1
$\begingroup$

I am analyzing the transition of the bond prices in the affine models in the form of $P(t,T)=e^{-A(t,T)-B(t,T)r_t}$

using the property that the diffusion and the drift of an affine model can be presented as $$b(t,r)=b(t)+\beta(t)r_t \quad \quad (1) $$ $$\sigma^2(t,r)=a(t)+\alpha(t)r_t \quad \quad (2) $$

where the A and B satisfy the system of the below ordinary differential equations $$\partial_t A(t,T) = \frac{1}{2} \ a(t) \ B^2(t,T) - b(t) \ B(t,T), \quad A(T,T)=0 \quad \quad (3)$$ $$\partial_t B(t,T) = \frac{1}{2} \ \alpha(t) \ B^2(t,T) - \beta(t) \ B(t,T) -1 , \quad B(T,T)=0 \quad \quad (4)$$

I am trying to derive bond price P(t,T) for the Ho-Lee model. $dr_t=\mu t dt + \sigma dW_t^*$
My understanding is that $b=\mu t, \ \beta=0, \ a=\sigma^2, \alpha=0$ will be the parameters for the eq (3) and (4), which gives

$$\partial A(t,T) = \frac{\sigma^2}{2} B^2(t,T) - \mu t \ B(t,T) \quad \quad (5)$$ $$\partial B(t,T) = -1 \quad \quad (6)$$ and further it should yield $$B(t,T)=T-t \quad \quad (7)$$

The last step is unclear to me.
My questions
1) why T-t is is substituted for, not t or T
2) why the integration $\partial B=-1$ results in (T-t) and not -(T-t)? integrating a function f'(x)=-1 should give f(x)=-x, shouldn't it?

$$A(t,T)=\frac{-\sigma^2}{6}(T-t)^3 + \int_t^T b(s)(T-s)ds \quad \quad (8)$$ giving the final equation $$P(t,T)=exp \ \big{(}\frac{-\sigma}{6}(T-t)^3 -\frac{\mu}{6} T^3 + \frac{1}{2} \mu T t^2 -\frac{1}{3} \mu t^3 - (T-t)r_t \big{)} \quad \quad (9)$$

Question 3)
How the integral $\int_t^T b(s)(T-s)ds$ gives $\frac{\mu}{6} T^3 - \frac{1}{2} \mu T t^2 + \frac{1}{3} \mu t^3$. I can't get the algebra here. Is this an ordinary integral?

$\endgroup$

1 Answer 1

2
$\begingroup$

From Equation (6), $B(t,T)=-t+c(T)$ for some function $c(T)$. $1=P(t,t)=e^{-A(t,t)-(c(t)-t)r_t}$ or $A(t,t)+(c(t)-t)r_t=0,\,\forall (r_t,t)$. So $c(t)=t, A(t,t)=0,\forall t$.

For Equation (8) you have missed the square on $\sigma$ and a factor of $\frac13$. Then you just need to substitute in the function for $b(s)$ and integrate the following to get the correct result. $$A(t,T) = -\frac{\sigma^2}{2}\int_t^T(T-s)^2ds +\mu\int_t^T s(T-s)ds.$$

$\endgroup$
10
  • $\begingroup$ did you mean P(T,T), A(T,T)? and than c(T)=T ? the notation is confusing $\endgroup$
    – Michal
    Commented May 4, 2016 at 17:22
  • 1
    $\begingroup$ @Michal: Does it make any difference? The variables are place holders. You may as well write $P(x,x)$ and $c(x)=x$ or use any other symbols, Latin, Greek, Chinese, Arabic or Hobbitian. $\endgroup$
    – Hans
    Commented May 4, 2016 at 17:57
  • $\begingroup$ ok, understood. One more question to that. Is the integral in the (8) an ordinary one. I can't get the right form of the P(t,T). Could you clarify that part please? $\endgroup$
    – Michal
    Commented May 4, 2016 at 19:21
  • 1
    $\begingroup$ The solution manual solution is correct. Do the second integral in my answer. $\endgroup$
    – Hans
    Commented May 4, 2016 at 23:45
  • 1
    $\begingroup$ Let us continue this discussion in chat. $\endgroup$
    – Hans
    Commented May 5, 2016 at 0:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.