|
A Few Comments About the Limitations of Credit Scoring
By Michael C. Dennis MBA,
CBF
The theory behind credit risk scoring software
is that the existence of certain risk factors, or combinations
of risk factors, will result in a greater likelihood of serious
payment delinquency or payment default. The stated advantages
-
Faster decisions.
-
More consistent decisions.
-
Better decisions resulting in increased sales and profits.
-
Increased productivity in the credit department.
-
The ability to identify customers that can be offered
higher credit limits at relatively low risk to the company.
-
Flexibility built into the model so that the credit manager
can alter certain parameters and increase or decrease the
amount of credit risk the scoring model will consider acceptable,
or unacceptable.
Much has been said and written about the advantages
and benefits of credit scoring applied to commercial credit risks,
but little has been written about the limitations of credit scoring
models. Here are a few comments about the problems, limitations
and disadvantages of credit scoring models:
-
A properly designed credit scoring system
allows creditors to evaluate thousands of applications consistently,
impartially and quickly. If this is true, then the opposite
must also be true. A poorly designed credit scoring system
can evaluate thousands of applicants and can make the wrong
recommendation every time.
-
Credit risk can never be measured precisely,
and any model that says it can is wrong.
-
Credit risk can change almost overnight. Example:
The owner of a business dies and there is no one qualified
to replace him.
-
Credit managers should be able to override
the credit score and its credit recommendation. However,
doing so is difficult to justify if there is a serious payment
problem, or worse a bad debt loss. For this reason, credit
professionals are reticent to override the scoring model
even when they believe the "recommendation" is wrong.
-
Internally developed credit scoring models
often lack sophistication and usually have not been subjected
to critical analysis of the statistical significance of the
factors used to develop a credit score and a credit recommendation...but
-
Professionally designed, tested and validated
credit scoring models can be expensive, and can be hard to
customize. As a result, the credit scores these programs
generate may not mimic the decision making style and the
risk tolerance of the companies that purchase them - and
therefore they do not produce the desired results.
-
Some professionally designed models only provide
the credit manager with a numerical score. With this limited
information, it can be quite difficult for the credit manager
to explain a negative credit decision [based only on a numerical
score] to an irate credit applicant, or to an active customer.
|
|
|
|
|