total marks = 35 + 22 + 18 + 15 + 35 + 5 + 15 + 15 = 160

5 35 marks

2 5
.

2 5
.

In
a) time series
analysis ,
a
stationary series

is a series whose statistical properties remain

constant over time . The series has :

Ye =
series

E (Yt) = 1 constant ..
independent of time

var(Yt) = o constant :·
independent of time

Cov(YE < Ye-s) = Cov(Yt-Ye s) + : correlation between

observations at different lags
C time intervals) remains the

same over time

Series has if it satisfies
weak
stationarity
above three
properties .

The distribution
strong stationarity :
of the
& cries remains the same over time .

b) Ye and
Suppose Xt are
non-stationary
and Yt is
regressed on Xt
.

Ye = Y ty + Vit UzyUztNN(O , On

Xt = Xt -
1 + Vzt

ye = a + BX + + Ut

Ho : B = 0
insignificant
H1 : B0
significant
If yt and not related
in
reality X+ are

to each other but regular - test rejects Ho

to conclude a
relationship between yt and Xt

then this is spurious.

regression

lead
Non-stationarity can to a spurious
relationship between variables.

any relationship
that
stationarity
identified
ensures

This allows for

is
genuine .
interpretation of
a
meaningful the series .
 Ye and
c) Suppose Xt are
non-stationary
and Yt is
regressed on Xt .
Suppose Ye
and Xt are related to each other then

we call them cointegrated

y+ = x + Bx+ + 4t

Run the Alt WU Et

regression = +

Ho : V = 0 >
spurious
Hi : ULO > cointegrated

If to then conclude
we
reject we can

that the variables are

cointegrated.

d) Unit root test is a test to check

for stationarity
Y = Y+ + et

Ho : p= L
non-stationary
11 : 04 >
stationary
The test statistic tall -
t ,
is compared
to the Dickey Fuller critical values · Whenever
↓ is less than the critical value the we
conclude
Reject
Ho to
stationary series
 e) If YUI(l) and Xu I(l) and this leads
to etw[(0) then Xt and Ye are

cointegrated . We use the error correction

short-run and
model to
analyze the

long-run relationship between Yt and X

Model is given as follows :

Ye = 8 + &y+ - + Sox+ + f, xt 1 - + Vt

Ye = 8 + &y+ - + Sox+ + f, xt 1 - + Vt

& y= = f + 10 -

1)yz+ Sox + S, x+ + Vt

=> S + 10 -
1)y +- 1 + foxt -
Soxz-1 + foxt + + f(x + + + Vt

= d + (0 1) YE-
+
+ 50x =
xt + +
(f0 + 81)xt -
1 + ut

- 8 + (0 -
1) YE+ 80Axz +
(f0 + f 1 )x + - 1 + vz

* Ye = 8 + (0 1)yz +
-
+ (f0 +f , (xz 1 + 80Bxt + Vt

(0-1)8 Y fotfl JoDt

= + + x +

o -1
*Yz =
=

(1 -
0)yt -1 -
B1 -
B2xt - + 80Dxt +V

first difference
-

first difference at xE

S
ofYt
error correction term

In the L R > ALEO VE

.

AYE = = 0

7 0 =
(0 1)(yz
- -1 - B1 - Baxt -

1)
> YE-1 = Bi + B2 Xt-1

In the S . R > AYt Dut Ve to

, ,

*Yz =
=

(1 -
0)yt - 1 -
B1 -
B2xt - + 80Dxt +V

=
-
a (YE -1 -
B1 -
B2Ut+ )+ 50Dxt + vz

0 -1 coefficient of Yt- 7 0 -110 it OLI

L
stationary
- cointegrated
04/ -
cointegrated
0 > 1 >
- not cointegrated
f) AR(2) model is as follows

Ye = a +
Byt -
1 + Yyz - - + Et

2 = constant

B,Y =
autoregressive coefficients
Et = error term

we assume that this is

stationary
no
multicolinearity and EENN1O-0)

This is a statistical model used to

forecast and analyze time series data.

to
It uses past values predict future
values

9) MA(2) model is as follows

Ye = c + Et + Viet+ + (2Ez =
2

c =
constant term
Et = error term
U <2 coefficients
=
moving average

we assume that this is

stationary
no
multicolinearity and EENN1O-0)
 MAC2) uses the last two errors
(residuals)
from last two
predictions to forecast
future values,

4) A series is Integrated of first order > I(l)

>
when it is
stationary after
taking
first difference
g
e .

y+ =
Y+ 1 + 2 t EtNN(OyE]

Ayt = Et

It = Et where It = AYt

ze is
stationary This shows that YE
III) becomes
is since it
stationary
upon taking first difference e.

