I am not turned on by reduced-form models or structural models. Surely there are other ways of modelling risky debt. Another thing -what is the best and simplest way of calculating probability of default? Has anyone done a probit regression to obtain the probability?

Amni, I agree that there are problems with both approaches. That's why a lot of work has been done and is being done on hybrid models that try to take the best of each type; also on reconciling the predictions of one type of model against the other. But out of interest, in what sense would a probit model not be a reduced-form model?

Hi Johnny, thanks for replying. Probit is a reduced form, I was just wondering which paper/study I can refer to for examples.I'm trying to study credit spread in the advent of bankruptcy, namely Enron and Worldcom. There is very little literature out there about this (that I know of), and looking at the credit spread of Enron for example, (which I obtained using spots between the company and Treasury); the spreads looked a little fishy to me. Therefore, I would like to get a rough measure of probability of default by performing probit by using factors such as the spreads, equity price, stock market price and leverage.As I truly belong to the world of the uninitiated; do you have any ideas how I can proceed?

QuoteOriginally posted by: amniI am not turned on by reduced-form models or structural models. Surely there are other ways of modelling risky debt. Another thing -what is the best and simplest way of calculating probability of default? Has anyone done a probit regression to obtain the probability?Amni,My guess is that you are put off by specific formulations of reduced-form and structural models. These are both very general ways of modeling default risk, and can be adapted to conform to just about any view you may have regarding the nature of default risk. The Merton approach (KMV, et al) is the most popular structural model, but a probit regression would be another type of structural model. (Johnny, I don't think it would be reduced-form because you would postulate that default hazard rate is a function of certain observable or latent variables, just like Merton's model. In fact, you could probably replicate the results of the contingent claims model through a purely statistical model.)I suppose that you could, in an APT framework, simply infer the market view on default probability from spreads on bonds or default swaps. However, I wouldn't consider that a model. It would just be extracting the consensus view from price information.I think that between structural and reduced-form, you have enough flexibility to model default risk in just about any manner you want. I have played around with structural models based purely on cash flows and liquidity, ignoring long-term debt and firm value completely. Instead of Merton's strategic default, the idea was to predict cash-cruch induced, involuntary default. Unfortunately, the quality of data I had access to did not allow me to really verify whether there was any merit to the methodology.Reduced-form models are probably even more flexible, but the big danger is over-fitting. I think they are more appropriate for approximating default risk over medium to large portfolios of issuers than for modeling single-issuer defaults.Anyway, I'd be interested to hear if you come up with any results using a non-standard model.