Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
+ + =
2
2
2 *
1
u
v
a
a
(7)
Finally, equation (7) is the solution of equation (6). The parameter
*
depends on the
source of economic uncertainty (the variability of output gap and inflation) and also on the
preference of policy maker in running monetary policy. By substituting equation (7) to
equation (5), we will obtain the optimal nominal interest rate.
In order to implement the model on Indonesian economy and the way Bank Indonesia set
its optimal interest rate, we need to estimate equation (3) and (4) in the model. The Guender
model lays freedom to do empirical testing to gain the following optimal parameter: (a
coefficient relating the interest rate and output gap), a (a coefficient relating inflation and
output gap), the stochastic disturbance variance of each equation, and the central bank
preference parameter .
7
3. The Model
The goal of this research is to develop the Guender model until we achieve an optimal
nominal interest rate for Bank Indonesias target of operation. The characteristics of the model
that we applied are several. First, the instrument used as an intermediate target is the nominal
interest rate, leaving no space for a debate whether money supply or interest rate to be used as
the chosen instrument. Second, the model is constructed only for countries applying the
inflation targeting policy. Finally, the model assumes that economic agents are forward
looking in forming their expectation, applying the rational expectation. However, the Guender
model assumes a closed economy, so the model has not included the exchange rate fluctuation
explicitly. The exchange rate fluctuation is our improvement over the Guender model.
The model characteristics follow the current world economy condition, where many
countries central bank apply the inflation targeting policy and set the nominal interest rate as
the intermediate target. Meanwhile, the rational expectation theory is also a modern theory in
explaining how people form their expectation, leaving the adaptive expectation theory behind.
In explaining the central bank characteristics, besides the output and inflation stabilization, the
exchange rate stabilization could also be another main focus of central bank policy.
The model applied in this research is to compile equation (3) to (5) in a system of equation
and after that we add the following equation (8):
( )
t t t t t t
w r r e E e + =
+
*
1
(8)
Variable e
t
is the real exchange rate as a function of the difference between real domestic
interest rate (r) and foreign interest rate (r*).
We take the first order condition of central bank objective function in equation (6) to get a
parameter determining the optimal interest rate. By following the flow of the model, we can
determine whether the actual Bank Indonesia interest rate policy (called the BI Rate) is
higher or lower or close to the estimated optimal value.
In short, we write down the system of equation as follows:
1. ( )
t t t t t t t t
v e e E b y E r y + + =
+ + 1 1
IS Relation
2.
t t t t t
u ay E + + =
+1
Forward Looking Phillips Curve
3. ( )
T
t t
r r + = Instrument Rule
8
4. ( )
t t t t t t
w r r e E e + =
+
*
1
Interest Rate Parity Condition
The purpose of the system of equation is to find the optimal solution value for parameter .
Having gained the estimated value of parameter and the estimated value of other coefficients,
we will gain the estimated value of the optimal central bank interest rate.
An estimation problem with a forward-looking model is the difference of time period
between a dependent variable with the equations error term. The existence of time period
differences cause estimation bias on variance of errors. In fact, variance of errors is the main
component to gain the estimated instrument rule. To overcome the time period difference
problem, Rudd & Whelan (2005) express the following forward looking equation:
t t
b
t t
f
t
x E + + =
+ 1 1
(9)
in term of:
1 1 1 + +
+ + + =
t t t
b
t
f
t
x (10)
Which are:
1 + t
( the actual inflation at time t+1),
1 + t
: (the error of expectation),
t
x : ( the
output gap or unemployment rate).
According to the theory of rational expectation, the error term at equation (10) cannot be
determined at time t, so that the coefficient
f
can be estimated consistently by using the same
variable at time period t or before, as a coefficient for
1 + t
. We estimate the coefficient
f
using the following method:
First, we do a backward-looking estimation to gain estimated values of
1
+ t
towards the
following equation:
t t t t
z x
3 2 1 1 1
+ + =
+
(11)
In equation (11), variable
t
z is other unknown variable stimulating inflation.
