You are on page 1of 21

Thoughts on Economics

79

Thoughts on Economics
Vol. 22, No. 04

Forecasting Inflation of Bangladesh Using VAR and


VEC Models
Md. Nezum Uddin*
ABSTRACT
This study examines the medium term forecasting of inflation rate for
the Bangladesh economy using four macroeconomic variables
employing annual data from 1981 to 2011.The methodology
employed in this paper uses unit root test and Johansens cointegation
tests followed by vector error correction (VEC) model, variance
decompositions (VDCs) and impulse response functions (IRFs)
techniques to examine dynamic relationship among the variables.
Three different vector error correction (VEC) models are estimated
starting from a two-variable model including money supply and CPI,
and sub-sequentially adding some financial variables such as real
GDP and nominal exchange rates. The analysis also uses diagnostic
tests, Portmanteau test and ARCH (multivariate) test, for choosing
comparatively more stable VAR (Vector Autoregressive) or VEC
models. The results show that there exists a statistically significant
long-run positive relationship between inflation and broad money
supply for the country as indicated by a statistically significant longrun positive relationship between CPI and broad money supply. The
empirical evidence also demonstrates that the shock in LRGDP and
LNER has no significant impact on LCPI. This also suggests that
Real GDP and nominal exchange rate has no forecasting power of
inflation in Bangladesh. With a view to containing medium-term
inflationary pressure, the continuation of existing cautious policy
stance of Bangladesh Bank may therefore be recommended. These
results have also important policy implications for both domestic
policy makers and development partners working .for the country.
Keywords: Cointegration, VAR, VEC, Macroeconomic variables,
Diagnostic checking, Variance decompositions (VDCs), Impulse
response functions (IRFs), Forecasting.

1. INTRODUCTION

80

Forecasting Inflation of Bangladesh ..

Economic forecasting is a rapidly developing field with wide applicability in


business and government. The increasing complexity of markets is fueling
the demand for professionals who possess an understanding of the
forecasting needs of organizations, the econometric tools to solve
forecasting problems, and the necessary computer skills to generate optimal
forecasts. Successful implementation and persuasion of monetary policy
largely depends on the efficiency and accuracy of forecasting major
macroeconomic events like inflation.
Rising rate of inflation has become a serious concern in Bangladesh in
recent years. The impact of rising inflation rate is being felt almost
everywhere. The prices of essential commodities has gone up, and so has the
cost of living. Inflation in FY11 rose by 8.8 percent from 7.3 percent in the
previous year. Point to point basis CPI inflation rose to 10.2 percent at the
end of FY11 from 8.7 percent at the end of FY10.
In this backdrop the paper attempts to measure and justify the overall stance
of monetary policy, considering not only the development of inflation, but
also other variables such as Broad Money, Real GDP, and Nominal
Exchange rate. Forecasts of inflation will represent a key ingredient in
designing policies which are geared toward the achievement of price
stability.
2. RELATED LITERATURE
In the context of both developed and developing countries, there have been
extensive theoretical and empirical researches to date that attempt to focus
on the relationship between inflation and relevant economic variables and
also forecasting of them. This section presents a brief review of some
available research works.
Callen and Chang (1999) forecast inflationary trends in India by estimating
two models that describe the inflationary process in India one is based on a
monetary approach and the other on output gap model. Callen and Chang
(1999) find that two monetary aggregates (M1 and M3) contain best
information about future inflation. The output gap model does not perform
well on Indian data. Hoffmaister (1999) makes an empirical exploration in
order to assess the predictability of inflation in Korea. He uses the VAR
model to calculate impulse responses to exogenous monetary policy. He
finds that inflation in Korea is as predictable as it was in inflation targeting
countries prior to their adoption of inflation targeting and concludes that the
empirical evidence supports the feasibility of inflation targeting in Korea.
Mallik and Chowdhury (2001) examine the short-run and long-run
dynamics of the relationship between inflation and economic growth for four

