Hypothetical Model to Predict PTPTN Defaulter – An Illustration

In regression analysis, logistic regression (“logit”) uses a logistic function to model a binary dependent variable.  It is used in various fields, including machine learning, most medical fields, and social sciences (Read more here).  In a business application, it can be used to predict the likelihood of a homeowner defaulting on a mortgage, by inputting homeowner’s data such as age, income, credit scores and other relevant information (Read more here).

Data such as school exams results, courses offered by University, type of schools, urban or rural, number of siblings, attendance records and other relevant personal data could be used as input for the logit model.

We do not have the information so this article will use “artificial” data to illustrate the capability of the logit model.  For simplicity, only five (5) independent variables are considered in the logit model.

·        Number of A’s in Sijil Pelajaran Malaysia (SPM) = X₁

·        Course offered by University – Science (When X₂ = X₃ = X₄ = 0)

·        Course offered by University – Technology = X₂

·        Course offered by University – Art = X₃

·        Courser offered by University – Business = X₄

The following picture is the artificial data for the logit model.  A total of 500 rows of data were created.  The first column shows whether the student will default the loan, 1 means default while 0 means no default.  Second column is the number of A’s that the student scored in SPM.  Column 3 to 5 are the courses offered by University to that student (triple zero means the course offered is Science).  While the last 4 columns are the equations for logit model.

Once the regression coefficients are determined by using Maximum Likelihood Estimation (MLE), the probability equation is

For example, let’s say a student scored 9 A’s in SPM and was offered a Science course in University, the chances for this student to default PTPTN loan is 0.24.  On the other hand, a student who scored 4 A’s in SPM and was offered an Arts course in University, the default probability is higher at 0.88.

So how could we use this model?  For those students who have higher default rate, a guarantor might be required.  If a complete credit rating system is established, government could potentially offload their financial burden to the private sector.  The above framework is possibly useful for PTPTN in evaluating applicants for its loans.