Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of...

37
Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon, June 3, 2015

Transcript of Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of...

Page 1: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Tradable Volatility Estimates

Bruno DupireHead of Quantitative Research

Bloomberg L.P.

Conference in Honor of Steve ShreveCarnegie Mellon, June 3, 2015

Page 2: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Outline

I. Trading Volatility Estimates

II. Tradable Estimates

III. A New Hi-Lo Based Tradable Estimate

Page 3: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

I. Trading Volatility Estimates

Page 4: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Historical volatility tends to depend on the sampling frequency

SPX 2006 to 2011 Data

Page 5: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Historical volatility tends to depend on the sampling frequency

SPX 2006 Data

Page 6: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Historical volatility tends to depend on the sampling frequency

SPX 2007 Data

Page 7: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Historical volatility tends to depend on the sampling frequency

SPX 2008 Data

Page 8: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Historical volatility tends to depend on the sampling frequency

SPX 2009 Data

Page 9: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Historical volatility tends to depend on the sampling frequency

SPX 2010 Data

Page 10: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Historical volatility tends to depend on the sampling frequency

SPX 2011 Data

Page 11: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Historical Vol / Historical Vol Arbitrage

If weekly historical vol < daily historical vol :

buy strip of T options, Δ-hedge daily

sell strip of T options, Δ-hedge weekly

Adding up :

do not buy or sell any option;

play intra-week mean reversion until T;

final P&L :

Page 12: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Daily / Weekly Vol Arbitrage

Page 13: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Intra week mean reversion strategy

Long

Short Short

Long

Page 14: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

1 day/1 week volatility tradeMore profitable when vol is high

Page 15: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

500 stocks cross section

Page 16: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Daily – Weekly Vol Spread for S&P 500 constituents

Jan 2008 to Apr 2015

Page 17: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Daily – Weekly Vol Spread for S&P 500 constituents

Jan 2008 to Apr 2015

Page 18: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Covariation Trading

Page 19: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Historical Cov / Historical Cov Arbitrage

Page 20: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

II. Tradable Estimates

Page 21: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Some Definitions(Normal as opposed to Lognormal convention)

TradableUnbiased

Page 22: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Discretized Quadratic Variation

Tradable unbiased estimate

Δ hedge of a parabola

Costless stock trading

Page 23: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Classical Estimator

Page 24: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Move-Based Estimator

For option pricing, a volatility estimate should reflect the cost of hedging

Hedge according to spot moves (move based) better than hedge at fixed times (time based)

Page 25: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Move-Based Estimator

noise is smoothed out by phase averaging

Page 26: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

OHLC

Open

High

Low

Close

Page 27: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Hi-Lo Estimates

Page 28: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

The Uncertainty of the Night

SPX Index GS US Equity

Page 29: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

But…

§ Hi-Lo Estimators are less useful than the classical one because they cannot be traded

§ Highs and Lows are always observed after the fact

§ However… we introduce a tradable Hi-Lo based estimator

Page 30: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

III. A New Hi-Lo Based Tradable Estimate

Page 31: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Averaging Down

Page 32: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

A Side Note

Page 33: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

A New Estimator

Page 34: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Trading Strategy

(Mid – Close)/Range

Page 35: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Comparison of Estimators

Page 36: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Conclusion

§ Most common volatility trade is historical vs implied

§ With tradable estimates: can trade historical vs historical, without any option market

§ Hi-Lo estimates are usually not tradable…

§ Not because they depend on Hi-Lo but because they do not depend on Close!

§ We introduce a new tradable estimate based on Hi-Lo (and Close!)

Page 37: Tradable Volatility Estimates - CMU · Tradable Volatility Estimates Bruno Dupire Head of Quantitative Research Bloomberg L.P. Conference in Honor of Steve Shreve Carnegie Mellon,

Thank You