Model uncertainty, recalibration, and the emergence of delta-vega hedging (Q2412385)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Model uncertainty, recalibration, and the emergence of delta-vega hedging
scientific article

    Statements

    Model uncertainty, recalibration, and the emergence of delta-vega hedging (English)
    0 references
    0 references
    0 references
    23 October 2017
    0 references
    The authors consider an agent who sells a non-traded option \(V\) on a stock \(S\) maturing at \(T\) with payoff depending only on the value of \(S\) at \(T\). The agent has access to three liquidly traded securities to hedge the exposure: \(S\), a vanilla call option \(C\) on \(S\), and a bank account with zero interest rate. The market price \(C(t)\) of the call option is assumed to be \(C_{BS}(t,S(t),\Sigma(t))\) where \(C_{BS}\) is the Black-Scholes price with implied volatility as third argument. The agent is uncertain about the dynamics of \(S(t)\) and \(\Sigma(t)\) and assumes \[ dS(t) = S(t) \sigma_{P}(t)dW_{0}(t), \] \[ d\Sigma(t) = \nu_{P}(t) dt + \eta_{P}(t)dW_{0}(t) + \sqrt{\xi_{P}(t)} dW_{1}(t), \] where \((W_{0},W_{1})\) is a Brownian motion in \(\mathbb{R}^{2}\) and the processes \(\zeta_{P} = (\sigma_{P},\nu_{P},\eta_{P},\xi_{P})\) are chosen to satisfy a zero-drift condition ensuring \(C(t)\) is a local martingale. The Black-Scholes model corresponds to \(\zeta_{P_{0}} = (\Sigma,0,0,0)\). To hedge the exposure to \(V\), the agent trades \(S\) and \(C\) using a self-financing strategy \((\theta,\phi)\), so that the price process \(Y(t)\) of the hedged position satisfies \[ dY_{\theta,\phi}(t) = \theta(t)dS(t) + \phi(t)dC(t) - dV_{BS}(t,S(t),\Sigma(t)), \] where \(V_{BS}\) is the Black-Scholes price of \(V\). I.e., \(S\) and \(C\) are marked to market, while \(V\) is marked to model. At maturity, the payoff is \(V_{BS}(T,S(T),\Sigma(T))\), which depends only on \(S(T)\). The agent aims to maximize its expected utility \(U\) from the terminal value, and takes models less seriously the more they deviate from the reference Black-Scholes model: \[ v(\psi) = \sup_{\theta,\phi} \inf_{P} \mathbb{E}_{P}\left[U(Y_{\theta,\phi}(T)) + \frac{1}{2\psi}\int_{0}^{T}U'(Y_{\theta,\phi}(t))\left|\zeta_{P}(t) - \zeta_{P_{0}}(t)\right|^2dt\right], \] where \(\psi>0\) measures the uncertainty aversion of the agent. The authors obtain a hedging strategy \((\theta^{\star},\phi^{\star})\), a family of models \(P(\psi)\), and \(w\geq 0\) such that as \(\psi \rightarrow 0\), \[ \begin{aligned} v(\psi) &= \mathbb{E}_{P(\psi)}\left[U(Y_{\theta^{\star},\phi^{\star}}(T)) + \frac{1}{2\psi}\int_{0}^{T}U'(Y_{\theta^{\star},\phi^{\star}}(t))\left|\zeta_{P(\psi)}(t) - \zeta_{P_{0}}(t)\right|^{2}dt\right] + o(\psi)\\ &= U(Y_{0}) - U'(Y_{0})w\psi + o(\psi).\end{aligned} \] The hedging strategies are \[ \phi^{\star}(t) = \frac{\partial_{3}V_{BS}(t,S(t),\Sigma(t))}{\partial_{3}C_{BS}(t,S(t),\Sigma(t))}, \] \[ \theta^{\star}(t) = \partial_{2}V_{BS}(t,S(t),\Sigma(t)) - \phi^{\star}(t)\partial_{2}C_{BS}(t,S(t),\Sigma(t)); \] i.e., call options are purchased to hedge the vega of the position, while the stock is used to hedge the delta. The strategies are independent of \(U\) and \(\psi\). The parameter \(w\) is related to the expected cost of the trading strategy in the Black-Scholes model, and in particular satisfies that the indifference ask price for \(V\) -- i.e.\ the price at which the agent is indifferent between keeping a flat position and changing the position by selling \(V\) for that price -- is \(V_{BS}(0,S(0),\Sigma(0)) + w\psi + o(\psi)\). The results can be interpreted as a justification for the delta-vega hedging based on dynamically recalibrated volatility parameters in the Black-Scholes framework.
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    delta-vega hedging
    0 references
    dynamically recalibrated implied volatility
    0 references
    Black-Scholes model
    0 references
    self-financing strategy
    0 references
    model uncertainty
    0 references
    Hamilton-Jacobi-Bellman-Isaacs equation
    0 references
    stochastic differential game
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references