 Introduction

 Causes of
 Autocorrelation OLS
 Estimation
BLUE Estimator
 Consequences of using
 OLS Detecting
Autocorrelation
1.Introduction

Autocorrelation occurs in time-series studies
when the errors associated with a given time
period carry over into future time periods.

For example, if we are predicting the growth of
stock dividends, an overestimate in one year is
likely to lead to overestimates in succeeding
years.
1.Introduction

Times series data follow a natural
ordering over time.


It is likely that such data exhibit
intercorrelation, especially if the time
interval between successive
observations is short, such as weeks or
days.
1. Introduction

We expect stock market prices to move or
move down for several days in succession.

In situation like this, the assumption of no auto
or serial correlation in the error term that
underlies the CLRM will be violated.

We experience autocorrelation when
E(uiu j ) ≠ 0
1. Introduction

Sometimes the term autocorrelation is used
interchangeably.

However, some authors prefer to distinguish between
them.

For example, Tintner defines autocorrelation as ‘lag
correlation of a given series within itself, lagged by a
number of times units’ whereas serial correlation is the
‘lag correlation between two different series’.

We will use both term simultaneously in this lecture.
1. Introduction

There are different types of serial correlation.
With first-order serial correlation, errors in
one time period are correlated directly with
errors in the ensuing time period.


With positive serial correlation, errors in one
time period are positively correlated with errors
in the next time period.
2. Causes of Autocorrelation
 Inertia - Macroeconomics data experience
cycles/business cycles.

Specification Bias- Excluded variable
Appropriate equation:
Yt = β1 + β 2 X 2t + β 3 X 3t + β 4 X 4t + ut
Estimated equation
Yt = β1 + β 2 X 2t + β 3 X 3t + vt
Estimating the second equation implies
vt = β 4 X 4t + ut
2. Causes of Autocorrelation

 Functional
Specification Bias-Incorrect
Form
Y = β+ βX + βX 2 + v t
t 1 2 2t 3 2t

Yt = β1 + β 2 X 2t + ut

u = βX 2 + v t
t 3 2t
2. Causes of Autocorrelation


Cobweb Phenomenon
In agricultural market, the supply reacts to
price with a lag of one time period because
supply decisions take time to implement. This
is known as the cobweb phenomenon.
Thus, at the beginning of this year’s planting
of crops, farmers are influenced by the price
prevailing last year.
2. Causes of Autocorrelation

Lags
Consumptiont = β1 + β 2 Consumptiont −1 + ut
The aboveequation is known as auto
regression because one of the explanatory
variables is the lagged value of the dependent
variable.
If you neglect the lagged the resulting error
term will reflect a systematic pattern due to the
influence of lagged consumption on current
consumption.
2. Causes of Autocorrelation


Data Manipulation
Yt −1 = β1 + β 2 X t −1
Yt = β 1 + β 2 X t + u t +u
t −1

ΔYt = β 2 ΔX t + vt
This equation is known as the first difference form
and dynamic regression model. The previous
equation is known as the level form.
Note that the error term in the first equation is not
auto correlated but it can be shown that the error
term in the first difference form is auto correlated.
2. Causes of Autocorrelation

 Nonstationarity
When dealing with time series data, we
should check whether the given time series is
stationary.
A time series is stationary if its characteristics
(e.g. mean, variance and covariance) are time
variant; that is, they do not change over time.
If that is not the case, we have a non
stationary time series.
 Suppose Yt is related to X2t and
X3t, but we wrongfully do not include
X3t in our model.

The effect of X3t will be captured by

the disturbances ut. If X3t like many
economic series exhibit a trend over
time, then X3t depends on X3t-1,X3t
-2and so on. Similarly then ut depends
on ut-1, ut-2 and so on.
Suppose Yt is related to X2t with a
quadratic relationship:
Yt=β1+β2X22t+ut
but we wrongfully assume and estimate
a straight line:
Yt=β1+β2X2t+ut

Then the error term obtained from the

straight line will depend on X22t.
Suppose a company updates its
inventory at a given period in time.

If a systematic error occurred then the

cumulative inventory stock will exhibit
accumulated measurement errors.

These errors will show up as an auto

correlated procedure.
The simplest and most commonly
observed is the first-order
autocorrelation.
Consider the multiple regression
model:
Yt=β1+β2X2t+β3X3t+β4X4t+…+βkXkt+ut

in which the current observation of the

error term ut is a function of the
previous (lagged) observation of the
error term:
ut=ρut-1+et
 The coefficient ρis called the
first-order autocorrelation coefficient
-1 to
and +1. values from
takes
It is obvious that the size of ρ will
determine the strength of serial
correlation.

