determinants of technical efficiency of rice farms in ... · determinants of technical efficiency...
TRANSCRIPT
DETERMINANTS OF TECHNICAL
EFFICIENCY OF RICE FARMS IN NORTH-
CENTRAL AND NORTH-WESTERN REGIONS
IN BANGLADESH
Stefan Bäckman
University of Helsinki, Finland
K.M. Zahidul Islam
University of Helsinki, Finland
John Sumelius
University of Helsinki, Finland
ABSTRACT
This paper estimates a quadratic stochastic frontier production function to examine the
determinants of technical efficiency in rice farming in Bangladesh using the computer program
FRONTIER 4.1. Primary data has been collected using multi-stage random sampling technique
from twelve villages in north-central and north-western regions in Bangladesh. Rice cultivation
displayed much variability in technical efficiency ranging from 0.16 to 0.94 with mean technical
efficiency of 0.83 which suggested substantial gains in output with available resources and existing
technologies. The analysis of the determinants of technical efficiency revealed that the age and
education of the household heads, availability of off-farm incomes, land fragmentation, access to
microfinance, extension visits, and regional variation were the major factors that caused efficiency
differentials among the farm households studied. Hence, the study proposes strategies such as
providing better extension services and farmer training programs, ensuring access to agricultural
microfinance, reducing land fragmentation and raising educational level of the farmers to enhance
technical efficiency.
JEL Classification: C21, Q12, D24
Key Words: Quadratic stochastic frontier production function, Technical efficiency,
Bangladesh, multi-stage random sampling, agricultural microfinance
Corresponding Author’s Email Address: [email protected]
INTRODUCTION
Agriculture is the most important sector in the economy of Bangladesh as it contributes
about 21% of the gross domestic product (GDP) and 50% of overall employment
(Bangladesh Agricultural Census 2008). The dominant food crop of Bangladesh is rice.
Rice accounts for 94% of the cereals consumed, supplies 68% of the protein in the
national diet, accounts for approximately 78% of the value of agricultural output, and
30% of consumer spending (Ahmed and Haggblade, 2000). It also accounts for 94% of
the total crops produced (Bangladesh Economic Review, 2009) and 76.62% of the
cropped area (BBS, 2006). In Bangladesh, 88.44% of the total households are located in
rural areas and they are more or less dependent on agriculture for a living (Bangladesh
Agricultural Census, 2008). Agriculture provides the basic food for the survival of the
74
subsistence farmers in Bangladesh. Subsistence farmers account for the greatest
proportion of those engaged in farming. Bangladesh agriculture already operates at its
land frontier and there is little or no scope to expand the cultivable land to meet the
increasing demand for food requirements for its ever-increasing population (Rahman,
2003). Moreover, high population growth, frequent crop failures resulting from flooding
or droughts put pressure for intensification of land use. So, this country simply cannot
afford any diminution in the productivity of its limited 8.44 million hectare arable lands
(BBS, 2006). We also need to maintain optimum productivity of our existing cultivable
lands in order to get increased yields.
The main agricultural products are rice, jute, sugarcane, potatoes, vegetables,
oilseeds, pulses and tea. Three rice crops are grown during the crop cycle beginning in
April- the 'Aus' (spring) crop, the 'Aman' (summer) crop, and the 'Boro' (winter) crop.
The first two are traditional1, rain-fed crops, whereas the Boro crop is the High Yield
Variety (HYV2). However, there is evidence that actual farm yields for both higher
yielding and traditional varieties show considerable shortfalls in yield from those attained
by experimental station levels; which gives rise to a „yield gap‟(De Datta et al., 1978) of
approximately 40% to 50% (BRRI, 2000; Sattar, 2000). The average rice yield in
Bangladesh is 2.74 tonnes/ha (BBS, 2008), which is much lower than those of other
Asian countries. The potential gain from closing this yield gap is higher for Bangladesh
compared to other Asian countries such as China, Korea, Indonesia, Myanmar, Nepal and
Vietnam (Pingali et al., 1997). This „yield gap‟ indicates a difference in productivity
between „best practice‟ and on other less efficient farms that operate with comparable
resource constraints under similar circumstances (Wadud, 1999; Villano, 2005). The
difference between the actual and technically feasible output for most crops implies great
potential for increasing food and agriculture production through improvements in
productivity. For a resource scarce country such as Bangladesh where opportunities to
develop and adopt new technologies are rare, empirical investigations of technical
efficiency and its determinants in rice farming are a dire necessity. Such studies help to
determine the level at which farmers are using existing technologies, and also explore the
possibility of raising the productivity by increasing the efficiency. A great deal of
empirical studies (e.g. Sharma et al. 1999, Nyemeck et al., 2003; Tzouvelekas et al.,;
Villano, 2005; Kalirajan, 1984; Kumbhkar, 1987; Battese et al., 1996; Coelli and Battese,
1996; Battese and Coelli, 1992 ; Binam et al., 2004; Bravo-Ureta and Pinheiro, 1997;
Wang et al., 1996b) have been conducted in other countries to measure the technical
efficiency by using production function, mathematical programming technique, panel
data, as well as using the cross-section data.
Determinants of inefficiency include some exogenous variables that have some
impacts on efficiency. Examples of such influences are age of the farmer, the education
level of the farmer, the size of the farm, access to credit, land tenure, farmers‟ capabilities
to use the inputs and so forth. The measurement of efficiency entails the determination of
factors influencing the overall efficiency. The most common approach to do this is the
determination of an inefficiency index (considered as the dependent variable) and then
regress the dependent variable against a set of explanatory variables considered to affect
the efficiency levels. Kumbhakar et al. (1991) proposed that the determinants of
inefficiency should be estimated simultaneously by noting that the two-stage procedures
introduce some bias in estimation. In the two-stage approach, efficiency scores
75
determined in the first stage regression are regressed by background and production
environment related factors (Pitt and Lee, 1981). This approach contains serious
problems concerning assumptions made for the non-negative random variable, ui.
Moreover, the second stage specification conflicts with the assumption that uis are
independent and identically distributed. This second stage was criticised by Battese and
Coelli (1995) and Wang and Schmidt (2002). Like Kumbhakar et al. (1991), Battese and
Coelli (1995), Huang and Liu (1994) also proposed similar models for incorporating
technical inefficiency effects.
