Equity derivative strategy

2012 Q1 update and Trading Volatility

Colin Bennett
(+34) 91 28 93056
cdbennett@[Link]

1
 Contents

VOLATILITY TRADING FOR DIRECTIONAL INVESTORS

Call overwriting

Protection buying

Choosing strike of option

ESSENTIAL FACTS OF VOLATILITY TRADING

Volatility is not as expensive as you think

Hedging equity with volatility

Stretching Black-Scholes

Variable annuity and structured products impact on the market

ADVANCED VOLATILITY TRADING

Dividends and correlation

Advanced volatility measures

Term structure and skew

2
Call overwriting can yield enhance returns Call overwriting

Reasons why volatility is usually overpriced Call overwriting

Return
150%
140%
 Demand for protection 130% Enhance performance
selling OTM calls when
120%
volatility is high
 Unwillingness to sell low premium (near dated) options 110%

100%
90%
 High gamma of near dated options has gap risk premium 80%
70%

 Index implied lifted by structured products 60%

50%
50% 60% 70% 80% 90% 100% 110% 120% 130% 140% 150%
Equity Equity - call Strike
Call overwriting improves portfolio performance
S&P500 1m ATM call overwriting performance
Price (rebased)
 Selling expensive implieds can lift performance, but note that 1000
BXM is a total return index, so needs to be compared
900
the delta of position is lower which reduces benefit of equity to S&P500 total return index for a fair comparison
800
risk premium 700
600

 On balance call overwriting is a winning strategy in most 500

market environments (except in very bullish markets) 400

300

200
 BXM index gives performance of S&P500 1m ATM call 100
overwriting BUT is total return (need to compare it to SPXT not 0
1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012
SPX)
BXM (1m 100%Buy Write) S&P500 S&P500 total return

3
Overwriting with 1 month 104% strike is best Call overwriting

Strike of optimal strategy depends on period of time examined

 Overwriting with near dated options outperform as can sell 12 one month options in a year, but only 4 three month
options. BUT selling multiple short dated options can be seen as more risky (if markets rise one month, then fall)

 Equities must have a realistic positive return during back test period (negative return optimum strike is < ATM). In
these periods a strike of 103-104% is best for 1 month SX5E options (107-108% for 3 month options).

 Strike should be higher for higher volatility stocks (rule of thumb is use c25% delta calls)
Call overwriting 3.0%

104%
103% 2.5%

Call overwriting return - index return

105%
102%
106%
2.0%
101%
Only upside risk is
108%
1.5% reduced (use
100%
110%
1.0%
Sortino ratio rather
Exact peak strike for overwriting than standard dev)
depends on period of backtest 0.5%

Index
0.0%
-8% -7% -6% -5% -4% -3% -2% -1% 0%
Call overw riting volatility - index volatility

4
Performance depends on market environment Call overwriting

Overwriting performs best on an index rather than single stock

 Overwriting with 1 month ATM call overwriting has outperformed, but there were periods where it underperformed

 As index implieds are more overpriced than stocks (implied correlation too high) best to overwrite using indices

 As call overwriting less attractive for single stocks, there is greater chance enhanced call overwriting can lift returns
Relative performance (rebased)
140
2009
S&P500 1m
Call ATM performance
overwriting call overwriting
depends on performance
market environment since 1988 trough

130
2003
trough

120
Start of late
Call overwriting 90's bull market
outperforms
110 Credit
crunch

100
Asian TMT
crisis peak
90
Call ovewriting
underperforms
80
Outperform Significantly Breakeven Significantly Underperform Significantly Significantly
Underperform Outperform Outperform Underperform
70
1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
BXM / S&P500 total return

5
Markets can crash, correct or enter bear market Protection buying

Examining previous declines is relevant to current crisis

 DAX declines since 1959 can be grouped into 3 categories: crash, correction and bear market

 Crash has a high annualised decline (c90%) for a period of 3 months or less

 Bear markets are multiple year declines of 23% or more

 Corrections are remaining declines of up to a year and up to 22%

Type of correction protection is required against, can help determine which strategy to choose

