International Journal of Business and Economics Research

2016; 5(5): 149-154

http://www.sciencepublishinggroup.com/j/ijber
doi: 10.11648/j.ijber.20160505.13
ISSN: 2328-7543 (Print); ISSN: 2328-756X (Online)

A Test for the Cobb Douglas Production Function in

Manufacturing Sector: The Case of Bangladesh
Shaiara Husain*, Md. Shahidul Islam
Department of Economics, Southeast University, Dhaka, Bangladesh

Email address:
shuma153@yahoo.com (S. Husain)
*
Corresponding author

To cite this article:

Shaiara Husain, Md. Shahidul Islam. A Test for the Cobb Douglas Production Function in Manufacturing Sector: The Case of Bangladesh.
International Journal of Business and Economics Research. Vol. 5, No. 5, 2016, pp. 149-154. doi: 10.11648/j.ijber.20160505.13

Received: August 16, 2016; Accepted: September 1, 2016; Published: September 21, 2016

Abstract: The manufacturing sector has been a major driver of GDP growth in Bangladesh. The sector recorded an
impressive average annual growth during last a couple of decades. This study is an attempt to test empirically the Cobb-
Douglas production function for manufacturing sector of Bangladesh. Data of total value of output, total asset, total liabilities,
number of permanent workers etc. of about six major types of industries including Garments, Textiles, Food & Food
Processing, Leather & Leather Products, Electronics and Chemicals & Pharmaceuticals have been used. Researchers use total
value of output as Y, total liabilities as K, and number of permanent workers (including officials managers, personal) as L.
Study finds that coefficient for K and L are 0.49 and 0.51 respectively for entire manufacturing sector. In case of Garments,
coefficient for K is 0.30 and L is 0.61 implying that labor is more productive than capital. The statement is also true for Textile
sector and Leather & leather products. Capital is, on other hand, more productive than labor in Food & food processing
industries, Electronics, Chemicals & Pharmaceuticals sectors. The coefficients are significant and do not suffer from MC and
also not from Autocorrelation. To correct heteroscedasticity, WLS method (weighted Least Square) has been adopted. Findings
reveal that Cobb-Douglas production function is applicable and exhibits increasing returns to scale in the context of
manufacturing sector of Bangladesh.
Keywords: Manufacturing Sector, Cobb-Douglas Production Function, Productivity, Labor, Capital

The popularity of this function is due to the fact, it is

1. Introduction amenable to easy mathematical manipulations and it satisfies
Given the ambitious growth targets of the vision 2021 of the restrictions imposed on production function. These
Bangladesh, it is considered very important to give serious restrictions are: (1) both inputs are essential for production
attention to the manufacturing sector for employment such that no output can be produced without using at least
expansion, productivity enhancement and increasing per some of both inputs; (2) The marginal product of each input
capita income oriented to reduction of poverty in the country. is positive; (3) the marginal product diminishes with
The sector recorded an impressive average annual growth increased use of each of the inputs; (4) The marginal product
during a last couple of decades. Given the still modest share of an input increases with increased use of the other input.
in GDP, manufacturing sector has considerable scope to lead Because of above logical and consistent characteristics,
the growth process. Cobb-Douglas production function has been considered to be
Production function provides quantitative link between tested in manufacturing sector of Bangladesh. There does not
inputs and outputs. Economists prefer to use the Cobb- appear to be overwhelming evidence that would lead one to
Douglas functional form Y = f (K, L) = AKαLβ, where Y is choose the CES over the Cobb-Douglas for forecasting GDP
the output, K is the capital, and L is the labor, α+β=1 α, β≥0 and income shares. When empirical estimates are restricted to
K, L>0 because it provides a relatively accurate description the Cobb-Douglas form, the fit tends to be quite good [1].
of the economy and is very easy to work with algebraically. Recently, businessmen as well as industrialists are very
much concerned about the theory of firm in order to make
150 Shaiara Husain and Md. Shahidul Islam: A Test for the Cobb Douglas Production Function in
Manufacturing Sector: The Case of Bangladesh

