Lecture 8

Stephen G. Hall

ARCH and GARCH

REFS

Athoroughintroduction
ARCHModelsBollerslevT,EngleRFandNelsonDB
HandbookofEconometricsvol4.orUCSDDiscussionpaper
no93.49.(availableonmywebsite)

Aquicksurvey
Cuthbertson Hall and Taylor
Until the early 80s econometrics had focused almost solely
on modelling the means of series, ie their actual values.
Recently however we have focused increasingly on the
importance of volatility, its determinates and its effects on
meanvalues.

A key distinction is between the conditional and

unconditionalvariance.

the unconditional variance is just the standard measure of

thevariance

var(x) =E(x -E(x))2

the conditional variance is the measure of our uncertainty
aboutavariablegivenamodelandaninformationset.

cond var(x) =E(x-E(x| ))2 Conditional

variance
this is the true measure of uncertainty

variance

mean
StylisedFactsofassetreturns

i) Thick tails, they tend to be leptokurtic

ii)Volatility clustering, Mandelbrot, large changes tend to be

followed by large changes of either sign
iii)Leverage Effects, refers to the tendency for changes in stock
prices to be negatively correlated with changes in volatility.
iv)Non-trading period effects. when a market is closed information
seems to accumulate at a different rate to when it is open. eg stock
price volatility on Monday is not three times the volatility on
Tuesday.
v) Forcastableevents,volatilityishighatregulartimessuch
asnewsannouncementsorotherexpectedevents,orevenat
certaintimesofday,eglessvolatileintheearlyafternoon.
vi)Volatility and serial correlation. There is a suggestion of an
inverse relationship between the two.

vii) Co-movements in volatility. There is considerable evidence

that volatility is positively correlated across assets in a market and
even across markets
Engle(1982)ARCHModel
Auto-Regressive Conditional Heteroscedasticity

Yt X t t t ~ N (0, t )
2

i 2 t i ( L) 2
2
t
i 1

t t t
2 2
define

( L) t
2 2
t

an AR(q) model for squared innovations.

note as we are dealing with a variance

0 i 0 all i
even though the errors may be serially uncorrelated they are not
independent, there will be volatility clustering and fat tails.

if the standardised residuals

zt t / t
are normal then the fourth moment for an ARCH(1) is

E ( t ) / E ( t ) 3(1 ) /(1 3 ) if 3 1
4 2 2 2 2 2
GARCH (Bollerslev(1986))

In empirical work with ARCH models high q is often required, a

more parsimonious representation is the Generalised ARCH model
q p
2 t i 2 t i j 2 t j
i 1 j 1

( L) ( L) 2 2

define t t 2 2
t

2
t ( ( L) ( L)) ( L) t 2

whichisanARMA(max(p,q),p)modelforthesquared
innovations.
Thisiscovariancestationaryifalltherootsof

( L) ( L) 1

lieoutsidetheunitcircle,thisoftenamountsto

(1) (1) 1

If this becomes an equality then we have an Integrated GARCH

model (IGARCH)
NelsonsEGARCHmodel

thiscapturesbothsizeandsigneffectsinanon-linear
formulation

q p
log( 2 t ) i ( zt i (| zt i | E | zt i |)) j log( 2 t j )
i 1 j 1
Non-linearARCHmodelNARCH

thisthenmakesthevariancedependonboththesizeandthe
signofthevariancewhichhelpstocaptureleveragetype
effects.

q p

t i | t i | j

t j
i 1 j 1
ThresholdARCH(TARCH)

Largeeventstohaveaneffectbutnoeffectfromsmall
events

q p
2 t ( i I ( t i 0) i ( t i 0)) 2 t i ) j 2 t j
i 1 j 1

Manyotherversionsarepossiblebyaddingminor
asymmetriesornon-linearitiesinavarietyofways.
Allofthesearesimplyestimatedbymaximumlikelihood
usingthesamebasiclikelihoodfunction,assuming
normality,

T
log( L) ( log( t ) t / t )
2 2 2

i 1
ARCHinMEAN(G)ARCH-M

Manyclassicareasoffinancesuggestthatthemeanofa
relationshipwillbeaffectedbythevolatilityoruncertaintyof
aseries.EngleLilienandRobins(1987)allowforthis
explicitlyusinganARCHframework.

