Methodology and Definitions

This report uses mixed methods research to examine housing loss resulting from evictions and mortgage foreclosures across the Sun Belt. In addition to a brief section on findings that are generalizable across the region, we focus on case studies of seven U.S. counties:

1. Clark County, Nevada
2. Forsyth County, North Carolina
3. Harris County, Texas
4. Maricopa County, Arizona
5. Miami-Dade County, Florida
6. Norfolk City, Virginia¹
7. Orange County, Florida

In each case study location, we analyzed geospatial and case data on evictions and mortgage foreclosures at the census tract level to visualize where housing loss was occurring. We utilized American Community Survey (ACS) data from the U.S. Census Bureau and tested for statistical relationships between housing loss and a number of socioeconomic variables to visualize who was most impacted by housing loss. And, we conducted key informant interviews (KIIs) to better understand the causes of home loss and the consequences of displacement, particularly within the context of the COVID-19 pandemic. These KIIs, along with desk research, helped to provide geographic, demographic, economic, social, political, and historical context to our housing-related findings.

Housing Loss Rate: Prior studies tend to examine different mechanisms of housing loss in silos. Eviction and mortgage foreclosure are analyzed separately, rather than as components of the same, broader problem. The processes of eviction and foreclosure may be different, yet the underlying causes are often the same, and each results in displacement and trauma—financially, physically, and emotionally. In most U.S. cities, a worker on minimum wage will be unable to make housing payments—either rent or a mortgage. The impact on a child who switches schools three times a year due to housing instability is likely similar whether home loss occurs via eviction or foreclosure. And residents of a blighted neighborhood likely do not care if their block of empty homes is a result of foreclosures or chronic evictions.

New America’s Displaced in America report addressed this problem by introducing a new measure that captures the overall magnitude of both evictions and mortgage foreclosures: the housing loss rate.

The housing loss rate combines the total number of evictions and the total number of mortgage foreclosures for a given geography, and then normalizes that sum by the total number of renters and the total number of homeowners with a mortgage within that given geography. The resulting rate shows the scale of housing loss within a given geography in relation to the number of households who could potentially experience such a loss.

Households: In keeping with the definition above, unless specified otherwise, "households" in this report refers to the combination of renters and homeowners with a mortgages, and not all households in the county. Specifically, we are excluding homeowners who have paid off their mortgage. These homeowners may be vulnerable to tax lien foreclosure, which we do not explore in this report, but they are generally not vulnerable to mortgage foreclosure or to eviction.

Case Study Selection Criteria: We decided to focus on the Sun Belt because of its rapidly growing and diversifying population, relatively high rates of pre-pandemic housing insecurity, and high rates of unemployment and housing insecurity during the pandemic. Within the Sun Belt we selected five locations based on a number of requirements:

1. An early assessment of the prevalence of housing loss and instability, based on our Displaced in America report.
2. The availability of granular and mappable data for evictions and mortgage foreclosures.
3. The availability to leverage existing networks in order to engage with local stakeholders and conduct qualitative analysis.
4. Regional variation, in order to account for differences in economics, politics, demographics, and histories.

In addition to the five newly-selected locations, we updated our findings from two of our Displaced in America case studies that happened to be in the Sun Belt: Maricopa County, Arizona and Forsyth County, North Carolina.

Quantitative Methodology: Together with our data science and visualization partner, DataKind, we located, cleaned, standardized, and visualized data on evictions and mortgage foreclosures for each case study location. In our analysis, we tested for any statistical relationships between housing loss and a number of socioeconomic variables via correlation analysis using five-year (2012-2016) ACS estimates from the U.S. Census Bureau. Where statistical relationships from the correlation analysis were strong, we visualized housing loss and that variable through maps and scatter plots. Note that for this analysis, we use the median housing loss rate such that 50 percent of census tracts fall below and above this metric, whereas in the rest of the report, we use the mean housing loss rate.

Project Data Sources

Qualitative Methodology: From September to December 2020, a group of researchers comprising doctoral and undergraduate university students and New America staff conducted key informant interviews (KIIs) with a range of stakeholders, including government officials, housing advocates, real estate developers, journalists, lawyers, service providers, and community members in each case study county, in order to gain an in-depth understanding of local issues related to housing loss. Questions were developed collaboratively, and focused on:

• How often residents lost their homes;
• Who was most at-risk of losing their home;
• Where within the relevant county this loss was most acute;
• Why people were losing their homes;
• What happened after they did; and
• How the COVID-19 pandemic was impacting housing stability in the case study location.

