Scribd
Upload
16 views

Write Up On A RCH and GARCH Model

Uploaded by

GOURAB GHOSH
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
16 views

Write Up On A RCH and GARCH Model

Uploaded by

GOURAB GHOSH
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 4

ARCH and GARCH Model

ARCH MODEL ( Auto Regressive Conditional Hetroscedastic Model)

What is an ARCH Model ?

ARCH is a statistical model to analyze volatility in time series in order to forecast future
volatility. This model is mainly used in the financial world, to estimate risk, by providing a
model of volatility that are closely resembles real markets. ARCH Model shows that it may
be possible that periods of high volatility are followed by more high volatility and periods
of low volatility are followed by more low volatility. In practice, this means that volatility of
variance tends to cluster, which is useful to investors when considering the risk of holding
an asset over different time periods. The ARCH concept was developed by Robert F. Engle
in the 1980s. ARCH immediately improved financial modeling, resulting in Engle winning
the Nobel Prize in Economics, in 2003.

Consider the K variable model

Yt = β1+ β2X2t+ β3X3t+……………………..+ βkXkt+ ut …………. (1)

Where ut is normally distributed with zero mean and

Variance = α0 + α1ut-12 ……………………………(2)

Mean (Yt) = β1+ β2X2t+ β3X3t+……………………..+ βkXkt …..1(a)

According to (2), variance of u at time t is not constant, as made in the usual regression
analysis, but in turn depends on squared disturbance at t-1, thereby giving appearance of serial
correlation.

Variance of ut is conditional; conditional on the squared disturbance at t-1, i.e., ut-12

In general, the variance of u at time t may depend not only on one lagged squared
disturbance term but also on several lagged squared disturbance terms as follows:

Var (ut ) = σ t 2 = α0 + α1ut-12 + α2ut-22 + …………. + αput-p 2 ………………(3)

Equation (2) is ARCH Model of order 1

Equation ( 3) is ARCH Model of order p

Mean Equation: Equation (1a ) is called mean equation .


Variance Equation: Equation (2) or (3) is called variance equation

Method of testing the presence of ARCH

The presence of ARCH is tested by examining the validity of null hypothesis

Ho = α1 = α2 = …….αp =0

Test proposed by Engle (1982)

Since we do not directly observe σ t 2 , to test this hypothesis, Engle proposed running the
following auxiliary regression :

e t 2 = α0 + α1et-12 + α2et-22 + …………. + αpet-p 2 ………………. (4)

where, e t = Y t – YHAT

Where, YHAT is the estimated value of Y from the regression (1) by using OLS.

The null hypothesis may be tested by

(i) usual F test


(ii) Alternatively, one can compute TR2, where R2, is the coefficient of determination from
the auxiliary regression (4). T is the number of observations. It has been shown that for
large sample TR2, has a χ2 distribution with degrees of freedom equal to p, where p
is the number of lags in (4).

Rejection rule

If the estimated value of the test statistics, TR2 = χ* is greater than the tabulated value of χ2 at
the chosen level of significance and the degrees of freedom p, we reject H0, and conclude that
ARCH is present.

Estimation and inference

The ARCH models are estimated by the method of maximum likelihood (ML).

Properties of ARCH estimators:

It can be shown that:

(i) The estimator of ARCH coefficients are normally distributed in large sample,
(ii) t- statistics in large samples have standard normal distributions and
(iii) Confidence intervals can be constructed as ML estimate ± standard errors.
GARCH Model (Generalised Auto Regressive Conditional Hetroscedastic Model )

One of the limitations of ARCH specification as used in equation (3) is that it


looked more like a moving average specification than auto regression. From this
a new idea was born which was to include the lagged conditional variance terms
as autoregressive term. This idea was worked out by Bollerslev(1986) who
developed GARCH Model.

The simplest GARCH (1,1) model is written as :

Var (ut ) = σ t 2 = α0 + α1ut-12 + β1 σ t-1 2 + ……………………(5)

In this model, the conditional variance of u at time t depends not only on the
squared terms of the previous period as in ARCH (1), but also on the conditional variance of
the previous period.

(5) is called GARCH (1,1) as it contained only one lag of the squared terms of the previous
period, and only one lag of the conditional variance of the previous period.

Generalized GARCH Model: GARCH (p,q)

The model can be generalized to GARCH (p,q) model where there will be p lagged terms of
squared errors and q terms of the lagged conditional variances, so that

Var (ut)
p q
=σt2=α0+ ∑ ❑αiut-i2+ ∑ ❑ βi σ t-i 2 + ……………………(5)
i=1 i=1

If we set p=1 and q= 0, then we have GARCH ( 1, 0) Model; the expression of


GARCH ( 1, 0) is same as the first order ARCH Model; i.e. ARCH (1) .

Hence when all βi = 0 , the GARCH p, q model reduces to ARCH(p) model.

Estimation and inference of GARCH Model

As in the case of ARCH Model, the GARCH models are estimated by the method of
maximum likelihood (ML).

Properties of GARCH estimators:

It can be shown that:


(i) The estimator of GARCH coefficients are normally distributed in large sample,
(ii) t- statistics in large samples have standard normal distributions and
(iii) Confidence intervals can be constructed as ML estimate± standard errors.

Advantage of GARCH Model

The advantage of the GARCH model is that it is much more flexible and more capable of
matching a wide variety of patterns, in particular, financial volatility.

Generalization of GARCH Model

The GARCH model also got revised subsequently , and we have models like

 GARCH-M, (GARCH Mean),


 T-GARCH (Threshold GARCH) ,
 EGARCH Exponential GARCH etc.

I am not discussing it here as it is not under the preview of present syllabus.

References

W.Enders : Applied Econometric Time Series, Wiley student edition, 2004.

You might also like