Industries
I. INTRODUCTION
three different years (1981, 1987 and 1991) by using four-digit firm level
data. Assuming the Cobb-Douglas production technology as the appropriate
Downloaded By: [Hiroshima University] At: 03:36 20 June 2008
description of the data set and applying the corrected least squares (COLS)
regression, Salim finds both productive capacity realisation and total factor
productivity to have improved over time, and that openness was a
significant determinant of capacity realisation for the food processing
industry and a sub sector of the textile industry – jute. However, an
improvement in capacity realisation and/or total factor productivity per se
does not imply an improvement in technical efficiency. The present study
thus represents the first attempt of its kind in the context of the Bangladesh
manufacturing sector.
The data used in this study are compiled from two main official sources of
the Government of Bangladesh, namely, the Bangladesh Statistical
Yearbook (various issues) and the Report on Census of Manufacturing
Industries of Bangladesh known as CMI (various issues). The CMI data are
based on the yearly census conducted across private and public enterprises
employing 10 or more people. Both the sources routinely publish data
according to the International Standard Industrial Classification (ISIC).
Table A1 in the Appendix presents the descriptions of the 25 industries
chosen for this study. The CMI data are available for the period 1974–96
except 1995 as no survey was undertaken for the financial year 1994–5. On
the other hand, a consistent database for all the 25 industries considered in
this study are available only from 1978. The study, therefore, chooses the
sample period 1978–94.
The variables used for the empirical analysis are defined as follows.
Output is represented by the gross value added rather than gross output.
One important reason for the preference of value added over gross output is
that it allows comparisons between firms, which may be heterogeneous in
raw material use [Griliches and Ringstad, 1971]. The use of gross output
necessitates the inclusion of raw material as an input variable in the model,
which might obscure the role of physical and human capital in productivity
growth. Secondly, unlike gross output, the value added accounts for the
differences and changes in the quality of inputs [Salim and Kalirajan,
1999]. While net value added might be more appropriate than gross value
added, the CMI estimates of the former are likely to be flawed because of
the arbitrary nature of deductions from gross output [Salim, 1999].
Capital is defined as the gross fixed assets representing the aggregate
book values of land, buildings, machinery, tools, transport, and office
equipment. In empirical analysis, capital is often represented by the
LIBERALISATION AND EFFICIENCY I N BANGLADESH 93
replacement cost of existing machinery and equipment. In the present case,
we follow the argument of Salim and Kalirajan [1999] justifying the use of
Downloaded By: [Hiroshima University] At: 03:36 20 June 2008
fixed capital assets for Bangladesh on the ground that in a country like
Bangladesh, ‘…capital stock is more often used at approximately constant
levels of efficiency for a period far beyond the accounting life measured by
normal depreciation until it is eventually discarded or sold for scrap’ [Salim
and Kalirajan, 1999: 363]. Labour is defined as the number of employees.
Empirical studies have alternatively used the number of employees and the
number of man-hours for labour inputs. It is, however, highly debatable as
to which measure performs better in empirical research.3 The sources of the
data this study utilises measure labour inputs in terms of production and
non-production workers. Thus labour is represented by the sum total of
production and non-production workers.
Capital deepening is defined as the ratio of capital to labour and is used
as a proxy for import liberalisation. From the theoretical point view, this can
be considered reasonable as reductions in tariff rates and quantitative
restrictions may lead to an increase in imported capital. It may be pointed
out here that reductions in tariff rates and quantitative restrictions have
constituted an important element in the consolidation and restructuring of
the Bangladesh import regime.
Export orientation is defined as the ratio of annual export to output of
each industry. Since the CMI does not record the share of export of the
individual firms surveyed, this study uses the ratio of the overall exports of
a three-digit industry to the respective level of output. The export figures are
constructed from the relevant four-digit level entries within each three-digit
level industry. The latter can be justified as a proxy for the former since the
CMI covers more than 60 per cent of total manufacturing establishments.
Other variables considered in the study are the proportion of non-production
workers in total employment, and intermediate inputs, the latter being
defined as gross output less gross value added.
The output variable is deflated by the wholesale price index of industrial
products, capital is deflated by the wholesale price index of manufacturing
that excludes fuel and lighting, and intermediate inputs is deflated by the
wholesale price index of raw materials. All the variables are expressed in
1990 prices.
