ARCH and MGARCH models

Christopher F Baum

EC 823: Applied Econometrics

Boston College, Spring 2014

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 1 / 38
 ARCH models Single-equation models

ARCH models

Heteroskedasticity can occur in time series models, just as it may in a

cross-sectional context. It has the same consequences: the OLS point
estimates are unbiased and consistent, but their standard errors will be
inconsistent, as will hypothesis test statistics and confidence intervals.

We may prevent that loss of consistency by using

heteroskedasticity-robust standard errors. The “Newey–West” or HAC
standard errors available from newey in the OLS context or ivreg2 in
the instrumental variables context will be robust to arbitrary
heteroskedasticity in the error process as well as serial correlation.

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 2 / 38
 ARCH models Single-equation models

The most common model of heteroskedasticity employed in the time

series context is that of autoregressive conditional heteroskedasticity,
or ARCH. As proposed by Nobel laureate Robert Engle in 1982, an
ARCH model starts from the premise that we have a static regression
model
yt = β0 + β1 zt + ut
and all of the Gauss–Markov assumptions hold, so that the OLS
estimators are BLUE. This implies that Var (ut |Z ) is constant. But even
when this unconditional variance of ut is constant, we may have time
variation in the conditional variance of ut :

E(ut2 |ut−1 , ut−2 , . . . ) = E(ut2 |ut−1 ) = α0 + α1 ut−1

2

so that the conditional variance of ut is a linear function of the squared

value of its predecessor.

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 3 / 38
 ARCH models Single-equation models

If the original ut process is serially uncorrelated, the variance

conditioned on a single lag is identical to that conditioned on the entire
history of the series. We can rewrite this as
2
ht = α0 + α1 ut−1
√
where ut = ht vt , vt ∼ (0, 1). This formulation represents the
ARCH(1) model, in which a single lagged u 2 enters the ARCH
equation. A higher-order ARCH equation would include additional lags
of u 2 . To ensure a positive variance, α0 > 0 and α1 > 0. When α1 > 0,
the squared errors are positively serially correlated even though the ut
themselves are not.

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 4 / 38
 ARCH models Single-equation models

Since we could estimate this equation and derive OLS b which are
BLUE, why should we be concerned about ARCH? First, we could
derive consistent estimates of b which are asymptotically more
efficient than the OLS estimates, since the ARCH structure is no
longer a linear model.

Second, the dynamics of the conditional variance are important in

many contexts: particularly financial models, in which movements in
volatility are themselves important. Many researchers have found
“ARCH effects" in higher-frequency financial data, and to the extent to
which they are present, we may want to take advantage of them. We
may test for the existence of ARCH effects in the residuals of a time
series regression by using the command estat archlm. The null
hypothesis is that of no ARCH effects; a rejection of the null implies the
existence of significant ARCH effects, or persistence in the squared
errors.

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 5 / 38
 ARCH models Single-equation models

The ARCH model is inherently nonlinear. If we assume that the ut are

distributed Normally, we may use a maximum likelihood procedure
such as that implemented in Stata’s arch command to jointly estimate
its mean and conditional variance equation.

The ARCH model has been extended to a generalized form which has
proven to be much more appropriate in many contexts. In the simplest
example, we may write
2
ht = α0 + α1 ut−1 + γ1 ht−1

which is known as the GARCH(1,1) model since it involves a single lag

of both the ARCH term and the conditional variance term. We must
impose the additional constraint that γ1 > 0 to ensure a positive
variance.

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 6 / 38
 ARCH models Single-equation models

We may also have a so-called ARCH-in-mean model, in which the ht

term itself enters the regression equation. This sort of model would be
relevant if we had a theory that suggests that the level of a variable
might depend on its variance, which may be very plausible in financial
markets contexts or in terms of, say, inflation, where we often presume
that the level of inflation may be linked to inflation volatility. In such
instances we may want to specify a ARCH- or GARCH-in-mean model
and consider interactions of this sort in the conditional mean (level)
equation.

