An Economic Analysis of Groundwater Development vis-à-vis Resource Use Efficiency in Tank Command Areas.

An Economic Analysis of Groundwater Development visvis Resource Use Efficiency in Tank Command Areas. Free Online Research Papers ABSTRACT: Out of total farmers 75 per cent were cultivating rice in tank command areas. The rest of the farmers cultivated other crops and hence resource use efficiency has been estimated for rice crop alone. Among the rice growers 27 per cent of farmers have raised rice with tank water alone while the rest applied both tank water and well water. The analysis was done for two situations i) tank water alone and ii) tank cum well water application. The total cost of rice cultivation using only tank water was Rs 16016/- per hectare and tank cum well water situation, the total cost of rice cultivation was Rs 24628/- per hectare. The Mean Technical Efficiency (MTE) was calculated to be 0.3996 for only tank water using farmers. It indicated that the technical efficiency of rice farmers were only 39.96 per cent and yield of rice could be increased by 60.04 per cent more by adopting a technically efficient plan without any increase in cost. The Mean Technical Efficiency (MTE) was calculated to be 0.6248 for tank cum well water users. It indicated that technical efficiency of rice farmers was only 62.48 per cent and yield of rice could be increased by 37.52 per cent more by adopting a technically efficient plan without any increase in cost. Introduction The important factor in agricultural development in India is going to be efficient use of available water resources for crop production. The increasing need for crop production due to growing population led to the rapid expansion of irrigation throughout the world. Historically, groundwater is an important source of irrigation of India. Its contribution in enhancing agricultural production was better realized during the green revolution period. However, in the race for increasing agricultural production, its over–exploitation and mismanagement had resulted in several problems like fluctuations in water table and increase in depth of wells. Indian agriculture received the highest priority in irrigation development in successive Five-Year Plans. The irrigated area increased from 20 million ha in 1950-51 to 57.3 million ha in 1999-2000. With the advent of new agricultural technology in mid 1960s, several incentives (like electrification, bank credit, etc.) were given to the groundwater development. The area under groundwater irrigation, which was less than six million ha in 1959-60, went up to 18 million ha in 1980-81 to 33 million ha in 1999-2000. More than half of the total irrigation is done through groundwater. The share of groundwater in total irrigated area increased from 30 per cent in 1960-61 to 58.77 per cent in 1999-2000. Most of the groundwater development came through private investment. The area under groundwater is increasing progressively as this is the most reliable and cost-effective source of irrigation (Joshi, 2002). This paper is based on the MSc (Agri.) of the first author. Thesis was submitted by the author (Venkatesh.G. 2003) to the Department of Agricultural Economics, Tamil Nadu Agricultural University (TNAU), Coimbatore. The major sources of irrigation in India are tanks, canals and wells. The tanks have existed in India from time immemorial, and have been an important source of irrigation, particularly, in South India where it accounts for about one-third of the rice irrigated area. (Palanisami et al., 2001) The recent estimate places the actual number of tanks in Tamil Nadu at 34,000, the remaining 5,000 plus has just disappeared over the past 15 years, so because of a variety of reasons during the 1980’s. Though there are several studies on tank irrigation and its problems, studies on groundwater development and resource use efficiency of rice in tank command areas are limited. However, in this paper we attempt to study the resource use efficiency in rice cultivation and returns to supplemental irrigation in tank command areas. In particular, we employ the stochastic frontier production techniques to measure technical efficiency of rice. The study proceeds