The researchers provided us with recordings and transcripts of the KIIs, a written summary of each interview, and a summary of findings.

Definitions: For definitions of eviction and foreclosure, and the way in which we chose to measure these two mechanisms of loss, please see the methodology section of our Displaced in America report. Our only significant deviation from the methodology that we used in the Displaced in America study is that in the current report we did not include short sales in our definition of foreclosure. Short sales occur when a distressed homeowner gains permission from their lender to sell their property for an amount less than what is due on the mortgage. While some data collectors do include short sales in their foreclosure counts, we decided to exclude this mechanism because we felt that, for all of its negative impacts on the homeowner, it did not amount to a forced displacement. As a result of this choice, our foreclosure figures are generally lower than the figures we found in the national section of our Displaced in America report.

Data Notes and Caveats

Comparisons across Timeframes and Counties: Counties vary greatly in how they collect, store, and share eviction and foreclosure data, and even in how they define evictions and foreclosures. As a result, while comparisons of eviction and foreclosure rates across counties have some utility, they remain limited.

Undercounting: The eviction data included in this report are for formal evictions, or evictions carried out through the court system. We know through our qualitative interviews that informal evictions, which occur outside the legal system through buyouts or illegal lockouts for example, can be just as common, especially in Black and Latinx communities. As such, the housing loss occurring in communities across the country is likely greater than what is in this report.

Average Rates: In a small number of census tracts across our case studies, housing loss data was available for some years and unavailable for other years. Because we could not determine whether data was simply missing or that housing loss did not occur, we decided to calculate average rates only for those years in which data was available.

Clark County and Norfolk City Eviction Findings: In Clark County and in Norfolk City we were only able to source eviction data at the zip code level. In Norfolk City, we analyze evictions at the zip code level. In Clark County, our census tract visualizations (on average there are approximately two census tracts in every zip code) are estimates derived from an API service provided by the U.S. Department of Housing and Urban Development’s (HUD) Office of Policy Development and Research (PD&R), which is available here. This service allows the user to input a zip code and receive as an output the list of census tracts which comprise that zip code. Furthermore, the service provides information about the fraction of residential area found in each census tract, which allowed us to proportionally assign eviction records according to the amount of residential area. This is useful in that it enabled us to estimate evictions at a more granular level than what exists in our source data; however, it also means that evictions are calculated and visualized proportionately to their area, which may conceal census tracts with high numbers of evictions relative to their populations, or over-estimate evictions in census tracts with very large residential land areas but which are, in actuality, sparsely populated.

Additionally, while Clark County evictions are processed through eleven justice courts, we were only able to obtain eviction data from the Las Vegas Justice Court. This court accounts for 83 percent of eviction filings, however it must be noted that a significant number of evictions are missing from our data.

Orange County Calculations: We also produced some eviction estimates for Orange County census tracts. Our source data contains eviction filings from all of 2017, September through December 2018, and January through August 2019, but was missing filings from the first eight months of 2018 and the last four months of 2019. However, given that we had 24 months’ worth of data, and could not identify census tracts whose eviction totals were likely to vary widely year to year, we were able to take monthly eviction filing counts provided by the Orange County Clerk’s office and use information about the relative proportion of evictions for each tract to assign filings to tracts for the months for which we lacked individual filings.

Estimating Evictions or Eviction Filings: For Clark County, Norfolk City, and Orange County we were able to source eviction filings, but these filing records did not indicate whether the eviction case was decided in favor of the plaintiff (landlord) or defendant (tenant). Similarly, for Forsyth County, our source data contained dispositions in favor of the plaintiff, but not the number of eviction filings. Since not every eviction filing results in an eviction, for these cases we also used census tract-level eviction rate data from the Princeton Eviction Lab (averaged across 2014 through 2016) to estimate the number of evictions for those counties for which we had filings, but not evictions, and to estimate the number of filings for the county for which we had evictions, but not filings.