Scribd
Upload
73 views

Breusch Godfrey Test of Autocorrelation

The document discusses the Breusch-Godfrey (BG) Test for autocorrelation, which allows for nonstochastic regressors and higher-order autoregressive schemes. It outlines the steps for conducting the test, including estimating a regression model and analyzing residuals to determine the presence of serial correlation. Practical considerations and limitations of the test, such as the specification of lag length and assumptions about homoskedasticity, are also addressed.

Uploaded by

raghavsharma123er5
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
73 views

Breusch Godfrey Test of Autocorrelation

The document discusses the Breusch-Godfrey (BG) Test for autocorrelation, which allows for nonstochastic regressors and higher-order autoregressive schemes. It outlines the steps for conducting the test, including estimating a regression model and analyzing residuals to determine the presence of serial correlation. Practical considerations and limitations of the test, such as the specification of lag length and assumptions about homoskedasticity, are also addressed.

Uploaded by

raghavsharma123er5
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 7

Introduction BG Test Practical Points References

Breusch-Godfrey Test of Autocorrelation

Anuj Goyal

Assistant Professor
Shaheed Bhagat Singh College
University of Delhi

May 23, 2025


Introduction BG Test Practical Points References

Introduction

Breusch (1978) and Godfrey (1978) have developed a test for


autocorrelation that is general in the sense that it allows for:
1. nonstochastic regressors, such as the lagged values of the
regressand;
2. higher-order autoregressive schemes (so far, we have only
looked at AR(1) scheme; but it allows for AR(1), AR(2), etc.);
and
3. simple or higher-order moving averages of white noise error
terms (for instance, in the model Yt = β1 + β2 Xt + ut , the
error term ut can be represented as ut = ϵt + λ1 ϵt−1 + λ2 ϵt−2 ,
which represents a three-period moving average of the white
noise error term ϵt ).
Introduction BG Test Practical Points References

The BG Test (also known as the LM test)-I


To illustrate, we use the simple two-variable regression model
(though multiple regressors can be there as well as the lagged
values of the regressand can be used as regressors in the model).
• Let our model be:

Yt = β1 + β2 Xt + ut (1)
• Assume that the error term ut follows the kth order
autoregressive, AR(k), scheme:

ut = ρ1 ut−1 + ... + ρk ut−k + ϵt (2)

where ϵt is a white noise error term. Notice that Eq(2) is


simply an extension of AR(1) scheme earlier discussed in the
classroom.
• The null hypothesis H0 : ρ1 = ... = ρk = 0. That is, there is
no serial correlation of any order.
Introduction BG Test Practical Points References

The BG Test-II
The BG-Test involves the following steps:
1. Estimate Eq.(1) by OLS and obtain uˆt .
2. Regress uˆt on Xt and ut−1
ˆ ,...,ut−k
ˆ . Note that to run this
regression, we have only (n − k) observations (since we have k
lagged values of the residuals). In short, we run the following
regression:
uˆt = α1 + α2 Xt + ρ1 ut−1
ˆ + ... + ρk ut−k
ˆ + ϵt (3)
and obtain R 2 from this auxiliary regression.
3. If the sample size is large enough (technically, infinite),
Breusch (1978) and Godfrey (1978) have shown that
(n − k)R 2 ∼ χ2k (4)
That is, asymptotically, (n − k)R 2 obtained from the auxiliary
regression (3) follows the chi-square distribution with k
degrees of freedom. If (n − k)R 2 > χ2k , we reject H0 .
Introduction BG Test Practical Points References

Some Practical Points about the BG Test-I


The following practical points about the BG test may be noted:
• The reason that the original regressor X is included in the
auxiliary regression (3) is to allow for the fact that X may not
be strictly nonstochastic. But if it is strictly nonstochastic, it
may be omitted from the auxiliary regression. You may refer to
Wooldridge (2003, p. 386) to explore this further.
• The regressors included in the regression model may contain
lagged values of the regressand Y (that is, Yt−1 , Yt−2 , etc
may appear as explanatory variables in the model). In
contrast, in the Durbin-Watson test that we discussed in the
classroom, one of the restrictions is that there cannot be any
lagged values of the regressand in the model.
• In the regression (3), ϵt satisfies all the classical assumptions.
• If k = 1 (that is, if it is the case of first-order autoregression),
then the BG test is known as Durbin’s M test.
Introduction BG Test Practical Points References

Some Practical Points about the BG Test-II

• A drawback of the BG test is that the value of k (the length of


the lag) cannot be specified a priori, and would require some
experimentation. Sometimes, we could use some information
criteria (such as Akaike and Schwarz information criteria; you
are encouraged to read the review paper by Stoica and Selén
(2004) to explore various information criteria) to select the lag
length.
• Given the values of the X variable(s) and the lagged values of
u, the test assumes that u in Eq. (2) is homoskedastic.
Introduction BG Test Practical Points References

Breusch, T. S. (1978).Testing for Autocorrelation in Dynamic


Linear Models. Australian Economic Papers, 17.
https://doi.org/https://doi.org/10.1111/j.1467-
8454.1978.tb00635.x
Godfrey, L. G. (1978).Testing Against General Autoregressive and
Moving Average Error Models when the Regressors include
Lagged Dependent Variables. Econometrica, 46.
https://doi.org/https://doi.org/10.2307/1913829
Stoica, P., & Selén, Y. (2004).Model-Order Selection: A Review of
Information Criterion Rules. IEEE Signal Processing
Magazine, 21, 36–47.
https://api.semanticscholar.org/CorpusID:17338979
Wooldridge, J. M. (2003). Introductory Econometrics: A Modern
Approach. South-Western Publishing Co.

You might also like