Sarhad J. Agric. Vol. 23, No. 1, 2007

MEASUREMENT OF ECONOMIC EFFICIENCY IN AGRICULTURE OF PESHAWAR DISTRICT Muhammad Saeed and Nasser Ali Khan ABSTRACT The aim of the present research is to quantify the existing productive potentials at the farm level and pinpoint the causes of inefficiency in agriculture of Peshawar District. A total of 210 farmers, (105 each small and large) randomly selected from four villages of sample district. Farm specific efficiency is estimated using Cobb Douglas Production function. The findings show that the farmers are inefficient in the use of their resource. In percentage terms the average inefficiency is 11.5% with a range of 3.0% to 18.2%. The main socio-economic causes of inefficiency are lack of education, financial resources and fragmented as well as large farm size. The study recommends that the inefficiency can be improved through strengthening of the social sector development mainly education and financial support to the farming community.

INTRODUCTION Restoring growth and rising income levels of the farming community is very important for broad based poverty reduction in Pakistan in general and in NWFP in particular. The majority (over 80%) lives in rural areas with a high dependence on farming. Agriculture is the main source of livelihood for these people. The development of this sector is not only important for enhancement of income in rural areas but also for the development of industrial sector as well. A number of authors have noticed the interdependence of agriculture and manufacturing sector. For example Ranis et al. (1961) claimed that the manufacturing sector can not continue to expand without the increase in agricultural productivity. Similarly, Jorgenson (1961) mentioned that economic growth, which breaks the vicious circle of poverty, can not be achieved without rapid improvement in agricultural technology in conjunction with a decline in the birth rate. Kuznets (1976) suggested that without a marked improvement in agricultural productivity, the economies of Europe and United States could not have attained such high country wide rates of growth. Schultz (1964) also pointed out that economic growth can not be attained with out modernizing the traditional agriculture. Ruttan (1987) argued that genuine economic development can not be attained without a concomitant for improvement of productivity in the agriculture sector. Agriculture provides food to the whole population of Pakistan. Though there is food

scarcity in Pakistan, yet measures are being taken to make the country self-sufficient in food. Agriculture plays a vital role in the economy of Pakistan. It has been realized that if Pakistan’s economy is to develop and has to function smoothly, agriculture sector will have to be given high priority in allocation of funds. Realizing the importance of agriculture and to achieve the goal of maximum efficiency in farm production the government has pursued multiple measure in the last few decades including the introduction of land reforms, changes in agricultural tax structure, the establishment of an Integrated Rural Development Programme (IRDP) and cooperative societies, the promotion of research on various aspects of high yielding varieties of seed, and an improvement in extension services for the dissemination of agricultural knowledge and the propagation of modern techniques of agricultural production. Price incentives and input subsidies are, however, two important instruments of economic policy for raising farm production. To save farmers from future uncertainties and to encourage them to supply adequate quantities of agricultural commodities, the government has been extending to them the facilities of price guarantees in the form of minimum procurement prices for various farm products. Pakistan's agricultural productivity is well below the potential level. So long as the yield of crops remains low, the great bulk of land will continue

Department of Economics, University of Peshawar, Peshawar – Pakistan.

198

Muhammad Saeed and Naseer Ali Khan.. Measurement of economic efficiency in agriculture in Peshawar….

to be devoted to the growing of subsistence cereals for the human population and there will be very little possibility of crop diversification to include more nutritious foods and commercial crops. Hence poverty and underdevelopment in Pakistan appear to be closely related to the very low level of productivity in agriculture, and policies for improvement are necessary (Shah and Khan, 1983). There is a great need to accelerate agricultural growth and to increase yields to achieve the objective of raising the living standard of the people. The most obvious way to increase land productivity is to use increasing quantities of agricultural inputs (fertilizer, improved seed, pesticides etc.) on the existing land. Developing new technologies which increase the productivity of land as well as other inputs is another and even better way. Although technological advancement has been identified as a very important source of growth, its development takes time and requires considerable investment (Greene, 1980). Inefficiency across farms occurs due to some farms being unable to produce a maximum output with given inputs (technical inefficiency) or a failure to combine inputs in the correct proportion at given factor prices (allocative inefficiency) or to random factors beyond the control of the farmers. The two types of inefficiencies are different from each other in nature and they cannot be examined by the traditional production function approach (Jondrow et al. 1982).

and farm specific inefficiency has not been estimated in detail (Shah, 1994). Therefore, this study is undertaken to accomplish the said task. Objectives of the Study The broad objective of the study is to quantify the existing productive potential at the farm level and to pinpoint the causes of inefficiency, in order to formulate better agricultural policy. The specific objectives of the study are: i.