Second, we use the estimated values of
1
+ t
on the second stage regression between the
current inflation
t
as the independent variable towards the expected future inflation
(approximated by
1
+ t
), the lag of inflation rate
1 t
and the driving variable
t
x , reaching the
next regression equation:
t t t
b
t
f
t
x + + + =
+ 1 1
(12)
9
We estimate the Forward Looking Phillip Curve by using the two-stage regression method. We
also use the same method to estimate equation (3), (4) dan (8).
4. Solving the model
The model is solved by using a post-putative solution for the three endogenous variables:
t t t t
w u v y
13 12 11 10
+ + + = (13)
t t t t
w u v
23 22 21 20
+ + + = (14)
t t t t
w u v e
33 32 31 30
+ + + = (15)
Readers may find the mathematical derivation for the optimal solution in Appendix 1. the
solution of equation (6) is the parameter , derived mathematically with the following step: (i)
Substitute one equation to another to get the reduced form for output, inflation and exchange
rate. (ii) Determine the coefficient of error term and the intercept of each reduced form
equation of output, inflation and exchange rate. (iii) Create the equation of output variance and
inflation variance by using the reduced form equation. (iv) Take the first order equation for
equation (6) from the equation of output variance and inflation variance created in the previous
step (v) Generate the optimal value of parameter , whose value depends on other parameters
in the structural equation. The other parameters are: the source of uncertainty in the economy
(the output gap variability and inflation variability) and the preference of the policy maker.
(vi) Substitute the optimal parameter into equation (5) to get the value of optimal nominal
interest rate.
Having included the interest rate parity equation, the solution for the objective function is:
( ) ( )
( ) ( ) ( ) ( ) | |
2 2 2 2 2 2 2 2 2
2
1
1
1
w u v
b a b b a
b a
L
Minimize
+ + + + + +
+ +
= (16)
The solution for the first order condition of equation (16) is the optimal value of the following
instrument rule parameter:
( )
)
+ + +
+
=
2
2
2
2
2
2 *
1
u
w
u
v
b a
b
a
(17)
10
Readers may see how equation (17) differs from equation (7). Equation (17) explicitly
include the variance of errors of the exchange rate (
2
w
), derived from the interest rate parity
equation. Equation (17) stated that the optimal monetary policy not only affected by the
fluctuation of output and inflation, by also by the exchange rate fluctuation.
We apply the constructed model to explain the behavior of Indonesian economy and the
way Bank Indonesia is supposed to set the optimal interest rate. In that case, we need to
estimate equation (3), (4) and (8) empirically. The parameters to be estimated from the three
equations are: (a coefficient relating the interest rate and the output gap), a (a coefficient
relating the inflation and the output gap), (a coefficient relating the exchange rate with
domestic and international interest rate parity), the variance of stochastic disturbance for each
equation, and the central bank preference parameter . A necessary condition for the estimated
model is that it has a forward-looking explanatory variable.
5. The Empirical Evidence
To determine the optimal interest rate, we start from doing estimation towards three
structural equations, equation (3), equation (4) and equation (8). The best model we estimate is
the one making parameters of the three equations statistically significant. The estimation
results by using statistical software EVIEWS are shown in Appendix 2.
We do a two-stage regression for each equation. At the first stage, we estimate the expected
value of each dependent variable, then use the forecasted expected value at the first stage as the
proxy for the expected value at the second stage. Rudd and Whelan (2005) apply a two-stage
regression to avoid the model misspecification problem.
We apply the logarithm transformation for each variable in the structural equation to make
the value of estimated variance homogenous and no more dependent on the unit of data . The
Forward Looking Phillips Curve equation is being estimated by using logarithmic CPI variable
instead of inflation data, in order to generate the variance of error homogenous among
equations. The estimated parameter value of equation (3), (4), and (8) above is as follows:
a : 0.0185
b : 0.2725
: 0.1002
: -0.0353
11
The three estimated structural equation can be formulated as:
( )
t t t t t t t t
v e e E y E r y + + =
+ + 1 1
* 2725 . 0 * 1002 . 0 (18)
t t t t t
u y E + + =
+
* 0185 . 0
1
(19)
( )
t t t t t t
w r r e E e + + =
+
*
1
* 0353 . 0 (20)
We use the sum squared residual (SSE) from each equation to estimate variance of error term
of each structural equation, following the definition: variance of error term = SSE / (T-M).
