Methodology 2.1 Matching Procedure To construct the control group, I follow the procedure of (Christensen et al., 2023) and use a nearest neighbor matching algorithm to identify 5 cities and townships to use for comparison to Rockford. These places are chosen based on a set of demographic and economic characteristics that capture both the current state of Rockford and its development over time. I restrict the choice set to those places which have between half and twice the population of Rockford (2,500 to 11,000 residents, respectively). Then, the cities closest matches are determined by the following criterion: the places population, percent of vacant houses, median household income, unemployment rate, and the change in household income and population since 1970. The 5 closest matches will be used in the estimation. In order of match distance, they are: Lawrence Township, Ohio; Madison Township, North Carolina; Jefferson Township, Pennsylvania; LaGrange Township, Ohio; and Prairie Township, Indiana. The covariate balance is shown in Table