Average Average Average annualised Duration Decline

Duration decline decline range range

Crash 1 month 31% 96% < 3 months 19% - 39%

Correction 3 months 14% 58% <= 1 year 10% - 22%

Bear market 2.4 years 44% 26% 1-5 years 23% - 73%

6
Option structures incorporate delta and vol view Protection buying

Choice of protection strategy depends on type of decline to be hedged

 Short dated puts are most appropriate for crashes, (rolling) put spreads are best for corrections (and bear markets)

 Protection has to be paid for through premium, loss of upside (collar) or potential losses on downside (1x2 put spd)

 Bullish strategies are the reverse of protection strategies. View on equity and volatility markets guides choice of
optimum option structure

Implied expensive

Volatility View
Bearish Bullish

Market View Market

View

Implied cheap

7
ITM options trade like a future Choosing strike of option

ITM options have highest delta, hence highest return if investor is confident
 Typically investors trade ATM or OTM options as they are cheapest

 Highest return for a given market move occurs for ITM options, as their higher delta more than outweighs their
higher cost (ITM options are similar to futures)

 ITM options do not have much convexity (compared to ATM), hence is a risky strategy as the high cost of ITM
options could be lost
Profit of 1 year call if markets rise 10%
Return ITM options have highest profit
60%

50%

40%
OTM options have low
30% profit due to low delta

20%

10%

0% 0%

4%

8%

2%

6%
%

%
60

64

68

72

76

80

84

88

92

96

10

10

10

11

11
Strike

8
 Contents

VOLATILITY TRADING FOR DIRECTIONAL INVESTORS

Call overwriting

Protection buying

Choosing strike of option

ESSENTIAL FACTS OF VOLATILITY TRADING

Volatility is not as expensive as you think

Hedging equity with volatility

Stretching Black-Scholes

Variable annuity and structured products impact on the market

ADVANCED VOLATILITY TRADING

Dividends and correlation

Advanced volatility measures

Term structure and skew

9
Implied should be above realised Volatility is not as expensive as you think

Assuming a positive equity risk premium, implied vol should be above realised
 Implied volatility is on average 1-2 pts above realised volatility

 Short volatility strategies are effectively long equity risk (assuming negative spot vol correlation)

 If long equity is expected to earn more than the risk free rate (i.e. positive equity risk premium) then short volatility
should also be profitable (as exposed to the same risk)

 Fair value of implied volatility is therefore above realised volatility

Structured products selling variance contain equity risk

 Shorting implied volatility is an opportunity, but returns are likely to be similar to going long equity

 There are many structured products based on selling variance swaps, their returns have suffered in the downturn
as volatility spiked as equities fell

10
Vol is not as expensive as you think Hedging equity with volatility

Implied volatility is negatively correlated to equity market, but is a poor hedge

 There is an R2 of 56% between weekly SX5E and vStoxx returns. Hence implied volatility can be used as a low
cost hedge.

 Strategy is not zero cost as implied volatility is on average expensive (trades above average realised volatility) as
short vol is implicitly long equity risk (and equities have an equity risk premium)

 Hedging with volatility (or variance swaps) is less effective than with futures, as volatility has less than 100%
correlation with equities and is expensive (on average)

SX5E and vStoxx weekly returns SX5E hedged with futures or variance swap
vStoxx Return 100% SX5E
100% 6%

80% 5%
2
R = 0.56 4%
60%
3%
Risk free rate Add increasing
40% 2%
(SX5E 100% hedged amount of variance
20% 1% swaps to 100% SX5E
with futures)
SX5E 0%
0%
-1% 0% 5% 10% 15% 20% 25%
-30% -20% -10% 0% 10% 20% Volatility
-20% -2%

-40% -3%
-4%
SX5E + 1 year variance swap SX5E + futures

11
Continuous delta hedging with known volatility Stretching Black-Scholes

Black-Scholes has unrealistic assumptions Continuous delta hedging with known vol
5
As underlying has constant
volatility, the amount earned from
 Known future realised volatility 4
gamma is exactly equal to the
loss of time value (theta)
3
 Ability to hedge continuously
2

 Black-Scholes also assumes volatility is constant, but results 1

Gamma
are the same if this condition is relaxed 0
90% Theta 95% 100% 105% 110%
-1
Straddle T=0 Straddle T=1