Thoughts on Economics

81

South Asian economies: Bangladesh, India, Pakistan, and Sri Lanka


applying co-integration and error correction models. They find that the
relationship between inflation and economic growth is positive and
statistically significant for all four countries and the sensitivity of growth to
changes in inflation rates is smaller than that of inflation to changes in
growth rates.
Sweidan (2004) examines whether the relationship between inflation and
economic growth has a structural breakpoint effect or not for the Jordanian
economy from the period between 1970 and 2003. He finds that this relation
tends to be positive and significant below an inflation rate of 2-percent and
the structural breakpoint effect occurs at an inflation rate equal to 2-percent.
Mubarik (2005) estimates the threshold level of inflation for Pakistan using
an annual data set from the period between 1973 and 2000. He employed the
Granger Causality test as an application of the threshold model and finally,
the relevant sensitivity analysis of the model. His estimation of the threshold
model suggests that an inflation rate beyond 9-percent is detrimental for the
economic growth of Pakistan. Thereafter Hafer and Hein (2005) compare
the relative efficiency of the widely used interest rate based forecasting
model and univariate time series model based on monthly data from the
United States, Belgium, Canada, England, France and Germany for the
sample period from 1978 to 1986. Using monthly data on the Euro rates and
the consumer price index (CPI) their results indicate that time series forecast
of inflation model produces equal or lower forecast errors and has unbiased
predictions than the interest rate based forecasts. Daoud et al (2008)
examine the long and short run equilibrium relationships between prices in
Palestine, money supply and prices in Israel by applying time series analysis
that includes cointegration for single equation, Johanson procedure for
multivariate cointegration system, and Granger causality analysis. The data
cover the periods 1970 to 2007. The result of the Granger causality test
results show that the prices in WB Granger cause for the prices in GS while
this is not true in the opposite direction.

3. OBJECTIVE OF THE STUDY


The main objective of the paper is to forecast Bangladeshs inflation using
VAR and VEC model. To fulfill the same, the following sub-objectives are
shown:

82

Forecasting Inflation of Bangladesh ..

Checking for the cointegration of the variable, if any, tested by


Johansen cointegration test
Applying VAR or VEC techniques to display the economic relationship
among consumer price index, nominal exchange rate, real GDP, broad
money supply in Bangladesh
Forecasting the data using best fitted model
Finally, making recommendations that can assist future policy in
Bangladesh

4. DATA AND ESTIMATION METHODOLOGY


4.1 DATA

This study has made use of secondary data analysis, and data on the CPI,
broad money supply and the nominal exchange rate are collected from
Bangladesh Bank, Monthly Economic Trend, and on real GDP are collected
from various issues of Statistical Yearbook of Bangladesh, Bangladesh
Bureau of Statistics, and Bangladesh Economic Review. To estimate the
autoregressive parameters of the models, annual data from 1981 to 2011 are
used. Here, real GDP is calculated at 1995-96 constant market prices, while
the base year for CPI is 1995-96, i.e., 1995-96 = 100. All the variables are in
logarithm form. Definitions of the variables are, LCPI = Log of Consumer
Price Index, LM2 = Log of Broad money supply, LRGDP= Log of Real
GDP, LNER = Log of nominal exchange rate
4.2 RESEARCH METHODS
In this paper a series of tests such as unit root, cointegration, diagnostics test,
vector error correction models (VECM) have been carried out. Granger
causality test was also conducted to test the causality. These tests examine
both long-run and short-run relationships between consumer price index and
the economic variables. The VECM analyses provide some support for the
argument that the lagged values of macroeconomic variables have a
significant influence on the inflation. They also provide important
information to forecast future values of the inflation. Econometric Software
Programming R is used for finding the estimates.
4.3 LAG LENGTH SELECTION

Thoughts on Economics

83

The lag length for selected macroeconomic variables may be determined


using model selection criteria. The three most common information criteria
are the Akaike (AIC), Schwarz-Bayesian (BIC) and Hannan-Quinn (HQ).
The AIC criterion asymptotically overestimates the order with positive
probability, whereas the BIC and HQ criteria estimate the order consistently
under fairly general conditions if the true order p is less than or equal to p max.
From Table 4.1, Log of real GDP has lagged one and all other selected
variables have lagged two.
Table 4.1 Lag Length Selection
LM2
Lag

AIC

-7.090

-7.238

-7.167

-7.151

-7.073

BIC

-7.050

-7.184

-7.099

-7.069

-6.978

HQ

-6.944

-7.043

-6.923

-6.858

-6.7318

LRGDP
Lag

AIC

-9.845

-9.767

-9.698

-9.619

-9.689

BIC

-9.804

-9.713

-9.630

-9.538

-9.594

HQ

-9.698

-9.572

-9.454

-9.3267

-9.3478

LNER
Lag

AIC

-6.8877

-7.359

-7.3014

-7.2376

-7.327

BIC

-6.847

-7.3053

-7.2338

-7.156

-7.232

HQ

-6.741

-7.164

-7.0576

-6.9451

-6.9859

LCPI
Lag

4.4

AIC

-7.539

-7.838

-7.7616

-7.8135

-7.7548

BIC

-7.49849

-7.784

-7.694

-7.7323

-7.660

HQ

-7.3927

-7.6436

-7.5178

-7.5209

-7.4135

ORDER

STATIONARITY

OF

DIFFERENCING

AND

TESTING

FOR

84

Forecasting Inflation of Bangladesh ..

