# Justification for substituting "Itô differentials"

I'm reading Shreve's Stochastic Calculus for Finance, Volume II. In it, he uses the stochastic differential notation. For example, he may write

$$\mathrm{d}X(t) = \sigma(t)\mathrm{d}W(t)+\alpha(t)\mathrm{d}t\tag{ 1}$$

to mean (formally)

$$\int_{t=s_0}^{t=s_1}\mathrm{d}X(t) = \int_{t=s_0}^{t=s_1}\sigma(t)\mathrm{d}W(t)+\int_{t=s_0}^{t=s_1}\alpha(t)\mathrm{d}t\text{.}\tag{2}$$

This is fine and makes sense. The problem is that he will often substitute the LHS of $$(1)$$ for the RHS of $$(1)$$ in other stochastic differential equations. For example, given the function $$S(t)$$, he may use $$(1)$$ to justify the equality

$$S(t) \mathrm{d}X(t) = S(t)\sigma(t)\mathrm{d}W(t)+S(t)\alpha(t)\mathrm{d}t\text{,}\tag{3}$$

which is equivalent to

$$\int_{t=s_0}^{t=s_1} S(t) \mathrm{d}X(t) = \int_{t=s_0}^{t=s_1} S(t)\sigma(t)\mathrm{d}W(t)+\int_{t=s_0}^{t=s_1} S(t)\alpha(t)\mathrm{d}t\text{.}\tag{4}$$

This trick appears to be purely symbolic. Intuitively, it makes sense. However, I am having trouble proving that you can make this substitution. Is there a proof that you can make these kinds of substitutions (in this example or in general)?

I don't see where in the book Shreve justifies (3) directly from (1). But he would have almost surely referred to the discretized versions of the "differentials". That is:

$$X(t_{i+1})-X(t_{i}) \approx \sigma(t_i)(W(t_{i+1})-W(t_{i}))+\alpha(t_{i})(t_{i+1}-t_{i})$$

implies $$S(t_{i})(X(t_{i+1})-X(t_{i})) \approx S(t_{i})\sigma(t_i)(W(t_{i+1})-W(t_{i}))+S(t_{i})\alpha(t_i)(t_{i+1}-t_{i}).$$

We would then take summations and limits (needed in an integration mechanism) on both sides and justify why a stochastic integral wrt to an Ito process integrator can be defined as in (4), which is Definition 4.4.5 in the book, given after Shreve had already defined stochastic integrals wrt to $$W(t)$$ and $$t$$ integrators.

• "I don't see where..." This question is based on (4.4.27), but a better example is from (5.2.20) to (5.2.22). I don't consider this an answer to the question because I am looking for a proof. Apr 29 at 2:29
• Proof of what? (2) and (4) are both definitions, one defines the integrator, the other defines the integral wrt that integrator.
– ir7
Apr 29 at 2:56
• (4.4.27) is a formal application of Ito-Doeblin 4.4.6 for (functions of) Ito processes (generalizing 4.4.1) to $f(x)=S(0)\exp(x)$, with $X(t)$ Ito process and $S(t)=f(X(t))$.
– ir7
Apr 29 at 3:16
• Once one also accepts (4.8.12), the multidimensional Ito-Doeblin theorem, and its consequence (a couple of phrases below it) the Ito's product rule, derivations like (5.2.20), $d(D(t)S(t))$, come relatively easy (you can, of course, post them as separate questions, if you think the application of Ito-Doeblin doesn't seem straightforward). [They are not a matter of stochastic integral definitions, in formal or informal shapes.]
– ir7
Apr 29 at 3:28