Before we conclude this section, a word on the limitations of the CMI
data. The reliability of the CMI data in general is questionable as they suffer
from various problems such as the under coverage, under-reporting or
misreporting, and measurement errors. While the problems of random errors
in data observation, especially of the dependent variable can be taken care
of by applying appropriate econometric technique such as the stochastic
frontier production function approach [Caves and Barton 1990], the
94 THE J OURNAL OF D E V E L O P ME N T ST U D I E S
Yit ¼ b0 þ bK Kit þ bL Lit þ bT Tit þ bKK ðKit Þ2 þ bLL ðLit Þ2 þ bTT ðTit Þ2
(1)
þbKL ðKit :Lit Þ þ bKT ðKit :Tit Þ þ bLT ðLit :Tit Þ þ vit uit ;
where:
Yit = the natural logarithm of value added for the i-th industry in the t-th year
of observation;
Kit = the natural logarithm of capital for the i-th industry in the t-th year of
observation;
Lit = the natural logarithm of labour for the i-th industry in the t-th year of
observation;
T = time;
vit = random variables assumed to be iid with mean zero and variance s2v .
uit = non-negative random variables assumed to be independently
distributed with mean, µit, and variance, σ2, where
where:
INPit = the natural logarithm of intermediate inputs for the i-th industry in
Downloaded By: [Hiroshima University] At: 03:36 20 June 2008
V. EMPIRICAL RESULTS
inefficiency over time. But only the dummies for the years 1983 and 1989
through 1994 have coefficients that are significant at 10 per cent level or
Downloaded By: [Hiroshima University] At: 03:36 20 June 2008
less. It follows from above that not all the individual parameters estimates
in the inefficiency model are statistically significant. But a decision to drop
a particular explanatory variable from the model must be based on tests of
hypotheses involving sets of parameters. Table 1 below presents the results
of hypotheses tests concerning some of the parameters as well as the
functional form of the production technology, the distributional form of the
inefficiency effects, and the technical change on the basis of the generalised
likelihood-ratio statistics.
To begin with, the null hypothesis that the Cobb-Douglas production
frontier is an adequate representation of the data is rejected at the 5 per cent
level of significance given the assumption of the Translog stochastic
production frontier, which implies that input and substitution elasticities
TAB L E 1
G E N E R A L I S E D L I KE L I HOOD RAT I O T E S T S OF N U LL H Y PO TH ESES FO R
PARAMETERS IN THE STOCHASTIC FRONTIER PRODUCTION FUNCTION FOR
THE BANGLADESH MANUFACTURING
Notes: Critical values for the tests of hypotheses, excepting γ = 0, are obtained from the
appropriate chi-square distribution. The critical value for testing the null hypothesis of γ
= 0 is taken from Kodde and Palm, 1986.
* indicates that the value of the generalised likelihood-ratio statistic exceeds the critical
value at 5 per cent level of significance.
LIBERALISATION AND EFFICIENCY I N BANGLADESH 99
vary across industries. Similarly, the null hypothesis of no technical
efficiency effects is rejected, although the value of γ is relatively low (0.25)
Downloaded By: [Hiroshima University] At: 03:36 20 June 2008
where the subscript j denotes the j-th industry. The variables and the random
terms have the same explanations as in equations (1) and (2). The choice of
100 THE J OURNAL OF D E V E L O P ME N T ST U D I E S
which has a χ2 distribution with 264 (=11 x 24) degrees of freedom. The
likelihood ratio (LR) statistic is calculated to be 106.18 which is far less
than the χ 2critical (= 302.90) at the 5 per cent level of significance. Thus, the
null hypothesis of the same frontier models for all the individual industries
cannot be rejected. The results, therefore, justify the specification of the
translog frontier model as a common production technology for the pooled
data.
Note: TE1 and TE2 denote respectively simple and weighted average estimates.
time in terms of technical efficiency. This implies that the benefits of trade
liberalisation were not confined only to the export-oriented industries rather
Downloaded By: [Hiroshima University] At: 03:36 20 June 2008
they spread across the board. At the same time, it must also be noted that in
general the import-substituting industries operate at lower levels of
efficiencies.
An Alternative Assessment
The main focus of this study is to examine if trade liberalisation enhanced
the technical efficiencies of the Bangladesh manufacturing industries. As
mentioned earlier, one way to assess the effects of trade liberalisation on
technical efficiency is to compare the estimates of technical efficiency on a
‘before and after’ basis for the manufacturing sector as a whole and/or the
individual industries. Now that the Bangladesh trade policy regime has gone
through three clearly distinct phases, namely, the pre-liberalisation period,
the transition period, and the post-transition period, it is worthwhile to
compare the changes in technical efficiency across these time periods. As
presented in columns [2]–[4] of Table 2 below, the (simple) average overall
technical efficiency is 0.414 for the period 1978–82, 0.572 for the period
1983–91 and about 0.700 for the period 1992–4. The corresponding
weighted average estimates are 0.478, 0.635 and 0.744 respectively. The
median estimates for the three phases are respectively 0.420, 0.550 and
0.686. Thus, there is a clear indication of an improvement in technical
efficiency of the Bangladesh manufacturing sector through phases of the
external trade policy regime.
