Using Natural Language Processing to Construct a National Zoning and Land Use Database

Area of Science:

• Urban Planning
• Public Policy
• Data Science

Background:

• Zoning and land use policies in the U.S. are associated with high housing costs and residential segregation.
• Existing zoning and land use data rely on costly, time-intensive surveys with limitations like low response rates and measurement errors.

Purpose of the Study:

• To develop a comprehensive and accurate National Zoning and Land Use Database.
• To introduce a parsimonious measure of exclusionary zoning, the Zoning Restrictiveness Index.
• To overcome the limitations of traditional survey-based data collection methods.

Main Methods:

• Utilized natural language processing (NLP) techniques on publicly available administrative data.
• Constructed a new database encompassing a near-universe of municipalities.
• Validated the database against existing indices like the Wharton Residential Land Use Regulatory Index and the National Longitudinal Land Use Survey.

Main Results:

• The new database and Zoning Restrictiveness Index demonstrate consistency with established datasets.
• The NLP approach captured previously omitted elements of land use policy.
• Comprehensive land use regulations were revealed for municipalities in the San Francisco and Houston areas.

Conclusions:

• The National Zoning and Land Use Database provides a more accurate, up-to-date, and longitudinal resource for studying zoning and land use.
• Publicly available code and data facilitate replication and future updates, ensuring data accuracy nationwide.
• This resource can better inform policy decisions related to housing costs and residential segregation.