# Nexus between poverty and productive efficiency among the farming households.

Background to the StudyPoverty has become the most prioritised socioeconomic development goal set by almost all of the developing economies. The phrase "make poverty history" has become a hackneyed term for development stakeholders such as donors, philanthropic institutions, and politicians. However, the key issue is identifying strategies to combat poverty, which may vary across countries and continents. Increasing productive efficiency in agriculture has become one of the major development policy options in many agriculture-based economies in the context of reducing poverty.

The World Bank links poverty with the absolute living standard of people and identifies poverty as the failure to achieve a minimum standard of living in the society (WB, 1989). Lipton and Ravallion (1995, p. 2553) assert that 'poverty exists when one or more persons fall short of a level of economic welfare deemed to constitute a reasonable minimum, either in some absolute sense or standards of a specific society'. Sen (1985; 1993) links poverty to human freedoms, entitlements, and empowerment and argues that poverty depends on individuals' command over different goods and services. Among all the prevailing definitions of poverty, a common definition is that a household is considered to be poor if it falls under a given level of welfare threshold.

Alleviation of poverty is the biggest challenge facing many developing countries, including Nepal. Particularly in Nepal, aggregate poverty reduced significantly over the last decade, however, the decline was slower in rural areas compared to urban areas. As the growth was greater in the rich community, overall economic growth rate increased and overall poverty reduced resulting a rise in inequality. Land appears to be one of the basic assets that poor people depend on to make a living. Furthermore, land is the main source of income and consumption for the majority of Nepalese people. However, the level of poverty is the highest among agricultural wage labourers (54%), followed by self-employed agricultural households (42%) (WB, 2006).

The relationship between land and poverty is embedded in Nepalese society which is highly hierarchical in terms of production relations. However, this resource is arguably misallocated. Access to, and control of, agricultural land is largely determined by existing power relations between different classes of Nepalese society, rather than its productive use as a resource. For instance, those who have more land do not know how to use it most efficiently and those who know how to use it do not have an adequate amount of land (NPC, 1998). Therefore, the agricultural productivity of the country is much lower than other countries in the region. This situation is considered to be an obstruction to agricultural development. Thus, an agrarian reform programme through effective policies, may increase productive efficiency, reduce poverty and promote equity.

Many efforts have already been made with the objectives of alleviating poverty and inequality. The land reform programme of 1964 had a higher social and psychological aim, but had little impact on changing the pre-existing power and discriminatory social and agrarian structure. In 1994, the High Level Land Reform Commission submitted a report to implement land reforms but the report was not implemented. Another land reform programme was announced in 2001, focusing on amending the Land Act 1964, to reduce the legal size of land holding per family. However, these were simply political pronouncements, with no assessment of the socioeconomic implications, particularly productive efficiency.

The prospect of an expansion in the amount of agricultural land in Nepal is virtually non-existent. The manufacturing sector is still infant and unorganised. So, in the long term, increased food production to meet the growing population will have to come through improvements in agricultural productivity. A Nepalese Government policy document estimates that there is the opportunity to increase farm production three- to four-fold through land and agrarian reform (NPC, 1998). In this perspective, the formulation and implementation of a reform programme based on a comprehensive understanding of production economics may be the most favoured policy option. It is therefore important to empirically investigate whether farmers can improve their productive efficiency by using existing technology more efficiently. In this context, an in-depth empirical study on technical efficiency in agriculture such as this seems to be crucial.

Very limited research measuring efficiency has been carried out in Nepalese context, either using old secondary data or using primary data, focusing on a single crop production and covering a small village or a very small part of the country (Adhikari and Bjorndal, 2012, Adhikari and Bjorndal, 2009; Adhikari, 2009; Dhungana, Nuthall and Nartea, 2004; Ali, 1996; Belbase and Grabowski, 1985). Against this background, the overall objective of this study is to explore the ways to increase agricultural efficiency and reduce poverty in Nepal; the specific objective is to measure the nature and extent of technical efficiency in agricultural production system in Chitwan district and trace possible sources of inefficiency in the context of alleviating poverty and promoting equity. This study using cross section primary data measures productive efficiency of aggregate farm production in Chit-wan, Nepal. Furthermore, constructing output variables incorporating all crops and other farm related products produced by sample households with in a whole year, this study analyses productive efficiency using stochastic distance function (SDF) model in the context of increasing technical efficiency and thereby reducing poverty and inequality.

Data and the Study Area

The cross sectional data used for this study was collected from primary sources using personal interview method. Chitwan district was selected purposively as the study area which is located in Narayani zone of the central development region of Nepal. The main reasons for selecting this district were researcher's familiarity with the study area and his previous work experience in the district. Furthermore, as the district is relatively advanced in agricultural production and farmers are more conscious, it is considered that the district is very conducive to analyse land related issues and hence derive land reform policy implications to the largest extent possible for the whole country.

Taking the district as a population, sample households were selected from the district using two stage area sampling2 procedure. In the first stage three village development committees (VDCs) namely Meghauli, Bachhauli and Birendranagar were selected randomly among the total 36 VDCs of Chitwan district. In the second stage a total of 50 households from each VDC were chosen randomly. The unit of information were household heads. The selected household heads were interviewed personally with the help of a specially structured questionnaire. Necessary information including detailed input and output data, range of household-specific information (socio-economic) and information on quality of land and additional variables were collected. The field survey was conducted in January and February, 2010.

Construction of Variables

Output Variables

Household outputs can be classified into the following categories: cereal crops; cash crops; and other outputs. Each output variable is measured in terms of Nepalese rupees (Rs.) and is obtained by multiplying the physical quantity by its respective average price. Thus all the outputs are aggregated in monetary terms. The three output variables are as follows:

[q.sub.1] = cereal crops: this group includes paddy, maize, wheat, and other cereal crops and accounts for 49.05% of total agricultural produce, [q.sub.2] = cash crops: this consists of all cash crops including oil seed crops and winter and summer vegetables. Altogether, they comprise 31.09% of total agricultural output in Nepal. [q.sub.3] = other outputs: this includes pulses, fruits, thatch, fodder trees, bamboo, income from other agriculture related activities such as draft animals, sales of crop by-products such as straw and husk and income from livestock.