as follows Section I explains the methodology used in the study. Data, model and variables are discussed in section II. Section III provides the empirical results and the final section (IV) summarizes the findings and suggests policy implications. Section I Methodology The measurement of efficiency was the main motivation for the study of frontier. The technical efficiency literature begins with Farrell (1957), employed a deterministic approach in which he estimated a cost frontier by using linear programming (LP), requiring all observations to lie on or above the frontier. Aigner and Chu (1968) translated Farrell’s cost frontier into a production frontier, since outlier observations under a deterministic approach seriously affect the problem, by using a probabilistic frontier function. This approach deletes outlier observations, one at a time, to avoid spurious errors due to extreme observations, until the estimated coefficients stabilize. Then, Timmer’s (1971) approach yields a frontier, which is probabilistic rather than deterministic or stochastic. Later Aigner et al., (1977) developed a stochastic frontier model and key feature of the model was that the disturbance term is composed of two parts, one symmetric and the other â₠¬Ëœone-sided’. A symmetric disturbance term is normally distributed component permits random variation of the frontier across firms and captures the effect of measurement error, other statistical noises and random shocks outside the firm’s control. A one-sided error component captures the effect of inefficiency relative to the stochastic frontier. Parameters of the stochastic frontier may be estimated by the Maximum Likelihood Estimate (MLE) or Corrected Ordinary Least Square (COLS), method if the probability function for symmetric and one-sided components of the error term is specified. A number of comprehensive literature reviews are available, such as Battese (1992), Kalirajan Shand (1994), Mythilli Shanmugam (2000), Tim Coelli et al., (2002) and Shanmugam K.R (2003) A (linear) stochastic frontier model is specified as Y = f (X1,X2,†¦Ã¢â‚¬ ¦Xn) + (v ? u) Where, v is the symmetric error component causing the deterministic part of the production frontier f (X1, X2†¦Xn) to vary across the firms. Technical efficiency relative to the stochastic production frontier is captured by the one-sided error component (? depending on whether one specifies a production or cost frontier), u?0. Given the density functions for u and v the frontier function defined above may be estimated by maximum likelihood techniques. While several distributions can be considered for the term u, the statistical estimation of the frontier model combining both u and v usually leads only to the estimation of average technical efficiency of the sample observations since their combined effects could not be separated under such general assumptions. However, individual observation specific-technical efficiency measures are more useful from a policy viewpoint. The approach to identify firm specific technical efficiency requires some estimators that allow for separating the effects of the one-sided error term u from the combined effects of u and v using the estimated frontier functions. Therefore, the problem is to predict ui under the assumption that ui+vi is known. The best predictor of an unknown random variable (ui) under the value of the combined random variables ui+vi is the minimum mean squared error predictor given by the conditional expectation of ui. Assuming a half normal distribution for ui and normal distribution for vi, the frontier model becomes Y=f (X1,X2,†¦Ã¢â‚¬ ¦Xn) + (v ? u) where, u = ? u ? and u ?N ?0, ?2u ? and v ? N ?0, ?2v ? The components of the disturbance term are assumed to be independent and the frontier is assumed to be linear in above case. (In case of multiplicative models the ?(v-u) component is expressed as exp (v-u)). Now, the firm or observation specific ui can be estimated as E? u i? ( ui + vi )?= ?u ?v /? ? f(.) / (1-F(.) -?( ui + vi ) / / (1-? )? 1/2? Where f(.) and F(.) are standard normal density and distribution functions evaluated at ?(ui+vi) / ? ? ? ? /1-? ? 1/2, ? = ?2u / ?2 and ?2= ?2u+?2v Alternatively, E (u? e) =/ (1+?2)? ? f (E? / ?) / F (E? / ?) E? / Where ?