Presentation PGR9 (2)
-
Upload
mwaseem2011 -
Category
Documents
-
view
19 -
download
0
description
Transcript of Presentation PGR9 (2)
-
Three Essays about Options on Leveraged ExchangeTraded Funds
Yahua Xu
Auckland University of Technology
July 1, 2015
Yahua Xu (AUT) PGR9 July 1, 2015 1 / 53
-
Background Knowledge
Exchange Traded Fund (ETF): An ETF is an investment fundtraded on stock exchanges. It has advantages such as low costs, taxefficiency and stock-like features.
Leveraged ETF (LETF): A LETF is a special type of ETF anddesigned to track a multiple daily return of a target asset.
LETFs tracking two important indices:S&P 500 Index.VIX Short Term Futures Index.
Yahua Xu (AUT) PGR9 July 1, 2015 2 / 53
-
Background Knowledge
Option: A call (put) option gives the holder the right to buy (sell) anunderlying asset at a specified strike price on a specified date:
Payoff of a Call Option: max{st k , 0}.Payoff of a Put Option: max{k st , 0}.
LETF Option: The underlying asset is a LETF.
LETF Option Market:The option on equity LETF was launched in 2011.The option on volatility LETF was launched in 2012.
In my research, I am going to analyze dynamics of the two specified LETFs
and price options written on them.
Yahua Xu (AUT) PGR9 July 1, 2015 3 / 53
-
Background Knowledge
It is from a regulatory notice about the LETFs of the Financial IndustryRegulatory Authority (FIRA) published in June 2009:
While such products may be useful in some sophisticated tradingstrategies, they are highly complex financial instruments that aretypically designed to achieve their stated objectives on a daily basis.Due to the effects of compounding, their performance over longerperiods of time can differ significantly from their stated dailyobjective.
Yahua Xu (AUT) PGR9 July 1, 2015 4 / 53
-
Analyzing Objectives
S&P 500 Index
Options
VIX Index
LETFs
UPRO(+3)
Options
SSO(+2)
Options
SPY(+1)
Options
SH(1)
Options
SDS(2)
Options
SPXU(3)
Options
Yahua Xu (AUT) PGR9 July 1, 2015 5 / 53
-
Analyzing Objectives
VIX Index
VIX Futures
VIX Short Term Futures Index
LETFs
UVXY(+2)
Options
VIXY(+1)
Options
SVXY(1)
Options
Yahua Xu (AUT) PGR9 July 1, 2015 6 / 53
-
Equity LETF Prices
Figure: Equity LETF Prices
0
10
20
30
40
50
60
70
LETFs Price
SPXU (-3) SDS (-2) SH (-1) SPY (+1) SSO (+2) UPRO (+3)
Note. Times series for equity LETF from 2011/06/01 to 2012/06/01.
Yahua Xu (AUT) PGR9 July 1, 2015 7 / 53
-
Research Questions (RQs)
RQ1
How is the empirical performance of the Heston model in regard ofpricing options on equity LETFs?
RQ2
What is the proper dynamic model for volatility LETFs? How is theempirical performance of the model in regard of pricing options onvolatility LETFs?
RQ3
What are the option pricing formulas when jump risk is added intostochastic volatility framework?
Yahua Xu (AUT) PGR9 July 1, 2015 8 / 53
-
Thesis Structure
Stochastic Volatility Framework
Chapter2
Logarithmic Model with StochasticVolatility (LRSV) for Volatility LETFs
Proposing LRSV for the volatilityLETFs;
Empirical analysis about theperformance of LRSV
Chapter1
Heston Model for Equity LETFs
Empirical analysis about theperformance of Heston
Yahua Xu (AUT) PGR9 July 1, 2015 9 / 53
-
Thesis Structure
Stochastic Volatility + Jump Framework
Chapter 3
LRSV + Jump (Volatility LETFs)
LRSV +Stochastic Jump
LRSV +Constant Jump
Heston + Jump (Equity LETFs)
Heston +Stochastic Jump
Heston +Constant Jump
Yahua Xu (AUT) PGR9 July 1, 2015 10 / 53
-
Volatility Skew
Volatility Skew/Smile is the phenomena of options impliedvolatilities varying with moneyness.
It is one of the most stylized facts observed in the option market:
Equity options exhibit negative volatility skew.Volatility options exhibit positive volatility skew.
The stochastic volatility models are capable of capturing the volatilityskew observed in option market, as documented in the literature suchas Heston(1993) and Bates(1996).