A series is Integrated of second order - F(2)

>
when it is
stationary after
taking
second difference
g
e .

* DYE = 8 AYE-1 + HE UEN NLO L Oil)

80 that DAY+
stationary
Ze is
implies =

Hence if ZE = ADYt is
stationary it
shows that ye is I(2) since it becomes
difference
stationary upon taking second
 Note that :

2t =
By = =
(yz -

Yz -

1)
* DYz = ze =
z+ + =
(yt -
yz -

1) -

(Ye -1 -
Ye -

2)
=> AYE-AYt-1

single difference > DYE = YE- YE-1

double difference > ABYE =

DYt-AYt-1

* DYE = DYE-AY--

=
(0 1) Ye+
-
+ Ex -
(0 1)yz -
-
2 -
Et - 1

=
( 1)- Yz +1 -
Ye -2 + 2t -
Et -

=
(0 -
1) AYe - 1 + Dzt

= 25 ye + Ut

-DYE = WAY#1 + ME > run

to find after
regression stationary
second difference .
5 22

2 marks
3 2

4 3 3

18
4
marks

10

Aus za) et = a+ peteut VanN(0-2)

et = a + Plet + Vt

( -

p)et = a + vt

2t = a + 1 + x + pi + . . .

ut

1- P

Ct E a + v = + 0Vt + + p7vt = + ....

1- P

Elet) = a #f (time)
1 =
p

Var(et) = E Altime)
1 -
07
Cov(e +, (+ ) = Cov(a + Pet + + Vt yet -

1)
=> & Var(et+ ) = 0 Er #f (time)
I -
p2

satisfies
Hence this weak
stationarity
conditions

Ans 2 b) et = a+ Pety Ut

E(et) constant time

=
> it

independent

i) Ezet = et at time t
everything up to
and
including et is known

ii) Et-y et = Et =
y a + ve + put+ + 1V = z + ...

1 -
D

= a + pvzy + 15Vt 5 + 1Vt -

6 + --

1- P

before and
Elet) for all t
including t-4 is

known

iii) Eto CE =
Elet) = a

1 -
P
 "
iv) E (ez) = Var(et) + Elez)

= E + 12 > time independent

I -
p2

EE(e) =

Ona I -

v) Corlete) = Eleze) -
Elee) Eleza)

# (etez + ) =
Cov)et , et +
) + E(et)E(ez+ )

) ah
Cov(a
=>
+ pe ++ + v=
,
(t + +

(l 1)2
-

= PVar(et-1) + a
2

1 p
-

2
-
po + a

1 -
32 I -
p

vi Eletet 2) -
=
Covlet ,
et -z) + E(et) Elet -2)

= Cov(a + Pet - + Vt -
2t - 2 + a

1 -
P
 2
= P (OV(et-- et =

2) + a

1 -
P

prvar(ez-2)
2
=> + a

1 -
0

[
D2 of ↓ a

l -
02 1 -
s

2
= p
-
8 ↓ Ch

1 -
12 1 -
C
 Aus 3a) Impulse response question

yt =
2y= + + 44 MY NN(0 + 1)

i) one time shock to use

Y
ii) permanent shock to u t

Suppose for ease that yo = 0

i) ni = 1 n = H =.... = 0

Y1 = 2y0 + 41 = 0+ 1 =

Yz =
2 Y1 + 4z = 2(1) +0 = 2

Yy 2(2) 4 2
=
2yz + 43 = + 0 = =

Yy my 2(4) 0 8 23
=
2y3 + = + = =

(1 -
2 -
22 -
23 -
24
,
25 ..... )

ii) (1 = 4z = un = . . .
= 1

Y1 = 0 +1 = 1

Y2 = 2(1) + 1 = 3

Y3 = 2(3) +1 = 7

Yy = 2(7) +1 = 15

) /
<
3> L
>
15 > 31637127 7 000)
 Ans 3t) one time shock to ut

uY = 1422 =
43 = um =.... = 0

Suppose for ease that yo0 ro = 0 and

uy
there is no output shock = 0

0
5yz- 0 24t + + uYz - 0
5 ye- 0 2rt +
yt yt
.
=
. - .
=
.

rt = 0 2 yt+ .
+ 0 .
5 + HE

r1 1
= 0 .