Variable T = the amount of observation and M = the amount of estimated parameter. The
estimated values of the three variance of error are:
2
v
:
0.9937
2
u
:
0.3403
2
w
:
0.8312
Next, we substitute the values of estimated parameters and the value of variance of errors to get
the value of * in equation (17), equal to 0.318. Therefore, we conclude that the instrument
rule as the guidance for Indonesian monetary policy based on 1993-2006 data:
( )
T
t t
r r + = * 0.318 (18)
Equation (18) shows that there is a roughly 32% deviation of the difference between
inflation target and actual inflation, for the optimal interest rate from the long run interest rate.
The higher the difference between inflation target and actual inflation is, the wider the gap
between the long run interest rate and the optimal interest rate that Bank Indonesia is supposed
to achieve.
As example, in May 2008, Bank Indonesia set BI rate at 8.25%, increase by 25 basis point
from the previous period, 8%. If we follow the instrument rule we generate in equation (18),
and assume that the inflation target and the output growth target have the same value, with the
inflation target as much as 6.5%, BI is supposed to set the interest rate at 8.78% instead of
8.25%.
12
Table 1
Several Values of Interest Rate and Inflation
Based on the Instrument Rule
Variable Optimal Value Actual Value
Optimal interest rate 8.78 8.25
Inflation target 6.50 8.10
-- -- --
actual inflation (yoy) -- 8.96
previous period interest rate -- 8.00
Source : Bank Indonesia and estimation result
The instrument rule performed in equation (18) emphasizes the importance of achieving the
inflation target. The instrument rule provide space for the actual inflation to be higher 0.8%
than the inflation target, assuming that the current Bank Indonesia interest rate policy were
optimal. By that assumption, the increase of 25 basis point interest rate taken by yang Bank
Indonesia may only decrease inflation rate to the level 8.16%, not to its target level; i.e. 5+1%.
Our next discussion will be based on two simulations. First, a simulation that explains the
importance of the central bank to focus to its goal, whether the goal is output stability or
inflation stability. Second, a historical review comparing the set BI rate so far with the optimal
interest rate generated by the instrument rule in equation (18).
The focus of the central bank is shown by the parameter in the objective function
) ( ) (
t t
V y V L + = . The value = 1 shows that the central bank set the same weight between
output stability and inflation stability. The higher the value of parameter is, the more
attentive the bank central towards inflation stability. The simulation on many values of
parameter and resulted on a linear function between the two parameters, as shown in
Figure 1:
13
Figure 1
The Linear Function of Parameter toward Parameter
Notes:
= a parameter indicates the focus of a central bank on controlling inflation. The higher the value of the
parameter indicates that the central bank prefer controlling inflation to promoting economic growth.
= a parameter indicates the function of the gap between actual inflation with target inflation towards the
difference between optimal and actual interest rate.
Figure 1 shows that the higher the focus of the central bank towards inflation stability is, the
larger the role of differential proportion of inflation target and actual inflation on optimal
interest rate. It means that to achieve the inflation target, Bank Indonesia must set a higher BI
rate. The implication is that, should Bank Indonesia set a too low inflation target, it will have
heavier burden to pay the interest on Certificate of Bank Indonesia (Sertifikat Bank Indonesia-
SBI), the central bank securities instrument to do open market operation.
