Abstract
The present study makes an attempt to analyse farm level cost inefficiency of maize farming and its determinants in different agro-climatic regions of Sikkim. The primary data for the study were collected during the third and fourth quarter of 2018 from different agro-climatic regions of Sikkim. Both data envelopment and stochastic frontier analysis were used for measurement of the farm level inefficiency across different agro-climatic regions of the study area. Based on the Cobb–Douglas cost function for maize output, the article simultaneously estimated stochastic frontier cost function and examined the effect of exogenous factors on farm level cost inefficiency. The results of this study showed that, on an average, the farmer incurred cost which was 8 per cent to 72 per cent above the minimum cost defined by the best practice frontier. Further, cost inefficiency was relatively higher among the farmers in temperate agro-climatic region. Greater cost inefficiency seems to be directly associated with remoteness of farmland from input market. The study also found that the additional years of farming experience and farming in the rented plots were useful in reducing cost inefficiency.
Keywords
Introduction
Agriculture is an important sector in Sikkim’s economy with nearly 64 per cent of the population dependent on agriculture as a source of livelihood (IBEF 2018). During 2017–2018, the contribution of agriculture, forestry and fishery to the gross state domestic product (GSDP) of Sikkim was 10.33 per cent at current price (GoS 2019). Maize is the dominant crop of Sikkim sharing 28.61 per cent of gross cropped area 1 of the state. In 2015–2016, area under maize farming was 38,960 ha and produced 68,310 tons of maize in the state (GoS 2017). Besides an assurance of food security among the farming community, the cultivation of maize serves as a staple food for the rural population, feed and fodder for animals in Sikkim (Basnet et al. 2003; Borah et al. 2012). The productivity of maize has remained low in Sikkim with an average of 1,660.18 kg/ha, against the national figure of 3,800.39 kg/ha during 2003–2004 to 2015–2016 (GoI 2019; GoS 2017). A study by Guha and Gosh (2017) found that the productivity of maize remained fairly constant across altitudes of Sikkim and was lower than the productivity of Darjeeling hills. Sikkim being not self-sufficient in food production given the limited availability of cultivable land and lower level of productivity (Subba 2009), there is a need for improving the output of maize. Improved productivity of maize is pivotal to meet the increasing demand for food, and feed and fodder for animals in Sikkim. Appropriate combination of inputs may assist the farm in attaining optimal output level at least cost, which may help in addressing the issue of food security. Therefore, it is imperative to investigate factors obstructing efficiency of farmers. Improved understanding of cost inefficiency and its determinants can greatly assist the policymakers for implementing suitable policies for farm efficiency.
The ability of a farming system to obtain maximum output from a certain number of inputs represents production cost efficiency. Thus, the farmers’ ability in using inputs to produce certain level of output at a certain level of technology is cost efficiency. Hence, the ability of a firm to produce a given level output using cost minimising input prices is cost efficiency otherwise referred as allocative efficiency (Coelli et al. 2005).