2y0 + 0 .
50 + = 1 Y1 = 0 .

5 y0 - 0 .
200 = 0

rz = 0 .
241 + 0 .
5r + 0 X2 = 0 .

541 - 0 .
25/

= 0 .
2(0) + 0 .
5(1) = 0 5 .
= 0 .
5(0) - 0 .
2(1) = - 0 2 .

ry = 0 .

2y2 + 0 52 . + 0
y3 =
0 .

542 - 0 .
25r

= 0 .

2) -
0 .

2) + 0 .

5(0 5) .
= 0 .

57 -
0 .
2) - 0 .

2(0 5) .

-
=
0 . 1 -
0 .
1 = -0 . 2

-
= 0 .
04 + 0 25 .
= 0 21 .

yy = 0
543 0
253
. .
-

ru = 0 .

2y3 + 0 .
53 + 0

= 0 .
5) -
0 .

2) - 0 .
2(0 .
21)
= 0 .

2) -
0 .

2) + 0 .

5(0 21) .
=
- 0 .
1 - 0 . 042

= =
0 .
04 + 0 . 105 = 0 . 065 =
-
0 . 142
This
process is continued and values

potted to obtain trajectories

YE (0 ,
0
,
-0 . 2
,
-0 .
2 - -
0 .
142 ,
...
)

re (0 ,
1
,
0 5 .

,
0 . 21 -
0 . 065- ...
)

Since the shock is one-time thus over

time
Ye and It will back to
converge

their initial values yo and to.

Method 2

0
5yz- 0 24t + + uYz uY 0
yt
.
=
.

rt = 0 2 yt+
.
+ 0 .
5 + HE

YE 0 5 .
- 0 2
.
YE 1 uYt
#
-

↓E 0 2
. 0 .
5 rt - u

Y I 0 5 .
- 0 2 .
Yo uY,
#
=

↓, 0 2 . 0 .
5 To ur ,
 O
Y 0 5 .
- 0 2 .
G
#
S
=
=

↓
↓, 0 2 . 0 .
5 O A

Y2 0 5 .
- 0 2 .

YI # u]
=

r2 0 2 . 0 .
5
VI u2

0 5 0 2
Y2 .
- . O

=
=

r2 0 2 0 5 I
. .

0 5 0 2 0 2
43
.
- . - .

-
=
-

53 0 2
. 0 .
5 0 5 .

0 2
0 5 - .
0 2 0 142
XY
.
- .
= .

=
=

ry 0 2 . 0 .
5 0 .
21 0 . 065
 -

· (15 marks)

12

Ye = (t + It =
Bo + B yz , + It + Mt I

Endogenous = Yt- It

Hence we need two equations thus

that It = Lot kYt + Vt where K is
suppose
a constant

Y= = Bo + B Yt , + It + Mt I

It =
xo+ kYt + VE 2

Bo
(l-Bi) Yz =
It = + Mt

-
kYt + It = 2 + Vt

1 - BI -
1 Yt Bo t
Ut
-

-
k I Go Vt
It
 I
Yt 1 BI I
-

-
1 Bo 1 BI
-

1
-

Ut
-
-

t
-

k I
k I
It Go
-
-

Vt

I I Bo I
I Ut
=
I + I

(1 B1) 1 k 1 BI 20 1- B1 K 12 1- BI Ut
-
- -
-

Bo + 20 HE + Vt
= 1 + I
1- Bi -
K KBo + 20 -

20 I
B 1 -
Bi -
K KUt + VE-BIVE

Ye (Bo+ 20) (UE + Ut) 1- Bi- K

= 1 -
B1 - K t

(kUE + UE-Birt) 1-B1 - K

It IkBotho-GoBi) l-Bi-k

reduced form equation > run

regression
YE to &
Eit
=

Wi Est
It

b) we obtaine the estimates -E ,

Ere ,
Gert
through which we can derive values of
the unknown constants Go , Bo , K.