Yahua Xu (AUT) PGR9 July 1, 2015 11 / 53
-
Volatility Skew
Figure: Negative Volatility Skew for Equity Option
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
SPY(+1)
Note. Smile for the maturity 0.14 for the LETF SPY(+1) for the day 2011/10/24.
Yahua Xu (AUT) PGR9 July 1, 2015 12 / 53
-
Volatility Skew
Figure: Positive Skew for Volatility Option
0.00
0.50
1.00
1.50
2.00
2.50
VIXY(+1)
Note. Smile for the maturity 0.14 for the LETF VIXY(+1) for the day 2011/10/24.
Yahua Xu (AUT) PGR9 July 1, 2015 13 / 53
-
Chapter I
Chapter I:
Options on Equity LETFs, Calibration and Error
Analysis
Yahua Xu (AUT) PGR9 July 1, 2015 14 / 53
-
Chapter I: Literature Review
Avellaneda and Zhang (2010) is the first to study the dynamics of theequity LETF returns from a theoretical perspective:
The equity LETFs fail to reproduce the multiple return of theunderlying over a long period of time.
The long-run return of a LETF is path-dependent.
They proposed the Heston model for the dynamics of equity LETFs.
Yahua Xu (AUT) PGR9 July 1, 2015 15 / 53
-
Chapter I: Literature Review
The other studies such as Ahn et al. (2012) and Deng et al. (2014) alsoemploy the Heston model for pricing options on the equity LETFs:
Zhang (2010) only utilizes one days underlying ETF option data tocalibrate the Heston model. The information of model fitting is onlyshowed by figures.
Ahn et al. (2012) only utilizes the option data of June, 14, 2013. Inthe calibration, they fix the correlation coefficient without givingthe reason.
Deng et al. (2014) also only uses one days option data, withoutreporting any pricing error.
Yahua Xu (AUT) PGR9 July 1, 2015 16 / 53
-
Chapter I: Literature Review
How to systematically evaluate the performance of an option pricingmodel?
The empirical study of Bakshi et al. (1997) on the performance ofseveral S&P 500 option pricing models provides a good reference:
The daily calibration is rolled based on a set of time series data.Estimated parameters and pricing errors are both provided.
Yahua Xu (AUT) PGR9 July 1, 2015 17 / 53
-
Chapter I: Model and Option Pricing
Assume the price of the underlying asset st follows the Heston model:
dstst
= (r q)dt +vtd1t ,dvt = ( vt)dt + vtd2t ,
with initial value s0 known.
t =(1t ,
2t
)t0 is a Brownian motion with correlation coefficient .
Yahua Xu (AUT) PGR9 July 1, 2015 18 / 53
-
Chapter I: Model and Option Pricing
For a LETF with leverage ratio m, the price lt is related to st as follows:
dltlt
= mdstst
+ (1m)rdt,dvt = ( vt)dt + vtd2t .
Moreover, the log return of lt can be expressed in terms of st , vt and m:
d ln lt = md ln st m2 m
2vtdt
volatility bias
+(1m)rdt,
When |m| > 1, the volatility bias is negative.
Yahua Xu (AUT) PGR9 July 1, 2015 19 / 53
-
Chapter I: Model and Option Pricing
The characteristic function of ln lt can be written as:
X (z) = E[eiz ln lt
l0
]= E
[eiz(m ln st
s0m2m
2
t0 vudu+(1m)rt)
]= e i(1m)zrtE
[eizm ln st
s0iz m2m
2
t0 vudu
]= e i(1m)zrt
(t, zm, z
[m
2 m2
])= e i(1m)zrt+A(t)+B(t)v0+izmx .
where the expressions of A(t) and B(t) can be computed and their valuesare based on the values of z , m, r , q, , , , .