Inputs Variables

Consistent with the literature on production economics, a rural household in Nepal needs three essential inputs to generate outputs: land, materials, and labour. Six production variables are defined and included in the empirical model. The variables are constructed so as to encompass the inputs used in agricultural production systems in Nepal.

[x.sub.1] = farmed land (Bigha): this is taken as the area of farmed or operated land in Bigha. [x.sub.2] = human labour (days): this is measured as the total number of worker days that family members and other hired labour work in the farm. Males and females are treated as equals. It is believed that gender does not make a significant difference in the contribution to agricultural production in Nepal. [X.sub.1] = purchased inputs (Rs.): this is the total expenditure (Rs.) on seeds, fertilizer (DAP, urea, potash, zink) and pesticides. [X.sub.4] = farm yard manure (Rs.): Almost all farms in Chitwan use farm yard manure as one of the important inputs. This is the total expenditure (Rs.) on manure which is constructed multiplying the total unit used, by its respective market price, [x.sub.5] = Capital service (Rs.): The capital service is included in the model as a proxy of risk aversion. First, values of domestic animals are aggregated based on their respective prices. The capital service variable is then constructed by calculating as 10 percent (the official nominal interest rate) of the total amount of farm assets, [x.sub.6] = other costs (Rs.): this variable covers input expenses not included in previous variables. The variable is constructed by adding the costs of irrigation, transportation and storage facilities and others.

Farm-specific Variables

The stochastic frontier methodology used in this study makes it possible to evaluate factors related with inefficiency. To do so, several socioeconomic and technical variables are incorporated into the model based on the literature and data availability. Thus, in addition to the conventional input variables described above, six relevant farm-specific variables are included in the inefficiency effect model as follows:

[z.sub.1] = land quality: value of land per bigha (Rs) is included as a proxy of land quality. This is measured as the market value of owned land. Agriculture output [yield] depends mainly on the quality of land. Generally, the availability of irrigation facilities is the primary factor differentiating land quality. [Z.sub.2] = extension service (number): this is the number of visits to the farm made by agricultural extension agents during the last agricultural year. [Z.sub.3] = age of the household head (years): this is intended to represent the experience of the farm manager. [Z.sub.4] = family bead's education (years): this is the level of formal education (in terms of years) obtained by the household head and is a proxy for the farm manager's skills.

Locational Dummies

The data is collected from three different parts of Chitwan district. Each location is diversified with respect to the mode of agricultural production. These locations may have significant impacts on technical efficiency. Three dummy variables are included to account for possible location fixed effects. Birendranagar is treated as a reference or default location and two locational dummies are specified as follows: [Z.sub.5] = Meghauli dummy, i.e., the value is 1 if the farm is located in Meghauli and 0 otherwise. [Z.sub.6] = Bachhauli dummy, i.e., the value is 1 if the farm is located in Bachhauli and 0 otherwise.

The Empirical Model

The principal methodology employed in this study to measure technical efficiency is the stochastic distance function (SDF) approach. The main reason for this specification is that agriculture in developing countries shows substantial variability in production due to random factors, including resource availability, missing variables, environmental influences, weather, and measurement errors. Consequently, the frontier and technical efficiency results derived from deterministic methodologies could lead to biased estimates because those methods do not address stochasticity in the empirical model. Agriculture is also a joint production system where multiple outputs are produced by using multiple inputs. Previous stochastic frontier analyses are based on a single output or aggregated single index output, implicitly assuming that the weight of all products is equal. To overcome this problem, the distance function technique is applied to estimate the stochastic frontier that can accommodate the multi-output multi-input problem. Under these circumstances, the SDF method is appropriate to study technical efficiency for two main reasons: (i) it allows for multiple outputs and inputs which are essential for the study of farm diversification effects on productivity and (ii) it can incorporate directly variables affecting inefficiency. In addition, SDFs have major strengths over other multi-output techniques commonly applied in agriculture, such as the deterministic DEA method, which includes the ability to deal with stochastic noise; accommodating traditional hypothesis testing; and allowing for single-step estimation of the inefficiency effects (Kumbhakar and Lovell, 2000).

Depending on the nature of the production system, a SDF can be formulated with an input and/or an output orientation. The input orientation gives the proportional reduction in all inputs that would bring a farm to the frontier isoquant while the output model reflects the proportional increase in outputs attainable by moving to the production possibilities frontier holding input quantities constant. If the application of inputs is more flexible than the outputs produced, the best choice is an output-oriented specification (Coelli et al., 2005, Paul and Nehring, 2005). On the other hand, if inputs are essentially fixed, then output composition is the primary economic performance determinant and an input-oriented specification is preferable. In Nepalese agriculture the balance of inputs used is more flexible than outputs and therefore an output distance function is specified. The translog SDF methodology employed largely follows Pascoe, Koundouri and Bjorndal (2007), O'Donnell and Coelli (2005), Paul and Nehring (2005), and Coelli and Perelrnan (1996, 2000). Those authors extended their methodology from that of Shephard (1970).

Given a production possibility frontier, the distance between the frontier and a specific farm is a function of the vector of inputs used, 'x', and the level of outputs produced, 'y'. Consider a case of a multiple output multiple input production function, where a farm uses the p x 1 input vector x = ([x.sub.1] ... [X.sub.p])' to produce the M x 1 output vector q = ([q.sub.1] ..., [q.sub.M])', the production relation can then be explained by the technology set as follows.