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 7 / 38
 ARCH models Alternative GARCH specifications

Alternative GARCH specifications

A huge literature on alternative GARCH specifications exists; many of

these models are preprogrammed in Stata’s arch command, and
references for their analytical derivation are given in the Stata manual.

One of particular interest is Nelson’s (1991) exponential GARCH, or

EGARCH. He proposed:
∞
X    

log ht = η + λj t−j − E t−j + θt−j
   
j=1

which is then parameterized as a rational lag of two finite–order

polynomials, just as in Bollerslev’s GARCH.

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 8 / 38
 ARCH models Alternative GARCH specifications

Advantages of the EGARCH specification include the positive nature of

ht irregardless of the estimated parameters, and the asymmetric
nature of the impact of innovations: with θ 6= 0, a positive shock will
have a different effect on volatility than will a negative shock, mirroring
findings in equity market research about the impact of “bad news” and
“good news” on market volatility. For instance, a simple EGARCH(1,1)
model will provide a variance equation such as
 p 
log ht = −δ0 + δ1 zt−1 + δ2 zt−1 − 2/π  + δ3 log ht−1
 

where zt = t /σt , which is distributed as N(0, 1).

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 9 / 38
 ARCH models Alternative GARCH specifications

Nelson’s model is only one of several extensions of GARCH that allow

for asymmetry, or consider nonlinearities in the process generating the
conditional variance: for instance, the threshold ARCH model of
Zakoian (1990) and the Glosten et al. model (1993).

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 10 / 38
 ARCH models Implementation

Stata 12 provides a suite of commands to estimate time series models

in the ARCH (Autoregressive Conditional Heteroskedasticity) family.
The command arch is used to estimate single-equation models. Its
options allow the specification of over a dozen models from the
literature, including ARCH, GARCH, ARCH-in-mean, GARCH with
ARMA errors, EGARCH (exponential GARCH), TARCH (threshold
ARCH), GJR (Glosten et al., 1993), SAARCH (simple asymmetric
ARCH), PARCH (power ARCH), NARCH (nonlinear ARCH), APARCH
(asymmetric power ARCH) and NPARCH (nonlinear power ARCH).

Errors may be specified as Gaussian, t, or GED (generalized error

distribution).

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 11 / 38
 ARCH models Implementation

To estimate an ARCH model, you give the arch varname command,

followed by (optionally) the independent variables in the mean
equation and the options indicating the type of model. For instance, to
fit a GARCH(1,1) to the mean regression of cpi on wage,
arch cpi wage, arch(1) garch(1)

It is important to note that a GARCH(2,1) model would be specified

with the option arch(1/2). If the option was given as arch(2), only
the second-order term would be included in the conditional variance
equation.

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 12 / 38
 ARCH models Implementation

A test for ARCH effects in a linear regression can be conducted with

the estat archlm command. Using Stata’s urate dataset of
monthly unemployment rates for several US states:
. webuse urates, clear
. qui reg D.tenn LD.tenn
. estat archlm, lags(3)
LM test for autoregressive conditional heteroskedasticity (ARCH)

lags(p) chi2 df Prob > chi2

3 11.195 3 0.0107

H0: no ARCH effects vs. H1: ARCH(p) disturbance

The LM test indicates the presence of significant ARCH effects.

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 13 / 38
 ARCH models Implementation

We estimate a GARCH(1,1) model:

. arch D.tenn LD.tenn, arch(1) garch(1) nolog vsquish
ARCH family regression
Sample: 1978m3 - 2003m12 Number of obs = 310
Distribution: Gaussian Wald chi2(1) = 9.39
Log likelihood = 127.4172 Prob > chi2 = 0.0022

OPG
D.tenn Coef. Std. Err. z P>|z| [95% Conf. Interval]

tenn
tenn
LD. .2129528 .0694996 3.06 0.002 .076736 .3491695
_cons -.0155809 .0085746 -1.82 0.069 -.0323868 .0012251