In order to examine further the validity of the claim above, we provide
an alternative assessment of the same by constructing three separate sub-
panels for the three phases of the Bangladesh international trade regime,
namely 1978–82 (pre-liberalisation period), 1983–91 (transition period) and
TAB L E 2
AVERAGE TECHNICAL E FFICIENCY OF THE BANGLADESH MANUFACTURING
S E C TO R ACROS S P HAS E S OF T HE T RADE PO LIC Y REG IME
in Section IV for each sub-panel. The parameters estimates and the overall
mean efficiencies for the sub-panels are presented in Table A4 in the
Appendix. Looking at the inefficiency model in each sub-panel, it appears
that all the four industry-specific variables either in levels or in squares or
in combination with another variable have contributed to the reduction in
technical inefficiencies. It would be interesting to compare the mean and
individual technical efficiencies based on the sub-panels estimation with
those based on the full panel estimation. The summary statistics are
presented in the last three columns of Table 2. The simple average technical
efficiencies for the three periods are 0.392, 0.531 and 0.554 in ascending
order of the sub-panels. Although, these figures do not exactly match with
the corresponding estimates based on the full panel estimation, they clearly
complement the latter in terms of the direction of changes in technical
efficiencies. Similar observations hold for the weighted average and the
median technical efficiency estimates.
Figure 2 presents the yearly estimates of the simple average, weighted
average and the median estimates of the overall technical efficiencies
obtained from the sub-panels estimation. The simple average (TE1) and the
median estimates show a clear upward tendency throughout the sample
period resembling the pattern of the full panel estimation (as in Figure 1).
F I GURE 2
TECHNICAL E FFICIENCY OF BANGLADESH MANUFACTURING, 1978–94
(BASED ON THE SUB - PA NE L S E S T IMATIO N )
104 THE J OURNAL OF D E V E L O P ME N T ST U D I E S
rise thereafter while the full panel estimates show an upward tendency over
the entire sample period.
With few exceptions, as presented in Table A5 in the Appendix, the
technical efficiencies of the individual industries based on the sub-panels
compare quite well in terms of the direction of changes with those based on
the full panel estimation. Of the exceptional cases, the most contrasting
results are obtained for the non-metallic mineral products with the full panel
estimation showing more or less a downward tendency while the sub-panels
estimation showing a clear upward movement. Other exceptions include: (a)
beverages (falling throughout and quite sharply after 1984 as opposed to
rising (until 1980) and then falling steadily in the full panel estimation); (b)
leather and leather products (more or less constant throughout the sample
period as opposed to an increasing tendency, especially after 1985, in the
full panel estimation); and (c) iron and steel basic industries (high and
steady throughout as opposed to relatively low technical efficiencies in the
first two years with full panel estimation).
is rejected for all the sub-panels on the basis of the likelihood ratio test. The
test statistics are respectively 41.94, 51.56 and 17.48 for the three sub-
panels, which are to be compared with the χ2critical = 12.59. Similarly, the
null hypothesis of no inefficiency effects (γ = 0) is also rejected for each
sub-panel. As presented in Table A4, the LR test statistics of the one sided
error for the three sub-panels are 153.22, 125.90 and 125.24 respectively
while the corresponding values of the χ2critical at the 5 per cent level of
significance are respectively 29.545, 41.977 and 26.983 (from Kodde and
Palm [1986]).
VI. CONCLUSION
This study has undertaken a panel data approach to measure the technical
efficiency of the Bangladesh manufacturing sector as a whole and the
individual technical efficiencies of the majority of the three-digit level
industries. The main objective has been to check if the manufacturing sector
as well as its constituent meso level industries have benefited from
microeconomic reforms in the Bangladesh external trade sector that took
place between 1982 and 1991. The findings of the study can be summarised
as follows. First, alternative measures of the overall technical efficiency
based on the full panel estimation show a rising tendency over time, which
also have support from the overall technical efficiency estimates based on
the three sub-panels representing different phases of the Bangladesh
external trade regime. This is complemented by the technical efficiency
estimates of the individual industries under alternative schemes. Second,
export orientation and capital deepening, both representing trade
liberalisation, appear to be associated with reductions in technical
inefficiencies. The same also applies to the other two industry-specific
variables – intermediate inputs, and proportion of non-production workers
to total employment. Third, with reference to the full panel estimation, most
of the industries (22 out of a total of 25) have experienced rises in technical
efficiencies over time either significantly or marginally. Fourth, majority of
the export-oriented industries have either maintained a high degree of
technical efficiency or gained over time. Fifth, some of the import-
substituting industries also significantly benefited from the trade policy
shift, although import-substituting industries in general have lower
technical efficiencies. Thus, if inter-temporal increases in technical
efficiencies are described as due to trade liberalisation, arguably, the
benefits of trade liberalisation have spread across the board. Sixth, the study
finds no (Hick’s) neutral technical change to have occurred in the
106 THE J OURNAL OF D E V E L O P ME N T ST U D I E S
have occurred in the manufacturing industries. Finally, the study rejects the
Cobb Douglas production technology as an adequate description of the
database used given the assumption of the Translog production technology.
