以下是引用stonexu1984在2007-5-25 1:20:00的发言:
谢谢两位, 请问warecucff, 这么说即期利率还是个三因素的线形模型, 关键是如何用kalman filter 获得这三因素对吧? 即使有三因素了,那么affine model的系数,rho0 rho' 是从哪获得呢? (有三因素,系数才能得到r(t))
那么bloomberg上的1-yr,3-yr,10yr...之类的是即期利率么? 如果是的话那不是能直接得到了么?
when you pricing a bond[zero coupon bond for example and also for conventionality], what do you take as the discount factor? The short rate, and now we assume there are three latent factors [not always, as some studies find two factors could explain 99% variation for certain countries]explaining linearly the dynamics of short rate, we could price the bond prices by substituting the equations of factors and short rates into the bond pricing equation. Therefore you have the analytical solutions for the bond prices. These analytical solutions contains the roh0 and rho you mentioned above, and it is just for illustrating these analytical solutions, when it comes to implementing the model, you solve backwards from bond prices.
A good model usually has few structural parameters that enable varieties, those restrictions pn model only help to generate reasonable results but not to complicate it. You will get to know better when you are trying to replicate the empirical.
I don't know what they offer from bloomberg, but conventionaly you should find something called yield rate of zero coupon bond for different maturities, and these yield rates are mostly interpolated from other type of coupon bonds. Check Piazzesi(2003) for data selections.
Finally, Campbell, J.Y., A.W. Lo and A.C. MacKinlay (1997) The Econometrics of Financial Markets. Princeton, NJ: Princeton University Press
Good luck.