ARCH
arch
L1. .1929262 .0675544 2.86 0.004 .0605219 .3253305
garch
L1. .7138542 .0923551 7.73 0.000 .5328415 .894867
_cons .0028566 .0016481 1.73 0.083 -.0003736 .0060868

Following estimation, we may use predict with the variance option

to produce the conditional variance series.
Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 14 / 38
 ARCH models Implementation

Conditional variance from GARCH(1,1)

.1
Conditional variance, one-step

.08

.06

.04

.02

1980m1 1985m1 1990m1 1995m1 2000m1 2005m1

Month

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 15 / 38
 ARCH models Implementation

We may also fit a model with additional variables in the mean equation:
. arch D.tenn LD.tenn LD.indiana LD.arkansas, arch(1) garch(1) nolog vsquish
ARCH family regression
Sample: 1978m3 - 2003m12 Number of obs = 310
Distribution: Gaussian Wald chi2(3) = 41.31
Log likelihood = 135.1611 Prob > chi2 = 0.0000

OPG
D.tenn Coef. Std. Err. z P>|z| [95% Conf. Interval]

tenn
tenn
LD. .1459972 .0723994 2.02 0.044 .004097 .2878974
indiana
LD. .1751591 .047494 3.69 0.000 .0820727 .2682455
arkansas
LD. .1170958 .0757688 1.55 0.122 -.0314083 .2655999
_cons -.0078106 .0087075 -0.90 0.370 -.0248769 .0092558

ARCH
arch
L1. .1627143 .0712808 2.28 0.022 .0230064 .3024221
garch
L1. .6793291 .1388493 4.89 0.000 .4071896 .9514687
_cons .0042064 .0026923 1.56 0.118 -.0010704 .0094832

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 16 / 38
 ARCH models Implementation

Following estimation, we may test hypotheses on the coefficients of the

conditional variance equation: for instance, that they sum to unity,
indicating integrated GARCH:
. test [ARCH]L.arch + [ARCH]L.garch == 1
( 1) [ARCH]L.arch + [ARCH]L.garch = 1
chi2( 1) = 2.30
Prob > chi2 = 0.1297

In this case, that hypothesis cannot be rejected at 90%.

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 17 / 38
 ARCH models Multiple-equation models

Multiple-equation GARCH models

Multivariate GARCH models allow the conditional covariance matrix of

the dependent variables to follow a flexible dynamic structure and allow
the conditional mean to follow a vector autoregressive (VAR) structure.

The general MGARCH model can be written as

yt = Cxt + εt
1/2
εt = Ht νt

where yt is a m-vector of dependent variables, C is a m × k parameter

matrix, xt is a k-vector of explanatory variables, possibly including lags
1/2
of yt , Ht is the Cholesky factor of the time-varying conditional
covariance matrix Ht , and νt is a m-vector of zero-mean, unit-variance
i.i.d. innovations.

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 18 / 38
 ARCH models Multiple-equation models

In this general framework, Ht is a matrix generalization of univariate

GARCH models. For example, a general MGARCH(1,1)) model may
be written as:

vech(Ht ) = s + A vech(εt−1 ε0t−1 ) + B vech(Ht−1 )

where the vech(·) function returns a vector containing the unique
elements of its matrix argument. The various parameterizations of
MGARCH provide alternative restrictions on H, the conditional
covariance matrix, which must be positive definite for all t.

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 19 / 38
 ARCH models Implementation

Implementation

Stata’s mgarch command estimates multivariate GARCH models,

allowing both the conditional mean and conditional covariance matrix
to be dynamic. Four commonly used parameterizations are supported:
the diagonal vech (DVECH) model
the constant conditional correlation (CCC) model
the dynamic conditional correlation (DCC) model
the varying conditional correlation (VCC) model

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 20 / 38
 ARCH models Parameterizations

Alternative parameterizations differ in terms of flexibility, allowing for

more complex H processes, and parsimony, allowing the model to be
specified with fewer parameters.

