Stochastic local volatility (SLV) and local volatility model calibration
The process of fitting and parameterizing a hybrid stochastic-local volatility model to match observed market option prices (e.g., the implied volatility surface) to ensure accurate pricing and hedging of exotic derivatives.
This skill is critical for structuring and trading desks to price complex path-dependent derivatives accurately, minimizing model risk and P&L volatility. Mastery directly translates into competitive pricing, effective risk management, and the ability to capture trades that pure local volatility or stochastic volatility models misprice.
1Careers
1Categories
9.0 Avg Demand
25%
Avg AI Risk
How to Learn Stochastic local volatility (SLV) and local volatility model calibration
Focus on 1) Understanding the fundamental difference between local volatility (Dupire) and stochastic volatility (Heston) models. 2) Mastering the core numerical methods for calibrating a pure local volatility model to vanilla options using finite differences or trees. 3) Studying the Heston model calibration to understand the dynamics of mean-reverting variance.
Progress to implementing a basic SLV model (e.g., the Schöbel-Zhu or Heston-inspired local vol). Practice calibrating to a realistic volatility surface using a Levenberg-Marquardt or differential evolution optimizer. Key pitfalls: mis-specifying the mixing parameter lambda, failing to enforce numerical stability of the Fokker-Planck PDE, and poor choice of the leverage function truncation.
Master the calibration of the full mixing parameter λ and the correlation ρ to fit exotic instruments (e.g., barriers, cliquets). Integrate real-time calibration pipelines with market data feeds. Focus on robust backtesting of model performance against P&L, and develop frameworks for dynamic hedging using SLV Greeks, which are complex due to the model's non-Markovian nature.
Practice Projects
Beginner
Project
Calibrate a Pure Local Volatility Model to Vanilla Options
Scenario
You are given a set of European option prices (calls/puts) for a single underlying (e.g., SPX) across strikes and maturities. Your task is to build a local volatility surface that perfectly matches these prices.
How to Execute
1. Write code to compute the risk-neutral density from the call price surface using finite differences on the Dupire formula. 2. Use the density to back out the local variance σ²(K,T). 3. Smooth the resulting surface using a parametric form (e.g., quadratic) or a robust interpolation scheme. 4. Validate by re-pricing the input options with your local vol model and comparing the implied volatilities.
Intermediate
Project
Build and Calibrate an SLV Model with a Fixed Mixing Parameter
Scenario
You need to price a set of barrier options for a client. The market vanilla surface is given. You decide to use an SLV model with a fixed λ=0.5 to capture the forward volatility skew dynamics better than pure local vol.
How to Execute
1. Implement the Heston stochastic volatility component. 2. Derive the leverage function L(t,S) from the target local volatility and the Heston instantaneous variance by solving the Riccati equation or via Monte Carlo simulation. 3. Implement a Monte Carlo engine for the SDE: dS = rS dt + L(t,S)√V S dW1; dV = κ(θ-V)dt + ξ√V dW2, with correlation ρ. 4. Calibrate the Heston parameters (κ, θ, ξ, ρ, V0) to match the vanilla surface, then fix λ and re-derive L(t,S) for final pricing.
Advanced
Case Study/Exercise
Optimize λ and ρ to Fit Exotic Structures and Analyze Hedge Effectiveness
Scenario
A trading desk is pricing a 5-year quarterly cliquet (accumulator) on an index. The market has a pronounced forward skew. Pure local vol underestimates the cliquet's value due to its convexity, while pure stochastic vol overestimates it. You must calibrate an SLV model to a joint set of vanilla options and a few key cliquet prices to find the market-consistent λ and ρ.
How to Execute
1. Set up a calibration objective function that minimizes the weighted sum of squared errors on both vanilla implied vols and exotic instrument prices. 2. Use a global optimizer (e.g., CMA-ES, differential evolution) to search over λ and ρ, while re-calibrating the Heston parameters and leverage function at each step. 3. Compute the model's Greeks (Delta, Gamma) for the cliquet via pathwise derivatives or likelihood ratio method. 4. Perform a backtest: simulate hedging the cliquet using the SLV Greeks and compare the residual P&L volatility against hedging using a pure local vol or pure stochastic vol model.
QuantLib provides core infrastructure for stochastic processes, PDE solvers, and Monte Carlo. NLopt/SciPy are used for the complex, multi-dimensional calibration of model parameters to market data.
Model-Specific Implementations
Custom Monte Carlo (C++/Python)Finite Difference PDE SolversRiccati Equation Solvers (for characteristic functions)
Monte Carlo is essential for pricing path-dependent exotics under SLV. PDE solvers are used for fast calibration of vanilla options. Riccati solvers are used for fast computation of the leverage function in some formulations.
Used to validate model performance by comparing model-implied prices and Greeks against real market movements and historical hedging outcomes.
Interview Questions
Answer Strategy
The strategy is to demonstrate a structured, practical workflow that balances theoretical rigor with numerical practicality. Start with data preparation, then explain the sequential calibration steps, highlight numerical challenges, and conclude with validation. A strong answer will mention the leverage function derivation and the treatment of the mixing parameter λ as a crucial step.
Answer Strategy
This tests diagnostic skills and model risk understanding. A good answer will list 3-4 plausible sources: 1) Calibration error to the vanilla surface (model not capturing the true risk-neutral distribution), 2) Incorrect estimation of the mixing parameter λ or correlation ρ, 3) Numerical errors in computing the leverage function L(t,S), 4) Greeks computation errors due to the model's complexity, 5) Market impact or transaction costs not in the model. The investigation plan should involve isolating each component: re-calibrating on denser data, checking numerical stability, and comparing Greeks from different approximation methods.
Careers That Require Stochastic local volatility (SLV) and local volatility model calibration