Itô Product and Quotient Rules

Motivation: why this matters in quant finance

Discounted prices, self-financing portfolios, and numeraire-relative assets are products and ratios of stochastic processes. Ordinary product and quotient rules miss the covariation term, and that missing term changes drift calculations.

The simplest finance example is discounting. If $S_t$ follows a Brownian price model and $D_t=e^{-rt}$ is the discount factor, the product $D_tS_t$ is the object that should become a martingale under the risk-neutral measure. To check that, the product differential must be computed correctly.

Lawler derives the product rule from the same principle as Itô's lemma: terms involving products of Brownian increments can survive at order

dt

. The quotient rule is then just Itô's formula applied to

X_t/Y_t

, with the same covariation bookkeeping.

The informal idea

In ordinary calculus,

d(fg)=f\,dg+g\,df

because the leftover product $df\,dg$ is of order $(dt)^2$ and vanishes. In Brownian calculus, a diffusion increment has a $dW_t$ part, and $dW_t\,dW_t$ contributes like $dt$ . The leftover product is no longer negligible.

The rule is therefore simple to remember:

d(X_tY_t)=X_t\,dY_t+Y_t\,dX_t+\text{shared noise correction}.

That shared-noise correction is $d\langle X,Y\rangle_t$ . It is zero when one factor has finite variation, and non-zero when both factors load on the same Brownian shock.

Formal definitions

Suppose two Itô processes are driven by the same Brownian motion:

dX_t=H_t\,dt+A_t\,dW_t,\qquad dY_t=K_t\,dt+C_t\,dW_t.

Their covariation is

\langle X,Y\rangle_t =\lim\sum_j\left(X_{t_j}-X_{t_{j-1}}\right) \left(Y_{t_j}-Y_{t_{j-1}}\right),

and Lawler computes

d\langle X,Y\rangle_t=A_tC_t\,dt.

The stochastic product rule is

d(X_tY_t)=X_t\,dY_t+Y_t\,dX_t+d\langle X,Y\rangle_t.

Equivalently,

\begin{aligned} d(X_tY_t) &=\left(X_tK_t+Y_tH_t+A_tC_t\right)dt\\ &\quad+\left(X_tC_t+Y_tA_t\right)dW_t. \end{aligned}

The quotient rule follows from Itô's formula applied to $f(x,y)=x/y$ on the region $y\ne 0$ . If $Y_t$ stays away from zero over the interval of interest, then

d\left(\frac{X_t}{Y_t}\right) =\frac{1}{Y_t}\,dX_t-\frac{X_t}{Y_t^2}\,dY_t -\frac{1}{Y_t^2}\,d\langle X,Y\rangle_t +\frac{X_t}{Y_t^3}\,d\langle Y\rangle_t.

This is not a separate primitive theorem; it is the multivariable Itô formula specialized to a ratio.

Key properties

Shared Brownian shocks create the correction

The extra term is

d\langle X,Y\rangle_t=A_tC_t\,dt.

It depends only on the diffusion coefficients, not on the drift coefficients. For two assets driven by the same risk factor, the covariance exposure enters product dynamics immediately.

Finite-variation factors behave classically

If $Y_t$ has no Brownian term, then $C_t=0$ and

d\langle X,Y\rangle_t=0.

The product rule reduces to the ordinary one. Discounting by a deterministic money-market account is the main finance case.

Quotients carry two correction terms

For $X_t/Y_t$ , the denominator's own quadratic variation contributes

\frac{X_t}{Y_t^3}\,d\langle Y\rangle_t,

and the interaction between numerator and denominator contributes

-\frac{1}{Y_t^2}\,d\langle X,Y\rangle_t.

These terms are the source of many drift adjustments in numeraire changes and relative-price calculations.

Product rules can also be derived from Itô's formula

Take $f(x,y)=xy$ in the two-variable Itô formula. Since $f_{xy}=1$ and $f_{xx}=f_{yy}=0$ , the only second-order term is the cross-variation term. This matches Lawler's direct differential derivation.