=?2u / ?2v One advantage of estimating the frontier production that is possible to find out whether the farmers deviation of yield from frontier is mainly because they did not use the best practical technique or due to external random factors. Thus, one can say whether the difference between the actual yield obtained and the frontier yield, if any, occurred accidentally or not. Following Battese and Coelli (1988), when output is measured in logarithms, the farm-specific technical efficiency can be estimated as: TEi = Exp (-ui) i = 1,2,3†¦n, 0 ? TEi ?1 The variance ratio ?, explaining the total variation in output from the frontier level of output attributed to technical efficiencies, can be computed as: ?= ?2u /?2 Where ?2 = ?2u+?2v and 0 ? ? ? 1 ? is an indicator of relative variability of ui and vi met differentiates the actual yield obtained from the frontier. There are two interesting points about ? 1) When ?2v is tends to zero, which implied that vi is the predominant error, then the ?=1. This means that the farmer’s yield difference from the maximum feasible yield mainly because he did not use the best practice technique. 2) When ?2u are tends to zero, which implies that the symmetric error term vi the predominant error, ? is tending to zero. This means that the farmer’s yield difference from the frontier yield is mainly because of either technical error or external factors not under his control. Direct estimates of the stochastic production function frontier model may be obtained by Maximum Likelihood Estimator (MLE) method. In this study MLE method is used to estimate (as was used by Olsen et al., (1980): and Banik Arindam (1994)). Measurement of technical efficiency has been attempted across crops such as Rice (e.g. Kalirajan Shand 1994; Mythili Shanmugam 2000); tea (e.g. Hazarika Subramanian 1999); rice, groundnut and cotton (Shanmugam 2003); and coffee, orange, banana and pepper (e.g. Venkatesh et al., 2005). Section II Data model and variables used in the study The study area was Sivaganga District, Southern Region of Tamil Nadu, which has more number of tanks, has been purposefully selected as study area. Multi-stage Stratified Random sampling was used. In study area Sivaganga District, Sivaganga Taluk (Stage I) was selected and in that taluk four tanks were selected from PWD management and two were selected from PU maintenance based on command area of the tank (Stage II). So, six villages are benefited by the chosen tanks, namely Namanur, Kovanur, Panaiur, Mudikondon, Valuthani and Salur. Twenty farmers from each of the mentioned villages were randomly selected (Stage III). On the total 120 respondents were interviewed. Rice was the major cereal crop in this district. Therefore, rice crop was chosen for further analysis. The survey was conducted during the year 2002-2003. The empirical model consists of single stage. In that stage, the stochastic frontier production function was estimated. For that purpose, the Cobb-Douglas production function was employed and which is given by: Cobb-Douglas production function was used to estimate the resource use efficiency. Y = bo X1b1X2b2X3b3X4b4 X5b5U Where, Y = Rice yield in quintals per ha X1 = Area under rice in ha. X2 = Fertilizer applied (N+P+K kgs per ha) X3 = Labour mandays per ha X4 = Expenditure on bullock, machinery power, seeds and pesticides (Rs. per ha) X5 = Irrigation (ha cm) bo = Intercept bi = 1,2,3,4, and 5 are production elasticities. U = Error term Section III Empirical Results Distribution of Land Holdings of the Sample Farmers in the Study Area It could be seen from the Table 1 that out of 120 sample farmers 56.7 per cent were marginal farmers, while 32.5 per cent were small farmers and the remaining 10.8 per cent belonged to big farmers’ group. Table 1. Distribution of Land Holdings of the Sample Farmers in the Study Area Name of the Village Category of farmers Marginal (2.5ha) Namanur 6 10 4 Kovanur 15 3 2 Mudikondum 10 7 3 Panaiyur 11 8 1 Valuthani 9 8 3 Salur 17 3 0 Total 68 (56.7) 39 (32.5) 13 (10.8) Figures in parentheses indicate percentage to total Distance of Sample Farms from Sluice of Tanks Distance from sluice is very important