pricei = β + β sqfti + β agei + β baths + e: Question 1 (7 marks)

You might also like

Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 11

Question 1 (7 marks)

(a)(2 marks) Estimate the following model using least squares. Report the result in full AND
provide Eviews output.

pricei = β1 + β2sqfti + β3agei + β4bathsi + ei

Method: Least Squares


Date: 10/08/21 Time: 12:30
Sample: 1 900
Included observations: 900

Variable Coefficient Std. Error t-Statistic Prob.

C -62.46140 9.182401 -6.802295 0.0000


SQFT 7.916557 0.348160 22.73831 0.0000
AGE -0.577210 0.138497 -4.167680 0.0000
BATHS 23.13072 5.660916 4.086038 0.0000

R-squared 0.629735 Mean dependent var 151.2452


Adjusted R-squared 0.628495 S.D. dependent var 111.4407
S.E. of regression 67.92448 Akaike info criterion 11.27910
Sum squared resid 4133906. Schwarz criterion 11.30045
Log likelihood -5071.597 Hannan-Quinn criter. 11.28726
F-statistic 507.9623 Durbin-Watson stat 1.979329
Prob(F-statistic) 0.000000

(b)(5 marks) Interpret the estimated coefficient for sqft, age and baths. Is the sign of
each coefficient what you expected? Why?

The 𝑏2 = 7.916557 suggests that, holding all other regressors constant, 100 square feet
increasing in total living area is associated with an increase in the selling price of house of
7916.557 dollars. The positive sign for 𝑏2 is according to our expectation because as
square feet of living area increases it also tends to increase selling price of living area.
.
The 𝑏3 = −0.577suggests that, holding all other regressors constant, with every 1 year
increase in the age of house, the selling price of total living area will decrease by -577.210
dollars The expected sign for 𝑏3 is expected because value of house depreciates with its
age in years.

The 𝑏4 = 23.1307 implies that, holding all other regressors constant, an additional
bathrooms will increase the selling price of houses. The beta coefficient is 23.1307 which
means that if 1 additional bathroom is establishing in the house it will increase seliing
price of houses by 23130.7 dollars.

Question 2 (6 marks): FOR ETF2100 ONLY

Based on the model in question 1(a) above, test at the 5% level of significance the null
hypothesis that a decrease in total living area by 100 square feet has the same effect on
house price as a 10 year increase in the house age, other things being constant. Use a t-
statistic approach and write down all the steps used to conduct your test. You can use
Eviews to calculate the test statistic and obtain the t-critical value.

To test the claim, the relevant hypotheses are

𝐻o: 𝛽2+10 𝛽3 = 0

𝐻a: 𝛽2+10 𝛽3 ≠ 0

b2 + 10b3 = 7.9166 – 10 x 0.5772 = 2.1446

se(b2 + 10b3) = √1.9532(b2+ 10 b3 )

var ( b 2+ b 3 )=var ( b2|x ) +102 var ( b 3 ) +2∗10 xcov (b 2 ,b 3)


var ( b 2+ b 3 )=var ( b2|x ) +100 var ( b 3 )+ 20 xcov (b 2, b 3)
var ( b 2+ b 3 )=0.1212+100∗0.0192+20∗ (−0.0044 )=1.9532

se(b2 + 10b3) = √1.9532 = 1.3976

The test statistic is


b 2−B 2
t= ~ t(N-2)
se(b 2)
2.14446
t=
1.3976
t=1.5345
For 𝛼 = 0.10, tc=t(0.975,896)= -1.6547 [Eviews command “scalar tcrit1
=@qtdist(0.975,896)”]

The test statistic follows the t-distribution with 896 degrees of freedom.
Critical value: ±1.962615

T statistic < t-critical


1.5345< 1.962615
Therefore we do not reject the null hypothesis.

Since 𝑡-statistics= 1.5345 is less than 𝑡-critical = 1.962615, we do not reject 𝐻o and
conclude that the estimated relationship does not support the claim that an increase
in total living area by 100 square feet does not have the same effect on house price
as a 10 years decrease in the house age, other things being constant.

Question 3 (15 marks)


(a) (2 marks) You suspect that the change in price associated with an extra square feet of
house size depends on how old the house is. Extend the model in question 1(a) to allow
for this (write the new model down). Estimate this model and include your Eviews output.

pricei = β1 + β2sqfti + β3agei + β4bathsi + β4age*bathsi +ei


(b) (2 marks) Comment on the significance of all the coefficients in the extended model you
just estimate for question 3(a) using the p-value approach. (Don't forget to state the
significance level at which the coefficients are significant, e.g. 1, 5 or 10 percent)

The p-value of all the coefficients is less than 0.01, on that basis all the coefficient
are statistically significant at 1%. Thus, there is enough evidence available to
conclude that all the coefficients have significant effect on the house’s price at 1%
significance level.

(c)(3 marks) Using this extended model, write down the expressions for the marginal effect
∂E(p where X denote all observations on sqft and age. Interpret this expression for a
^ri
ce|X)

∂sqft
house with age equal to a particular level, say age0. Note, there is no need to put in any numbers
here and you can use bk to denote estimate for βk.

