ECONOMETRIC METHODS (AE/EE 06)

INTERNAL ASSESSMENT I

February 28, 2019

Max. Marks: 20 Max. Time: 1.5 Hours

Answer ALL Questions.

1. Given the OLS assumptions, show that Var(Yi ) = σu2 , where, σu2 is the error variance in the
regression model: Yi = β0 + β1 Xi + ui . (2 Marks)

Proof. We know
Var(Yi |Xi ) = E(Yi2 |Xi ) − [E(Yi |Xi )]2 (1)
If Yi = β0 + β1 Xi + ui ,

Yi2 = (β0 + β1 Xi + ui )2
= β02 + β12 Xi2 + 2β0 β1 Xi + u2i + 2β0 ui + 2β1 Xi ui
Thus,

E(Yi2 |Xi ) = E(β02 + β12 Xi2 + 2β0 β1 Xi + u2i + 2β0 ui + 2β1 Xi ui |Xi )
= E(β02 |Xi ) + E(β12 Xi2 |Xi ) + E(2β0 β1 Xi |Xi ) + E(u2i |Xi ) + E(2β0 ui |Xi ) + E(2β1 Xi ui |Xi )
= β02 + β12 Xi2 + 2β0 β1 Xi + E(u2i |Xi ) + 2β0 E(ui |Xi ) + 2β1 Xi E(ui |Xi )
= β02 + β12 Xi2 + 2β0 β1 Xi + E(u2i |Xi )
(2)
where the third equality follows from applying the Law of Iterated Expectations and the fourth
equality follows from the OLS Assumption 1: E(ui |Xi ) = 0. Now, given Yi = β0 + β1 Xi + ui ,
we know E(Yi |Xi ) = E(β0 + β1 Xi + ui |Xi ) = E(β0 + β1 Xi |Xi ) + E(ui |Xi ) = β0 + β1 Xi (since,
E(ui |Xi ) = 0). Thus,

[E(Yi |Xi )]2 = (β0 + β1 Xi )2

(3)
= β02 + β12 Xi2 + 2β0 β1 Xi
Substituting Equation (2) and Equation (3) in Equation (1), we get:

Var(Yi |Xi ) = E(Yi2 |Xi ) − [E(Yi |Xi )]2

= E(u2i |Xi )
= Var(ui |Xi )
However, by OLS Assumption 3, we know that error variance is homoskedastic. That is Var(ui |Xi ) =
σu2 . Thus, Var(Yi |Xi ) = Var(ui |Xi ) = σu2 

2. For the tobacco industry in Andhra Pradesh, the estimated supply function is given by

Ŷi = 33.75 + 3.25Xi

where Yi is quantity supplied in tons and Xi is market price in Rs. (in thousands). A total volume
of 756 tons of raw tobacco is produced by 12 factories that are sampled. The average price facing
the factories is $9. If average price facing the industry increases by 1% what will be the percentage
increase in the average quantity supplied by the industry? (2 Marks)
1
 ECONOMETRIC METHODS (AE/EE 06) INTERNAL ASSESSMENT I 2

12
X
Solution. Total volume produced by 12 factories is 756 tons. That is Yi = 756. Thus,
i=1
756
Ȳ = 12 = 63. Given average price is $9, that is X̄ = 9. We know that elasticity is
 
∂ Ŷi X̄ 9
ηs = = 3.25 = 0.4643
∂Xi Ȳ 63
That is, 1 % increaes in price of tobacco will lead to 0.4643% increaes in the quantity supplied of
tobacco.

3. A linear regression yields β̂1 = 0. Show that R2 = 0. (1 Marks)

Proof. Given the estimated model Ŷi = β̂0 + β̂1 Xi , if β̂1 = 0, then Ŷi = β̂0 . Further, since
Ȳ = β̂0 + β̂1 X̄, with βˆ1 = 0, we have Ȳ = β̂0 . Thus, we have Ŷi = Ȳ or (Ŷi − Ȳ ) = 0 ⇒
Xn
(Ŷi − Ȳ )2 = 0 ⇒ (Ŷi − Ȳ )2 = 0 or ESS = 0. We know R2 = ESS T SS . If ESS = 0, then
i=1
R2 = 0. 

4. A linear regression yields R2 = 0. Does this imply that β̂1 = 0? (2 Marks)

P 2
Proof. R2 = ESS
= P(ŷi −ȳ)2 . We know ŷi = β̂0 + β̂1 xi and ȳ = β̂0 + β̂1 x̄. Thus,
T SS (yi −ȳ)
n
X
(ŷi − ȳ)2
i=1
R2 = n
X
(yi − ȳ)2
i=1
n
X
(β̂0 + β̂1 xi − β̂0 − β̂1 x̄)2
i=1
= n
X
(yi − ȳ)2
i=1
n
X
(β̂1 xi − β̂1 x̄)2
i=1
= n
X
(yi − ȳ)2
i=1
n
X
[β̂1 (xi − x̄)]2
i=1
= n
X
(yi − ȳ)2
i=1
Xn
2
β̂1 (xi − x̄)2
i=1
= n
X
(yi − ȳ)2
i=1

Xn
If R2 = 0 then the numerator β̂ 2 (xi − x̄)2 = 0. This would be true if either β̂1 = 0 or
i=1
2
P
i (xi − x̄) = 0. However, the latter expression being a sum of positive (squared) expressions,
will be zero only when ∀ i, xi − x̄ = 0 or ∀ i xi = x̄. Thus, it is not necessary that if R2 = 0, then
it implies β̂1 = 0. 

