This article was downloaded by: [University of Guelph]

On: 30 May 2012, At: 08:53

Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Applied Economics Letters

Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/rael20

Determinants of the long-run growth rate of

Bangladesh
a a
B. Bhaskara Rao & Gazi Hassan
a
School of Economics and Finance, University of Western Sydney, Sydney, Australia

Available online: 13 Jan 2011

To cite this article: B. Bhaskara Rao & Gazi Hassan (2011): Determinants of the long-run growth rate of Bangladesh, Applied
Economics Letters, 18:7, 655-658

To link to this article: http://dx.doi.org/10.1080/13504851003800760

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to
anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should
be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims,
proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in
connection with or arising out of the use of this material.
 Applied Economics Letters, 2011, 18, 655–658

Determinants of the long-run

growth rate of Bangladesh
B. Bhaskara Rao and Gazi Hassan*
School of Economics and Finance, University of Western Sydney, Sydney,
Australia

This article estimates Total Factor Productivity (TFP) for Bangladesh and
analyses its key determinants. According to the Solow (1956) growth model,
long-run growth rate equals TFP. Estimated b-coefficients show that trade
Downloaded by [University of Guelph] at 08:53 30 May 2012

openness, foreign direct investment and development of financial sector

increase TFP.

I. Introduction where Y = output, A = stock of knowledge, K = stock

This article analyses some key determinants of the long-
run growth rate of Bangladesh. Our framework uses
theoretical insights from the Solow (1956) growth
model and the growth accounting framework in Solow
(1957). Our empirical methodology is based on the
extensions to these works by Mankiw et al. (1992) and
Senhadji (2000). Empirical results for Bangladesh are
estimated for the period 1970 to 2007.
In Section II a production function for Bangladesh is
estimated. Using the estimated factor elasticities, a
growth-accounting exercise is conducted next to esti-
mate Total Factor Productivity (TFP). In Section III
the Autoregressive Distributed Lag (ARDL) approach
is used to estimate the long-run relationship between
TFP and its key determinants. Policy implications and
conclusions are discussed in Section IV.
II. The Production Function and TFP
The following specification of the human capital aug-
mented Cobb–Douglas production function with con-
stant returns to scale is used. A similar function is also
used by Senhadji (2000).
Yt ¼ At Kta ðHt · Lt Þð1aÞ
of capital, H = an index of human capital formation
through education and L = employment. Unit root
tests for these variables did not yield unequivocal
results. Although Y and (H · L) are found to be I(1)
in levels by the Augmented Dickey–Fuller (ADF) and
Kwiatkowski–Phillips–Schmidt–Shin (KPSS) tests, K
is found to be I(1) by only the Elliot, Rothenberg and
Stock (ERS) point optimal test. As there is some doubt
on the order of integration of these variables, we shall
use the ARDL approach, based on the bounds test, for
estimation. It is well known that the bounds test does
not require pretesting of the variables and is popular in
many applications. The implication is that this techni-
que can be used if the variables are I(0) and/or I(1).
Using this approach we found that Y, L · H and K are
cointegrated. The computed test statistic F(2, 29) =
5.456 exceeds the upper bound critical value of 4.855
at the 5% level and the null of no cointegration is
rejected. The estimated value of a, with the t-ratio in
brackets, is 0.468 (15.30) and is highly significant.1
Using this value, TFP is estimated as follows:
TFP ;
ln Y  0:468
ln K
ð2Þ
 ð1  0:468Þð
ln H þ
ln LÞ
The average TFP before 1990 was -0.6% and this has
ð1Þ increased to near 0 during 2000 to 2007. During the

*Corresponding author. E-mail: gazi.hassan@gmail.com

1
Details of the unit root and bounds test results are not reported to conserve space and may be obtained from us. Definitions of
the variables and data sources can also be obtained from us. Due to the limited space these are not included.