There have been very few studies undertaken in Bangladesh that measured the
determinants of technical efficiency. Khan et al. (2010) investigated technical efficiency
of a sample of 150 Bangladeshi rice farmers. Separate Cobb-Douglas production frontiers
were estimated for boro and aman rice producers. The mean technical efficiency scores
reported were 95% and 91% respectively and the result indicated that farmers‟ education
had a significant influence on technical efficiency of boro rice producers. Rahman and
Rahman (2009) examined how the land fragmentation and resource ownership (land,
animal power and family labor) affected productivity and technical efficiency of rice
producers in Bangladesh, using survey data from farms. They estimated the mean
technical efficiency to be 91 % and the efficiency differentials were markedly influenced
by land fragmentation and resource ownership. Asadullah and Rahman (2009) examined
the influence of education on farm production efficiency for a large dataset obtained from
141 villages and their analyses revealed that household education significantly reduced
production inefficiencies. Wadud (2003) used both Data Envelopment Analysis (DEA)
and Stochastic Frontier Approaches (SFA) to examine the technical, allocative, and
economic efficiency of a sample taken from 150 farm households and found high level of
technical efficiency. The technical efficiency was explained by land degradation and
irrigation infrastructure. Coelli et al. (2002) used DEA and examined technical, allocative,
costs, and scale efficiencies for the modern Aman and modern Boro rice from a total of
406 sample households. They reported technical efficiency of 66% for Aman rice
whereas a technical efficiency of 69% was reported for Boro rice. Sharif and Dar (1996)
examined how education, growing experience, and farm size influenced technical
efficiency for HYV Boro rice using a two step procedure, and found that education was
positively related to technical efficiency. However, there have been no studies on farm
level technical efficiency and its determinants focusing on financial factor by way of
microfinance. This study introduced a new explanatory variable named access to
agricultural microfinance which was not examined in the previous studies as a potential
determinant of efficiency in rice farming. The study thus attempts to test the hypothesis
that access to agricultural microfinance affects rice production efficiency.
Given that little attention has been devoted to quantify and identify the
determinants of technical efficiency, the present study aims to estimate the determinants
of technical efficiency and each factor‟s contribution to inefficiency. The present study
chooses the appropriate functional form of the inefficiency component and a suitable
production function model that fit the data most based on several empirical hypotheses.
Another justification of this study is the introduction of a flexible production function
rather than the commonly used SFA using Cobb-Douglas and or DEA in estimating
technical efficiency. Further, proposing microfinance as a determinant of technical
efficiency of farmers in Bangladesh is a substantially different policy variable. The
76
present paper contributes to the literature in three ways. First, we incorporated the whole
farm rice production in the analysis and in doing so we assumed that the economic
situation of a farmer is better represented by aggregate production of crops, second, we
estimated and identified the determinants of the whole farm rather than for a specific rice
crop and thus gave recommendations to the policy makers to formulate policies that
improve farm productivity, third, our data reinforces some theoretical arguments that
extension visits, education, access to finance and regional variation may have on farm
productivity and efficiency. An understanding of these relationships can provide the
policy makers with information about the nature of the problems facing the rice farms in
Bangladesh and to design programs that improve efficiency.
The rest of the paper is organized as follows: Section two outlines the theoretical
model. Section three describes the methodology, study areas, survey method, and list of
variables of collected data. Section four specifies the models and the results are discussed
in section five. Section six concludes.
ANALYTICAL FRAMEWORK
According to Farrell‟s (1957) model, technical efficiency (TE) is defined as the ability of
a farm to obtain the best production from a given set of inputs (output-increasing
oriented), or alternatively as the measure of the ability to use the minimum feasible
amount of inputs to produce a certain level of output (input-saving oriented) (Greene,
1980; Atkinson and Cornwell, 1994). Consequently, technical inefficiency is defined as
the extent to which firms fail to reach the optimal production. Farrell (1957) proposed to
measure TE of a farm by comparing its observed output to that output which could be
produced by a fully efficient farm, given the same bundle of inputs. Aigner et al. (1977)
and Meeusen and van den Broeck (1977) independently proposed the stochastic frontier
(SF) production function to account for the presence of measurement errors and other
noise in the data, which are beyond the control of managers. Farmers in general operate
under uncertainty and therefore, the present study employs a stochastic production
frontier approach for measuring TE. Following Battese and Coelli (1995), the following
stochastic frontier production function and inefficiency effects model are estimated
simultaneously using single stage with the computer program, FRONTIER 4.1,
developed by Coelli (1996).
Following their specification, we specify the general SF model defined as:
iii xfy );( i = 1, 2, …, N (1)
Where, yi is the revenue from rice for the i
th
farm, xi is a vector of k inputs (or
cost of inputs), β is a vector of unknown parameters to be estimated, )(f is a suitable
functional form for the frontier (Cobb-Douglas, translog or quadratic), εi is an error term,
and N is the total number of observations. The stochastic frontier production is also called
„composite error‟ model, because it postulates that the error term εi is decomposed into
two components: a stochastic random error component (random shocks/white noise) and
a technical inefficiency component defined as follows:
77
εi ii uv
(2)
Where vi is a symmetrical two sided normally distributed random error that
captures the stochastic effects outside the farmers‟ control (for example, weather, natural
disasters, omitted variables, luck, exogenous shocks, measurement errors, and other
statistical noise). It is identically, independently and normally distributed vi~iid N (0, 2
v ), independent of the uis. Thus, vi, allows the production frontier to vary across farms,
or over time for the same farms and therefore, the production frontier is stochastic in
nature. The term ui (asymmetric non-negative error term) is a one sided (u
i ≥ 0) efficiency
component that captures the technical inefficiency of the ith
farm. This may follow a half-
normal, exponential, truncated-normal or gamma distribution (Stevenson, 1980; Aigner et
al., 1977; 1990; Meeusen and Broeck, 1977). In this study we assumed that ui follows the
exponential distribution as was done in various published studies in applied stochastic
frontier literature. It is obtained by the truncation at zero of the normal distribution with
mean μ, and variance (2
u ). If μ is pre-assigned to be zero, then the distribution is half-
normal. The variance parameters of the model are parameterized as:
, 22222
suuvs So that 0 ≤ γ ≤ 1 (3)
The parameter γ must lie between 0 and 1. Here, 2
s denotes the total variation
in the dependent variable due to technical inefficiency (2
u ) and random shocks (2
v )
together. The gamma ( ) parameter explains the impact of inefficiency on output. The
maximum likelihood estimation (MLE) of equation (1) provides consistent estimators for
β, γ, and 2
s parameters. Aigner et al. (1977) expressed the likelihood function in terms
of the two variance parameters, 222
vus and vu / . Battese and Corra
(1977) suggested that the parameter, 22
/ su , be used because it has a value
between zero and 1 and this property permits to obtain a suitable starting value for an
iterative maximization process, whereas the -parameter could be any non-negative
value. A value of closer to zero implies that much of the variation of the observed
output from frontier output is due to random stochastic effects, whereas a value of
closer to one implies proportion of the random variation in output explained by
inefficiency effects or differences in technical efficiency.
