Probability of Default

Overview


The probability of default samples use the following dataset



which can be found at the following link: data


The data represents monthly snapshots of a collection of loans. Each snapshot gives a date when the snapshot was taken. It contains an ID that identifies each loan. Each loan has a rating, either A,B, or C, which represents the degree of riskiness of each loan based on an assessment of the bankers. A borrower's rating can change over time.

Each loan also has three numbers assigned to, here called factors. The factors represent measurements taken of the borrower. These numbers could be things like income/interest or other business measurements indicating health of the borrower. The factors can change over time.

Lastly, each snapshot record contains an indicator as to whether the loan has defaulted in that period or not. Default is represented as a 0 if no default has occurred or a 1 if default has occurred. Once a loan has defaulted, no further snapshots are taken of that loan. That is, the default record is the last record in the dataset for any given loan.

Demos and Tutorials


PD from loan snapshot data.
Calculates a probability of default as an average default rate.
source
PD by rating
Shows how to derive a pd value from loan defaults and a set of factors.
source
PD over Time
Shows how to derive a pd value from loan defaults and a set of factors.
source
PD Factors
Shows how to derive a pd value from loan defaults and a set of factors.
source
Transition Matrix
Constructs a Markov Model of Transition Matrices from loan data with ratings.
source