Using Discriminant Analysis with Receivables

RATIONALE

Managing the Credit Decision is essentially an exercise into looking into the future. When deciding to grant credit we can think of the new customer as falling into one of the following classes:

The customer will turn out to be good and pay all bills
The customer will be bad and will cause us to lose money

If the firm collects exactly the same data from each potential customer and retains the data in a database along with a binaryn code for the experience, then new customers data can be compared with the history. The statistical method used for this type of classification procedure is called Disciminant Analysis.

View Sample Customer Data

Notice that the binary value has been added to the customer data based on the firm's experience with the customer.

METHODOLOGY

Follow these steps:

Start the SPSS Program
Open the spreadsheet file customerdata.xls specifying variable names in the first row
Click on Analysis-Classify-Discriminant
The dialog box setup appears. You must specify the grouping variable and the independent variables and also what kind of output you want.
The grouping variable is goodbad
specify the range as 0 1
add the other variables to the independent list. Do not add ACCT
Click on classify then select
select summary table and the three plot options
click continue
click ok
Your output appears
What you need to copy to the clipboard is the little table that shows the "standardized canonical discriminant function coefficients"
Paste from the clipboard to the spreadsheet
Look at the charts and the classification results

USING THE STATISTICS

Open your spreadsheet and follow these steps.

Fill in Column G with the calculated z-scores for each old customer. Multiply each variable value which has been divided by the mean of all values by the discriminant coefficient. Add up all four terms.
Compare the z-scores for the 0 customers with the 1 customers
Use the z-function to calculate z-scores for new customers and make decisions.