In Figure 1, if the distance from origin till R is considered as r1 and the distance from origin till Q is considered as r2 then CE in polar form is defined as
Since the pioneering work of Aigner et al. (1977) and Meeusen and Broeck (1977), the study of efficiency of production and its measurement has been a popular area of research. The literature on efficiency in agriculture is spread over various dimensions. For example, Pradhan and Mukharjee (2018) stressed on improvement in farmer’s education and irrigation facility for minimising technical inefficiency of Indian agriculture. Zamani et al. (2019) identified that the mean efficiency of cooperative farming was higher than non-cooperative farming in Azarbaijan. While measuring the technical efficiency of maize cultivation in Nigeria, Ajao et al. (2005) established that the efficiency of mechanised farm was higher than non-mechanised farm. The increased efficiency of commercial maize production in South Africa was driven by lower levels of intermediate input use (Bach et al. 1998). Seyoum et al. (1998) argued that mean technical efficiency of maize farmers within the Sasakawa-Global 2000 project was better than farmers outside the project in eastern Ethiopia. Aye and Mungatana (2012) pointed that efficiency gain of Nigerian maize farming was considerably influenced by seed, inorganic fertiliser, size of holding, education, access to extension service and credit. It was argued that the small maize farmers were technically more efficient in Zimbabwe (Etienne et al. 2019). Again, taking account of cost efficiency in agriculture, existing studies have come up with mixed findings in the case of farm size. Small farmers of Pakistan’s agriculture sector were more efficient than large farmers (Parikh et al. 1995), while reverse statement about agricultural efficiency was made by Paul et al. (2004) in USA; Rungsuriyawiboon and Hockmann (2015) in Poland; Asogwa et al. (2011) in Nigeria; and Ali et al. (1996) in Pakistan. In recent years, studies on cost efficiency in agriculture include a variety of crops such as paddy, maize and other crops. Taking account of paddy farming, it was found that the cost inefficiency of Bangladesh and South Korea was 0.44 (Coelli et al. 2002; Nguyen et al. 2012). Wadud (2003) traced that the land fragmentation, inadequate irrigation infrastructure and environmental factors were responsible for the cost inefficiency among paddy farmers in Bangladesh. Evidence suggested the greater cost inefficiency of wet season paddy farming households to that of dry season farming households in Cambodia (Rido 2014). Tang et al. (2018) favoured agricultural services, viz. machinery and financial facilities, for recovering cost efficiency of paddy farming in China. Other studies on cost efficiency include the tobacco, onion and dairy farming. The cost inefficiency was observed to be lower among larger farms compare with smaller farms in tobacco plantations of Malwai (Hazarika and Alwang 2003) and dairy farms in the USA (Tauer and Misra 2006). Gebremariam et al. (2019) outlined that the efficiency of onion farming in Ethiopia was gender sensitive with male outperformed the female headed farms. With special reference to cost efficiency of maize farming, the scope for minimising cost inefficiency in eastern Ethiopia was reported by Ahmed et al. (2018), as the output was largely influenced by allocative efficiency. Contrasting argument was observed by Ogundari et al. (2006) and Abdulai et al. (2017) while studying cost efficiency of maize farming in Nigeria and northern Ghana, respectively. However, few studies mention the requisite for minimising cost inefficiency. It was noted that, production environment, soil suitability and stability of mean temperature were required for minimising cost inefficiency of Bangladesh’s maize farming (Rahaman et al. 2002). The cost efficiency of maize farming in Nepal was largely reliant on the provision of farmer’s education and younger generation’s participation in farming (Paudel and Matsuoka 2009). Abdulai and Abudulai (2016) suggested the improvement of road and transport services for cost efficiency of maize farming in Zambia. A detailed understanding of productivity and efficiency of different agricultural crops were published in various government and micro level studies. However, an attempt to evaluate the cost inefficiency of maize farming in the Eastern Himalayan state of Sikkim is conspicuous by its absence. Taking into consideration the aforementioned facts, the present study is conducted for evaluating the cost inefficiency of maize farming and its determinants in different agro-climatic regions of Sikkim. The rest of the article is organised as follows. The second section explains the study area, data and sample of the article. The third section describes the methodology of the article. The empirical analysis and results are presented in the fourth section. Conclusions and policy implications are covered in the fifth section.