The importance of the trade variables, especially export orientation, in
the reduction of technical inefficiency suggests that improvement in
technical efficiency of the Bangladesh manufacturing sector may be
attributed to the competitive push that trade liberalisation inflicted to the
domestic industries. First, industries with higher export orientation are
exposed to greater international competition than industries with lower
export orientation and/or the import substituting industries. International
competitiveness help reduce ‘X-inefficiency’ of the export industries by
forcing them to utilise a higher proportion of their productive capacities
and/or adopt new technologies [Nishimizu and Robinson, 1984]. In the
context of Bangladesh, Salim [1999] finds openness as an important
determinant of capacity realisation for some of the key manufacturing
industries. These results are very well complemented by the present study.
Second, as mentioned in Section II, the new growth theories emphasize that
trade openness provides the domestic producers access to imported capital
embodying new technologies, which in turn enhance capacity utilisation
and technological progress [Grossman and Helpman, 1991]. The
significance of capital deepening as a determinant of technical efficiency
and the indication of a possible (non-neutral) technical change in the present
case implicate an improvement in capacity utilisation as well as the
occurrence of technological progress in the Bangladesh manufacturing
sector. The importance of the proportion of non-production workers in total
employment, which emphasizes the role of human capital, also points to the
competitive push argument. As pointed out earlier, non-production workers
help reduce inefficiency by greater acquisition of new technologies and
combining the productive resources more effectively. Several empirical
studies indicate that various measures of the real effective exchange rate of
the Bangladesh currency have undergone depreciation over time. As a
result, the anti-export bias has also shown a downward tendency through
time [Rahman, 1995; Hossain and Karunaratne, 2002]. This forms another
dimension of international competitiveness facing the Bangladesh export
industries. As such, it is reasonable to conclude that competitive push has
played an important role in enhancing the technical efficiency of the
Bangladesh manufacturing sector.
LIBERALISATION AND EFFICIENCY I N BANGLADESH 107
NOTES
2. The presence of heterogeneity may render the pooling of the aggregative data and, therefore,
the assumption of a single production frontier inappropriate for purposes of estimating
technical efficiencies. In the present case, as presented in Section V, the results based on a
generalised likelihood ratio test suggests that a common production technology can indeed
be applied to the pooled data used in this study.
3. Denison [1961] finds better results by including man-hours worked in the production
function while Apsden [1990] argues that hours worked may be subject to sampling error as
they are affected by holidays, strikes as well as the lack of a standard unit of measurement.
4. This study avoids the description of these alternative techniques since they are well
documented in existing literature ( see, for examples, Bauer [1990] and Kalirajan and Shand
[1999].
5. For a detailed list of the relative weaknesses of the two models, see Coelli, Rao and Battese
[1998].
6. The table also presents the estimated results based on the Cobb-Douglas production function.
7. The authors gratefully acknowledge the suggestion made by an anonymous reviewer of this
journal on this procedure, and to Professor Tim Coelli for the clarification on the hypothesis
testing.
8. As Coelli, Rao and Battese [1998] point out, the simple average or arithmetic mean may not
be the best estimator if the firms in the sample have significant size differences and/or if the
sample is not constructed by simple random sampling. This study uses the amount of
intermediate inputs used as weights.
REFERENCES
Aghion, P. and P. Howitt, 1998, Endogenous Growth Theory, Cambridge, MA: MIT Press.
Aghion, P., and P.A. Howitt, 1992, ‘A Model of Growth with Creative Destruction’,
Econometrica, Vol.60, pp.323–51.
Aigner, D.J., Lovell, C.A.K. and P. Schmidt, 1977, ‘Formulation and Estimation of the Stochastic
Frontier Production Function Models’, Journal of Econometrics, Vol.6, pp.21–37.
Apsden, C., 1990, ‘Estimates of Multifactor Productivity, Australia’, Occasional Papers,
Canberra: Australian Bureau of Statistics.
