High-cost debt and perceived creditworthiness: Evidence from the UK

https://doi.org/10.1016/j.jfineco.2021.07.005Get rights and content

Abstract

We show that high-cost debt exacerbates financial constraints by affecting lenders’ perception of credit risk. Using data from a high-cost lender in the UK, we show that high-cost credit reduces applicants’ credit score and future bank credit even though it does not affect future debt repayment. These effects are not present among borrowers who are already tagged as high risk at application, consistent with high-cost credit affecting lenders’ beliefs about borrowers’ creditworthiness. The results highlight a novel channel through which high-cost credit can harm consumers’ financial health: a self-reinforcing stigma of high risk.

Introduction

“Some lenders might see the fact that you’ve taken out a payday loan as a sign that your finances are under pressure.”

— James Jones, Head of Consumer Affairs, Experian UK.

Credit cards, bank overdrafts, and payday loans are relatively high interest rate sources of short-term credit. High-cost credit attracts the attention of academics and regulators because of its potentially harmful effects on the welfare of financially unsophisticated borrowers. Understanding the mechanisms through which the use of high-cost credit can affect future financial health is essential to evaluate the consequences and desirability of policy proposals to regulate it.

This paper proposes and provides evidence of a novel mechanism through which the use of high-cost credit can affect borrowers’ financial health. If the typical high-cost borrower has a relatively high risk of default, then lenders will pool borrowers who use high-cost credit with other high-cost, poor credit risk borrowers, independent of their actual risk of default. As a consequence of this mechanism, which we refer to as the perceived creditworthiness channel, the cost of borrowing will increase and access to credit will suffer.1

Providing evidence of the effect of high-cost credit on lenders’ perception of creditworthiness is a challenging task. Lenders will correctly revise downwards the expected creditworthiness of a borrower who takes up a high-cost loan if riskier borrowers self-select into high-cost credit or if high-cost credit causes more default. The empirical challenge addressed in this paper is to demonstrate that high-cost credit affects access to credit through its direct effect on lenders’ perceptions, even when alternative channels of influence that affect access to credit indirectly are possible.

To address this challenge, we measure and compare the effects of the use of high-cost credit across the distribution of borrowers’ ex-ante credit scores, our measure of creditworthiness as perceived by lenders. We show that taking up a high-cost loan causes an immediate and persistent drop in credit scores, reduced access to standard sources of credit, and no increase in default rates among borrowers with relatively higher credit scores. On the other hand, take-up of a high-cost loan does not affect the credit score or access to credit of borrowers who have a low credit score at the time of application and who are more exposed to the indirect effects of high-cost credit (e.g., more vulnerable to liquidity shocks or more prone to strategic default). We conclude that the drop in the credit scores of higher-score borrowers is evidence of the perceived creditworthiness mechanism. Although potentially counterintuitive at first sight, the rationale behind this heterogeneity is well grounded in theory, since it is a direct implication of Bayesian updating of lenders’ prior beliefs about borrower risk.

We combine the credit bureau data of all applicants, approved and rejected, to a high-cost lender in the UK (the “Lender”) with two research designs. The first research design, which measures the impact of take-up on borrowers with relatively higher credit scores, exploits the fact that applicants to the Lender are assigned quasi-randomly to loan officers of different systematic propensity to approve loans-different “leniency”-within a branch. We measure loan officer leniency using leave-one-out fixed effects and use it as an instrument for loan take-up, an approach similar to that used in measuring the procontinuation attitude of bankruptcy judges (see, e.g., Chang, Schoar, 2008, Dobbie, Song, 2015; and Bernstein et al., 2019). The second design, which is aimed at obtaining precise estimates of the causal effect of take-up on low-score borrowers, is a fuzzy regression discontinuity design (RDD) around the minimum credit score eligibility threshold imposed by the Lender for loan approval.

We first show that taking up a high-cost loan causes a 24-point reduction (off a mean of 539) in borrowers’ credit score within the same quarter of application but only among relatively high-score applicants. This decline in credit score persists for at least four quarters. For low-score borrowers, taking up a high-cost loan has no effect on borrowers’ perceived creditworthiness at any horizon. The evidence is consistent with Bayesian updating when take-up reflects information that differs from the credit bureau’s prior: only borrowers with sufficiently high initial credit scores experience a decline in their perceived creditworthiness. This Bayesian argument is not unique to our setting: for example, Dobbie et al. (2020) find that the removal of the bankruptcy flag from credit records in the US has a larger positive impact on the credit scores of individuals who are more creditworthy based on their demographic characteristics.