correct decisions regarding what items, how much and how estimates of technical, allocative and economic efficiencies.
to produce them. All these decisions are directly related with The study shows that, on average, sawmills in Ondo and
the cost considerations and market situations where the firm Osun states have high technical, allocative and economic
is to be operated. In this regard, this paper should be helpful efficiencies. In both states, sawmills have the potential to
in suggesting the most suitable functional form of production either increase output using the same level of input or to
process for the major manufacturing industries of developing reduce the present level of input use for the same level of
countries like Bangladesh. This study is an attempt to test output.
empirically the Cobb-Douglas production function for [6] investigates the technical efficiency of selected
manufacturing sector of Bangladesh with the validity of all manufacturing industries of Bangladesh using a stochastic
its assumptions. frontier production function approach. A feasible Cobb-
Douglas stochastic frontier production function, which has
2. Review of Literature time-varying technical inefficiency effects, is estimated. In
the study, researchers have analyzed the stochastic frontier
[2] considered the manufacturing sector of south-west production function using panel data in selected
region of Bangladesh and apply Cobb-Douglas production to manufacturing industries in Bangladesh. Two alternative
estimate the productivity, allocative efficiency and measuring distributions are used to model the random inefficiency term:
returns to scale. The study showed cement, jute, and textile a truncated normal distribution and a half-normal
manufacturing firms having decreasing returns to scale distribution. Researchers have observed that the estimated
whereas fertilizers and seafood processing firms have values of the time-varying inefficiency parameter are positive
increasing returns to scale. The estimated value of marginal for both the truncated and the half normal distribution. These
productivity and allocative efficiency revealed that labor indicate that technical inefficiency has declined over the
productivity of all sorts of manufacturing firms is greater reference period. Tests for different null hypotheses involved
than that of capital productivity. Therefore, additional in the stochastic frontier production function show that the
employment can be generated by utilizing labor intensive technical inefficiency effects for the selected manufacturing
technique. If investors invest in labor intensive technique, industries in Bangladesh are significant. It has been found
they can expect a positive return from investment. that the mean efficiencies according to the truncated and the
Regression analysis and hypothesis testing of the study show half normal distributions are 0.4022 and 0.5557 respectively.
that labor input has a significant impact on the production of Here it should be noted that although the growth in technical
these manufacturing firms. The estimation of Cobb- Douglas efficiency is statistically significant over time as tested by the
production function which showed that labor and capital null hypothesis, the rate of increase in technical efficiency
inputs influence the total variation of production of cement, has been very slow over time in Bangladesh.
fertilizer, jute, sea foods and textile firms respectively [7] noticed increasing concern of businessmen as well as
0.927%, 0.917%, 0.90%, 0.928% and 0.851% respectively. industrialists about the theory of firm in order to make
The study also used marginal productivity of labor and correct decisions regarding what items, how much and how
capital by taking partial derivation of Cobb-Douglas to produce them. Researchers wanted to suggest the most
production function which shows that the marginal suitable functional form of production process for the major
productivity of labor of these firms is greater than that of manufacturing industries in Bangladesh. The study considers
capital. Therefore, these firms could reduce production cost Cobb-Douglas (C-D) production function with additive error
by shifting resources from capital intensive technique to and multiplicative error term. The main purpose of the study
labor intensive technique and at the same time they can is to select the appropriate Cobb-Douglas production model
generate additional employment opportunities. for measuring the production process of some selected
[3] shows that Cobb-Douglas Production function is the manufacturing industries in Bangladesh. Researchers use
only production function which has the property of a constant different model selection criteria to compare the Cobb-
functional distribution of income of the factors of production. Douglas production function with additive error term to
The estimation of the parameters of an aggregate production Cobb-Douglas production function with multiplicative error
function is central in work on growth and productivity. term.
[4] estimates Cobb-Douglas Production function and [8] derived the exact distribution of the indirect least
detects that industries in Bangladesh have potential allocative square estimators of the coefficients of the Cobb-Douglas
efficiency as they are using at least some capital. By production function. Researcher developed a model in which
estimating the marginal productivity of labor and capital, a Cobb- Douglas type function is coupled with simultaneous
study explains that allocative efficiency is achievable through multiplicative and additive errors. The specification was a
appropriate pricing of capital and its proper disbursement natural generalization of the “pure” models in which either
among the proprietorships. additive or multiplicative stochastic terms were introduced.
[5] deals with the issue of improving economic efficiency Researcher used a maximum likelihood approach to the
in sawn wood in Ondo and Osun states, southwest Nigeria. estimation of a Cobb-Douglas type model when the model
The stochastic frontier approach is used to estimate a self included both multiplicative and additive disturbance terms.
dual Cobb-Douglas production function which gives Because of the complexity of the likelihood function,
 International Journal of Business and Economics Research 2016; 5(5): 149-154 151