Bangladesh Bureau of Statistics, Report on Census of Manufacturing Industries of Bangladesh
(various issues), Statistics Division, Ministry of lanning, Government of Bangladesh.
Bangladesh Bureau of Statistics, Statistical Yearbook of Bangladesh (various issues), Statistics
Division, Ministry of Planning, Government of Bangladesh.
Bangladesh, Government of, Detailed Estimates of Revenue and Receipts (various issues),
Ministry of Finance.
Battese, G.E. and T.J. Coelli, 1993, ‘A Stochastic Frontier Production Function Incorporating a
Model for Technical Inefficiency Effects’, Working Papers in Econometrics and Applied
Statistics, No. 69, Department of Econometrics, Armidale: University of New England.
Battese, G.E. and T.J. Coelli, 1995, ‘A Model for Technical Inefficiency Effects in a Stochastic
Frontier Production Function for Panel Data’, Empirical Economics, Vol.20, pp.325–32.
Bauer, P.W., 1990, ‘Recent Developments in the Economic Estimation of Frontier’, Journal of
Econometrics, Vol.46, pp.39–56.
Bhagwati, J.N., 1988, ‘Export-Promoting Trade Strategy: Issues and Evidence’, World Bank
Research Observer, Vol.1, pp.27–58.
Brada, J.C., King, A.E. and C.Y. Ma, 1997, ‘Industrial Economics of the Transition: Determinants
of Enterprise Efficiency in Czechoslovakia and Hungary’, Oxford Economic Papers, Vol.49,
No. 1, pp.4–27.
Cambell, C., 1984, ‘Correlates of Residual Growth in New Zealand Manufacturing Industries,
1952–73’, Paper presented to Economics Section ANZAAS, Canberra.
108 THE J OURNAL OF D E V E L O P ME N T ST U D I E S
Caves, R.E. and D.R. Barton, 1990, Efficiency in U.S. Manufacturing Industries, Cambridge,
MA: MIT Press.
Charnes, A., Cooper, W.W., Lewin, A.Y. and L.M. Seiford, 1995, Data Envelopment Analysis:
Downloaded By: [Hiroshima University] At: 03:36 20 June 2008
APPENDIX
TABL E A1
D E S C R I P T I O N O F T HE T HRE E - DI GI T L E VE L I NDUS TRIES A C C O R D IN G TO ISTC
Downloaded By: [Hiroshima University] At: 03:36 20 June 2008
Frontier function
constant β0 5.27 0.99 5.31*
Kt: capital βK 1.39 0.29 4.73*
Lt: labour βL -0.85 0.32 2.68*
Tt: time βT 0.0087 0.0056 1.56
Kt2 βKK 0.031 0.026 1.19
Lt2 βLL 0.2047 0.0402 5.09*
Tt2 βTT 0.0039 0.0121 0.33
(Kt) x (Lt) βKL -0.130 0.046 2.81*
(Kt) x (Tt) βKT 0.013 0.015 0.86
(Lt) x (Tt) βLT 0.0049 0.0141 0.35
Inefficiency model
constant δ0 3.08 0.99 3.09*
INPt δ1 -0.68 0.26 2.57*
KDt δ2 0.091 0.364 0.25
XORt δ3 -0.57 0.31 1.81**
NPLt δ4 0.15 0.45 0.32
INPt2 δ5 -0.044 0.023 1.95**
KDtv δ6 -0.094 0.055 1.71**
XORt2 δ7 -0.050 0.043 1.17
NPLt2 δ8 -0.220 0.076 2.88*
(INPt) x (KDt) δ9 -0.150 0.056 2.65*
(INPt) x (XORt) δ10 0.035 0.033 1.07
(INPt) x (NPLt) δ11 0.010 0.064 0.14
(KDt) x (XORt) δ12 -0.101 0.064 1.57
(KDt) x (NPLt) δ13 -0.31 0.11 2.88*
(XORt) x (NPLt) δ14 0.274 0.10 2.66*
X30
Dj’s(time dummies) dt § -1.11 not available not available
t¼15
Variance parameters
δs2 = δ2 + δv2 0.25 0.03 8.74*
γ = δv / (δ2 + δv2) 0.20 0.10 1.96**
Log-likelihood -295.52
LR-test of the one-sided error*** 244.49*
Mean TE 0.548
Notes: S.E. = standard error. Standard errors are rounded to two-significant digits after the
decimal point; *significant at 5 per cent level or less; **significant at 10 per cent level or less;
*** tests the null hypothesis of no inefficiency effects ( =0); § the time dummies for the years
1983 and 1989 to 1994 are found significant at 10 per cent level or less.