P&L known under Black-Scholes Profit (& loss) from delta hedging
P&L (%)
50
 Payout of delta hedged option is known under Black-Scholes Hedging with delta
40
calculated using known
volatility means profit (or 30
 Profit (or loss) is value of option using future (known) volatility loss) is constant 20
less price paid for option (i.e. value of option using implied 10

volatility) 0
-10 -8 -6 -4 -2 0 2 4 6 8 10
-10
Realised vol - implied vol (%)
-20
 P&L of delta hedging an option under Black-Scholes vs
-30
“Realised vol – implied vol” is a straight line
-40

-50

12
Continuous delta hedging with unknown volatility Stretching Black-Scholes

If future volatility is unknown delta is incorrect Continuous delta hedging with unknown vol
5
Delta hedging a cheap option with
 If future volatility is unknown delta has to be calculated using 4
delta calculated from implied
the implied volatility volatility (as volatility is unknown)
is always profitable, but profits
3
are spot dependent
 Delta calculated is only correct if future volatility = implied
volatility (i.e. when option trading at fair price). Hence P&L line 2

has to go through origin 1

Small Large
 If there is a difference between future realised volatility and 0
profit profit

implied volatility then the payout is uncertain 90% 95% 100% 105% 110%
-1
Straddle T=0 Straddle T=1

Profit (& loss) from delta hedging

P&L (%)
Always profit from delta hedging cheap option When realised volatility =
50
40
implied volatility, delta from
30
 While the exact profit is uncertain, delta hedging a cheap implied volatility is correct
hence profit is always 0 20
option will always give a profit (P&L line has to go through 10
origin) 0
-10 -8 -6 -4 -2 -10 0 2 4 6 8 10
 While bigger the difference between implied and realised -20 Realised vol - implied vol (%)
means the delta is less accurate, this effect is dwarfed by the -30
additional cheapness of the option -40
-50
Average profit Profit +/- 1σ

13
Discrete delta hedging with known volatility Stretching Black-Scholes

Continuous trading is unrealistic assumption Discrete delta hedging

5

 Presence of weekends and less than 24 hour trading makes 4 Changing hedging
frequency has changed
continuous trading an unrealisic assumption the profit made
3

 Trading costs make it unlikely delta hedging is done 2

continuously throughout the day
1

 Ideal time to delta hedge is before an uncertain 0

annoucement or potential change in direction of market 90% 95% 100% 105% 110%
-1
Initial delta hedged straddle Straddle rehedged after +5% move

Discrete hedging introduces noise Profit (& loss) from delta hedging
P&L (%)
50
 Noise of discrete delta hedging is independent of how cheap Hedging a 4x frequency
40
(or rich) the option is halves the noise from discrete
30
delta hedging
20
 It is possible to lose money when buying a cheap option 10
when discrete delta hedging 0
-10 -5 -10 0 5 10
Realised vol - implied vol (%)
-20
 Noise from discrete hedging is halved if frequency of delta -30
hedging is 4x as frequent -40
-50

 σP&L = σ x Vega x √(π/[4N]) Average profit Profit +/- 1σ Profit +/- 1σ with 4x frequency

14
Discrete delta hedging with unknown volatility Stretching Black-Scholes

Real life has unknown vol and discrete hedging Discrete delta hedging
150 4000
Implied vol at inception = 22%
 Errors in real life are combination of unknown volatility and Realised vol over life = 42%
100 3500
discrete hedging
50 3000
 Possible to lose money when delta hedging a discrete option
0 2500
 If you had bought an ATM option in April 2008 with 22% Apr-08 May-08 Jun-08 Jul-08 Aug-08 Sep-08 Oct-08 Nov-08 Dec-08

implied, you would have lost money despite realised being -50 2000
almost twice as large (42%)
-100 1500

 Loss due to fact delta was up to 24% different (when using P&L using delta from future vol P&L using delta from implied vol SX5E (RHS)

future realised volatility instead of implied)

Profit (& loss) from delta hedging
P&L (%)
 As market declined before vol spiked, delta using implied was 50
Errors from discretely hedging
near 100% but should be less. Hence trader bought too many with unknown volatility is the
40

futures to delta hedge and suffered loss. sum of error due to discrete 30

hedging and error due to 20

unknown volatility 10

Should calculate delta using expected vol -10 -8 -6 -4 -2

0
-10 0 2 4 6 8 10

-20 Realised vol - implied vol (%)

