Presentation is loading. Please wait.

Presentation is loading. Please wait.

Analyzing Credit Behavior of an Asset Class using Historical Data: A Case Study April 22, 2013.

Similar presentations


Presentation on theme: "Analyzing Credit Behavior of an Asset Class using Historical Data: A Case Study April 22, 2013."— Presentation transcript:

1 Analyzing Credit Behavior of an Asset Class using Historical Data: A Case Study April 22, 2013

2 Acknowledgements ► My sincere thanks to Bama for letting me use certain slides from her presentation

3 In memory of… all the ping pong balls sacrificed in the North-West corner Image source: http://cargocollective.com/rmattgarcia/Ping-Pong-Trophy-1

4 Agenda ► Asset Class and Data ► Scope of the study ► Methodology and findings ► Application

5 Asset class and Data ► More than 115000 loans covering three asset classes: Two-wheeler loans, Commercial Vehicle loans, and MSME loans ► Following details were available over a period of 5 years: ► Monthly billing and recovery ► Loan amount ► Borrower Occupation ► Borrower income ► LTV (for a certain asset class) ► Tenure ► City/State ► Only such loans were considered for the study where the repayment information is available for at least 12 months ► More than 67000 loans were selected for the study ► We would refer the three asset classes as A, B and C (in random order)

6 Scope of the Study ► Estimate transition probabilities to migrate from one PAR state to another ► Observe payment behaviour to understand the time a loan typically spends as PAR0 or PAR30 and whether loans default early or late in their ‘lifetime/tenure’ ► Observed transition probabilities for sub-portfolios of loans with certain seasoning, loan tenure and LTV for collateralized loans

7 Methodology ► Estimating transition proportions: proxy for probabilities ► Suppose ‘x’ loans were observed in dpd-1 bucket during any time of the loan ► Of these ‘x’ loans, ‘y’ loans were also observed to be in dpd-30 bucket ► So, x/y is the proportion of loans which ever moved from dpd-1 to dpd-30

8 Methodology ► Estimating transition proportions: proxy for probabilities ► Suppose ‘x’ loans were observed in dpd-1 bucket during any time of the loan ► Of these ‘x’ loans, ‘y’ loans were also observed to be in dpd-30 bucket ► So, x/y is the proportion of loans which ever moved from dpd-1 to dpd-30 Figure 1: Transition Probabilities

9 Methodology ► Estimating time-to-PAR ► For each loan which was ever observed in dpd-30, determine the time elapsed before it first entered the dpd-30 bucket ► Time elapsed could be observed in absolute scale (e.g. months) or in scale relative to its tenure (e.g. 30% of loan tenure) ► A frequency distribution of the time elapsed for all such loans could be useful

10 Methodology ► Estimating time-to-PAR ► For each loan which was ever observed in dpd-30, determine the time elapsed before it first entered the dpd-30 bucket ► Time elapsed could be observed in absolute scale (e.g. months) or in scale relative to its tenure (e.g. 30% of loan tenure) ► A frequency distribution of the time elapsed for all such loans could be useful Figure 3: Time to Hit PAR30 Asset AAsset BAsset C

11 Methodology ► Estimating Time-spent-in-delinquency ► As per the latest observation, a loan could be ‘Current/Mature’ or ‘Delinquent/Default’ ► For each ‘Current’ loan, determine the time spent as delinquent (Note that a loan, current now, could have been delinquent in the past) ► Similarly for each delinquent loan (no matter dpd-1 or dpd-180 or default), determine the time spent as delinquent ► ‘Time’ could be absolute or relative (though relative helps in comparing across tenures)

12 Methodology ► Estimating Time-spent-in-delinquency ► As per the latest observation, a loan could be ‘Current/Mature’ or ‘Delinquent/Default’ ► For each ‘Current’ loan, determine the time spent as delinquent (Note that a loan, current now, could have been delinquent in the past) ► Similarly for each delinquent loan (no matter dpd-1 or dpd-180 or default), determine the time spent as delinquent ► ‘Time’ could be absolute or relative (though relative helps in comparing across tenures) Figure 2: Life Spent in Delinquency Asset AAsset BAsset C

13 Application ► Consider a portfolio of loans of asset class A ► ‘Time-to-Default’ analysis would provide us the distribution of defaults over the tenure of loans (say 80% dpd-1 or PAR0 cases occur after 3 month or 25% of loan tenure and rarely after 75% of loan tenure) ► Use transition proportions to calculate losses and recoveries in subsequent periods ► Stress the assumptions on transition proportions and ‘time-to-default’ to test the portfolio

14 Thank you


Download ppt "Analyzing Credit Behavior of an Asset Class using Historical Data: A Case Study April 22, 2013."

Similar presentations


Ads by Google