A More Effective Approach to Credit Card Debt Collection
© 2016 EPFL
Delinquent credit card accounts are a billion dollar problem for banks. One challenge is deciding what resources should be committed to chasing a debt that may never be recovered. Traditional debt scoring methods, used to help assess the probability of payment, have proved useful but not that effective. However, a new dynamic scoring approach combining existing account details, ongoing payment history, and economic data, offers a substantial improvement on returns.
High levels of consumer spending and debt often go hand-in-hand. Take the US, for example. During 2015 consumer spending in the US was roughly $11.3 trillion, while household debt at the end of that year was some $12 trillion. A significant proportion of consumer purchases are made on credit, US credit card balances in the last quarter of 2015 amounted to $714 billion. Unfortunately, though, not all credit cards balances are paid up on time. Credit card delinquency, where payments have been overdue for a specific period, has become a very expensive problem both for firms issuing credit cards, individuals unable to make card repayments, and the broader economy.
While credit providers have their methods of rating delinquent accounts and collecting outstanding payments these are often unsophisticated and ineffective. Now though, with their Dynamic Collectability Scoring methodology, Professor Thomas Weber at EPFL and Professor Naveed Chehrazi at McCombs School of Business, University of Texas, may have discovered a more effective approach to dealing with credit card delinquency.
Debt chasing
At any one time a significant percentage of credit card accounts have outstanding balances. In the US, for example, according to the Federal Reserve Bank's Quarterly Report on Household Debt and Credit, credit card balances in the third quarter of 2015 were $714 billion, of which about $125bn was outstanding for 30 days or more.
Usually, an internal collections team will try to collect outstanding payments. If the collections team is unsuccessful, after an agreed period of time, say 60 days, an overdue account will be declared delinquent by the issuer. The account may then be outsourced to a series of external collections agencies (involving multiple collection agencies fosters competition and improves the potential return) until the debt is either collected in part or whole, or written off.
Alternatively, if no payment is obtained, delinquent accounts can be bundled up, hived off, and sold to collection firms. Banks are able to unload nonperforming consumer loans from their balance sheets, while the collection firms can take a more proactive and aggressive approach to collecting outstanding balances. In total, a huge amount of debt is passed on to credit card collection agencies.
One challenge for the banks and collection agencies is deciding the level of resources that should be committed to chasing a particular credit card account. Many factors affect the likelihood of repayment being made on an outstanding debt. These include account specific data about the debtor such as legal status, other loans held, other credit lines, credit rating (FICO score), repayment history and demographic information. Macroeconomic indicators are also relevant, such as interest and inflation rates, GDP, and stock market performance -the idea being that if economic well-being is high, the likelihood of payments in the long run increases, and if economic circumstances are difficult, payment prospects decrease. And there is also the collection strategy and actions likely to be pursued to consider, from phone calls to court action.
Banks assign pre-computed scores to outstanding accounts that attempt to reflect these factors and provide an indication of the probability of collecting the debt. However, these static scoring methods are often poor at predicting repayment behavior. So when a bank approached Weber and Chehrazi for their assistance in optimizing its collection processes, they decided to adopt a different approach to scoring these delinquent accounts.
Their new scoring methodology builds on developments made in probability theory and statistical methods between the 1960s and early 2000s and relates to the analysis of random sequences of points correlated in time and space, such as clusters of lightning strikes, for example. In particular it relates to a mathematical construct called a self-exciting point process, where the occurrence of past events makes the occurrence of future events more probable, at least in the short run. Its use in finance is comparatively recent and in the context of credit collection completely new.
The major difference is that the new approach is dynamic. It focuses on two dimensions, in particular, balance and intensity: how much is owed and how likely it is that the next due repayment arrives. When a repayment is made the likelihood of another repayment increases. This is combined with the account specific data and macroeconomic factors to produce a Dynamic Collectability Score (DCS) that allows an employee of a bank or collection agency to calculate how much the account is worth - the value of the expected discounted repayment that you can get from the account - at any particular point in time.
A dynamic advantage
The DCS is useful for several reasons. The DCS provides a more accurate idea of the value of an account at a particular point in time. In doing so it helps collection professionals make better decisions about whether to walk away from the debt, or continue the collection effort, and if so what amount to settle for. It also allows banks to optimize commission rates paid to collections agencies.
At the same time it helps banks manage their regulatory risk management obligations under Basel II and III. The DCS can be used to more accurately assess loss given default (LGD) - the expected loss incurred in the event of a default - and enable banks to adjust their capital reserves to more closely reflect the real business risk.
And, by building a more accurate picture of bank payoffs in the event of default, this will inform risk underwriting decisions more generally. It will also allow banks to optimize their collection strategy to maximize expected returns from any account, an area of ongoing study that Weber and Chehrazi's most recent research focuses on.
The DCS approach is an exciting development in delinquent credit card debt management; a more effective method of predicting debtor behavior and the chances of recovering outstanding balances. It is about forty per cent more effective in its predictive power than the standard non-dynamic approach, say Weber and Chehrazi.
Consequently, by optimizing the collections process in this way, the model provides a means of increasing the asset quality of outstanding consumer loans, in terms of a higher expected return and a lower risk. Based on follow-on work (currently under review) by Chehrazi and Weber, together with Peter Glynn from Stanford University, the collection optimization can almost double the expected amount collected while at the same time cutting the risk by about 50%.
Next steps
Weber and Chehrazi are now engaged in further research projects on the topic and looking for banks and collection agencies to collaborate with. Given the potential of the DCS to improve the balance sheet, they are unlikely to be short of offers.