To examine the extent of farm specific technical inefficiency in the use of agricultural resources.

ii.

To determine the factors responsible for inefficiency.

iii.

To suggest policy measures to improve agricultural productivity.

MATERIALS AND METHODS The geographical scope of the study was restricted to Peshawar District in 2005. For empirical estimation primary data was collected through structured questionnaire, which was designed to cover various aspects of farm operations and through face to face interview. The sample in each village was divided into small and large farms of major crops i.e. wheat, maize, sugarcane etc. of district Peshawar and on the tenurial group it was distributed among owners and pure tenants. Then households in each village from various categories were randomly selected. The distribution of the sample respondents by size and tenurial status is given in Table I.

Most of the previous studies in NWFP have used the traditional approach for estimating efficiency Table I Distribution of respondents by size and tenurial status Tenurial Status Small Owners 52 Tenants 53 Total 105

The study interest was to group the farmers into small and large, so the data collected in the same way. But for econometric analysis it is important to test the validity whether these group farmers are really on different production function, the Chow test was used. To test the hypothesis two types of

Size Large 53 52 105

All 105 105 210

regressions need to be estimated, one is a restricted regression assuming no farm size effect and the other is an unrestricted equation allowing for the farm size effect.

199

Sarhad J. Agric. Vol. 23, No. 1, 2007

N1,N2= Number of observations in small and large farms. K = Number of regression parameters including the constant term.

The appropriate test applied here is the Chow, Ftest given as follows:

F* =

[ RP − R S − RL ] / K ……(1) ( RS + RL ) /[ N 1 + N 2 − 2 K ]

The estimated value of F* was compared with the tabulated value for estimating the different production function for small and large farmers. For this purpose the Cobb-Douglas Production Function was specified for the small, large farms and for the pooled data.

Where RP = residual sum of squares from pooled regression (small and large farms). Residual sum of squares from the RS = equation for small farms. RL = Residual sum of squares from the equation for large farms.

Ln Y = αo+β1LnSD+β2LnMN+β3LnFR + B4 LnIR + B5 Ln HL+ B6 Ln BO + B7 Ln BH + B8 Ln TR + e …….. (2)

Y=

where is the aggregated value of output (crops only) and by-product (If the variation is very low in any input, those will be dropped). Farm Yard Manure in 50 kg bags. Fertilizer in 50 kg bags. Family labour in hours. The use of own bullock power in

The OLS estimates of the equation (2) for small farms, large farms and for the aggregate data are presented in the following Table II. The R2 is fairly high and indicates that more than half of the variation in the output is due to the explanatory variables. The residual sums of squares (RSS) are also presented in the table. This will be used to conduct the Chow, F test. The value of the Chow F is 1.86, which is insignificant at 5% level confirming that small and large farmers are not different in their production process.

MN= FR= FL= BO= hours. TR= Tractor use in hours. e= Others factors which are not visible e.g. rain, draughts etc Table II

Results of Cobb-Douglas production function

Parameters

Coefficients Small Farmers 5.054 0.0202 0.0650 0.0324 0.0470 0.0261 0.54 0.321

αo β1 β2 β3 β4 β5 R2 RSS

T-ratio 80.32* 2.50* 5.31* 2.17* 4.14* 3.30*

Large Farmers αo β1 β2 β3 β4 β5 R2 RSS

4.321 0.0359 0.0492 0.0397 0.0556 0.0457 0.68 0.211

40.33* 2.82* 2.08* 1.78* 4.17* 3.998*

All Farmers (Pool Data) αo β1 β2 β3 β4 β5 RSS

3.212 0.346 0.0284 0.0497 0.578 0.447 0.562

•

= Significant at least at 5% level.

65.21* 2.72* 3.32* 4.82* 2.41* 1.95*

Muhammad Saeed and Naseer Ali Khan.. Measurement of economic efficiency in agriculture in Peshawar….