The oldest and simplest parameterization is the diagonal vech

(DVECH) of Bollerslev, Engle, Wooldridge (JPE, 1988), which restricts
the A and B matrices to be diagonal. The number of parameters grows
rapidly with the size of the model. For instance, there are 3m(m + 1)/2
parameters in a DVECH(1, 1) with m series.

Despite the large number of parameters, the diagonal structure implies

that each conditional variance and covariance depends only on its own
past, and not on past values of other elements. For a DVECH(1,1),

hij,t = sij + aij εi,t−1 εj,t−1 + bij hij,t−1

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 21 / 38
 ARCH models Parameterizations

Conditional correlation models

Conditional correlation (CC) models use nonlinear combinations of

univariate GARCH models to represent the conditional covariances in
H. They often have less difficulty with satisfying the restrictions on the
estimated H, and their number of parameters grows more slowly than
in the DVECH specification.

In CC models, Ht is decomposed into a matrix of conditional

correlations Rt and a diagonal matrix of conditional variances, Dt :
1/2 1/2
Ht = Dt Rt Dt

implying that hij,t = ρij,t σi,t σj,t , where σi,t is modeled as a univariate
GARCH process. The CC models differ in how they parameterize Rt .

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 22 / 38
 ARCH models Parameterizations

The constant CC model of Bollerslev (REStat, 1990) specifies the

correlation matrix as time invariant:
q
hij,t = ρij hii,t hjj,t

where the diagonal elements follow univariate GARCH processes, and

ρij is a time-invariant weight.

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 23 / 38
 ARCH models Parameterizations

Engle’s (JBES, 2002) extension, the dynamic CC model, allows the

conditional correlations (technically, quasicorrelations) to follow a
GARCH(1,1)-like process:
q
hij,t = ρij,t hii,t hjj,t

where now the ρ parameters follow a dynamic process.

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 24 / 38
 ARCH models Parameterizations

Tse and Tsui’s (JBES, 2002) variant, the varying CC model, expresses
the conditional correlations using a time-invariant component, a
measure of recent correlations among the residuals, and last period’s
values. It differs from the DCC model in terms of the dynamic process
followed by the ρ parameters.

In Stata, the four MGARCH specifications are invoked with the mgarch
command, with a first argument being the model specification: dvech,
ccc, dcc or vcc.

To illustrate, we use Stata’s stocks dataset, and model daily Toyota

and Honda equity returns as AR(1) processes with the ccc and dcc
specifications.

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 25 / 38
 ARCH models Parameterizations

The estimated Toyota mean and conditional variance equations:

. webuse stocks, clear
(Data from Yahoo! Finance)
. mgarch ccc (toyota honda = L.toyota L.honda), arch(1) garch(1) nolog vsquish
Constant conditional correlation MGARCH model
Sample: 1 - 2015 Number of obs = 2014
Distribution: Gaussian Wald chi2(4) = 4.34
Log likelihood = 11602.61 Prob > chi2 = 0.3620

Coef. Std. Err. z P>|z| [95% Conf. Interval]

toyota
toyota
L1. -.03374 .032697 -1.03 0.302 -.097825 .030345
honda
L1. -.005188 .0288975 -0.18 0.858 -.0618261 .0514502
_cons .0004523 .0003094 1.46 0.144 -.0001542 .0010587

ARCH_toyota
arch
L1. .0661046 .0095018 6.96 0.000 .0474814 .0847279
garch
L1. .916793 .0117942 77.73 0.000 .8936769 .9399092
_cons 4.50e-06 1.19e-06 3.78 0.000 2.17e-06 6.83e-06

...

