Pricing a Heat Rate Option on Two Independent Lognormal Assets
Summary
The document derives a Black–Scholes-style valuation for a heat rate-linked payoff equal to the positive part of the difference between two independent lognormal asset values. It represents each asset with a standard normal variable, then changes numéraire to the asset appearing as the subtracted term. A rotation of the normal variables and a shift of their means reduce the two-dimensional expectation to a one-dimensional call payoff. The resulting effective underlying has the first asset’s par forward, the second asset’s par forward acts as the strike, and the effective volatility combines both lognormal volatilities.
The derivation gives an analytical route in the independent case, avoiding direct numerical integration of the original payoff. It states that correlation can also be handled with adjusted forward and volatility terms, and points toward quanto option pricing for that extension, but does not derive those adjustments. The setup is simplified: discounting and broader market conventions are not developed, and the result relies on the stated lognormal assumptions.
Key ideas
- The payoff is the positive part of the difference between two lognormal asset values.
- Changing numéraire and rotating the normal variables reduces the expectation to a one-dimensional call valuation.
- The effective strike is the second asset’s par forward, and the effective volatility combines both assets’ volatilities.
- The derivation assumes independence; correlated assets require adjusted terms that are not worked out here.
- The result depends on a simplified lognormal pricing setup.
Tags
Full text
# pricing of heat rate-linked derivative
# pricing of heat rate-linked derivative
It's a simplified model. Suppose $U_t$ is a random variables subject to Lognormal($x_1$, $z_1^2$)distribution. $V_t$ is a random variables subject to Lognormal($x_2$, $z_2^2$)distribution. Suppose they are independent here. The payoff of the heat rate-linked derivatives is $\max(U_T - V_T, 0)$. How to price this option? It's a integration stuff.
## Answer by achille hui (score 5, accepted)
https://quant.stackexchange.com/a/7982
This might be a surprise to you, you can evaluate the option using Black Scholes.
The key concept is change your numéraire from dollar to the asset associated with $V$. The $V$ in your payout $\max(U_t-V_t,0)$ will effectively get replaced by a constant, the par forward of asset $V$ at maturity $t$.
Since $U_t$ and $V_t$ are independent, you can parametrize them by two standard normal random variables $\eta_1$, $\eta_2$ with mean $0$ and standard derivation $1$:
$$U_t = e^{x_1 + z_1 \eta_1}\quad\text{ and }\quad V_t = e^{x_2 + z_2 \eta_2}$$
Let $(\cdots)^{+}$ stands for the function $\max(\cdots,0)$, the future value of the option is given by the integral:
$$\begin{align}\text{F.V.} = & \int ( U_t - V_t )^{+} \exp( -\frac{\eta_1^2 + \eta_2^2}{2}) \frac{d\eta_1 d\eta_2}{2\pi}\\ = &\int ( e^{x_1 + z_1 \eta_1} - e^{x_2 + z_2 \eta_2} )^{+} \exp( -\frac{\eta_1^2 + \eta_2^2}{2}) \frac{d\eta_1 d\eta_2}{2\pi}\\ = &\int ( e^{x_1 + ( z_1 \eta_1 - z_2 \eta_2 ) } - e^{x_2} )^{+} \exp( z_2\eta_2 -\frac{\eta_1^2 + \eta_2^2}{2}) \frac{d\eta_1 d\eta_2}{2\pi}\tag{*1} \end{align}$$
Let $$z = \sqrt{z_1^2+z_2^2}\quad\text{ and }\quad\begin{cases}u = \frac{z_1\eta_1 - z_2\eta_2}{z}\\ \\ v = \frac{z_2\eta_1 + z_1\eta_2}{z}\end{cases} \Longleftrightarrow \begin{cases}\eta_1 = \frac{z_1 u + z_2 v}{z}\\ \\ \eta_2 = \frac{z_1 v - z_2 u }{z}\end{cases} $$
It is easy to check:
$$ u^2 + v^2 = \eta_1^2 + \eta_2^2 \quad\text{ and }\quad du dv = d\eta_1 d\eta_2$$
Let $U_F = e^{x_1 + \frac{z_1^2}{2}}$ and $V_F = e^{x_2 + \frac{z_2^2}{2}}$ be the par forward of asset $U$ and $V$ at maturity $t$. We can rewrite $(*1)$ as: $$ \begin{align} &\int ( e^{x_1 + z u } - e^{x_2} )^{+} \exp\left( \frac{z_2(z_1 v - z_2 u)}{z} -\frac{u^2 + v^2}{2}\right) \frac{du dv}{2\pi}\\ = & \int ( e^{x_1 + z u } - e^{x_2} )^{+} \exp\left( \frac{z_2^2}{2}-\frac{(u + (z_2^2/z))^2 + ( v - (z_1z_2/z))^2}{2}\right) \frac{du dv}{2\pi}\\ = & \int ( e^{\tilde{x}_1 + z \tilde{u}} - e^{\tilde{x}_2} )^{+} e^{-\frac{\tilde{u}^2}{2}} \frac{d\tilde{u}}{\sqrt{2\pi}} \quad\text{ where } \begin{cases} \tilde{u}\; = u + (z_2^2/z)\\ \tilde{x}_1 = x_1 - z \frac{z_2^2}{z} + \frac{z_2^2}{2} = \log U_F - \frac{z^2}{2}\\ \tilde{x}_2 = x_2 + \frac{z_2^2}{2} = \log V_F \end{cases}\end{align}$$ As a result, we have: $$\text{F.V.} = \int ( U_F\,e^{z\tilde{u} - \frac{z^2}{2}} - V_F )^{+} e^{-\frac{\tilde{u}^2}{2}} \frac{d\tilde{u}}{\sqrt{2\pi}}\tag{*2}$$
This is nothing but the future value of a call option with strike $V_F$ on an asset with par forward $U_F$ and standard derivation $z$ at maturity. You can finish the integral using Black Scholes.
If $U_t$ and $V_t$ are not independent to each other, you can still transform F.V. to an integral of the form $(*2)$. The only difference is $U_F$ and $z$ there will be adjusted by some factors.
If you want to learn how to deal with the case with correlation, pickup any standard textbook on option pricing and look for the pricing of quanto option. The issues you encountered in pricing a quanto option is similar to the one you need to price your heat-linked option under the log normal model.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.