Subprime Consumer Credit Demand Evidence from a Lender's pricing E periment Sule alan College of Administrative Sciences and Economics, Koc University and CFAP Faculty of Business, Economics and Statistics, University of vienna and CEPR Using a unique panel data set from a U. K credit card company, we sensitivity of subprime credit card borrowers. In addition to all individual transactions and tails of a randomized interest rate experiment conducted by the lender on existing(inframarginal)loans For the whole sample, we estimate a statistically significant f3. 4 reduction in monthly credit demand in response to a five percentage point in interest rates. This aggregate response is small, but it masks very intere heterogeneity in the sample. We find that only low-risk borrowers who fully utilize their credit cards lower their credit demand significantly when faced with an increase rates. We also document that a five percentage point significant additional revenue for the lender without inducing delinquency over a short horizon. (JEL Dll, D12. D14) Borrowing rates affect firms'and households demand for credit. Quantifying such effects. that is estimating credit demand elasticities. has become an increasingly important academic endeavour. At the microlevel, lenders nterested in gauging these elasticities as an input to their optimal loan pricing strategies. At the macrolevel, knowledge of these elasticities is essential for understanding the transmission of monetary policy. Moreover, they can be We are grateful to the lender for sly providing us with the data We especially thank the lender's database ow, Nick Souleles and Philip Vermeulen, as well as participants of the Cambridge-Wharton conference, BCL- ECB conference cal Studies(IFS) and Bundesbank lysis and Policy( CFAP) and th ongyi Loranth, University of Vienna, 72 Bruenner Strasse 10 Vienna, Austria; telephone +431-4277-38052. E-mail: gyoengyi loranth@univie.acat. O The Author 2013. Published by Oxford University Press on behalf of The Society for Financial Studies. AllrightsreservedForPermissionspleasee-mail:journalspermissions@oup.com do:10.1093/rfs/hht029 Advance Access publication June 7, 2013 Downloadedfromhttps://academic.oupcam/rfs/article-abstract/26/9/2353/166253 e
[16:30 29/7/2013 RFS-hht029.tex] Page: 2353 2353–2374 Subprime Consumer Credit Demand: Evidence from a Lender’s Pricing Experiment Sule Alan College of Administrative Sciences and Economics, Koc University and CFAP Gyongyi Loranth Faculty of Business, Economics and Statistics, University of Vienna and CEPR Using a unique panel data set from a U.K. credit card company, we analyze the interest rate sensitivity of subprime credit card borrowers. In addition to all individual transactions and loan terms, we have access to details of a randomized interest rate experiment conducted by the lender on existing (inframarginal) loans. For the whole sample, we estimate a statistically significant £3.4 reduction in monthly credit demand in response to a five percentage point increase in interest rates. This aggregate response is small, but it masks very interesting heterogeneity in the sample. We find that only low-risk borrowers who fully utilize their credit cards lower their credit demand significantly when faced with an increase in interest rates. We also document that a five percentage point increase in interest rates generates significant additional revenue for the lender without inducing delinquency over a short horizon. (JEL D11, D12, D14) Borrowing rates affect firms’ and households’ demand for credit. Quantifying such effects, that is estimating credit demand elasticities, has become an increasingly important academic endeavour. At the microlevel, lenders are interested in gauging these elasticities as an input to their optimal loan pricing strategies. At the macrolevel, knowledge of these elasticities is essential for understanding the transmission of monetary policy. Moreover, they can be We are grateful to the lender for generously providing us with the data. We especially thank the lender’s database managers, who volunteered a great deal of assistance.We thank the editor (Alexander Ljungqvist) and the referees for their excellent comments. For comments and suggestions on the earlier versions, we are very grateful to Thomas Crossley, Alejandro Cunat, and Christian Laux. We also thank Orazio Attanasio, James Banks, Martin Browning, Russell Cooper, John Gathergood, Mark Jenkins, Soren Leth-Petersen, Valerie Lechene, Hamish Low, Nick Souleles and Philip Vermeulen, as well as participants of the Cambridge-Wharton conference, BCLECB conference, the EFA 2011 conference, and seminar participants at University of Cambridge, Koc University, University of Copenhagen, European University Institute, the Institute for Fiscal Studies (IFS), and BundesbankECB. This research is funded by the Cambridge Center for Financial Analysis and Policy (CFAP) and the University of Vienna. Send correspondence to Gyongyi Loranth, University of Vienna, 72 Bruenner Strasse, 1210 Vienna, Austria; telephone +431-4277-38052. E-mail: gyoengyi.loranth@univie.ac.at. © The Author 2013. Published by Oxford University Press on behalf of The Society for Financial Studies. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com. doi:10.1093/rfs/hht029 Advance Access publication June 7, 2013 Downloaded from https://academic.oup.com/rfs/article-abstract/26/9/2353/1662534 by Fudan University user on 14 December 2017
The Review of Financial Studies/v 26 n 9 2013 informative about whether households are credit constrained or whether they borrow responsibly and understand the basic credit terms offered to them. The latter point is particularly important because recent research documents low debt literacy and high financial vulnerability among a large number of households(see Lusardi and Tufano 2009). Such households are the primary concern of this paper. We estimate the sensitivity of credit demand to a large interest rate hike for individuals who are deemed to be subprime borrowers. We do this using a unique panel data set on credit card transactions from a private lender. Our lender serves only the subprime market in the United Kingdom. The strength of the paper relative to previous related studies is that we have access to a large exogenous change in interest rates. This variation is generated by the lender's randomized price experiment. To conduct the experiment, the lender classifies clients according to a behavior score that is designed to measure a clients riskness (low, medium, or high) and their utilization of credit cards (ow, medium, or high). This 3x3 classification produces nine"cells, and in five of these, the lender conducts a randomized experiment with a five percentage point interest rate increase. This setting not only allows us to dentify the causal effect of borrowing costs on credit demand for inframarginal loans but also gives us the opportunity to assess heterogeneity in treatment effects Subprime borrowers are commonly presumed to be credit constrained implying that they will not reduce borrowing in response to an interest rate increase. This argument lends itself to the conclusion that the interest rate increase necessarily leads to higher interest charges(revenue) for the lender and a faster debt accumulation for the borrowers. For the whole sample, we estimate a statistically significant f3.4 reduction in monthly credit demand in response to a five percentage point increase in interest rates. This aggregate response is small. We find no effect of the interest rate increase on the short-run probability of a client becoming delinquent. Together, the small reduction in monthly credit demand and the lack of an increase in delinquencies mean that the interest rate increase does lead to higher interest charges for the lender and no reduction in the stock of debt for the borrowers The finding that there is no reduction in debt despite the reduction in monthly credit demand is due to the fact that the increased interest rate applies to the entire stock of accumulated debt This overall pictur important heterogeneity the sample. We find credit demand reductions that are neither statistically nor economically different from zero among borrowers with high utilization rates and medium to high default risk. This is consistent with these particular borrowers being credit constrained. On the other hand, we estimate a ot disclose the name of the company. We will refer to it hereafter as the 2354 Downloadedfromhttps://academic.oupcam/rfs/article-abstract/26/9/2353/166253