SURVEY DATA
The Study Areas and Sampling Methods
Data were collected from twelve villages in north-west and north-central regions in
Bangladesh through a survey conducted in June-August 2009. These regions were
78
selected due to their high levels of poverty and good agricultural potential. For
microfinance users, data were collected with the help of Microfinance Institutions‟ (MFIs)
clients‟ lists. Personal interviews were conducted for both the users and non users of
microfinance to collect the data. We interviewed 180 agricultural microfinance borrowers
and 180 non-borrowers (the control group) of agricultural microfinance who operated
farm land between 0.2 to 1 hectare. This land holding criteria was set by the microfinance
institutions while granting agricultural microfinance loan. Non-borrowers are selected
based on similar land holdings and socio-economic background to provide a control
group for comparison with borrowers. As the entire sample is used in explaining
efficiency among the sampled farms, there are no sample-selection issues as well.
Farmers having land more than 1 hectare but taking microfinance exclusively for
agriculture were also considered. Data were collected from the farmers producing Boro,
Aman, and Aus rice crop from the selected areas. As most farmers in Bangladesh are
illiterate, most of them, with some exceptions, do not keep any vouchers or written record
of input prices as well as do not maintain any written documents about input- output data.
With a view to minimizing errors stemming from reliance on farmers‟ memory, data were
collected immediately after the harvest from the three growing seasons.
In conducting the research, multistage sampling technique was used. The first
stage was the purposive selection of two districts (Mymensingh and Sherpur) form north-
central and four districts (Rajshahi, Naogoan, Dinajpur, and Gaibandha) from the north-
west region in Bangladesh. The second stage involved the identification of those farmers
who had taken microfinance specially allocated for agricultural production. Finally, a
multi-stage proportional random sampling method was used to select 60 households (30
from microfinance borrowers and 30 from non-borrowers of microfinance) from each
district, thus a total of 360 households were surveyed.
Description of the Data
Output is defined as the market value of the aggregated rice production in the
survey period. Rice output prices were gathered from individual farms. All rice (Boro,
Aman, Aus) produced on the sample farms were aggregated into one output value (Taka3).
Land represents the total amount of land (own-cultivated land, sharecropping land, and
rented/leased land) used for rice production and was measured in hectare. Labor includes
both family (imputed for hired labor) and hired labor utilized for pre and post planting
operations and harvesting excluding threshing. It was measured in annual labor-days used
for rice production. Fertilizers include all sorts of organic and inorganic fertilizers used
by the farm households for rice production. It represents the total cost of fertilizer
measured by market prices. Seeds included all seeds used in rice production and was
measured in Taka. If seedlings were purchased, it was converted into equivalent amount
of seeds to compute the seed price.
79
TABLE 1. DESCRIPTIVE STATISTICS
Variables Unit Mean Std. Dev. Minimum Maximum
Output Taka 88326 98980 4200 795000
Land Hectare 1.25 1.64 0.09 13.47
Seeds Taka 3039 6550 70 102080
Fertilizers Taka 12584 19544 0 180200
Labor Days 199 175 12 1274
Irrigation Taka 7326 11125 0 155000
Pesticides Taka 1334 2670 0 27000
Other variable
costs
Taka 2976 3780 0 36960
Tractor & animal
power
Taka 5922 7688 160 67200
Capital Taka 42117 82303 40 902100
Age of farmer Years 42 13 17 85
Education Years 4.99 5 0 16
Extension No. 6 7 0 24
Off-farm income Taka 61151 78094 0 600000
Experience Years 23 13 1 70
Numbers of plots No. 4 2 1 10
Source: Computed by the authors.
Irrigation represents the total irrigation costs for rice production. This cost is estimated
from total rice land irrigated and it was measured in Taka. Tilling includes the total land
tilled with tractor and or bullocks. It represents the total cost of tilling measured in Taka.
Other costs include pesticide, seed bed preparation, and crop transportation costs and it
was measured in Taka. Capital is the sum of farm tools, machineries and animal power
used in rice production and was measured in Taka. A large set of data were also collected
about the farmers‟ socio-economic characteristics and other aspects such as the farmer‟s
age, years of schooling, access to credit, numbers of contacts with extension agents,
wealth, investment, institutional constraints to get loan, land ownership etc.
Some basic characteristics of the sample farms are presented in Table 1. It is
evident that farms were small in terms of output and total area farmed. On average each
farm produces rice worth Taka 88326 and it is highly variable ranging from Taka 4200 to
Taka 795000. Farm operators averaged 42 years old and it ranged from 17 years to 85
years. Approximately 98% of the farm households were adult. Their experience in rice
farming was vast and it ranged from 1 year to 70 years while their education level was
moderate.
80
MODEL SPECIFICATION
Stochastic Frontier Production (SFP)
We specify a log-quadratic production function as introduced by Chu, Aigner and
Frankel (1970) to estimate the stochastic frontier production function. We used a less
restrictive log-quadratic specification that takes into account both the Cobb-Douglas
specification and translog second order (excluding the cross-term) log-linear form.
The following quadratic model was specified in this study:
8
1
16
9
2
0 )(loglnlnj j
iiijjijji uvxxy (4)
Where yi represents the value of rice output of ith
farm and j is the jth
input used
in production. ln = natural logarithms , X1 = Total Land used for rice production; X2 =
Farm Capital ; X3 = Total labor days used ; X4 = Costs of Fertilizers; X5 = Irrigation costs ;
X6 = Seeds costs ; X7 = Tractor and animal power costs, and X8 = other variable costs .
Inefficiency Model for the Cross Section Data
The technical inefficiency (u
i) could be estimated by subtracting TE from unity. The
function determining the technical inefficiency effect is defined in its general form as a
linear function of socio-economic and management factors. It can be defined in the
following equation:
8
1
0
k
ikkiu (5)
Where, ui is the technical inefficiency effect, δk is the coefficient of explanatory
variables. The Zi variables represent the socio-economic characteristics of the farm
explaining inefficiency and may not be functions of y. We proposed that the technical
inefficiency could be explained by the following determinants:
Zi1
= Age of the household head (years); Zi2
= Education (number of years of schooling of
the farmer); Zi3
= Experience (years); Zi4
= Off-farm income (in Taka); Zi5
= Land
fragmentation (it includes the total number of plots operated); Zi6
= Extension visits
(number); Zi7
= Access to microfinance (A dummy variable to measure the influence of
microfinance on efficiency. Value is 0 if the farmer had cash credit in the last 12 months
prior to the survey from microfinance institutions exclusively for agriculture, otherwise 1)
and Zi8
= Region (A dummy variable. It takes a value of 1 if the region is northwest and 0
otherwise).
81
Hypotheses Tests
The following tests have been carried out for testing the functional forms, inefficiency
effects, and determinants of coefficients for rice farmers in the study areas:
(1) Frontier model specification for the data is Cobb-Douglas production function.
That is :0H C-D ( )0............: 4490 H is an adequate representation
of the production Function.
(2) Frontier model specification for the data is a Quadratic production function.
That is 0......: 1690 H is an adequate representation of the production
function. Here 169...... represent the quadratic terms.
(3) Frontier model specification for the data is a Translog production function. That
is 0......: 4490 H is an adequate representation of the production
Function. Here 449...... represent the quadratic terms and also the cross
terms.
