# A dataset of US precinct votes allocated to Census geographies with precision

## TL;DR

This paper introduces a dataset that links U.S. voting precincts to census geographies, enabling analysis of voter behavior alongside demographic and environmental data.

## Contribution

The paper introduces a novel method, RLCR, for allocating precinct votes to census geographies with higher accuracy than existing methods.

## Key findings

- The RLCR method outperforms surface area and imperviousness methods in vote allocation accuracy.
- The datasets cover the 2016 and 2020 U.S. general elections and enable integration with demographic and health data sources.
- Validation tests using census blocks and voter-level data confirm the superior performance of RLCR.

## Abstract

Voting precincts are the finest spatial units for recording U.S. election results, while census geographies, including block groups, census tracts, and ZIP Code Tabulation Areas (ZCTAs), provide administrative data on demographic, economic, health, and environmental factors. This paper presents datasets that link precinct-level voting records to census geographies with precision. The allocation assumes votes within a precinct are proportional to household population, with population distributed from block groups to overlapping precinct fractions using the Regional Land Cover Regression (RLCR) method. Datasets based on surface area and imperviousness methods are also provided, but RLCR outperforms them across multiple error metrics in validation tests using census blocks and voter-level data. Covering the 2016 and 2020 U.S. general elections, these datasets facilitate merging voting records with sources like the American Community Survey, CDC Places, and IRS Statistics of Income to explore voter behavior across various contexts.