Yes it is possible to retrieve structural

from RFE explained above
parameters as .
 ·
(3)

(0)
135 marks (
(3)

(2 5).
(2 5)
.

(4)
(2 5)
.

(2 5)
.

(5)

a) Yt GDPE
=

explain
xt =
Consumption +

1 B12 YE Ye Yin

yesu
B10
b) =
+

B21 I Xt Bro Y21 Y2z CE-

 -
Y I B12 I B12
"
Yo You B12 "We
yes
1
= Bio
t

Xt B2 1
Bro B21 1
Y21 Y2z CE
B2 1 unt

1 B1z Bio
1 -
B1z uYz
-

1 - BLE Y Y12
= 1 + 1 1
Yet A
1 B12B21 -
B21 I B20 1- BIzB21 B21 1 1 - BIzB2-B2 I ne
Y21 Yuz
- -

Let

A1z uYz
Yt Gil all aiz Yt-1 DII
= + +

20 Az1 922 Xt b2 ↓ He
Xt
-
1 zz

Hence we
regress the
following RFE

Gil all Q12 Yt Elt

Yt 1
-

= + +

20 921 922 Xt -
1
Est
Xt

c) Eit = biute + brune

I I uYe-Biz u
1-Biz Bri 1 -
BizB2l

Eit = I ut =
B1 Ye
1-Biz B21

Est = baut + bau

= -
B2 uYt + 1 We
1- BizBzI 1- B12B21
 22t =
1 Ut -
BaU
Y
t

1- B12B21

f) Eit =
f(ui = -um ) =

+
-um)
22t =

g(u =

Hence we have an identification

problem. A shock inIt can come

from u4t and a shock in Est

can come from Utt . Hence we fail

to
identify the source of the shock.

we can solve the identification problem

and
by imposing
restrictions .
short-run long-run

uYz
d)i)uz = u
Y
-
upuN(0 + 1)
u4t

EU
O
F(ut) = =

*
Var (u z) I
Var(4t) = =

Var (44) I
 ii) E(zt) =
E(z t) ,

E(E2t)

=>
(u* =
B1 UE) =
6

I
I

E(U-B2 U =)
1- BizBri *
E

Var(Et) = 1
1 + (Biz)2

(l-BizB21)
2
1 +
(B21)

(1 Bir) (-BizBzi)
<
=
+

Bai)/(1-BizBzi)
"

(1 +

f) i) -
Variance Covariance = EIUEU +T)
U
matrixe
= (u - u2)uY
un
=

=
Gut Outux
=
I O

E I
GuuY Guy

12 =
E(z etT)
+ = E But (But)T
 = BEIMHET) BT = BIBT = BBT

Et = But

1 -
B1z
B =
bu biz - I

B2i I
buz
-

b21 1 - BIzB2

1 =
B12 I B2
BBT
-

= 1

21-BizBar)"
I
- B2 -
B12

I
1 + Bi22 -
Bri - Bin
-

(1-BizBri) -
B2-Bir Bri+1

Eziez 1 + Bi22 -
Bri - Bin

-Bri]2
=
=

-
B2l-Biz Bri+ 1
Grizz Ger

L
L
known unknown

9) E =
1+ Biz

(l-BizBzi)2
012 =
-

(B2l + Bir)
1- Biz B21

02 = 1 + B2
(1-B12B2)
 No . of equations = 3

No .
of unknowns = 2

We do not need to
impose restrictions
any
to find the unknown constants Bir and B21 .

