Ebook Optimizing The Collections Process In Consumer Credit
Consumer credit has been little researched in banking and finance certainly when compared with the level of research directed at corporate credit. In particular there has been no modelling undertaken to improve the management of the operations involved in recovering some or all of the debt incurred when a borrower defaults. Yet the US sub prime mortgage crisis and its knock on effects has made bankers and ordinary citizens realise how important is consumer credit. One important issue which has arisen in consumer credit within the last few years, because of changes in regulation and the increase in the default rate on credit cards to above 5%, is how to maximize the amount recovered from consumer debts. This leads into how best to manage the recoveries and collection process, particularly for unsecured credit since the process in secured credit can be modeled by a two stage process – what is the chance the lender will have to take possession of the security and how much will the lender get for the security.
This paper seeks to model the collections process of unsecured consumer credit debt. In this process the collections department of the lender has a number of actions it can take to secure some repayment of the debt. These can range from telephone calls and gentle reminder letters to more formal letters, getting agreement to rescheduled repayment patterns with some repayment being made immediately. If this fails the collectors occasionally can make home visits but more generally seek legal redress involving court proceedings and the use of bailiffs. The questions are which actions should be taken and how long should a particular course of action be undertaken before trying another action. These problem can be modeled in two different ways. In the first case, the decisions are made each period in the light of the individual debtor’s repayment performance up to that point. Such decisions would be modeled using stochastic dynamic programming. In the second case, the decision made is which action perform next and how long to undertake it, and this depends only on the debtor’s repayment performance on previous actions and assumes a deterministic “average” recovery profile under the current action. This gives rise to a deterministic dynamic programme. This has the advantage that you can easily calculate what are the optimal actions that will be applied to the “average” debtor and what is the average cash flow of recoveries that will follow from these actions.
These are exactly the forecasts those who have to manage a collections process need to make prospectively when they are deciding the resources needed in the collections department and reporting to those who have to deal with provisioning ( setting aside money to cover lender expected losses in the future) how much of the bad debt they expect to recover. In the former case they want to know how many staff they will require and how many will need familiarity with the legal side of debt recoveries. So knowing which actions will be undertaken and how long on average they will last for the typical debtor makes such resource calculations possible. Similarly knowing the repayment pattern for the average debtor allows estimates of total cash flow from recoveries to be calculated. We concentrate on this second model in this paper. There has been very little analytic modeling of the collections process for any form of lending until the advent of the new Basel Accord which came into operation in 2007. This changed the way regulators determined how much capital banks have to hold to cover against credit risk. It required banks to estimate for each segment of their loan portfolio three quantities PD, the probability of default ( what proportion of the portfolio will default ), EAD the exposure at default ( the amount of money that could be defaulted upon within the segment) and, LGD, the loss given default ( the percentage of any default that is not recovered eventually), (Bennett et al 2005). The capital the banks have to set aside for the credit risk of this loan segment is then f(PD).EAD.LGD where f(.) is defined in the Basel regulations. LGD is closely related to recovery rate, RR, the percentage of the default amount that is recovered since LGD =1-RR. It is partly this regulatory emphasis on percentage of loss recovered that makes lenders’ collections departments measure their performance in terms of recovery rates rather than total amount recovered. They also use recovery rate because it is a measure which has the same bounds ( apparently 0 and 1) on all loans, and so it is easier to empirically estimate recovery rate distributions. Also, as most countries have limits on the amount of unsecured credit that can be offered, the default amounts of unsecured credit are not that different from one another. Empirically one can have recovery rates less than 0 if interest is charged on defaulted loans and nothing is repaid; recovery rates greater than 1 can occur if all that interest and the original default amount is paid off.
Previously there had been some work on estimating recovery rates in corporate lending since these affect the price of risky bonds. The edited book by Altman et al (Altman et al 2002) outlines, the mainly, regression based models that seek to relate recovery rates to economic factors and characteristics of the loan and the defaulter. The work on modeling the collections process for secured consumer lending (Lucas 2006) is directly motivated by Basel and so is more interested in estimating how much of the debt would be collected rather than optimizing the collections process.
For unsecured consumer credit, Matuszyk et al (Matuszyk et al 2007) have recognized that the recovery rate depends both on decisions by the lender as well as the uncertainty about the borrower’s ability and intention to repay. They though look at models which support the strategic level decision of whether to collect the debt in house, use an agent or sell off the debt. Makuch et al (Makuch et al 1992) addressed bad consumer credit management for General Electric Co in its GE Capital which provides credit card services, building a probabilistic account flow model to optimize resource allocation through linear programming. Otherwise the only analysis tends to be on very specific issues within the process such as how one could use text mining of the recorded conversation between the collector and the defaulter to identify whether the defaulter is likely to repay (Chin and Kotak 2006). The books by McNab and Wynn (McNab and Wunn 2000) and Anderson (Anderson 2007) describe the process and the sorts of actions that can be undertaken but do not model the process. In other areas there have been some attempts to model an operations process so as to optimize the outcome. One of the nearest to the work here is (Yu and Gittins 2008).
This looks at how many staff should be put in the different stages of a pharmaceutical Research and development operations. There the time of the different operations depends on the number of staff assigned to it. In our model the decisions are how long to run each different action for. In section two we introduce the model, while in section three we prove some results concerning the form of the optimal recovery rate and the optimal collections process under certain special conditions. Section four describes the case study of applying this model to real collections data, while the final section draws some conclusions and indicates how one could develop more detailed models of the collections process. These would be aimed at customizing the actions for each particular debtor whereas in this model we assume we are dealing with a homogeneous population.
Download
PDF Ebook Optimizing The Collections Process In Consumer Credit
Posted in :