[16:30 29/7/2013 RFS-hht029.tex] Page: 2354 2353–2374 The Review of Financial Studies / v 26 n 9 2013 informative about whether households are credit constrained or whether they borrow responsibly and understand the basic credit terms offered to them. The latter point is particularly important because recent research documents low debt literacy and high financial vulnerability among a large number of households (see Lusardi and Tufano 2009). Such households are the primary concern of this paper. We estimate the sensitivity of credit demand to a large interest rate hike for individuals who are deemed to be subprime borrowers. We do this using a unique panel data set on credit card transactions from a private lender. Our lender serves only the subprime market in the United Kingdom.1 The strength of the paper relative to previous related studies is that we have access to a large exogenous change in interest rates. This variation is generated by the lender’s randomized price experiment. To conduct the experiment, the lender classifies clients according to a behavior score that is designed to measure a client’s riskness (low, medium, or high) and their utilization of credit cards (low, medium, or high). This 3×3 classification produces nine “cells,” and in five of these, the lender conducts a randomized experiment with a five percentage point interest rate increase. This setting not only allows us to identify the causal effect of borrowing costs on credit demand for inframarginal loans but also gives us the opportunity to assess heterogeneity in treatment effects. Subprime borrowers are commonly presumed to be credit constrained, implying that they will not reduce borrowing in response to an interest rate increase. This argument lends itself to the conclusion that the interest rate increase necessarily leads to higher interest charges (revenue) for the lender and a faster debt accumulation for the borrowers. For the whole sample, we estimate a statistically significant £3.4 reduction in monthly credit demand in response to a five percentage point increase in interest rates. This aggregate response is small. We find no effect of the interest rate increase on the short-run probability of a client becoming delinquent. Together, the small reduction in monthly credit demand and the lack of an increase in delinquencies mean that the interest rate increase does lead to higher interest charges for the lender and no reduction in the stock of debt for the borrowers. The finding that there is no reduction in debt despite the reduction in monthly credit demand is due to the fact that the increased interest rate applies to the entire stock of accumulated debt. This overall picture does, however, mask some important heterogeneity in the sample. We find credit demand reductions that are neither statistically nor economically different from zero among borrowers with high utilization rates and medium to high default risk. This is consistent with these particular borrowers being credit constrained. On the other hand, we estimate a 1 For confidentiality reasons, we do not disclose the name of the company. We will refer to it hereafter as the “lender.” 2354 Downloaded from https://academic.oup.com/rfs/article-abstract/26/9/2353/1662534 by Fudan University user on 14 December 2017
Subprime Consumer Credit Demand statistically significant E9.0 reduction in monthly credit demand for borrowers with high utilization rates and low default risk. However, even for this group the response to the interest rate increase is not strong enough to lower the interest charges. In fact, despite their efforts, treated individuals in this group pay a 10%o higher interest charges relative to controls Borrowers with moderate utilization rate and "low default risk' also exhibit no sensitivity to higher interest rates. This is at first sight surprising because the unused borrowing capacity of individuals in this group suggests they are not credit constrained. These borrowers, however, had an increase in their credit limit just prior to the experiment. This increase make them appear to be borrowers who do not fully utilize their credit cards Hence, a potential interpretation of their behavior might be that the increase in credit limit relaxed their previously binding credit constraint. Their sitivity to the interest rate increase dir debt accumulation (E71, corresponding with a8.5% increase in total debt outstanding relative to the control) over three months following the interest rate Increase Estimating interest rate sensitivity of credit demand using survey data has been challenging for researchers This is because the cross-sectional variation in interest rates is likely to be endogenous to borrowing and repayment behaviors through unobservable characteristics of the borrowers Previous studies tried to overcome this challenge by using quasiexperimental designs. However, these research designs require strong identification assumptions. The experimental setting of our data gives us clean identification of credit demand elasticities without resorting to such assumptions. Moreover, the interest rate increase in ur data is substantial ( five percentage points)and the experimental sample size is large enough that we can be confident of detecting any economically Our study concerns a subset of households in a developed economy that are considered to be financially vulnerable. The U. K. credit market is a highly sophisticated market in which lenders have access to advanced risk pricing technologies. Such an environment allows access to formal credit (albeit at a high price) for households who would otherwise be rationed out. This access can provide insurance to temporary disruptions in households'income(such as 2 Attanasio, Goldberg, and Kyriazidou(2008)estimate interest rate elasticities of car loan demand by exploiting 四 nts for borowing rates. Adams. Eina v. and Levin (2oo9 se data on a u.s. privat sensitivity to loa tudy in which the authors find evidence of significant elasticity of credit card debt with respect to interest rates. Examples of identification assumptions include general exclusion restrictions for IV methods and common trend assumptions for difference-in-differences met 2355 Downloadedfromhttps://academic.oupcam/rfs/article-abstract/26/9/2353/166253 e