S = {(x, q): x can produce q} (1)

This set consists of all input output sets (x, q), such that 'x' can produce 'q'. This production function can also be represented using a technical transformation function. The functional relationship defined by the set, S, may be correspondingly defined in terms of the output set, P(x), which represents the set of all output vectors. The vector q can be produced by using the input vector, x. In notational expression, the output set is defined by

P (x) = {q: x can produce q} = {q : (x, q) [member of] S} (2)

The output sets are sometimes identified as production possibility sets associated with various input vectors x. Following O'Donnell and Coelli (2005) and Fare and Primont (1995), the production technology can be assumed to satisfy a standard set of axioms including convexity, strong disposability, closedness and boundedness. This production relation can also be explained in terms of the output distance function:

D(x, q) = min {[delta] : [delta] > 0, (q/[delta] [member of] S} (3)

Some simple properties of D(x, q), derived from the axioms on the technology set as listed in Fare et al. (1985) and Coelli et al. (2005), can be specified: i. D(x, 0) = 0 for all non-negative x; ii. D(x, q) is non-decreasing in q and non-increasing in x; iii. D(x, q) is linearly homogeneous in q; iv. D(x, q) is quasi-convex in x and convex in q; v. If q belongs to the production possibility set of x (i.e., q [member of] P(x)), then D(x, q) [less than or equal to] 1; and vi. Distance is equal to unity (i.e., D(x, q) = 1) if q belongs to the frontier of the production possibility set.

The distance measure D(x, q) is "the inverse of the factor by which the production of all output quantities could be increased while still remaining within the feasible production set, for the given input level" (O'Donnell and Coelli, 2005, p 497). The distance function measure is therefore equivalent to a Farrell-type output-oriented measure of technical efficiency.

From an empirical point of view, to estimate a distance function (or frontier) it is necessary to specify a functional form that adequately captures the relationship between inputs and outputs. Most recent studies applying the distance function approach have made use of the translog form because imposing linear homogeneity in output is impossible for the other flexible functional forms (Pascoe et al, 2007; Irz and Thirtle, 2005; Paul and Nehring, 2005; O'Donnell and Coelli, 2005). Following these studies, the distance function model in this study is specified by using a translog functional form.

The translog output distance function for "M "outputs and "P*' inputs can be specified as:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (4)

where the [a.sub.o], [a.sub.m], [a.sub.mn], [b.sub.p], [b.sub.pj], and [g.sub.pm] are unknown parameters and In represents the natural logarithm. From Euler's theorem the homogeneity of degree one in outputs implies:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (5)

that will be satisfied if

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (6)

The symmetry restrictions require [a.sub.mn] = [a.sub.nm] and [b.sub.pj] = [b.sub.jp] for all m, n, j and p.

Following Lovell et al. (1994), we impose the homogeneity constraint in the model. "Substituting these constraints into the distance function is equivalent to normalising by one of the outputs" (O'Donnell and Coelli, 2005; 499). If output M is chosen to normalise, Equation (4) becomes:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (7)

where [q.sup.*.sub.m] = [q.sub.m]/[q.sub.M]. Equation (7) can be written in more compact form as

ln(D/[q.sub.M] = TL (x, q/[q.sub.M], [beta]), (8)

or

ln(D)-ln([q.sub.M]) = TL (x, q/[q.sub.M], [beta]) (9)

where TL (.) represents to the translog function and [beta] is the vectors of a, b and g parameters.

Rewriting the equation by substituting -ln(D) = -u as a one sided error term, captures the effects of inefficiency

[-lnq.sub.M] = TL (x, q/[q.sub.M], [beta]) -u (10)

A symmetric error term, v, can be added in this model to address the effects of data noise. Then the translog model is

-ln [q.sub.M] = TL (x, q/[q.sub.M], [beta])-u + v (11)

The parameters of this model can be estimated using maximum likelihood assuming that 'u' is a non-positive random variable independently distributed as truncations at zero of N (0, [[sigma].sup.2.sub.u]) and 'v' is an independently and identically distributed random variable which is N (0, [[sigma].sup.2.sub.u]). Equation (11) can equivalently be specified as

ln [q.sub.M] = TL (x, q/[q.sub.M], [beta]) - u + v (12)

This translog stochastic distance function model is in a normal stochastic frontier form with a two-part error term. As in the ordinary stochastic frontier model, the 'u' in this model is the deviation from the frontier and 'v' is a random error.

The translog distance can be written as:

ln([q.sup.*.sub.M]) = TL (x, q/[q.sub.M], [beta]) + v) (13)

Equation (5.12) may be rewritten using equation (13)

In [q.sub.M] = In [q.sup.*.sub.M] -u (14)

or

ln ([q.sub.M]/[q.sup.*.sub.M]) = (-u) (15)

This illustrates that the technical efficiency (TE) of a farm is the ratio of its mean production to the corresponding mean production if the farm utilised its levels of inputs most efficiently (Battese and Coelli, 1988), i.e.:

TE = [q.sub.M]/[q.sup.*.sub.M] = exp (-u) (16)

This takes values between 0 and 1, with TE = 1 indicating that the farm is fully efficient. To sum up, the difference between [q.sub.M] and [q.sup.*.sub.M] is embedded in u. If u = 0, then [q.sub.M] equals to [q.sup.*.sub.M] implying that the production unit lies on the frontier. In this condition, the farm is technically efficient. If u > 0, the level of the farm's production lies somewhere below the frontier, implying that the farm is technically inefficient.

The technical inefficiency distribution parameter, 'u, can be a function of various operational and farm-specific variables hypothesised as follows

[u.sub.i] = [[delta].sub.0]+[8.summation over p=1][delta] [p.sup.Z] pi + [w.sub.i] (17)

where [z.sub.i] is a 1 x p vector of various farm specific variables which may influence efficiency of a farm, [delta] is a set of parameters to be estimated and [w.sub.i]'s are the random variables defined by the truncation of the normal distribution with mean 0 and variance [[sigma].sup.2.sub.u], such that the point of truncation is [-z.sub.i][delta] i.e., [w.sub.i] [less than or equal to] - [z.sub.i][delta]. These assumptions are consistent with [u.sub.i] being a non-negative truncation of the N([z.sub.i][delta], [[sigma].sub.u.sup.2] distribution (Battese and Coelli, 1995).

