Logical Postulates of Consumer Credit Risk: Credit Tractatus Logico-Philosophicus

 

By Dhruv Sharma, Arlington VA based Independent Scholar; and George Overstreet, UVA.

 

This article discusses credit risk from the viewpoint of a set of logical postulates for credit risk.  To date logic based credit risk theory is lacking. This article is a step toward this direction.

 

The following is a series of self-evident logic based assertions about consumer credit risk in the spirit of Wittgenstein’s Tractatus Logico-Philosophicus.  This is a theoretical and logical piece based on the fundamentals of cash flow as a driver of consumer behavior and credit risk.

 

1: Free cash flow of borrower affects capacity to pay back loan

 

2: All consumer behavior has a potential for interaction with a borrower's free cash flows

 

3: the fact that revolving lines have a lower default rate compared to installment loan is tied to free cash flows as well.  In a revolving line a borrower has greater flexibility in juggling their cash flows and can take longer to default while increasing the loan amount. thus revolving lines although have lower default rates than installment loans they have greater exposure as the loss given default which may occur later will be of greater magnitude given the negative amortization which occurs if borrowers choose to keep balances growing until default.

 

4: Using credit scoring of variables which reflect borrower cash flows can perform as well as free cash flows in steady periods of economic activity as credit variables are known for multi-collinearity and it is very possible there is some consumer behavior pattern of variables which are correlated with cash flows which can yield accurate true positive rates of detecting bad loans.

 

5: There is an inherent bias in credit scoring which can make credit scores appear to outperform cash flow underwriting in the short term due to the fact that additional borrowings are tied to credit score.  This means if one has a high credit score they will qualify for multiple lines of credit from various lenders and because of this additional ability to borrow the borrower will be less likely to default than if the borrower had been denied credit.

 

Thus a borrower who has a high credit score and a poor cash flow situation can appear as a good loan for a time and also make the credit score appear to be a stronger predictor than cash flow.

 

6:  Given the entanglement issue between credit score and free cash flows and the expected interactions between the 2 models it makes sense to explicitly account for these relationships during credit score development.

 

7: there is also a chicken and egg problem with free cash flows and credit scores.  Before credit scoring took hold loans were based on sound underwriting based on cash flow proxy estimates and thus the credit people acquire which is used to build models is based on loans which traditionally had passed free cash flow or at a minimum some financial ratio tests.  Thus no can fairly say the underwriting loans based on cash flow can be replaced by credit scores.  In fact FICOs position is that lenders who used scores within a framework of sound underwriting are performing fine during the credit crisis.

 

 

8: That said the problem of forecasting uncertain cash flows, given uncertainty in borrower consumption behavior, and macro-economic risks makes the forecasting of borrower cash flows a complex task.  To incorporate uncertainty one must consider simulation and shocks to borrower income during the underwriting process.

 

9: It is well known the life events which threaten consumer well being comprised of divorce, health, death, and unemployment.  It is also well known that occupation, and education background affect expected income and income growth.

 

10: credit scores are able to account for behavioral problems which manifest in lack of control on the part of the borrower over time and are usually time weighted to allow borrowers to prove improvement or to minimize the long term impact of temporary adverse conditions. 

 

11: Free cash flows can be as effective in predicting bad loans as credit scores but the false positives for credit score tend to be lower as the credit score accounts for additional borrowings conditioned on itself which might not be warranted based on cash flows.

 

 

12: if all lenders use cash flow standards using the same assumptions then the impact of the credit score allowing for higher than warranted additional borrowings would diminish and free cash flow estimates would outperform credit scores.   This again would contain some reflexivity as if no one is willing to lend additional money then once a borrower runs out of free cash flow they will default as expected. (Barring a lottery winning)

 

13: The use of credit scores and free cash flows together should outperform either.  The most direct method of combining the 2 approaches would be to build predictive models for propensity to consumer and include them in the cash flow equations.

 

 

14: To ensure soundness and robustness of free cash flow based decisioning it is important to include constraints on borrower well being from a moral and social viewpoint.  Such constraints can be used to handle risks of life events.  A simple example of this is requiring the borrower to be able to maintain 6 months worth of liquid reserves to deal with life events.  If a loan has higher risk characteristics requiring 1 year or more of reserves might be prudent.

 

Thus free cash flow based lending can lend itself to optimizing loan terms to enhance the quality of the loan and borrower well being.

 

Another aspect to consider explicitly is the borrower's ability to save.  Using Savings and free cash flow rate one can infer the savings rate for a borrower using the following equation:

 

Free cash flow X (1+ savings rate)^(months on job)=savings

Using this equation one can solve for the savings rate for the borrower using the free cash flows and current point in time savings.

 

A borrower's ability to save itself is a strong sign of character and sound mind.

 

A draft Cash flow model for consumer loans:

NPV income*(stability function)-fixed expense-variable costs-savings rate-consumption function +NPV of savings+ additional borrowings

 

 

The savings rate and consumption function can be an ideal place for credit scoring to be added to strengthen the free cash flow model.

 

The stability function can be used to simulate loss of job and unstable cash flows.  For example self employed borrower loans which tend to have 3 times higher default rates can be underwritten using an adjustment for income stability.  For example one could simulate loss of income by 50% and see whether the borrower can still qualify for the loans using cash flows conditioned on loss of income.  This would better reflect the cash flow position of higher risk borrowers.

 

This estimate of income instability is another function which could be specified using credit scoring based models.