A National Synthetic Populations Dataset for the United States

Area of Science:

• Computational Social Science
• Demographic Modeling
• Geographic Information Systems (GIS)

Background:

• Accurate, spatially explicit population data is crucial for studying community-level effects.
• Traditional methods often lack the granularity needed for detailed neighborhood analysis.
• Synthetic data generation offers a solution for creating statistically robust, geographically precise datasets.

Purpose of the Study:

• To demonstrate a workflow for generating spatially explicit synthetic populations for the United States.
• To create a dataset representing 2019 U.S. households and individuals at a granular level.
• To enable novel research into community and neighborhood dynamics through synthetic data.

Main Methods:

• Utilized U.S. Census American Community Survey (ACS) 5-year estimates (2015-2019).
• Employed Iterative Proportional Fitting (IPF) to generate synthetic population counts.
• Spatially allocated households using Environmental Protection Agency (EPA) Integrated Climate and Land Use Scenarios (ICLUS) data.

Main Results:

• Generated a synthetic dataset with over 120 million households and 303 million individuals.
• Created a spatially explicit dataset by integrating household distribution estimates.
• Achieved strong validation against original census variables, with Pearson's r correlation coefficients often exceeding 0.99.

Conclusions:

• The demonstrated workflow successfully produces statistically accurate and spatially explicit synthetic population data.
• This methodology provides a valuable resource for researchers investigating neighborhood-level phenomena.
• The high correlation with census data confirms the reliability and utility of the generated synthetic population.