This study employed a maximum-likelihood econometric method to estimate translog stochastic distance function model in search of impacts on the farm level technical efficiency. For this purpose Frontier 4.1 econometric software developed by Coelli (1996) was used to estimate the model.

Findings and Discussion

Stochastic distance frontier

Table 1 presents the maximum likelihood (ML) parameter estimates for the estimated SDF model. One of the requirements for the parametric empirical analysis is the selection of an appropriate functional form. The Cobb-Douglas and the translog functions are the two most popular functional forms used in measuring technical efficiency. Flexibility, homogeneity and ease for calculation are some of the characteristics of an ideal functional form. The Cobb-Douglas function is relatively simple, easy to estimate and interpret because in the logarithmic form it is linear and parsimonious in parameters. This function has been a popular choice in production analyses for many years. It may be a good approximation for production processes for which factors are imperfect substitutes over the entire range of input values. The function has convex isoquant. However, its simplicity is associated with a number of restrictive properties (Coelli, 1995).

Most notably, returns to scale are restricted to having the same value across all farms in the sample, and elasticities of substitution are assumed equal to one (Coelli et al., 2005). The function does not allow for technically independent or competitive factors, nor does it allow for first stage and third stage along with second stage. Furthermore, the Cobb-Douglas transformation function is not an acceptable model in a purely competitive farm because it is not concave in the output dimensions (Chambers, 1988). For example, this would imply that the output transformation curve would have an isoquant shape, which is convex to the origin rather than concave. On the other hand, the translog is a flexible form that can offer second-order elasticities to the functional form. Greene (1980) was the first to propose using the translog functional form to estimate inefficiency. Since then the translog functional form has been used in many applied works (e.g. Lovell, Richardson, Travers, and Wood, 1994; Coelli and Perelman, 1996). Although the estimation of the translog form is more difficult, as it needs to estimate more parameters, it has been widely used in empirical studies because it satisfies all the requirements discussed earlier. In this study, a translog functional form is selected to estimation the empirical model.

Following common practice, all variables are normalised by their geometric mean. Therefore, the first-order coefficients can be interpreted as partial production elasticities evaluated at the geometric mean (GM). It is useful to point out that, because of the structure of a distance function, the partial output elasticity corresponds to its estimated coefficient (Coelli and Perelman, 1999). Furthermore, in order to qualify as a well-behaved model, the SDF needs to be non-decreasing in inputs and decreasing in outputs. Table 1 shows that, at the GM, the estimated model satisfies these conditions.

Two hypotheses have been tested with regard to the model specification. The first is a technical inefficiency test, with null hypothesis [H.sub.0]: [gamma] = 0 and the alternative hypothesis [H.sub.1] : [gamma] > 0. The values for the parameters [[sigma].sup.2] and [gamma] are reported at the end of Table 1. The parameter [gamma] is statistically significant at the 1% level, with an estimated value of 0.66. These results indicate that inefficiency is highly significant among the studied households. In addition, the null hypothesis that [gamma] = 0 is rejected, confirming that technical efficiency is stochastic. The test implies that inefficiency exists in the production system and that specification of the SDF model is justified.

The second hypothesis tested is the choice of the functional form, Cobb-Douglas vs. translog. The null hypothesis is rejected, implying that the translog frontier is preferred and captures the production behaviour in Nepalese agriculture.

Returns to Scale

The sum of the first-order input elasticities measures the distance function-based scale economy. The sum measures the percentage change in output if all inputs were changed proportionally. If this estimate is equal to 1, it implies constant returns to scale. If the sum is less (greater) than 1, then, the returns to scale are decreasing (increasing). The sum of the first order input elasticities in this model is equal to 0.77, i. e., less than 1. As the absolute value of the computed t-statistic 2.04 is greater than the critical t-value at the 1% level of significance, the null hypothesis of constant returns to scale is rejected. This illustrates the existence of decreasing returns to scale prevailing in the agriculture of Chitwan. This suggests that productivity gains could be achieved by reducing the size of the farm (Gilligan, 1998). Thus, reducing farm size by breaking up large farms could lead to increased aggregate productivity.

Production Elasticity

In the output distance function model, each of the first order output elasticities with respect to input provide the specific productive contribution to total output. Such elasticities represent the returns to or output contributions from [X.sub.k] changes, similar to output elasticities from production function estimation. The first-order elasticities of the translog distance-function model can also be decomposed into second-order effects to reflect input or output composition changes as scale expands (Paul and Nehring, 2005). These measures present further insights into the production systems. The second-order elasticities provide production complementarities or substitutions among the variables. A negative sign on the elasticity implies a substitute, whereas a positive sign reflects a complement.

In the output translog distance function, the partial derivative of the output with respect to the [m.sup.th] output provides the ratio of the shadow prices of [q.sub.M] and [q.sub.m]. It reflects the slope of the production possibility curve or the marginal rate of transformation between [q.sub.M] and [q.sub.m].

The one sided error term, u, which is the deviation of a particular observation from the estimated frontier, provides the level of technical inefficiency. The inefficiency measures provide the percentages by which production could be increased, or input use reduced, to reach the production frontier.

An output oriented translog distance function can be said to be well behaved if the function is monotonically increasing and concave in input quantifies (Kumbhakar, 1994). Monotonicity implies positive elasticities of inputs within the data range. The complete regression results of the output oriented distance function model across the entire sample are reported in Table 1.

The signs of the first order output elasticities are negative and statistically significant indicating that the transformation curve has a concave shape. The cross and squared output terms are significant across specifications, and many cross-input terms are also significant. The result thus indicates the possibility of substitution among output variables.

