Responsible Lending and Affordability. An Experian white paper Simon Harben, Bureau Analytics, Decision Analytics

Transcription

1 Responsible Lending and Affordability An Experian white paper Simon Harben, Bureau Analytics, Decision Analytics September 2008

2

3 Executive Summary There is more pressure than ever on the Credit Industry to practice responsible lending in all its dealings with consumers. In the UK this pressure is not only coming from the industry s governing bodies and consumer support groups, it is coming increasingly from the investment community and the UK Government as well. Although this paper describes a UK-based responsible lending approach, the lessons learned from the UK have implications for many other developed - and developing - consumer credit markets. This paper starts by examining recent trends in indebtedness and provides an overview of the main factors driving the responsible lending debate in the UK. It then describes how an automated responsible lending solution can be delivered using a new generic mechanism for estimating disposable income and assessing consumer affordability. Illustrations of how this new responsible lending mechanism works are provided for both mortgage lending and unsecured lending in the prime sector. Responsible lending in the unsecured non-prime sector is also discussed. The ability to deliver truly automated responsible lending decisions has implications for all consumer credit markets and some recommendations are given for how this approach could be applied outside the UK. Responsible Lending & Affordability

4

5 Contents 1. Introduction Insolvency and bankruptcy - the UK context The regulators view 4 2. Responsible lending and the role of credit scoring 5 3. Assessing disposable income and affordability Estimating disposable income Assessing affordability Automated responsible lending 6 4. Responsible lending in the mortgage sector Mortgage quotations Mortgage applications 9 5. Responsible lending in the unsecured prime sector Responsible lending in the unsecured non-prime sector Implications for application form design Assessing affordability outside the UK Estimating disposable income The Affordability Index Appendices 14 9.A Appendix A - The Consumer Indebtedness Index 14 9.B Appendix B - Using Experian s affordability metrics 15 9.C Appendix C - Credit rating questionnaire 16 9.D Appendix D - Maximum expected income algorithm About the author About Experian 18 Responsible Lending & Affordability 3

6 1. Introduction 1.1 Insolvency and bankruptcy - the UK context The number of people experiencing severe financial difficulties in the UK has grown significantly over recent years. Figure 1 shows the increase in insolvencies in the consumer sector in England and Wales since Although these numbers now seem to have peaked, there are still 25,000 individual insolvencies in England and Wales every 3 months. Coupled with this, the number of UK properties being repossessed jumped to 18,900 (up 48% year-onyear) in the first half of And these numbers only reflect a fraction of the UK s consumer debt problem. For example, the Citizens Advice Bureau deals with well over 1million consumers with debt problems every year. But what are the root causes of the UK s consumer debt problem? It is now widely accepted that one of the main contributing factors has been the number of consumers that have over borrowed - particularly on unsecured credit. As a consequence of this, the UK credit industry is now looking to place even more emphasis on affordability when lending to consumers. And this means that there is a pressing need for a means of delivering this capability into the highly automated systems that are now making the vast majority of consumer lending decisions in the UK. 1.2 The regulators view Over the last 2-3 years there has been a greatly increased focus on responsible lending from the consumer credit industry s regulatory bodies. The Office of Fair Trading (OFT) has a clear remit in this area from the 2006 Consumer Credit Act (CCA). Under the new CCA, the OFT now looks at business practices that appear to be:... deceitful or aggressive or otherwise unfair or improper (whether unlawful of not). And the OFT considers irresponsible lending to be a species of this:... the business Figure 1 The number of individual insolvencies in England and Wales (seasonally adjusted). Source: Insolvency Service practices which the OFT may consider to be deceitful or aggressive (etc.)... include practices in the carrying on of a consumer credit business that appear to the OFT to involve irresponsible lending. The 2006 CCA also includes an unfair relationships test that can now be applied to assess whether a lender has treated a borrower unfairly. And, according to the Department for Business, Enterprise and Regulatory Reform (BERR), fairness considerations include a borrower s age, experience and financial capability. The Financial Services Authority (FSA), which regulates the provision of financial services in the UK, is also pushing for a more responsible approach to consumer lending, particularly in the mortgage sector: (1) A firm must be able to show that before deciding to enter into, or making a further advance on, a regulated mortgage contract, or home purchase plan, account was taken of the customer s ability to repay. (2) A mortgage lender must make an adequate record to demonstrate that it has taken account of the customer s ability to repay for each regulated mortgage contract that it enters into and each further advance that it provides on a regulated mortgage contract. The record must be retained for a year from the date at which the regulated mortgage contract is entered into or the further advance is provided. (Mortgage & Home Finance: Conduct of Business Sourcebook) And the British Banker s Association (BBA) has introduced a new Banking code that also commits to delivering more responsible lending practices: Before we lend you any money or increase your overdraft, or other borrowing, we will assess whether we feel you will be able to repay it. At the same time there is also government recognition that borrowers need to be responsible too meaning that they have to be realistic about what level of borrowing they can support. To this end, government has stepped up its efforts to educate consumers about credit and financial management. But this has to be viewed as a longer term solution, and, in the meantime, there is clearly increasing internal and external pressure on UK lenders to not only lend responsibly but also to be seen to be doing so. 4

