Technical Efficiency and Production Risk of Maize Production: Evidence from Ghana

This paper analyzed maize production efficiency in Ghana due to differences in efficiency. The stochastic frontier model with flexible risk properties is applied with 232 farms from the Brong-Ahafo Region. Findings of the study were the translog model best fits the mean output function, whilst the input variables: seed, herbicide, land, labour and cost of intermediate inputs influenced maize output at decreasing returns to scale. The study also found seed and labor inputs reduced production risk, whilst land and cost of intermediate inputs increased the risk. The average technical efficiency estimate was 62% and the combined farm specific factors explained the variation in technical efficiency. This study concludes, on the average 38% of potential output is lost due to technical inefficiency and production risk in inputs and the use of the best farm practices produce maize efficiently.


INTRODUCTION
World production of maize amounted to 875, 226,630 tons in 2012 mainly from United States, China and Brazil [1]. Africa contributes small fraction of the total supply [2]. Maize consumption per capita is highest between 52 to 328 g/person/day as a staple in Africa. Ghana's per capita consumption of mainly white maize, increased from 38.4 kg in 1980 to 43.8 kilograms in 2011 [3]. The current average yield of maize in Ghana is estimated to be 1.9 t/ha [2] against achievable yields of 6t/ha. Similarly maize yields for Burkina Faso, Togo, Cote D'Ivoire are 1.59t/ha, 1.19t/ha and 2.06 t/ha and this have been very erratic over the years [2]. The worst yields are decreasing in Kenya, Morocco and Rwanda whilst population is growing meanwhile the crop constitute about 5-51% of calorie intake. But in Asia and other parts of Africa yields are consistently increasing that is Ethiopia, Angola, and South Africa. Thus, in some parts of Africa deviation of observed maize yields from the achievable yield is worst due to constraints from poor physical structures, weather, pest and disease incidence and socio economic characteristics of the farmers. Consequently, the supply of maize is not enough to meet its higher demand from growing population [4].
Technical efficiency analysis is of paramount importance to increase maize productivity and contribute to the attainment of food security and income generation. In addition, production risk in inputs influences the production structure and subsequently the technical efficiency estimates [5,6,7,8,9]. However, the conventional stochastic frontier model neglects the role of the inputs towards risk. A comprehensive analysis of production risk in input and technical efficiency of maize production has not been properly addressed in Ghana. Such studies could contribute to policy formulation on maize production [10,11,12,13,14]. The study assessed technical efficiency and production risk of selected maize farms in Brong-Ahafo Region of Ghana.

Study Area
The study was based on farm level data on maize production in the Brong-Ahafo region of Ghana. Maize is grown in two seasons but mostly cultivated in the first season with the onset of rains. Major season cultivation usually starts from March to June and a short dry-spell which occurs in July provides suitable conditions for harvesting and sun-drying. The minor season follows in August till November. Nkoranza, Kintampo North and South, Wenchi Districts as part of the study area are found in the transition zone of Ghana whereas Sunyani West and Berekum Districts are located in the semideciduous forest zone.

Theoretical Framework
The method of analysis proposed for this study is consistent with the stochastic frontier approach which was independently proposed by [15] and Meeusen and [16]. However, this model proposes inputs have similar effect on mean and variance outputs. But, [17], production function proposed separate effects of the inputs on the mean and variance outputs whilst [5] further incorporates technical inefficiency model.
Following [5] the production process is represented below as; The technical efficiency of the i-th farm is given by equation (3) which is consistent with [5] specification of technical efficiency. (1) ; , : 1 , 0 : : The variance of output or production risk is given by, The marginal effect of the input variables on the production risk is given as;

Empirical Model Specification
The empirical application of this study is consistent with models developed by [5,15,16,17] Deterministic part of the production frontier in equation (1) assumed a translog model in equation (8). Cobb-Douglas model specified as: The error term is specified as; the frontier output elasticity. Man days for labour have been calculated with the formula in line with [18] and [19]. One adult male working for 8 hours equals one man day; one female and one child (< 18years) working for 8 hours equals 0.75 and 0.5 man days respectively.
The linear production risk function is specified as; (10) Where ' m x s represent the input variables, as described in Table 1. The technical inefficiency effects were given by;

Statement of Hypothesis
The following hypotheses were considered for

Data and Sampling Technique
This study used cross sectional data from 232 maize farms, which is a fair representation of the maize farms in the region. Multi-stage sampling procedure was employed for the farm survey to obtain the data on the relevant variables for the study including output and input variables as well as the farm specific variables. Within each district three major communities with varying intensity of maize production were selected from which the maize farm households are selected randomly. The farmers are distributed within the districts as 50, 50, 47, 39 and 46 for Sunyani West, Nkoranza South, Kintampo North and South Wenchi and Berekum districts respectively which occur in the transition and semi-deciduous zones as soil and weather characteristics are favorable for optimum maize production.

Summary Statistics of the Output and the Input Variables
The study demonstrated that output range between (337.5 -6750) kg/ha at the mean of 1957.506 kg/ha with standard deviation of 1027.74 kg/ha (Table 1). Maize producers obtained yields of 3.3-6tons/ha of which the production technology becomes fairly represented for the region. The average yield of 1957 kg/ha of maize implies significant number of farmers obtain yields below the maximum yield per hectare but considering all the inputs in the production process the frontier output is not known thus, this study seeks to estimate the determinants of technical efficiency.