As expected, all input elasticities with respect to output are positive and with the exception of one all are highly significant. Thus, the model demonstrates a well behaved production technology. The positive signs of these elasticities indicate that farms can increase output by using more of these inputs. In output oriented translog distance function the production elasticities indicate how overall output changes with the variation in an individual input, keeping other input and output ratios constant, which is similar to output elasticities in production function estimation (Paul and Nehring, 2005). For instance, the elasticity of capital service (0.21) indicates that a 1% rise in capital service would increase overall output by 0.21%. In other words, it seems possible to increase output by increasing farm capital and maintaining the existing levels of other inputs. Similarly, other production elasticities imply that an increase in these inputs will also increase output.

The elasticity for farm capital (0.21) and farmyard manure (0.19) exhibit positive, relatively large and increasing partial production elasticities, indicating that both inputs have a positive impact on total household productivity. The high elasticity of the capital indicates that farm capital is the most important input determining yields in Nepalese agriculture followed by farm yard manure. The elasticity for human labour (0.11), farm land (0.09), and other inputs (0.08) imply that increase in the amount of these variables can increase the total output. The estimated elasticity for purchased inputs (0.7) is also positive but statistically not significant. This implies that these inputs are already used in maximum level and an increase in these inputs will not contribute to total output.

Diversification economies (DEs) are measured as the second derivative of the SDF with respect to outputs (Coelli and Fleming, 2004). Indices of DEs reflect the gain or loss in total output achievable from the reallocation of resources among different products (Morrison Paul et al., 2000; Coelli and Fleming, 2004). More precisely, if the DE index for a pair of outputs is positive then an increase in one raises the marginal product of the other, suggesting that the two outputs are complements. By contrast, a negative sign implies substitutability. DEs are positive between cereal crop and cash crop, and between cash crops and other farm incomes. Thus, cereal crop households could expect gains in productivity by engaging in (additional) cash crops and other incomes oriented households could improve productivity by diversifying towards cash crops.

Technical Efficiency

As stated earlier, the technical efficiency (TE) of a farm is the ratio of its mean production to the corresponding mean production if the farm utilised its levels of inputs efficiently (Battese and Coelli, 1988). This means TE = [q.sub.M] /[q.sup.*.sub.M] = exp (-u) where exp denotes the exponential operator. The difference between [q.sub.M] and [q.sup.*.sub.M] is embedded in u. If u = 0, then [q.sub.M] equals to [q.sup.*.sub.M] implying that the production unit lies on the frontier. In this condition, the farm is technically efficient. If u > 0, the level of the farm's production lies somewhere below the frontier, implying that the farm is technically inefficient. The estimated technical efficiency scores range widely from 0.08 to 0.96, with a mean efficiency score of 0.75. The estimated average efficiency score 0.75 indicates that typical Nepalese farms can increase agricultural production by 25% adopting the technology and the techniques used by the "best practice" farms. Alternatively, on average, there is the potential to achieve the existing level of output by reducing 25% of their inputs. This indicates that a high degree of technical inefficiency is present relative to the best performing farms. It follows that a large proportion of farms operate far from the efficient frontier, implying a substantial scope for improving productivity using the existing level of inputs and resources efficiently.

Table 2 presents the distribution of technical efficiency scores. The table shows that approximately 70% of the farmers achieve technical efficiency levels of 70% or higher and only 30% of farms were operating in the less than 70% technical efficiency range. The highest relative frequency of the technical efficiency index is found in the 81-90% range, followed by 91-100% and 71-80%.

The mean technical efficiency of 75% is consistent with other studies using cross section data. It is also similar to the average efficiency score calculated by Bravo-Ureta and Pinheiro (1993). The authors found the average efficiency score to be 70%, derived from 30 studies conducted by various authors in developing countries using the stochastic frontier and cross section data. Similar results have been found for Nepal.

Factors Determining Inefficiency

The results of the technical inefficiency estimates are presented at the end of Table 1. Following common practice, the interpretation of the parameters is performed with respect to their effect on technical efficiency. In general, the sign of all the coefficients are as expected. Land quality ([z.sub.1]) has a negative sign and it is statistically significant. It indicates a positive effect on efficiency (or negative effect on inefficiency). The largest absolute value of value among the regression coefficients suggests that land quality might be the most important determinant of efficiency. A higher value of land may reflect a higher quality of land (a proxy for land quality) that would be expected to be more efficient from the production point of view. This implies that households with a higher quality of land are more efficient than those having low quality land.

The variable for extension service ([z.sub.2]) reflects the influence of the government extension programme. The extension variable has positive effects on efficiency and it is statistically significant. This indicates that access to extension services in agriculture is likely to promote efficiency.

Older farmers are likely to have more farming experience than the younger entrants and hence less inefficient. However, it is also likely that they might be more conservative and thus less receptive to modern and newly introduced agricultural technology. In this case there would be greater inefficiency in their production. In this study, the household head's age has a positive effect on efficiency but the estimated coefficient is not statistically significant.

The coefficient of the family head's education ([z.sub.4]) is expected to have a negative sign. By intuition, one can expect that greater levels of formal education tend to be more efficient in agricultural production. This implies that farmers with more education respond more rapidly when new technology becomes available and attempt to produce an output level closer to the frontier. As expected, the results reveal positive relationships between the level of education of the household head and technical efficiency but this is not statistically significant. This implies that educated people are not interested to be involved in agricultural profession.

The location dummy variables [z.sub.5] (Meghauli dummy) and [z.sub.6] (Bachhauli) are expected to have positive because these regions mainly rely on conventional farming practices that emphasise producing staple food crops and considered to be less conducive to high value agricultural production. As expected, the signs of these dummies are positive. This implies that the dummy variables for the Bachhauli and Meghauli have a negative effect on efficiency. The case could have been different, had farming in these regions been diversified to non-staple agricultural crops, for instance vegetables and cash crops. This also indicates that the Birendranagar dummy, taken as the base case, has a positive effect on efficiency.

To sum up, farmers with higher quality of land, access to extension service and who live in the Birendranagar location have a higher level of technical efficiency than the farmers not possessing those attributes.