Downloaded By: [Hiroshima University] At: 03:36 20 June 2008
TABLE A 3
TECHNICAL E FFICIENCY E STIMATES OF VARIOUS THREE-DIGIT MANUFACTURING INDUSTRIES OF BANGLADESH, 1978–1994
112
(BASED ON THE FULL PANEL E STIMAT ION)
Industry
Code/year 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994
311+312 0.87 0.89 0.89 0.91 0.93 0.92 0.95 0.93 0.92 0.94 0.95 0.97 0.98 0.97 0.97 0.97 0.97
313 0.58 0.72 0.82 0.57 0.42 0.47 0.45 0.29 0.26 0.31 0.26 0.37 0.26 0.22 0.35 0.39 0.36
314 0.65 0.79 0.85 0.71 0.83 0.80 0.85 0.82 0.85 0.93 0.87 0.90 0.92 0.93 0.93 0.96 0.93
321+322 0.92 0.92 0.93 0.93 0.93 0.93 0.95 0.95 0.94 0.94 0.95 0.96 0.96 0.96 0.97 0.96 0.97
323 0.10 0.17 0.11 0.07 0.15 0.25 0.36 0.30 0.29 0.36 0.67 0.94 0.94 0.95 0.95 0.96 0.95
324 0.49 0.68 0.54 0.57 0.54 0.50 0.76 0.64 0.67 0.75 0.78 0.92 0.91 0.89 0.90 0.92 0.90
325 0.21 0.38 0.20 0.15 0.24 0.53 0.37 0.38 0.41 0.40 0.37 0.64 0.90 0.49 0.51 0.59 0.30
326 0.41 0.38 0.43 0.54 0.39 0.34 0.52 0.63 0.56 0.58 0.58 0.62 0.65 0.57 0.55 0.64 0.56
331 0.13 0.11 0.14 0.17 0.14 0.16 0.44 0.27 0.28 0.31 0.33 0.32 0.55 0.46 0.47 0.49 0.45
THE J OURNAL OF D E V E L O P ME N T ST U D I E S
332 0.08 0.13 0.12 0.17 0.19 0.52 0.25 0.31 0.30 0.29 0.32 0.46 0.23 0.27 0.27 0.45 0.37
341 0.28 0.35 0.37 0.47 0.49 0.47 0.71 0.64 0.67 0.56 0.76 0.90 0.90 0.91 0.92 0.89 0.92
342 0.18 0.20 0.22 0.28 0.28 0.35 0.33 0.34 0.34 0.36 0.38 0.61 0.65 0.53 0.71 0.86 0.72
351 0.75 0.77 0.87 0.93 0.92 0.87 0.94 0.93 0.92 0.83 0.86 0.94 0.92 0.93 0.93 0.95 0.93
352 0.29 0.42 0.42 0.49 0.52 0.53 0.78 0.65 0.83 0.85 0.77 0.90 0.92 0.92 0.91 0.90 0.91
353 0.56 0.64 0.66 0.75 0.79 0.64 0.83 0.82 0.77 0.76 0.78 0.86 0.85 0.82 0.86 0.90 0.86
356 0.16 0.17 0.20 0.22 0.25 0.14 0.17 0.16 0.16 0.18 0.20 0.34 0.46 0.30 0.54 0.56 0.43
357 0.06 0.08 0.08 0.10 0.11 0.11 0.13 0.19 0.26 0.29 0.32 0.40 0.62 0.41 0.55 0.64 0.49
361 0.08 0.12 0.12 0.13 0.14 0.13 0.18 0.18 0.17 0.21 0.24 0.36 0.35 0.35 0.35 0.35 0.42
362 0.12 0.13 0.16 0.16 0.21 0.24 0.27 0.25 0.20 0.18 0.18 0.18 0.23 0.23 0.27 0.30 0.27
369 0.23 0.26 0.32 0.47 0.42 0.40 0.46 0.43 0.41 0.32 0.34 0.51 0.51 0.43 0.52 0.55 0.52
371+372 0.41 0.68 0.70 0.90 0.91 0.81 0.93 0.83 0.80 0.83 0.83 0.95 0.94 0.93 0.95 0.95 0.95
381+382 0.23 0.26 0.31 0.36 0.31 0.31 0.38 0.36 0.34 0.37 0.40 0.71 0.89 0.93 0.91 0.86 0.88
383 0.12 0.17 0.19 0.25 0.26 0.26 0.38 0.33 0.31 0.32 0.32 0.30 0.31 0.28 0.30 0.46 0.30
384 0.32 0.40 0.35 0.39 0.39 0.42 0.60 0.61 0.72 0.78 0.81 0.87 0.89 0.85 0.76 0.80 0.75
385 0.31 0.30 0.51 0.63 0.47 0.41 0.77 0.67 0.63 0.75 0.66 0.69 0.89 0.82 0.81 0.91 0.83
TE1 0.342 0.405 0.420 0.453 0.449 0.460 0.550 0.517 0.522 0.536 0.557 0.665 0.689 0.654 0.686 0.731 0.682
S.D. 0.25 0.27 0.28 0.28 0.28 0.25 0.28 0.26 0.27 0.27 0.26 0.28 0.27 0.28 0.25 0.24 0.25
TE2 0.379 0.490 0.489 0.519 0.510 0.506 0.618 0.560 0.578 0.581 0.588 0.760 0.770 0.757 0.747 0.739 0.747
Median 0.282 0.351 0.356 0.468 0.392 0.422 0.459 0.435 0.413 0.405 0.580 0.711 0.846 0.824 0.766 0.861 0.754
Note: TE1 and TE2 represent respectively the simple and the weighted average of the efficiency estimates for the manufacturing sector
as a whole and S.D stands for the standard deviation from the mean value (TE1).