7 2. Responsible lending and the role of credit scoring Credit scoring has been used successfully for many years for assessing the creditworthiness of new applicants for credit. For the majority of lenders, one of the key benefits of credit scoring was that it could remove the need for a long drawn out income and affordability check, greatly speeding up the time to process each credit application. But credit scoring has tended to focus on establishing an individual s propensity to pay rather than their ability to pay, and while credit scoring does deliver many of the elements required for responsible lending, it also has some limitations: Credit scoring is good at identifying individuals that are likely to experience financial difficulties in the future e.g. young people living in rented accommodation that have a short time in employment. Credit scoring is also good at identifying consumers that are already having repayment problems e.g. consumers with accounts in arrears on the Credit Bureau. More recently, scoring models have also been used to identify consumers that are keeping up with their credit commitments, but that are actually highly indebted. (Experian s Consumer Indebtedness Index was developed to do just this - see Appendix A). But, because credit scoring has no specific income dimension, consumers with negative affordability can still be given additional credit based on their credit score alone. All of which implies that lenders need to incorporate a reliable affordability check into their decision-making processes to complement their credit scores and policy rules. And, while leaving the affordability assessment to an underwriter to make subjectively is one option, it is rarely going to be the best solution. Credit scoring has tended to focus on establishing propensity to pay rather than their ability to pay. Responsible Lending & Affordability 5