These results suggest that policy makers in Nepal need to understand that there is a high degree of inefficiency in the agricultural production systems. Where inefficient households are able to surplus some resources, they could be used to make additional income to enhance household welfare. For instance, surplus labour could be diverted to off-farm employment where an opportunity exists. Households could use the additional income to acquire new technologies including improved seeds, effective use of compost or organic fertilizer, and new agricultural implements. Further, they could invest in land improvement. All this would lead to improved technical efficiency and thereby household welfare. Increasing household welfare is an effective way of alleviating poverty.

The factors that significantly influence farmers' resource allocation decisions differ widely among individual farmers. The effectiveness of new policies designed to increase efficiency and productivity may depend largely on the extent to which such differences are recognised. In a broad sense, inefficiency should not be viewed as just a result of the differences in the use of input quantities. Institutional factors including extension systems, education, research and general policies are also important. Efficiency enhancing policies must be flexible enough to accommodate these realities.

In the past, various research studies have been conducted using different methodologies in many developing countries but only a few recent studies have measured technical efficiency using a stochastic distance function approach (Paul and Nehring, 2005; Coelli and Perelman, 2000). However, none of these studies have analysed the determinants of efficiency. Applying a distance function methodology, this study examined for the first time the determinants of technical inefficiencies in Nepalese agricultural production system. This may be the novelty of this research work.

Conclusions and Policy Implications

The empirical results showed that the variation in output among agricultural farms in Nepal is due to differences in technical efficiency. The estimated model showed significant levels of technical inefficiency among the sampled households. This suggests that opportunity exists to expand household production using the current level of inputs and technologies already available in the regions. Variations in amounts of production inputs have a significant influence on the level of production and efficiency across farm households. The empirical results indicate that land quality, measured by the value of cultivated land per Bigha and farm extension visits are key factors associated with higher levels of technical efficiency. These results also suggest that crop diversification which involves a substitution of one crop for other agricultural products seems to be a desirable path for producers of subsistence crops. Decreasing returns to scale also suggest that productivity gains can be achieved by reducing the size of larger farms. Based on the findings, the following policy implications can be derived with regard to increasing efficiency so as to reduce poverty and promote equity.

In view of the limited arable land and other resources, satisfying the increased demand for food through domestic production must come through improvements in productivity, from technological progress or increases in technical efficiency at the farm level. Technical progress relates to the development and adoption of modern technologies, whereas technical efficiency refers to the farmer's ability to achieve maximum output from a given set of inputs by using available productive technology efficiently. Given the existing production technology in Nepal, there is limited prospect of technical progress. In this context, the policy makers need to understand that an increase in technical efficiency is relatively cost effective and therefore government policies should be directed towards this.

This study shows that given the present state of agricultural technology, farms have a potential for enhancing productivity by increased use of some traditional inputs. Capital service is identified as the main factor for determining yields in agriculture. Therefore, government policy should give a high priority to increasing farm assets with regard to draft animals. Farm yard manure is another important input to increase productivity. In the context of substantial use of chemical fertilizers, households should be encouraged to prepare and effective use of compost or organic fertilizer in larger quantity. In the same way, government policy should facilitate the supply of and access to required capital, high quality seeds and other inputs for farmers.

Determinants of efficiency are the shifting factors of the production frontier. The quality of land is recognised as the most influential determinants of efficiency. Irrigation is the main factor for determining quality of agricultural land. Therefore, government policy should give a high priority to increasing irrigation facilities. The findings suggest that agriculture extension programmes have a significant effect on technical efficiency. Therefore, government policies should also be targeted to increased access to extension service for farmers along with enhancement in the level of education, training and knowledge of farmers. These types of policies and practices could contribute to increased technical efficiency. Government policies should also facilitate the private sector to come forward and assist in diffusing modern technologies through extension and training, so that farmers can apply available agricultural technology more efficiently and escape from poverty trap.

Decreasing returns to scale also suggest that productivity gains can be achieved by reducing the size of larger farms and maintaining appropriate farm size. Policies targeted at creating viable farm size by breaking up large farms and the merger of small farms might have beneficial effects on efficiency, although this issue may need to be studied further.

The existence of a high degree of technical inefficiency also suggests that farmers' resource allocation decisions differ widely among individual farmers. Farmers' interactions with each other should have some beneficial effect towards catching up on new technology. Producers' organisations can also improve efficiency in the delivery of government support services and empower them to get involved in many activities.

Higher efficiency gives subsistence farmers the opportunity not only to produce the current level of outputs with less inputs, but also allows them to free up some of their land which can then be shifted towards cash crops and thus a more diversified income stream. Implementing such an option can enhance household liquidity and can gradually open up further opportunities which could reduce the vicious circle of rural poverty. However, this strategy also requires that explicit attention be given to facilitating market access to peasant farmers for their growing and diversified output mix.

The most important factor preventing the achievement of sustainable economic development in many developing countries is the vicious cycle of rural poverty which forces farmers to overuse their land leading to its degradation followed by lower agricultural productivity and more poverty. The findings of this study are particularly relevant in the process of resource management and poverty alleviation. The empirical evidence suggests that policies directed to controlling farm-level land degradation using right proportion of farmyard manure, and other agricultural inputs can improve agricultural efficiency. Moreover, crop diversification which involves a substitution of one crop for other agricultural product seems to be a desirable path for producers of subsistence crops.

The analysis clearly demonstrates that technical efficiency varies significantly across farms. The effectiveness of new policies designed to increase efficiency and productivity may depend largely on the extent to which such differences are recognised. Efficiency improvement policies should be flexible enough to accommodate these realities. For instance, younger and older household managers, educated and uneducated, with and without capital, with irrigated land and rain-fed land, might comprise sub-groups with small, medium and large farms located in different regions. Therefore, policies targeting separate groups, rather than 'one size fits all', will be an effective approach to improve efficiency and productivity. In the same way, recognising farmers who are inefficient in using some resources such as chemical fertilisers would be useful in treating them separately for intervention purposes.

