I am trying to fit a model for a loan portfolio which have been acquired over several years, each of the loan has one of the several term periods (different performance window) over which the loan needs to be paid off. A few number of loans have charged off before the term expired and a few after the term.

The event is chargeoff (1) or not (0).

How should I define time to event? For chargeoffs, the time to event is Time to chargeoff. For non-chargeoffs, should the time be till the end of their respective term or end of the maximum term of all the terms? I tried former definition, and used Cox proportional hazard model in SAS, but my expect vs actual survival curves are way off. Any help is appreciated.


  • $\begingroup$ Can you clarify the question? You have some loans for which the event ("charge off") has occurred and some for which the event has not occurred. For those for which the event has occurred, as I understand it, the time till the event is t(charge off) - t(loan disbursed). For others, there is no charge off event... but you still want to define a time till charge off? Are you trying to estimate the time till charge off? Are you assuming that all loans in the portfolio will be charged off? (A charged off loan is one that the lender declares noncollectable, right?) $\endgroup$ – roundsquare Jul 27 '16 at 21:09
  • $\begingroup$ Thanks. What should I put for non-chargeoffs in Time variable. Should I leave it as missing or Time to payoff or time to max payoff of all terms? $\endgroup$ – Zenvega Jul 28 '16 at 1:08
  • $\begingroup$ It is impossible to give you an answer without more information. It all depends on what you are trying to do with the data. $\endgroup$ – roundsquare Jul 28 '16 at 3:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.