[16:30 29/7/2013 RFS-hht029.tex] Page: 2355 2353–2374 Subprime Consumer Credit Demand statistically significant £9.0 reduction in monthly credit demand for borrowers with high utilization rates and low default risk. However, even for this group, the response to the interest rate increase is not strong enough to lower the interest charges. In fact, despite their efforts, treated individuals in this group pay a 10% higher interest charges relative to controls. Borrowers with moderate utilization rate and “low default risk” also exhibit no sensitivity to higher interest rates. This is at first sight surprising because the unused borrowing capacity of individuals in this group suggests they are not credit constrained. These borrowers, however, had an increase in their credit limit just prior to the experiment. This increase makes them appear to be borrowers who do not fully utilize their credit cards. Hence, a potential interpretation of their behavior might be that the increase in credit limit relaxed their previously binding credit constraint. Their insensitivity to the interest rate increase directly translates into a significant debt accumulation (£71, corresponding with a 8.5% increase in total debt outstanding relative to the control) over three months following the interest rate increase. Estimating interest rate sensitivity of credit demand using survey data has been challenging for researchers. This is because the cross-sectional variation in interest rates is likely to be endogenous to borrowing and repayment behaviors through unobservable characteristics of the borrowers. Previous studies tried to overcome this challenge by using quasiexperimental designs.2 However, these research designs require strong identification assumptions. The experimental setting of our data gives us clean identification of credit demand elasticities without resorting to such assumptions.3 Moreover, the interest rate increase in our data is substantial (five percentage points) and the experimental sample size is large enough that we can be confident of detecting any economically significant effects. Our study concerns a subset of households in a developed economy that are considered to be financially vulnerable. The U.K. credit market is a highly sophisticated market in which lenders have access to advanced risk pricing technologies. Such an environment allows access to formal credit (albeit at a high price) for households who would otherwise be rationed out. This access can provide insurance to temporary disruptions in households’income (such as 2 Attanasio, Goldberg, and Kyriazidou (2008) estimate interest rate elasticities of car loan demand by exploiting the tax reform of 1986 in the United States. Alessie, Hochguertel, and Weber (2005) analyze the same issue using a similar design. Gross and Souleles (2002) use the U.S. Credit Bureau data and propose some firmspecific practices as instruments for borrowing rates. Adams, Einav, and Levin (2009) use data on a U.S. private subprime auto loan company. The general conclusion drawn from the studies is that there seems to be no sensitivity to borrowing rates among low-income households. However, such households display some sensitivity to loan features related to liquidity, such as down payment requirements, credit limits, and loan maturities. This finding is interpreted as the presence of binding liquidity constraints. The exception is the Gross and Souleles (2002) study in which the authors find evidence of significant elasticity of credit card debt with respect to interest rates. 3 Examples of identification assumptions include general exclusion restrictions for IV methods and common trend assumptions for difference-in-differences methods. 2355 Downloaded from https://academic.oup.com/rfs/article-abstract/26/9/2353/1662534 by Fudan University user on 14 December 2017
The Review of Financial Studies /v 26 n 2013 unemployment and sickness)and therefore can be beneficial. * However, access to high cost credit can pose a danger for financially fragile households if they borrow too much, relative to their means. The evidence reported in this paper provides novel insights on(1) the prevalence of liquidity constraints and(2) the mechanism of debt accumulation among subprime borrowers in developed economies. Such insights are critical to the development of public policy and consumer protection actions targeting financially vulnerable households in the United Kingdom and other developed economies From a policy point of view, the results illustrate(1) the vulnerability of subprime borrowers to interest rate increases and (2)that interest rate increases would be profitable for the lender for almost all types of borrowers studied.6 Whereas imposing interest rate caps might be an unpalatable option for a policy maker(because it could result in credit rationing), a range of other policy interventions might aid these individuals. These include restrictions on credit limit increases(particularly, limit increases initiated solely by the lender) and higher required minimum payments. A policy that requires lenders to fully explain and illustrate the consequences of higher interest rates on debt on might also be beneficial The rest of the paper is organized as follows. We provide a brief overview of the U. K. credit card market in the next section. In Section 2, we present our data and the experimental design. In Section 3, we motivate our outcome rariable and assess the magnitude of expected response to the experiment. We present and discuss the results in Section 4, and Section 5 concludes 1. Subprime Credit Card Market in the United Kingdom Credit cards have steadily grown in importance as a payment device in all industrialized countries. As of 2007, it is estimated that approximately 70 million credit cards were in issue in the United Kingdom. These cards were responsible for 22.4%o of the total consumer transactions, which stood at f540 billion in 2007(see Data monitor 2008). Moreover, borrowing on credit cards(revolving credit card debt from one month to the next and therefore incurring interest charges) has grown rapidly over the last few decades in the 4 Karlan and Zinman (2010) show that access to consumer credit even at very high rates can be beneficial.The ndomly assigned marginal loans produced significant net benefits for borrowers across a wide range of outcomes South afric emand elasticities in South Africa and Bangladesh, respectively. Karlan and maturity. Dehejia, ry, and Morduch(2012)estimate subtantial interest rate sensitivity among the poor. 6 A caveat applies to this result as the implications of the lender's profitability are based on short-run estimates. It is plausible that a permanent increase in interest rates has different long 2356 Downloadedfromhttps://academic.oupcam/rfs/article-abstract/26/9/2353/166253 e