References

Adhikari, C. B. and Bjorndal, T. 2012. "Analyses of Technical Efficiency Using SDF and DEA Models: Evidence from Nepalese Agriculture." Applied Economics, 44:3297-3308.

Adhikari, C. B. 2011. "Determinants of Growth Rate of Per Capita Income and Manufacturing Output in Process of Economic Development in Nepal", Contributions to Nepalese Studies, 38(1): 55-74.

Adhikari, C. B. 2010. "Reform as an Effective Approach to Alleviate Rural Poverty and Inequality in Nepal". Nepalese Journal of Development and Rural Studies, 7 (2): 9-33.

Adhikari, C. B. 2009. Poverty Linkages to Productive Efficiency and Access to Land: Illustration from Nepal. Germany: Lambert Academic Publishing.

Adhikari, C. B. and Bjorndal, T. 2009. "Can Land Reform be an Effective Approach to Alleviate Poverty and Inequality in Nepal?" SNF Working Paper No. 30, Institute for Research in Economics and Business Administration, Bergen.

Adhikari, C. B. and Bjorndal, T. 2009. "Measuring the Extent of Technical Inefficiency in Nepalese Agriculture using SDF and DEA Models." SNF Working Paper 28, Institute for Research in Economics and Business Administration, Bergen.

Adhikari, C. B. and Chatfield, P. 2008. "The Role of Land Reform in Reducing Poverty: Evidence from Nepal." Paper Presented at the Third Annual Himalayan Policy Research Conference Nepal Study Centre, Wisconsin, Madison, USA.

Ali, M. 1996. "Quantifying the Socio-Economics Determinants of Sustainable Crop Production: An Application to Wheat Cultivation in the Terai of Nepal." Agricultural Economics 14: 45-60.

Belbase, K. and Grabowski, R. 1985. "Technical Efficiency in Nepalese Agriculture." Journal of Developing Areas, 19, 515-526.

Chambers, R. G. 1988. Applied production Analysis: A Dual Approach. Cambridge: Cambridge University Press.

Coelli, T. 1996. "A Guide to FRONTIER Version 4.1: A Computer Program for Stochastic Frontier Production and Cost Function Estimation," CEPA Working Paper, 07/96. Department of Econometrics, University of New England, Armidale, Australia.

Coelli, T. and Fleming, E. 2004. "Diversification Economies and Specialization Efficiencies in a Mixed Food and Coffee Smalholder Farming in Papua New Guinea." Agricultural Economics 31: 229-239.

Coelli, T. and Perelman, S. 1999. "A Comparison of Parametric and Non-parametric Distance Functions: with Application to European Railways." European Journal of Operational Research 117: 326-339.

Coelli, T. J. 1995. "Recent Development in Frontier Modelling and Efficiency Measurement". Australian Journal of Agricultural Economics 39:219-245.

Coelli, T. J. and Perelman, S. 1996. "Efficiency Measurement, Multi-output Technologies and Distance Functions: with Application to European Railways." CREPP Working Paper 96/05, University of Liege.

Coelli, T. J. and Perelman, S. 2000. "Technical Efficiency of European Railways: A Distance Function Approach." Applied Economics 32: 1967-76.

Coelli, T. J., Ran, D. S. P., O'Donnell, C. J. and Battese, G. E. 2005. An Introduction to Efficiency and Productivity Analysis. New York: Springer.

Dhnngana, B. R., Nuthall P. L. and Nartea, G. V. 2004. "Measuring the Economic Inefficiency of Nepalese Rice Farms Using Data Envelopment Analysis." The Austrilian Journal of Agricultural and Resource Economics 48: 347-369.

Fare, R. S. and Primont, D. 1995. Multi-Output Production and Duality, Theory and Applications. Boston: Kluwer Academic Publishers.

Fare, R. S., Grosskopf, S. and Lovell, C. A. K 1985. The Measurement of Efficiency of Production. Boston: Kluwer-Nijhoff Publishing.

Gilligan D. O. 1998. "Farm Size, Productivity, and Economic Efficiency: Accounting for Differences in Efficiency of Farms by Size in Honduras." Paper Presented at AAEA.

Greene, W. H. 1980. "On the Estimation of a Flexible Frontier Production Model." Journal of Econometrics 13:101-116.

Irz, X. and Thirtle, C. 2005. "Dual Technological Development in Botswana Agriculture: A Stochastic Input Distance Function Approach." Journal of Agricultural Economics 55: 455-478.

Kumbhakar, S. C. 1994. "Efficiency Estimation in a Profit Maximising Model Using Flexible Production Function." Agricultural Economics 10: 143-152.

Kumbhakar, S. C. and Lovell, C. A. K. 2000. Stochastic Frontier Analysis. Cambridge: Cambridge University Press.

Lipton, M. and Ravallion, M. 1995. "Poverty and Policy". In: Behrman, J. and Srinivasan, T. N. (eds.). Handbook of Development Economics. Amsterdam: Elsevier.

Lovell, C. A. K., Richardson, S. Travers, P. and Wood, L. L. 1994. "Resources and Functionings: A New View of Inequality in Australia." In: Eichhom, W. (ed.). Models and Measurement off Welfare and Inequality. Berlin: Springer-Verlag Press.

Morrison Paul, C., Johnston, W. and Frengley, G. 2000. "Efficiency in New Zealand sheep and beef farming: The impact of regulatory reform." The Review of Economics and Statistics 82: 325-337.

National Planning Commission (NPC) 1998. The Ninth Plan. Kathmandu: NPC, Nepal.

O'Donnell, C. J. And Coelli, T. J. 2005. "A Bayesian Approach to Imposing Curvature on Istance Functions." Journal of Econometrics 126: 493-523.

Pascoe, S., Koundouri, P. and Bjorndal, T. 2007. "Estimating Targeting Ability in Multi-Species Fisheries: A Primal Multi-Output Distance Function Approach." Land Economics 83: 382-397.

Paul, C. J. M. and Nehring, R. 2005. "Product Diversification, Production Systems, and Economic Performance in U. S. Agricultural production." Journal of Econometrics 126: 525-548.

