The relationship between the housing and labor market crises and doubling up: an MSA-level analysis, 2005–2011
In addition, a few previous studies have sought to explicitly link changes in the housing market and/or the labor market during the recent crisis to changes in household formation at the MSA level.16 For example, Timothy Dunne used MSA-level data on people ages 18 to 34 to investigate the correlation between household formation (e.g., headship rates17 and number of households) and labor market conditions and the correlation between household formation and housing prices.18 Although he found that doubling up is associated to some extent with both a weak housing market and a weak labor market, he did not probe further. Gary Painter also used data on 80 MSAs from the American Community Survey (ACS) for the period 2005–2008 to examine changes in household headship, homeownership, and overcrowding within a dwelling for MSAs grouped by immigrant status.19 His study found that headship rates and overcrowding rose while homeownership declined for all groups examined, although to differing extents.
Finally, it is worthwhile pointing to a few studies that have used MSA-level data to investigate associations between the housing crisis and other outcomes. Lisa Dettling and Melissa Kearney, for instance, used MSA-level data to examine the relationship between variations in housing prices and fertility during the recent crisis.20 In addition, a number of studies have looked at spillover and contagion effects of the foreclosure crisis that accompanied the burst of the housing bubble.21
To sum up, this article builds upon earlier studies of subnational housing and labor markets and documents geographic differences in the timing and severity of the housing and labor market crises of the late 2000s. It then takes a further step and examines associations between these crises and household formation.
Data and methodology
The analysis uses data on 353 MSAs, with a focus on 12 MSAs, for the period from 2005 to 2010 (and, where data were available, to 2011).22 The rationale for the selection of these specific MSAs is discussed shortly. Table 1 summarizes the key indicators for household formation and housing and labor market conditions, along with their data sources. Labor market conditions are principally measured using BLS data on nonfarm payroll employment.23 Overall conditions in the housing market are measured using the FHFA House Price Index for single-family units.24 In the analysis, the index is set to 100, with 2005 as the base year. These data are available at the quarterly level through 2011. Previous studies examining changes in overall housing conditions across MSAs have similarly used these data or relied on a smaller set of MSAs included in the Case–Shiller indices.25
|Household formation and homeownership|
|Size of household||ACS|
Nonrelatives in family households (percent)
|Nonrelatives living in family households as a percentage of total people in family households||ACS|
Unmarried households (percent)
|Unmarried (opposite-sex) partner households as a percentage of total households||ACS|
Homeownership rate (percent)
|Owner-occupied households as a percentage of total households||ACS|
|Housing market conditions|
House Price Index (2005 = 100)
|Prices for single-family-unit transactions serviced by Fannie Mae or Freddie Mac||FHFA|
Foreclosure rate (percent)
|(number of foreclosures / number of loans) x 100; measure is akin to a default rate||CoreLogic|
|Labor market conditions|
|Nonfarm payroll employment, age 16 and over||BLS|
|Rate calculated for civilian noninstitutional population, age 16 and over||BLS|
Some attention is also focused on foreclosures, given the acute impact they had on particular housing submarkets (e.g., subprime lending). Foreclosures are measured using proprietary data obtained from CoreLogic, which includes 85 percent of foreclosures and first lien loans.26 CoreLogic defines a foreclosure as a situation in which an owner’s right to a property is terminated, usually because of default. The foreclosure rate is calculated here as foreclosures per number of loans (multiplied by 100). Foreclosure data are available at the monthly level, although in many of the analyses they are aggregated to the quarterly (or annual) level for comparisons with data from other sources. For all nonannual data, seasonal adjustment is undertaken using a locally weighted regression.27
20 Lisa J. Dettling and Melissa S. Kearney, “House prices and birth rates: the impact of the real estate market on the decision to have a baby,” working paper 17485 (Cambridge, MA, National Bureau of Economic Research, October 2011).
21 For example, see Atif R. Mian, Amir Sufi, and Francesco Trebbi, “Foreclosures, house prices, and the real economy,” working paper 16685 (Cambridge, MA, National Bureau of Economic Research, January 2011); William H. Rogers, “Declining foreclosure neighborhood effects over time,” Housing Policy Debate 20, September 2010, pp. 687–706; William H. Rogers and William Winter, “The impact of foreclosures on neighboring housing sales,” Journal of Real Estate Research 31, no. 4, 2009, pp. 455–479; and John P. Harding, Eric Rosenblatt, and Vincent W. Yao, “The contagion effect of foreclosed properties,” Journal of Urban Economics 66, 2009, pp. 164–178.
22 The FHFA House Price Index series includes a total of 367 MSAs (metropolitan, micropolitan, and metro divisions). For this article, the figure is slightly reduced to 353 because of data availability for selected years in two other databases: CoreLogic and the annual ACS.
23 The advantage of looking at employment rather than unemployment is that unemployed status depends on job-search activity. Only figure 1 provides information on both employment and the unemployment rate.
24 FHFA constant-quality housing prices are constructed using a repeat-sales method (similar to the Case-Shiller index), where sales observations are not included unless the housing unit has been sold at least twice in the data series. One characteristic of a repeat-sales index is that previously published values are subject to change over time, as existing housing units are sold a second time. The present analysis uses data collected from FHFA in the summer of 2012. Thus, the rates of change calculated here may not precisely match reported rates in current or future FHFA publications.
25 See Cohen et al., “The boom and bust of U.S. housing prices;” and Sinai, “House price moments in boom–bust cycles.”
26 For a comparison with other sources of foreclosure data, including RealtyTrac, see “Sources of foreclosure data,” MBA Research DataNotes (Mortgage Bankers Association, 2008), http://www.mortgagebankers.org/files/Research/July2008SourcesofForeclosureData.pdf. The report by the Mortgage Bankers Association points to an important drawback of the oft-used RealtyTrac measure: the measure is subject to double counting. Although RealtyTrac only counts one filing per month (even if more than one occurs), if filings for the same property occur in different months, RealtyTrac counts all such events.
27 Robert B. Cleveland, William S. Cleveland, Jean E. McRae, and Irma Terpenning, “STL: a seasonal-trend decomposition procedure based on loess,” Journal of Official Statistics 6, 1990, pp. 3–73.