Probability of Default (PD) Modelling for
Credit Cards & Personal Loans
- The Client is one of the universal banks in the South east Asian region.
- It has strong customer base in assets and liabilities side.
- The Client wanted to predict the probability of Credit Cards and Personal Loans products defaulting on their loan obligations by considering several factors such as credit history, financial statements, and economic indicators.
- Data extraction for PD of Cards and personal loans.
- Building models using macroeconomic data.
- Preliminary tests to determine expected correlation, stationarity, optimal lag and cointegration.
- Johansen cointegration test used to decide between VAR or VECM model.
- Model diagnostics ensure candidate model meets criteria.
- The best model chosen among candidate models.
- The models predicted the possibility of a borrower defaulting on their loan and cards.
- Developed a reliable and easy-to-understand model that guides good lending decisions.
- Update the model continuously to reflect changes in the economy and borrower behavior.