Yahua Xu (AUT) PGR9 July 1, 2015 20 / 53
-
Chapter I: Model and Option Pricing
The option pricing formula can be obtained by inverting the characteristicfunction via Fourier Transform:
c(t, k,m) =kert
pi
0
Re
[X (z)e
iz ln k 1iz(iz 1)
]dzR ,
Yahua Xu (AUT) PGR9 July 1, 2015 21 / 53
-
Chapter I: Data
We have a rich dataset which contains prices of options on a sextet ofequity LETFs. The dataset spans from March 24, 2011 to February 16,2015. The equity LETFs offered by Proshares are as follows:
Table 1: Equity LETFs
Fund Name Ticker Name Leverage Ratio
UltraPro Short S&P 500 ETF SPXU -3UltraShort S&P 500 ETF SDS -2Short S&P 500 ETF SH -1SPDR S&P 500 ETF SPY +1Ultra S&P 500 ETF SSO +2UltraPro S&P 500 ETF UPRO +3
Yahua Xu (AUT) PGR9 July 1, 2015 22 / 53
-
Chapter I: Calibration
We adopt the norm-in-price criterion. At day t, we separately calibrateoptions on the LETFs with different leverage ratios. The objectivefunction is:
minvt ,,,,
1
Nj
i
(Vmarket(ti ,Ki ,mj) Vmodel(ti ,Ki ,mj)
)2,
Yahua Xu (AUT) PGR9 July 1, 2015 23 / 53
-
Chapter I: Calibration
For a given day t:
If the estimated parameters are the same, the model is good!
If different, analyze the distribution of pricing errors corresponding todifferent leverage ratios. It is quite likely the larger the absolute valueof the leverage, the greater the pricing errors are.
Yahua Xu (AUT) PGR9 July 1, 2015 24 / 53
-
Chapter I: Calibration
A rolling calibration using several days option data:
We make a comparison of parameters (, , , ) obtained from eachdays calibration to see if they are stable over time.
If the estimated parameters are stable over time, then the pricingmodel is a robust.
If not, the jump risk may be needed to improve the modelperformance.
Yahua Xu (AUT) PGR9 July 1, 2015 25 / 53
-
Chapter I: Contribution
It is the first to provide a systematic and comprehensive analysis forthe fit of the Heston model in pricing options on equity LETFs.
Yahua Xu (AUT) PGR9 July 1, 2015 26 / 53
-
Chapter II
Chapter II:
Options on Volatility LETFs with Positive
Volatility Skew
Yahua Xu (AUT) PGR9 July 1, 2015 27 / 53
-
Chapter II: Options on Volatility LETFs with PositiveVolatility Skew
The underlying index for the volatility LETFs is the VIX Short-TermFutures Index.
The VIX futures index has a close relationship with the VIX index.
The modeling of VIX dynamics can serve as a good reference for thedynamic model of volatility LETFs.
Yahua Xu (AUT) PGR9 July 1, 2015 28 / 53
-
Chapter II: Literature Review
Detemple and Osakwe (2000) proposed a mean-reverting logarithmicprocess for the VIX dynamics.
Sepp (2008) suggests incorporating stochastic volatility into the VIXdynamics to model the feature of positive skew in VIX options.
Kaeck and Alexander (2010) developed a stochastic volatility ofvolatility model for the logarithmic VIX dynamics.
Yahua Xu (AUT) PGR9 July 1, 2015 29 / 53
-
Chapter II: Literature Review
Bao et al. (2012) adopts the model of Kaeck and Alexander (2010)for the dynamics of VXX (an ETN similar to VIXY (+1) tracking thedaily return of VIX Short Term Futures Index):
Bao et al. (2012) only focuses on the dynamics of the unleveragedvolatility ETF.No research about the volatility Leveraged ETFs has been carried out.
Based on Bao et al.(2012), a logarithmic stochastic volatility modelfor the volatility LETFs is proposed, which is the first model of thiskind.
Yahua Xu (AUT) PGR9 July 1, 2015 30 / 53
-
Chapter II: Model and Option Pricing
Assume the spot price of the VIX Short Term Future Index Vt follows thelogarithmic model with stochastic volatility (LRSV):
d lnVt = ( lnVt)dt +vtd1t ,dvt = v (v vt)dt + vvtd2t ,
with initial value V0 known.
t =(1t ,
2t
)t0 is a Brownian motion with correlation coefficient .
Yahua Xu (AUT) PGR9 July 1, 2015 31 / 53
-
Chapter II: Model and Option Pricing
For a LETF with leverage ratio m, the return xt is based on the underlyingasset return Vt :
dxtxt
= mdVtVt
+ (1m)rdt,dvt = v (v vt)dt + vvtd2t ,
with the initial value x0 = x known.
Moreover, the log return of xt can be expressed in terms of Vt , vt and m:
d ln xt = md lnVt m2 m
2vtdt
volatility bias
+(1m)rdt,
When |m| > 1, the volatility bias is negative.
