Classifying Opportunity Zones—A Model-Based Clustering Approach

Jamaal Green
University of Pennsylvania

Wei Shi
Travelers Insurance

Objective: Opportunity Zones (OZs) are the first major place-based economic development policy from the federal government in nearly two decades. To date, confusion persists among planners and policymakers in some places as to what features of OZ tracts matter for their inclusion, and, secondly, what features of OZ tracts make them attractive targets for potential investment. The authors developed a typology of OZ tracts in order to offer planners and policymakers alternative ways of organizing a highly variable set of tracts.

Methods: This study employs model-based clustering, also known as latent class analysis, to develop a typology OZ tracts from the population of all eligible tracts in the United States. The authors use publicly available data from the U.S. Census Bureau and Urban Institute in developing the typology. Descriptive statistics and graphics are presented on the clusters. Using Portland, Oregon, as an example city, the authors present a cartographic exploration of the resulting typology.

Results: OZs present with immense variation across clusters. Some clusters, specifically cluster 3 and 9, are less poor, have a greater number of jobs and higher development potential than other clusters. Additionally, these exceptional clusters have disproportionate rates of final OZ designation compared to other clusters. In Portland, these less distressed clusters make up the majority of ultimately designated OZ tracts in the city and are concentrated in the downtown area compared to the more deprived eastern part of the city.

Conclusions: We find that OZ designation is disproportionately seen in particular clusters that are relatively less deprived than the larger population of eligible tracts. Cluster analysis as well as other forms of exploratory or inductive analyses can offer planners and policymakers a better understanding of their local development context as well as offering a more coherent understanding of a widely variant set of tracts.

OZs, the newest federal government place-based economic development tool since the New Markets Tax Credit in the early 2000s, has reportedly marshaled more than $50 billion in investment in the 2 years since its passage (Drucker and Lipton, 2019). Opportunity zones allow investors to defer taxes on their capital gains if they invest in qualified Opportunity Zone funds in development-starved census tracts.

Recent investigations show a disproportionate amount of investment being steered into a minority of tracts that formally qualified for the program based on their income but are not suffering from a lack of development (Buhayar and Leatherby, 2019; Drucker and Lipton, 2019; Ernsthausen and Elliott, 2019).

A central tension in those articles concerning Opportunity Zone investment is that the Tax Cut and Jobs Act of 2017 used a broad qualifying rule for Opportunity Zone designation based only on tract income to maximize flexibility. It resulted in variations within designated Opportunity Zones in terms of their socioeconomic characteristics but also redevelopment attractiveness. An important issue for economic development researchers and analysts is to find alternative ways of organizing Opportunity Zones into more useful categories of analysis than simply qualified or non-qualified Opportunity Zone designations.

This paper presents model-based clustering, also known as latent class analysis. This unsupervised machine learning technique is one way to address the difficulties of classifying designated Opportunity Zone tracts. The remainder of this article will offer background on some troubling OZ issues, a description of latent class analysis through model-based clustering, and the results of cluster analysis and its relationship with Opportunity Zone designation. The findings contribute to a better understanding of the variation of eligible tracts and what features make the zones attractive for designation.