(4) There is no inefficiency effect that is 0....: 8100 H
(5) The coefficients of determinants of inefficiency model equals zero that is
0.. 810 H
All the above hypotheses were tested using the generalized Likelihood Ratio
(LR) which is defined as: -2[L(H0)-L(H1)], where L(H0) and L(H1) are the values of
the likelihood function of the frontier model under null hypothesis and alternative
hypothesis respectively. The null hypothesis was rejected when CRL2
.2 . If the null
hypothesis was true, the test statistic had approximately a 2 -distribution or mixed
2 -
distribution with degrees of freedom equal to the difference between the number of
parameters specified in the null hypothesis and alternative hypothesis. If there is no
inefficiency effect, 0....: 810 H , then the test statistic is distributed
like a mixed 2 -distribution with degrees of freedom equal to 9. All the hypotheses are
conducted assuming 05.0 . Thus, if the 2 statistic exceeds the 95% point for the
appropriate 2 -distribution, the null hypothesis would be rejected. The critical value for
the Likelihood ratio for was obtained from Table 1 of Kodde and Palm (1986).
Output Elasticity ( j )
The rice output elasticity for land, labour, fertilizers, irrigation, seeds, tractor hours,
capital, and other variable costs are included in the regression of interest. The output
elasticity ( j ) with respect to inputs were computed for the quadratic model as follows:
82
y
x
x
xf ij
ij
ij
j
),ˆ( ) (6)
For quadratic terms, we may represent the above elasticity in the following equation:
j =y
xx
ij
ijjj )ˆ2ˆ( )8( (7)
Where, ijx is the mean of thj input, y is the mean production estimated at mean
inputs, j is the estimated coefficient of the X term and )8(ˆ
j is the estimated
coefficient of the 2X term.
Returns to Scale (RTS)
Returns to scale is equal to the sum of marginal production elasticises of each input. It is
defined in the following equation:
RTS =
8
1j
j (8)
RESULTS AND DISCUSSIONS
The MLE of the parameters of the Cobb-Douglas stochastic frontier production function,
the quadratic production function, and the translog model were obtained using computer
program NLOGIT 4.0 (Greene, 2007). The results are presented in Table 2. To select the
most suitable model (Cobb-Douglas, quadratic, or translog) we tried all models with
different distribution assumptions of the error component (ui) and tested all models with
the results of Log likelihood at the predetermined critical value ( )71.295.0),1(2 to reject
or accept one model over another. First, we tested the Cobb-Douglas with the translog to
determine whether Cobb-Douglas fitted the data by using the likelihood ratio test4. We
rejected the null hypothesis and excluded this model from further consideration. Finally,
we compared the translog model with the quadratic function and found that linear-
quadratic model fitted the data well with the expected signs for production coefficients
and with the results of hypothesis, which assumed the error term to be exponential. The
estimates of the quadratic stochastic frontier production are presented in Table 2. The
result revealed that with the exception of fertilizer all the explanatory variables conform
to prior expectation of signs of the coefficients for the quadratic production function with
nine coefficients significant at different significance levels and suggesting that model
fits the data well.
83
TABLE 2. MAXIMUM LIKELIHOOD ESTIMATES OF COBB-DOUGLAS AND
QUADRATIC STOCHASTIC FRONTIER PRODUCTION FUNCTION
Variables Parameters Cobb-Douglas ML Estimaste Quadratic ML Estimates
Intercept
0 4.665*** (21.59) 1.329 (1.13)
Ln (Land)
1 0.526*** (14.21) 0.742***(3.04)
Ln (Capital)
2 -0.016** (-2.00) 0.075 (1.14)
Ln (Labor)
3 0.160*** (4.21) 1.079***(3.07)
Ln (Fertilizer)
4 0.021 (1.00) -0.085**(2.36)
Ln (Irrigation)
5 0.030*** (2.73) 0.017 (1.06)
Ln (Seed)
6 0.103*** (3.96) 0.198(0.188)
Ln (Tractor and bullock)
7 0.1020*** (3.52) 0.631***(2.67)
Ln (Other variable costs)
8 0.102*** (4.86) 0.011(0.31)
[Ln (Land)]2
9 -0.023(-1.00)
[Ln (Capital)]2
10 -0.005 (0.004)
[Ln (Labor)]2
11 -0.068***(-2.72)
[Ln (Fertilizer)]2
12 0.010*** (3.33)
[Ln (Irrigation)]2
13 0.0037* (1.95)
[Ln (Seed)]2
14 -0.008(0.66)
[Ln (Tractor)]2
15 -0.034**(-2.27)
[Ln (Other variable costs)]2
16 0.008***(2.67)
Variance parameters
)/( 222vuu
0.475 0.2529
)(222
vus 0.20
0.119
Log Likelihood -152.87 -125.843
2
v 0.0889
2
u 0.0301
Inefficiency effects
Constant
0 -2.184 (-0.54)
Age
1 0.003 (0.17)
Education
2 0.017 (1.13)
Experience
3 -0.057***(-2.71)
Off-farm Income
4 0.001 (1.00)
84
Number of plots
5 0.004**(2.00)
Extension Visits
6 -0.024 (-0.88)
Access to microfinance
7 0.132 (0.62)
Region
8 0.154 (0.38)
Source: Computed by the authors.
Notes: t-statistics are in parentheses; *, **, *** indicate significance at 10%, 5% and 1% level, respectively, Log Likelihood
under OLS estimates is -127.790
The significant positive coefficients of land, labour, and tractor imply that as
each of these variables is increased, rice production also increases. One explanation of
negative coefficient of fertilizer may be due to the wrong application leading to excessive
use of urea as source of N fertilizer since it is relatively cheap and use very little of
expensive fertilizers like P and K. Lack of farmers‟ knowledge about the need for
balancing the application of fertilizer is another plausible reason of this negative sign.
Government subsidy for fertilizer in Bangladesh may also encourage the farmers to use
too much urea and it may have long term damaging effects on the long-term productivity
of soil. Since the fertilizer dealers are more responsive than the government to local
fertilizer requirements and preferences, government may encourage the dealers to guide
and motivate the farmers in maintaining an optimum nutrient balance on the farms while
selling fertilizers to the farmers. Other independent variables such as seeds, irrigation,
and other variables costs have positive coefficients but are insignificant under quadratic
production function. For labor, the poor performance is attributed to high average man
days of labour (199). This is an indication of over-utilization of labor as is typical of
developing countries.