to get water for field and also the distance decides the number of supplemental irrigation to be applied. The requirement of supplemental irrigation is less if the fields are nearer to head of the Tank and vice versa. The distances of farmers’ field from sluice of tanks are presented in Table 2. The distribution of farmers among the head, middle and tail end reach of the tank sluices were 38, 37 and 45 respectively. This clearly showed that majority of the farmers field were located at tail end of the tank sluice and the rest were equally distributed between head and middle reach from sluice. Table 2. Distance of Sample Farms from Sluice of Tanks Villages Head (< 400 m) Middle (401-800 m) Tail (> 801 m) Namanur 5 5 10 Kovanur 6 7 7 Mudikondum 8 7 5 Panaiyur 4 6 10 Valuthani 9 6 5 Salur 6 6 8 Total 38 37 45 Well Details of Sample Farmers Details about wells owned by farmers are furnished in Table 3. Generally farmers owned open wells and open cum bore wells. Namanur village had more number of open wells numbering 10, while it was only one in Mudikondum village. The open cum bore wells were maximum in Salur village and they were least in Kovanur village. The average depth of wells was the highest in Salur village (18.8 m) and the least in Mudikondum village (12.7 m) Average pumping hours of irrigation water from wells was the highest in Salur village (8 hours) during monsoon season and the least in Panaiyur village (5.10 hours per day). During non-monsoon season, the average pumping hours were the highest in Mudikondum village with 3.10 hours per day while it was the least in Panaiyur village (2.05 hours per day). The difference in pumping hours between monsoons was the highest in Salur village (5.50 hours per day) and the least in Namanur village (3.00 hours per day). Table 3. Well Details of Sample Farmers Name of the Villages No of Open wells No of Open cum bore wells Total no of wells Average depth of well (m) Average pumping (in hours/day) Monsoon Season (Sep-Dec) Non season (Jan-Aug) Difference between Monsoon and non-monsoon seasons Namanur 10 6 16 13.0 5.15 2.15 3.00 Kovanur 5 1 6 13.6 6.00 2.00 4.00 Mudikondum 1 8 9 12.7 7.20 3.10 4.10 Panaiyur 7 4 11 13.6 5.10 2.05 3.05 Valuthani 7 2 9 13.6 7.15 2.33 4.82 Salur 0 13 13 18.8 8.00 2.50 5.50 Periodicity of Digging of Wells It could be seen from the Table 4, that during 1980-90’s 38.1 per cent of wells were dug by the farmers. Twenty-four wells out of 63 wells were dug during this period. Next to this 31.7 per cent of wells were dug during 1970-80’s. During 1990-2000, 20.6 per cent of wells were dug and the rest were dug before 1970’s. There was no well digging activity after 2000. Thus, two third of wells were dug during 1970 to 1990 and thereafter there had been a slow down in well digging activity. Table 4. Periodicity of Digging of Wells (No. of wells) Name of the Village Before 1970 1970-1980 1980-1990 1990-2000 After 2000 Total Namanur 2 6 5 3 16 Kovanur 3 1 2 6 Mudikondam 3 5 1 9 Panaiyur 2 3 4 2 11 Valuthani 1 2 5 8 Salur 1 3 4 5 13 Total 6 20 24 13 63 Per cent (9.5) (31.7) (38.1) (20.6) (100.0) Average Annual Decline of Water Table It could be seen from the Table 5 that the average annual decline of water table was the highest in Salur village, with 0.396 m which indicated that more number of farmers resorted to groundwater use in that village. In Panaiyur and Kovanur villages the annual decline in groundwater table was 0.280 m. The least groundwater decline was recorded in Mudikondum with 0.163 m during the reference period. Table 5. Average Depth and Decline of Water Table Name of the Village Water Table (in mm) Average annual decline of water table (in meters) 1990 1995 2003 Namanur 9.7 11.2 13.0 0.256 Kovanur 10.0 10.9 13.6 0.280 Mudikondum 10.6 11.2 12.7 0.163 Panaiyur 10.0 11.5 13.6 0.280 Valuthani 10.2 11.2 13.6 0.221 Salur 13.6 15.8 18.8 0.396 Sample Households’ Participation in Groundwater Sales It could be seen from Table 6 that in all the selected villages, own well water users were more in number because of demand for water for own-cultivation. Groundwater sellers sold it to neighbours because of the following reasons. 