8 3. Assessing disposable income and affordability 3.1 Estimating disposable income Experian has been working on an algorithm for estimating Effective Disposable Income (EDI) for some time. The traditional approach to calculating a monthly EDI has been to combine the following components: Net Monthly Income (NMI) [supplied via application form] Monthly Mortgage/Rent [from Credit Bureau or application form or modelled] Monthly Credit Commitments [from Credit Bureau] Monthly Expenditure [modelled] In the UK, monthly expenditure (and monthly rent) models can be developed using data from the UK Government s Office for National Statistics (ONS) Expenditure & Food Survey. However, for highly credit active households in particular, there is inevitably a high degree of overlap between monthly expenditure and monthly credit commitments. This, in turn, can lead to significant underestimates of EDI for these credit active cases - which are just the cases that require a reliable EDI to enable their affordability to be assessed accurately. Following a good deal of research into this problem, Experian has developed an heuristic approach that allows for this effect by amalgamating monthly mortgage or rent, monthly credit commitments (MCC) and monthly expenditure (MEX) into Monthly Outgoings, along the following lines: Monthly Outgoings = Monthly Mortgage/Rent + ƒ (MCC, MEX) where the function, ƒ, is directly related to household indebtedness. Effective Disposable Income can then be calculated using: EDI = Net Monthly Income - Monthly Outgoings Section 4 provides some illustrations of how this new EDI approach works in practice. 3.2 Assessing affordability While the EDI provides a useful new tool for new business decisionmaking - particularly in the secured lending sector - it does not appear to provide a general predictor of credit risk. This is probably not that surprising since income information rarely, if ever, makes a significant contribution to a credit scorecard. However, a useful new risk predictor has been produced by adding key application form information and other financial metrics to the Consumer Indebtedness model (ref. Appendix A). On the model development sample, the resulting affordability score had a Gini Coefficient of 72 - a 12% increase over the Indebtedness score. To make it more useable, the affordability score was converted into an Affordability Index (AI) in the range 1-99, as follows: 1 - Very low affordability (high risk) Very high affordability (very low risk) Section 5 illustrates how the AI can be used in practice. 3.3 Automated responsible lending The availability of a reliable EDI and AI means that whenever income is supplied as part of the credit application decision-making process, this process can now incorporate an automated affordability check, along the following lines: Apply primary credit decision criteria: Calculate the applicant s credit score, based on application form and Credit Bureau information (including an indicator of consumer indebtedness when available); Apply policy decline rules (e.g. 1 or more court judgements or previous defaults) and a low score cut-off; Decline cases with a low Affordability Index; Refer cases for identity authentication or fraud checks; Apply other refer rules to cases with a moderate AI (or low EDI); Accept all other cases. Appendix B provides more information on the practical use of Experian s Affordability Metrics within a lender s application processing environment. 6

9 4. Responsible lending in the mortgage sector Responsible lending is - more than ever - a key issue for all mortgage providers. And a key element of responsible lending is the ability to assess an individual s disposable income. Experian s new EDI calculation was developed primarily for use in the mortgage sector, specifically to replace the traditional income multiples approach. 4.1 Mortgage quotations Income multiples have been used for many years in the mortgage industry to give an indication of how much to lend. The standard multiples used to be around 3.5 times gross annual income for single applicants and 2.75 for joint applicants. But, as UK house prices have continued to rise, income multiples have risen to stay in step. However, an increasing number of UK lenders are now looking to move to affordability based systems. These reflect an individual s income and expenditure, which, of course, is also the aim of the EDI calculation described in Section 3.1. As long as a Credit Bureau quotation search takes place as part of the mortgage quotation process, only the following personal information has to be supplied to calculate an EDI: Main and joint applicant gross annual incomes Main and joint applicant ages Marital status Number of dependants This allows Monthly Expenditure to be generated from the MEX model set, so Monthly Outgoings can then be calculated as described in Section 3.1 (with Monthly Mortgage/ Rent set to zero): Monthly Outgoings = ƒ (MCC, MEX) With Net Monthly Income generated from the stated Gross Annual Income(s), EDI can also be calculated using: EDI = Net Monthly Income - Monthly Outgoings Based on the relevant monthly interest rate (R), the mortgage term in months (T) and the EDI, a maximum (repayment) mortgage amount (A) can then be calculated using the following equation: A = EDI [(1 + R) T - 1] / [R (1 + R) T ], subject to the Lender s maximum multiple amount. If a mortgage amount has been requested as part of the quotation, then this can be agreed if it is reasonably close to A. If no actual amount has been requested, then A can be used in the mortgage quotation, provided that it is above a realistic threshold level. The following figures compare maximum mortgage amounts calculated based on income multiples and EDI s for some different applicant profiles (based on a 25 year term and an APR of 6%). Responsible Lending & Affordability 7

10 Figure 2 Maximum mortgage amounts for one adult with no dependants Figure 3 Maximum mortgage amounts for one adult with two dependants 8

