The following is the conclusion of the working paper “The Demographics of Loan Delinquency: Tipping Points or the Tip of the Iceberg?” by William R. Emmons and Lowell Ricketts, both of the St. Louis Fed’s Center for Household Financial Stability. This working paper was prepared for a conference called “America’s Debt Problem: How Too Much Debt Is Hurting U.S. Household and Holding Back the U.S. Economy,” to be held Thursday in New York.
Demographic characteristics such as age, education, and race or ethnicity are correlated with a host of economic and financial behaviors and outcomes, including borrowing and delinquency rates. If demographic factors truly didn’t matter—if all individuals and families faced identical choices and opportunities in what we called a Demographics-Don’t-Matter (DDM) framework—then these correlations simply might reflect a connection between differences in risk preferences and differences in age, education, and race or ethnicity.
Groups that more frequently miss loan payments, for example, might just prefer “living close to the edge” financially so that a given economic or financial shock is more likely to represent a tipping point into delinquency. In principle, any member of a delinquency-prone demographic group could mimic the behavior of families that aren’t as likely to be delinquent; most simply choose not to do so.
An alternative approach to understanding demographic correlations with delinquency rates is to assume that all groups do not face identical choices and opportunities. Instead, for structural or historical reasons, some individuals and families may live with more risk than others.
Young people, for example, may face greater immediate financial challenges with higher stakes than do older people; those with low levels of education may face fewer economic opportunities than those with high skills; and black and Latino families may start their financial lives at an inherent disadvantage.
To proxy for these structural or systemic factors, we partition observable financial and nonfinancial choices and behaviors into peer-group effects and individual choices relative to those peer-group means. This is a Peer-Groups-Matter (PGM) framework.
Using Survey of Consumer Finances data for the period 1989-2013, we find limited support for the DDM hypothesis. In other words, if we assume that all families face the same opportunities and choices, then differences in a large set of observable financial and nonfinancial variables largely—but not completely—eliminate the role of demographic characteristics in predicting loan delinquency. Under this theoretical framework, groups with high delinquency risks have chosen to take on more risk; they could, if they chose, mimic low-risk groups and eliminate the gap in delinquency rates.
When we partition the potential explanatory variables into peer-group and idiosyncratic parts, we find strong support for a PGM hypothesis. Even after controlling for idiosyncratic choices of assets and liabilities, family structure and propensities to be lucky, a family’s stage of life and whether it is black remain highly significant predictors of delinquency. Delinquency-prone demographic groups may face greater risk for structural or systemic reasons rather than simply having a taste for more risk.
The implications for research and policy are clear. The DDM framework that dominates research and most credit-market policies receives limited empirical support. Even controlling for “good vs. bad choices and behavior,” being old is an advantage and being black is a disadvantage in credit markets.
The PGM framework that recognizes inherent differences in exposure to risk is strongly supported. Age and race matter; individual choices and behavior may not be enough to overcome them.
Research and policy that ignore these results subject some demographic groups to greater risk even while “blaming the victim.” Future research and policy could usefully explore how framing of choice and opportunity affect our conclusions about how credit and other markets actually do and should operate.