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 26 / 38
 ARCH models Parameterizations

The estimated Honda mean and conditional variance equations, and

correlation estimate:

honda
toyota
L1. -.0066352 .0343028 -0.19 0.847 -.0738675 .0605971
honda
L1. -.0332976 .0316213 -1.05 0.292 -.0952743 .028679
_cons .0006128 .0003394 1.81 0.071 -.0000524 .0012781

ARCH_honda
arch
L1. .0498417 .0080311 6.21 0.000 .0341009 .0655824
garch
L1. .9321435 .0111601 83.52 0.000 .9102701 .9540168
_cons 5.26e-06 1.41e-06 3.73 0.000 2.50e-06 8.02e-06

Correlation
toyota
honda .7176095 .0108477 66.15 0.000 .6963483 .7388707

In this CCC specification, the sizable correlation indicates the

interaction between the two equations’ error processes.

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 27 / 38
 ARCH models Parameterizations

In the DCC model, the diagonal elements of Ht are modeled as

univariate GARCH models. The off-diagonal elements are modeled as
nonlinear functions of the diagonal terms:
q
hij,t = ρij,t hii,t hjj,t

where ρij,t follows a dynamic process, rather than being constrained to

be constant as in the CCC specification.

Two additional parameters, λ1 and λ2 , are adjustment parameters that

govern the evolution of the conditional quasicorrelations. They must be
positive and sum to less than one. A test for the sum of these
parameters equalling zero tests the DCC model against the special
case of the CCC model.

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 28 / 38
 ARCH models Parameterizations

The DCC model may be written as

yt = Cxt + t
1/2
t = Ht νt
1/2 1/2
Ht = Dt Rt Dt
Rt = diag(Qt )−1/2 Qt diag(Qt )−1/2
Qt = (1 − λ1 − λ2 )R + λ1 ˜t−1 ˜0t−1 + λ2 Qt−1

where Dt is a diagonal matrix of conditional variances,

Rt is a matrix of conditional quasicorrelations,
−1/2
and ˜t is a vector of standardized residuals, Dt t .
R is a weighted average of the unconditional VCE of the standardized
residuals and the unconditional mean of Qt .

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 29 / 38
 ARCH models Parameterizations

With the DCC specification:

. mgarch dcc (toyota honda = L.toyota L.honda), arch(1) garch(1) nolog vsquish
Dynamic conditional correlation MGARCH model
Sample: 1 - 2015 Number of obs = 2014
Distribution: Gaussian Wald chi2(4) = 4.81
Log likelihood = 11624.54 Prob > chi2 = 0.3074

Coef. Std. Err. z P>|z| [95% Conf. Interval]

toyota
toyota
L1. -.0346653 .0319267 -1.09 0.278 -.0972404 .0279098
honda
L1. -.0069742 .0284872 -0.24 0.807 -.0628081 .0488597
_cons .000373 .0003108 1.20 0.230 -.0002362 .0009821

ARCH_toyota
arch
L1. .0629146 .0093309 6.74 0.000 .0446263 .0812029
garch
L1. .9208039 .0116908 78.76 0.000 .8978904 .9437175
_cons 4.32e-06 1.16e-06 3.72 0.000 2.04e-06 6.60e-06

...

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 30 / 38
 ARCH models Parameterizations

honda
toyota
L1. .0030378 .0339118 0.09 0.929 -.0634281 .0695036
honda
L1. -.0367691 .0316091 -1.16 0.245 -.0987219 .0251836
_cons .0005624 .000341 1.65 0.099 -.0001059 .0012307

ARCH_honda
arch
L1. .0536899 .008511 6.31 0.000 .0370087 .0703711
garch
L1. .928433 .0115932 80.08 0.000 .9057107 .9511554
_cons 5.43e-06 1.44e-06 3.77 0.000 2.61e-06 8.26e-06

Correlation
toyota
honda .7264858 .0132659 54.76 0.000 .7004852 .7524864

Adjustment
lambda1 .0528653 .014217 3.72 0.000 .0250005 .0807301
lambda2 .746622 .0746374 10.00 0.000 .6003354 .8929085

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 31 / 38
 ARCH models Parameterizations

In both the CCC and DCC specifications, the mean equations indicate
that lagged daily returns of both stocks are not significant determinants
of current returns, as is implied by efficient markets theory.

