WHEAT IN BANGLADESH: YIELD GROWTH, PRODUCTION PERFORMANCE AND DETERMINANTS - Soybean and Wheat Crops: Growth, Fertilization, and Yield - page 213

in this study includes truly exogenous (e.g., weather), quasi-fixed characteristics (e.g., soil
types and land types) as well as combinations of exogenous shocks and managerial response
(e.g., pest and weed infestation). Both Sherlund et al., (2002) and Rahman and Hasan (2008)
showed that the production inputs were found to be strongly correlated with the variables that
condition the production environment of the farmers.
Parameter estimates of the restricted Translog stochastic production frontier are reported
in Table 3 using the Maximum Likelihood Estimation (MLE) procedure in STATA Version 8
(STATA Corp, 2003). A series of hypothesis tests using Likelihood Ratio (LR) test statistic
were conducted regarding the model choice, inclusion of environmental variables and
determinants of inefficiency, the results of which are presented in Table 4. The first test of
hypothesis is the choice of the functional form, i.e., Cobb-Douglas vs Translog functional
form. The result indicates that non-linearities in the production function is present and, hence,
the choice of flexible Translog functional form is a better representation of the true
production structure as compared to a more restricted Cobb-Douglas form. The next test of
hypotheses that 'the environmental variables are jointly zero' is also strongly rejected
indicating that environmental production conditions significantly affect productivity. The test
of γ is also strongly rejected suggesting presence of technical inefficiency, and hence the
application of stochastic production frontier model is justified (Table 4).
All five input variables significantly influence wheat pr oductivity as expected. The input
variables were mean corrected prior to estimation

X − . Therefore, the coefficients on
the first order terms of the input variables can be read directly as elasticities. Land is the most
dominant input followed by fertilizers, labour, animal/mechanical power services and
irrigation. The hypothesis of constant returns to scale in wheat production is rejected in
favour of decreasing returns to scale (Table 5) implying that farmers are not operating at the
optimal scale.
Poor land types, delay in sowing and poor soil quality significantly reduces productivity,
variables that are typically omitted in most studies. Since variables representing
environmental production conditions were incorporated in the model, the responsiveness of
the key production inputs on wheat productivity is likely to be more accurate. Both Sherlund
et al., (2002) and Rahman and Hasan (2008) reported positive response of output and fall in
inputs of labour and/or fertilizers when controlled for environmental production conditions.
Geography does matter in wheat production performance. Wheat production is significantly
lower in Jamalpur regions, although it is an intensive agricultural region but not a typical
wheat growing region as compared to Rajshahi and Dinajpur.

( X )

ij j

Production Efficiency

The mean technical efficiency level in wheat production is estimated at 83% which
implies that production can be increased by 20% [{(0.83-1.00)/0.83}*100] with efficiency
improvements. The minimum score is 66% and the maximum is 99%. The mean estimate is
comparable to estimates for other developing countries. For example, technical efficiency in
wheat production varies between 57.0-78.9% in Pakistan (Battese et al., 1996), 81.0-93.4%
in India (Singh, et al., 2004) and 91.0-93.0% in Iran (Bakhsoodeh and Thomson, 2001),
respectively.