block groups）——这一具备定期更新人口估算值的最小普查单元——的家庭人口数，将选票分配至投票区与普查边界相交形成的分片区域中。本分配流程共采用三种分配策略：面积加权法（areal weighting）、不透水面加权法（impervious surface weighting）以及分区土地覆盖回归（Regionalized Land Cover Regression，RLCR），并提供了三种方法的全部计算结果。其中，基于选民级真实数据验证的结果显示，分区土地覆盖回归法的准确率最高，因此推荐作为分析的首选版本。其余两种方法可用于开展稳健性检验或敏感性分析。本数据集目前涵盖2016年与2020年美国大选数据，且可通过GEOID字段与美国社区调查（American Community Survey，ACS）、CDC PLACES、美国国税局收入统计（IRS Statistics of Income，SOI）等其他数据集实现无缝对接。