However if we impose the short-run

restriction that Br = O so that yet has

no immediate
impact on
yo
then we

can obtain value of unknown B21

quickly as follows

E =
1+ Bizz > Gi =
1

(l-BizBzi)2
02 = -(B2l + Bir) < Giz =
-

B2
1- Biz B21

82 2 = It B2 7 = 1+ B2
(1-B12B2)

The SR restrictions helps us identify the

source of the shock which ut only
is
.
Since Et = But where we can write
our model in matrix form

zt = 00 + Q1 Zt+ + But
If a shock ut hits in to it's
cumulative
impact on It in the

long-run is
given by

Zetta = But + 0 , But + 0 , But + -

L >

impact int impact in t +

Stable VAR > Zetta = PBE

j = 0

= BEE
1 -

&

=
(-9)" Bet

C=
(1-4) "B captures the cumulative

impact of Ut on zt from t till o

To achieve identification we defince

BBT((1 17T
"
- = (CT = (1 (1)
-
-
0, )

-((l ]T
+
=
(1 4)
- -
4, )
 &. =
(l-0)c > achieved identification

Matrixe B is not
triangular like it is

with zero contemporaneous restrictions

identification-

In the Long-run restrictions

we
impose

on Matrix .
C

zt = 00 + Q1 Zt+ + But

Long-run : Zt = Zt + = 2

z -
0, 2 =
00 + But

(I-01) z = 00 + But

2 = (I-0)"00 +
(I-0)"But
C = (I =1) "B impose LR restrictions

on this matrix
 >
- stationary at 1%
significance
>
-
non-stationary at 10 % Significance
15 marks)

(5)

40) 115 marks

Ans6 DYE = a+
YYt + + Vt

Ho :
Y = 0
non-stationary
H : Y10 stationary

tan =
-
3 . 178 <= 3 43 . Ho
Reject >
stationary
tall = -1 9757-2 57
.
. Do not Reject Ho >
non-stationaly
 Ans (a) Ut
yt = x+ y+ + +

Y = a + yo + v

Yz = a + y1 + Vz = a+
(x + yo + v) + uz

= a + a + yo +1 + vz

Y3 = a + yz + Vy = a + (a + 2 + yo + V1 + vz) + V3

= a + x + 2 + yo + y1 + u + u3

t
+2
Yz = + yo +

i= 1
Vi

Ye = a + y+ + + v=
x yz = ta + yo +
1 F(yt) = Ex + yo + 2 F(vi)
= ta + yo +0

= Ex + yo = f (time)
t

2 Var (YE) = var(vi) = =Ev =

fetime)
i=1

The and
series
non-stationary
is can

become
stationary upon differencing
Ut
Yt = x+ y+ - +
 YE YE -
= G + Ut

* Ye = a+ Vz

E(DYt) = < # f (time)

Var(DYz) = En 7 (time)

Cov(y5 Yes)
- = Ev #f (time)

discussed
b) This is
already in Q1 b) < c) < <
 15 marks (

(5)

(10)

25
A =

5 15

2 5
270 = 30 - 25 = 570
7

515

principal minors are positive as required

251
B =

5 15 I

I I 2

25
270- =
5707
515

25
230 1) 5(10 1) +
1(5 15)
-
= - -
-

5 151
112
= 2(29) -

5(9) -
10 = 370

principal minors are positive as required

 25
4) i) FLEET) = 2 = Bu
5 15

bi bit b21
BBT =
O

D21 b22 6 b2z

= Di bl b21

bzibil b22+ b22t

bu = z > bu = 2

bllbri = 5 - b2 = 5

ba + 622 = 15 - baz = 15- 25 - 5

2 2

ii) EleET) = 251

5 15 I

I I 2

bi O bil 121 ba
⑧
BBT =

b 21 b22 b32
b2z E O

bu b32 b33 O G bus

 2
bi , bil D2l bu +31
=

bzi bi bar 122 + bzikbi + bazba32

basi bl bibli + zabat bartba+ has

di = 2 > bu = 2

↓ 11k2 = 5 > b2 =
5
2

bi b31 = 1 > b3 =
12

bribi =
5 > b2 = 5 2

bri + ba = 15 7 b2z =
15 - 25
2
= 5/2

b3i bi = 1 >
b = 12

buib31 bazbzu b32 1

51 +
+ = 1 = =

> b32 = 2 1 - 5

5 2

7 b32 =
-
3 2 -
=
3

2 5
16
b3 b21
, + barbz =
1 7 D3l7Dbzib2bz2 calculated
above

bi + barbar = 2 bab = 2- bar-bar

= 2 -
1 - 9
2 10

- 20 -
5 - 9 = 6

10 10

= 3

< 7