Housing Landscape 2015 Methodology

This report is based on American Community Survey (ACS) data collected by the U.S. Census Bureau in 2010, 2011, 2012 and 2013. Estimates in this report were generated using Public-Use Microdata Sample (PUMS) population and housing files made publicly available by the Census Bureau. Each file includes roughly 40 percent of the full ACS sample for its respective year, resulting in over three million records in each population file and over one million records in each housing file. There is a unique identifier that links individuals in the population file to households in the housing file. The only geographic identifiers are the state, the census region and the Public-Use Microdata Area (PUMA) of residence. PUMAs are locally defined geographic areas that allow researchers to produce socioeconomic and demographic estimates with ACS data for sub-state geographies.[1] Each PUMA has a minimum population of 100,000.

The remainder of this section explains how the PUMS files and constituent variables were used to develop the estimates in this report.

Metropolitan Area Estimates: The ACS PUMS files were used to generate metropolitan area statistics by associating each PUMA with the metropolitan area (or non-metropolitan area) in which it is located.[2]  These PUMA-to-metropolitan area relationships were generated using the Missouri Census Data Center’s MABLE/Geocorr12 online application.[3] Because not all PUMAs are entirely contained within a metropolitan area, each PUMA was assigned to a metro area if at least 50 percent of its housing units fell within the area’s boundaries. PUMAs that did not fall at least 50 percent within a metropolitan area were coded as non-metropolitan.

One consequence of using this “50 percent rule” is that where metropolitan area and PUMA boundaries are not coterminous, either too few or too many households are assigned to the metro area (i.e., if a PUMA falls 75 percent within a metro area, all of its households are considered to reside in the metro area even though 25 percent, in actuality, do not). However, the PUMAs do a sufficiently good job of approximating the metropolitan areas.

Household Income Relative to Area Median Income: For each household assigned to a metropolitan area, household income (variable HINCP in the PUMS file) was compared to the area’s median family income estimate (ACS detailed table B19113), adjusted for household size.[4], [5], [6] The ratio of household income to this AMI was used to determine the income category for each household, as well as to identify whether or not the household met the income requirements for working household classification (i.e., <=120 percent of the AMI). Non-metropolitan AMIs for each state were derived from the household records classified as non-metropolitan in the PUMS files. The households reporting zero or negative income in each of the study years (roughly 1.7 million households in 2013) were excluded from these analyses.

Housing Costs: The PUMS housing files include two variables that aggregate monthly housing costs for owners and renters. For owner-occupied households, this variable (SMOCP) includes first and second mortgage payments, property taxes, insurance, homeowner association fees and utilities; for renter-occupied households, this variable (GRNTP) includes cash rent and utility costs. This analysis used the Census Bureau’s aggregation for owner-occupied households but replaced the renter housing cost aggregation with a custom-calculated variable. This was necessary because the PUMS housing file does not aggregate housing costs for renters that do not pay cash rent, even if they pay utilities. Because using the PUMS variable would have excluded these households from the analysis, a replacement variable was calculated that sums utility costs for renter-occupied households that do not pay cash rent. 

 

Archived methodologies:

Housing Landscape 2014

Housing Landscape 2013 

 


[1] The 2013 ACS PUMS files use the redefined 2010 PUMAs which differ from the PUMAs in 2011 and earlier ACS PUMS files.   

[2] Metropolitan area definitions are consistent with those defined by the Office of Management and Budget in Revised Delineations of Metropolitan Statistical Areas, Micropolitan Statistical Areas, and Combined Statistical Areas, and Guidance on Uses of the Delineations of those Areas, OMB Bulletin No. 13-01, issued February 28, 2013 (available at http://www.whitehouse.gov/sites/default/files/omb/bulletins/2013/b-13-01.pdf) .

[3] Available at http://mcdc.missouri.edu/websas/geocorr12.html.

[4] Similar to the way HUD develops income limits for households of various sizes, median family income is used as the benchmark to which the income of a four-person household is compared. Incomes of larger households are compared to an upwardly adjusted median family income, and the benchmark for smaller households is adjusted downward. For a detailed description of the adjustments used by HUD and in this report, see p. 12 in HUD’s FY2014 HUD Income Limits Briefing Material, available at http://www.huduser.org/portal/datasets/il/il14/IncomeLimitsBriefingMaterial_FY14_v2.pdf. 

[5] Median family incomes for non-metropolitan areas in each state were derived from the household records classified as non-metropolitan in the PUMS files.

[6] Housing Landscape 2015 utilizes a methodology first implemented in Housing Landscape 2012.  In Housing Landscape 2011 and previous Housing Landscape reports, we adjusted income using the income adjustment variable (ADJINC) when calculating the housing-cost-to-income ratio (HCIR). The Census Bureau no longer advises making this adjustment and thus we have discontinued it. Consequently, the results from the current Housing Landscape report are not comparable to reports issued prior to Housing Landscape 2012.