返回题库

BS 波动率错配的对冲误差

Black Scholes Robustness

专题
Finance / 金融
难度
L4

题目详情

金融数学题:BS 波动率错配的对冲误差。

英文原题

You sell a European option for which you estimated the volatility to be σt\sigma_{t} . What is the PnL error if you hedge this option until expiry and that the realized volatility turns out to be σrσt\sigma_{r} \neq \sigma_{t} ?

解析

用交易波动率 σt\sigma_t 做 delta 对冲,但真实实现波动率为 σr\sigma_r

经典 tracking error(忽略融资/离散误差)主导项来自 gamma:

PnL 误差  120TΓtSt2(σr2σt2)dt.\boxed{\text{PnL 误差}\ \approx\ \frac12\int_0^T \Gamma_t\,S_t^2\bigl(\sigma_r^2-\sigma_t^2\bigr)\,dt}.

其中 Γt=SSV(St,t)\Gamma_t=\partial_{SS}V(S_t,t)


英文解析

This classic result was first discussed by El Karoui, Jeanblanc and Shreve (1996) and Carr and Madan (1997) and is known as the PnL tracking error formula or Black Scholes robustness formula. Let VV be the price of the option. It is designed at inception to verify the Black Scholes equation with σt\sigma_{t}

Vt+12VSSσt2S2+rSVSrV=0V_{t} + \frac{1}{2} V_{SS}\sigma_{t}^{2}S^{2} + rSV_{S} - rV = 0

On another hand, we apply Itô's lemma to VV , which evolves during the life of the trade at the realized volatility σr\sigma_{r}

dV=Vtdt+VsdS+12Vssσr2S2dtdV = V_{t}dt + V_{sdS} + \frac{1}{2} V_{ss\sigma_{r}^{2}S^{2}dt}

We isolate VtV_{t} in (14), substitute it in (15) and integrate from 0 to TT

V(T)V(0)=0TrVdt0TrSVSdt+0TVSds+0T12VSB(σrσt)2S2dtV(T) - V(0) = \int_{0}^{T}rVdt - \int_{0}^{T}rSV_{Sdt} + \int_{0}^{T}V_{Sds} + \int_{0}^{T}\frac{1}{2} V_{SB}(\sigma_{r} - \sigma_{t})^{2}S^{2}dt

We see that the 2 first integrals correspond to the cost of holding a position in cash equal to (VSVS)(V - SV_{S}) , which is the position in cash after selling the option at VV and hedging it with a VSV_{S} position in stock

I1=0TrVdtrSVSdtI_{1} = \int_{0}^{T}rVdt - rSV_{S}dt

The third integral is the change in value of a VSV_{S} position in stock

I2=0TVSdSI_{2} = \int_{0}^{T}V_{S}dS

And we find that the difference in value of the european claim is equal to its expected hedging cost (I1+I2)(I_{1} + I_{2}) plus an error term

PnLerror=0T12VSS(σrσt)2S2dt\mathrm{PnL_{error}} = \int_{0}^{T}\frac{1}{2} V_{SS}(\sigma_{r} - \sigma_{t})^{2}S^{2}dt