Mapping Manufactured Housing Nationwide

Noah J. Durst
Nithya Mylakumar
Angela Perez
Michigan State University

Angelica Gacis
Data Science Consultant

This study uses building footprints from Microsoft and OpenStreetMap and the Python package momepy to measure the shape, size, and placement of buildings and their 5, 10, and 20 nearest neighbors across the continental United States. Using estimates of building and neighborhood morphology and machine learning, we predict whether each building is a singlewide manufactured home and whether it is in a manufactured home park, informal or manufactured home subdivision, or another setting. We describe the methods used to create these predictions and discuss issues of model performance and their implications for future research, compare our estimates with the locations of manufactured homes documented in the American Community Survey and with government and private registries of these communities, illustrate their distribution nationwide, and present descriptive statistics on their demographic and socioeconomic characteristics. Our findings illustrate that manufactured home parks are more common in Midwestern and Northeastern states, whereas informal or manufactured home subdivisions are more common in Southern and Western states. We find that both neighborhoods are demographically diverse but economically disadvantaged. We conclude by briefly discussing the implications of our research for state and federal housing policy.