Analysis of Productive Efficiency
The result of the TE estimates is presented in Table 3. The TE analysis revealed that
technical efficiency score of sample farms varied from 16.22% to 94.47%, with the mean
efficiency level being 83%. The mean technical efficiency implies that the average farm
produces 83% of the maximum attainable with given input levels. This variation is also
confirmed by the value of gamma ( ) that is 0.25. The gamma value of 0.25 suggests
that 25% variation in output is due to the differences in technical efficiencies of farm
household in Bangladesh. This finding establishes the fact that inefficiencies exist in the
sampled farmers. Moreover, the corresponding variance-ratio parameter5,
* , implies that
11% differences between observed and maximum frontier output for rice farming is due
to the existing differences in efficiency among the sample farms.
85
TABLE 3. SUMMARY OF TECHNICAL EFFICIENCY OF THE RICE
FARMERS
Efficiency level Frequency Percentage
0.10-0.19 1 0.28
0.2-0.29 0 0
0.3-0.39 4 1.11
0.4-0.49 2 0.56
0.5-0.59 6 1.67
0.6-0.69 13 3.61
0.7-0.79 65 18.05
0.8-0.89 222 61.67
0.9-0.99 47 13.05
Total farms 360 100
Mean 82.65
Standard Deviation 9.84
Min. 16.22
Max. 94.47
Source: Computed by the authors.
The indices of TE indicates that if the average farmer of the sample could
achieve the TE level of its most efficient counterpart, then average farmers could increase
their output by 12% approximately [that is, 1-(83/94)]. Similarly the most technically
inefficient farmer could increase the production by 83% approximately [that is, 1-(16/94)]
if he/she could increase the level of TE to his/her most efficient counterpart. Since the
mean TE is 83%, it can be deduced that 17% of the output is lost due to the inefficiency
in rice producing system or in the inefficiency among the sampled farmers or both
combined. It also indicates that small farms in the study area, on average, can gain output
growth at least by 12% through the improvements in the technical efficiency. For a land
scarce country like Bangladesh this gain in growth will help much to ensure food security
in the country. These findings may invite attention of the policy makers to improve the
efficiency of the farmers through adoption of right policies. Results of the Hypotheses Test
The formulation and results of different hypotheses (model selection, inefficiency effect,
determinants of coefficients) are presented in Table 4. All the hypotheses are tested by
using generalized likelihood-ratio (LR). The first hypothesis relates to the
appropriateness of the Cobb-Douglas functional form in preference to translog model.
The computed LR statistic exceeded the tabulated value of 2 at 5% significance level.
So, we rejected the null hypothesis by indicating that the translog functional form is a
better representation of the data.
86
TABLE 4. SUMMARY OF HYPOTHESES FOR PARAMETERS OF
STOCHASTIC FRONTIER AND INEFFICIENCY EFFECTS MODELS
Null Hypotheses L(H0 L(H1 ) LR 2 critical
value
Decision
1.Production Function is
Cobb-Douglas
( )0............: 4490 H
-152.867 -111.849 82.04 49.23 Reject H0
2. Production Function is
Cobb-Douglas
( 0......: 1690 H )
-152.867 -125.843 54.05 14.85 Reject H0
3. Production function is
Quadratic
0......: 4490 H
-125.843 -111.849 27.99 41.98 Accept H0
4. There is no inefficiency
effect
(H0: =0)
-127.790 -125.843 3.89 2.71 Reject H0
5. The coefficients of
determinants of inefficiency
model equals zero
0.... 8100 H
-116.24 -125.843 19.21 16.27 Reject H0
Source: Computed by the authors.
The second hypothesis relates to the appropriateness of the Cobb-Douglas in
preference to the quadratic functional form. This hypothesis was also rejected at 5% level
of significance and indicated that quadratic functional form is a better formulation than
the Cobb-Douglas functional form. The third hypothesis relates to the appropriateness of
the quadratic functional form in preference to the translog functional form. The computed
LR statistic fell below the tabulated value of 2 at 5 % significance level and we failed
to reject the null hypothesis indicating that the quadratic functional form was the best fit
for the data. Therefore, we selected this functional form in our analysis. The fourth
hypothesis stated that 0 , is rejected at the 5% level of significance confirming that
inefficiencies exist and are indeed stochastic (LR statistic 3.89> 71.22
95.0,1 ). The fifth
hypothesis that dd 0...00 , which means that the technical inefficiency
effects were not related to the variables specified in the inefficiency effect model, is also
rejected at the 5% level of significance (LR statistic 19.21> 27.162
95.0,9 ). Thus the
observed inefficiency among the rice farmers in Bangladesh can be attributed to the
variables specified in the model and the variables play a significant role in explaining the
observed inefficiency.
87
Elasticities and Returns to Scale
Table 5 reports output elasticity estimates with respect to eight production inputs used
and were evaluated at the sample means. Furthermore, the last column of the same table
gives the scale elasticities for combined inputs. Scale elasticity exceeds unity thus leading
to the conclusion that rice producers operate in the region of increasing returns to scale.
The sample mean of RTS, 1.04, indicates that the farmers could be made scale efficient
by providing more input to produce more output with the exception of capital (tractors,
buildings, machineries).
TABLE 5. ELASTICITIES AND RETURN TO SCALE OF THE QUADRATIC
FRONTIER PRODUCTION FUNCTION
Independent Variable Mean Value Elasticities RTS
Land 1.25 ha 0.505 1.04
Labor 199 labor days 0.106
Fertilizer Taka 12584 0.093
Irrigation Taka 7326 0.076
Seeds Taka 3039 0.080
Tractors and animal power Taka 5922 0.070
Capital Taka 42117 -0.016
Other variable costs Taka 2976 0.128
Source: Computed by the authors.
The negative sign of capital implies low marginal increments to total output if more
capital is provided. One explanation may be that farmers in Bangladesh are mostly
subsistence farmers and operate very small size of land (Table 1). This leads to increasing
the opportunity cost of capital items like tractor and other expensive cultivating and
harvesting machineries. The elasticity of output with respect to land is the highest among
all the inputs, which demonstrates the importance of scarce land in boosting rice
production in Bangladesh. It is concluded that land had the major effect on the total value
of rice production. The policy implication of this finding is that government could give
incentives and encouragement to the farmers to keep their existing arable land and bring
the remaining fallow land under cultivation, if any. Elasticity of labor is the third highest
but excess use of the labor exerts negative impacts on output as is observed from the
second order of labor. Both fertilizers and irrigation should be utilized efficiently to
ensure optimum growing conditions of land since their inappropriate utilization may have
far reaching impacts through degradation of land and its soil quality.
88
Factors Explaining Inefficiency
The parameters of the explanatory variables in the inefficiency model were
simultaneously estimated in a single stage using computer program, FRONTIER 4.1. The
dependent variable of the model was inefficiency and the negative signs imply that an
increase in the explanatory variable would decrease the corresponding level of
inefficiency. Lower part of Table 2 shows the coefficients of explanatory variables in the
inefficiency model. The results show that most of the signs related to inefficiency
determinants were as expected. The parameter estimates showed that factors such as age,
education, number of plots, region (dummy variable), access to microfinance (dummy
variable), and off-farm income were positively related with inefficiency while extension
visits and experience were negatively related to inefficiency.