maximization algorithm had been used. transformation of this model is as follows:

[9] estimated production functions for Indian
manufacturing industries and to find whether plausible and LnY = LnA+ α Ln K+ β LnL+uiL (2)
meaningful estimates can be obtained for returns to scale, = B+ α Ln K+ β LnL+ ui,where B= Ln A (3)
substitution, distribution, and efficiency parameters. Data had
been collected through surveys specially designed for Thus written, the model is linear in the parameters B, α
estimating the levels of technical efficiency (TE) and and β and is therefore a linear regression model. It may be
variations in TE only in small-scale industrial units. mentioned here that though it is nonlinear in the variables Y
Researchers found little variation in TEs across industry and K, L but linear in the logs of these variables. In short,
groups and a high level of average TE in each industry equation (3) is a log log, double log, or log linear model.
groups. The empirical analysis has been carried out to study α, the
[10] found that, the Cobb-Douglas production function (partial) elasticity of output with respect to the capital input,
with additive error performs better for the selected and β, the partial elasticity with respect to labor for the whole
manufacturing industries based on the data under study industries.
period. Thus the strictly nonlinear models (which are In this part, informationwill be examined in regression of
nonlinear with additive error terms) seemed to be better than whole industry about the possibility of violation of CLRM
intrinsically linear model (which are nonlinear with assumption, ie Multicollinearity (MC) Autocorrelation, &
multiplicative error terms). Researchers estimated the Heteroscedasticty.
parameters of the Cobb-Douglas production function with It is expected that the result will not be affected by MC.
additive errors by using optimization subroutine. The Though it is a matter of suspicion that as the study is about
estimates showed economies of scale in the manufacturing of Cobb-Douglas production function, there might exist a linear
drugs & pharmaceuticals, Furniture & fixtures, Iron & steel relationship between two inputs, such as in a garments
basic, Leather footwear, Fabricated metal products, Plastic factory a women worker may operate only one sewing
products, Printing & publications and Tobacco. There are machine. In this regard, MC might be a simple phenomenon.
diseconomies of scale in the Beverage, Chemical, Glass & As the study has a large sample size consist of different type
glass products, Leather & leather products, Paper & paper of industries and some of them are labor intensive (ex:
products, Textile, Wood & crock products industries and garments) and some are capital intensive (Ex: Electronics,
Transport equipment. Chemical & Pharmaceutical), linear relationship in
regression of whole industrial sector is expected to be
prevented and degree of collinearity is expected to be
3. Methodology and Data reduced.
In this study, data are collected from Bangladesh Bureau It is also expected that the original equation is not affected
of Statistics (Survey of manufacturing industries 2012) and by Autocorrelation. Actually there is no reason to believe that
World Bank on various types of industries depending on their disturbance term of a garments industry can be affected by
performance, employment size, disturbance due to various disturbance of a leather industry or electronics industry
national and international affairs, inputs, imports, exports, because the data are cross sectional. It has been reconfirmed
nature of ownership (pvt or plc or state owned) etc. for the by checking DW statistic.
year 2012. From a total number of 1003 firms including But it’s thought that heteroscedastictiy might exist in present
Garments, Textiles, Food & Food Processing, Leather & case as there is a big heterogeneity in samples. Since the data
Leather Products, Electronics, Chemicals & Pharmaceuticals; contain information about six broad categories of industries
979 samples are collected. For the Exporter Industries, and some of these are exporters, production may be affected by
sample size is 467 and for the non exporter industries, sample national and international event. To be clear, BPG test and
size is 512. Data on about six major type of industries are White test have been done due to their suitability.
available. Information about exporter industries (among Finally it has been tried to remove heteroscedasticityby
these) and non exporter industries are also available. applying WLS. Weight (1/σi)has been derived by calculating
The study uses total value of output as Y, total liabilities as σi of different industries (for every type separately) and also
K, and number of permanent worker (including officials, for Exporter industries & Non exporter industries separately.
managers, personal) as L. The following regression has been run:
The Cobb-Douglas Production Function:
(Ln Y)/ σi = B/ σi+ α {(LnK)/ σi} + β {(LnL)/ σi}+ui/ σi (4)
The Cobb-Douglas Production, in its stochastic form, may
be expressed as The hypothesis of constant returns to scale (CRS) using
α β U Restricted Least Square Model (RLS) and F test have also
Y= f (K, L) = AK L e (1)
been examined.
Where, Y= Output, K= capital, L=Labor, U= Stochastic
disturbance term, e = Base of natural logarithm. 4. Findings
From this equation it is clear that the relationship between
output and the two inputs is nonlinear. However, log The following regression equationsare obtained for the
152 Shaiara Husain and Md. Shahidul Islam: A Test for the Cobb Douglas Production Function in
Manufacturing Sector: The Case of Bangladesh

