Methodology

Data Sets

American Community Survey Public Use Microdata

The American Community Survey (ACS) program of the U.S. Census Bureau is the primary data source for estimates presented in this tool. Through the ACS, the Census Bureau surveys roughly 3 million households nationwide each year to assess key housing, demographic, and economic characteristics. Data from this survey sample — which represents about 2.5 percent of households nationwide — are used to produce estimates reflecting one year’s and five years’ worth of survey data for geographies of various sizes. Additionally, the Census Bureau releases a set of anonymized individual responses to ACS questionnaires in what are known as Public Use Microdata Sample (PUMS) files. This analysis relies on one-year (2017 and 2022) and five-year (2013–2017 and 2018–2022) PUMS files, accessed via the Minnesota Population Center IPUMS-USA database (IPUMS-USA, University of Minnesota, www.ipums.org).

The lowest level of geography at which PUMS data are available is a specialized census geography called a Public Use Microdata Area (PUMA).¹ These are areas with populations of at least 100,000 that were developed to preserve the privacy of individual respondents. Densely populated geographies, such as the city of Philadelphia, are split into multiple PUMAs that fit together within their boundaries, whereas PUMAs in less dense areas generally contain multiple counties or parts of counties. In this analysis, PUMAs were matched to metropolitan statistical area (MSA) and state boundaries using information provided by the Missouri State Data Center’s MABLE/Geocorr2K and MABLE/Geocorr12 applications.²

The determination of whether to use the one-year or five-year ACS files was made based on the number of renter households in each geography. The one-year files were used for geographies with at least 200,000 renter households, and the five-year files were used for all other geographies. The geographies considered for inclusion in this analysis were states, Metropolitan Statistical Areas (MSAs), Metropolitan Divisions (MDs), and populous counties within multicounty MSAs. Changes in the PUMA geographies following the 2020 Census were implemented in the 2022 ACS PUMS, presenting major cross-year comparability challenges for this update. As a result, estimates were only produced for geographies that could be wholly or very closely reconstructed with both sets of PUMA geographies. The included geographies and ACS samples are summarized below.

One-Year ACS Files

• States: New Jersey, Pennsylvania
• Metropolitan Divisions: Philadelphia, PA
• Counties: Philadelphia, PA

Five-Year ACS Files

• States: Delaware
• Metropolitan Areas: Allentown-Bethlehem-Easton, PA-NJ; Atlantic City, NJ*; Dover, DE; Erie, PA; East Stroudsburg, PA; Harrisburg-Carlisle, PA; Johnstown, PA; Lancaster, PA; Lebanon, PA; Reading, PA; State College, PA; Trenton-Princeton, NJ; York-Hanover, PA
• Metropolitan Divisions: Camden, NJ; Montgomery-Bucks-Chester, PA
• Counties: Allegheny, PA; New Castle, DE

* Estimates omit a small portion of Atlantic County that falls within the 34-02600 (2010) and 34-02601 (2020) PUMAs, representing approximately 4 percent of households in the MSA.

National Housing Preservation Database

Supplementary data on the federally assisted rental stock was accessed through the National Housing Preservation Database (NHPD).³ The NHPD integrates data from various federal housing programs, such as those administered by the U.S. Department of Housing and Urban Development (HUD), the U. S. Department of Agriculture (USDA), and the Low-Income Housing Tax Credit (LIHTC) program.⁴ The address-level database provides detailed information on property and subsidy characteristics, including the number and types of subsidies attached to a property, the number of units assisted by each program, and relevant contract expiration or loan maturity dates. The NHPD includes geographic identifiers corresponding to the state, county, and metropolitan area in which the property is located. The database is intended to be used by state and local stakeholders to identify assisted properties at risk of exiting the affordable stock. The NHPD is updated quarterly; data from the August 2024 update were used in this analysis.