[16:30 29/7/2013 RFS-hht029.tex] Page: 2356 2353–2374 The Review of Financial Studies / v 26 n 9 2013 unemployment and sickness) and therefore can be beneficial.4 However, access to high cost credit can pose a danger for financially fragile households if they borrow too much, relative to their means. The evidence reported in this paper provides novel insights on (1) the prevalence of liquidity constraints and (2) the mechanism of debt accumulation among subprime borrowers in developed economies. Such insights are critical to the development of public policy and consumer protection actions targeting financially vulnerable households in the United Kingdom and other developed economies.5 From a policy point of view, the results illustrate (1) the vulnerability of subprime borrowers to interest rate increases and (2) that interest rate increases would be profitable for the lender for almost all types of borrowers studied.6 Whereas imposing interest rate caps might be an unpalatable option for a policy maker (because it could result in credit rationing), a range of other policy interventions might aid these individuals. These include restrictions on credit limit increases (particularly, limit increases initiated solely by the lender) and higher required minimum payments. A policy that requires lenders to fully explain and illustrate the consequences of higher interest rates on debt accumulation might also be beneficial. The rest of the paper is organized as follows. We provide a brief overview of the U.K. credit card market in the next section. In Section 2, we present our data and the experimental design. In Section 3, we motivate our outcome variable and assess the magnitude of expected response to the experiment. We present and discuss the results in Section 4, and Section 5 concludes. 1. Subprime Credit Card Market in the United Kingdom Credit cards have steadily grown in importance as a payment device in all industrialized countries. As of 2007, it is estimated that approximately 70 million credit cards were in issue in the United Kingdom. These cards were responsible for 22.4% of the total consumer transactions, which stood at £540 billion in 2007 (see Data monitor 2008). Moreover, borrowing on credit cards (revolving credit card debt from one month to the next and therefore incurring interest charges) has grown rapidly over the last few decades in the 4 Karlan and Zinman (2010) show that access to consumer credit even at very high rates can be beneficial. The randomly assigned marginal loans produced significant net benefits for borrowers across a wide range of outcomes in South Africa. 5 The evidence on the credit elasticities of financially vulnarable households is limited, but there is a large body of academic literature on estimating credit elasticities in developing countries. Using a field experiement, Karlan and Zinman (2008) and, using between-branch variation, Dehejia, Montgomery, and Morduch (2012) provide evidence on the size of credit demand elasticities in South Africa and Bangladesh, respectively. Karlan and Zinman (2008) estimates modest interest rate sensitivity of the demand for new term loans in South Africa, with demand apparently more sensitive to loan maturity. Dehejia, Montgomery, and Morduch (2012) estimate subtantial interest rate sensitivity among the poor. 6 A caveat applies to this result as the implications of the lender’s profitability are based on short-run estimates. It is plausible that a permanent increase in interest rates has different long-run consequences, such as default or driving away clients. 2356 Downloaded from https://academic.oup.com/rfs/article-abstract/26/9/2353/1662534 by Fudan University user on 14 December 2017
Subprime Consumer Credit Demand United Kingdom attracting much attention from consumer protection groups, regulatory bodies, and, of course, the media. In 2007, total credit card debt stood at around f65 billion, representing approximately 30%o of consumer credit in the United Kingdom Consumers who are not considered suitable for unsecured credit by mainstream issuers comprise the U.K. nonstandard credit card market. The term"subprime "refers to a subsection of the nonstandard market in the United Kingdom. This subsection usually consists of individuals with adverse credit histories, that is, individuals with an even higher risk of default than the typical nonstandard individual Individuals deemed to be subprime borrowers are more difficult to evaluate in terms of default risk. This can be because of volatile income(e.g, many in this category are self-employed), low income (e.g unemployment ), lack of credit history in the United Kingdom, or impaired credit history due to past defaults or mortgage arrears. Therefore, lenders targeting this segment(such as our lender) invest heavily in advanced risk pricing technologies to combat the adverse effect of delinquencies and defaults The lenders randomized price experiments are part of its risk pricing practice 2. Data and Experimental Design Our data set is provided to us by a private credit card issuer, who is one of the major players in the subprime segment of the U. K. market. The data set comprises all individual transactions, including purchases, cash advances, payments, interest charges, and fees. We also have income, age, marital status, and home ownership reported by individuals at the application stage Our lender has routinely performed randomized interest rate experiments on ubsamples of clients since 2006. The main reason for these experiments is to establish sensitivity to interest rates as part of the companys risk pricing practice. Each experiment lasted between 3-6 months, and the lender initiated another experiment immediately following the previous one. Interest rate changes were permanent until the next change took effect. All interest rate experiments were designed based on ex-ante-determined blocks, which we will explain in greater detail later. The lender agreed to provide us with two of the experiments, called Phase 2 and Phase 6. As the later experiment, Phase 6, involved a much larger number of individuals and a much higher intensity of treatment(five percentage point increase in interest rates for all treated individuals), we chose to use these data. The experiment involved 39, 883 individuals. The randomization was done in November 2007, and the interest rate changes were communicated to the individuals allocated to treatment groups in January 2008. The interest rate changes were implemented 7 The main reason to fall into the subprime catagory is a County Court Judgement(CC))record. County Court judgement refers to an adverse ruling of the County Court against a person who has not satisfied debt payments with their creditors. An adverse ruling remains on the individuals record for six years from the date of judgement 2357 Downloadedfromhttps://academic.oupcam/rfs/article-abstract/26/9/2353/166253
[16:30 29/7/2013 RFS-hht029.tex] Page: 2357 2353–2374 Subprime Consumer Credit Demand United Kingdom attracting much attention from consumer protection groups, regulatory bodies, and, of course, the media. In 2007, total credit card debt stood at around £65 billion, representing approximately 30% of consumer credit in the United Kingdom. Consumers who are not considered suitable for unsecured credit by mainstream issuers comprise the U.K. nonstandard credit card market. The term “subprime” refers to a subsection of the nonstandard market in the United Kingdom. This subsection usually consists of individuals with adverse credit histories, that is, individuals with an even higher risk of default than the typical nonstandard individual. Individuals deemed to be subprime borrowers are more difficult to evaluate in terms of default risk. This can be because of volatile income (e.g, many in this category are self-employed), low income (e.g., unemployment), lack of credit history in the United Kingdom, or impaired credit history due to past defaults or mortgage arrears.7 Therefore, lenders targeting this segment (such as our lender) invest heavily in advanced risk pricing technologies to combat the adverse effect of delinquencies and defaults. The lender’s randomized price experiments are part of its risk pricing practice. 