I am doing my master thesis on MBS modelling.which practical models and inputs would you recommend?thank youJulien

Model what? Default? loss?

To make my question more precise:for pricing MBS, what do you think about pricing MBS as :a portfolio of index amortizing swaps?reduced form models (intensity of default as a function of an incentive to refinance)?an optimal recurcive refinancing?and if you have an idea, which gives the best results?thanksJulien

It greatly depend on what type of MBS you're talking. For instance, if you want to price agency MBS (GNMA, FNMA, GHLMC type), then you don't have to worry about credit risk issues as these securities are guaranteed by the US Treasury. then the only thing you'll have to do is to find a prepayment model (Davidson model is one example) to extract prepayment speed over various time horizons. From there, project the expected cash flows over the life time of the MBS and discount these cash flows (and sum them up).Then, getting the price is straightforward.Hope this helps.Regards.

Yes, I want to model agency MBS. For the moment I still don't have data so I am trying to see in which direction to go until I get the data.

Check out articles by Andrew Davidson. You can also check out the MBS book by Edited by Fabazi

Try that link to find Davidson's prepayment model :http://www.ad-co.com/prepayment_models/mbs_pp_model.htm

I made a 3-factor Monte Carlo model for a federally backed MBS last year for a masters paper last year. I used Stanley Pliska's paper "A 3-Factor Valuation Model for Mortgage-Backed Securities" (http://tigger.uic.edu/~srpliska/) as the mathematical guideline. I used MATLAB to apply it and it worked pretty well, though a trully affective model would have many more prepayment factors and not exclude the possibilty of partial prepayment.

Hi pdb36,I wanted to know if your masters paper was available ?Is it possible to have a look at it or could you upload it ?I think it would be a good resource to this forum.Thanks !Regards.

QuoteOriginally posted by: pdb36I made a 3-factor Monte Carlo model for a federally backed MBS last year for a masters paper last year. I used Stanley Pliska's paper "A 3-Factor Valuation Model for Mortgage-Backed Securities" (http://tigger.uic.edu/~srpliska/) as the mathematical guideline. I used MATLAB to apply it and it worked pretty well, though a trully affective model would have many more prepayment factors and not exclude the possibilty of partial prepayment.I implemented it also, but did you test it on data?