ECON2280: Introductory Econometrics

Assignment 2 (Due Monday Mar. 18, 2024)

February 27, 2024

Guidelines
You are free to work on this assignment in groups (of up to 3 students) but
you are strongly encouraged to work out these problems individually before
discussing answers with your group. Each group should submit a single copy of
answers to these questions onto Moodle. Write your student ID and the names of
any student you worked with on this assignment. Names cannot be added after the
assignment is submitted. The whole assignment must be typed. Equations must be
typed with software for typing equations, such as the Equation Editor in MS Word.
Late submission or handwritten assignments will not be accepted and a zero score
will be given.

Questions
1 [30 points] Continue from Question 3 of Assignment 1.
(a) (21 points) Given the data (copied below), compute the following statistics:
(i) Total Sum of Squares (TSS), (ii) Explained Sum of Squares (ESS), (iii) Residual
Sum of Squares (RSS), (iv) R2 , (v) Adjusted R2 , (vi) Variance of 1 , (vii) Standard
error of 1 .
(b) (5 points) What might be a reason for which the estimator of 1 is biased
downwards? Explain.
(c) (4 points) If the reason you gave in part (b) is correct, does this mean that
the estimated coefficient of the class size is definitely smaller than the true e↵ect?
Explain.

1
 1 .
a)
可
=
*
88 82 + 75+ 73 +57 ) / 5 =
15
)
T TSS = Ʃ- yp
(
(

35} =+ (82 35)

2 )
=
( 88 - -

+ 95 (15 " ]+73-T5 ) 57 -75

^+

=
169 + 49f 4 +324 =54
o
The
" Assignment
多,
:modo
1
ii)
regesso
;
?
M

y .5 s, x;
-

Y :
=

performanlei ,
×T =
sizei

Xi y 订 Yi gp
( -^
Yi y (

20 90 . 3 88 /90 3 .
-
÷
75) 234 09 .
/88 90 3) ≡ 5 29
-
.
.

.
(

40 80 .
1
8z 1 - 75 =
26 . 01 82 80 - =
3 61.

50 75 75 /75 -

75) == 0
( 75 =
0

69 9
2 2
60 .
巧 (69 9 75)
.
- = 26 01.

9
7369 . =
9 61.

=
800 59 7 . 57 (
≥=
5 7 75} 234 09 -

. 2 57 59 ( -
)
7
. 29

Sum :
520 25 8 .

Yg )
≥
ii ) ESS =

=
520 . 2

/
=
: τi ) RSS - ;) ISS ESStRss
=
= y

=
25 8 . 546 =
5202+RSS
RSS =
25 .
iv ) R2 =
ESS/ TSS
520 2
953
.

= =
0 .

546

v ) adjusted R
2
=
1 -

nk 1 -
(1 -
R
=
)

|
2 )
=
1 -

(1 -

R
6 937
立 -

vi )Var (β ) =
TSSX

σ
“
=
RS 的 = 25 =
86
比 2 灣
.
-

TSSx
"
Ʃ 2
"

( xT
" "
= -
π:) = (20 -
50) + (40 -

( 50)
50) +50 -
+ (60 .

)
50 + (80 50)≥=2000
-

8 6
Var ( ,β )
.

:
=
可死
o0
=
0 0043 .
、
》
 ^
se ( β, ) (
vir ) β
ioo
:
1

= 4s

=
0 .
06561

b) There might an omitted

regressor .

Say , the true model is

YI
=
Jot hix thaxatu .

Now, assumeXa Ts omifted and Ts orrelated to Xi

x2 =γ +γix1 tv
0

= do +dix ha ( Jo + )
∴
YI t ,
a, t r tu

y
=
(
ho α 50 ) + (
e + h ,) x
γ , + hrtu )

Cumparnmdel
ehur : Yi =Bo Tβ 14 t E

.
: β , = { 1 tof 1
When there dowmward bias
is on β1 ,

I ,

- 31 α25 > =
0

∴ . One Possibility is that we omiffed a variable that is

Positively arrelated 4, ( ie , sze:) and fi lie. perforrMancei

)
with

c
子
 Table 1: Class size and student performance

Index Code Size Grade

1 ECON1210 20 88
2 ECON1220 40 82
3 ECON1280 50 75
4 ECON2262 60 73
5 ECON2280 80 57

2 [25 points] The following equation describes the median housing price in a com-
munity in terms of amount of pollution (nox for nitrous oxide) and the average number
of rooms in houses in the community (rooms):

log(price) = 0 + 1 log(nox) + 2 rooms + ui

(a) (5 points) What are the probable signs of 1 and 2 ? Explain.

(b) (5 points) Why might nox [or more precisely, log(nox)] and rooms be neg-
atively correlated? If this is the case, does the simple regression of log(price) on
log(nox) produce an upward or downward biased estimator of 1 ? Explain. [Note:
upward means ˆ1 > 1 and downward means ˆ1 < 1 .]
Using the real-world data, the following equations were estimated:

\ = 11.71
log(price) 1.043 log(nox), n = 506, R2 = 0.264
\ = 9.23
log(price) 0.718 log(nox) + 0.306rooms, n = 506, R2 = 0.514

(c) (5 points) Interpret the two R2 . Why can’t you use the R2 to compare the
two equations?
(d) (5 points) Compute the adjusted R2 and compare the explanatory power of
2
the two equations. [Hint: R̄2 = 1 (1 nR k)(n1 1) ]
(e) (5 points) The estimate of the elasticity of price with respect to nox in the
simple regression is more negative than the estimate in the multiple regression. Is
this relationship what you would have predicted, given your answer to part (b)? Does
this mean that 0.718 is definitely closer to the true elasticity than 1.043? Explain.

