Autocorrelation

Some Questions To Deal With

• What Happens If the Error Terms Are Correlated?

• What is the nature of autocorrelation?
• What are the theoretical and practical consequences of
autocorrelation?
• How does one know that there is autocorrelation in any
given situation?
• How does one remedy the problem of autocorrelation?
Introduction
• In cross-section studies, data are often collected on the basis of a random sample
of cross-sectional units, such as households or firms.
• Therefore, there is no prior reason to believe that the error term pertaining to one
household or firm is correlated with the error term of another household or firm.
• If by chance such a correlation is observed in cross-sectional units, it is called
spatial autocorrelation, that is, correlation in space rather than over time.
• Autocorrelation:
If we are dealing with time series data, for the observations in such data follow a
natural ordering over time so that successive observations are likely to exhibit
intercorrelations, especially if the time interval between successive observations
is short, such as a day, a week, or a month rather than a year.
 What is Autocorrelation and Its Nature?

• Put simply, the classical model assumes that the disturbance term relating to any observation is not
influenced by the disturbance term relating to any other observation.
• For example (time series auto correlation), if we are dealing with quarterly time series data
involving the regression of output on labor and capital inputs and if, say, there is a labor strike
affecting output in one quarter, there is no reason to believe that this disruption will be carried over to
the next quarter. That is, if output is lower this quarter, there is no reason to expect it to be lower next
quarter.
• Another Example (Spatial auto correlation): if we are dealing with cross-sectional data involving
the regression of family consumption expenditure on family income, the effect of an increase of one
family’s income on its consumption expenditure is not expected to affect the consumption
expenditure of another family.
Cont…

• In this situation, the disruption caused by a strike this quarter may very well affect
output next quarter, or the increases in the consumption expenditure of one family
may very well prompt another family to increase its consumption expenditure.
Difference Between Autocorrelation and Serial Correlation

• Although it is now a common practice to treat the terms autocorrelation

and serial correlation synonymously, some authors prefer to distinguish
the two terms.
• Tintner (1965) defines autocorrelation as “lag correlation of a given
series with itself, lagged by a number of time units,’’ whereas serial
correlation refers to “lag correlation between two different series.”
• Thus, correlation between two time series such as 𝒖𝟏 , 𝒖𝟐 , ..., 𝒖𝟏𝟎 and
𝒖𝟐 , 𝒖𝟑 , ..., 𝒖𝟏𝟏 , where the former is the latter series lagged by one time
period, is autocorrelation, whereas correlation between time series such
as 𝒖𝟏 , 𝒖𝟐 , ..., 𝒖𝟏𝟎 and 𝒗𝟐 , 𝒗𝟑 , ..., 𝒗𝟏𝟏 , where u and v are two different
time series, is called serial correlation.
Patterns of Auto- and Non-Autocorrelation,
Why Does Serial Correlation Occur?
Why Does Serial Correlation Occur?
Why Does Serial Correlation Occur?

• Lags:
o In a time series regression of consumption expenditure on income, it is not
uncommon to find that the consumption expenditure in the current period depends,
among other things, on the consumption expenditure of the previous period.

o Above regression is known as autoregression because one of the explanatory

variables is the lagged value of the dependent variable.
o The rationale for the above model is that consumers do not change their
consumption habits readily for psychological, technological, or institutional reasons.
o Now if we neglect the lagged term in Eq. (12.1.7), the resulting error term will
reflect a systematic pattern due to the influence of lagged consumption on current
consumption.
Why Does Serial Correlation Occur?

• Manipulation of Data:
o In time series regressions involving quarterly data, such data are usually derived from the monthly
data by simply adding three monthly observations and dividing the sum by 3.
o This averaging introduces smoothness into the data by dampening the fluctuations in the monthly
data.
o Therefore, the graph plotting the quarterly data looks much smoother than the monthly data, and this
smoothness may itself lend to a systematic pattern in the disturbances, thereby introducing
autocorrelation.
o Another source of manipulation is interpolation or extrapolation of data.
o For example, the Census of Population is conducted every 10 years in this country, the last being in
2000 and the one before that in 1990. Now if there is a need to obtain data for some year within the
inter-census period 1990–2000, the common practice is to interpolate on the basis of some ad hoc
assumptions.
o All such data “massaging’’ techniques might impose upon the data a systematic pattern that might not
exist in the original data.
Why Does Serial Correlation Occur?
Why Does Serial Correlation Occur?