11 Figure 4 Maximum mortgage amounts for two adults with no dependants Figure 5 Maximum mortgage amounts for two adults with two dependants For the cases with no dependants, the EDI-based mortgage amounts are equivalent to or exceed the traditional multiple amounts for incomes of 25k+. At the other end of the scale, families with 2 (or more) dependants have much less disposable income and generally have lower EDI-based mortgage amounts as a result. Experian has also derived a Credit Rating Questionnaire that can be used to generate an accurate credit risk assessment as part of the mortgage quotation process without a full Credit Bureau search being made - Appendix C provides details. 4.2 Mortgage applications A similar approach to that described in Section 4.1 can be taken for mortgage applications, with a full Credit Bureau Search being made in this case. The resulting EDI (and the Affordability Index) can then be used in the final mortgage application decision along the lines described in Section 3.3. Responsible Lending & Affordability 9

12 5. Responsible lending in the unsecured prime sector To ensure that the EDI calculation is as accurate as possible in this scenario, the following personal details are required: Main and joint applicant (or partner s) incomes. (If partner s income is not requested for couples then the EDI calculation is clearly less reliable in this case.) Marital status Number of dependants Monthly mortgage / rent - this can be estimated if it is not asked as part of the credit application process. Again, Net Monthly Income is calculated and Monthly Expenditure is generated from the MEX model set. The Credit Bureau search supplies MCC to complete the dataset for the full EDI calculation: EDI = NMI - MMR - ƒ (MCC, MEX) The AI can then be calculated and cross-tabulated with the existing application score, as Figure 6 illustrates: Figure 6 Cross-tabulating the Affordability index with the existing application score The main use of the AI is then to identify the two groups highlighted: High risk accepts that should be declined; Low risk refers that should be auto-accepted. Using this approach on a recent sample of (prime) personal loan applications produced the following results: Increase in declines of 2% Bad debt reduction of 12.5% Reduction in referrals of 27% 10

13 6. Responsible lending in the unsecured non-prime sector It is probably not surprising that the Estimated Disposable Income calculation described in Section 3.1 generates very few cases with a positive EDI for households with gross annual incomes of less than 15k. In fact, the whole concept of responsible lending is very difficult to apply quantitatively to low income families. As an illustration of this, the latest Expenditure & Food survey shows clearly that average monthly expenditure regularly exceeds average monthly income for lower income families with 2 or more children: This, allied to the fact that up to 50% of sub-prime borrowers only have state benefits as a source of income, makes it impossible to develop an automated responsible lending solution that is sensitive enough for the non-prime sector. Instead, the existing - much more qualitative - practices of using face-to-face interviews and/or home visits to assess affordability need to continue, and in some cases may need to be tightened in light of the UK s continuing consumer debt problem. Table 1 Average monthly household expenditure. Source: ONS (EFS) and Experian The concept of responsible lending is very difficult to apply quantitively to low income families. Responsible Lending & Affordability 11

14 7. Implications for application form design Clearly, a reliable calculation of disposable monthly income depends largely on the information going into it. But, at the moment, there can be a disconnect between the household level EDI calculation and the personal details requested on a credit application form. Partner/joint applicant income information enables the disposable income calculation to accurately represent the complete household position. Age and employment type/status are used in Experian s new Maximum Expected Income algorithm to check that the supplied Incomes are not unrealistic - see Appendix D for details. To make the EDI calculation as accurate as possible, the following information should always be requested at the point of application: Main applicant details Age Gross annual (or net monthly) income Marital status Number of dependants Employment type/status Time in employment Accommodation status (non-mortgage applications only) Monthly mortgage / rent (non-mortgage applications only) Partner (or joint applicant) details Age Income Employment type/status 12

