You are on page 1of 5

Koyck

Polynomial DL (PDL)

Pongsa Pornchaiwiseskul
Faculty of Economics
Chulalongkorn University
(c) Pongsa Pornchaiwiseskul, Faculty of Economics,
Chulalongkorn University

(c) Pongsa Pornchaiwiseskul, Faculty of Economics,


Chulalongkorn University

A form of short-term model


Explanatory variables

lagged values of dependent variable


current and lagged values of exogenous
variables

Three components

Auto-Regressive part
Distributed-Lag part
random part

(c) Pongsa Pornchaiwiseskul, Faculty of Economics,


Chulalongkorn University

Printed with FinePrint - purchase at www.fineprint.com

(c) Pongsa Pornchaiwiseskul, Faculty of Economics,


Chulalongkorn University

Yt = 0 + 1Yt 1 + 2Yt 2 + ... + pYt p


+ 10 X 1t + 11 X 1,t 1 + ... + 1, p1 X 1,t p1
+ 20 X 2t + 21 X 2,t 1 + ... + 2, p 2 X 2,t p 2
M
+ K 0 X Kt + K 1 X K ,t 1 + ... + K , pK X K ,t pK
+ t
(c) Pongsa Pornchaiwiseskul, Faculty of Economics,
Chulalongkorn University

k0

k1

(c) Pongsa Pornchaiwiseskul, Faculty of Economics,


Chulalongkorn University

Printed with FinePrint - purchase at www.fineprint.com

(c) Pongsa Pornchaiwiseskul, Faculty of Economics,


Chulalongkorn University

No lagged Y on the right-hand side.


Only one X
error terms could be ARMA
Declining effect over infinite lag or
pk =
Restriction on kj

Effect of current Xk at time t on Y is


distributed over pk periods.
k0 = Current or short-run effect of Xk on Y
kj = j-period delayed effect of Xk on Y
k = total or long-run effect of Xk on Y
where = + + ... +
k

Note that OLS is valid as lagged Y can be


uncorrelated with the error term. If the
random part is auto-correlated error terms,
OLS becomes invalid. No problem with
DL models (without AR part).
Without restriction on parameters, there
will be too many parameters (),
especially for very long lag.

j = ( ) j 0 , 0 < < 1, j = 0,1,...

k , pk

(c) Pongsa Pornchaiwiseskul, Faculty of Economics,


Chulalongkorn University

Estimation
Step 0 Guess
Step 1 Estimate ( , , ) for

kj
k0

Yt Yt 1 = 0 (1 ) + 0 X t + t t 1

Step 2 Use new and go back to


Step 1 until convergence

Lag (j)
(c) Pongsa Pornchaiwiseskul, Faculty of Economics,
Chulalongkorn University

= 0 + 0 X t + 0 X t 1 + 2 0 X t 2 + ... + t
= 0 + 0 (1 + L + 2 L2 + ...) X t + t
= 0 + 0 (1 L) 1 X t + t
(1 L)Yt = 0 (1 ) + 0 X t + (1 L) t

j = 0 + 1 j + 2 j 2 + ... + m j m

Yt Yt 1 = 0 (1 ) + 0 X t + t t 1
Yt = 0 (1 ) + Yt 1 + 0 X t + t t 1

Printed with FinePrint - purchase at www.fineprint.com

11

No lagged Y on the right-hand side.


Allow more than one X
error terms could be ARMA
polynomial effect over finite lag
Restriction on kj

Yt = 0 + 0 X t + 1 X t 1 + ... + p X t p + t

(c) Pongsa Pornchaiwiseskul, Faculty of Economics,


Chulalongkorn University

(c) Pongsa Pornchaiwiseskul, Faculty of Economics,


Chulalongkorn University

m << p, j = 0,1,..., p
10

(c) Pongsa Pornchaiwiseskul, Faculty of Economics,


Chulalongkorn University

12

Yt = 0 + 0 X t + 1 X t 1 + ... + p X t p + t

Yt = 0 + 0 Z 0t + 1Z1t + 2 Z 2t + ... + m Z mt + t

= 0 + 0 X t
+ ( 0 + 1 + 2 + ... + m ) X t 1

==> Unrestricted OLS is BLUE

+ ( 0 + 2 1 + 4 2 + ... + 2 m m ) X t 2
+ ( 0 + 3 1 + 9 2 + ... + 3m m ) X t 3
M
+ ( 0 + p 1 + p 2 2 + ... + p m m ) X t p
+t

(c) Pongsa Pornchaiwiseskul, Faculty of Economics,


Chulalongkorn University

13

Yt = 0

Z 0t

+ 0 (X t + X t 1 + ... + X t p )

+ 1 (X t 1 + 2 X t 2 + ... + pX t p )

(
(X

+ 2 X t 1 + 4 X t 2 + ... + p 2 X t p
+3
M

3
+
8
+
...
+
X
p
X t p
t 1
t 2

+t

(c) Pongsa Pornchaiwiseskul, Faculty of Economics,


Chulalongkorn University

Printed with FinePrint - purchase at www.fineprint.com

15

Case 1 = 0
Restriction

)
)

+ m X t 1 + 2 m X t 2 + ... + p m X t p

(c) Pongsa Pornchaiwiseskul, Faculty of Economics,


Chulalongkorn University

Z mt

)
14

0 1 + 2 + ... + (1) m m = 0

Effect starts from zero


==> Restricted LS is BLUE
(c) Pongsa Pornchaiwiseskul, Faculty of Economics,
Chulalongkorn University

16

Case p +1 = 0
Restriction

0 + ( p + 1) 1 + ( p + 1) 2 2
+ ... + ( p + 1) m m = 0

Effect dies down to zero.


==> Restricted LS is BLUE
(c) Pongsa Pornchaiwiseskul, Faculty of Economics,
Chulalongkorn University

17

Case 1 = 0, p +1 = 0
Restrictions

0 1 + 2 + ... + (1) m m = 0
0 + ( p + 1) 1 + ( p + 1) 2 2
+ ... + ( p + 1) m m = 0

Effect
starts
from
zero
and
finally
back
to zero. ==> Restricted LS is BLUE
(c) Pongsa Pornchaiwiseskul, Faculty of Economics,
Chulalongkorn University

Printed with FinePrint - purchase at www.fineprint.com

18