whole Industrial Sector of Bangladesh. (N=979): Table 7. Chemicals & Pharmaceuticals Industry.

se (.506) (.066) (.100)

Table 1. Whole Industrial Sector of Bangladesh.
t (3.392) (9.096) (5.721)
se (0.164) (0.019) (0.025) R2=0.828, DW=1.812, VIF=2.015, TOL=0.496, F=190.797
t (20.67) (25.31) (19.88)
R2=0.714, DW=1.78, VIF=1.501, TOL=.666, F=1216.110 LnY = 1.716 +0.597 Lnk +0.573LnL + Ui (11)

LnY = 3.381+0.489 Lnk+0.505 LnL+ Ui (5) For the Exporter Industries (N=467) the equation is,

For the Garments Industry (N=304) the equation is, Table 8. Exporter Industry.

SE (0.304) (0.026) (0.036)

Table 2. Garments Industry.
T (12.951) (17.892) (12.783)
se (0.398) (0.040) (0.057) R2=0.566, DW=1.64, VIF=1.046, TOL=.956, F=303.066
t (11.95) (7.608) (10.661)
R2=0.492, DW=1.904, VIF=1.216, TOL=.823, F=145.810
LnY = 3935+ 0.470 Lnk+0.457LnL+ Ui (12)

For the Non Exporter Industries (N=502) the equation is,

LnY = 4.756+0.301 Lnk+0.609 LnL+ Ui (6)
Table 9. Non-exporter Industry.
For the Textiles (N=253) the equation is,
se (.216) (.031) (.045)
Table 3. Textile Industry. t (14.993) (16.498) (10.083)
R2=0.734,DW=1.70, VIF=2.223, TOL=0.45, F=689.382
se (0.313) (0.041) (0.062)
t (9.423) (10.061) (11.481) LnY = 3.240 +0.517 Lnk+0.457LnL+ Ui (13)
R2=0.77, DW=1.66, VIF=2.061, TOL=.485, F=411.038
On the basis of above information it may be said that, the
LnY = 2.954 +0.409 Lnk +0.712LnL+ Ui (7) coefficients are significant and do not suffer from
multicollinearity, at least do not from perfect
For the Food & Food Processing industries (N=143) the
multicollinearity, which is evident from the value of VIF &
equation is,
TOL and also free from the problem of autocorrelation as the
Table 4. Food & Food Processing Industry. values of DW statistic near about 2, and the study is dealing
with cross section data. However, it was suspected that the
se (.436) (.059) (.083) original equation would demonstrate regression
t (6.511) (11.877) (2.256) heterogeneityin type of industries. So information (residuals)
R2=0.68, DW=1.8, VIF=1.656,TOL=.604, F=148.938 have been tested formally and; BPG (Breusch-Pagan-
Godfrey) test and White’s General’ Heteroscedastictiy test
LnY = 2.838 + 0.696 Lnk +0.188 LnL + Ui (8) have been followed.
For the Leather & Leather Products (Sample size = N=95) Through BPG test, it is seen that observed Θ (ESS1/2)
the equation is, which follows χ2m-1 distribution is 24.612 which is
significant at both 5% & 1% level of significance. So, null
Table 5. Leather & Leather Products Industry. hypothesis of homoscedasticity can easily be rejected.
Through White test, it is observed that the value of
se (.694) (.074) (.117)
n R2 (which asymptotically approaches to χ2 df) is 71.467.
t (6.128) (5.653) (4.419)
The 5% critical value for 5 df is 11.075, the 10% critical
R2=0.54, DW=1.9, VIF=1.372,TOL=.729, F=53.158
value is 9.2363 and the 25% critical value is 6.62568. It can
be said on the basis of these information that there is
LnY = 4.253 + 0.418 Lnk +0.517 LnL + Ui (9)
heteroscedasticity.
For the Electronics (N=91) the equation is, To correct heteroscedasticity, WLS (weighted Least
Square) methodhas been followed. At first stage, the outcome
Table 6. Electronic Industry. has been derived according to the assumption of
se (.426) (.065) (.093) heteroscedasticity which may be a result of heterogeneity
t (6.478) (9.33) (4.176)
(due to types) of industries. Then the new result (weighted) is
as follows.
R2=0.787, DW=2.24, VIF=2.031, TOL=0.498, F= 162.542
Table 10. Whole Industry sector of Bangladesh (Corrected for
LnY = 2.761 + 0.609 Lnk +0.389LnL + Ui (10) Heteroscedasticity).

For the Chemicals & Pharmaceuticals (N=82) the equation se (.193) (.0253) (.031)
is, t (12.854) (24.312) (22.007)
R2=0.848 DW=1.54 VIF=2.810 TOL=0.356
 International Journal of Business and Economics Research 2016; 5(5): 149-154 153

