Choosing Models for Equity Volatility Surfaces and Historical IV Fitting
Summary
The document distinguishes pricing models used for equity options from parametric fits used to store and retrieve implied volatility data. It identifies Black–Scholes, including time-dependent versions, and local volatility as common industry choices, with stochastic rates, discrete dividends, and quanto effects often included in practical pricing. Stochastic volatility may also be used depending on the payoff and risks being managed; jump models are described as less common and potentially relevant for some variance products.
For historical IV queries, the response recommends fitting a parametric implied volatility surface rather than treating the task as option pricing. SVI or SSVI are given as examples that can represent varied market shapes, compressing each maturity's surface into parameters and potentially producing arbitrage-free surfaces. The document offers practitioner guidance rather than comparative empirical results. It does not specify calibration procedures, accuracy benchmarks, or how to choose piecewise segments, so the fit's precision and suitability would need to be assessed on the user's data.
Key ideas
- Industry equity option pricing commonly uses Black–Scholes variants and local volatility models.
- Practical pricing may incorporate stochastic rates, discrete dividends, and quanto effects.
- The appropriate model depends on the payoff being sold and the risks being managed.
- Parametric implied volatility fits such as SVI or SSVI can compactly represent a surface by maturity.
- Arbitrage-free behavior is a potential benefit of suitable SVI or SSVI surface construction.
Tags
Full text
# Commonly used vol surface calibration model in the industry # Commonly used vol surface calibration model in the industry I have 2 questions: - What is the most commonly used equity option pricing model? I learned jump diffusion at school, read about Hensen and a few other models online. - I am actually only calibrating the surface for easy retrieval of historical data. Right now to answer questions like "what is the implied vol of an option with strike = 105% of forward price and time = 3 month, for each day for the past 10 years", I literally have to interpolate on 3650 different option chains. Since each option chain is a 2000 * 4 grid, the process is slow. I am hoping that after calibrating, I only have to look at the parameters, instead of the entire surface. But since calibrating loses accuracy, my second question is, is it viable to increase accuracy by cutting the surface into a few segments, and then calibrating each segment separately? I am honestly only looking for a function that fits through the surface, and I have no problem making that function piece-wise since any fitting attempt is better than my current method of using the entire surface. ## Answer by vanna (score 6, accepted) https://quant.stackexchange.com/a/16662 The most used equity volatility models in the industry are the Black-Scholes model (including its time dependent version) and the local volatility model. It always come along with stochastic rates, discrete dividends and quanto effects (a must-have when pricing even simple payoffs) so the calibration/pricing process is much more involved than what you might expect. Of course stochastic volatility are used but it all boils down to the payoffs you are selling to clients and the subsequent risks you are managing. Jumps are less common to my knowledge but might be used in some specific cases (exotic variance products). As for your question if you are just computing implied volatilities for the sake of storing data then you are not using any model (pricing any other payoff) so I assume this is not related to your first question. What you might want to do is however to fit a parametrical implied volatility surface to your data set. Have a look at Gatheral's SVI/SSVI for example. This is basically a parametrical function that can fit many market shapes (so you only have to store some parameters per maturity) but most importantly can lead to arbitrage free surfaces.
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.