∂ E ( price∨x )
=b2 +b5 agei
∂ sqft
Interpretation: An increase in total living area by 100 square is associated with
b 2+b 5 agei increase/decrease in the price of house in thousand dollars.

(d)(8 marks) Find point estimates AND 95% interval estimates for the marginal effect of an
extra hundred square feet of total living area on house price for houses that are (i) 2 years old,
and (ii) 45 years old. How do these estimates change as age increases? [You can use Eviews to
find the appropriate standard errors].

∂ E ( price∨x )
=b2 +b5 agei
∂ sqft

var ( b2 +b5 agei )=var ( b 2) + age2i var ( 5 ) +102 var ( b 3 ) +2∗age❑i ∗cov (b 2 , b 5)

var ( b2 +b5 agei )=0.418722+ age2i 0.01252+ 2∗age ❑


i ∗0.0471

se(b2 +b5 agei )=√ b2 +b5 agei

i) 2 years old,

b2 + b5 agei = 11.1795 – 0.1533x 2 = 10.8729

Se(b2 + b5 age) = √0.41872 + 4 × 0.01252 + 2 × 2 × 0.0471 = 0.6036

Tc = t(0.975, 896) = 1.9626


Point estimate for marginal effect of extra 100 square feet on house’s price is:

= 10.8729 ± 1.9626 x 0.6036 = [9.6883, 12.0575]

For 2-year-old house, extra hundred square feet of total living area will increase house price
between $9,688.3 and $12,057.5, by holding all other factors constant

ii) 45 years old

b2 + b5 agei = 11.1795 – 0.1533x 45 = 4.281

Se(b2 + b5 age) = √0.41872 + 45 × 45 × 0.01252 + 2 × 45 × 0.0471 = 2.17502

4.281 ± 1.9626 x 2.17502 = [0.0123, 8.5497]


For 45-year-old house, extra hundred square feet of total living area will increase house price
between $12.3 to $8,549.7 holding all other things constant.

As age of house increases, for hundred square feet increase in the total house’s size, the price of
the old house will increase by a smaller amount of money comparing to the new house.

(a)(1 marks) Estimate the following regression model by least squares. Provide Eviews output.
No need to report the result in full in equation form.

bweighti = β1 + β2smoke2i + β3smoke3i + β4smoke4i + β5magei + β6mmariedi + ei


Dependent Variable: BWEIGHT
Method: Least Squares
Date: 10/09/21 Time: 01:16
Sample: 1 1200
Included observations: 1200

Variable Coefficient Std. Error t-Statistic Prob.

C 3161.192 85.08063 37.15525 0.0000


SMOKE2 -183.3664 80.18980 -2.286655 0.0224
SMOKE3 -210.4482 67.67266 -3.109797 0.0019
SMOKE4 -271.5894 64.99755 -4.178457 0.0000
MAGE 5.419030 3.394939 1.596208 0.1107
MMARRIED 151.3034 43.01231 3.517676 0.0005

R-squared 0.055642 Mean dependent var 3369.426


Adjusted R-squared 0.051688 S.D. dependent var 597.4775
S.E. of regression 581.8316 Akaike info criterion 15.57523
Sum squared resid 4.04E+08 Schwarz criterion 15.60068
Log likelihood -9339.136 Hannan-Quinn criter. 15.58481
F-statistic 14.07026 Durbin-Watson stat 1.980587
Prob(F-statistic) 0.000000

(b)(3 marks) Interpret the estimated coefficients of smoke2, smoke3 and smoke4
The smoke2 = -183.3664 suggests that, holding all other regressors constant, increasing
mother smokes 1-5 cigarettes per day is associated with decrease in the birth weight of infants by
183.3664 grams.
.
The smoke3 = −210.4482 suggests that, holding all other regressors constant, increasing
mother smokes 6-10 cigarettes per day is associated with decrease in the birth weight of infants
by 210.4482 grams.

The smoke4 = -271.589 implies that, holding all other regressors constant, increasing mother
smokes 11 or more cigarettes per day is associated with decrease in the birth weight of infants by
271.589 grams.

(c)(6 marks) Using the F-test at 1% significance level, test the hypothesis that mother’s smoking
behavior does not affect the birthweight of her baby. [Hint: All smoking dummies should be
considered here.] You must write out the test in full including all the steps. Don’t forget to
write down the restricted model. Compute the F-statistic using the Wald test function in Eviews
and show the Eviews Wald test output screenshot (no need to manually estimating both the
restricted and unrestricted models separately here). Also use Eviews to find the exact F critical
value.

H0 : β2 = β3= β4 or = 0 (mother’s smoking behavior does not affect the birthweight of her baby)
H1 : Atleast one is not equal−to zero (mother’s smoking behavior affect the birthweight of her
baby

Dependent Variable: BWEIGHT


Method: Least Squares
Date: 10/09/21 Time: 03:35
Sample: 1 1200
Included observations: 1200

Variable Coefficient Std. Error t-Statistic Prob.

