You are on page 1of 5

Assignment on Introduction to Econometrics for Finance

Part-I: Workout on violations


2. A financial analysts is interested in the relationship between earning per share (EPS) and
sales. He collects information on these two variables for 10-years period and estimates
the regression: EPS=+sales+ui. From this regression he computes the error for each
company.
year et
1
130.5
2
-540.35
3
150.10
4
-240.32
5
100.24
6
-10.11
7
83.35
8
-90.30
9
34.04
10
-127.20
Does autocorrelation appear to be a problem in this regression?
Let us check for autocorrelation by using Durbin Watson
d= (t-t-1)2
t2
d=t2+t-12 - 2tt-1
or d2(1- tt-1)
t2
t2
Year
1
2
3
4
5
6
7
8
9
10

et

et2

et-et-1

130.5
-540.35
150.10
-240.32
100.24
-10.11
83.35
-90.30
34.04
-127.20
Total

17030.25

291978.1
22530.01
57753.7
10048.06
102.2121
6947.223
8154.09
1158.722
16179.84
431882.2

-670.85
690.45
-390.42
340.56
-110.35
93.46
-173.65
124.34
-161.24

(et-et-1)2

et*et-1

450039.7
476721.2
152427.8
115981.1
12177.12
8734.772
30154.32
15460.44
25998.34
1287695

-70515.7
-81106.5
-36072
-24089.7
-1013.43
-842.669
-7526.51
-3073.81
-4329.89
-228570

d= (t-t-1)2
t2
d=(et-et-1)2 / et2

d=1287695/431882.2

Econometrics for finance


Page 1

tt-1
t2

Assignment on Introduction to Econometrics for Finance


d=3.0

Using the row () formulas to compute d approximately yields the same results.
= et*et-1/ et2

= -228570/431882.2

d= 2(1-)

d= 2(1-(-0.5347)

= -0.5457

d= 3.1

Let us develop null hypothesis in order to know the existence of the autocorrelation in the above data;
Ho: there is no positive autocorrelation
Ho: there no negative autocorrelation
Summary of the solutions
Since -1<<1, this implies that 0<d<4

According to the Durbin Watson if there is negative autocorrelation, a positive t will


tend to be followed by a negative t and vice versa so that | t - t-1 | will usually be greater
than |t|. Therefore, the numerator of d will be comparatively larger than the denominator.
In a data the d is approaches to 4 therefore the data has negative autocorrelation. Again,
we reject null hypothesis.

Par- II: Workout and interpretation


Question #2. For the study of the relationship between consumption expenditure (C) and income
(Y) in a hypothetical town, the data collected from 10 families are shown below.
C
100
120
80
100
500
120
50
600
150
250
Y
80
90
80
70
300
75
30
350
70
150

In this question, we taken the consumption as dependent variable because of the consumption
level is depend on the level of income. Again, the income was taken as independent variable
because it determines the consumption level.

No.
`1
2
3
4

Yi
80
90
80
70

Ci
100
120
80
100

207
207
207
207

129.5
129.5
129.5
129.5

(Ci-)
-107
-87
-127
-107

(Yi-)
-49.5
-39.5
-49.5
-59.5

Econometrics for finance


Page 2

(Ci-)2
11449
7569
16129
11449

(Yi-)2
2450.25
1560.25
2450.25
3540.25

(Ci-)(Yi-)
5296.5
3436.5
6286.5
6366.5

Assignment on Introduction to Econometrics for Finance


5
6
7
8
9
10
Total
Mea

300
75
30
350
70
150
1295

500
120
50
600
150
250
2070

207
207
207
207
207
207
_

129.5
129.5
129.5
129.5
129.5
129.5
_

293
-87
-157
393
-57
43
-0-

170.5
-54.5
-99.5
220.5
-59.5
20.5
-0-

85849
7569
24649
154449
3249
1849
324210

29070.25
2970.25
9900.25
48620.25
3540.25
420.25
104522.5

49956.5
4741.5
15621.5
86656.5
3391.5
881.5
182635

129.5

207

Required
A. Fit the regression equation.
1= (-C) (Y-)= 182635/104522.5= 1.747327131

0=- 1
0=207-129.5(1.747327131)

(Yi-)2

o=207-226.2788634
o= -19.2788634
Therefore the regression equation Model: = -19.279 + 1.747+
B. calculates variances and standard errors of the parameters estimates.
TSS= (Ci-)2=324210
RSS=TSS-ESS

ESS= 1 (Ci-) (Yi-)=1.747* 182635 =319063.345


RSS=324210-319063.345

2=RSS/N-2

2= 5146.655/8

RSS=5146.655

2= 643.332

Var(1)= 2/(Yi-)2= 643.332/104522.5= 0.006155


Se (1)=var(1)=0.006155= 0.0785
Var()= Yi2* 2/n* (Yi-)2=272,225*643.332/10*104522.5
=167.55
Se (o)= 167.55 =12.94
To know the significance of consumption on income let us compare the standard error of
beta one to the half of beta coefficient.
Se (1)<1/2 1
se(1)=0.0785
1/2 1 =1.747*0.5 = 0.876
Therefore, the consumption expenditure has significant effect on the income.
Again, let us develop null hypothesis and compare the t-calculated with t-tabulated.
Ho: 1= 0 or the estimates has no significant.
Econometrics for finance
Page 3

Assignment on Introduction to Econometrics for Finance


T-calculated = 1/ se(1) =1.747/0.0785 =22.25
T-tabulated at 5% confidence interval and 8 degree of freedom:= 2.31. So, t-calculated is
greater than t-tabulated.
Therefore, the null hypothesis is rejected and, the estimate has significant.
C. The R-square
R2=ESS/TSS=319063.345/182635=0.984, which means, 98.4%
Therefore, r2 is interpreted as 98.4%of the variability between the two variables has been
accounted for, and the remaining 1.6% of the variability is still unaccounted for. Or 98.4% are
due to explained variables in consumption and 1.6% is due to unexplained or other variables.
D.
Using the spearmans rank correlation coefficient, show whether there is
heteroskedasticity problem or not.
Our regression equation model is = -19.279 + 1.747+
= C-
Y

= C-

80
90
80
70
300
75
30
350
70
150

100
120
80
100
500
120
50
600
150
250

20.481
17.951
40.481
3.011
4.821
8.254
16.869
7.829
46.989
7.229

rank of rank

y
8
5.5
7
7
9
5.5
1
2
2
9
5
4
6
1
4
10
10
3
3
8

d= -y

d2

2.5
0
3.5
-1
-7
1
5
-6
7
-5

6.25
0
12.25
1
49
1
25
36
49
25
204.5

According to the spearman correlation coefficient the formula

or r =

6204.5
2
10(10 1)

=1- 1.2394

Econometrics for finance


Page 4

= -0.234

Assignment on Introduction to Econometrics for Finance


The r tabulated at 5% confidence interval is 0.714, which exceeds the r calculated.
If we consider the direction to the negative, r- tabulated is less than r-calculated.
Therefore, there is heteroscedastcity problem.

t= -0.2348/1-(-0.234)2

t= -0.5363

The same is true for t-test.

Econometrics for finance


Page 5

You might also like