In Table 4.2, results of the unit root tests on concerned variables have been
reported. The tests for non-stationarity show that LCPI is non-stationary
based on ADF, PP, and KPSS tests. The first difference of LCPI is stationary
based on PP, and KPSS tests although the ADF test fails. Since the PP and
KPSS tests are preferable to ADF it can be concluded that LCPI is also
stationary, I (1). The results also shows that at 5% significance level all the
other variables are non stationary at levels since the calculated tau values are
less in absolute terms than their critical values and the variables are found to
be stationary only when tested at first difference. Thus, they are integrated of
order one -I (1). Each of these variables becomes stationary if it is
differenced once.
Table 4.2 Unit Root Test Statistics
Variab
les

LCPI

LRGD
P

LM2

LNER

Level/di
fference

L
a
g
s

ADF

PP

KPSS

Intercep
t

Intercept
& Trend

Interce
pt

Intercep
t
&
Trend

Intercept

Intercept
& Trend

DECISIO
N

Level

-1.22

-3.2118

-2.90

-3.2911

1.0802

0.2282

Non
stationary

First
Differen
ce

-2.13

-1.5514

-3.65*

-3.5418*

0.5205***

0.1938***

Stationary

Level

2.177

2.3703

0.4841

-3.15

1.0834

0.242

Non
stationary

First
Differen
ce

1.37

-0.5595

-8.69*

-11.75*

0.1375

0.06*

Stationary

Level

0.0344

-2.2516

-0.984

-2.3617

1.0907

0.1442***

Non
stationary

First
Differen
ce

-4.032*

3.2358**

-3.08*

-3.23**

0.2172*

0.1745***

Stationary

Level

-2.0322

-3.5534

1.874

-2.4843

0.4317*

0.0715***

Non
stationary

First
Differen
ce

-3.4622*

-3.3897**

3.00*

-2.9827

0.1375**

0.0569***

Stationary

Notes:
1. L means the series in log level, * and ** means significant at 5-percent and
10- percent levels, respectively.

Thoughts on Economics

85

2. Lag length for ADF tests have been decided on the basis of AIC.
3. Maximum Bandwidth for PP and KPSS tests have been decided on the basis
of Newey-West (1994).
4. All tests have been performed on the basis of 5-percent significance level
using R version 2.12.2
5. The ADF and PP tests are based on the null hypothesis of unit roots while
the KPSS test assumes the null hypothesis of stationarity.
6. For KPSS test, *, ** & ***means the null hypothesis accepted at 1percent,
-percent and 10- percent levels, respectively.
4.5 GRANGER CAUSALITY TESTS
Table 4.3 shows the results of the Granger causality test conducted for the
respective variables of the model. The null hypothesis in each case is that the
variable under consideration does not Granger cause the other variable.
Table 4.3 is analyzed at 5% significance level. The tests found that money
supply does Granger cause consumer price index (inflation rate). It is
evident, therefore, that the direction of causality runs from money supply to
consumer price index. Hence the causality here is unidirectional.
Bidirectional causality was found between LGDP and LCPI (inflation rate).
Unidirectional causality was found between LNER and LCPI (inflation rate).
It is evident, therefore, that the direction of causality runs from nominal
exchange rate to money supply. Hence, causality is unidirectional. The test
also found that nominal exchange rate does Granger cause real GDP. Hence,
causality is unidirectional.
Table 4.3 Results of Pair wise Granger causality Tests

86

Forecasting Inflation of Bangladesh ..

Null Hypothesis:

