AUTOCORRELATION

Submitted

BY

Sidra shahzadi

ROLL # 23011720-015

COURSE CODE: FIN-753

COURSE TITLE

QUANTITATIVE TECHNIQUES IN FINANCE

Submitted To

Dr. Adnan Bashir

Program: MPhil in Management Sciences (Finance)

DEPARTMENT OF MANAGEMENT SCIENCES

1
 Example 7.1
Detection & Removal of Autocorrelation
Detection of Autocorrelation
Regression Model
In this model Investment is dependent variable while income and interest rate are independent
variables.
Number of Observation= 30.
I = a + b1R + b2Y

Dependent Variable: I
Method: Least Squares
Date: 10/30/24 Time: 21:07
Sample: 1 30
Included observations: 30
Coefficient Std. Error t-
Statistic Prob.
Variable

C 6.224938 2.510894 0.0197

2.479172
R -0.184196 0.126416 -1.457068 0.1566
Y 0.769911 0.071791 10.72442 0.0000

R-squared 0.816282 Mean dependent var 20.22200

Adjusted R-squared 0.802673 S.D. dependent var 7.495569

S.E. of regression 3.329642 Akaike info criterion 5.338246
Sum squared resid 299.3358 Schwarz criterion 5.478366
Log likelihood -77.07369 Hannan-Quinn criter. 5.383071
F-statistic 59.98221 Durbin-Watson stat 0.852153
Prob(F-statistic) 0.000000

2
Interpretation
As we now that range of Durbin Watson statistics is between 0 to 4. If Durbin Waston is equal 2,
it means that there is no autocorrelation. If Durbin Waston is less than 2, it shows positive
autocorrelation and if Durbin Waston is more than 2, it shows negative autocorrelation. Above
regression table shows that Durbin Waston statistics is 0.852153 which is less than 2 which
means that it shows positive autocorrelation.

Use Serial Correlation LM Test To check severity, that we move to removals or not for that
we apply serial correlation LM test.
Breusch-Godfrey Serial Correlation LM Test:
F-statistic 13.04955 Prob. F(1,26) 0.0013
Obs*R-squared 10.02538 Prob. Chi-Square(1) 0.0015

Test Equation:
Dependent Variable: RESID
Method: Least Squares
Date: 10/30/24 Time: 21:24
Sample: 1 30 Included observations: 30
Presample missing value lagged residuals set to zero.
t-
Variable Coefficient Std. Error Statistic Prob.

C 1.365569 2.121810 0.643587 0.5255

R -0.082614 0.107576 -0.767957 0.4494

Y -0.007865 0.059735 -0.131664 0.8963
RESID(-1) 0.595067 0.164728 3.612417 0.0013

R-squared 0.334179 Mean dependent var

-1.55 E-15
Adjusted R-squared 0.257354 S.D. dependent var 3.212775

3
S.E. of regression 2.768671 Akaike info criterion 4.998178
Sum squared resid 199.3040 Schwarz criterion 5.185004
Log likelihood -70.97266 Hannan-Quinn criter. 5.057945
F-statistic 4.349852 Durbin-Watson stat 1.695546
Prob(F-statistic) 0.013046

Interpretation
If prob value is greater than 0.05 which means that we can accept the hypothesis which shows
the no autocorrelation while if the prob value is less than 0.05, it means that we can reject
hypothesis because autocorrelation exist. Above table shows that prob value is 0.0013 which is
less than 0.05 due to which we reject the hypothesis because autocorrelation exist and we move
to removals.

Removals
There are two removals for autocorrelation.
1. Addition of relevant variables
2. HAC Test
For removal, we use HAC Test.

