Applying Itô's Product Rule to Self-Financing Portfolios
Summary
The document works through the cash-account holdings needed to keep a portfolio self-financing under two choices for the risky-asset holding: the time integral of the asset price, and the asset price itself. It starts from portfolio value as cash holdings times the bank account plus risky holdings times the asset price, then imposes the self-financing condition. Applying Itô's product rule shows which differentials and cross-variation terms enter each case.
For the integrated-price holding, the risky position has finite variation, leading to a cash-account adjustment driven by the squared asset price. When the holding equals the asset price, the quadratic variation term contributes; under diffusion dynamics, the resulting cash process includes both drift and a Brownian component. The derivation assumes an exponentially growing bank account, diffusion cash holdings, and standard asset dynamics where time increments have zero cross variation with the asset price. These assumptions limit the formulas' scope.
Key ideas
- A self-financing portfolio equates changes in value to gains from the existing asset holdings.
- Itô's product rule includes cross-variation terms when portfolio holdings and asset prices are stochastic.
- The time integral of the asset price has differential equal to the current price times the time increment.
- If the risky holding equals the asset price, the asset's quadratic variation affects the cash position.
- The derivations assume diffusion processes and a continuously compounded bank account.
Tags
Full text
# Self finance conditions - proof check
# Self finance conditions - proof check
Find expressions for the process $\psi=(\psi(t),\ 0\leq t\leq T)$ , so the portfolio $(\phi,\ \psi)$ is self-financing when:
(1) $\phi(t)= \int_{0}^{t}S_{s}ds $
(2) $\phi(t)=S_{t}$
where $\phi(t)$ is an Ito process.
Where is the error in my Ito product rule? I can't figure out what the cross-variation terms should be. Any help would be greatly appreciated.
## 1)
Given: $$V_t=A_t\psi_t + S_t\phi_t$$
We have the below condition: $$\phi_t= \int_{0}^{t}S_{s}ds$$
condition for a portfolio to be self financing: $$dV_t= \psi_t dA_t +\phi_t dS_t$$
Input our condition of: $$ \phi_t= \int_{0}^{t}S_{s}ds $$
$$dV_t= \psi_t dA_t + (\int_{0}^{t}S_{s}ds)dS_t$$
## 2)
given: $$V_t=A_t\psi_t + S_t\phi_t$$
We have the below condition: $$\phi_t= S_t$$
So We now have: $$V_t=A_t\psi_t + S_t^{2}$$
See below for correct solution
## Answer by Daneel Olivaw (score 1, accepted)
https://quant.stackexchange.com/a/38771
Let me define $B_t=A_t=e^{rt}$ $-$ to avoid confusing it with the geometric average $1/t\int S_u\text{d}u$. Your portfolio value is: $$ V_t =\psi_tB_t+\phi_tS_t $$ To be self-financing we need to enforce one of the following equivalent conditions: $$\begin{align} & \text{[1]} \quad \text{d}V_t =\psi_t\text{d}B_t+\phi_t\text{d}S_t \\[3pt] & \text{[2]} \quad B_t\text{d}\psi_t+\text{d}\psi_t\text{d}B_t+S_t\text{d}\phi_t+\text{d}\phi_t\text{d}S_t=0 \end{align}$$
You haven't specified any dynamics for the asset $S_t$ but we will assume that the cross-term $\text{d}t\text{d}S_t$ is equal to $0$ which is true for all common models such as Black-Scholes or Heston. We will also assume the process $\psi_t$ is a diffusion: $$\text{d}\psi_t=a_{\psi}(t,S_t)\text{d}t+b_{\psi}(t,S_t)\text{d}W_t$$
Case 1: $\phi_t=\int_0^t S_u\text{d}u$
Let us first compute the differential of $\phi_t$. Note $\phi_t=\phi(t)$ is a function of time $t$ thus: $$\begin{align} \text{d}\phi_t=\text{d}\left(\int_0^tS_u\text{d}u\right)=S_t\text{d}t \end{align}$$
Hence: $$\begin{align} B_t\text{d}\psi_t+\text{d}\psi_t\text{d}B_t+S_t\text{d}\phi_t+\text{d}\phi_t\text{d}S_t & = B_t\text{d}\psi_t+\text{d}\psi_t\text{d}B_t+S_t^2\text{d}t+S_t\text{d}S_t\text{d}t \\[3pt] & = B_t\text{d}\psi_t+rB_t\text{d}\psi_t\text{d}t+S_t^2\text{d}t \end{align}$$
Because the process $\psi_t$ is a diffusion, the cross-term $\text{d}\psi_t\text{d}t$ should also be null, hence: $$\begin{align} B_t\text{d}\psi_t+rB_t\text{d}\psi_t\text{d}t+S_t^2\text{d}t & = B_t\text{d}\psi_t+S_t^2\text{d}t \end{align}$$
Therefore: $$ \psi_t=\psi_0-\int_0^te^{-ru}S^2_u\text{d}u$$
Case 2: $\phi_t=S_t$
$$\begin{align} B_t\text{d}\psi_t+\text{d}\psi_t\text{d}B_t+S_t\text{d}\phi_t+\text{d}\phi_t\text{d}S_t & = B_t\text{d}\psi_t+\underbrace{\text{d}\psi_t\text{d}B_t}_{0}+S_t\text{d}S_t+(\text{d}S_t)^2 \\[3pt] & = B_t\text{d}\psi_t+S_t\text{d}S_t+(\text{d}S_t)^2 \end{align}$$
Therefore: $$ \psi_t=\psi_0-\int_0^te^{-ru}S_u\text{d}S_u-\int_0^te^{-ru}(\text{d}S_u)^2$$
If we assume $S_t$ follows a diffusion of the form: $$ \text{d}S_t = a_S(t,S_t)\text{d}t+b_S(t,S_t)\text{d}W_t$$
Then: $$ \psi_t=\psi_0-\int_0^te^{-ru}\left(S_ua_S(u,S_u)+b^2_S(u,S_u)\right)\text{d}u-\int_0^te^{-ru}S_ub_S(u,S_u)\text{d}W_u$$Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.