Applied Economics Letters ISSN 1350–4851 print/ISSN 1466–4291 online # 2011 Taylor & Francis 655
http://www.informaworld.com
DOI: 10.1080/13504851003800760
 656
2000s, per capita income grew at an average rate of 2.5%.
This was due to factor accumulation, which is difficult to
sustain over longer periods because of negligible TFP. To
double the per capita income from US$428 in 2007 in
about 15 years, it is necessary to increase the per capita
income growth rate at least to 4.5%. A policy option is to
increase TFP to 2.5% and also to ensure that factor
accumulation contributes the balance of 2%.2
III. Determinants of TFP
Many empirical works on growth claim that their spe-
cifications are based on one or another endogenous
growth model. In our view, the specifications in most
of these works are arbitrary and their results need to be
interpreted cautiously. Commenting on the arbitrary
nature of specifications, Easterly et al. (2004) have
Downloaded by [University of Guelph] at 08:53 30 May 2012
B. B. Rao and G. Hassan
positive or negative effects, (8) a dummy variable to
capture the effects of reforms since the late 1980s
(DUM90, +) and (9) the rate of inflation (PRAT, -).
The expected signs for these variables are shown in
parentheses. Definitions of the variables and sources of
data may be obtained from the authors.
Our aim is to estimate a cointegrating equation, if it
exists, to estimate the long-run relationship between TFP
and its key determinants. When these variables are tested
for unit roots, TFP and PRAT are found to be stationary
whereas all other variables are nonstationary. Instead of
conducting a variety of unit root tests with many alter-
native options, which seems to have become currently
prolific and perhaps a less meaningful activity, we shall
use the ARDL approach for cointegration, which, as
noted earlier, does not need pretesting of the variables.
Furthermore, we shall use the standardized variables to
estimate the b-coefficients for easy comparisons of the

observed that ‘This literature has the usual limitations
of choosing a specification without clear guidance from
theory, which often means there are more plausible
specifications than there are data points in the sample.’
Consequently, Durlauf et al. (2005) have found that
the number of potential TFP improving variables in
various empirical works is as many as 145. Therefore,
it is hard to select a handful of key determinants of TFP
from such a large list of variables. Nevertheless, because
data on country-specific time series are limited, we can
select only a few key determinants of TFP. We selected
nine such potential variables that affect TFP. These
are the ratios to Gross Domestic Product (GDP) of
(1) foreign direct investment (FDIRAT, +), (2) overseas
development aid (ODARAT, +), (3) M2 (M2RAT, +)
as a proxy for development of the financial sector,
(4) remittances by emigrant workers (REMRAT, )
which is a significant proportion to GDP in Bangladesh,
(5) trade openness (TRAT, +) proxied with the ratio of
imports plus exports to GDP and (6) current government
expenditure (GRAT, -). Three other explanatory vari-
ables are (7) a time trend (T, ) to capture the effects of
other trended but ignored variables which may have
relative contribution of these variables to TFP. With only
38 observations it is not possible to determine the optimal
lag order for Vector Autoregression (VAR) with 10 vari-
ables even if we use an initial lag order of 2. Therefore, we
have used only three crucial ratios namely, FDIRAT,
GRAT and TRAT, starting with five lags, to determine
the order of VAR. The Schwarz Bayesian Criterion
(SBIC) indicated that a second-order VAR is optimal.3
We also made a small change to the sequence of four
steps in the ARDL approach to cointegration. First, the
optimal order of the unrestricted VAR is selected. Second,
using this order, a cointegration test is conducted. Third,
if there is cointegration, the long-run cointegrating coeffi-
cients are estimated with the bounds test and finally, with
the lagged Error Correction Term (ECT) and an ARDL
specification, the final dynamic structure for the changes
in the dependent variable is estimated. Given that we have
nine potential variables to explain TFP it is convenient to
follow steps 1 and 3 to start with and conduct the coin-
tegration test with only the significant variables found in
step 3. Using a second-order VAR, the estimated equa-
tion in step 3, after deleting the insignificant variables one
at a time, is as follows4:

TFPz ¼1:828  0:164T þ 0:776DUM90  0:913GRATz  0:755PRATz

ð3:24Þ ð4:68Þ ð12:37Þ ð3:49Þ ð4:61Þ
ð3Þ
þ 0:462TRATz þ 0:217FDIRATz þ 0:402M2RATz
ð3:42Þ ð2:08Þ ð1:78Þ

