You are here:
Comparison of historical redlining HOLC grade exposure assessment metrics for different spatial scales
Citation:
Luben, Tom, M. Jimenez, A. Wilkie, C. Gray, AND K. Rappazzo. Comparison of historical redlining HOLC grade exposure assessment metrics for different spatial scales. Society for Epidemiologic Research (SER) Annual Meeting, Portland, OR, June 13 - 16, 2023.
Impact/Purpose:
We evaluate how assignment of HOLC grades differs across spatial scales within five historically redlined cities in North Carolina (NC).
Description:
Background and Aim: Around the mid-1930s, the Home Owners’ Loan Corporation (HOLC) ranked urban neighborhoods as least to most desirable in terms of perceived mortgage stability. When evaluating the impacts of this racist practice (i.e., redlining), studies have not evaluated potential misclassification to HOLC grades when individual-level locational information is unavailable. We evaluate how assignment of HOLC grades differs across spatial scales within five historically redlined cities in North Carolina (NC). Methods: Using geocoded NC birth records (2003-2015), we created and compared HOLC grade designations at spatial scales of individual residential address, census block group (CBG), census tract (CT), and ZIP code. At the 4 different scales, addresses were assigned HOLC grade A (Best), B (Still desirable), C (Declining), or D (Hazardous). For residential address, we used a “point-in-polygon” approach to assign geocoded addresses to an HOLC grade. HOLC grade was assigned to CBG, CT, and ZIP code based on the grade of the HOLC polygon(s) covering the majority of the land area. We used Pearson correlation coefficients (rP) and ranked classifications (Kappa statistics) to compare the designations made at different spatial scales. Results: Our analysis included 39330 residential addresses; 3311 (8.4%) in grade A, 5776 (14.7%) in grade B, 20204 (51.4%) in grade C, and 10039 (25.5%) in grade D. Assignments to HOLC grade using CBG or CT were well-correlated with HOLC grade using residential address (rP=0.77 for CT and rP=0.78 for CBG); correlations were lower for ZIP code (rP=0.56). The weighted kappa statistic showed a greater degree of agreement to HOLC designation when comparing CBG to residential address compared to CT or ZIP code (Figure 1). Conclusions: This comparison of HOLC grade assignment metrics suggests that the use of CBG or CT, rather than residential address, may be feasible. The use of ZIP code is less feasible in North Carolina.