 If volatility is seen as 5pts too cheap, should bump vol surface -30
5pts to calculate deltas -40
-50
Average profit Profit +/- 1σ
 Error is most significant for long vol, as markets tend to decline
when vol rises
15
Variable annuity often give investors a “put” option Variable annuity hedging

Variable annuities often sold with protection to make them more attractive
 With fixed annuities, the insurance company invests proceeds and guarantees a fixed return

 Variable annuities allow the purchaser to pick the investments, but leaves investor exposed to the downside

 To make variable annuities more attractive they were often sold with forms of downside protection

Hedging of variable annuities lifts index term structure and skew

 While products can be up to 20+ years long, position are dynamically hedged with 3-5 years puts for liquidity
reasons

 When modelling dynamic strategies, future implied volatility is modelled with a confidence interval, e.g. 95% to
ensure only 1 in 20 chance of a loss. As volatility rose to levels greater than seen in great depression, cost of
hedging has weighted on margins

 The constant bid from variable annuity hedging lifts term structure and skew, particularly for the S&P500 (but also
for other major indices due to relative value traders)

 Volker rule prohibits proprietary trading, which has reduced the number of counterparties for long dated protection
causing skew to rise (particularly at the far end of volatility surfaces)

16
Structured products can cause volatility overshoot Structured products

Hedging of structured products can exaggerate implied volatility moves

 Sale of structured products causes investment banks to be short skew (vanna) and short vega convexity (volga)

 When markets decline, the skew skew position causes sellers to become short vol. To hedge this position traders
buy volatility, lifting implieds.

 As implieds rise, the short vol position increases in size due to vega convexity. Traders then have to buy more vol,
causing a structured product vicious circle and an implied volatility overshoot
1. Market declines 2. Traders become short 3. Traders buy vol
vol as are short skew
Price Implied Vol

Vicious Circle

Time 80% 90% 100% 110% Strike

4. Vega convexity means

traders become shorter vol
as volatility rises

17
 Contents

VOLATILITY TRADING FOR DIRECTIONAL INVESTORS

Call overwriting

Protection buying

Choosing strike of option

ESSENTIAL FACTS OF VOLATILITY TRADING

Volatility is not as expensive as you think

Hedging equity with volatility

Stretching Black-Scholes

Variable annuity and structured products impact on the market

ADVANCED VOLATILITY TRADING

Dividends and correlation

Advanced volatility measures

Term structure and skew

18
Dividends have lower vol, but higher skew Dividends and correlation

Realised dividends are less volatile than equities SX5E 2010 Dividends vs Spot
2010 dividends
200
 Constant dividend yield implies dividends have same vol surface 180
Long SX5E 2010 dividends traded similar to long
as equities 160
SX5E and short SX5E 3000 strike put
140
120
 Realised dividend volatility is 50-70% of equity volatility as: A) 100
companies suppress dividend volatility due to less than 100% 80

payout and; B) equity volatility is too high compared to 60

40
fundamentals 20
0
0 1000 2000 3000 4000 5000
 ATM dividend volatility should be lower than equity ATM implied
2010 (Jan-08 to Sep-08) 2010 (Oct-08 onwards) SX5E

Dividends have higher skew than equities Dividend volatility surface

30%
28%
 While realised dividends are less volatile than equities, implied Low strike implieds are greater
26%
dividends can be more volatile due to imbalances caused by 24%
than high strike implieds

structured product sellers (get longer dividends as market falls) 22%

Implied vol
20%
 Underlying of options on dividends is implied dividends (realised 18%
dividends cannot be traded) 16%
14%

 Dividends have higher skew (3rd moment) than equities as 12%

10%
dividends are cut to zero before equity prices reach zero
35 40 45 50 55 60 65

Negative skew Strike (€)

19
Vega weighted dispersion is best Dividends and correlation

Dispersion traders need to decide how to weight short index & long single stock legs
 Theta (or correlation) weighted: Vega x volatility is equal for both legs. This weighting assumes implieds move by same percentage
amount (eg. if index vol is 20% and increases to 30%, single stock vol of 25% rises to 37.5%). This is the purest dispersion trade as
payout = difference between realised correlation and implied correlation MULTIPLIED by weighted average variance of stocks. Due to
the payout being multiplied by weighted average variance dispersion is short vol of vol (as correlation is correlated to volatility).