Figure 2
Inflation, BI Rate and Optimal Interest Rate
Based on 6.5% Inflation Target y-o-y
Period August 9 April 3, 2008
= 0.0704 + 0.2468
-
0.1
0.1
0.2
0.2
0.3
0.3
0.4
0.4
0.5
0 1 2
14
0
2
4
6
8
10
12
14
16
18
9
-
A
u
g
-
0
5
6
-
S
e
p
-
0
5
4
-
O
c
t
-
0
5
1
-
O
c
t
-
0
5
6
-
D
e
c
-
0
5
9
-
J
a
n
-
0
6
7
-
F
e
b
-
0
6
7
-
M
a
r
-
0
6
5
-
A
p
r
-
0
6
9
-
M
a
y
-
6
-
J
u
n
-
0
6
6
-
J
u
l
-
0
6
8
-
A
u
g
-
0
6
5
-
S
e
p
-
0
6
5
-
O
c
t
-
0
6
7
-
N
o
v
-
0
6
7
-
D
e
c
-
0
6
4
-
J
a
n
-
0
7
6
-
F
e
b
-
0
7
6
-
M
a
r
-
0
7
5
-
A
p
r
-
0
7
8
-
M
a
y
-
7
-
J
u
n
-
0
7
5
-
J
u
l
-
0
7
7
-
A
u
g
-
0
7
6
-
S
e
p
-
0
7
8
-
O
c
t
-
0
7
6
-
N
o
v
-
0
7
6
-
D
e
c
-
0
7
8
-
J
a
n
-
0
8
6
-
F
e
b
-
0
8
6
-
M
a
r
-
0
8
3
-
A
p
r
-
0
8
0
2
4
6
8
10
12
14
16
18
20
Actual BI Rate Optimal Interest Rate Inf lation
Source: www.bi.go.id
Figure 2 shows a simulation between the BI rate, optimal interest rate and actual inflation
since Bank Indonesia announced BI rate for the first time (August 9, 2005) up to date (April 3,
2008). There are two assumptions regarding how the figure is made. First, we assume that
Bank Indonesia set the same weight between output stability and inflation stability ( = 1).
Second, they set inflation target at 6.5%. We may see from Figure 2 that the higher inflation
rate is, the wider the gap between BI rate and the optimal interest rate.
6. Conclusion
The step taken by Bank Indonesia to set the inflation targeting policy framework
following the worldwide monetary policy - is accurate. However, for the case of Indonesia,
Bank Indonesia does not have an instrument rule as a guidance for policy implementation yet.
The purpose of this paper is to construct a monetary policy rule for Indonesia, developed from
the Guender optimal rule (2003) and from the empirical estimation of the model. Several
conclusions from the empirical estimation result are as follows.
First, the policy based interest rate has different direction with the inflation rate. However,
the changes of the focus of the central banks policy, whether they concern more either on
stability inflation or on output stability, neither have large nor significant impact on the
changes of the interest rate policy (BI rate in this case). Therefore, it is time for Bank Indonesia
15
to pay more concern on the growth of output, since it has only small impact on the decrease of
interest rate.
Second, the gap between actual inflation with inflation target contributes as high as 32% to
the increase of interest rate. As the gap increases by 1%, the optimal interest rate Bank
Indonesia is supposed to set increases by 32 basis points (0.32 %). The increase of BI rate as a
short-term policy instrument is mostly lower than the optimal interest rate generated by the
instrument rule of our research. In addition, Bank Indonesia need to choose well-timing to
increase the interest rate, as well as concern on increasing a higher long-run interest rate that is
valid in longer term.
Finally, the inflation target set by Bank Indonesia is relatively too low, make it difficult to
achieve. Therefore, Bank Indonesia needs to evaluate the target and needs more often and
regularly announce the inflation target, according to the past and present economic condition,
and the future expectation.
16
References
BEECHEY, MEREDITH, NARGIS BHARUCHA, ADAM CAGLIARINI, DAVID GRUEN
and CHRISTOPHER THOMPSON (2000), A Small Model of the Australian
Macroeconomy, Economic Research Department. Reserves Bank of Australia. Research
Discussion Paper.
DEBELLE, GUY and JENNY WILKINSON (2002), Inflation Targeting and The Inflation
Process: Some Lessons from an Open Economy, Economic Research Department.
Reserves Bank of Australia. Research Discussion Paper.
GUENDER, ALFRED V. (2003), Optimal Monetary Policy under Inflation Targeting Based
on an Instrument Rule, Economic Letters.78 (2003) 55-58.
LEITEMO, KAI & ULF SDERSTRM. (2005), Robust Monetary Policy in A Small Open
Economy, Bank of Finland Discussion Papers .20.
MAJARDI, FAJAR (2004), Pembentukan Model Makroekonomi Skala Kecil (SSMX),
Bagian Studi Sektor Rill. Direktorat Riset Ekonomi dan Kebijakan Moneter. Bank
Indonesia.
RUDD, JEREMY & KARL WHELAN (2005), New Test of The New-Keynesian Phillips
Curve, Journal of Monetary Economics. 52(2005) 1167-1181.