5. Suppse that a researcher, using data on class size (CS) and average test scores from 50 M.A.
classes, estimates the OLS regression: (5 Marks)
\ = 640.3 − 4.93 × CS
T estScore
 ECONOMETRIC METHODS (AE/EE 06) INTERNAL ASSESSMENT I 3

with R2 = 0.11, SER=8.7.

(a) A classroom has 25 students. What is the regression’s prediction for that classroom’s average
test score?
\ CS=25 = 640.3 − 4.93 × 25 = 517.05
Ans: T estScore|
(b) Last year a classroom had 21 students, and this year it has 24 students. What is the regres-
sion’s prediction for the change in the classroom average test score?
Ans: ∆T \ \ CS=21 −T estScore|
estScore = T estScore| \ CS=24 = (640.3 − 4.93 × 21) − (640.3 −
4.93 × 24) = 14.79
(c) The sample average class size across the 50 classrooms is 22.8. What is the sample average
of the test scores across the 50 classrooms?
Ans: We know T estScore = 640.3−4.93×CS. Substituting CS = 22.8, T estScore = 527.896
(d) What is the sample standard deviation of test scores across the 50 classrooms?
SSR
Ans: We know SER = σ̂=8.7. However, SER2 =σ̂ 2 = n−2 ⇒ SSR = (n − 2)σ̂ 2 = (50 −
SSR SSR
2) × (8.7)2 = 3633.12. From the formula for R2 , we know R2 = 1 − T SS ⇒ T SS = 1−R2 =
3633.12 3633.12
1−0.11 = 0.89 = 4082.16.
T SS 4082.16
The sample variance is s2Y = n−1 = 49 = 83.31. Thus, standard deviation is sY =
p
s2Y = 9.127.
6. For a sample of 506 communities in Boston area, the following model estimates the median housing
price (price) in the community to various community characteristics: nox - the amount of nitrogen
oxide in the air (in ppm); dist - a weighted distance of the community from five emplyment centres
(in miles); rooms - is the average number of rooms in houses in the community and stratio is the
average student-teacher ratio of the schools in the community. The estimated model is
\ = 11.08 − 0.954 log(nox) − 0.134 log(dist) + 0.255 rooms − 0.052 stratio
log(price)
(0.32) (0.117) (0.043) (0.019) (0.006)

2
with n = 506 and R = 0.581 (5 Marks)
(a) Interpret the coefficient β1 , the estimated parameter corresponding to nox .

Solution. 1 unit increaes in nox pollution reduces house price by 0.954 percent.

(b) Suppose we wish to test the hypothesis H0 : β1 = −1 against the alternative H1 : β1 6= −1.
Interpret the hypothesis and its alternative.

Solution. Null hypothesis: elasticity of house price to nox pollution is 1, that is for 1%
increase in nox pollution, house price decreases by 1%. Alternative hypothesis: elasticity
of house price to nox pollution is not 1, that is 1% increase in nox pollution, house price
decreases by any value other than 1%.

(c) Carry out the test and interpret the result.

Solution. The t-stat for the test under H0 is given by

β̂1 − β1 −0.954 + 1
tcal = = = 0.393
SE(β̂1 ) 0.117
Now, the two-sided tcrit at 5% level of significance, with n − k − 1 = 504 df is 1.96. Since,
tcal < tcrit , we fail to reject the null hypothesis, suggesting that there exist little evidence
that the elasticity is different from -1.

7. In the case of a multiple regression model with k regressors, (3 Marks)

y = Xβ + u
 ECONOMETRIC METHODS (AE/EE 06) INTERNAL ASSESSMENT I 4

if the estimated coefficient is β̂ = (X 0 X)−1 (X 0 y), then show that

Var(β̂) = σ 2 (X 0 X)−1

Proof. We know that

β̂ = (X 0 X)−1 (X 0 y) (4)
But y = Xβ + u. So substituing for y in (4), we get

β̂ = (X 0 X)−1 X 0 (Xβ + u)
= (X 0 X)−1 X 0 Xβ + (X 0 X)−1 X 0 u
(5)
= β + (X 0 X)−1 X 0 u
β̂ − β = (X 0 X)−1 X 0 u

Now, the variance of the estimated parameter vector Var(β̂) is

Var(β̂) = E[(β̂ − β)(β̂ − β)0 ]

Now, substituting from Eq (5), we get

Var(β̂) = E[(β̂ − β)(β̂ − β)0 ]

= E[(X 0 X)−1 X 0 uu0 X(X 0 X)−1 ]
= (X 0 X)−1 X 0 E(uu0 )X(X 0 X)−1
= (X 0 X)−1 X 0 σ 2 X(X 0 X)−1
= σ 2 (X 0 X)−1 X 0 X(X 0 X)−1
= σ 2 (X 0 X)−1
Hence, proved.