2
This is a suggested option, which we think is a pragmatic option. However, we think that to increase TFP to a higher level some
difficult and large changes in policy measures are necessary.
3
Akaike’s Information Criteria (AIC) indicated a higher order for VAR but this is not a pragmatic option. Our procedure for
the optimal lag is perhaps better than the arbitrary selection of one or two periods in many works citing data limitations as the
reason.
4
First, we delete a variable with the smallest t-ratio and below the 5% critical value. The equation is reestimated and another
insignificant variable is removed. This procedure is continued until all the retained coefficients are significant.
 Determinants of TFP in Bangladesh
Standardized variables are denoted with the Z
subscript. t-Ratios are below the coefficients in the
brackets. These estimates are impressive and all the
coefficients have the expected signs and are significant
at the conventional levels of 5 or 10%. The positive
and significant coefficient for DUM90 implies that
reforms in Bangladesh have enhanced TFP. The nega-
tive effects of GRAT and PRAT are larger in absolute
value than the positive effects of TRAT, M2RAT and
FDIRAT.
Negative and significant coefficient for trend
indicates that it is necessary to implement policies
to improve the overall efficiency in Bangladesh.
Although some of these negative TFP effects can be
offset by increasing the ratios with positive coeffi-
cients, decreasing the negative effects is more amen-
able to quickly respond to policy changes. For
example, it may be relatively quick to reduce govern-
Downloaded by [University of Guelph] at 08:53 30 May 2012
ment expenditure and inflation rate than to increase
foreign direct investment or speed up the progress of
the financial sector.
We reestimated Equation 3 by allowing for the
negative effects of a major natural calamity during
2004 by adding an additional dummy DUM04 to cap-
ture this effect. These estimates are as follows:
TFPz ¼ 3:019  0:163T þ 0:785DUM90  0:662DUM04  0:822GRATz
ð5:61Þ ð5:43Þ ð4:067Þ
 0:728PRATz þ 0:383TRATz þ 0:228FDIRATz þ 0:490M2RATz
ð5:20Þ ð3:25Þ
Addition of this dummy made only marginal changes
to the estimated coefficients but the coefficient of
M2RAT has increased and become significant at the
5% level. This is not unexpected because various sec-
tors try to improve efficiency to meet emergency needs
and the financial sector might have responded posi-
tively because of the large amounts of aid money it
might have received from the emigrant workers and
international agencies.
We shall use the variables in Equation 4 with the
two dummy variables and a second-order VAR for
testing cointegration with the bounds test. The com-
puted test statistic, after adding the relevant one-
period lagged level variables, is F(6, 13) = 24.827
and exceeds the upper bound value of 4.088 at the
5% level. Therefore, the null of no cointegration can
be rejected and it may be concluded that in
Bangladesh there is a robust long-run relationship
between TFP (which is also its steady-state growth
rate) and economic reforms, progress of the financial
657
sector, trade openness, foreign direct investment, cur-
rent government expenditure and inflation.
IV. Policy Implications and Conclusion
In this article we found some key determinants of TFP
for Bangladesh. The ratios to GDP of trade openness,
foreign direct investment and progress in the financial
sector have significant positive effects. Economic
reforms from 1990 have also had a significant positive
effect. Government current expenditure and inflation
rate have relatively large negative effects on TFP.
Although the average rate of growth of per capita
output during 2000 to 2007 was 2.5%, the average
TFP was negligible. Therefore, it is not possible to
sustain this growth rate in the long run because it is
entirely because of factor accumulation. To increase
this growth rate to 4.5% to double the per capita
income in 15 years, it is necessary to increase TFP to
2.5% and maintain the contribution of factor accu-
mulation to growth by 2%.
How can this be achieved? Alternative scenarios are
possible but we opt for the following. If GRAT and
PRAT can be decreased by 10% from their 2000 to
2007 averages and TRAT, FDIRAT and M2RAT can
ð2:76Þ ð3:67Þ
ð4Þ
ð2:53Þ ð2:45Þ
be increased by 7.5% from their averages, TFP can be
increased to 2.5% to sustain growth of per capita
income at 2.5%. If the three variables, TRAT,
FDIRAT and M2RAT, are increased by another per-
centage point, TFP will be 3%. Factor accumulation
should fill the gap between these rates and the target
4.5% per capita growth because increase of TFP
further may require harsher measures to reduce gov-
ernment expenditure and anti-inflation policies.
These policies to increase TFP can be implemented
in the short to medium terms. Furthermore, their
transitional growth effects, ignored in this short arti-
cle, will also be significant. Hopefully our framework
and results will also interest policy makers of other
developing countries.
References
Durlauf, S., Johnson, P. and Temple, J. (2005) Growth
econometrics, in Handbook of Economic Growth, Vol.
 658
1, Chap. 8 (Eds.) P. Aghion, and S. Durlauf, North
Holland, Amsterdam, pp. 555–677.
Easterly, W. Levine, R. and Roodman, D. (2004) New data,
new doubts: a comment on burnside and Dollar’s ‘Aid,
Policies, and Growth’ (2000), American Economic
Review, 94, 774–80.
Mankiw, N. G., Romer, D. and Weil, D. (1992) A contribu-
tion to the empirics of economic growth, Quarterly
Journal of Economics, 107, 407–37.
Downloaded by [University of Guelph] at 08:53 30 May 2012
B. B. Rao and G. Hassan
Senhadji, A. (2000) Sources of economic growth: an exten-
sive growth accounting exercise, IMF Staff Papers, 47,
129–57.
Solow, R. (1956) A contribution to the theory of eco-
nomic growth, Quarterly Journal of Economics, 70,
65–94.
Solow, R. (1957) Technological change and the aggregate
production function, Review of Economics and
Statistics, 39, 312–20.