15 8. Assessing affordability outside the UK The problem of UK consumer indebtedness has been widely publicised, but it is certainly not solely a UK problem. The following provides some general guidelines for how the EDI/AI approach can be applied in other markets where there is an increasing consumer indebtedness problem. 8.1 Estimating disposable income Although the form of the EDI calculation described in Section 3.1 should have fairly general application, its components may have to be derived differently in other markets. For example: 8.2 The Affordability Index In the UK, the AI was developed by adding key application form data and other financial metrics to the existing Consumer Indebtedness model (ref. Section 3.2) This approach should work equally well outside the UK, but only in markets that share both positive and negative credit account information via the local Credit Bureau(s). Net monthly (household) income Applicants income is generally supplied on their credit application form, but household income should also be requested, particularly in developing markets where the household is very likely to take collective responsibility for a debt. Monthly mortgage/rent This can be derived from the local Credit Bureau if it holds mortgage data, or it can be modelled based on national survey data. The simplest approach is probably just to ask this on the application form, although the answer is not always going to be completely reliable. Monthly credit commitments This is definitely more reliable if it is derived independently via the local Credit Bureau, but it too can be obtained from the application form if it is not available otherwise. Monthly expenditure This is also best derived independently because it is notoriously difficult to obtain an accurate estimate of expenditure from consumers themselves. National survey data is used to estimate this in the UK primarily as a function of Income and household composition. A similar approach is recommended elsewhere even if this means representatives of the local credit industry producing and distributing their own survey specifically for this purpose. The approach should work equally well outside the UK, but only in markets that share both positive and negative credit information. Responsible Lending & Affordability 13

16 9. Appendices A: The Consumer Indebtedness Index Experian has conducted a good deal of research into assessing consumer indebtedness over the last 6 years. This has mainly been aimed at answering the following question: Because total unsecured debt is not a good predictor of risk (and therefore measure of indebtedness) on its own, what are the main factors that should be used to measure an individual s indebtedness? Experian s analysis has focussed on credit applications from individuals that were showing no signs of payment difficultly, but that were fairly heavy users of credit. The ultimate aim of the analysis was to derive a statistical model of Probability (Good) based on a set of factors that were specific to this type of credit user, with P(Good) measuring the probability that an individual will continue to make all their credit repayments over the following 2 years. The main focus of Experian s indebtedness research was then to look for new ways of combining the Experian credit bureau data available on these individuals into a meaningful set of predictors of P(Good). The P(Good) model that results from this approach works extremely well on the target group of consumers, with an uplift in predictive power of over 30%. Table 2 shows how this Indebtedness score predicts the likelihood of repayment problems on these apparently good consumers. To summarise, the best predictors of indebtedness were found to be: Table 2 The indebtedness score predicts the likelihood of repayment problems on apparently good consumers. Source: Experian By contrast, Total Unsecured lending (TUL) is much flatter (i.e. has very limited predictive power) as the following Table shows: Table 3 Total Unsecured Lending is more limited in terms of predictive power Source: Experian The number of active credit accounts in use. Compared to a single account with a balance of 20,000, finding 20 accounts each with a balance of 1,000 was found to be a much better measure of true indebtedness. The number of revolving credit accounts in use and the limit utilisation on these accounts. Given the relatively high cost of credit on most revolving accounts, these types of indicator were very predictive of an individual s indebtedness. The type of neighbourhood the consumer lives in. When an individual lives in a neighbourhood with lots of previous debt problems, it s very likely that they, too, are going to struggle with their debts. Experian converts the resulting Indebtedness Score into a Consumer Indebtedness Index (CII), so that an individual s CII directly corresponds to the likelihood that they will experience credit repayment problems as a result of their current level of indebtedness. The CII is now widely used as an additional credit risk assessment tool throughout the UK consumer credit industry. In time it is anticipated that widespread use of the CII should minimise the number of highly indebted cases that are given additional credit. Indeed, the Government s Advisory Group on Overindebtedness cited the CII development as a key partnership action for enabling responsible lending. 14

17 B: Using Experian s affordability metrics Experian s latest set of Affordability metrics (i.e. EDI & AI) are available as a standalone Strategy Management module. This can then be embedded within any UK lender s application process flow, along the following lines: Alternatively, the algorithm and scoring models that form the basis of the EDI and AI calculations can be coded within a lender s own Strategy Management system or within any other business rules engine. Responsible Lending & Affordability 15