yt xt 2
t t

q p
2
t i 2
t i j 2
t j
i 1 j 1

typicallyeitherthevarianceorthestandarddeviationare
includedinthemeanrelationship.
oftenfinancestressestheimportanceofcovarianceterms.
Theabovemodelcanhandlethisifyisavectorandwe
interpretthevariancetermasacompletecovariancematrix.
Thewholeanalysiscarriesoverintoasystemframework
 Nonnormalityassumptions

WhilethebasicGARCHmodelallowsacertainamountof
leptokurticbehaviourthisisofteninsufficienttoexplainreal
worlddata.Someauthorsthereforeassumearangeof
distributionsotherthannormalitywhichhelptoallowforthe
fattailsinthedistribution.
v

tDistribution
Thetdistributionhasadegreesoffreedomparameterwhich
allowsgreaterkurtosis.Thetlikelihoodfunctionis

lt ln((0.5(v 1))(0.5v) 1 (v 2) 1 / 2 (1 zt (v 2) 1 ) ( v1) / 2 ) 0.5 ln( 2 t )

where Fis the gamma function and v is the degrees of

v
freedom asthistendstothenormal distribution
IGARCH.
ThestandardGARCHmodel

2
t ( L) ( L)
2 2

iscovariancestationaryif

(1) (1) 1

ButStrictstationaritydoesnotrequiresuchastringent
restriction(Thatisthattheunconditionalvariancedoesnot
dependont),infactweoftenfindinestimationthat

(1) (1) 1
thisisthentermedanIntegratedGARCHmodel(IGARCH),
Nelsonhasestablishedthatasthissatisfiestherequirement
forstrictstationarityitisawelldefinedmodel.

HoweverwemaysuspectthatIGARCHismoreaproductof
omittedstructuralbreaksthantheresultoftrueIGARCH
behavior.
MultivariateModels

IngeneraltheGarchmodellingframeworkmaybeeasily
extendedtoamultivariateframeworkwhere

Et ( t ' t ) t

howevertherearesomepracticalproblemsinthechoiceof
theparameterisationofthevarianceprocess.
AdirectextensionoftheGARCHmodelwouldinvolveavery
largenumberofparameters.

Theconditionalvariancecouldeasilybecomenegativeeven
whenalltheparametersarepositive.

Thechosenparameterisationshouldallowcausalitybetween
variances.
VectorARCH
let vech denote the matrix stacking operation

a b
vech (a b d )
b d

ageneralextensionoftheGARCHmodelwouldthenbe

vech( t ) W A( L)vech( t 1 't 1 ) B( L)vech(t 1 )

thisquicklyproduceshugenumbersofparameters,for
p=q=1andn=5thereare465parameterstoestimatehere.
OnesimplificationusedistheDiagonalGARCHmodelwhere
AandBaretakentobediagonal,butthisassumesaway
causalityinvariancesandco-persistence.Weneedstill
furthercomplexrestrictionstoensurepositivedefiniteness
inthecovariancematrix.

Amoretractablealternativeistostate

q p
t V 'V A'i t i 't i Ai B ' j t j B j
i 1 j 1

wecanfurtherreducetheparameterisationbymakingAand
Bdiagonal.
FactorARCH
SupposeavectorofNserieshasacommonfactorstructure.
Suchas;

yt B t t

wherearethecommonfactorsand

~ N (0, ) E ( 't ) t
thentheconditionalcovariancematrixofyisgivenby

Covt 1 ( yt ) t t B t B'
Or k
t t i 'i k
i 1
Sogivenasetoffactorswemayestimateaparsimonious
modelforthecovariancematrixoncewehaveparameterized

Oneassumptionisthatweobserveasetoffactorswhich
causethevariance,thenwecansimplyusethese.E.G.GDP,
interestrates,exchangerates,etc.

anotherassumptionisthateachfactorhasaunivariate
GARCHrepresentation.

K K
t k ( k t 1 't 1 'k ) k ( k t 1 'k )
k 1 k 1