Analysis

ACS estimates of median family income (MFI) were used to categorize the individual housing unit/household records contained in the PUMS file into relative affordability and income categories.  PUMAs that were wholly contained within an MSA were assigned the MFI of that MSA. Single-county nonmetro PUMAs were assigned the county’s MFI. PUMAs containing multiple nonmetro counties or parts of counties were assigned the PUMA’s MFI. Last, PUMAs containing a mix of MSA geographies or MSA and nonmetro geographies were assigned a weighted average of the MSA(s) and/or county MFIs based on the proportion of housing units falling within each. The time periods of the MFI estimates used for MSA-level analyses correspond to those of the PUMS file (e.g., if analysis of the State College MSA uses the five-year ACS PUMS, the five-year ACS MFI estimate was used as the underlying PUMA income). To produce one-year rental housing estimates for New Jersey and Pennsylvania, one-year estimates were used for all PUMAs and aggregated to develop estimates for these larger geographies.

By comparing household-level income reported in the PUMS file with the MFI for the broader geographic region, renter households were divided into the following income categories to enable closer analysis of different income segments:

• Extremely low-income (ELI): households with incomes ≤30 percent of the regional MFI
• Very low-income (VLI): households with incomes 31–50 percent of the regional MFI
• Low-income (LI): households with incomes 51–80 percent of the regional MFI
• Not low-income: households with incomes >80 percent of the regional MFI

Households reporting zero or negative income that paid more than the Fair Market Rent (FMR)⁵ for their units during the survey year were reclassified as “not low-income,” since their income situation was assumed to be either temporary or not indicative of broader economic hardship. To assign FMRs to PUMAs, county-level FMRs were merged to PUMA boundaries based on the survey year, then the appropriate FMR was assigned to housing units based on the number of bedrooms. For PUMAs covering multiple counties or MSAs, the highest applicable FMR value was assigned. For five-year ACS PUMS files, the FMR that corresponds to the actual year the household was surveyed was applied, then adjusted for inflation to be consistent with the dollar values in which estimates are reported (e.g., for households in the 2018–2022 files, FMRs were adjusted to 2022 dollars). In accordance with the FMR methodology of the U.S. Department of Housing and Urban Development (HUD), FMRs for unit sizes larger than four bedrooms were calculated by adding 15 percent for each extra bedroom to the four-bedroom FMR.

To account for the variation in income needs of households of various sizes, we adopted HUD’s approach to adjusting MFI estimates based on household size.⁶ An unadjusted MFI was used to categorize households of four, with a downward adjustment of 10 percent for each person fewer than four. For each person in a household exceeding four, the MFI was adjusted upward by 8 percent. Table 2 summarizes the adjustments to households consisting of one to eight residents.

Once households were sorted into income categories, the ratio of their monthly gross housing costs to their monthly income was calculated to assess the variation in housing cost burdens across income levels. For units in which utility costs were separate from rent, gross housing costs were calculated as the sum of rent and utility costs. Households for which monthly gross rent exceeded 30 percent of monthly income were considered burdened by their housing costs, whereas those paying more than 50 percent of their monthly income were considered severely cost burdened.

The housing units themselves — including those that were vacant — were also sorted into affordability categories to enable analysis of the supply of affordable rental units. Utility costs for vacant units were imputed based on the median utility costs for similarly sized units in the associated state for the given sample year and added to the reported contract rent to produce an estimate of gross rent. To be considered affordable, monthly gross rent could not exceed 30 percent of monthly household income. Units were categorized using the MFI-based income thresholds outlined previously. Again, adjustments were made to reflect the variation in income sufficiency for different household sizes. Following HUD’s methodology, the applicable household size was inferred from the number of bedrooms in a unit. Efficiencies were assumed to accommodate one person, one-bedroom units were assumed to accommodate 1.5 persons, and each additional bedroom was assumed to accommodate an additional 1.5 persons.

It may be instructive to include an example of how a unit would be categorized. Assume there is a two-bedroom unit with a gross rent of $1,000 in an area with an MFI of $70,000. The applicable income thresholds in that community would be:

• ELI threshold: 30% × $70,000 × 90%⁷ = $18,900
• VLI threshold: 50% × $70,000 × 90% = $31,500
• LI threshold: 80% × $70,000 × 90% = $50,400

At a rent of $1,000 per month, the unit is affordable at an annual income of $36,000 ($1,000 monthly rent/30 percent to calculate the monthly income needed to afford the unit, multiplied by 12 months to produce the annual income, multiplied by 90 percent to adjust for a three-person household), placing it above the ELI and VLI thresholds but affordable to a three-person household in the LI category.