Study Area, Data and the Sample
The present study was based on primary data collected from the maize producing farm households by using a multistage sampling technique during July–November 2018. Since the sowing and harvesting period of maize in Sikkim ranges between February and July or April and September, therefore, the present study used the output and input data of the previous farming season (February–September 2017). In the first stage, among five agro-climatic regions of Sikkim, three were selected, viz. tropical, sub-tropical and temperate for their suitability 2 in maize farming. In the second stage, out of four districts, two were selected, viz. East and South, 3 for their importance in maize farming in the state. The selected districts fall under all the three agro-climatic zones as mentioned earlier (Figure 2). In the third stage, three development blocks were selected randomly from each sample district. They were Pakyong, Rhenock and Khamdong from East district, and Namchi, Ravangla and Temi Tarku from South district. In the fourth stage, two villages were randomly selected from each block. In the fifth stage, 6–11 per cent of farm households from each village were selected randomly for the primary survey. Therefore, a total of 200 farm households were interviewed by using a pre-tested question schedule. The survey collected data on output quantities of maize, cost of inputs used in its cultivation, background and different characteristics of the sampled farm households, enabling factor such as distance to input market, and locational characteristics like agro-climatic parameters.
The descriptive statistics of variables are listed in Table 1. As shown in Table 1, the average sample farm incurred ₹71,169.46/ha as total cost for producing 1,007.57 kg/ha of maize output. The cost of production seems to be similar across the farms as evident from the low value of standard deviation. About 34.63 per cent of the total cost of production was shared by hired labour. For an average farm, the amount of expenditure on organic fertiliser was ₹17,980.25/ha, accounting for 25.26 per cent of the total cost of production. About 3.66 per cent of the total cost of production was spent on bullock for land preparation. The average expenditure of leased in farmer was ₹10,604.64/ha of land in payment of rent, which was 14.90 per cent of the total cost of production. For the sampled farm of the present study, the expenditure on farm machinery accounted for 4.55 per cent of the total cost of production. The expenditure on seed accounted for 2.56 per cent of the total cost of production as significant section of farmers used their own seeds for farming practices, out of previous harvest stock.

Descriptive Statistics
As Table 1 shows, the average years of schooling of farmers was 5 years. The average age of farmers in the study area was 52 years, which implies that they attained the age of middle adulthood. The average distance of the market from the farmland across the sampled households of the study area was 15.47 km. Nearly 81 per cent of the farmers in the study area initiated maize farming in their own land and rest being on leased form of tenurial arrangement. As regards to the differential exposure to agro-climatic condition, it was found that 34 and 33 per cent of the sampled farms were located in temperate and sub-tropical agro-climatic region, respectively. The extent of crop diversification 4 was observed to be low among the sampled farms with crop diversification index (CDI) value of 0.31 in temperate and sub-tropical agro-climatic region, while it was 0.28 in tropical agro-climatic region (Appendix 1).
Methodology and Model
The present study applied both non-parametric data envelopment analysis (DEA) and parametric stochastic frontier analysis (SFA) approaches for measurement of farm level efficiency across different agro-climatic regions of the study area.
The DEA method was initially formulated by Charnes et al. (1978) assuming constant returns to scale (CRS), which had an input orientation. Subsequently Banker et al. (1984) have considered alternative sets of assumptions, who proposed a variable returns to scale (VRS) model. Assume that the ith farm uses xi = [xki] of inputs k (k = 1, 2, …, 6) and produces a single output qi. The (k × n) input matrix denoted by X and the (1 × n) output vector is denoted by Y for all n farms, where n = 200. Following Charnes et al. (1978), the technical efficiency (TE) can be estimated by solving the optimisation problem based on the DEA model as follows:
where θ is a scalar and λ is a N × 1 vector of constants. The VRS DEA frontier can be formulated by including the convexity constraint Ω₹ λ = 1 in Equation (1), where Ω is (n × 1) vector of ones. The economic efficiency (EE) can be estimated by using the following cost minimisation DEA model:
where pi is the vector of input prices for ith farm; xi is cost minimising input vector, given the prices of the inputs and the levels of output yi. The EE is measured by the ratio of minimum cost to observed cost,
Inability to accommodate the data noise is the limitation of DEA technique, which is better handled by SFA. The SFA was independently put forward by Aigner et al. (1977) and Meeusen and Van den Broeck (1977), and it specifies a production function of cross-sectional data with an error term comprised of random effect and technical inefficiency component. The frontier production function is widely used for the efficiency measurement in agriculture due to its capacity to accommodate statistical noise, and its parametric specification of technology. The cost efficiency of the sampled maize farming households is tested using a stochastic frontier cost function, and inefficiency scores are obtained for them. The functional specification of cost frontier following Coelli, Rahman, and Thirtle (2005) used in the study is
where Ci is the observed cost of the ith farmer, Pni is the nth input price; Xmi is the mth output; and C (P1i, P2i, …, Pni; X1i, X2i, …, Xmi) gives minimum cost of producing outputs X1i, X2i, …, Xmi when a producer incurs prices P1i, P2i, …, Pni.