2. Data and Experimental Design Our data set is provided to us by a private credit card issuer, who is one of the major players in the subprime segment of the U.K. market. The data set comprises all individual transactions, including purchases, cash advances, payments, interest charges, and fees. We also have income, age, marital status, and home ownership reported by individuals at the application stage. Our lender has routinely performed randomized interest rate experiments on subsamples of clients since 2006. The main reason for these experiments is to establish sensitivity to interest rates as part of the company’s risk pricing practice. Each experiment lasted between 3–6 months, and the lender initiated another experiment immediately following the previous one. Interest rate changes were permanent until the next change took effect. All interest rate experiments were designed based on ex-ante-determined blocks, which we will explain in greater detail later. The lender agreed to provide us with two of the experiments, called Phase 2 and Phase 6. As the later experiment, Phase 6, involved a much larger number of individuals and a much higher intensity of treatment (five percentage point increase in interest rates for all treated individuals), we chose to use these data. The experiment involved 39,883 individuals. The randomization was done in November 2007, and the interest rate changes were communicated to the individuals allocated to treatment groups in January 2008. The interest rate changes were implemented 7 The main reason to fall into the subprime catagory is a County Court Judgement (CCJ) record. County Court Judgement refers to an adverse ruling of the County Court against a person who has not satisfied debt payments with their creditors. An adverse ruling remains on the individual’s record for six years from the date of judgement. 2357 Downloaded from https://academic.oup.com/rfs/article-abstract/26/9/2353/1662534 by Fudan University user on 14 December 2017
The Review of Financial Studies/v 26 n 9 2013 Descriptive statistics Mean Median tatement balance(E) 633 Debt(E 615.6 Interest rate(%) Income(E) 0.26604 15,910 41.0 Hom No other card (%) ation(November 2007). The total number of individuals is 39, 883. variables edit limit, interest rates, self-reported income, and age. The table also reports the mposition of the sample in terms of marital and employment status, home ownershi in February 2008. We have data until May 2008, so we can measure the effect of interest rate changes over the three months following the implementation, that is, from February to April 2008. We lose one month because of lagging for the construction of our outcome variables The experimental sample was not chosen from the lender's full client base Accounts that are flagged for reasons such as default, several months of delinquency, or inactivity are excluded before the selection of the sample. Furthermore the lender excluded individuals who have been with the lender for less than seven months at the time of the design. Table 1 presents the characteristics of the individuals in the sample at the time of the randomization (November 2007) The median income reported at the time of application is f15, 500. Given that the median individual income for the United Kingdom is about f19,000 individuals in our sample represent the lower end of the income distribution The average monthly utilization rate, defined as outstanding monthly balance divided by the credit limit, is about 79.4%o with the median value of 94.8%. The average utilization rate for all U.K. credit card borrowers is approximately 34%(see the Data monitor[2008]report). Interest rates and credit limits are the two other variables highlighting the differences between our average borrower versus the average U. K. borrower. The mean(median) interest rate is 30.9% 30.0% pa). These interest rates are significantly higher than the rates on typical U.K. credit cards(approximately 15%0-18% pa). The mean(median) credit limit is f1, 082(f1, 000), which is much lower than the average U. K credit card limit of f5. 129 in 2007 As Table I shows, the average monthly purchase value is about 576 with the median value of fo. It is worth drawing attention to the size of revolving debt in the table. This figure is calculated as the balance appearing on the 2358 Downloadedfromhttps://academic.oupcam/rfs/article-abstract/26/9/2353/166253
[16:30 29/7/2013 RFS-hht029.tex] Page: 2358 2353–2374 The Review of Financial Studies / v 26 n 9 2013 Table 1 Descriptive statistics Mean Median SD Utilization rate (%) 79.4 94.8 33.4 Statement balance (£) 848.7 726.6 633.4 Debt (£) 743.2 628.5 615.6 New transactions (£) 76.1 0.0 182.4 Credit limit (£) 1,182 1,000 796.8 Interest rate (%) 30.9 30.0 2.3 Income (£) 17,866 15,500 15,910 Age 42.1 41.0 11.8 Married (%) 56 – – Employed (%) 63 – – Self employed (%) 10 – – Home owner (%) 34 – – No other card (%) 42 – – The table presents the descriptive statistics of the individuals in the sample at the time of the randomization (November 2007). The total number of individuals is 39,883. Variables include utilization rate, statement balance, outstanding credit card debt, new transactions, credit limit, interest rates, self-reported income, and age. The table also reports the composition of the sample in terms of marital and employment status, home ownership, and ownership of other credit cards. in February 2008. We have data until May 2008, so we can measure the effect of interest rate changes over the three months following the implementation, that is, from February to April 2008. We lose one month because of lagging for the construction of our outcome variables. The experimental sample was not chosen from the lender’s full client base. Accounts that are flagged for reasons such as default, several months of delinquency, or inactivity are excluded before the selection of the sample. Furthermore, the lender excluded individuals who have been with the lender for less than seven months at the time of the design. Table 1 presents the characteristics of the individuals in the sample at the time of the randomization (November 2007). The median income reported at the time of application is £15,500. Given that the median individual income for the United Kingdom is about £19,000, individuals in our sample represent the lower end of the income distribution. The average monthly utilization rate, defined as outstanding monthly balance divided by the credit limit, is about 79.4% with the median value of 94.8%. The average utilization rate for all U.K. credit card borrowers is approximately 34% (see the Data monitor [2008] report). Interest rates and credit limits are the two other variables highlighting the differences between our average borrower versus the average U.K. borrower. The mean (median) interest rate is 30.9% pa (30.0% pa). These interest rates are significantly higher than the rates on typical U.K. credit cards (approximately 15%–18% pa). The mean (median) credit limit is £1,082 (£1,000), which is much lower than the average U.K. credit card limit of £5,129 in 2007. As Table 1 shows, the average monthly purchase value is about £76 with the median value of £0. It is worth drawing attention to the size of revolving debt in the table. This figure is calculated as the balance appearing on the 2358 Downloaded from https://academic.oup.com/rfs/article-abstract/26/9/2353/1662534 by Fudan University user on 14 December 2017