Proof of error term 𝒗𝒕 in Eq. (4) is autocorrelated.

Why Does Serial Correlation Occur?
Mean, Variance, and Covariance of error term (𝒖𝒕 ) and OLS estimator (𝜷𝟐 )
Cont..
Consequences of Using OLS in the Presence of Autocorrelation

• As in the case of heteroscedasticity, in the presence of autocorrelation the

OLS estimators are still linear unbiased as well as consistent and
asymptotically normally distributed, but they are no longer efficient (i.e.,
minimum variance).
• What then happens to our usual hypothesis testing procedures if we
continue to use the OLS estimators?
OLS Estimation Allowing for Autocorrelation

The implication of this finding for hypothesis

testing is that we are likely to declare a coefficient
statistically insignificant even though in fact (i.e.,
based on the correct GLS procedure) it may be.

In the fig since b2 lies in the OLS confidence

interval, we could accept the hypothesis that true
β2 is zero with 95 percent confidence.

But if we were to use the (correct) GLS confidence

interval, we could reject the null hypothesis that
true β2 is zero, forb2 lies in the region of rejection.
OLS Estimation Disregarding Autocorrelation
How do we know if our data suffer from autocorrelation??
How do we know if our data suffer from autocorrelation??
Double-Log Model
Linear Model

Qualitatively, both the models give similar results. In both cases the estimated
coefficients are “highly” significant, as indicated by the high t values.

• How reliable are the results given in the above two models if there is autocorrelation?
• As stated previously, if there is autocorrelation, the estimated standard errors are biased, as a result of
which the estimated t ratios are unreliable.
• We obviously need to find out if our data suffer from autocorrelation.
Detecting Autocorrelation

1. Graphical Method
2. Runs Test
3. Durbin–Watson d Test
4. Breusch–Godfrey (BG) Test
Detecting Autocorrelation: Graphical Method

𝒖𝟐𝒕 ) against time can

ෝ 𝒕 or (ෝ
• A visual examination of 𝒖
provide useful information about autocorrelation.
That is called time sequence plot
• Alternatively, we can plot the standardized residuals
against time which is shown in figure.

• Examining the time sequence plot given in Figure, we

ෝ 𝒕 and the standardized 𝒖
observe that both 𝒖 ෝ 𝒕 exhibit a
pattern suggesting that perhaps 𝒖𝒕 are not random.
Detecting Autocorrelation: Graphical Method

• To see this differently, we

can plot 𝒖 ෝ 𝒕 against 𝒖
ෝ 𝒕−𝟏 ,
that is, plot the residuals at
time t against their value at
time (t - 1), a kind of
empirical test of the AR(1)
scheme.
• If the residuals are
nonrandom, we should
obtain pictures similar to
those shown in following
Figure.
Detecting Autocorrelation: Runs Test

• If we carefully examine this Figure, we

notice a peculiar feature: Initially, we have
several residuals that are negative, then
there is a series of positive residuals, and
then there are several residuals that are
negative.
• If these residuals were purely random,
could we observe such a pattern?
• Intuitively, it seems unlikely.
• This intuition can be checked by the so-
called runs test, sometimes also known as
the Geary test, a nonparametric test.
Detecting Autocorrelation: Explaining Runs Test
• To explain the runs test, let us simply note down
the signs (+ or -) of the residuals obtained from
the regression, which are given in the first column
of Table.
• Thus there are 8 negative residuals, followed by
21 positive residuals, followed by 11 negative
residuals, followed by 3 positive residuals,
followed by 3 negative residuals, for a total of 46
observations.
• We now define a run as an uninterrupted
sequence of one symbol or attribute, such as + or
−. We further define the length of a run as the
number of elements in it.
• In the sequence shown above in bracket, there are
5 runs: a run of 8 minuses (i.e., of length 8), a run
of 21 pluses (i.e., of length 21), a run of 11
minuses (i.e., of length 11), a run of 3 pluses (i.e.,
of length 3), and a run of 3 minuses (i.e., of
length 3).
Cont…
Are the 5 runs observed in our example
consisting of 46 observations too many
or too few compared with the number
of runs expected in a strictly random
sequence of 46 observations?