F-Statistic

p-value

LCPI does not Granger cause LRGDP

2.0461

0.0314*

LRGDP does not Granger cause LCPI

0.2084

0.0427*

LCPI does not Granger cause LNER

0.8233

0.4456

LNER does not Granger cause LCPI

3.2118

0.0498*

LCPI does not Granger cause LM2

0.8806

0.4217

LM2 does not Granger cause LCPI

5.018

0.0108*

LRGDP does not Granger cause LNER

0.2473

0.7820

LNER does
LRGDP

5.5143

0.0073*

LRGDP does not Granger cause LM2

7.5977

0.0014*

LM2 does not Granger cause LRGDP

0.2377

0.7894

LNER does not Granger cause LM2

1.2925

0.2848

LM2does not Granger cause LNER

1.7277

0.1895

not

Granger

cause

4.6 MODEL SPECIFICATION


This study examines both long run and short-run relationship among
consumption price index and the macroeconomic variables and forecast on
them. Here three different models have been considered. In the model,
consumption price index (CPI) is considered as dependent variable and real
gross domestic product (RGDP), broad money (M2), the nominal exchange
rate (NER) as independent variable. All variables are transformed into
natural logs. We expect that there will be long-run cointegrating
relationships between the variables consistent with macroeconomic
fundamentals. Further, there should be short run relationships between these
variables. The forms of these variables in the model are in the order of
integration one I (1), which is explained in Table-4.2 and Table-4.3. So the
models are as follows.
LCPIt=
1+11
LM2
.1
LCPIt=
2+21
LM2t+
22
.2

LRGDPt

e1t
+

e2t

Thoughts on Economics

87

LCPIt= 3+31 LM2t + 32LRGDPt + 33 LNERt+ e3t


3
4.7 DIAGNOSTIC TESTS OF VAR (P)
Given the diagnostic test results from table 4.4, we conclude that a VAR (1)
specification might be too restrictive for all three models. Although some of
the stability tests do indicate deviations from parameter constancy, the timeinvariant specification of the VAR (2) for model-1 and model-3 and VAR (3)
for model-2 will be maintained as tentative candidates for the following
cointegration analysis.
Table 4.4 Diagnostic tests of VAR (p)

Variable

lag

Model

LCPI,

-1

LM2

LCPI,
Model
-2

-3

Test

p~value

MARC
H

p~value

Portmant
eau Test

p~value

p=3

3.3182

0.5061

52.2033

0.2143

42.8488

0.8132

p=2

3.6733

0.452

47.944

0.3543

46.2872

0.8194

p=1

15.879

0.003186

39.9147

0.6867

61.8743

0.409

p =4

0.1081

0.9908

126

0.9992

106.1218

0.5331

p=3

6.1175

0.1060

132

0.9971

114.8545

0.5388

p=2

23.534

3.125e-05

138

0.9913

105.3269

0.9096

p=1

18.874

0.0002903

144

0.9775

124.7901

0.7247

p =4

10.810

0.2127

210

232.3488

0.02478

p=3

6.7492

0.5639

220

201.4034

0.6157

LM2,
LRGDP

Model

JB-

LCPI,
LM2,
LRGDP,

88

Forecasting Inflation of Bangladesh ..

LNER

p=2

21.627

0.1356

230

179.3016

0.9874

p=1

53.567

8.377e-09

240

222.0211

0.7914

4.8 JOHANSEN TEST FOR CO-INTEGRATION


From table 4.5 the computed the maximum Eigen value test statistics and
their corresponding critical values indicate that the null hypothesis of no cointegration (r = 0) can be rejected under both of these tests at both 5-percent
and 1-percent levels of significance. According to the maximum Eigen value
test we conclude that there is one co-integration equation at 1- percent levels
of significance for model-3 (p = 2). This implies a long-run relationship
between inflation and broad money in Bangladesh for model-1, a long-run
relationship among inflation, broad money and real GDP in Bangladesh for
model-2, and a long-run relationship among inflation, broad money, real
GDP and the nominal exchange rate in Bangladesh for model-3.

Table 4.5 Johansen Test for Co-integration (Maximum eigen value Test)
Test
Variable

H0

H1

Statistics

Critical Values

Conclusion

Thoughts on Economics

89

r<=1

r=2

23.16

28.76

23.11

25.54

30.34

r<=2

r=3

11.98

14.98

16.85

18.96

23.65

r<=3

r=4

10.22

12.01

10.49

12.25

16.26

4.9 SHORT-RUN DYNAMIC ADJUSTMENT USING VECM FOR


MODEL-1
The Vector Error Correction Model (VECM) allows us to make inferences
on the long-run impacts of the variables in levels to those in differences. The
variables of the model-1 are one co-integrating equation; then there exists an
error correction model (i.e VEC (2) with r=1) that may take the following
form
Model---1

LCPIt=11+11
LCPIt-1+
1a

12

LM2t=12+21
LCPIt-1+
..1b
2t

12

LM2t-1+
LM2

Dt
+

t-1

+1t
Dt+

The estimated coefficients of the error correction term (long-run effects) and
the lagged values of the two series (short-run effects) are presented in Table
4.6.
Table 4.6.The Error Correction Model-1
Variables equations

Error Correction cointEq1

LM2t-1

LM2t

LCPIt

-0.3039

0.02501

(0.0590)

(0.03733)

{-5.151*}

{-0.670}

[2.83e-05]

[0.50918]

2.6801

-0.2063633

(0.5103)

(0.3228424)

{5.252 *}

{-0.639}

[2.20e-05]

[0.52874]

0.1794739

0.2075785

(0.1297705)

(0.0821030)

90

Forecasting Inflation of Bangladesh ..