Following Battese and Coelli (1995), the study used stochastic frontier cost function with the behaviour inefficiency component. The model is implicitly expressed as:
where Ci is total cost of production; f is suitable functional form such as the Cobb–Dogulas; Pi is vector variable of input prices; Xi is value of output; δ is parameter to be estimated; vi is random disturbance cost due to the factor outside the scope of farmers with vi (vi ~ iidN (0, σv2)); ui is one sided disturbance to represent cost inefficiency with ui (ui ~ iidN+ (μi, σu2)), which is non-negative. vi represents random error like measurement errors, specification errors and random shocks, that are not under the control of a producer. ui represents the cost inefficiency of the ith farmer that results from managerial problems and co-ordination issues at work. The ui is assumed to follow a truncated normal distribution. Since inefficiencies normally assumed to increase cost so the two error terms are processed by positive sign.
The farm operating above the cost frontier is defined by the cost function. Assumption of AE leads the ui closer to the cost of technical inefficiency. Furthermore, cost efficiency (CE) of an individual farm is defined as:
CE = observed cost/minimum cost
where CE takes the value of 1 with cost efficient farm. The Cobb–Dogulas stochastic cost frontier applied in this study with empirical model being formulated as:
where, i = 1, 2, 3, …, 200 maize farming households; Ci stands for total production cost of maize for the ith farmer, since fixed cost of production is negligible in the short run, the study uses only variable cost of production per hectare as a proxy for total production cost; P1i stands for expenses on land incurred by the ith farm households, which is measured as the amount of money or equivalent paid as the rent for the use of land during the last maize farming season; P2i expenditure on seed incurred by the ith farm households; P3i is expenses on wage payment by the ith farm households, here wage calculations are done based on expenses on hired labour and imputed value of family labour; P4i is expenses incurred on use of bullock for ploughing the land by the ith farm households; P5i is expenses incurred on use of organic fertiliser by the ith farm households, here organic fertiliser includes bio-fertiliser such as castol cake, cow dung, green manure and chicken manure; P6i is for expenses incurred on use of machinery by the ith farm households, here expenses on machinery includes all items such as sickle, trowel, plough axe, hoes, rake, tractor, water pump, etc.; Xi is the quantity of maize produced by the ith farming households.
The self-duality nature of cost function urged the choice of Cobb–Dogulas functional form for analysis. With the focus of the study being to identify the factors determining the cost inefficiency of maize farming households in Sikkim, hence the model with mean of the pre-truncated inefficiency distribution (Huang and Liu 1994; Kumbhakar et al. 1991) being formulated as:
where Z1i is age of the ith farmer; Z2i is years of schooling (formal education) of the ith farmer; Z3i is years of maize farming experience of the ith farmer; Z4i is farm land distance of the ith farming households to input market; Z5i is extent of crop diversification by the ith farming households; Tenui is the Tenureship status of the ith farming households; and Relii is the religion of the ith farming households. The Temp stands for the temperate agro-climatic region (R1) of the study, while Sub-Trop represents the sub-tropical agro-climatic region (R2).