LIBERALISATION AND EFFICIENCY I N BANGLADESH 113
TABL E A4
M A X I M U M L I K E L I HOOD E S T I MAT E S ( ML E ) F OR PARA METER S O F TR A N SLO G
(T L ) S TO C H A S T I C F RONT I E R P RODUCT I ON F UNCT IO N S FO R TH E SU B -PA N ELS
Downloaded By: [Hiroshima University] At: 03:36 20 June 2008
Variable
description Sub-Panel t Sub-Panel t Sub-Panel t
in natural logs Parameter 1978–82 statistic 1983–91 statistic 1992–94 statistic
Frontier function
constant β0 2.01 1.04 6.85 4.74* 7.28 5.47*
Kt: capital βK 3.72 6.51* 1.57 3.29* 1.51 4.22*
Lt: labour βL -1.89 3.82* -1.39 3.06* -1.58 3.82*
Tt: time βT 0.01 0.35 0.02 0.48 0.02 1.61
Kt2 βKK 0.10 1.09 -0.04 1.34 0.07 3.25*
Lt2 βLL 0.39 3.54* 0.11 2.16* 0.23 6.04*
Tt2 βTT -0.01 0.23 0.00 0.31 -0.00 0.49
(Kt) x (Lt) βKL -0.66 2.68* -0.06 1.79* -0.29 5.70*
(Kt) x (Tt) βKT 0.04 1.04 -0.01 0.27 0.01 1.65
(Lt) x (Tt) βLT 0.03 0.65 -0.01 0.32 -0.01 0.74
Inefficiency model
constant δ0 -0.36 0.16 2.37 1.99** 6.16 5.96*
INPt δ1 1.85 2.99* 0.14 0.40 -0.90 3.82*
KDt δ2 0.23 3.22* -0.14 0.25 -0.83 1.08
XORt δ3 -0.06 1.38 -0.09 1.62 -1.07 3.44*
NPLt δ4 1.35 1.39 0.25 0.41 0.22 0.45
INPt2 δ5 -0.22 5.08* -0.03 1.70** 0.03 1.88**
KDt2 δ6 0.09 0.75 -0.22 2.59* 0.01 0.29
XORt2 δ7 -0.13 1.12 -0.18 1.90** -0.06 1.80*
NPLt2 δ8 -0.33 3.53* -0.26 1.94** -0.19 3.63*
(INPt) x (KDt) δ9 -0.24 3.74* 0.02 0.28 -0.07 1.27
(INPt) x (XORt) δ10 0.02 0.29 -0.16 1.84* -0.01 4.13*
(INPt) x (NPLt) δ11 -0.28 2.44* -0.04 0.41 0.01 0.07
(KDt) x (XORt) δ12 0.27 1.81* 0.07 0.78 0.09 1.47
(KDt) x (NPLt) δ13 -0.59 2.83* -0.47 2.08* -0.26 3.04*
(XORt) x (NPLt) δ14 0.39 1.59 0.20 1.56 0.31 2.98
Dj’s X
T1
(time dummies) dt 0.21 not -0.48 not -1.07 not
t¼15 available available available
Variance parameters
σs2 = σ2 + σv2 0.19 6.20* 0.23 7.41* 0.22 11.21*
γ = σ2 / (σ2 + σv2) 0.54 2.82* 0.14 2.02* 0.11 4.37*
Log-likelihood -64.09 -144.68 -55.39
LR-test of the one-sided
error*** 153.22* 125.90* 125.24*
Mean TE 0.392 0.531 0.554
Note: *significant at 5 per cent level or less; **significant at 10 per cent level or less; ***tests
the null hypothesis of no inefficiency effects (γ =0).
