# Quantitative Finance I

Modeling Volatility (Lecture 4)
Winter Semester 2011/2012 by Lukas Vacha and Jozef Baruník
* If viewed in .pdf format - for full functionality use Mathematica 7 notebook (.nb) version of this .pdf

Characteristics of Volatility
Volatility is one of the most important concepts in finance, as it measures the risk of financial assets. Altough volatility is not directly observable, and can be just estimated, it has some important characteristics: (i) volatility clusters - there are periods of high/low volatility (ii) volatility evolves over time continuously (jumps in volatility are rare) (iii) volatility does not diverge to infinity (iv) negative and positive price shocks tend to have different impacts on volatility (leverage effect) (v) heavy-tailed, non-gausian distribution

† Historical Volatility
Most straightforward way to measure volatility is to estimate time-series of variance on "rolling samples" For zero-mean variable (let us say return), this would be:
2 2 2 s2 = Irt-1 + rt-2 + ... + rt-q M ë q t

, where q is the latest observation used. This method may produce abrupt changes in estimates.

† Exponential volatility
Alternatively, exponential moving average (EMA) estimator of volatility can be used. It uses all data points, and recent observations carry larger weights. (weights are exponentially decreasing).
2 s2 = H1 - lL rt-1 + ls2 , t t-1

where initial value could be unconditional variance in a historical sample. In most financial applications, l=0.94 is used.

† Example on S&P 500 index data (change ticker in the code to get other data)
Note that characteristics of volatility can be observed from returns of S&P 500. Moreover, ACF plot suggests us, that returns are not serially correlated, but ACF plot of squared returns shows, that they are not independent. By modeling volatility, we will attempt to capture this kind of dependece.

08 0.1 0.02 0.06 0.2 QF_I_Lecture4.05 -0.2 0.04 0.1 -0.0 -0.10 -0.1 0.05 0.2 0 5 10 15 20 25 30 ACF of r2 0.9L 0.00 -0.00 1980 1990 2000 .10 0.15 -0.0 -0.1 -0.2 0.20 1980 1990 2000 2010 ACF of r 0.cdf Prices 1500 1000 500 0 1980 1990 2000 2010 Returns 0.2 0 5 10 15 20 25 30 † Conditional Standard Deviation estimated by EMA (or EWMA) S&P standard deviation estimate with EMA Hl=0.

a1 L. but it is dependent. with 3 a1 2 < 1 r3 kurtHet L = EAet 4 E EAet E 2 2 =3 1-a1 2 1-3 a1 2 . 1L s2 = a0 + a1 s2 . † Kurtosis of ARCH(1) processes The fourth moment of ARCH(1) process: EAet 4 E = 3 a0 2 H1-a1 L 2 1-a1 2 1-3 a1 2 . 1 Thus excess kurtosis is positive and the tail distribution of at is heavier than normal distribution. (i. process is stationary. ARCH(1) model is most common from ARCH(m) models in financial theory. Basic idea is that the mean-corrected return is serially uncorrelated. the model can be expressed in terms of information set available in time t: Ft (Engle 1982).QF_I_Lecture4. unconditional variance is a0 ê H1 . a1 ¥ 0. Conditional heteroscedastic models are used for modeling s2 as an volatility estimate t by "allowing" heteroscedasticity (time-variation) and capturing that dependence. ARCH(1) The first model that provides a systematic framework for volatility modeling is Autoregressive conditional heteroscedasticity (ARCH) model. or with minor lower order serial correlations. Unconditional mean of at is zero. but dependent. et is conditionally normally distributed) yt Ft-1 ~ NH0. ht L 2 ht = a0 + a1 yt-1 Example of other notation: yt = c + et st et ~ NH0. if a1 < 1. and that the dependence can be described by a simple function of lagged values: at = st et s2 = a0 + a1 a2 t t-1 Adding the assumption of normality.cdf 3 Conditional Heteroscedastic Models Log returns of an asset at time is either serially uncorrelated. t t-1 where a0 > 0.e. the fourth moment is finite if 3 a2 < 1.