For consistently estimating the TE scores of every farmer and the marginal effects of exogenous factors on them, Equations (6) and (7) are simultaneously estimated using maximum likelihood method. This method is an improvement over the two-step’s method used in literature (Kalirajan and Shand 1985), as this method allows consistent estimation of the technical inefficiency terms (and parameters) even if they are correlated with the inputs and incorporates the non-positive nature of the inefficiency values.
In order for determining marginal effects of cost inefficiency, a linear regression model is not appropriate, as the dependent variable is bounded between 0 and 1. In such formulation of dependent variable, we see clusters of observations in both the ends, in which the dependent variable takes the value of 0 and 1. Hence, a Tobit model with censoring on both sides has been formulated. The model is formulated with the help of latent variable
The Qi is the vector of all explanatory variables as used in Equation (7). The observed dependent variable Yi (i.e., value of cost inefficiency score for the ith sample farm arrived from SFA, and DEA estimates assuming CRS and VRS in production) is linked to the latent variable
The random disturbance εi is assumed to be independently normally distributed with zero mean. The fitted model (8) has been estimated using the maximum likelihood method.
Empirical Analysis
Maximum Likelihood Estimates of the Stochastic Frontier Cost Function Model
However, the cost elasticity of production with respect to expenditure on seeds is found to be negatively significant. This may be because a significant section (89%) of farmers in the study area uses their own seed from the treasury of the previous harvest for cultivation of maize. The estimated cost elasticity of production with respect to expenditure on seeds is negative, which implies that 1 per cent increase in expenditure on seeds lead to a reduction in the cost of production by 0.19 per cent. Thus, increase expenditure on seeds help in reducing cost of production. Contrasting result was found by Aye and Mungatana (2012). An increase in payment for the use of land, labour and farm machinery by 1 per cent leads to increase in the cost of production by 0.15, 0.36 and 0.05 per cent, respectively. In addition, an increase in the level of maize output by 1 per cent leads to increase in cost of production by 0.43 per cent.
Before proceeding to the parameter estimate of the technical inefficiency model, it is important to evaluate the presence of technical inefficiency in the production model. The estimated likelihood ratio statistic is 36.75, substantially greater than
Generalised Likelihood Ratio Test of Hypothesis for Inefficiency Effect Model
The inverse coefficient of cost elasticities with respect to maize output gives the scale effect (SE) among the maize farms. The computed value of the SE is 2.32, which confirms that there is a positive economies of scale. Thus, during the course of maize production, 1 per cent increase in total production cost leads to 2.33 per cent increase in total maize output. Thus, irrespective of the area under maize production, an average maize farmer in the study area experiences a decrease in the total production cost. Hence, the sampled maize farmers experience increasing returns to scale in the stage I of the production surface. Since fixed resources are abundant relative to the variable resources, so stage I of the production can be traced as sub-optimal (Reddy et al. 2004).
Marginal Effects on Cost Inefficiency of Maize Farming
Frequency Distribution of Cost Efficiency Scores
The marginal effect of distance to input market under SFA is in confirmatory with the cost inefficiency estimates as shown in Table 2. Thus, an increase distance of farmland from input market is associated with higher levels of cost inefficiency among the sampled farms. The Buddhist farmers are observed to be more cost inefficient. The religious beliefs influence economic outcomes by affecting personal traits such as honesty, thrift, willingness to work hard and openness to strangers (Barro and McCleary 2003; Chen 2005). It was found that community influences agricultural performance, in which, Muslim farmers were found to be more efficient than non-Muslim farmers of Assam (Bhattacharyya and Mandal 2016). The coefficient of locational dummy (temperate agro-climatic region) is found to be positively significant in both SFA and DEA. Such a result is consistent with the estimates of the cost inefficiency effect model as presented in Table 2. The estimated locational dummy of sub-tropical agro-climatic region is negatively significant in SFA, implying cost inefficiency is lower among the farmers undertaking farming in such location. Hence, the farmers in higher elevation area are subject of greater cost inefficiency.
