Skip to content
All library documents

Choosing Between SABR and Dupire Local Volatility Models

Article Quant Q&A · Author: victorinux
Read the original

Summary

The document raises the practical choice between SABR, including its arbitrage-free variant, and Dupire local volatility for pricing products. It frames the trade-off as SABR's ability to represent market smiles realistically versus the arbitrage-free construction of Dupire, while questioning whether the latter can produce unrealistic smile dynamics. It also asks how stochastic-volatility extensions alter the comparison.

This is a question rather than a resolved analysis: it offers no product-specific recommendations, market-practice evidence, or references, and does not establish that either model is preferable in particular settings. Its value is in identifying model-selection dimensions—fit to observed smiles, arbitrage behavior, and the added complexity of stochastic volatility—that practitioners need to assess for a product and use case.

Key ideas

  • The document contrasts SABR smile realism with the arbitrage-free construction attributed to Dupire local volatility.
  • It asks which product families are typically priced with each model but supplies no answer.
  • It raises stochastic-volatility extensions as a further model-selection consideration.
  • The post gives no empirical evidence or references for current market practice.

Tags

OptionsVolatilityDerivatives pricing
Full text
# SABR vs Dupire: when to use what?


# SABR vs Dupire: when to use what?












I was wondering for which products one would use the (AF)SABR model and for which ones Dupire's Local Volatility model.

If I understand correctly, Dupire is by construction Arbitrage-Free but produces an unrealistic smile. On the other hand, SABR seems to give a more realistic smile but might produce arbitrage. With these considerations at hand, which family of products would one price with SABR and which one with Dupire? Can someone give an idea of what are the "trending" market practices (and some reference would be great there!)?

Also, how do AFSABR and Dupire with stochastic volatility compare? My impression is that they should ameliorate the problem of their respective "classical" versions but, if so, how would one choose when to use what?

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.