3 [30 points] Let ˆ0 and ˆ1 be the OLS intercept and slope estimators of the simple
linear regression model, respectively, and let ū be the sample average of the errors
(not the residuals!). P
(a) (5 points) Show that ˆ1 can be written as ˆ1 = 1 + ni=1 wi ui , where wi = SST
di
x
and di = xi x̄. P
(b) (5 points) Use part (a), along with ni=1 wi = 0, to show that ˆ1 and ū are
uncorrelated. [Hint: You are being asked to show that E[( ˆ1 1 ) · ū] = 0.]

2
eA

)
2
C 1R measuves the offif of model
goodness a ,

The second model Mdudesanextrarariable . ?willincreusein general

R

even if it duls hothave real explanatory . THus, itTsnotfair to

2 power
compare the R of the 2 model ,

d)
Model 1 :
RE = 1
- HR 5,
]
品 !
= 0 .
2625

Model 2 :
RE =
1- ( -2
1
R

5062
)
-1
= 052

Model hasa 2

stronger explanatny power

,
 M
Ʃ (xi π) (yi
- -

可 )
[ Ex( )= EXT -
- Ex

3 )
a β =

Ʃ(
X xR
= RX -

=
Ʃ-
xi ( βo+β xi + Ui -

可
) =
0 J
SSTX
Ʃβ (λi- π +β , Ʃ4 -] xi +Ʃ(xi π)
)
i
-

Ui 可 ε (Xπ
-
-
π )
=

SSTK

=
β, Ʃπ xP -

+ βπ EX π) +ƩxiuT
SSTX

=
β. SSTx + / : π)ui
Ex -

SSTX

=β+ 之 ω i =β itEurwi

b> E( [β -

β ,) ,

U: ) =
E ( Eu: wi ) a )
=
π EEEui ωi )
=
π E ( E) uω i+ Enω i) )
Cor ( ui xi )
E(
,
)
=
+ a Ʃ w:
ssTx
=
a ( O+ 0 )

= O

c ) Bi =

y -
π T
β
= 坑 Ʃ Yi -

π (β , + Ewini )

=
E( β +β o xi + Ui ) -

πβ , + π ( Ewini )
=
Bo + R π+ a- πβ . +λ ( β^ -β )

N
Bot π -

[ -

β 1)
d) Var ( β) o
=
Var ( Bo +a -
- )π ) β

=
Varla ) t VarliB Bia andB areancorrelated
Varlh Ʃ ui ) =Var β
(^-
(
=
)
+π β
)

( πPVur ( β^
)
Var (Eu: ) +

=
者 ( no2 ) +^
作
=
兵 +
SSTX
二
父
、
方
氏
。
父
一
方
 (c) (5 points) Show that ˆ0 can be written as ˆ0 = 0 + ū ( ˆ1 1 )x̄.
ˆ
(d) (5 points) Use parts (b) and (c) to show that Var( 0 ) = /n + 2 (x̄2 )/SSTx .
2

(e) (5 points) Do the algebra to simplify the expression in part (d) to show that

2 1
P
n
n x2i
Var( ˆ0 ) = P
n
i=1

(xi x̄)2
i=1
P
. [Hint: SSTx /n = n 1 ni=1 x2i (x̄)2 .]
(f ) (5 points) Show that, when estimating 0 , it is best
Pto have x̄P= 0. What is
ˆ
Var( 0 ) in this case? [Hint: For any sample of numbers, i=1 xi > ni=1 (xi x̄)2 ,
n 2

with equality only if x̄ = 0.]

4 (15 points) John is interested in understanding the determinants of housing prices

in Hong Kong. He has collected data on all the apartments sold in 2023. The data
contains the following variables:

• price: transaction price in millions of HKD

• sqft: size of the apartment in square feet

• ycons: year of construction

• age: age of the apartment

• distMTR: distance to the nearest MTR station in meters

He estimated the following models:

ln(price) = 0 + 1 sqf t +u

(a) (5 points) What are the potential problems with this model? How would
you suggest John to improve his model? [Hint: Think about the potential omitted
variable bias. Be specific about the direction of the bias]
(b) (5 points) John believe that both the age of the apartment and its year of
construction may be important determinants of the price. Should he include both
variables in the model? Why or why not?
(c) (5 points) Suppose that distM T R is uncorrelated with sqf t, do you think
that distM T R should be included in the model? Why or why not?

*** END OF ASSIGNMENT 2 ***

3
 ) Var ( i
β )
兵
如
e =
+
Fπ
SSTx
^ (六 ) SSTK
0

+ =☆
=

SSTx
+ { R☆
2

[2
)
" x] π^
n -

(
=

SSTX
' "
(

& n Ʃ xi
=

sSTX
'

=
σ n
^
ExjI

xixPestimate
(

f) 70 best Bo we wanttominimize Var (β ]

Since ( xi π) ≥≥0
2
ExT ε
-

≥
( P
Exi
when theequality holds when x
l , only
= 0
2

( -X
,

Exi
'

Varl βδ ) xiP
& n ε(
so that =
is minimized to
Ʃ (XT π ) -

& n+ ( . ) =π^ 11
Σ
小
 1
ikely to be
Priceisnaffected by multiplefathus
4

a ) However only fuctor 7s mladed

,
one
,

This lead to omitted variable Dias

may
,