Website of Bank Indonesia at www.bi.go.id
17
Appendix 1
Substitute Phillips Curve eq(4) to instrument rule eq (5)
(A1) ( )
T
t t t t t
u ay E r r + + + =
+1
Substitute eq (A1) to IS Relation eq (3)
(A2) ( ) ( ) ( )
t t t t t t
T
t t t t t
v e e E b y E u ay E r y + + + + + =
+ + + 1 1 1
Substitute eq (8) to eq (A2)
(A3) ( ) ( ) ( ) ( )
t t t t t t
T
t t t t t
v w r r b y E u ay E r y + + + + + =
+ +
*
1 1
Substitute eq (A1) to eq (A3)
(A4) ( ) ( )
1 1 + +
+ + + + =
t t
T
t t t t t
y E u ay E r y
( ) ( ) ( )
t t t
T
t t t t
v w r u ay E r b + + + +
+
*
1
Reduced Form IS Relation
(A5) ( ) ( ) ( ) ( )
1 1
1
+ +
+ + + = + +
t t
T
t t t t
y E u E r b a y
( ) ( ) ( )
t t t
T
t t t
v w r u E r b + + +
+
*
1
Putative Solutions for endogenous variables:
(A6)
t t t t
w u v y
13 12 11 10
+ + + =
(A7)
t t t t
w u v
23 22 21 20
+ + + =
(A8)
t t t t
w u v e
33 32 31 30
+ + + =
The Expected Values
(A9)
10 1
=
+ t t
y E
(A10)
20 1
=
+ t t
E
(A11)
30 1
=
+ t t
e E
Solution for parameters in eq (26)
(A12) 0
10
=
(A13)
( )
b a + +
=
1
1
11
18
(A14)
( )
( )
b a
b
+ +
+
=
1
12
(A15)
( )
b a
b
+ +
=
1
13
Substitute eq (21) to eq (4)
(A16) ( ) ( )
t t
T
t t t t t t t
w r u ay E r e E e + + + + =
+ +
*
1 1
Substitute eq (26) and the expected values to eq (29)
(A17) ( ) ( ) ( )
t t
T
t t t t t t t
w r u w u v a E r e + + + + + + + =
+
*
13 12 11 10 1 30
Solution for parameters in eq (28)
(A18)
( )
b a
a
+ +
=
1
31
(A19)
( )
b a + +
=
1
32
(A20)
( )
b a
ab
+ +
=
1
1
33
Substitute eq (26) and the expected values to eq (2)
(A21) ( )
t t t t t
u w u v a + + + + + =
13 12 11 10 20
Solution for parameters in equation eq (27)
(A22)
( )
b a
a
+ +
=
1
21
(A23)
( )
b a + +
=
1
1
22
(A24)
( )
b a
ab
+ +
=
1
21
19
The objective function is:
(A25)
( ) ( )
( ) ( ) ( ) ( ) | |
2 2 2 2 2 2 2 2 2
2
1
1
1
w u v
b a b b a
b a
L
Minimize
+ + + + + +
+ +
=
The policy maker is assumed to set the parameter value that minimizes the loss function. The
parameter optimal value of the instrument rule is:
(A26) ( )
)
+ + +
+
=
2
2
2
2
2
2 *
1
u
w
u
v
b a
b
a
20
Appendix 2
EVIEWS output of estimation of the IS Relation equation
Dependent Variable: GDPGAP
Method: Least Squares
Date: 05/26/08 Time: 11:23
Sample(adjusted): 1993:2 2006:4
Included observations: 55 after adjusting endpoints
Convergence achieved after 3 iterations
Variable Coefficient Std. Error t-Statistic Prob.
C 11921.81 2312.085 5.156304 0.0000
AR(1) 0.415229 0.124628 3.331742 0.0016
R-squared 0.173173 Mean dependent var 11963.38
Adjusted R-squared 0.157573 S.D. dependent var 10924.12
S.E. of regression 10026.58 Akaike info criterion 21.29955
Sum squared resid 5.33E+09 Schwarz criterion 21.37255
Log likelihood -583.7377 F-statistic 11.10050
Durbin-Watson stat 2.033944 Prob(F-statistic) 0.001578
Dependent Variable: LOG(ABS(GDPGAP-GDPGAPF(+1)))
Method: Least Squares
Date: 06/05/08 Time: 13:38
Sample(adjusted): 1993:1 2006:3
Included observations: 55 after adjusting endpoints
Variable Coefficient Std. Error t-Statistic Prob.