The age coefficient is positive but insignificant, which indicates that younger
farmers are more efficient than older farmers. This conforms to the results obtained by
Coelli and Battese (1996) and Battese and Coelli (1995). A possible explanation might be
that the adoption of new technology and managerial capability to carry out farming
activities decreases with age. However, its inclusion in the inefficiency model improved
the model‟s explanatory power.6 The analysis revealed that education, measured in terms
of years of schooling, had a statistically insignificant effect on technical inefficiency.
This result conforms to those obtained by Wadud (2003) and Coelli and Battese (1996). It
can be deduced that five or more years of formal education are required before increases
in efficiency can be observed. The off-farm income variable is positively and
insignificantly related with technical inefficiency. This indicates that higher off-farm
income increases the technical inefficiency of rice farmers. It also implies that the more
off-farm hours a producer works, the less time is devoted to farming, thus resulting in
higher technical inefficiency. This result is consistent with Abdulai and Eberlin (2001)
and Coelli et al. (2002). The estimated coefficient of farming experience had a significant
negative impact on technical inefficiency, which implies that rice farmers‟ expertise
assists them in ensuring the optimal timing and use of inputs and thereby reduces their
technical inefficiency. Another probable reason for the significant negative contribution
of experience on technical inefficiency could be that farmers with more years of
experience tend to gain more proficiency through „learning-by-doing‟ in uncertain
production environment. Several other empirical studies have also reported similar results
(Bozoglu and Chehan, 2007; Huffman 2001; Kalirajan and Flinn, 1983).
The estimated co-efficient of the number of plots operated by the farm
household is positive and significant. It implies that the more the lands are fragmented,
the more the technical inefficiency increases. That is farmers with less fragmented land
will operate at higher technical efficiency levels. This result is also consistent with those
of Wadud (2003), Wadud and White (2000) and Coelli and Battese (1996). Higher
technical efficiency associated with less fragmented land can be attributed to adopting
modern technologies and better farm practices such as the use of irrigation (Wadud,
1999). Extension visits were negatively related with technical efficiency. Although not
significant, however, the extension visits may be an important policy instrument by
which the government could raise agricultural productivity since the agricultural
extension visits enable the farmers to learn better farm management methods and more
efficient uses of limited resources. The policy implication of this finding is that the
89
government could support further the agricultural extension network in order to make the
interactions between the farm and extension agent more participative and field oriented
through practical demonstrations rather than just conveying some recommendations.
The coefficient of the access to microfinance (dummy variable) was positively
related to technical inefficiency. It indicates that those farmers who did not have
agricultural microfinance, tended to have higher technical inefficiency levels than their
peers. It implies that access to microfinance reduces the technical inefficiency of the
sample farms as the estimated average technical efficiency of microfinance borrowers
was 84% and for the non-borrowers the average technical efficiency was 81%. The
difference in mean technical efficiency is also significant at 5% significance level.7 The
surveyed farms in the study areas faced acute shortage of working capital for farming.
Average loan obtained by the microfinance borrowers for agriculture was Taka 16673
while the average demand was Taka 39383.8 The shortage of working capital due to the
increasing price of inputs as well as the low returns to farm produce resulted in high level
of technical inefficiency. Most rice farms faced negative cash flow during the planting
and growing period due to the time lag of purchasing the inputs and receiving the returns
long after the crops are harvested. Credit thereby helps to mitigate the financial constraint
and to reduce inefficiency. This finding also conforms to the results of Binam et al.
(2004). Credit also helps the farmers to increase farm revenue while lack of credit
decreases the efficiency of the farmers by limiting their adoption of high yielding
varieties and acquiring of information for increased productivity, a view supported by
Wozniak (1993).
Thus, improved access to agricultural microfinance remains an important issue
for improving the rural farm production efficiency in Bangladesh. Another implication of
this finding is that farmers who are indebted need to meet their repayment obligations and
this puts more pressure on the farmers to produce more output to repay the loan by
generating more cash. For microfinance borrowers, the future possibility of getting loan
depends on current repayment behaviour and this implicit pressure to repay the loan acts
as a catalyst to optimize the resources to produce more. The dummy variable region is
positively related with technical inefficiency. Thus farmers operating in the north western
region perform less efficiently compared to those of the north central region. This finding
reinforces the argument that regional concentration is a vital policy instrument that
should be addressed in formulating agricultural policy in Bangladesh.
CONCLUSION This study uses a quadratic stochastic frontier production function on survey data (2009)
to determine the technical efficiency and its determinants in rice production in north-
central and north-western region of Bangladesh. First, we draw conclusion on the
methodology choice of production technology. This was based on some selected
hypotheses and we concluded that traditional production function model was not
adequate for farm level analysis. Consequently, we proposed the quadratic stochastic
production function. Second, the results revealed that mean technical efficiency of farms
was 0.83, indicating that there are opportunities to gain substantial additional output or
decreases the inputs, given the existing technology and resource endowments of rice
farmers in the study areas.
90
The empirical results revealed that inefficiency exists in the rice production
systems and we found farmers‟ experience and extension visits negatively affected
technical inefficiency whereas factors such as access to microfinance, regional dummy,
off-farm income, age, education, and land fragmentation positively affected technical
inefficiency. In particular policies leading to granting access to microfinance, raising the
educational level of farmers, ensuring land preservation for agricultural purposes, and
ensuring sufficient returns to the farmers could be beneficial for reducing inefficiency in
rice production in Bangladesh. The findings of the relationship between microfinance and
technical efficiency suggest that improving greater access of farmers to agricultural
microfinance will improve production efficiency. Consequently, streamlining the
microfinance to the credit constraints farmers would be vital factor in increasing farm
technical efficiency and revenue. However, this is a multi-disciplinary work that needs to
be addressed out more rigorously by the government policy makers in collaboration with
Non Government Organizations (NGOs) and the donor agencies.
To improve farmers‟ access to microfinance, at first policies geared towards
addressing the features of agricultural microfinance products are vital. These policies
should include substantial modifications to conventional operational methodologies of
agricultural microfinance and should take into account the seasonality of crops and farm
incomes. Such modifications should match the heterogeneous households‟ demands and
enable farmers to afford credit by devising flexible repayment schedules. Second,
effective linkages between the rural MFIs with liquidity constraints and mainstream
banks with excess liquidity may minimize the demand-supply gap and ensure greater
access to microfinance for those farmers that are largely excluded or untapped by the
MFIs. Third, the establishment of „poor-friendly‟ microfinance banks to improve the
access of farmers to finance without collateral and at reasonable cost is suggested. The
delivery of such tailor-made agricultural microfinance that is backed up by direct support
from the government through regulatory framework and institutional innovations, would
improve farmers‟ access to microfinance. This in turn, may lead to more efficient
allocation of resources and increased production through improved efficiency.