Subprime Consumer Credit Demand November 2007 statement minus the payments made toward that balance in the following month(December 2007). This is the debt revolved from November to December, to which the interest charge is applied. The mean revolving debt in November 2007 is approximately f743, with the median value of 5628. This is quite a large figure given a monthly interest rate of about 2.5%o. It is clear that a significant portion of the individuals in our data set use their card for borrowing purposes. To be precise, approximately 81% of the individuals in our sample revolved debt every month between November 2007 and April 2008 2.1 Experimental design erhaps the most intriguing feature of our data is that the lender had changed Its clients interest rates through randomized trials since 2006. They carried out the randomization as a block design in which a sample of individuals were assigned to blocks(cells, henceforth) defined by the interaction of utilization rates and an internally developed behavior score that summarizes individuals,risk characteristics. Individuals were allocated into cells according to their utilization rates and behavioral scores as of november 2007. After the allocation, the randomization was performed within cells. Such designs well known in the statistical, medical, and experimental economics literatures Simple randomization to treatment and controls is rarely employed in real randomized control trials for a number of reasons. For example, block designs reduce the variance of the experimental estimates(see, e.g., List, Sadoff, and Wagner[2011]or Duflo, Glennerster, and Kremer [2006]). This design implies that within cells, there is no selection problem, and conditional on cells, interest rate changes are exogenous o Table 2 presents the cell design, the sample sizes of each cell, and the number of individuals allocated into the treatment and control groups. In each cell, individuals in the treatment group(approximately 93.5% of the individuals) received a five percentage point increase in interest rates. For example, cell 1 contains individuals who had high utilization rates and low behavior scores high default risk) in November 2007. In this cell, 4, 319 individuals were allocated in the treatment group, whereas 280 individuals were in the control group. Similarly, cell 9 contains individuals who had low utilization rates and high behavior score(low default risk). In this cell, 4,030 individuals received a five percentage point increase in interest rates, whereas 276 individuals were in the control group. Note that the control size is quite small. However, as we show and discuss in the results section, these data give us a reasonable statistical power to estimate the economically significant impact. Note also that a 50 /50 llocation to treatment and control is not necessary and in general not optimal (see List, Sadoff, and Wagner 2011). For cells 2, 3, 5, and 6, the lender did 8 Internally developed credit scoring systems eatures of our lender' s scoring system, but we were informed practice for credit card issuers. We do not know the exact ed that it is a continously updated, multivariate orithm Downloadedfromhttps://academic.oupcam/rfs/article-abstract/26/9/2353/166253
[16:30 29/7/2013 RFS-hht029.tex] Page: 2359 2353–2374 Subprime Consumer Credit Demand November 2007 statement minus the payments made toward that balance in the following month (December 2007). This is the debt revolved from November to December, to which the interest charge is applied. The mean revolving debt in November 2007 is approximately £743, with the median value of £628. This is quite a large figure given a monthly interest rate of about 2.5%. It is clear that a significant portion of the individuals in our data set use their card for borrowing purposes. To be precise, approximately 81% of the individuals in our sample revolved debt every month between November 2007 and April 2008. 2.1 Experimental design Perhaps the most intriguing feature of our data is that the lender had changed its clients’ interest rates through randomized trials since 2006. They carried out the randomization as a block design in which a sample of individuals were assigned to blocks (cells, henceforth) defined by the interaction of utilization rates and an internally developed behavior score that summarizes individuals’risk characteristics.8 Individuals were allocated into cells according to their utilization rates and behavioral scores as of November 2007. After the allocation, the randomization was performed within cells. Such designs are well known in the statistical, medical, and experimental economics literatures. Simple randomization to treatment and controls is rarely employed in real randomized control trials for a number of reasons. For example, block designs reduce the variance of the experimental estimates (see, e.g., List, Sadoff, and Wagner [2011] or Duflo, Glennerster, and Kremer [2006]). This design implies that within cells, there is no selection problem, and conditional on cells, interest rate changes are exogenous. Table 2 presents the cell design, the sample sizes of each cell, and the number of individuals allocated into the treatment and control groups. In each cell, individuals in the treatment group (approximately 93.5% of the individuals) received a five percentage point increase in interest rates. For example, cell 1 contains individuals who had high utilization rates and low behavior scores (high default risk) in November 2007. In this cell, 4,319 individuals were allocated in the treatment group, whereas 280 individuals were in the control group. Similarly, cell 9 contains individuals who had low utilization rates and high behavior score (low default risk). In this cell, 4,030 individuals received a five percentage point increase in interest rates, whereas 276 individuals were in the control group. Note that the control size is quite small. However, as we show and discuss in the results section, these data give us a reasonable statistical power to estimate the economically significant impact. Note also that a 50/50 allocation to treatment and control is not necessary and in general not optimal (see List, Sadoff, and Wagner 2011). For cells 2, 3, 5, and 6, the lender did 8 Internally developed credit scoring systems are general practice for credit card issuers. We do not know the exact features of our lender’s scoring system, but we were informed that it is a continously updated, multivariateprobit-type algorithm. 2359 Downloaded from https://academic.oup.com/rfs/article-abstract/26/9/2353/1662534 by Fudan University user on 14 December 2017
The Review of Financial Studies /v 26 n 2013 LL1CELL T=5pp T=5p T=5pp igh #C=2804C=573#C=95 CELL 2 CELLS T=3.252 #C=0 CELL3 CELL6 CELL 9 137#=1,065#T=4030 C=0 #C=276 vior Score(Bscore cell design of the experiment, the sample sizes of each cell, and the numbere tis deont (r)and control( C)groups. The lender classifies individuals according the clients riskiness(low, medium, or high) and their redit cards(low, medium, or high). In each cell, individuals in the treatment group(approximatel individuals)received a five percentage point in not allocate individuals to a control group, making them unavailable for our purposes 2.2 Implementation Unlike studies using randomized field experiments(mainly in development economics), we were not involved in the design orimplementation of the experi- ment on which our analysis is based. Although randomized experiments are now standard practice among credit card companies and they have every incentive to implement them correctly, we need to make sure that the randomization was carried out properly to ensure the internal validity of our results. We perform mean equality tests on a range of variables including,our outcome variable. These tests are carried out using the variables measured in November 2007(the date of the randomization). Table 3 presents the means of tested variables for the treated and control. The p-values obtained from mean equality tests are displayed in parentheses. As shown in the table, we could not detect any statistically significant difference between the treated and control groups in any cell(as would be expected when randomization is carried out Even though the randomization was carried out properly, there may be ther challenges to the internal validity of our experimental estimates. Sample attrition, for example, would be of particular concern if it were caused by the treatment. This could happen if the treatment initiated delinquency nd eventually default, making the remaining treatment sample no longer comparable to the control sample(a dynamic selection problem). If the treatment caused some accounts to be charged off, our treatment effect estimates may be biased toward finding insensitivity to interest rates. Alternatively, if the 2360 Downloadedfromhttps://academic.oupcam/rfs/article-abstract/26/9/2353/166253
