3. P.D.E.
There are three2 state variables in the model: observed St, unobserved vt, and observed t. Throughout, t
will stand for the current time, T will represent the time of expiration, and τ = T — t will be referred to as
the “time to expiration” .
It will be more convenient to operate with a different set of state variables: observed st = In St, unobserved
vt, and observed τ. Clearly, for fixed T there is a one-to-one correspondence between the two sets of state
variables.
Given equations (1) and (2), by Ito’s lemma:
dst
= (r — δ —2) dt + √v)dWιt,
(3)
dvt = (a — βvt) dt + y√vfdW2t∙
(4)
It is reasonable to model the value function of a derivative security, D, as a function of the state vari-
ables, D = D (St,vt,t) = D (est,vt,T — (T — t)) = D (st,vt,τ). D is hypothesized to be twice continuously
differentiable in (st,vt) and once continuously differentiable in τ.
Let и ∈ (t, T]. By the fundamental theorem, E [p(s">^T ∣^j = °^stβ^tt,τ). Taking и arbitrarily close
to t, E [dd^t) lʃt] = 0.
It follows that:
d D (st,vt, τ)
0 = EJ d—l-fi-trJ.
Mt
lʃt
= EJ
dD (st,vt,τ)
Mt
= Mt 1E [dD (st,Vt, τ) — D (st,Vt, τ) rdt∣Et].
(5)
Expressing dD (st,vt,τ) by Ito’s formula as:
dD =
= —Dtdt + Dsdst + Dvdvt + 2Dssd (s)t + 2Dvvd {v)t + Dsvd (s, v)t =
= —Dtdt + Ds [ɑ' — d —2) dt + √vtd^1t] + Dv [(a — βvt) dt + l√vtd^2t] +
+ 2 DssVtdt + 2 Dvv l 2vtdt + Dsv ElVtdt.
2For methodological reasons, I prefer to treat time as a separate state variable. Clearly, it has a deterministic and trivial law
of motion.