 Vega weighted: Vega is equal for both legs. This weighting assumes implieds move by the same absolute amount (eg. if index vol is
20% and increases to 30%, single stock vol of 25% rises to 35%). As correlation is correlated to volatility, the payout of a theta
weighted dispersion trade is (negatively) correlated to volatility. To remove this sensitivity it is better to go long more single stock
volatility, as vega weighted dispersion does. Arguably vega weighted dispersion has a single stock leg 2-5% too large, but over
hedging could be seen as an advantage (as funding etc could dry up in a crisis and the position exited prematurely).

 Gamma weighted: Rarely used, as difficult to justify using more single stock vega than index vega when stocks have higher volatility

Greek protectionTheta-weighted
Type of correction is required against, canVega-weighted
help determine whichGamma-weighted
strategy to choose

Theta 0 pay pay a lot

Vega Short 0 Long

Gamma Very short Short 0

Single-stock vega Less than index Equal to index More than index
20
Yang-Zhang is best measure for small samples Advanced volatility measures

Using intraday prices can improve historical volatility measurement

 When comparing volatility between regions, weekly volatility is better than daily to reduce effect of different time
zones. This is only appropriate for large data samples, if this is not available / practical an advanced volatility
measure is better.

 Close to close volatility needs c20 or more days of data to be accurate, for smaller periods e.g. 5 days close to
close volatility is very noisy. An advanced measuring Open (O), High (H), Low (L) and Close (C) is better for small
samples.

Estimate Prices Taken Handle Drift? Handle Overnight Jumps? Efficiency (max)

Close to close C No No 1

Parkinson HL No No 5.2

Garman-Klass OHLC No No 7.4

Rogers-Satchell OHLC Yes No 8

G-K Yang-Zhang ext OHLC No Yes 8

Yang-Zhang OHLC Yes Yes 14

21
Surfaces move by “square root of time” Term structure and skew

Volatility move weighted by square root of time is roughly constant

 Near dated implieds move more than far dated implieds. Can adjust whole surface by adjusting implieds for
maturity T by “one year implied vol move / Tp”. P is the power of the move.

 Typically volatility move weighted by square root of time is approximately constant (power 0.5).

 Surfaces also sometimes move in parallel (power 0).

 On average surfaces move power 0.44, hence usually square root of time but sometimes parallel.
Volatility moving by square root of time
23%
1 year implied moves half
22% amount of 3 month implied

21% +2%
Implied vol

+1%
20%
-1% -0.5%
19%
4 year implied moves half
18% amount of 1 year implied

17%
3 months 6 months 1 year 2 years 3 years 4 years
Rise in implied Flat term structure Fall in implied

22
Can compare different term structure & skew Term structure and skew

Term structures can be normalised Term structure (normalised)

Term structure x
√(T2T1)
 If assume term structure is a fixed vol for infinite maturity and a (√T2-√T1)
5
square root of time bump, then different term structures can be
compared 0

Dec-06

Apr-07

Aug-07

Dec-07

Apr-08

Aug-08

Dec-08

Apr-09

Aug-09

Dec-09

Apr-10

Aug-10

Dec-10

Apr-11

Aug-11

Dec-11
 Multiplying standard V2 – V1 term structure by -5

√(T2T1)/ (√T2- √ T1) allows different term structures to be -10

compared
-15

 Normalised term structure puts term structure in same units as -20

1 year – 3 month term structure 6 mths - 3 mths (normalised) 1 year - 6 mths (normalised)

Skew (normalised)
Skew In 2010 Q2 skew spiked, particularly at the far end,
(normalised √T) due to changes in US regulation
3.8
Skew multiplied by square root of time constant 3.6
3.4
3.2
 Skew is greater for near dated implieds than far dated
3
2.8
 Can compare different skews when multiply by square root of 2.6
time 2.4
2.2
2
Dec-09 Mar-10 Jun-10 Sep-10 Dec-10 Mar-11 Jun-11 Sep-11 Dec-11
3 month skew (90-100%) 6 month skew (90-100%)

23
24