There are very significant GARCH effects in both specifications. A

sizable correlation parameter appears, as it did in the CCC
specification. The magnitudes of the lambda parameters indicate that
the evolution of the conditional covariances depends more on their
past values than on lagged residuals’ innovations.

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 32 / 38
 ARCH models Parameterizations

The VCC model of Tse and Tsui can be written as

yt = Cxt + t
1/2
t = Ht νt
1/2 1/2
Ht = Dt Rt Dt
Rt = (1 − λ1 − λ2 )R + λ1 Ψt−1 + λ2 Rt−1

where Dt is a diagonal matrix of conditional variances,

Rt is a matrix of conditional correlations,
R is the matrix of means to which the dynamic process reverts, and
Ψt is the rolling estimator of the covariance matrix of the standardized
residuals ˜t .

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 33 / 38
 ARCH models Parameterizations

We illustrate the VCC model with two companies’ shares, assumed to

have no mean equation per previous findings, but with their ARCH and
GARCH parameters constrained to be equal.

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 34 / 38
 ARCH models Parameterizations

. constraint 1 _b[ARCH_toyota:L.arch] = _b[ARCH_nissan:L.arch]

. constraint 2 _b[ARCH_toyota:L.garch] = _b[ARCH_nissan:L.garch]
. mgarch vcc (toyota nissan =, noconstant), arch(1) garch(1) constraints(1 2) n
> olog vsquish
Varying conditional correlation MGARCH model
Sample: 1 - 2015 Number of obs = 2015
Distribution: Gaussian Wald chi2(.) = .
Log likelihood = 11282.46 Prob > chi2 = .
( 1) [ARCH_toyota]L.arch - [ARCH_nissan]L.arch = 0
( 2) [ARCH_toyota]L.garch - [ARCH_nissan]L.garch = 0

Coef. Std. Err. z P>|z| [95% Conf. Interval]

ARCH_toyota
arch
L1. .0797459 .0101634 7.85 0.000 .059826 .0996659
garch
L1. .9063808 .0118211 76.67 0.000 .883212 .9295497
_cons 4.24e-06 1.10e-06 3.85 0.000 2.08e-06 6.40e-06

ARCH_nissan
arch
L1. .0797459 .0101634 7.85 0.000 .059826 .0996659
garch
L1. .9063808 .0118211 76.67 0.000 .883212 .9295497
_cons 5.91e-06 1.47e-06 4.03 0.000 3.03e-06 8.79e-06

...
Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 35 / 38
 ARCH models Parameterizations

...
Correlation
toyota
nissan .6720056 .0162585 41.33 0.000 .6401394 .7038718

Adjustment
lambda1 .0343012 .0128097 2.68 0.007 .0091945 .0594078
lambda2 .7945548 .101067 7.86 0.000 .596467 .9926425

The validity of the constraints could be established with a likelihood

ratio test against the unconstrained model.

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 36 / 38
 ARCH models Parameterizations

We can produce predictions of the three series in the conditional VCE,

ex post and ex ante. Notice that the ex ante predictions (beyond the
sample period, ending in day 2015) quickly converge in the absence of
additional information, as these are dynamic forecasts.
. tsappend, add(50)
. predict H*, variance dynamic(2016)
. lab var H_toyota_toyota CV_Toy
. lab var H_nissan_nissan CV_Nis
. lab var H_nissan_toyota CCov_Toy_Nis
. lab var t "Trading Day"
. tsline H* in 1800/l, leg(rows(1)) xline(2015) ylab(,angle(0))

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 37 / 38
 ARCH models Parameterizations

.001

.0008

.0006

.0004

.0002

0
1800 1850 1900 1950 2000 2050
Trading Day

CV_Toy CCov_Toy_Nis CV_Nis

Christopher F Baum (BC / DIW) ARCH and MGARCH models Boston College, Spring 2014 38 / 38