Downloaded By: [Hiroshima University] At: 03:36 20 June 2008
TABLE A 5
TECHNICAL E FFICIENCY E STIMATES OF VARIOUS THREE-DIGIT MANUFACTURING INDUSTRIES OF BANGLADESH, 1978–1994
114
( BAS E D ON TH E SU B-PA N ELS ESTIMATIO N )
Industry
Code/year 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994
311+312 .95 .95 .92 .94 .95 .95 .97 .96 .94 .95 .95 .98 .98 .98 .97 .98 .96
313 .83 .78 .84 .61 .41 .57 .41 .21 .20 .24 .17 .19 .15 .11 .13 .16 .16
314 .76 .84 .78 .77 .82 .88 .88 .86 .86 .92 .75 .91 .95 .93 .93 .94 .91
321+322 .95 .96 .94 .95 .96 .96 .96 .97 .96 .96 .96 .97 .97 .97 .97 .96 .96
323 .10 .14 .12 .07 .15 .36 .60 .42 .42 .51 .88 .97 .97 .97 .96 .94 .91
324 .69 .74 .62 .49 .43 .50 .60 .49 .48 .54 .50 .78 .70 .57 .61 .56 .54
325 .20 .25 .28 .26 .37 .54 .28 .28 .30 .29 .22 .43 .37 .30 .33 .39 .34
326 .29 .22 .27 .26 .22 .38 .43 .54 .49 .48 .40 .28 .49 .33 .32 .37 .36
331 .10 .13 .14 .10 .10 .20 .40 .18 .25 .25 .23 .66 .61 .46 .49 .46 .52
THE J OURNAL OF D E V E L O P ME N T ST U D I E S
332 .11 .09 .07 .06 .17 .55 .42 .63 .49 .40 .45 .12 .19 .29 .31 .33 .39
341 .46 .53 .65 .55 .37 .42 .41 .42 .47 .38 .46 .66 .72 .70 .73 .75 .69
342 .07 .10 .10 .12 .14 .38 .30 .34 .42 .43 .40 .89 .63 .47 .49 .54 .47
351 .57 .66 .75 .75 .81 .83 .94 .93 .90 .75 .77 .92 .89 .82 .79 .76 .72
352 .52 .58 .51 .50 .40 .53 .61 .56 .77 .82 .56 .84 .90 .89 .88 .84 .78
353 .29 .37 .38 .41 .35 .91 .95 .95 .91 .89 .75 .74 .75 .55 .65 .74 .76
356 .09 .09 .10 .09 .11 .14 .14 .14 .14 .14 .14 .23 .29 .20 .27 .29 .34
357 .06 .05 .04 .06 .05 .10 .12 .16 .22 .24 .22 .24 .32 .21 .31 .28 .25
361 .06 .09 .09 .10 .08 .13 .19 .12 .13 .18 .20 .27 .29 .27 .28 .31 .29
362 .09 .10 .13 .11 .12 .27 .29 .26 .19 .13 .11 .13 .14 .12 .13 .14 .12
369 .29 .32 .32 .31 .28 .34 .36 .37 .35 .20 .19 .13 .11 .11 .14 .12 .16
371+372 .87 .95 .93 .72 .74 .65 .88 .61 .82 .81 .75 .94 .92 .86 .88 .89 .81
381+382 .09 .10 .10 .11 .25 .25 .27 .28 .26 .27 .26 .45 .73 .57 .66 .62 .54
383 .15 .19 .27 .30 .36 .21 .33 .31 .27 .24 .25 .19 .25 .15 .19 .22 .22
384 .41 .49 .57 .61 .60 .61 .75 .73 .85 .88 .85 .79 .84 .71 .73 .78 .70
385 .28 .38 .36 .39 .52 .62 .96 .94 .93 .94 .80 .47 .88 .55 .64 .77 .72
TE1 .368 .404 .411 .386 .390 .491 .538 .506 .521 .514 .489 .576 .610 .532 .552 .566 .545
S.D. .320 .320 .320 .306 .207 .265 .291 .292 .297 .302 .287 .3123 .300 .296 .293 .289 .272
TE2 .519 .546 .550 .631 .610 .549 .591 .540 .581 .609 .512 .703 .755 .697 .699 .651 .657
Median .280 . 250 .280 .300 .360 .500 .420 .420 .470 .430 .450 .660 .700 .550 .610 .560 .540
Note: TE1 and TE2 represent respectively the simple and the weighted average of the efficiency estimates for the manufacturing sector
as a whole and S.D stands for the standard deviation from the mean value (TE1).