{1.383}

{2.528*}

[0.179]

[0.01845]

0.2840767

0.4338898

(0.2080724)

(0.1316429)

{1.365}

{3.296*}

[0.185]

[0.00304]

Multiple R-squared:

0.9677

0.9323

Adjusted R-squared:

0.9623

0.921

179.8

82.59

p-value: < 2.2e-16

p-value: 1.135e-13

0.0321

0.02031

LCPIt-1

F-statistic:
Residual standard error:

Notes: Values in parentheses are (Std.Error), {t value}, [p-value],


respectively, and * & ** indicate significant at 1-percent and 5- percent
levels, respectively, comparing critical value of t-statistics.
The empirical results show the existence of short-run and long-run
relationships between CPI and M2 in Bangladesh. This also implies shortrun and long-run relationships between inflation and broad money supply in
the country. The estimated coefficients of the error correction term is
significant at 1-percent level from M2 to CPI and vice versa with
appropriate (i.e., negative) signs. This means that in the long-run if the two
series are out of equilibrium, M2 will adjust to reduce the equilibrium error
and vice versa is not significant in the model-1. In other words, it shows that
30 percent (error correction term -0.3039) of the deviation of the M2 from
its long run equilibrium level is corrected each year. The estimated results in
the ECM also show that a short-run change in CPI and M2 affect M2
positively and statistically insignificant, and vice versa is significant with
positive in sign.
Table 4.7. Diagnostic checking For VEC (2) with one cointegrating
equation and VEC (3) with one cointegrating equation for model-2
Name of the test

Chi-squared value

df

p-value

Thoughts on Economics

1. Normality test(JBTest)
2.

serial

correlation

test(Portmanteau Test)
3.ARCH (multivariate)

91

VEC (2)

VEC (3)

VEC (2)

VEC (3)

VEC (2)

VEV (3)

with r=1

with r=1

with r=1

with r=1

with r=1

with r=1

0.338

0.5926

6.8183

4.626

93.9629

106.36

129

120

0.9912

0.8085

138

132

180

180

0.9913

0.9971

The diagnostic tests for the VECM equation confirmed its coefficient
stability. The Jarque-Bera test did not reject the null hypothesis of a normal
distribution of the residuals (at 5 percent significance level) and the
Portmanteau test as well as the correlogram of squared residuals did not
show any autocorrelation or ARCH in the residuals. The short-run Dynamic
Adjustment using VEC for model-1 and model-3 has been also done,but it is
not shown in this paper.

5. ANALYSIS OF EMPIRICAL FINDINGS


Forecasted points estimates of each model by making necessary adjustments
based on the latest available information for the selected macroeconomic
variables are reported in Table 5.1. The forecasted outcome for inflation of
the current study from model-3 indicates that the rate of inflation will be
about 6.86 percent, 5.28 percent, 5.72 percent, 6.25 percent and 6.23
percent, respectively, in fiscal year 2011-12, 2012-13, 2013-14, 2014-15,
and 2015-16 as against 8.8 percent in fiscal year 2010-11, which are on
average the same as the predicted inflation as shown in the medium term
macroeconomic framework (Ministry of Finance, 2010). On the other hand,
the broad money supply is predicted to grow by 11.64 percent, 14.73
percent, 17.08 percent, 16.07 percent and 15.16 percent, respectively, in
fiscal year 2011-12, 2012-13, 2013-14, 2014-15,2015-16, as against 21.4
percent in fiscal year 2010-11. And the forecasted Real GDP growth will be

92

Forecasting Inflation of Bangladesh ..

about 6.94 percent, 6.43 percent, 5.81 percent, 5.55 percent and 5.50
percent, respectively, in fiscal year 2011-12, 2012-13, 2013-14, 2014-15,
2015-16 as against 6.7 percent in fiscal year 2010-11. And the forecasted
Period Average Exchange rate will be about 77.78 Taka/USD, 80.42
Taka/USD, 83.46 Taka/USD, 86.89 Taka/USD and 90.3 Taka/USD,
respectively, in fiscal year 2011-12, 2012-13, 2013-14, 2014-15, 2015-16 as
against 74.15 Taka/USD in fiscal year 2010-11. A gradual depreciation of
Taka-USD will occur in the medium term in line with current policy
intervention (Table 5.1).The findings of model-1 and model-2 shown in
Table 5.1

