The HPRM is based on the hypothesis that employment is the primary resource for maintaining housing. Therefore, pre-pandemic risk of forced migration through displacement or eviction was exacerbated by 2020 unemployment and COVID-19 infection. To calculate the HPRM score, we developed a point system for each variable of displacement, eviction, unemployment, and unemployment change and then summed the values to create the HPRM index.
We use US Census data to estimate 2019 displacement using the Urban Displacement Project’s Gentrification and Displacement methodology. From this methodology, we use the neighborhood typologies of “low income/susceptible to displacement” to give tracts 1 point on the HPRM index and “ongoing displacement,” 2 points. Tracts without either of these two typologies received 0 points.
Next, we use eviction data from the Eviction Study and The Eviction Lab to analyze eviction rates in our study areas. Because eviction data are sparse and largely unavailable, we developed a groundbreaking eviction risk estimation model that used complete eviction lockout data across 17 metros (Atlanta, Boston, Charleston, Charlotte, Chicago, Cleveland, Denver, Kansas City, Miami, Oklahoma City, Orlando, Raleigh, Richmond, Seattle Puget Sound, St. Louis, Tampa Bay, Virginia Beach) to extract the primary US Census variables that predict eviction. We controlled for over a hundred variables related to eviction using a Bayesian additive regression tree (BART) model, isolating the following primary tract factors: percent Black, rent, change in rent, percent with college degrees, percent of households with seniors, percent married, and building age. These variables proxy complex conditions that households face related to eviction risk. For example, percent Black is related to U.S. legacies of forced segregation and housing discrimination over time. We then calculated points for each tract based on the eviction rate for the related levels of each variable. For example, we found that tracts with less than a 9% Black population had, on average, less than 2% eviction rate (2% being the overall average eviction rate among our 17 metros of data) while tracts with 9% to 31% Black had an average 2% to 4% eviction rate, and above 31% had greater than 4% eviction rate. Therefore, tracts with a 0% to 9% Black population received 0 points, 9% to 31% received 1 point, and above 31% received 2 points. We then replicated this process with the other six variables and summed the tract scores together to create an eviction index. We then compared our index to actual eviction data to confirm our estimation. We found that tracts with 0 to 3 eviction points mostly had a less than 2% eviction rate, 4 to 5 points ranged with less than 2% to over 4% eviction rate, and tracts with 6 to 11 points had largely higher than 4% eviction rate. Using these results we then calculated this index for all tracts across our 53 metro areas. For the Seattle, San Francisco, and Chicago map, we relied on actual eviction data drawn from the Evictions Study and San Francisco eviction notice data portal.
For 2020 unemployment data, we used tract level estimates developed by Catalyst based on PUMS and weekly unemployment insurance claims data. We used 2019 ACS 5-year tract level unemployment data to calculate change in unemployment from 2019 to 2020. Unemployment rates of less than 8% received 0 points, 8% to 14% received 1 point, and greater than 14% received 2 points. Unemployment change less than 8 percentage points between 2019 and 2020 received 0 points while greater than 8 percentage point change received 1 point.
COVID-19 data were collected directly from the county health departments of San Francisco, California; Cook County, Illinois; King County, Washington; and Pierce County, Washington.