(Ln Y)/ σi = 2.476 (1/σi) + .553 {(LnK)/ σi} + .680 {(LnL)/ σi} (14) employment increases by 1 unit, production increases by
0.39unit remaining other variables constant. It reveals that
In second Stage, now the study is interested to derive capital is more productive than labor in Electronics sector of
corrected result according to the assumption of heterogeneity Bangladesh.
(due to exporter or non exporter). The findings indicate that labor should be employed
Table 11. Exporter or Non-exporter Industry (Corrected for more in Garments sector, Textile sector and Leather &
Heteroscedasticity). leather products in order to increase production (economic
growth) as well as economic development. On the other
se (.172) (.020) (.027)
hand, capital should be employed more in Food & food
t (14.854) (28.430) (22.898)
processing industries, Electronics, Chemicals &
R2=0.848 DW=1.5 VIF=2.008 TOL=0.479
pharmaceuticals sectors.
(Ln Y)/ σi = 2.556 (1/σi) +0.562 {(LnK)/ σi} + 0.607 {(LnL)/ σi} (15) This paper is based on the data set from the survey of
manufacturing sector 2012 where the data is being updated
All of the above parameters and coefficients are highly up to 2012 and six major types of industries are being taken
significant. The results now show that the industrial sector is into account. More recent data up to 2015 and including
characterized by increasing returns to scale. On this other major types of industries in the calculation can be an
circumstance, the study is now eager to test a hypothesis avenue for future research in this area.
(null) of constant returns to scale (CRS) through restricted
least square (RLS) especially by F testing, 6. Conclusion
F={RSSR-RSSUR)/m}/{RSSUR/n-k)}
From the findings of this study, it can be said that Cobb-
=(2074.3-1785.53)/(1785.53/976)= 288.77/1.83=157.8 Douglas production function is applicable in the context of
manufacturing sector of Bangladesh and revealing an
Where RSSR= residual sum of square of restricted model, interesting aspect that this sector exhibits increasing returns
RSSUR is the residual sum of square of unrestricted model, to scale. It means that increase in production is more than
m= number of restriction (1 in the present study), n= number increase in employment of inputs (capital, labor) in
of observations, k = number of parameter. The F value production process which is very much optimistic result for
follows the F distribution with 1 df in the numerator and future investment decisions both form the government and
976df in the denominator, which is clearly significant and the private sector which might be helpful for employment
null hypothesis of CRS can be rejected. So,It can be said that generation and for also potential economic development.
the industrial sector (as a whole) probably characterized by
IRS (increasing returns to scale) and using mere RLS may
not give proper information. References
[1] Miller, E. (2008) ‘An Assessment of CES and Cobb-Douglas
5. Economic Interpretation & Limitation Production Functions’, Congressional Budget Office.
Retrieved