Yahua Xu (AUT) PGR9 July 1, 2015 32 / 53
-
Chapter II: Model and Option Pricing
The characteristic function of ln xt can be written as:
X (z) = E[eiz ln xt
x0
]= E
[eiz(m ln Vt
V0m2m
2
t0 vudu+(1m)rt)
]= e i(1m)zrtE
[eizm ln Vt
V0iz m2m
2
t0 vudu
]= e i(1m)zrt
(t, zm, z
[m
2 m2
])= eA(t)+B(t)v+C(t)x .
The expressions of A(t), B(t) and C (t) can be computed and they arebased on the values of z , m, r , q, , , , v , v and v .
Yahua Xu (AUT) PGR9 July 1, 2015 33 / 53
-
Chapter II: Model and Option Pricing
The option pricing formula can be obtained by inverting the characteristicfunction via Fourier Transform:
c(t, k,m) =kert
pi
0
Re
[X (z)e
iz ln k 1iz(iz 1)
]dzR ,
Yahua Xu (AUT) PGR9 July 1, 2015 34 / 53
-
Chapter II: Data
We have a rich dataset which contains prices of options on a set ofvolatility LETFs. The dataset spans from March 19, 2012 to October 31,2014. The volatility LETFs offered by Proshares are as follows:
Table 1: Volatility LETFs
Fund Name Ticker Name Leverage Ratio
Short VIX Short-Term Futures ETF SVXY -1VIX Short-Term Futures ETF VIXY +1Ultra VIX Short-Term Futures ETF UVXY +2
Yahua Xu (AUT) PGR9 July 1, 2015 35 / 53
-
Chapter II: Calibration
The calibration procedure is similar to the previous part. Firstly, thenorm-in-price criterion is adopted. We implement calibration separately foroptions written on different LETFs on the same day t. The objectivefunction is:
minvt ,,,,v ,v ,v
1
Nj
i
(Vmarket(ti ,Ki ,mj) Vmodel(ti ,Ki ,mj)
)2,
Yahua Xu (AUT) PGR9 July 1, 2015 36 / 53
-
Chapter II: Calibration
For a given day t:
If the estimated parameters are the same, the model is good!
If different, analyze the distribution of pricing errors corresponding todifferent leverage ratios. It is quite likely the larger the absolute valueof the leverage, the greater the pricing errors are.
Yahua Xu (AUT) PGR9 July 1, 2015 37 / 53
-
Chapter II: Calibration
A rolling calibration using several days option data:
We make a comparison of parameters (vt , , , , v , v , v ) obtainedfrom each days calibration to see if they are stable over time.
If the estimated parameters are stable over time, then the pricingmodel is a robust.
If not, the jump risk may be needed to improve the modelperformance.
Yahua Xu (AUT) PGR9 July 1, 2015 38 / 53
-
Chapter II: Contribution
It is the first model proposed for dynamics of volatility LETFs.
An extended calibration on a time series of option data will be carriedout to assess the models performance.
Yahua Xu (AUT) PGR9 July 1, 2015 39 / 53
-
Chapter III
Chapter III:
Analytical Extensions of Pricing Options on Equity
and Volatility LETFs
Yahua Xu (AUT) PGR9 July 1, 2015 40 / 53
-
Chapter III: Analytical Extensions of Pricing Options onEquity and Volatility LETFs
In the first two chapters, we focus on utilizing the stochastic volatilitymodels addressing the characteristic of volatility skew observed in theLETF option market:
The Heston model is adopted for the dynamics of an equity LETF toincorporate the negative volatility skew.The LRSV model is adopted for the dynamics of a volatility LETF toincorporate the positive volatility skew.
Yahua Xu (AUT) PGR9 July 1, 2015 41 / 53
-
Chapter III: Literature Review
Some research declares that the inclusion of jumps into the stochasticvolatility model may better explain the volatility skew:
Bates (1996) is the first to introduce a jump component into thestochastic volatility model to analyze the Deutsche Mark Optionmarket.Bakshi et al. (1997) conjecture that jumps in volatility may benecessary to fully explain the volatility smile observed in the S&P 500option market.Duffie et al. (2000) also found that the inclusion of jumps improves themodel fitting.Broadie et al. (2007) utilize an extensive data sample of S&P 500futures option. Both time series and cross section test show that jumpin volatility can improve the model fit.
Yahua Xu (AUT) PGR9 July 1, 2015 42 / 53
-
Chapter III: Literature Review
The specification of jumps is furthermore important:
Typical jump models such as Bates (1996) usually assume the jumpintensity is constant.Bates (2000) found that the time-varying jump risk and stochasticvolatility work together to better explain the negative skewness.Eraker (2004) also assumes that the jump arrival intensity isstate-dependent.