[16:30 29/7/2013 RFS-hht029.tex] Page: 2360 2353–2374 The Review of Financial Studies / v 26 n 9 2013 Table 2 Descriptive statistics 100% Utilization Rate CELL 1 CELL 4 CELL 7 High T= 5pp T= 5pp T= 5pp #T= 4319 #T= 8,072 #T= 14,418 #C= 280 #C= 573 #C= 995 Mid CELL 2 CELL 5 CELL 8 T= 5pp T= 5pp T= 5pp #T= 281 #T= 3,252 #T= 6469 #C= 0 #C= 0 #C= 451 Low CELL 3 CELL 6 CELL 9 T= 5pp T= 5pp T= 5pp #T= 137 #T= 1,065 #T= 4,030 #C= 0 #C= 0 #C= 276 0 Low Mid High Behavior Score (Bscore) The matrix presents the cell design of the experiment, the sample sizes of each cell, and the number of individuals allocated into the treatment (T) and control (C) groups. The lender classifies individuals according to a behavior score (Bscore) that is designed to measure the client’s riskiness (low, medium, or high) and their utilization of credit cards (low, medium,or high). In each cell, individuals in the treatment group (approximately 93.5% of the individuals) received a five percentage point increase in interest rates (T= 5pp). not allocate individuals to a control group, making them unavailable for our purposes. 2.2 Implementation Unlike studies using randomized field experiments (mainly in development economics), we were not involved in the design or implementation of the experiment on which our analysis is based.Although randomized experiments are now standard practice among credit card companies and they have every incentive to implement them correctly, we need to make sure that the randomization was carried out properly to ensure the internal validity of our results. We perform mean equality tests on a range of variables including, our outcome variable. These tests are carried out using the variables measured in November 2007 (the date of the randomization). Table 3 presents the means of tested variables for the treated and control. The p-values obtained from mean equality tests are displayed in parentheses. As shown in the table, we could not detect any statistically significant difference between the treated and control groups in any cell (as would be expected when randomization is carried out correctly). Even though the randomization was carried out properly, there may be other challenges to the internal validity of our experimental estimates. Sample attrition, for example, would be of particular concern if it were caused by the treatment. This could happen if the treatment initiated delinquency and eventually default, making the remaining treatment sample no longer comparable to the control sample (a dynamic selection problem). If the treatment caused some accounts to be charged off, our treatment effect estimates may be biased toward finding insensitivity to interest rates. Alternatively, if the 2360 Downloaded from https://academic.oup.com/rfs/article-abstract/26/9/2353/1662534 by Fudan University user on 14 December 2017
Subprime Consumer Credit Demand variable Cell 4 Cell 7 Cell 8 Cell 9 Utilization Sc 10410597998.194.594.550.950.75.06540 77 8558668168271771772672674073 0.15) Debt(n) 78478278778710151,01865666742340.7 Credit limit(E) 8117998678761,1721,1791.5551.5481,6631,642 IcY 36.238.186.190.6153.1153.4625 Interest rates So 31.331.431.431.330.730830.830.9 (064) Interest charges()20.220.021.021025.725.617217.25 (00) 1767817,929175431708917.79417,1051807117,24118.69418.,163 (0.18 Net new B(E) 5.1-240-19.9-23. 3.785.2045.552.0701464 The table shows the mean values for treatment and control for our variables of interest and control vanables in the month of randomization(November 2007).T, treatment; C, control. P-values for equality tests are in parentheses treatment caused voluntary closures of accounts, our treatment effect estimates nay be biased toward finding sensitivity. With respect to the latter, we find that no account was closed within the sample period. For the former, recall that we can follow outcomes of the experiment only for three months. It is unlikely that we would see any default in such a short period, as it takes six months for the lender to charge off the delinquent account. The lender stops charging interest on the outstanding debt after four months of delinquency(by law).The defaulted debt is transferred to the collection agency after six months. However, we can explore whether the treatment induced intention to default looking into the number of delinquent months following the treatment If the treatment induces default, we may observe it as delinquency(missed monthly payments), beginning with the date the interest rate increase was communicated (January 2008 ). For this, we investigate whether there is any statistically significant difference between the treated and control groups in terms of falling into a delinquency cycle after the communication of the interest rate increase. We do not reject the hypothesis of equality and conclude that the treatment did not induce delinquency within the sample period. We will return to the implications of this issue later in the text. rtunately, we do not face this problem in our sample: all accounts that are allocated into treatment receive the change in interest rates Downloadedfromhttps://academic.oupcam/rfs/article-abstract/26/9/2353/166253
[16:30 29/7/2013 RFS-hht029.tex] Page: 2361 2353–2374 Subprime Consumer Credit Demand Table 3 Internal validity checks Variable Cell 1 Cell 4 Cell 7 Cell 8 Cell 9 TCTCTCTCT C Utilization % 104 105 97.9 98.1 94.5 94.5 50.9 50.7 5.06 5.40 (0.24) (0.18) (0.98) (0.77) (0.52) Bscore 585 586 681 682 717 717 726 726 740 739 (0.92) (0.54) (0.63) (0.85) (0.15) Debt (£) 784 782 787 787 1,015 1,018 656 667 42.3 40.7 (0.96) (0.99) (0.88) (0.64) (0.88) Credit limit (£) 811 799 867 876 1,172 1,179 1,555 1,548 1,663 1,642 (0.72) (0.69) (0.75) (0.86) (0.73) New transactions (£) 16.7 12.2 36.2 38.1 86.1 90.6 153.1 153.4 62.5 60.2 (0.18) (0.78) (0.44) (0.98) (0.80) Interest rates % 31.3 31.4 31.4 31.3 30.7 30.8 30.8 30.9 30.2 30.3 (0.55) (0.77) (0.20) (0.26) (0.64) Interest charges (£) 20.2 20.0 21.0 21.0 25.7 25.6 17.2 17.2 5.4 4.4 (0.76) (0.99) (0.94) (0.97) (0.09) Income (£) 17,678 17,929 17,543 17,089 17,794 17,105 18,071 17,241 18,694 18,163 (0.82) (0.44) (0.24) (0.18) (0.68) Net new B. (£) −25.1 −24.0 −19.9 −23.1 3.78 5.20 45.5 52.0 7.01 4.64 (0.24) (0.49) (0.50) (0.40) (0.80) The table shows the mean values for treatment and control for our variables of interest and control variables in the month of randomization (November 2007). T, treatment; C, control. P-values for equality tests are in parentheses. treatment caused voluntary closures of accounts, our treatment effect estimates may be biased toward finding sensitivity. With respect to the latter, we find that no account was closed within the sample period. For the former, recall that we