Table 5.1, Forecast value for selected three models

Forecast
value for
model-2
on

15.
9

408995
.7

6.2
2

16.
78

432959
.7

5.8
6

Infl
atio
n
Rat
e

Broad
Money
(M2)(Taka
in crore)

M2
Gro
wth

20112012

261.4
4

8.4
7

523996.
35

18.
95

20122013

282.6
8

8.1
3

611573.
24

16.
71

20132014

304.3
2

7.6
5

703139.
09

14.
97

20142015

326.0
8

7.1
5

801187.
27

13.
94

20152016

348.0
1

6.7
2

909872.
72

13.
57

20112012

260.7
5

8.1
8

510542.
93

20122013

279

596234.
9

Year

Forecast
value for
model-1
on
VEC(2)

Real GDP
(Taka in
crore)

Rea
l
GD
P
Gro
wth

CPI
(Base:
19951996)

NER

NE
R
Gro
wth

Thoughts on Economics

VEC(2)

Forecast
value for
model-2
on
VEC(3)

Forecast
value for
model-3
on
VEC(2)

93

201
3201
4

297.6
1

6.6
7

705674.
95

18.
36

456968
.6

5.5
5

20142015

318.0
8

6.8
8

838340.
82

18.
8

481128
.7

5.2
9

201
5201
6

340.7
5

7.1
3

991803.
22

18.
31

505816
.9

5.1
3

20112012

250.4

3.8
9

498380.
94

13.
13

412800
.6

7.2
1

20122013

259.7
5

3.7
3

577746.
23

15.
92

439340
.5

6.4
3

20132014

277.5
6

6.8
6

689119.
52

19.
28

464046
.5

5.6
2

20142015

298.9
4

7.7

809352.
64

17.
45

489544
.2

5.4
9

20152016

319.4
2

6.8
5

937423.
09

15.
82

516303
.5

5.4
7

20112012

257.5
6

6.8
6

491796.
36

11.
64

411790
.5

6.9
4

77.
78

4.9

20122013

271.1
5

5.2
8

564254.
12

14.
73

438283

6.4
3

80.
42

3.4

20132014

286.6
4

5.7
2

660650.
34

17.
08

463749
.6

5.8
1

83.
46

3.7
8

20142015

304.5
7

6.2
5

766814.
35

16.
07

489505

5.5
5

86.
89

4.1
1

20152016

323.5
5

6.2
3

883070.
22

15.
16

516442
.9

5.5

90.
3

3.9
3

The variance decomposition for model-3 is presented in Table-5.2. At the


one-year time horizon, money supply explains around 0.11%, real GDP
explains 0.91%, and period exchange rate explains 2.37% of the total
variation of inflation rate while 96.61% of variations in prices are explained
by their own shock. At the five- year horizon, money supply explains 2.87%,
real GDP explains 0.44%, and nominal exchange rate explains 13.51% of
the total variation of inflation rate and 83.18% of variations in prices are
explained by their own shock. At all time horizon, more than 4.56% of the
variance in prices is explained by money supply while their own innovations

94

Forecasting Inflation of Bangladesh ..

explain 84.94% and real GDP explain 0.6%, period exchange rate explains
9.90%.Therefore, The shock in LM2 has a positive impact on LCPI for all
time periods. The shock in Real GDP has a negative impact on LCPI for
medium term periods.
Table-5.2. Variance Decomposition for VEC model of LM2, LRGDP,
LNER and LCPI.
Variance Decomposition of LM2
Forecast