1) The lands belonging to small and marginal belonged to the poor farmers’ category do not have wells, in such a situation; well owners sold water to them. 2) The sellers reduced their own demand for water by reducing the number of irrigations, and the water thus saved was sold to other farmers. Cost of well water varied among the villages. Also, it depended on whether the well water was pumped with electric motor or oil engine. It ranged from Rs. 20/- to Rs. 50/- for wells fitted with electric motor and oil motor. The price of well water per hour was high in Namanur village, (Rs.25/-) and low in Salur village (Rs.15/-). In wells fitted with oil motor, the farmers sold well water for Rs. 35/- to Rs. 50/- per hour which was the highest in Namanur village (Rs.50/-) and the least in Salur village (Rs. 35). There was neither selling nor buying of well water in Kovanur and Mudikondum villages because of high salt content in water. Table 6. Sample Households’ Participation in Groundwater Sales Name of the Village Own users Sellers Buyers Water charge of irrigation water with Electric motor (Rs/hr) with Oil motor (Rs/hr) Namanur 9 7 4 25 50 Kovanur* 6 Mudikondam* 9 Panaiyur 7 4 8 15-20 40-50 Valuthani 9 4 6 20 40 Salur 5 5 7 15 35 * No groundwater market emerged Resource Use Efficiency Out of 120 farmers, 100 were cultivating rice in tank command areas. The rest of the farmers cultivated other crops and hence resource use efficiency has been restricted for rice crop alone. Among the rice growers, 27 per cent of farmers have raised rice with tank water alone while the rest applied both tank and well water. The Cobb-Douglas production function was estimated as specified for rice growers with tank water alone as well as tank water plus well water and the results are presented in Table 7. In case of tank water users alone, the co-efficient of multiple determinations was 0.897 which indicated that 89 per cent of variations in rice yield have been attributed by the independent variables included in the function and it was significant at one per cent probability level. Among the independent variables included in the function, area under rice and tank water application had significantly influenced rice yield at one per cent probability level. The partial regression coefficients revealed that elasticity of production for area under rice was 6.039 and 0.0393 for tank water application respectively. The production function estimated for rice growers applying both tank and well water revealed that 77.80 per cent variation in rice yield was explained by independent variables included in the function and the function as a whole was significant at one per cent probability level. Among the explanatory variables included, the area under rice and well water application significantly influenced the rice yield at one per cent probability level while the other expenditures significantly influenced the rice yield at five per cent probability level. This showed that the availability of well water had encouraged farmers to spend more on seed, pesticides and machineries. The estimated partial regression coefficients showed the elasticity of production due to land; well water application and other expenditures were respectively 2.598, 0.276 and 0.0007. The elasticities of production indicated that tank water, well water and other expenditures were less than one and were operating in the second zone of production. On the other hand, the elasticity of production for area under rice was more than one for both tank water users and tank and well water users. This showed that there is scope for increasing rice production through expansion of area in Sivaganga district provided the water is made available either in-sittu conditions or water application deliberately and crop management methods. Table 7. Cobb-Douglas Production Function for Farms using Tank Water alone and Tank cum Well Water Sl.No. Particulars Estimated partial regression co-efficients Tank water alone Tank and well water 1 Constant 5.4868 (10.2229) 5.6996 (6.2736) 2 Area under rice in ha 6.0359* (1.3542) 2.5984* (0.6899) 3 Fertilizer (N+P+K) in kg per ha 0.1090 (0.0398) 0.0934 (0.0189) 4 Labour man days per ha 0.2045 (0.1731) 0.1324 (0.173) 5 Expenditure on bullock, machine