To supplement the indicators produced using the ACS PUMS, the NHDP data were cleaned and processed to derive estimates related to the federally assisted housing stock. First, the full NHPD data set was downloaded and filtered to records in Delaware, New Jersey, and Pennsylvania.⁸ Properties with inactive or inconclusive subsidies were removed from the data set. The remaining properties were then classified into one of four subsidy types: HUD Public Housing, HUD Multifamily,⁹ USDA Multifamily,¹⁰ and LIHTC.¹¹

Owing to data constraints and the frequent use of multiple subsidy types within properties, it is not feasible to precisely estimate the total number of affordable units attributable to each federal program. To produce reasonable estimates based on available information, the following rules were implemented to classify the assisted stock: (1) Any property that is identified as being assisted through HUD Public Housing is classified as HUD Public Housing, regardless of whether additional subsidies are attached to the property; (2) for all other records, the subsidy category is assigned based on the program associated with last provided subsidy end date;¹² (3) properties assisted solely through Mod Rehab and/or Project-Based Vouchers (which do not have fixed terms/expiration dates) are categorized as HUD Multifamily.

Although the total number of units in a property is provided in the NHPD, owing to variations in program requirements, not all units in assisted properties are necessarily affordable to households with incomes ≤ 80 percent of regional MFI. The following steps were taken to estimate the number of affordable units: (1) for HUD Public Housing, the unit counts associated with each reported Public Housing subsidy were summed and compared with the total units reported at the property level; the lesser of these values was used; (2) for all other subsidy types, the reported number of assisted units in each program (excluding those assisted through FHA and USDA Section 538¹³) were summed and compared with the total units reported at the property level; the lesser of these values was used.

Last, to identify units with subsidies that are set to expire within the next 10 years, records for which the property-level “Last End Date” variable had a year value less than 2035 were flagged in the data set. All affordable units in these properties were categorized as expiring by 2035, although some subsidies attached to the property may expire sooner. Units categorized as HUD Public Housing were omitted from this calculation, even if there were other subsidies with expiration dates (e.g., LIHTC) attached to the property.

Estimates

While many of the estimates produced in this analysis were the result of simple cross tabulations (e.g., rates of cost burden by household income category), others required additional calculations. Following the methodology outlined in HUD’s “Worst Case Housing Needs”series of reports to Congress,¹⁴ the ratio of affordable and available units for households at or below the three income thresholds developed in the analysis (≤30 percent MFI, ≤50 percent MFI, and ≤80 percent MFI) were calculated to assess the sufficiency of the existing affordable rental stock relative to demand. A unit was considered affordable if gross rent would not exceed 30 percent of the monthly household income at the given income level (e.g., 30 percent of the monthly income of a household at 50 percent of the regional MFI). A unit was available if it was either vacant or currently occupied by a household at or below the given income level.

For example, if there were 1,000 renter households with income at or below 30 percent of the regional MFI and 900 units affordable to a household at that income threshold, there would be 90 units affordable for every 100 renter households.¹⁵ However, if 400 of those units were occupied by households with income greater than 30 percent of the regional MFI, there would only be 500 affordable and available units for households at or below the income threshold, or 50 for every 100 renter households.¹⁶ It is important to note that these ratios were cumulative for each income level, including all housing units and households that fell at or below a given affordability or income category.

Additionally, the overall deficit or surplus of units affordable and available at different income thresholds is reported for the most recent estimate year or period. These estimates were calculated by subtracting the number of households at or below a given income threshold from the number of available units at or below the corresponding affordability category. Like the ratios, these estimates are cumulative.

Last, to help inform preservation strategies aimed at the existing stock of housing in which low-income renters reside, household income categories were cross-tabulated by relevant unit characteristics. These included binned categories of the type of structure that the units were part of (e.g., single-family home, duplex, multifamily building, etc.) and the year the structures were built, both of which were available from the ACS PUMS.