• If there are too many runs, it would

mean that in our example the residuals
change sign frequently, thus indicating
negative serial correlation.

• Similarly, if there are too few runs,

they may suggest positive
autocorrelation. A priori, then, would
indicate positive correlation in the
residuals.
Obviously, this interval does not include 5. Hence, we can reject
the hypothesis that the residuals in our wages–productivity
regression are random with 95% confidence.
Durbin–Watson d Test
Important Assumptions of Durbin–Watson d Test
Durbin–Watson d Test: Decision Rule
Durbin–Watson d Test: Decision Rule
Breusch–Godfrey (BG) Test: A General Test of Autocorrelation
• Breusch and Godfrey have developed a test of 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, such as AR(1), AR(2), etc.; and

(3) simple or higher-order moving averages of white noise error terms, such as ε𝒕
• BG test is also known as the LM test as it is based on the Lagrange multiplier principle.
Steps in Breusch–Godfrey (BG) Test
Remedies
What to Do When You Find Autocorrelation?

We have four options as follows:

1. Try to find out if the autocorrelation is pure autocorrelation and not the result of
mis-specification of the model.
2. If it is pure autocorrelation, one can use appropriate transformation of the
original model so that in the transformed model we do not have the problem of
(pure) autocorrelation. As in the case of heteroscedasticity, we will have to use some
type of generalized least-square (GLS) method.
3. In large samples, we can use the Newey–West method to obtain standard errors of
OLS estimators that are corrected for autocorrelation. This method is actually an
extension of White’s heteroscedasticity-consistent standard errors method.
4. In some situations we can continue to use the OLS method.
Remedies
A. Model Mis-Specification versus Pure Autocorrelation
Cont…
Remedies
B. Correcting for (Pure) Autocorrelation: Generalized Least Squares (GLS) Method
Once we got to know that its pure autocorrelation not specification error, then what to do?

• The remedy depends on the knowledge one has about the nature of interdependence
among the disturbances, that is, knowledge about the structure of autocorrelation.
Remedies
(B.1) When ρ is Known
Cont…
 Remedies
(B.2) When ρ Is Not Known

Since the error term (ϵ𝒕 ) in last Equation is free from (first-order) serial correlation, to run the regression all one has to
do is form the first differences of both the regressand and regressor(s) and run the regression on these first differences.
Cont…
 Cont…

• Compared with the level form regression (level form), we see that the slope coefficient has not changed much,
but the r 𝟐 value has dropped considerably.
• This is generally the case because by taking the first differences we are essentially studying the behavior of
variables around their (linear) trend values.
• We cannot compare the r 𝟐 of Eq. (first difference) directly with that of the r 𝟐 of Eq. (level form) because the
dependent variables in the two models are different.
• Also, notice that compared with the original regression, the d value has increased dramatically, perhaps
indicating that there is little autocorrelation in the first difference regression.
Cont…

• If the original time series are nonstationary, very often their first differences become stationary.
• Therefore, first-difference transformation serves a dual purpose in that it might get rid of (first-order)
autocorrelation and also render the time series stationary.
Cont…

(B) ρ Estimated from the Residuals:

where 𝒖ෝ 𝒕 are the residuals obtained from the original (level form) regression and where 𝒗𝒕
are the error term of this regression.
Note that there is no need to introduce the intercept term in the above Equation, for
we know the OLS residuals sum to zero.

Use ρො in estimating the following equation

Cont…
(B) ρ Estimated from the Residuals: example

Use ρො in estimating the following equation

Cont…

(C) Iterative Methods of Estimating ρ:

The Newey–West Method of Correcting the OLS Standard Errors