power, seeds and pesticides (Rs per ha) 0.0019 (0.0007) 0.0007** (0.0003) 6 Tank Irrigation (ha cm) 0.0393* (0.0454) 0.0289 (0.130) 7 Well irrigation (ha cm) NA 0.2762* (0.0906) N 27 73 R2 0.897* 0.778* Figures in parentheses indicate standard errors * Significant at 1 % level of probability ** Significant at 5% level of probability Resource Use Efficiency of Rice Growers Resource use efficiency of rice growers have been worked out for the resources which had significantly influenced the rice yield (Table 8). The ratio of VMP of resource to their price indicated that for farmers using only tank water, both of the area and tank water resource are over utilized. The ratio of VMP of resources to their price estimated for farmers using tank cum well also indicated the over utilization of land and other expenditures whereas underutilization of well water. Table 8. Resource Use Efficiency of Rice Growers A. Tank water alone VMP Px VMP/Px Land 4.57 1500* 0.003 Tank Irrigation 0.99 4 0.25 B. Tank cum well water Land 2.5 1500* 0.002 Well Irrigation 15.16 15 1.01 Other expenditures. 0.58 912.0 0.0007 * Rental value of land was taken as the price of land Marginal product=Elasticity* Geometric mean VMP valued at output price of rice Maximum Likelihood Estimator Method for Production Function for Farms using Tank Water alone and Tank cum Well Water It could be seen from the Table 9 that the estimated discrepancy parameter (?) was 0.9703 and 0.9521 for tank water alone and tank cum well water application respectively. This implied that deviation in the output from the frontier yield was mainly due to technical inefficiency at the farmers’ level. The Mean Technical Efficiency was 0.3996 and 0.6248 respectively for tank water alone and tank cum well water applying farms. This implied that yield was 60 percent less than the maximum possible output for only tank water using farmers and 38 per cent less than the maximum possible output for tank cum well water using farmers. The low technical efficiency was due to inadequate water during crop period in the former category. Besides uncertainty in rainfall and poor filling of tanks had led to these problems. Table 9. Maximum Likelihood Estimator Method for Production Function for Farms using Tank Water alone and Tank cum Well Water Sl.No. Particulars Estimated partial regression coefficients Tank water alone Tank and well water 1 Constant 6.5177 (7.1483) 6.2553 (4.7527) 2 Area under paddy in ha 4.9367* (1.1734) 2.6494** (1.1245) 3 Fertilizer (N+P+K) in kg per ha 0.1035 (0.0246) 0.0604* (0.0156) 4 Labour man days per ha 0.1126** (0.0578) 0.0232** (0.0093) 5 Expenditure on bullock power, machine power, seeds and pesticides (Rs per ha) 0.0009 (0.0007) 0.0013* (0.0005) 6 Tank Irrigation (ha cm) 0.0429* (0.0129) 0.0304 (0.0689) 7 Well irrigation (ha cm) NA 0.6742* (0.1603) 8 ?2u 1.7776 0.6788 9 ?2v 0.0544 0.0342 10 ?=?u / ?v 5.7153 4.4559 11 ?=?2u/ (?2u +?2v) 0.9703 0.9521 12 MTE=1- ?u?2/? 0.3996 0.6248 Figures in parentheses indicate standard errors * 1% level of significant level ** 5% level of significant level NA : Not Applied Technical Efficiencies of Rice Growers The farm specific technical efficiency is furnished in Table 10. It was found that a majority of farmers (55.6 per cent) using only tank water were operating at 40-50 per cent technical efficiency level. On the contrary, majority of the farmers (52.1 per cent) using tank cum well water, were operating at 70-80 per cent technical efficiency and 15.1 per cent of farmers were operating most efficient category (80-90 percent). This indicated that there is scope to improve the productivity of the rice farmers. Identification of farms, which lead to variation in the farm specific technical efficiency, is an important issue for formulating strategies to increase the productivity. Farms Irrigated by Tank water alone Figure 1 Technical Efficiency of Tank water alone irrigated Farmers Farms Irrigated by Tank water and well water Figure 2 Technical Efficiency of Tank and well water irrigated Farmers Table 10. Technical Efficiencies of Rice Producing Farmers (in numbers) Sl.No. Technical efficiency of Rice growers Only tank water using farmers Tank and well water using farmers 1