LM2

LRGDP

LNER

LCPI

1.0000000

0.00000000

0.00000000

0.000000000

0.9281112

0.04691942

0.02190837

0.003060972

0.8730181

0.11087654

0.01199048

0.004114879

0.8243135

0.14594943

0.01454673

0.015190304

0.7932603

0.16294092

0.01909392

0.024704818

0.7711308

0.17530461

0.02219857

0.031366073

0.7518035

0.18553372

0.02552378

0.037139008

0.7351727

0.19332417

0.02914722

0.042355903

0.7215745

0.19918010

0.03247602

0.046769331

10

0.7104130

0.20382182

0.03534363

0.050421531

Horizon

Variance Decomposition of LGDP


Forecast

LM2

LRGDP

LNER

LCPI

0.6306442

0.3693558

0.000000000

0.000000000

0.6179559

0.3686398

0.005048734

0.008355577

0.6392246

0.3419106

0.004911274

0.013953483

0.6640160

0.3182072

0.003460403

0.014316414

0.6803065

0.3017387

0.004649569

0.013305183

0.6908565

0.2895759

0.007287709

0.012279853

0.6986915

0.2793220

0.010689863

0.011296607

0.7046832

0.2703145

0.014670275

0.010332046

0.7091762

0.2624614

0.018922521

0.009439930

10

0.7125851

0.2556078

0.023155961

0.008651171

Horizon

Variance Decomposition of LNER


Forecast

LM2

LRGDP

LNER

LCPI

Thoughts on Economics

95

Horizon
1

0.02073619

0.04086744

0.9383964

0.000000000

0.04963582

0.04745808

0.8989852

0.003920943

0.06558947

0.06173850

0.8652608

0.007411260

0.07131477

0.07378131

0.8474333

0.007470586

0.07439737

0.08127628

0.8378113

0.006515078

0.07734745

0.08639471

0.8306029

0.005654945

0.07997100

0.09052459

0.8245386

0.004965796

0.08205271

0.09394412

0.8196184

0.004384773

0.08372893

0.09672721

0.8156489

0.003894935

10

0.08514613

0.09901781

0.8123473

0.003488789

Variance Decomposition of LCPI


Forecast

LM2

LRGDP

LNER

LCPI

0.001113351

0.009082615

0.02367268

0.9661314

0.009912304

0.013689889

0.08065954

0.8957383

0.020136422

0.009216613

0.12917521

0.8414718

0.025051094

0.005884118

0.14077007

0.8282947

0.028676439

0.004392340

0.13514071

0.8317905

0.032599881

0.003880304

0.12682869

0.8366911

0.036470520

0.003996506

0.11905307

0.8404799

0.039907461

0.004515394

0.11176028

0.8438169

0.042920881

0.005230819

0.10501476

0.8468335

10

0.045602550

0.006018612

0.09900500

0.8493738

Horizon

Figure5.1 shows the impulse response functions (IRF) for four variable VEC
model. The IRF results show that, the shock in LM2 has a positive impact on
LCPI for all time periods. Therefore, a positive shock to money supply
persistently raises the CPI in the long run. The shock in LRGDP and LNER
has no significant impact on LCPI. This suggests that nominal exchange
rate has no forecasting power of inflation in Bangladesh. The response of
prices to their own shocks is again positive and significant at all time periods
and persistent indicating the inflation inertia. The same finding is also done
for model-1 and model-2, which is not reported in this paper.

96

Forecasting Inflation of Bangladesh ..

Figure 5.1 Impulse Responses for VEC (2) models of LM2, LRGDP,
LNER and LCPI
6. CONCLUSIONS AND POLICY IMPLICATIONS
The main objective of the study is to find out the significant factors behind
inflation in the context of Bangladesh during the period FY 1981- FY 2011
and forecasting the data to Medium-Term Outlook (FY 2012- FY 2016). In
this regard, the paper studies the historical trends of inflation and other
macroeconomic variables such as money supply growth, exchange rate
depreciation, real GDP and their relationship. It is observed that Bangladesh
experienced a moderate level of inflation during FY 1992 to 2006. On the
other hand, Bangladesh faced two digit inflation rate most of the time during
FY 1982 to 1991. In regard to cause and sources of inflation, the data reveal
that money supply growth, real GDP growth and exchange rate depreciation
appear significant. The empirical findings strongly support the historical
data that inflation in Bangladesh during FY 1981 FY 2009 was a monetary
phenomenon. The outlook of the model-1 is that the annual average inflation

Thoughts on Economics

97

will cruise within 8.47 percent in FY2011-12 and is forecast to fall gradually
to 6.72 percent by FY 2015-16 which is higher than our expectation, but
forecast values of inflation for model-2 and model-3 are stable. In particular,
VDCs support the view that money supply has an explanatory power of
forecasting the movements in consumer price index (CPI). Also, IRFs show
a positive influence of money supply on inflation rate, which is very much in
line with the outcome of VDCs. So authorities should be cautious while
increasing money supply in the market if they want order to control the
inflation any pressure in the country. From the analysis, we found that if the
government wants to control the inflation rate below 6% in the next five
year, supply of money should not exceed 13% each year. The results from
IRFs and VDCs suggest that the real GDP has a moderate short run negative
impact on inflation rate. So care should be taken so that real GDP does not
fall to keep the inflation as it is. It has also been found that exchange rate
depreciation has a small positive impact on inflation; monetary policy
authority should therefore remain vigilant to prevent erosion of the
exchange rate. In the absence of any direct controlling instrument,
Bangladesh Bank can initiate some case specific counter-action. For
example, it can undertake some responsibilities such as monitoring
modalities of Letter of Credit (L/C) operation so that market forces
determine the exchange rate in a process that remains free from much
speculative transactions.
Empirical results of this study suggest that direct linkages between monetary
policy instrument and inflation appear strong, stable and predictable in
Bangladesh. This study presents clear evidence that the contribution of
money supply is more significant than other variables such as real GDP and
nominal exchange rate.
REFERENCES
Akhtaruzzaman, Md. Inflation in the Open Economy: An Application of the Error
Correction Approach to the Recent Experience in Bangladesh, Working Paper
Series, WP 0602 (2005), Policy Analysis Unit (PAU), Research Department,
Bangladesh Bank.
Bangladesh Bank, Economic Trends Monthly Bulletine, Various Issues 1981-2011

