Local Volatility Calibration using the “ Most Likely Path ”

1. Local Volatility Calibrationusing the “Most Likely Path” 19 December 2006 for Computational Methods in Finance Prof Ali Hirsa/ Paris Pender

3. Option Data Extraction • Use “Option Metrics” from the WRDS (Wharton Data Research Services) • Option Metrics is a comprehensive source of historical price and implied volatility data for the US equity and index options markets. • Volatility Surface contains the interpolated volatility surface for each security on each day, using a methodology based on kernel smoothing algorithm.

4. Data Fields • We download the following fields from the database: • Days to Expiration. • Interpolated Implied Volatility • Implied Strike Price • Implied Premium. • Spot price.

5. Mechanism • A standard option is only included if there exists enough option price data on that date to accurately interpolate the required values. • We have designed a data processing module in Matlab that pulls this data in Matlab vectors and then fed into out local volatility processing engine. • The Matlab vectors contain implied volatility data only for OTM calls and puts.

6. Example OptionMetrics file

7. Calibration to SPX • Given a finite set of implied volatility ( )

8. Calibration to SPX • Given a finite set of implied volatility ( ) • We interpolate onto a “calibration grid” using Matlab’s gridfit function

9. Calibration to SPX • Given a finite set of implied volatility ( ) • We interpolate onto a “calibration grid” using Matlab’s gridfit function • This is the “market” implied volatility surface that use to calibrate on

10. Table of Call Prices Results Table of Put Prices

11. Results

12. Results

13. Overview of scheme Take market impled volatility surface as first guess of Local Vol Local volatility surface Converged. Stop!

14. Two Key Concepts • Most Likely Path: • Implied Volatility Proxy

15. Two Key Concepts • Most Likely Path \ • Definition: • Difficult to compute directly from the original local volatility dynamics: • Under simpler dynamics, however, we have a closed form solution: where

16. Two Key Concepts Recall: 1) Compute by our iterative algorithm 2) Compute by Monte-Carlo

17. Two Key ConceptsComparison of the most likely path • Using iterative algorithm (black) • Using Monte Carlo Simulation (blue) • They are very similar!

18. Two Key Concepts • Implied volatility proxy • This states that the BS implied volatility of an option with strike K and expiration T is given approximately by the path-integral from valuation date (t=0) to the expiration date (t = T) of the local volatility along the “‘most likely path”

19. How does our method work?? (1/5)

20. How does our method work?? (2/5) Based on a fixed-point iteration scheme: • Initialize • Repeat the following until convergence under

21. How does our method work?? (3/5) For each (K,T) on the calibration grid: • Get: a. initialize b. set c. set d. repeat (b-c), until converges in • Set

22. How does our method work?? (4/5)

23. How does our method work?? (5/5) Conclusion: The method is robust and calibration takes around 3 minutes

24. Overview of scheme Take market impled volatility surface as first guess of Local Vol Local volatility surface Converged. Stop!

25. Questions/ Comments Presentation by: Kwasi Danquah, Saurav Kasera, Brian Lee, Sonky Ung