C 8.475644 0.906749 9.347283 0.0000
LOG(ABS(USDF(+1)-
USD))
-0.272458 0.156727 -1.738426 0.0882
0.4*SBI -0.250437 0.074731 -3.351183 0.0015
0.6*SBC 0.256445 0.073092 3.508503 0.0010
R-squared 0.209304 Mean dependent var 8.515750
Adjusted R-squared 0.162792 S.D. dependent var 1.089477
S.E. of regression 0.996861 Akaike info criterion 2.901536
Sum squared resid 50.68032 Schwarz criterion 3.047524
Log likelihood -75.79224 F-statistic 4.500040
Durbin-Watson stat 1.693048 Prob(F-statistic) 0.007078
21
EVIEWS output of estimation of the Phillips Curve equation
Dependent Variable: CPI
Method: Least Squares
Date: 05/26/08 Time: 13:07
Sample(adjusted): 1993:3 2002:4
Included observations: 38 after adjusting endpoints
Variable Coefficient Std. Error t-Statistic Prob.
CPI(-1) 1.499453 0.147086 10.19441 0.0000
CPI(-2) -0.483853 0.151699 -3.189548 0.0029
R-squared 0.989905 Mean dependent var 60.88457
Adjusted R-squared 0.989625 S.D. dependent var 25.59550
S.E. of regression 2.607161 Akaike info criterion 4.805597
Sum squared resid 244.7024 Schwarz criterion 4.891785
Log likelihood -89.30634 Durbin-Watson stat 2.190726
Dependent Variable: LOG(ABS(CPI-CPIF(+1)))
Method: Least Squares
Date: 05/26/08 Time: 13:02
Sample(adjusted): 1993:3 2006:3
Included observations: 53 after adjusting endpoints
Convergence achieved after 9 iterations
Variable Coefficient Std. Error t-Statistic Prob.
LOG(GDPGAP) 0.050048 0.022769 2.198087 0.0325
AR(1) 0.630649 0.107693 5.855962 0.0000
R-squared 0.385247 Mean dependent var 0.564072
Adjusted R-squared 0.373193 S.D. dependent var 0.736845
S.E. of regression 0.583368 Akaike info criterion 1.797009
Sum squared resid 17.35625 Schwarz criterion 1.871360
Log likelihood -45.62075 Durbin-Watson stat 2.247950
Inverted AR Roots .63
22
EVIEWS output of estimation of the Interest Rate Parity equation
Dependent Variable: USD
Method: Least Squares
Date: 05/23/08 Time: 16:56
Sample: 1993:1 2006:4
Included observations: 56
Variable Coefficient Std. Error t-Statistic Prob.
SING 0.903575 0.238794 3.783906 0.0004
YEN 0.530061 0.164382 3.224573 0.0021
R-squared 0.961032 Mean dependent var 6738.769
Adjusted R-squared 0.960310 S.D. dependent var 3419.728
S.E. of regression 681.2910 Akaike info criterion 15.92092
Sum squared resid 25064500 Schwarz criterion 15.99325
Log likelihood -443.7857 Durbin-Watson stat 2.427764
Dependent Variable: LOG(ABS(USD-USDF(+1)))
Method: Least Squares
Date: 06/03/08 Time: 11:30
Sample(adjusted): 1993:1 2006:3
Included observations: 55 after adjusting endpoints
Variable Coefficient Std. Error t-Statistic Prob.
C 5.756505 0.177871 32.36329 0.0000
ABS(SBI-FEDRATE) 0.035255 0.011029 3.196471 0.0023
R-squared 0.161624 Mean dependent var 6.167423
Adjusted R-squared 0.145805 S.D. dependent var 0.986432
S.E. of regression 0.911686 Akaike info criterion 2.688644
Sum squared resid 44.05211 Schwarz criterion 2.761638
Log likelihood -71.93772 F-statistic 10.21743
Durbin-Watson stat 1.581512 Prob(F-statistic) 0.002346