98

Forecasting Inflation of Bangladesh ..

Bangladesh Bureau of Statistics. National Accounts Statistics of Bangladesh


(Revised Estimates, 1989-90 to 1998-99), Strengthening National Accounts and
Poverty Monitoring Project (SNAPMP), National Accounting Wing (NAW), BBS,
Ministry of Planning: Statistics Division (2000).
Begum, Nazma (1991). A Model of inflation for Bangladesh, Philippines Review
of Economics and Business, Vol. 28,No. 1, pp. 100-117
Bera, A. K. and C. M. Jarque.An Efficient Large-Sample Test for Normality of
Observations and Regression Residuals, Australian National University Working
Papers in Econometrics, Vol. 40 (1981), Canberra.
Callen, T. and D. Chang (September 1999), Modeling and Forecasting Inflation in
India, IMF Working Paper, WP/99/119.
Chowdhury, Abdur R., Minh Q. Dao and Abu N. M. Wahid (1995). Monetary
Policy, Output and Inflation in Bangladesh: A Dynalysis, Applied Economic
Letters, Vol. 2, No. 3, pp. 51-55.
Christoffersen, P., T. Slok and R. Wescott. Is Inflation Targeting Feasible in
Poland?, Economics of Transition Vol. 9, No. 1, 2001
Dickey, D. A. and W. A. Fuller. Distribution of the Estimators for Autoregressive
Time Series with a Unit Root, Journal of the American Statistical Association,
Vol. 74 (1979), pp. 427-431.
Engle, R. F. and C. W. J. Granger.Co-integration and Error Correction:
Representation, Estimation and Testing, Econometrica, Vol. 55 (1987), pp. 1-87.
Faria, J. R. and F. G. Carneiro. Does High Inflation Affect Growth in the Long and
Short-run?, Journal of Applied Economics, Vol. IV, No. 1 (2001), pp. 89-105.
Hoffimaister, A. W. Inflation Targeting in Koera: An Emprical Exploration, IMF
Working Paper WP/99/7, Washington D.C., August 1999.
Hossain, M. Akhtar (1995) Infaltion, Econimic Growth and the Balance of
Payments in Bangladesh: A Macroeconomic Study, Dhaka: The University Press
limited.
Hossain (2002) Exchange Rate Responses to inflation in Bangladesh, IMF
Working Paper WP/02/166: International Monetary Fund.
Johansen, S. 1991, Estimation and Hypothesis Testing of Cointegrating Vectors in
Gaussian Vector Autoregressive Models, Econometrica 59, pp. 1551-1580.
Johansen, S. and K. Juselius, 1990, Maximum Likelihood Estimation and
Inference on Cointegration- With Application to the Demand for Money, Oxford
Bulletin of Economics and Statistics 52, pp. 169-210.
Md. Nehal Ahmed and Mahmood Osman Imam, 2007, Macroeconomic Factors
and Bangladesh Stock Market: Impact Analysis through Co integration Approach

Thoughts on Economics

99

International Review of Business Research PapersVol. 3 No.5 November 2007


Pp.21-35.
Mubarik, Y. A. Inflation and Growth: An Estimate of the Threshold Level of
Inflation in Pakistan, State Bank of Pakistan Research Bulletin, Vol.1, No. 1-2
(2005), pp. 35-44.
Naka A., Mukherjee, T. and Tufte D., 1998, Macroeconomic Variables and the Sun
Tao (May 2004), Forecasting Thailands Core Inflation, IMF Working Paper,
WP/04/90.
Sweidan, O. D. Does Inflation Harm Economic Growth in Jordan? An
Econometric Analysis for the Period 1970-2000, International Journal of Applied
Econometrics and Quantitative Studies, Vol.1-2 (2004), pp. 41-66.

You might also like