Worked examples

Example 1: Brownian motion times geometric Brownian motion

Lawler's product-rule example takes $Y_t=W_t$ and lets $X_t$ satisfy

dX_t=mX_t\,dt+\sigma X_t\,dW_t.

Here

d\langle W,X\rangle_t=\sigma X_t\,dt.

Therefore

\begin{aligned} d(W_tX_t) &=W_t\,dX_t+X_t\,dW_t+d\langle W,X\rangle_t\\ &=W_t(mX_t\,dt+\sigma X_t\,dW_t)+X_t\,dW_t+\sigma X_t\,dt\\ &=X_t\left((mW_t+\sigma)dt+(\sigma W_t+1)dW_t\right). \end{aligned}

The $\sigma X_t\,dt$ term is the stochastic correction.

Example 2: discounted stock dynamics

Let

dS_t=\mu S_t\,dt+\sigma S_t\,dW_t, \qquad D_t=e^{-rt}.

Since $D_t$ has finite variation,

d\langle D,S\rangle_t=0.

The product rule gives

\begin{aligned} d(D_tS_t) &=D_t\,dS_t+S_t\,dD_t\\ &=D_t(\mu S_t\,dt+\sigma S_t\,dW_t)-rD_tS_t\,dt\\ &=D_tS_t\left((\mu-r)dt+\sigma\,dW_t\right). \end{aligned}

Under the risk-neutral drift $\mu=r$ , the drift vanishes and the discounted price is a martingale candidate.

Example 3: quotient of two diffusion factors

Suppose

dX_t=H_t\,dt+A_t\,dW_t,\qquad dY_t=K_t\,dt+C_t\,dW_t,

and $Y_t>0$ . Then

\begin{aligned} d\left(\frac{X_t}{Y_t}\right) &=\frac{1}{Y_t}(H_t\,dt+A_t\,dW_t) -\frac{X_t}{Y_t^2}(K_t\,dt+C_t\,dW_t)\\ &\quad-\frac{A_tC_t}{Y_t^2}\,dt +\frac{X_tC_t^2}{Y_t^3}\,dt. \end{aligned}

The first line is the ordinary quotient-rule shape. The second line is the Brownian correction. When $Y_t$ is deterministic, $C_t=0$ and the correction disappears.

Common confusions and pitfalls

"Products only need correction for nonlinear functions." A product is already a two-variable function with a mixed second derivative. If both factors have Brownian parts, the mixed term survives.

"The quotient rule can be copied from ordinary calculus." A noisy denominator has its own quadratic variation, and numerator-denominator covariation also matters.

"Every product gets a non-zero cross term." The term is $d\langle X,Y\rangle_t$ . It vanishes when one factor has finite variation or when the Brownian drivers are independent in the relevant component.

"The covariation term depends on drift." Drift terms are order $dt$ ; their products vanish at the quadratic-variation scale. The covariation term depends on diffusion coefficients.

"Discounting always adds an Itô correction." Deterministic discounting has finite variation, so $d\langle D,S\rangle_t=0$ . Corrections appear when the numeraire itself has diffusion risk.

Where this goes next

Itô's Lemma: Gives the general chain rule from which product and quotient rules are special cases.
Quadratic Variation: Defines the covariation term that product algebra must keep.
Geometric Brownian Motion: Provides the standard asset-price SDE used in product-rule examples.
Black-Scholes PDE: Uses discounted portfolio and option-value dynamics to remove risk and identify the pricing PDE.
Stratonovich Integrals: Contrasts Itô's left-endpoint algebra with a calculus closer to the ordinary chain rule.

References

Lawler, G. F. (2023). Stochastic Calculus: An Introduction with Applications. Ch. 3 §3.4 (More versions of Itô's formula), §3.6 (Covariation and the product rule), §3.7 (Several Brownian motions).