Next we show that taking up a high-cost loan reduces the number of bank credit lines by 24% four quarters after application. This decline is not driven by a reduced demand for credit, as take-up causes an increase in the average number of credit searches, which serves as a proxy for credit applications to standard lenders, by 170%. Borrowers with initially low perceived creditworthiness, on the other hand, neither search more for, nor are less likely to, obtain credit from banks following the take-up of high-cost debt. This result highlights the trade-off faced by high-cost borrowers with a relatively high perceived creditworthiness: alleviate short-term financial needs at the cost of constraining access to standard sources of financing. This result is also consistent with previous literature that shows lenders’ response to credit bureau data (e.g., see Musto, 2004, Liberman, 2016, Dobbie, Goldsmith-Pinkham, Mahoney, Song, 2020).

Finally, we evaluate whether the heterogeneous effects of take-up on perceived creditworthiness, search, and credit across the distribution of credit scores reflect heterogeneous causal effects on true borrower risk. Ex ante this is unlikely, as borrowers with low credit scores are likely to be financially vulnerable, meaning that repaying high interest rates should imply a higher burden on the household’s resources (see Skiba, Tobacman, 2019, Gathergood, Guttman-Kenney, Hunt, 2019). In addition, borrowers with low credit scores have a lower reputation cost from walking away without repaying, so moral hazard and strategic default should be more pronounced for them than for borrowers with higher scores (for evidence on moral hazard, see Karlan, Zinman, 2009, Dobbie, Skiba, 2013). In our setting, simple correlations are consistent with this view. For example, conditional on take-up, default on the Lender’s loan is negatively correlated with credit score. And indeed, we find that taking up a high-cost loan reduces the default probability of relatively higher-score borrowers in the quarter of application across all categories of credit, as reported by the credit bureau. In contrast, estimates for low-score borrowers show an increase, although insignificant, in the default rate.

Our results can be rationalized by a model of heterogeneous true credit risks within credit scores in the population. Conditional on their credit score, applicants to a high-cost lender are riskier than average, and this difference in true risk increases with credit scores. That is, while high-score applicants to a high-cost lender are likely to be riskier than non-applicants, low-score applicants and non-applicants are equally risky. Because applications to a high-cost lender are imperfectly or not at all observed by market participants, credit bureaus only update negatively the credit scores of borrowers, which affects future access to standard sources of credit even though borrowers and applicants have a similar true risk of default ex post. For example, according to aggregate data shared with us by the credit scoring agency, less than 60% of applicants to any high-cost credit provider in the UK follow up by taking up the loan. Since loan approval is not public information, lenders (and credit bureaus) cannot distinguish between applicants who were rejected by the lender (a potentially very bad signal) and those who were approved but subsequently opted to not take up the loan (a potentially positive signal). And since applicants may not be fully aware of the loan terms at the time of applying, not taking up a loan after seeing the terms is not necessarily a sign of poor credit quality.

The perceived creditworthiness mechanism can be self-reinforcing, as drops in credit scores restrict access to credit and lead to further drops in creditworthiness. In addition, it is profit-maximizing for lenders to use the high-cost loan flag in the credit history to make inferences about the borrower’s willingness or ability to repay. As a result, high-cost borrowing can cause poverty traps with negative long-term implications on consumer welfare even when borrowers understand the trade-offs involved in the use of high-cost credit (e.g., Manso, 2013 makes this point theoretically for the feedback effect of corporate credit ratings, and, more generally, a large body of theoretical work following Arrow, 1973 rationalizes statistical discrimination as an equilibrium with self-confirming beliefs). Further, our results suggest that the use of credit scores that are calibrated for the entire population of borrowers may hinder financial inclusion, particularly among relatively poorer applicants. These implications can be mitigated with policies that drop certain negative flags from borrowers’ past credit history (see, e.g., Bos, Nakamura, 2014, Liberman, Neilson, Opazo, Zimmerman, 2018).