Assuming total value ofoutput as Y, total liabilities as K, from:https://www.cbo.gov/sites/default/files/cbofiles/ftpdocs/9
and number of permanent workers (including official 4xx/doc9497/2008-05.pdf
managers, personal)as L, analysis finds that coefficient for [2] Hossain, M. S. & Islam, A. N. (2013) ‘Productivity Analysis
capital (K) is 0.49 and coefficient for labor (L) is 0.51 for and Measuring the Returns to Scale of Manufacturing Firms in
entire manufacturing sector. It means that in manufacturing the South-West Region of Bangladesh’, IOSR Journal Of
sector, if amount of capital increases by 1 unit, production Humanities And Social Science (IOSR-JHSS),vol.17, no.1, pp.
increases by 0.49 unit remaining other variables constant; 69-77.
and if amount of labor employment increases by 1 unit, [3] Hagendorf, K. (2009) ‘The Cobb-Douglas Production
production increases by 0.51 unit remaining other variables Function and Political Economy’, Journal of Political
constant. It reveals that labor is more productive than capital Economy, vol.69, no.8, pp. 69-79.
in manufacturing sector of Bangladesh. [4] Hossain, S. I. (1987) Allocative and Technical Efficiency: ‘A
In case of Garments, coefficient for K is 0.30 and L is Study of Rural Enterprises in Bangladesh’, The Journal of
0.61. It shows productivity of inputs that labor is more Developing Economics, vol.25, no.1, pp. 56-72.
productive than capital. The statement is also true for Textile
[5] Kehindi, A. L. & T. T. Awoyem. (2009) ‘Analysis of
sector and Leather & leather products. Capital is, on other Economics Efficiency in Sawn wood Production in Southwest
hand, more productive than labor in Food & food processing Nigeria’, Journal of Human Ecology, vol.26, no.3, pp. 175-
industries, Electronics, Chemicals & pharmaceuticals sectors. 183.
In Electronics sector, coefficient for K is 0.61 and L is 0.39.
[6] Baten, M. A., Rana, M. Das, S. &Khaleque, M. A. (2006)
It means that in Electronics sector, if amount of capital ‘Technical Efficiency of Some Selected Manufacturing
increases by 1 unit, production increases by 0.61unit Industries in Bangladesh: A Stochastic Frontier Analysis’,The
remaining other variables constant; and if amount of labor Lahore Journal of Economics, vol.11, no 2, pp 23-41.
154 Shaiara Husain and Md. Shahidul Islam: A Test for the Cobb Douglas Production Function in
Manufacturing Sector: The Case of Bangladesh

[7] Hossain, M. M., Majumder, A. K. &Basak, T. (2012) ‘An [9] Bhashin, V. K. & Seth, V. K. (1980) ‘Estimation of production
Application of Non-Linear Cobb-Douglas Production Functions for Indian Manufacturing Industries’, Indian
Function to Selected Manufacturing Industries in Bangladesh’, Journal of Industrial Relations, vol.15, no.3, pp. 395-409.
Open Journal of Statistics, vol. 2, pp. 460-468.
[10] Hossain, M. Z., Bhatti, M. I. & Ali, M. Z. (2004) ‘An
[8] Wu, D. M. (1975) ‘Estimation of the Cobb-Douglas Econometric Analysis of Some Major Manufacturing
Production Function’, Econometrica, vol. 43, no.4, pp. 739- Industries: A Case Study’,Managerial Auditing Journal,
744. vol.19, no.6, pp. 790-795.