Methods: This paper details the methods used in a large metropolitan area to develop a typology of landlords to guide policy makers, advocates and community outreach specialists working to implement a healthy housing agenda. Specifically, the paper will cover the technology used to aggregate public records across many government agencies and the data science tools used to uncover landlord patterns and practices in this sector. Using model based clustering and machine learning, the typical landlord profiles were identified. Geospatial analysis methods were used to examine the relationship between the density of properties by landlord type and the incidence of lead poisoning in young children.
Results: The landlord types were characterized by a combination of characteristics including numbers and types of properties owned, market valuation, corporate structure, housing code violations, frequency of evictions, tax delinquencies, and whether the landlord was local or an out of town business. Geospatial analysis revealed that neighborhoods differed in their mix of landlord types, and that the neighborhood landlord mix was predictive of rates of lead poisoning in children.
Conclusions: Housing quality problems are severe in the low cost rental market where small and under-capitalized landlords predominate. By using an empirically derived landlord typology, community advocates are able tailor their policy recommendations and outreach strategies to most effectively regulate and incentivize housing improvements in the low-cost rental market.