18 C: Credit rating questionnaire This has been derived by Experian to produce a reliable Credit Rating for use when a Credit Bureau search cannot be taken e.g. as part of a mortgage quotation process. A very accurate Credit Rating from 1 (low) to 5 (high) can be generated based on (truthful) answers to the following questions: How old are you? What is your employment status? (Full-time; Part-time etc) How many credit accounts have you applied for in the last 6 months? How many credit accounts do you have on which you owe money? Have you missed payments on any of your credit accounts in the last 6 months? (No; Only 1; 2 or more etc) In the last 6 years have you defaulted on a credit account, incurred a CCJ or had an IVA or been made bankrupt? Are you on this year s Electoral Roll at your current address? Would you say that most of your credit accounts have been opened in the last 18 months? Do you pay the full balance on all your credit and store cards every month? What is your Postcode? A final Credit Rating (from 1 to 5) is then generated based on a credit scoring model that scores the responses. D: Maximum expected income algorithm As part of its responsible lending research, Experian has developed a mechanism for checking whether or not a stated income is realistic by calculating a Maximum Expected Income (MEI). This mechanism estimates Gross Annual Income (GAI) using two new regression models, which were developed on the following types of credit applicant: Age & Stated Gross Annual Income 30k+ Age 31+ & Stated Gross Annual Income 30k+ Both models are based on a combination of application form and Credit Bureau characteristics. The MEI is then calculated using GAI, as follows: MEI = MAX (GAI * 1.2, GAI ) (In practice, the MEI can be adjusted further within a lender s own decisioning system to achieve their target referral rate precisely.) The MEI algorithm returns a maximum expected Gross Annual or Net Monthly Income depending on the income type supplied. And, as long as the key information is supplied for each applicant, MEI is calculated for both Main and Joint Applicants. 16

19 10. About the author Simon Harben graduated from Manchester University with an MSc in Statistics in In 1984 he was a founder member of the scoring and consultancy team that formed the nucleus for Experian Decision Analytics. Simon is now Head of Bureau Analytics within Experian Decision Analytics, and is closely involved with all new analytics developments within EDA. He has particular responsibility for the development of Experian UK s bureau-based credit, fraud and identity scoring models. For more information about Experian s responsible lending and Affordability Solutions, please contact your Experian Account Director. Responsible Lending & Affordability 17

20 11. About Experian Experian is a global leader in providing information, analytical and marketing services to organisations and consumers to help manage the risk and reward of commercial and financial decisions. Combining its unique information tools and deep understanding of individuals, markets and economies, Experian partners with organisations around the world to establish and strengthen customer relationships and provide their businesses with competitive advantage. For consumers, Experian delivers critical information that enables them to make financial and purchasing decisions with greater control and confidence. Clients include organisations from financial services, retail and catalogue, telecommunications, utilities, media, insurance, automotive, leisure, e- commerce, manufacturing, property and government sectors. Experian plc is listed on the London Stock Exchange (EXPN) and is a constituent of the FTSE 100 index. Experian has corporate headquarters in Dublin, Ireland and has operational headquarters in Costa Mesa, California and Nottingham, UK. The Group employs approximately 15,500 people in 38 countries worldwide, supporting clients in over 65 countries around the world. Continuing sales for the year ended 31 March 2008 were $4,059m ( 2,020m / 2,858m). For more information, visit The word Experian is a registered trademark in the EU and other countries and is owned by Experian Ltd and/or its associated companies. About Experian s Decision Analytics division Decision Analytics is the international division of Experian specialising in providing credit risk and fraud management consulting services and products. Over more than 30 years, it has developed its best practice analytical, consulting and product capabilities to support organisations to manage and optimise risk; prevent, detect and reduce fraud; meet regulatory obligations; and gain operational efficiencies throughout the customer relationship. With clients in more than 60 countries and offices in more than 30, the Decision Analytics division of Experian delivers experience and expertise developed from working with national and international organisations around the world across a wide range of industries and business size. For more information, visit the company s website on 18

21

22 Experian The word EXPERIAN and the graphical device are trade marks of Experian and/or its associated companies and may be registered in the EU, USA and other countries. The graphical device is a registered Community design in the EU. All rights reserved.