We can have three different cases

If ρ is zero, then we have no
autocorrelation. If ρ approaches unity,
the value of the previous observation
of the error becomes more important
in determining the value of the current
error and therefore high degree of
autocorrelation exists. In this case we
have positive autocorrelation.
If ρ approaches -1, we have high
degree of negative autocorrelation.
Second-order
when:
ut=ρ1ut-1+ ρ2ut-2+et
Third-order when
u =ρ u + ρ u +ρ u +e
t 1 t-1 2 t-2 3 t-3 t

p-th order when:

u =ρ u + ρ u +ρ u +…+
t 1 t-1 2 t-2 3 t-3

ρu +e
p t-p t

The OLS estimators are still unbiased and
consistent.
This is because both unbiasedness and

The OLS estimators
consistency will be inefficient
do not depend and 6
on assumption
therefore
which is inno longer
this caseBLUE.
violated.

The estimated variances of the regression
coefficients will be biased and inconsistent,
and therefore hypothesis testing is no longer
valid. In most of the cases, the R2 will be
overestimated and the t-statistics will tend to
be higher.
 There are two ways in general.
The first is the informal way
which is done through graphs and
therefore we call it the
graphicalmethod .
The second is through formal
tests for autocorrelation, like the
following ones:

The Durbin Watson Test

The Breusch-Godfrey Test

Run test
 The following assumptions
should be satisfied:

The regression model includes a
constant

Autocorrelation is assumed to be of
first-order only

The equation does not include a
lagged dependent variable as an
explanatory variable

Step 1: Estimate the model by OLS and
obtain the residuals

Step 2: Calculate the DW statistic

Step 3: Construct the table with the
calculated DW statistic and the dU, dL,
4-dU and 4-dL critical values.

Step 4: Conclude
 It is a Lagrange Multiplier Test that resolves
the drawbacks of the DW test.

Consider the model:
Yt=β1+β2X2t+β3X3t+β4X4t+…+βkX
kt
+ut
where:
u =ρ u + ρ u +ρ u +…+
t 1 t-1 2 t-2 3 t-3

ρu +e
p t-p t

Combining those two we
get:
Yt=β1+β2X2t+β3X3t+β4X4t+…+ Serial Correlation Topic Nine
 The null and the alternative hypotheses are:


H0:ρ1= ρ2=…= ρp=0 no autocorrelation

Ha:at least one of the ρ’s is not zero, thus, autocorrelation

Step 1: Estimate the model and obtain the residuals

Step 2: Run the full LM model with the number of lags used
being determined by the assumed order of autocorrelation.
Step 3: Compute the LM statistic = (n-ρ)R2from the LM
model and compare it with the chi-square critical value.
Step 4: Conclude

Serial Correlation Topic Nine

 We have two different
 cases:
When ρ is known

When ρ is unknown

Consider the model

Yt=β1+β2X2t+β3X3t+β4X4t+…+β
X +ut
k kt

where
u ρ u +e
t= 1 t-1 t

Serial Correlation Topic Nine


Write the model of
t-1Y: =β +β X +β X +β X +…+β X
t-1 1 2 2t-1 3 3t-1 4 4t-1 k kt-1

+u
t-1

Multiply both sides by ρto get

ρY = ρβ + ρβ X + ρβ X + ρβ X
t-1 1 2 2t-1 3 3t-1 4 4t-1

+…+ ρβkXkt-1+ ρut-1


Subtract those two equations:
Yt-ρYt-1= (1-ρ)β1+ β2(X2t-ρX2t-1)+ β3(X3t-
ρX3t-1)+
+…+ βk(Xkt-ρXkt-1)+(ut-ρut-1)
Topic Nine
 Where now the problem of
autocorrelation is resolved because et
is no longer autocorrelated.
Note that because from the
transformation we lose one
observation, in order to avoid that
loss we generate Y1 and Xi1 as
X*i1=Xi1 sqrt(1-ρ )
2

follows: Y*1=Y1 sqrt(1- ρ2)

This transformation is known as
the quasi- differencing or
generalised differencing.

The Cochrane-Orcuttiterative procedure.
Step 1: Estimate the regression and
obtain residuals
Step 2: Estimate ρ from regressing the
residuals to its lagged terms.
Step 3: Transform the original variables
as starred variables using the
obtained from step 2.
Step 4: Run the regression again
with the transformed variables
and obtain residuals.

Step 5 and on : Continue repeating

steps 2 to 4 for several rounds until
(stopping rule ) the estimates of from
two successive iterations differ by no
more than some preselected small
value, such as 0.001.