Policies leading to the improvement of farm education and land holding will be
favorable for improving the technical efficiency of farmers. More investments in
education in rural areas through private and public partnerships, initiating programs to
encourage those at school-going age and „food for education‟ programs may be harnessed
as a central ingredient in the development strategies. Moreover, the farmer field schools
(FFS) program, promoted by different development agencies including the World Bank,
may be rigorously implemented and practiced. This would help farmers develop their
„learning by doing‟ practices and improve their analytical and decision making skills that
contribute to adapting to improved farming technologies. Initiating well-designed adult
literacy programs that have direct impact on household production could also contribute
to ensuring basic literacy and numeracy skills for the farmers.
The land fragmentation problems in Bangladesh should be directed through
addressing the law of inheritance of parental property, developing the land market and
tracing the causes of such fragmentation. The broad policy and legal measures that may
be devised should include, inter alia, revising the laws of inheritance and land tenancy
91
and motivating the small and marginal farmers to consolidate their lands through creating
viable farms. Enlargement of farms through forming cooperatives, encouraging voluntary
exchange of plots to form larger unitary plots, motivating farmers to buy and enlarge
contiguous plots by selling discrete distant plots, passing legislation that supports such
consolidation and formulating national land use policy will restrict land fragmentation
and are recommended.
These measures, if addressed in national agricultural policy formulation, may
direct the farmers‟ production frontier upward in the long run, which may in turn, reduce
technical inefficiency on the one hand and lead to food security through increased
production on the other hand.
ENDNOTES
1Aus and Aman are local breed crops and typically known as traditional rice crop.
2Boro is the irrigated rice crop and typically known as High Yielding Variety rice.
3USD 1= Taka 69.15 approximately; Euro 1=Taka 93.52 (as of April 13, 2010).
4The likelihood-ratio test statistic, -2{ln[likelihood (H0)]-ln[likelihood(H1)]}, has
approximately chi-square distribution with parameter equal to the number of parameters assumed
to zero in the null hypothesis, (H0), provided.
5 is not equal to the ratio of variance of inefficiency to total residual variance. This is because
the variance of ui )(2
u is equal to [2]/)2( not
2 . The relative contribution of the
inefficiency effect of the total variance term )]2/)1(/[* (Coelli et al.,
1998).
6Age and experience are generally interrelated but their impacts on technical inefficiency are not
necessarily identical. In this analysis, the coefficient of age is positive while that of experience is
negative. This finding is in line with Coelli et al. (2002) and Bozoglu and Chehan (2007), who
found experience to be a better predictor of technical inefficiency than age for farm household
production efficiency.
7 The critical of t358 (0.05) is 1.96 and the t-test statistic is 3.56 and thereby suggesting significant
differences in averages of technical efficiency between microfinance borrowers and non-
borrowers of microfinance.
8 The results are not reported here but available on request from the authors.
92
REFERENCES
Abdulai, A. and Eberlin, R., “Technical Efficiency During Economic Reform in
Nicaragua: Evidence From Farm Household Survey Data”, Econom. Syst., 2001, Vol. 25,
No.2, pp.113-125.
Aigner, D., Lovell, K.C.A. and Schmidt, P., “Formulation and Estimation of
Stochastic Frontier Production Function Models”, Journal of Econometrics, 1977,Vol.6,
No.1, pp. 21-37.
Ahmed, R., Haggblade, S. and Chowdhury, T.E., “In Out of the Shadow of
Famine: Evolving Food Markets and Food Policy in Bangladesh”, 2000, pp. 1–20, Johns
Hopkins University Press.
Asadullah, M.N. and Rahman, S., “Farm Productivity and Efficiency in Rural
Bangladesh: The Role of Education Revisited”, Applied Economics, 2009, Vol. 41, No.1,
pp.17-33.
Atkinson, S. E. and Cornwell, C., “Estimation of Output and Input Technical
Efficiency Using a Flexible Functional Form and Panel Data”, International Economic
Review, 1994, Vol.35, No.1, pp. 245-255.
Bangladesh Agricultural Census, 2008. Dhaka: BBS
Bangladesh Bureau of Statistics, (2006), The Yearbook of Statistics, 2006. Dhaka: BBS.
Bangladesh Bureau of Statistics, (2008), The Yearbook of Statistics, 2008. Dhaka: BBS.
Bangladesh Economic Review, (2009), Ministry of Finance and Planning, The
Government of Bangladesh.
Bangladesh Rice Research Institute, (2000), Annual Report. Gaziupur: BRRI.
Battese, G.E. and Corra, G.S., “Estimation of a Production Frontier Model: With
Application to the Pastoral Zone of eastern Australia”, Australian Journal of
Agricultural Economics, 1977, Vol.21, No.3, pp.169-179.
Battese, G.E. and Coelli, T.J., “Frontier Production Functions, Technical
Efficiency and Panel Data with Application to Rice Farmers in India”, The Journal of
Productivity Analysis, 1992, Vol. 3, Nos.1-2, pp.153-169.
Battese, G.E. and Coelli T.J., “A Model for Technical Inefficiency Effects in
Stochastic Frontier Production Function for Panel Data”, Empirical Economics, 1995,
Vol.20, No.2, pp.325-332.
Battese, G.E., Malik, S.J. and Gill, M.A.,“An Investigation of Technical
Inefficiencies of Production of Wheat Farmers in Four Districts of Pakistan”, Journal of
Agricultural Economics, 1996, Vol. 47, No.1-4, pp. 37-49.
Binam, J.N., Toyne, J., Wandji, N., Nyambi, G. and Akoa, M., “Factors
Affecting the Technical Efficiency Among Smallholder Farmers in the Slah and Burn
agricultural Zone of Cameroon”, Food Policy, 2004,Vol. 29, No.5,pp. 531-545.
Bozoglu, M. and Ceyhan, V., “Measuring the Technical Efficiency and
Exploring the Inefficiency Determinants of Vegetable Farms in Samsun Province,
Turkey”, Agricultural Systems, 2007, Vol.94, No.3, pp.649-656.
Bravo-Ureta, B.E. and Pinheiro, A.E.,“Technical, Economic and Allocative
Efficiency in Peasant Farming: Evidence from the Dominican republic”, The Developing
Economies, 1997, Vol.35, No.1, pp. 48–67.
Chu, S.F., Aigner, D.J. and Frankel, M., “On the Quadratic Law of Production”,
Southern Econ, 1970,Vol.37, No.1, pp.32-39.
93
Coelli, T.J., “A guide to Frontier Version 4.1: A computer Program for Frontier
Production Function Estimation”, CEPA Working Papers, 1996, No. 7/96, ISBN 1 86389
4950, Department of Econometrics, University of New England, Armidale, pp.33.
Coelli, T.J. and Battese, G.E., “Identification of Factors which Influence the
Technical Efficiency of Indian Farmers”, Australian Journal of Agricultural Economics,
1996, Vol. 40, No.2, pp.103-128.