4 QF_I_Lecture4.4 Example : a1 = 0.5 .1 Example : a1 = 0.cdf Example : a1 = 0.

9 † Examples of ARCH(1) artificial processes Sample ARCHH1L ARCHH1L volatility ACF function of a2 t PACF function of a2 t a0 a1 0. a ¥ 0 for i > 0. t t-1 t-m where at is mean corrected return at = rt .QF_I_Lecture4. a > 0. + am a2 . s2 = a0 + a1 a2 + .823 New Random Case 10 5 0 -5 -10 0 100 200 300 400 500 Example ...82 0. {et < is i.Simulation of ARCH process ARCH(m) at = st et .cdf 5 Example : a1 = 0.mt . somteimes denoted as h . .i.d random variables with zero mean and variance 1.innovations.

that large past squared shocks 9a2 =i=1 imply large conditional variance s2 . a0 > 0. with the null hypothesis of no ARCH effects: 2 2 et = a0 + I⁄s=1 as et-s M + nt q H0 : a1 =.cdf at = st et ... ai ¥ 0 for i > 0. or t-i t volatility.6 QF_I_Lecture4.1< Sample ARCHHmL ARCHHmL volatility ACF function of a2 t PACF function of a2 t New Random Case 0.4 0.innovations. ai ¥ 0 for i > 0. One can easily observe.2 -0.mt . 0..i.d random variables with zero mean and variance 1.ARCH effect m † Examples of ARCH(m) artificial processes Note that you have to input a0 > 0.2 0. s2 = a0 + a1 a2 + . large shocks tend to be followed by large shocks . am < Input Parameters of ARCHHmL: am 80..4 0 5 10 15 20 25 30 Test for ARCH effects in time series We use Lagrange Multiplier test (LM test).5. t t-1 t-m where at is mean corrected return at = rt .R2 ~ c2 HpL .. + am a2 .0 -0. in form 8a0 . {et < is i. . somteimes denoted as ht .. Thus under ARCH. = as = 0 LM = N. a1 .

so unconditional variance of at is finite. It is assumed.s) As ARCH model requires many parameters to adequately describe volatility process of returns.. Note.mt be the mean-corrected log return.QF_I_Lecture4..iL = e2 h h+l-i if l . the 1-step ahead forecast of s2 is: h+1 s2 H1L = a0 + a1 e2 + . + am e2 h h h+1-m l-step ahead forecast will be: s2 HlL = a0 + ⁄m ai s2 Hl . a0 > 0. b j ¥ 0. s2 = a0 + ⁄m ai a2 + ⁄s b j s2 j . whereas its conditional variance s2 evolves ⁄i=1 t over time.cdf 7 Forecasting with ARCH Forecasts of the ARCH can be obtained recursively at origin h.iL. (ii) model gives us description of the conditional variance. that assets respond differently to positive and negative shocks.Generalized autoregressive conditional heteroscedasticity (GARCH) model. ai ¥ 0. and Hai + bi L < 1. but does not explain the real behavior and source of variance (iii) ARCH models tend to overpredict the volatility as they respond slowly to large isolated shocks. maxHm. t i=1 t-i j=1 twhere {et < is i. h i=1 h where s2 Hl . that mean equation is adequately described by ARMA. but empirical testing shows. Weaknesses of ARCH Models ARCH model has also following weaknesses: (i) assumes that positive and negative shocks have the same effect on volatility. that GARCH model reduces to a pure ARCH(m) model for s = 0.s) model is described by following equation: at = st et . GARCH(m.d random variables with zero mean and variance 1.i § 0. letting at = rt .i.sL . extension is used . GARCH models GARCH(m.

