Vulnerability of low-income homeownership in the United States: An analysis based on household liquidity

ABSTRACT

In the United States after the Great Recession, despite growing attention to low-income households’ homeownership vulnerability, the existing works tend to take specific angles and produce only piecemeal evidence. By placing liquid assets’ function of mediating financial hardships in the context of homeownership dynamics, I establish a synthesized conceptual framework. Based on data from the Panel Study of Income Dynamics (PSID), I put this framework to a test and find that liquid assets not only reduce the risk of homeownership exit in general but play a pivotal role in accounting for low-income borrowers’ elevated rates of exit. I discuss policy implications of the findings at the end of the paper.

KEYWORDS:

• Homeownership
• income
• liquid assets
• financial security
• social policy

Acknowledgments

The author thanks Michael Sherraden for inspiring the research idea, three anonymous reviewers for thoughtful comments and suggestions, and JUA editors for the generous help with revising and editing the manuscript.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Notes

1. The present study is designed within the conceptual boundaries of the “ownership society” paradigm—that is, homeownership is assumed to be a desirable mode of investment in the existing U.S. socioeconomic system. However, it shall be pointed out that skepticism on this assumed desirability is growing, especially given the increasing affordability issues facing low-income and minority homeowners (Pattillo, 2013; Taylor, 2019). As such, some housing scholars have raised their voices on giving up on the “ownership society” ideal and proposing to transition toward an European mode of rental system. Please see, Brenner et al. (2012) and Pattillo (2013) for detailed discussions.

2. Researchers often rely on mortgage foreclosure/default to proxy for involuntary homeownership exit, but this approach is far from perfect. For one thing, it leaves out the vast number of forced home sales that did not go through the foreclosure/default process (Sharp & Hall, 2014). For another thing, as discussed above, driving forces behind involuntary homeownership exit (e.g., health care expenditure) go far beyond mortgage-market dynamics, affecting mortgage holders as well as non-mortgage holders. Thus, a certain degree of assumption is inevitable when it comes to measuring involuntary homeownership exit.

3. I produced an alternative sample with the missing assets information of in-between years (e.g., 1985) being estimated based on that of adjacent years (Sharp et al., 2020), although this sample is not preferred due to its rough measurement of assets variables. Nevertheless, the two versions lead to consistent findings. Results are available upon request.

4. The target population of this study is homeowners, and I do not consider the exclusion of non-owners as a sample selection bias. As explained above, homeowners are a specific population who have the need, the will, and the resources to pass mortgage scrutiny. As such, the relationship between homeowners and non-owners is fundamentally different than that of a sampling bias situation (e.g., housewives’ earnings had they entered labor markets).

5. The original sample includes 37,933 person-spells that represent all home-owning families with any observed financial assets. As shown in Appendix B1, the missingness is not random but occurs disproportionately among those of weak socioeconomic characteristics. That said, findings of this study shall be interpreted under the assumption that they would stay unchanged, had missing cases been included. My confidence in this assumption is strengthened by additional analyses reported in Appendix B2, which demonstrate that this study’s major conclusions hold consistently across different scenarios of sub-samples where the missingness mainly occurs.

6. The PSID purposely oversamples low-income families to improve the ability to analyze issues where their representation is low (Wilson et al., 2015). For that reason, many PSID-based homeownership studies do not apply weights to the sample (Berger et al., 2015; Boehm & Schlottmann, 2004, 2009; Hall & Crowder, 2011; Sharp & Hall, 2014; Sharp et al., 2020), which is my choice as well. Nevertheless, I produced a version of results based on the PSID’s nationally representative SRC sample, which leads to consistent findings. (Please see Appendix G.)

7. This will not be an issue under either of the following two circumstances. First, the predictor variable is not time-varying in nature. For example, Lee et al. (2017) used the initial household forming status (in public housing or market-rate housing) to predict subsequent residential trajectories. Second, the time-varying relationship between the predictor and outcome is highly consistent. For example, consider the relationship between education and income—more education is consistently associated with higher income over time. In contrast, an increase in liquid wealth could be either a positive (e.g., reception of a gift) or a negative (e.g., liquidation of other assets due to financial stress) sign for homeownership retention.

8. Although liquid assets should not be operationalized as a time-varying variable, they could be used, in a longitudinal context, as a predictor for the duration of homeownership spell. This type of so-called survival/event-history analyses require different conceptual framing and data structure (e.g., tracking a set of young homeowners over an extended period) and are not included in this study. Nevertheless, my examination of the PSID data shows that event-history analyses lead to essentially the same conclusion—it is household liquidity that determines a homeowner’s duration of ownership spell, and not income. Results are available upon request.

9. For example, since the income-to-repayment ratio is a fundamental component of mortgage evaluation, I do not see it as an important source of low-income homeownership vulnerability—its effect should have been well taken into account. Instead, I focus on factors that tend to be neglected or misevaluated by the market (i.e., liquid assets).

10. It shall be noted that the PSID defines disability as any physical or nervous condition. So, it does not necessarily equate to not working. Nevertheless, I tried multiple measures of employment status, including designating disability as an independent category, all of which lead to consistent findings.

11. Unlike regression models with a continuous dependent variable where the total variance is fixed and the residual variance is flexible, the logistic regression has the opposite pattern. Thus, in a logistic regression context that is featured by a flexible total variance, it is problematic to compare coefficients/odds ratios across nested models, as the amount of variance explained by a set of independent variables is not absolute but relative to the corresponding total variance associated with that specific model (Mood, 2010). The KHD method is one of the newly developed techniques to handle this issue. For a detailed explanation on the KHB method, please see, Karlson et al. (2012).

12. All analyses are performed using Stata 14. All data sets, syntax, Stata outputs, and additional analyses are available upon request.

13. All charts are created using the “mcp” command of Stata. Please see, Royston (2013) and Williams (2021) for detailed explanations.

Additional information

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Notes on contributors

Chunhui Ren

Chunhui Ren is a visiting assistant professor in the Department of Sociology at University of Cincinnati and a faculty associate in the Center for Social Development at Washington University in St. Louis. He has been conducting multiple research projects on institutionalized racial economic inequality, sustainable homeownership and wealth building, and advanced statistical modeling. His works have appeared in journals such as Demography, Social Forces, Social Science Research, and Urban Research & Practice.