Testing of Hypothesis
The translog model is an adequate representation of the data, given its specification. Production risk in inputs and technical inefficiency are present and the estimated lambda is 1.7. Thus the variations in output due to technical inefficiency are relatively larger than the deviations in output from pure noise component of the composed error term. The study finds technical inefficiencies are explained by exogenous variables ( Table 2).

Frontier Estimates
The effects of the inputs conform to expectation on output. Output is mainly contributed by cost of intermediate input and seed. Additionally, land contributes technical efficiency gains as found in [21]. At the scale elasticity of 0.8%, output does not respond proportionally to input change. But, [12] results of maize production in Northern Ghana have indicated an increasing return to scale (Tables 4 and 5).

Production Risk
Production risk in inputs is significant in the production process with the exception of herbicide. Seed and labor reduce risk because seed as an input factor has the favorable characteristics to support its growth into maturity.
This contradicts with what [22] found in which seed was a risk increasing-input in rice production. Labour performs the best farm practices to support the farmer to achieve the expected output as way of reducing risk in the production process. This result is consistent with the findings of [7,8,22]. Risk averse farmers in pursuit of reducing their risk are expected to use more of seed and labour to better their situation which can alter the technical efficiency score.
Land and cost of intermediate inputs are positively related to production risk. Land might increase the risk of exposure of the crops from unfavorable weather conditions especially during the dry season. [9] study reveals greater area cultivated lead to increased output variability, possibly suggesting that larger farms are less able to react quickly to unfavourable weather conditions at harvest or planting times. On the other hand, [22] found land to be a risk-reducing input because the rice farmers had parceled their land into plots such that losses from one plot are compensated by gains in another due to differences of weather at the different plots.

Technical Efficiency Estimates
Maize production in the region is not technically efficient. The lowest efficiency score is 8% which is incomparable to the highest at 99%. On the average the farmers produce about 62% of the frontier output. Quite significant number of farmers obtains relatively higher efficiency scores (Fig. 1). The results might be similar to other areas of Ghana [10,12]. The study found that farmers in Nkoranza have a higher rate of technical efficiency due to their ability to apply the best farm practices more efficiently.

Determinants of Technical Inefficiency
Farm size reduces inefficiency in the production process. The reason might be such farmers adopted the best farm practices so as to achieve the frontier output [24,25]. [26] study indicated that soil conservation practices result to higher levels of technical efficiency among farmers but ploughing affected technical inefficiency positively. Location has been an important factor to determine efficiency because the level of efficiency at Sunyani West is significantly lower than the other districts. Similarly the efficiency of cocoa production varied by regions of Ghana [27] as well as rice production in South Korea [28].

Risk and Technical Efficiency
Technical efficiency estimates for the maize farms when production risk component is excluded ranged from 13% to 97%, with a

Determinants of Technical
reduces inefficiency in the production process. The reason might be such farmers adopted the best farm practices so as to achieve [24,25]. [26] study indicated that soil conservation practices result to higher ency among farmers but ploughing affected technical inefficiency positively. Location has been an important factor to determine efficiency because the level of efficiency at Sunyani West is significantly lower rly the efficiency of cocoa production varied by regions of Ghana [27] as well as rice production in South Korea [28].

Risk and Technical Efficiency
Technical efficiency estimates for the maize farms when production risk component is excluded ranged from 13% to 97%, with a sample mean of 76%. However, when the stochastic frontier model with flexible risk properties was considered, the technical efficiency estimates ranged from 8% to 99% with a mean of 62%, which is significantly differ from the 76%. Thus the technical efficiency estimates may be compromised when the production technology of the maize farms in the study area is modeled without the flexible risk component [5,6,8,9].

CONCLUSIONS AND POLICY RECOMMENDATIONS
This study has estimated stochastic frontier model with flexible risk properties. It revealed the input factors determined maize output as well as production risk. On average, maize production in the region has been technically inefficient and is dependent upon the application of best farm practices. It further predicted technical efficiency to reveal that technical efficiency estimates may 12  sample mean of 76%. However, when the stochastic frontier model with flexible risk properties was considered, the technical efficiency estimates ranged from 8% to 99% with a mean of 62%, which is significantly different technical efficiency estimates may be compromised when the production technology of the maize farms in the study area is modeled without the flexible risk

CONCLUSIONS AND POLICY
This study has estimated stochastic frontier model with flexible risk properties. It revealed the input factors determined maize output as well as production risk. On average, maize production in the region has been technically inefficient and is upon the application of best farm practices. It further predicted technical efficiency that technical efficiency estimates may 33 91-100 be compromised when the production technology is modeled without the flexible risk component. Farmers consider their land sizes before applying best farm practices. The study recommends policy to promote the application of best farm practices on small land holdings as well as bridging the gap in district level efficiency. Again efficient methods of ploughing to suit local conditions are recommended. Lastly, it is appropriate to incorporate production risk in technical efficiency analysis if the inputs are nonneutral in risk.