C 3161.192 85.08063 37.15525 0.0000


SMOKE2 -183.3664 80.18980 -2.286655 0.0224
SMOKE3 -210.4482 67.67266 -3.109797 0.0019
SMOKE4 -271.5894 64.99755 -4.178457 0.0000
MMARRIED 151.3034 43.01231 3.517676 0.0005
MAGE 5.419030 3.394939 1.596208 0.1107

R-squared 0.055642 Mean dependent var 3369.426


Adjusted R-squared 0.051688 S.D. dependent var 597.4775
S.E. of regression 581.8316 Akaike info criterion 15.57523
Sum squared resid 4.04E+08 Schwarz criterion 15.60068
Log likelihood -9339.136 Hannan-Quinn criter. 15.58481
F-statistic 14.07026 Durbin-Watson stat 1.980587
Prob(F-statistic) 0.000000
Wald Test:
Equation: Untitled

Test Statistic Value df Probability

F-statistic 9.086573 (3, 1194) 0.0000


Chi-square 27.25972 3 0.0000

Null Hypothesis: C(2)=C(3)=C(4)=0


Null Hypothesis Summary:

Normalized Restriction (= 0) Value Std. Err.

C(2) -183.3664 80.18980


C(3) -210.4482 67.67266
C(4) -271.5894 64.99755

Restrictions are linear in coefficients.

Dependent Variable: BWEIGHT


Method: Least Squares
Date: 10/09/21 Time: 03:37
Sample: 1 1200
Included observations: 1200

Variable Coefficient Std. Error t-Statistic Prob.

C 3076.677 83.75662 36.73353 0.0000


MMARRIED 206.9425 41.99636 4.927630 0.0000
MAGE 5.448010 3.420160 1.592911 0.1114

R-squared 0.034082 Mean dependent var 3369.426


Adjusted R-squared 0.032468 S.D. dependent var 597.4775
S.E. of regression 587.6980 Akaike info criterion 15.59280
Sum squared resid 4.13E+08 Schwarz criterion 15.60553
Log likelihood -9352.680 Hannan-Quinn criter. 15.59759
F-statistic 21.11779 Durbin-Watson stat 1.970287
Prob(F-statistic) 0.000000

The results of Wald test shows that F-statistics is 9.08657 with the p-value of 0.000 which is
significant at 5% level. Thus we reject the null hypothesis and concluded that mother’s
smoking behavior affect the birthweight of her baby.

d. (3 marks) What is the estimated difference in the expected birthweight of a baby whose
mother is not married and smokes 1-5 cigarettes a day relative to that of a baby whose
mother is of the same age, married, and smokes 11 or more cigarettes a day? Interpret
your answer in full.

b2 – b4 = −0.042− (−0.101)

= 0.056

It is estimated that estimated difference in the expected birthweight of a baby is 0.056


whose mother is not married and smokes 1-5 cigarettes a day relative to that of a baby
whose mother is of the same age, married, and smokes 11 or more cigarettes a day

(e)(1 marks) Now consider a similar model but with log of birthweight as dependent variable.
Estimate this model and provide Eviews output. No need to report the result in full in
equation form.

ln(bweight)i = β1 + β2smoke2i + β3smoke3i + β4smoke4i + β5magei + β6mmariedi + ei

Dependent Variable: LBWEIGHT


Method: Least Squares
Date: 10/09/21 Time: 02:25
Sample: 1 1200
Included observations: 1200

Variable Coefficient Std. Error t-Statistic Prob.

C 8.031185 0.032162 249.7111 0.0000


SMOKE2 -0.042403 0.030313 -1.398841 0.1621
SMOKE3 -0.065921 0.025581 -2.576909 0.0101
SMOKE4 -0.101766 0.024570 -4.141835 0.0000
MAGE 0.001614 0.001283 1.258010 0.2086
MMARRIED 0.058887 0.016259 3.621730 0.0003

R-squared 0.048302 Mean dependent var 8.101898


Adjusted R-squared 0.044317 S.D. dependent var 0.224984
S.E. of regression 0.219942 Akaike info criterion -0.185917
Sum squared resid 57.75919 Schwarz criterion -0.160467
Log likelihood 117.5504 Hannan-Quinn criter. -0.176330
F-statistic 12.11994 Durbin-Watson stat 1.888703
Prob(F-statistic) 0.000000

ln(bweight)i = 8.031—0.042smoke2i -0.066smoke3i -0.102smoke4i 0.001 β5magei +


0.059mmariedi + ei

(f)(3 marks) Using the results from part (e), interpret the coefficient of smoke4 using both the
“rough” calculation and the “exact” calculation. Report your answers to 2 decimal place.

Using rough calculations: We estimated that birth weight of infants is lower than by 10.2% for
the women who smoke 11 or more per day as compared to those who do not smoke.

Using exact calculation: The percentage birth weight of infants who smokes 11 or more
cigarettes per day as compared to the birth weight of those married women who do not smoke is
= (e(0.101) -1) *100
= 10.62%

We estimate that the percentage birth weight of infants who smokes 11 or more cigarettes per
day is 10.62% lower than the married women who do not smoke, keeping all other
characteristics the same.

You might also like