Dependent Variable: I
Method: Least Squares
Date: 10/30/24 Time: 21:29
Sample: 1 30
Included observations: 30
HAC standard errors & covariance (Bartlett kernel, Newey-West fixed
bandwidth = 4.0000)
Coefficient Std. Error t-
Statistic Prob.
Variable

C 6.224938 2.124542 0.0068

2.930014
R -0.184196 0.121236 -1.519318 0.1403
Y 0.769911 0.077516 9.932311 0.0000

4
R-squared 0.816282 Mean dependent var 20.22200

Adjusted R-squared 0.802673 S.D. dependent var 7.495569

S.E. of regression 3.329642 Akaike info criterion 5.338246
Sum squared resid 299.3358 Schwarz criterion 5.478366
Log likelihood -77.07369 Hannan-Quinn criter. 5.383071
F-statistic 59.98221 Durbin-Watson stat 0.852153
Prob(F-statistic) 0.000000 Wald F-statistic 57.35421
Prob(Wald F-
statistic) 0.000000

If standard error and significance value are not reliable than we apply HAC Test. HAC test
resolve the standard error and significance value issues which make the model more reliable and
fit.

Example 7.2
Detection & Removal of Autocorrelation
Detection of Autocorrelation
Regression Model
In this model quantity of a good produced during various years is dependent variable while price
of the good, amount of fertilizer used in the production of this good and amount of rainfall
during each production year are independent variables.

Number of Observation= 31.

Q = a + b1F + b2P + b3R

Dependent Variable: Q
Method: Least Squares
Date: 10/30/24 Time: 21:13
Sample (adjusted): 2 31
Included observations: 30 after adjustments

5
 Coefficient Std. Error
t-Statistic Prob.
Variable

C -11.59809 14.96435 -0.775048 0.4453

F 0.061583 0.080990 0.760379 0.4539

P 0.316229 0.087615 3.609282 0.0013
R 0.058138 0.008069 7.205443 0.0000

R-squared 0.705547 Mean depen dent var 34.22000 Adjusted

R-squared 0.671571 S.D. dependent var 3.647238 S.E. of regression
2.090185 Akaike info criterion 4.435948 Sum squared resid
113.5907 Schwarz criterion 4.622774 Log likelihood
-62.53922 Hannan-Quinn criter. 4.495715
F-statistic 20.76641 Durbin-Watson stat 1.805563
Prob(F-statistic) 0.000000

Interpretation
As we now that range of Durbin Watson statistics is between 0 to 4. If Durbin Waston is equal 2,
it means that there is no autocorrelation. If Durbin Waston is less than 2, it shows positive
autocorrelation and if Durbin Waston is more than 2, it shows negative autocorrelation. Above
regression table shows that Durbin Waston statistics is 1.805563 which is less than 2 which
means that it shows positive autocorrelation.

Use Serial Correlation LM Test

To check severity, that we move to removals or not for that we apply serial correlation LM test.
Breusch-Godfrey Serial Correlation LM Test:

F-statistic 0.037072 Prob. F(1,25) 0.8489

Obs*R-squared 0.044421 Prob. Chi-Square(1) 0.8331

6
Test Equation:
Dependent Variable: RESID
Method: Least Squares
Date: 10/30/24 Time: 21:25
Sample: 2 31 Included observations: 30
Presample missing value lagged residuals set to zero.
Coefficient Std. Error
t-Statistic Prob.
Variable

C 0.185311 15.27974 0.012128 0.9904

F -6.33E-05 0.082534 -0.000767 0.9994

P -0.001783 0.089763 -0.019861 0.9843
R 0.000107 0.008241 0.012980 0.9897
RESID(-1) 0.039252 0.203864 0.192542 0.8489

R-squared 0.001481 Mean depen dent var -6.84 E-15

Adjusted R-squared -0.158282 S.D. dependent var 1.979121
S.E. of regression 2.130000 Akaike info criterion 4.501133
Sum squared resid 113.4225 Schwarz criterion 4.734665
Log likelihood -62.51699 Hannan-Quinn criter. 4.575842 F-statistic
0.009268 Durbin-Watson stat 1.884547
Prob(F-statistic) 0.999817

Interpretation
If prob value is greater than 0.05 which means that we can accept the hypothesis which shows
the no autocorrelation while if the prob value is less than 0.05, it means that we can reject
hypothesis because autocorrelation exist. Above table shows that prob value is 0.8489 which is
greater than
0.05 due to which we accept the hypothesis because there is no autocorrelation exist.