(
2023-05)
An, Brian Y.; Jakabovics, Andrew
; Orlando, Anthony W.; Rodnyansky, Seva; Son, Eunjee
Scholars and practitioners are increasingly interested in understanding who owns real estate in communities and resultant implications for targeted planning approaches. Yet, practitioners lack an efficient and comprehensive methodology to assess landlords’ ownership scale, namely how many properties they own in a given geographic area. The existence of variegated ownership, multiple legal entities, siloed databases within government bureaucracies, and inconsistencies in spelling and documentation across data entries make it time-consuming and costly to determine the extent of real estate ownership by the same landlords. To address these challenges, this study provides a data-driven natural language processing solution. Using OpenRefine, an open-source software, we present a step-by-step, practice-oriented methodology for amassing data, cleaning textual inconsistencies, and clustering properties to uncover the truer ownership scale in local housing markets. Applied to a large U.S. urban county—Fulton, home to Atlanta, Georgia—our proposed methodology demonstrates its superior efficiency, comprehensiveness, and accuracy, compared to traditional approaches. Using code enforcement as a study frame, we then empirically examine a linkage between landlords’ ownership scale and their code violation patterns. With the proposed methodology in place, the analysis consistently shows that the ownership scale is related to both the likelihood and number of code violations. In contrast, the analysis misses such a critical linkage without applying the methodology. Our methodology yields practical implications regarding targeted code enforcement.