Table 5 shows the cost efficiency scores in different methodologies and scale assumptions. The predicted cost efficiency in SFA shows substantial variability (58%–100%) among the sampled farming households, with mean efficiency of 92 per cent and a standard deviation of 11 per cent. This means that an average farm in the study area incurs cost 8 per cent above the minimum cost defined by the frontier. Thus, in comparison to best practice, farms producing the same output and using the same technology, more than 8 per cent of costs are wasted in maize production. This reflects the presence of cost inefficiency in farm production and there is a scope for minimising the cost of production. For the overall sample farmers, nearly 82 per cent farmers belong to the cost efficiency interval of 80–100 per cent as per the SFA. However, the predicted mean efficiency for the overall sample farms in the study area is found to be 28 and 45 per cent, respectively, with the assumption of CRS and VRS, respectively. Thus, an average farm in the study location incurs cost about 72 and 55 per cent above the minimum cost defined by the frontier. Specifically, DEA tends to yield a lower mean and a higher standard deviation than SFA, unless there are some unusual statistical association between random error and true efficiency (Bauer et al. 1998).
The mean cost efficiency shows a very irregular pattern across the agro-climatic regions. There seems to be inverted U-shaped relationship between elevation level of farming and mean cost efficiency as per SFA and CRS assumption of the DEA (Table 5). The estimated average cost efficiency scores imply that farmers in sub-tropical region is more cost efficient. These results are in line with the study by Cowan et al. (1998), which found major increases in efficiency of subtropical dairy feeding systems in northern Australia. However, under the VRS assumption of DEA, the cost efficiency scores for different agro-climatic regions of Sikkim reveals U-shaped relationship with elevation areas of farming.
Conclusion and Policy Implications
The article makes an attempt to measure the cost inefficiency of maize farming and its determinants in different agro-climatic regions of Sikkim, India by applying the DEA and SFA approach. Using original farm level survey data, stochastic frontier cost function is estimated together with the inefficiency model. There is the presence of cost inefficiency among the sample farms of the present study, with mean cost inefficiency of 0.92 as per SFA estimates, while those obtained from DEA estimates in CRS and VRS are 0.28 and 0.45, respectively. Such results suggest that an average farm in the study area incurs cost that is 8 per cent to 72 per cent above the minimum cost defined by the best practice frontier while producing the same level of maize output. Hence, there is a scope for minimising cost of production in attaining a frontier level of output through better managerial practices and deliberately changing the exogenous factors that adversely influence cost efficiency of farmers. Except the cost of seed, bullock and organic fertiliser, the estimated coefficients of all input cost are in conformity with prior expectation.
The estimated inefficiency model shows that an additional year of maize farming experience helps in reducing cost inefficiencies of farmers. Greater cost inefficiency seems to be associated with remoteness of farmland from input market. The farmers in the rented plots are found to be more cost efficient. The farmers in temperate agro-climatic region are more cost inefficient compared with farmers in tropical and sub-tropical agro-climatic region. Thus, cost inefficiency is directly linked to the higher elevation area of maize farming in Sikkim.
It can be concluded that farmers operating in leased-in land are more cost efficient in maize cultivation. Hence, effective land reforms in this regard may prove to be useful to enhance maize production in Sikkim. Public and private initiative in opening up of farm input outlet at the village level may help the farmers in getting an easy access to farm inputs, which may indirectly influence the cost and farm output of maize in the state. The possibility of agro-climatic regional concentration of cost efficiency needs to be explained with appropriate region-wide information on soil characteristics of the farms, environmental condition and rainfall pattern. The present study could not control for the effect of such factors which might have influenced the efficiency level of farms in three different agro-climatic regions of the study area. This may be considered as a data limitation of the study.
Extent of Crop Diversification
Footnotes
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.
Funding
The study was funded by Sikkim University as a part of ‘University Research Award’, 2017–18.