However, evidence supporting the specification of jumps is mixed.Therefore, we are going to analyze both cases.
Yahua Xu (AUT) PGR9 July 1, 2015 43 / 53
-
Chapter III: Heston Model Extension with Constant JumpIntensity
The dynamics of the underlying asset price st is specified as follows:
dstst
= (r q)dt +vtd1t ,dvt = ( vt)dt + vtd2t + JdNt ,
with initial value s0 known.
1t and 2t has correlation coefficient , Nt is a Poisson Process with
constant intensity . The magnitude J of jumps is an exponential randomvariable, i.e. J exp ( 1J ).
Yahua Xu (AUT) PGR9 July 1, 2015 44 / 53
-
Chapter III: Heston Model Extension with Stochastic JumpIntensity
The dynamics of the underlying asset price st is specified as follows:
dstst
= (r q)dt +vtd1t ,dvt = ( vt)dt + vtd2t + JdNt ,dt = ( t)dt +
td
3t ,
with initial value s0 known.
1t and 2t has correlation coefficient , and
1t and
3t has correlation
coefficient 0. Nt is a Poisson Process with stochastic intensity t . Themagnitude J of jumps is an exponential random variable, i.e. J exp ( 1J ).
Yahua Xu (AUT) PGR9 July 1, 2015 45 / 53
-
Chapter III: Heston Model Extension with Stochastic JumpIntensity
We denote by lt the spot price of the LETF. The characteristic function ofln ltl0 can be written as:
X (z) = E[eiz ln lt
l0
]= E
[eiz(m ln st
s0m2m
2
t0 vudu+(1m)rt)
]= e i(1m)zrt
(t, zm, z
[m
2 m2
]).
Yahua Xu (AUT) PGR9 July 1, 2015 46 / 53
-
Chapter III: LRSV Extension with Constant Jump Intensity
We assume the underlying asset price Vt follows the logarithmic modelwith stochastic volatility (LRSV), and is specified as follows:
d lnVt = ( lnVt)dt +vtd1t ,dvt = v (v vt)dt + vvtd2t + JdNt ,
with initial value V0 known.
1t and 2t has correlation coefficient . Nt is a Poisson Process with
constant intensity . The magnitude J of jumps is an exponential randomvariable, i.e. J exp ( 1J ).
Yahua Xu (AUT) PGR9 July 1, 2015 47 / 53
-
Chapter III: LRSV Extension with Stochastic JumpIntensity
We assume the price Vt follows the logarithmic model with stochasticvolatility (LRSV), and is specified as follows:
d lnVt = ( lnVt)dt +vtd1t ,dvt = v (v vt)dt + vvtd2t + JdNt ,dt = ( t)dt +
td
3t ,
with initial value V0 known.
1t and 2t has correlation coefficient , and
1t and
3t has correlation
coefficient 0. Nt is a Poisson Process with stochastic intensity t . Themagnitude J of jumps is an exponential random variable, i.e. J exp ( 1J ).
Yahua Xu (AUT) PGR9 July 1, 2015 48 / 53
-
Chapter III: LRSV Extension with Stochastic JumpIntensity
We denote by lt the spot price of the LETF. The characteristic function ofln ltl0 can be written as:
X (z) = E[eiz ln lt
l0
]= E
[eiz(m ln st
s0m2m
2
t0 vudu+(1m)rt)
]= e i(1m)zrt
(t, zm, z
[m
2 m2
]).
Yahua Xu (AUT) PGR9 July 1, 2015 49 / 53
-
Chapter III: Option Pricing
The option pricing formula can be obtained by inverting the characteristicfunction via Fourier Transform:
c(t, k,m) =kert
pi
0
Re
[X (z)e
iz ln k 1iz(iz 1)
]dzR ,
Yahua Xu (AUT) PGR9 July 1, 2015 50 / 53
-
Chapter III: Sensitivity Analysis
We will carry out sensitivity analysis for the models to see the impactof the parameters.
Yahua Xu (AUT) PGR9 July 1, 2015 51 / 53
-
Chapter III: Contribution
It is the first to introduce jump risk to the stochastic volatilityframework for the dynamics of equity and volatility LETFs.
Yahua Xu (AUT) PGR9 July 1, 2015 52 / 53
-
Thank You
Yahua Xu (AUT) PGR9 July 1, 2015 53 / 53