Coelli, T. J., Prasada, R.D.S. and Battese, G.E., “Introduction to Efficiency and
Productivity Analysis”,1998, Kluver Academic Publisher.
Coelli, T.J., Rahman, S. and Thirtle, C., “Technical, Allocative, Cost and Scale
Efficiencies in Bangladesh Rice Cultivation: A Non-parametric Approach”, Journal of
Agricultural Economics”, 2002,Vol.53, No.3, pp.607-626.
De Datta, S.K., Gomez, K., Herdt, R.W. and Barker, R., “A Handbook on the
Methodology for an Integrated Experiment on Rice Yield Constraints”, 1978, Los Ban˜os.
Farrell, M.J., “The measurement of Production Efficiency”, Journal of Royal
Statistical Society, 1957 Series A,Vol.120, No.3, pp.253–290.
Greene, W., NLOGIT Version 4.0, Economic Software Inc, 2007.
Greene, W.H., “On the Estimation of a Flexible Frontier Production Model”,
Journal of Econometrics, 1980, Vol.13, No.1, pp.101-115.
Huang, CJ. and Liu JT., “Estimation of a Non-Neutral Stochastic Frontier
Production Function”, Journal of productivity Analysis, 1994, Vol.5, No.2, pp.171-180.
Huffman, W.E., “Human Capital: Education and Agriculture”. In Gardener,
G.L., Rausser, G.C., (Eds.), Handbook of Agricultural Economics, 2001, Vol. A., New
York, NY: North Holland, pp. 334-381.
Kalirajan, K. and Flinn, J.C., “The Measurement of Farm-Specific Technical
Efficiency”, Pakistan Journal of Applied Economics, 1983, Vol.2, No.2, pp.167-80.
Kalirajan, K., “Farm Specific Technical Efficiencies and Development Policies”,
Journal of Economic Studies, 1984, Vol.11, No.3, pp. 3-13.
Khan, A., Huda, F.A. and Alam, A.., “Farm Household Technical Efficiency: A
Study on Rice Producers in Selected Areas of Jamalpur District in Bangladesh”,
European Journal of Social Sciences, 2010, Vol.14 (2), pp. 262-271
Kodde, D.A. and Palm, F.C., “Wald Criteria for Jointly Testing Equality and
Inequality Restrictions”, Econometrica, 1986, Vol.54, No.5, pp. 1243-1248.
Kumbhakar, S.C., “The Specifications of Technical and Allocative Inefficiency
in Stochastic Production and Profit Function”, Journal of Econometrics, 1987, Vol.34,
No.3, pp.335-348.
Kumbhakar, S.C., Ghosh, S. and McGuckin, J.T., “A Generalized Production
Frontier Approach for Estimating Determinants of Inefficiency in U.S Dairy Farms”,
Journal of Business and Economic Statistics, 1991, Vol.9, No. 3, pp.279-286.
Meeusen, W. and van den Broeck, J., “Efficiency Estimation from Cobb-
Douglas Production with Composed Error”, International Economic Review, 1977,
Vol.18, No.2, pp. 435-444.
Nyemeck, B.J., Sylla, K., Diarra, I. and Nyambi, G., “Factors Affecting
Technical Efficiency Among Coffee Farmers in Coˆ te d_Ivoire: An Evidence of Centre
West Region”, African Development Review, 2003, Vol.15, No.1, pp. 66–76.
94
Pingali, P.L., Hossain, M. and Gerpacio, R.V., “The Asian Rice Bowls: The
Returning Crisis”?, 1997, CAB International.
Pitt, M.M. and Lee, L.F., “Measurement and Sources of Technical Inefficiency
in the Indonesian Weaving Industry”, Journal of Development of Economics, 1981, Vol.9,
No.1, pp.43-64.
Rahman, S. and Rahman, M., “Impact of Land Fragmentation and Resource
Ownership on Productivity and Efficiency: The Case of Rice Producers in Bangladesh”,
Land Use Policy, 2009, Vol.26, No.1, pp. 95-103
Rahman, S., “Profit Efficiency Among Bangladeshi Rice Farmers”, Food Policy,
2003, Vol.28, Nos.5-6, pp.483-503.
Sattar, A.,” Bridging the Rice Yield Gap in Bangladesh. In Bridging the Rice
Yield Gap in Asia and the Pacific”,2000, RAP Publication 2000/16.
Sharif, N.R. and Dar, A. A., “An Empirical Study of the Patterns and Sources of
Technical Inefficiency in Traditional and HYV Rice Cultivation in Bangladesh”, Journal
of Development Studies, 1996,Vol.32, No.4, pp. 612 -629.
Sharma, K.R., Leung, P. and Zalleski, H.M., “Technical, Allocative, and
Economic Efficiencies in Swine Production in Hawaii: A Comparison of Parametric and
Non-parametric Approaches”, Agricultural Economics, 1999, Vol.20, No.1, pp. 23–35.
Stevenson, R.F., “Likelihood Functions for Generalized Stochastic Frontier
Estimation, Journal of Econometrics”, 1980, Vol.13, No.1, pp. 57-66.
Tzouvelekas, V., Pantzios, C.J. and Fotopoulos, C., “Technical Efficiency of
Alternative Farming Systems: The Case of Greek Organic and Conventional Olive-
Growing Farms”, Food Policy, 2001, Vol.26, No.6, pp. 549–569.
Villano, A. R., “Technical Efficiency of Rainfed Rice Farms in the Philippines:
A Stochastic Frontier Production Function Approach”, 2005, Working Paper, School of
Economics, University of New England, Armidale, NSW, 2351.
Wadud, A., “Farm Efficiency in Bangladesh. Doctoral Thesis”, 1999,
Department of Agricultural Economics and Food Marketing, University of Newcastle
upon Tyne, UK.
Wadud, A. and White, B., “Farm Household Efficiency in Bangladesh: A
Comparison of Stochastic Frontier and DEA Models”, Applied Economics, 2000, 32,
1665-73.
Wadud, A.,“Technical, Allocative, and Economic Efficiency of Farms in
Bangladesh: A Stochastic Frontier and DEA Approach”, The Journal of Developing
Areas, 2003, Vol.37, No.1, pp.109-126.
Wang, J., Cramer, G. L. and Wailes, E. J., “Production efficiency of Chinese
Agriculture: Evidence from Rural Household Survey Data”, Agricultural Economics,
1996b,Vol. 15, No.1, pp.17-28.
Wang, H. and Schmidt, P., “One-Stage and Two-Stage Estimation Effects of
Exogenous Variables on Technical Efficiency Levels, Journal of Productivity Analysis,
2002, Vol.18, No.2, 129-144.
Wozniak, G.D., “Joint Information Acquisition and new Technology Adoption:
Later Versus Early Adoption”, Review of Economics and Statistics, 1993, Vol.75, No.3,
pp. 438-445.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.