can follow outcomes of the experiment only for three months. It is unlikely that we would see any default in such a short period, as it takes six months for the lender to charge off the delinquent account. The lender stops charging interest on the outstanding debt after four months of delinquency (by law). The defaulted debt is transferred to the collection agency after six months. However, we can explore whether the treatment induced intention to default by looking into the number of delinquent months following the treatment. If the treatment induces default, we may observe it as delinquency (missed monthly payments), beginning with the date the interest rate increase was communicated (January 2008). For this, we investigate whether there is any statistically significant difference between the treated and control groups in terms of falling into a delinquency cycle after the communication of the interest rate increase. We do not reject the hypothesis of equality and conclude that the treatment did not induce delinquency within the sample period.9 We will return to the implications of this issue later in the text. 9 Another problem common in randomized experiments is noncompliance, that is, the possibility that units allocated to the treatment group are not treated. This situation could arise in our case if, for example, some individuals that are allocated to a treatment group objected to the interest rate increase and the lender consequently reversed the change. Fortunately, we do not face this problem in our sample; all accounts that are allocated into treatment groups did receive the change in interest rates. 2361 Downloaded from https://academic.oup.com/rfs/article-abstract/26/9/2353/1662534 by Fudan University user on 14 December 2017
The Review of Financial Studies /v 26 n 9 2015 3. Outcome Variable and Expected Effects We begin our analysis by first characterizing our borrowers using a standard intertemporal consumption framework. We do this to motivate our outcome variable and to generate testable hypotheses. To this end, we argue that individuals in our data set are very likely to be net borrowers. A net borrower who is not liquidity constrained is expected to lower his credit demand(his consumption) when faced with an increase in borrowing rate, because this increase implies an increase in the price of today's consumption(a substitution effect). The subsequent decline in consumption is reinforced by the fact that the individual is now lifetime poorer as he carries forward a stock of debt(an income effect). On the other hand, a borrower who is constrained by his credit limit is not likely to change his consumption following a(small) interest rate change but he may react to large changes In this section, we outline our measure of credit card borrowing We then ddress the following question: What should be the expected response of an individual in our sample when faced with a five percentage point increase in his borrowing rate? An important feature of credit card debt is that changes in interest rates apply to the stock of debt hence when faced with an increase in interest rate, the individual,s debt automatically increases due to the additional interest charges. For this reason, we cannot use the stock of credit card debt as a choice variable. The actual monthly credit demand for a credit card user is monthly purchases on credit minus the subsequent payments made toward the outstanding balance. This difference constitutes the monthly ddition to the existing credit card debt that accrues interest; thus, it forms our outcome variable. We call our outcome variable"net new borrowing, NNB. and we define it as NNB+1=NTt+1-Pi+1 (1) where NTLi+ is new transactions made on credit between month t and 1+1 and Pi+1 is the payment made toward the outstanding balance at t +1. NTi. t+1 is interest exempt between period t and t+1, whereas if NT,1+1-Pi+1>0,the difference accrues interest charges until paid. Therefore, a positive(negative) value for NNB indicates an increase(decrease)in monthly credit demand. We expect the unconstrained borrower to reduce NNB when faced with an increase in borrowing rates I1 This is not the case for cash advances that are included in NT, in which case the interest charges resume as soon 12 For most credit card products, monthly payment Pi +1 is subject to P+12 Markk Br,时 2362 Downloadedfromhttps://academic.oupcam/rfs/article-abstract/26/9/2353/166253
[16:30 29/7/2013 RFS-hht029.tex] Page: 2362 2353–2374 The Review of Financial Studies / v 26 n 9 2013 3. Outcome Variable and Expected Effects We begin our analysis by first characterizing our borrowers using a standard intertemporal consumption framework. We do this to motivate our outcome variable and to generate testable hypotheses. To this end, we argue that individuals in our data set are very likely to be net borrowers. A net borrower who is not liquidity constrained is expected to lower his credit demand (his consumption) when faced with an increase in borrowing rate, because this increase implies an increase in the price of today’s consumption (a substitution effect). The subsequent decline in consumption is reinforced by the fact that the individual is now lifetime poorer as he carries forward a stock of debt (an income effect). On the other hand, a borrower who is constrained by his credit limit is not likely to change his consumption following a (small) interest rate change but he may react to large changes.10 In this section, we outline our measure of credit card borrowing. We then address the following question: What should be the expected response of an individual in our sample when faced with a five percentage point increase in his borrowing rate? An important feature of credit card debt is that changes in interest rates apply to the existing stock of debt. Hence, when faced with an increase in interest rate, the individual’s debt automatically increases due to the additional interest charges. For this reason, we cannot use the stock of credit card debt as a choice variable. The actual monthly credit demand for a credit card user is monthly purchases on credit minus the subsequent payments made toward the outstanding balance. This difference constitutes the monthly addition to the existing credit card debt that accrues interest; thus, it forms our outcome variable. We call our outcome variable “net new borrowing,” NNB, and we define it as NNBt+1 =NTt,t+1−Pt+1, (1) where NTt,t+1 is new transactions made on credit between month t and t +1, and Pt+1 is the payment made toward the outstanding balance at t +1. NTt,t+1 is interest exempt between period t and t +1,11 whereas if NTt,t+1−Pt+1 >0, the difference accrues interest charges until paid. Therefore, a positive (negative) value for NNB indicates an increase (decrease) in monthly credit demand. We expect the unconstrained borrower to reduce NNB when faced with an increase in borrowing rates.12 10 Note that the latter prediction refers to the strongest definition of liquidity constraints for which there is an actual quantity limit to borrowing. One can also extend the notion of liquidity constraint to individuals who face increasing borrowing cost with quantity demanded as in Pissarides (1978). 11 This is not the case for cash advances that are included in NT, in which case the interest charges resume as soon as the cash advance is made. 12 For most credit card products, monthly payment Pt+1 is subject to Pt+1 ≥Max[κBt,θ], 2362 Downloaded from https://academic.oup.com/rfs/article-abstract/26/9/2353/1662534 by Fudan University user on 14 December 2017