Our work contributes to an active literature that studies the effects of high-cost credit on financial and economic outcomes (see, e.g., Morse, 2011, Melzer, 2011, Bhutta, Skiba, Tobacman, 2015, Zaki, Gathergood, Guttman-Kenney, Hunt, 2019). We contribute to this literature by showing that high-cost credit may be costly even when it does not causally affect repayment. Moreover, our results imply that studies that measure the consequences of high-cost credit use for populations with ex-ante low credit scores, who are less likely to face the perceived creditworthiness mechanism (see, e.g., Gathergood et al., 2019), may capture a lower bound on the average effect on financial health. Our work also contributes to a literature on the information environment in consumer credit markets, which focuses on how information sharing may affect the equilibrium amount of lending while remaining silent on the specific mechanisms (e.g., Djankov, McLiesh, Shleifer, 2007, Jappelli, Pagano, 2002, De Janvry, McIntosh, Sadoulet, 2010, Liberman, Neilson, Opazo, Zimmerman, 2018). In a related paper, Garmaise and Natividad (2017) show empirically that changes in firms’ perceived creditworthiness affect firms’ default. Our results highlight a novel channel through which information sharing institutions may affect the allocation of credit among financially vulnerable households. Our findings are thus relevant for understanding the consequences of changes in the information sharing environment, such as the Consumer Financial Protection Bureau’s recent proposal to require lenders in the US payday credit market to share and use information from credit agencies.2

Section snippets

Empirical setting

The Lender is based in England and provides small short-term loans to subprime borrowers. Business is conducted through a chain of retail stores staffed by loan officers. Since the available loan products are prepackaged combinations of amount, rate, and maturity, loan officers can only influence the extensive margin: they decide whether or not to grant a loan. Loan officers have full discretion in the approval process for first-time applicants, and they are encouraged to use their judgment in

Empirical strategy

We start this section by developing a simple framework that provides the intuition behind our empirical strategy. Then we explain the two research designs, the leniency instrumental variable (IV) and the RDD, which allow us to overcome the endogeneity of loan take-up on financial outcomes.

Heterogeneous effects of high-cost debt

In this section we present the first two steps of our analysis. First, we verify that take-up of high-cost credit reduces credit scores, except for individuals with already low scores. Next, we verify the heterogeneous effects of high-cost debt on usage of and search for credit.

Additional evidence

Our main hypothesis is that when the use of high-cost credit is observable by lenders and credit bureaus, using high-cost credit is a negative signal that affects perceived creditworthiness and future outcomes. However, high-cost credit use may affect the incidence of other credit events, such as default, that also affect perceived creditworthiness (due to moral hazard or burden of repayment). We posit that, consistent with theory, this indirect channel should be stronger for the more

Conclusion

This paper highlights a novel mechanism through which the use of high-cost credit may affect borrowers’ future access to credit, which we denote as the perceived creditworthiness mechanism. Borrowers who take up a high-cost loan suffer an immediate decline in their credit rating that cannot be explained by their repayment behavior. After a year, borrowers search more for bank credit but have fewer bank credit accounts. By looking at borrowers who already have a poor credit score, we show that

References (32)

  • J.D. Angrist et al.

    Wanna get away? Regression discontinuity estimation of exam school effects away from the cutoff

    J. Am. Stat. Assoc.

    (2015)
  • K. Arrow

    The theory of discrimination

  • S. Bernstein et al.

    Asset allocation in bankruptcy

    J. Finance

    (2019)
  • N. Bhutta et al.

    Payday loan choices and consequences

    J. Money Credit Bank.

    (2015)
  • M. Bos et al.

    Should Defaults be Forgotten? Evidence from Variation in Removal of Negative Consumer credit information

    Unpublished working paper

    (2014)
  • S. Calonico et al.

    Robust nonparametric confidence intervals for regression-discontinuity designs

    Econometrica

    (2014)
  • Cited by (7)

    View all citing articles on Scopus

    Toni Whited was the editor for this article. Andres Liberman is at Betterfly; email: [email protected]. Daniel Paravisini is at London School of Economics; email: [email protected]. Vikram Pathania is at University of Sussex; email: [email protected]. This paper previously circulated with the title “High-Cost Debt and Borrower Reputation: Evidence from the UK” We thank the Lender and the Credit Bureau for providing the data. We thank the editor and an anonymous referee, Neil Bhutta, Jason Donaldson, Paul Goldsmith-Pinkham, Don Morgan, Adair Morse, Justin Murfin, Giorgia Piacentino, Adriano Rampini, Kelly Shue, Jeremy Tobacman, Ansgar Walther, and workshop participants at numerous seminars and conferences for useful comments and discussion. All errors and omissions are ours only.

    View full text