Rawlins, G. 1985. "Measuring the Impact of I. R. D. P. II upon the Technical Efficiency Level of Jamaican Peasant Farmers." Social and Economic Studies 34: 71-96.

Sawers L. 1998. "Farm Size, Productivity, and Public Policy in the Argentine Interior." Journal of Developing Areas 33: 121-149.

Sen, A. K. 1985. Commodities and Capabilities. Amsterdam: North Holland.

Sen, A. K. 1993. "Capability and Well-Being: The Quality of Life." In: Nussbaum M. C. and Sen, A. K. The Quality of Life. Oxford: The Clarendon Press.

Shepbard, R. W. 1970. Theory of Cost and Production Functions. New York: Princeton University Press.

World Bank (WB) 1989. Nepal: Policies for Improving Growth and Alleviating Poverty. Washington DC: World Bank.

World Bank (WB) 2006. Nepal Resilience amid Conflict: An Assessment of Poverty in Nepal, 1995-96 and 2003-04. Washington DC: The World Bank.

Chandra Bahadur Adhikari (1)

Notes

(1) Corresponding address: CNAS, Tribhuvan University, Kirtipur, Nepal. E-mail address: chandra.adhikari@gmail.com

(2) As its name implies two-stage area sampling procedure involves two different stages of random sampling. Stage 1 is a random selection of a sample of sub areas and stage 2 is a random selection of a sample of units from each chosen sub-area. Although, this sampling procedure generally requires more time and money this allows researcher to examine a wider, more representative geographic area, hence it will be statistically more efficient.

Table 1: Maximum-likelihood estimates for the SDF St. Variables Coeff. Err. t-ratio Constant 1.257 0.413 3.043 Cereal -0.604 0.031 -19.281 Cash -0.254 0.051 -4.955 Others -0.143 0.017 -8.465 Cereal-Cash 0.015 0.004 3.947 Cereal x Other -0.018 0.003 -6.571 CashxOther 0.004 0.002 2.037 [(Cereal).sup.2] -0.003 0.002 -2.000 [(Cash).sup.2] 0.001 0.000 4.604 [(Other).sup.2] 0.012 0.002 6.719 Land 0.091 0.008 11.192 Labour 0.113 0.009 12.429 Purchased 0.075 0.061 1.220 Manure 0.192 0.012 15.524 Capital 0.213 0.011 19.244 Others 0.085 0.012 7.054 LandxLabour 0.015 0.033 0.454 LandxPurchased -0.012 0.032 -0.368 LandxManure 0.043 0.024 1.774 LandxCapital 0.009 0.020 0.449 LandxOthers 0.008 0.020 0.400 LabourxPurchased 0.051 0.021 2.399 LabourxManure 0.015 0.018 0.826 LabourxCapital 0.021 0.016 1.295 LabourxOthers 0.031 0.016 1.908 PurchasedxManure 0.024 0.017 1.393 PurchasedxCapital 0.017 0.015 1.118 PurchasedxOthers 0.024 0.015 1.575 ManurexCapital 0.019 0.014 1.336 ManurexOthers 0.021 0.014 1.476 [(Land).sup.2] 0.086 0.069 1.242 [(Labour).sup.2] 0.059 0.027 2.165 [(Purchased).sup.2] 0.078 0.025 3.086 [(Manure).sup.2] 0.058 0.020 2.861 [(Capital).sup.2] 0.083 0.021 3.900 [(Others).sup.2] 0.028 0.019 1.456 LandxCereal -0.006 0.021 -0.276 LabourxCereal 0.039 0.016 2.398 PurchasedxCereal 0.024 0.016 1.479 ManurexCereal 0.016 0.015 1.053 CapitalxCereal -0.024 0.014 -1.685 Others-Cereal 0.022 0.014 1.546 LandxCash 0.030 0.019 1.560 LabourandxCash 0.025 0.015 1.641 PurchasedxCash 0.013 0.015 0.817 ManurexCash 0.027 0.014 1.895 CapitalxCash -0.002 0.013 -0.160 OthersxCash 0.013 0.013 0.985 LandxOthers 0.003 0.017 0.191 LabuorsxOthers 0.017 0.015 1.118 PurchasedxOthers 0.024 0.014 1.685 ManurexOthers 0.027 0.014 1.895 CapitalxOthers -0.005 0.013 -0.385 Othersx0thers 0.025 0.013 1.887 Constant (Z0) -1.520 0.525 -2.896 - Land ualit (Z1) -0.253 0.024 10.424 Extension (Z2.) -0.214 0.052 4.088 Age (Z3) -0.047 0.232 -0.203 Education (Z4) -0.025 0.024 -1.081 Megauli (Z5) 0.247 0.015 16.310 Bachhauli (Z6) 0.326 0.062 5.216 [[sigma].sup.2] = [sigma][v.sup.2]/ 2.425 0.907 2.673 [sigma][u.sup.2] [gamma] = [sigma][u.sup.2]/ 0.661 0.027 28.240 [[sigma].sup.2] Table 2: Distribution of Technical Technical Eff. Number of Farms Percentage of Farms Interval (%) 0-10 2 1.34 11-20 3 2.01 21-30 4 2.68 31-40 5 3.36 41-50 7 4.7 51-60 11 7.38 61-70 12 8.06 71-80 17 11.41 81-90 48 32.21 91-100 40 26.85 Total: 149 100 Mean TE 75

Printer friendly Cite/link Email Feedback | |

Author: | Adhikari, Chandra Bahadur |
---|---|

Publication: | Contributions to Nepalese Studies |

Article Type: | Report |

Geographic Code: | 9NEPA |

Date: | Jul 1, 2012 |

Words: | 8748 |

Previous Article: | Phonological comparison between Chhatthare Limbu and Athpahariya. |

Next Article: | Levels of users' participation and decision making process in community forest management of Nepal: a case study of Kaski district Nepal. |

Topics: |