2.1) GARCH(1.1L GARCHH1. t t-1 t-1 where Ha1 + b2 L < 1.05.1) artificial processes Note that condtition ⁄i=1 Hai + bi L < 1 must be met.1) process with a0 = 0. a1 = 0.1L volatility ACF function of a2 t PACF function of a2 t am bs 80.cdf † Example GARCH(m.3< 80.2.0 0 100 200 300 400 500 GARCH(1. thus volatility clustering is quite well captured.5. It takes form of: at = st et . as the dependencies are mostly very weak. s2 = a0 + ai a2 + b1 s2 .8 QF_I_Lecture4.1) is most common for financial applications.4 maxHm.s) artificial processes Sample GARCHH1.5 0. 0.sL . 0. Large past shocks of return and volatility in t-1 gives large shocks to volatility in time t. † Example GARCH(1. Initial values are set to simulate GARCH(1.5 1.0 0.1.2< New Random Case Export Simulated Series 1. 0. b1 = 0.

s2 = a0 + ai a2 + b1 s2 .1)-M artificial processes Note.581 0. To model this. t t-1 t-1 where m and c are constant. returns are positively skewed. that with positive risk premium c.1L GARCHH1. we use GARCH in mean (GARCH-M) model.1L volatility ACF function of a2 t PACF function of a2 t a0 a1 b1 0.296 New Random Case Export Simulated Series 10 5 0 -5 -10 0 100 200 300 400 500 GARCH-M model The return of a security may sometimes depend directly on volatility.QF_I_Lecture4.667 0. c is also called risk premium † Example GARCH(1.M is formalized as: rt = m + cs2 + at t at = st et . GARCH(1.cdf 9 Sample GARCHH1.1) . as they are positively related to its past volatility .

1). it is not so easy to determine the t order. .for most series. we remove also sample mean from the data.38 0.). or by Lagrange multiplier test (LM test). This can be done either using Ljung-Box statistics.) by ARMA model .1L-M ACF function of a2 t PACF function of a2 t risk premium c a0 a1 b1 -0.1). but empirical findings can serve again.10 QF_I_Lecture4. For GARCH. If the statistic is significant.5 0.e. PACF functions.375 0. and PACF of a2 is used to determine order of ARCH effect.318 New Random Case Export Simulated Series 2 0 -2 -4 0 100 200 300 400 500 ARIMA-GARCH Models Common way to build ARCH model is to remove any linear dependencies in the data (i. GARCH(2. or ACF. that in most financial applications. and use residuals for testing the ARCH effects.cdf Sample GARCHH1. we use lower order models such as GARCH(1. then conditional heteroscedasticity of at is detected.

b1 L a2 t t-1 t-1 . -0.Integrated GARCH IGARCH models are unit-root (integrated) GARCH models.9.1. 0.qL-GARCHH1.1L ACF ARIMA@p. their key feature is.1D parameters a0 -5 a1 -10 0 100 200 300 400 500 New Random Case b1 Export Simulated Series IGARCH .4< 80.d.cdf 11 † Example ARIMA(p.5< 5 0 GARCH@1.d)-GARCH(1.q. IGARCH(1.QF_I_Lecture4.dD parameters process is unit-root stationary 10 difference ARHpL parameters MAHqL parameters 80. that past squared shocks is persistent.1) is formalized as: at = st et . s2 = a0 + b1 s2 + H1 .1) artificial processes Sample ARIMAHp.

cdf † Example IGARCH(1. s2 M n Empirical example Homework #4 Deadline: Tue 15. t t-1 where 8ht < ~ iidH0.611 0.12 QF_I_Lecture4.1) artificial processes IGARCHH1.cz :] Exercise 1 [: Estimate ARMA-(G)ARCH model on the real stock market data (use dataset which contains (G)ARCH effect). 8nt < ~ iidI0. .2010. et = gIs2 M ht .10. 3:00 pm Homework may be returned in class.6 New Random Case Export Simulated Series 800 600 400 200 0 0 100 200 300 400 500 Stochastic Volatility Model rt = mt + et . or sent via email to vachal@utia. 1L.1L Simulated series Simulated Volatility a0 b1 0. t s2 = a0 + a1 s2 + nt .cas.

QF_I_Lecture4. .cdf 13 :] Exercise 1 [: Estimate ARMA-(G)ARCH model on the real stock market data (use dataset which contains (G)ARCH effect). * Please include your computations program with your results.