There are Different approaches to measure the efficiency. The approaches are broadly divided into two methods: a. Deterministic frontiers, and b. Stochastic parametric frontier. In this study, the stochastic frontier approach was used to measure the efficiency, which has advantage over the deterministic frontiers, because this separate the two type of errors, i.e. random error and technical inefficiency. The effort was made to use all the important inputs of agricultural activity, i.e. human labour, animal labour, fertilizer, tractor and manure use. On the other side all the possible factors of economic efficiency, i.e. demographic, institutional and resource factors are taken into consideration to explain the variation in efficiency. In this way, a more reasonable policy recommendation for the agriculture of Peshawar District will be suggested. Measurement of Cost Inefficiency Using the stochastic frontier approach a direct estimation of Cobb-Douglas Production function is estimated. An attempt is made to explain the estimated cost inefficiency for farms by the socioeconomic factors. Cobb-Douglas Cost Function Cobb-Douglas production, cost and profit functions are very popular in econometric literature. Econometric testing of various assumptions can provide a parsimonious form to represent the data in hand. Instead of estimating a direct Cobb-Douglas production function, duality approach is used where cost is related to input prices and output quantities. Binswanger (1978) suggests that for estimating pair wise substitution elasticities between inputs, translog cost function approach is superior to translog production function, since the estimation of substitution elasticities from cost function does not involve the inversion of a matrix as it is required from a production function. This is consistent with the assumptions that prices are exogenous in economics while quantity of inputs are endogenous in agriculture where there is a timelag between input use and output. The time-lag between sowing and harvesting and farmers' decision making on sowing and harvesting affects the amount of output a farmer will produce.

200

Allocation of land to a given crop determines potential output of a crop at the time of sowing. A Cobb-Douglas Cost function with the stochastic frontier where there are two error terms, one to represent the efficiency component and another the random term is shown below: log c = αo + αQ log Q + Σ αi log Pi+ ε I …..(3) where εi = ui + vi

i = 1, 2.... n

Where c is total cost, Q is the value of output, Pi are factor prices and ui and vi are efficiency and random errors, respectively. ui are assumed to be distributed half normal, exponential or truncated normal. This assumption of distribution is made for the sake of convenience. The compound disturbance is the sum of symmetrically distributed variables and a onesided one. vi is allowed to be normally distributed to reflect the random factors such as weather, and use a one-sided disturbance, ui to represent the inefficiency component. Each farm's efficient frontier is unique to that farm. A producer's costs may be forced up by weather, other random and uncontrollable events. These are not due to inefficiency. It is assumed u = │u│ and u ~ N(0,σ²u) (4) While v ~ N(0,σ²v), u and v are assumed to be independent of each other. If all of the disturbances in the model were strictly one-sided i.e vi=o, then we have only inefficiency disturbance term. For exponential model, the density function is f(u)=θe-θu and the means and variances are E(u) = 1/θ, V(u) = 1/θ2. The residual is computed by the formula E[u|v + u] for the cost frontier. Jondrow, Lovell, Materov and Schmidt (1982) derived the conditional mean of ui given εi. They show that: E[u| ε] = z+

σv φ(z/σv ) .......... .......... .......... .......... .......... .......... .......... .......... .....(5) Φ(z/σv )

Where z = ε - Θ σ 2v and φ and Φ are the standard normal and standard cumulative functions respectively. The maximum likelihood estimates of β, and σ2ε are the ones where the value of

201

Sarhad J. Agric. Vol. 23, No. 1, 2007

likelihood is maximized in a model y = β'x + v + u. These estimates are obtained by equating the first derivatives of likelihood with respect to β,

λ = σu

σv

and σ2ε to zero and solving these

simultaneously yield maximum likelihood estimators. Replacing εi by ^ e i in equation (5), estimates of ui were obtained. Exponential of these estimates will be eui .

.......... .......... .......... ............ .......... .......... .......... .......... .... (6) σε=σu+σv.......... 2

2

than its profit. It was not possible to present the farms specific inefficiency of the individual 210 farms. Therefore, frequency distribution is presented in Table V. The mean inefficiency was 11.5% with a maximum of 81.2% and with a minimum of 3%. Majority of the farms (17.8%) were inefficient in the range of (50-55). The next 16.19% were inefficient in the range of (15-20). It mean that large number of farms are not getting the frontier cost. In other words, enough farmers are producing below the best farms.

2

Estimation of the Function In this section, the Cobb-Douglas Cost function is specified and the estimated coefficients are presented in Table IV. The R2 was 0.55 and signs of the coefficients were in the line of economic theory. The estimated coefficients show the decreasing returns to scale. Moreover the value of λ was 3.84, which is significant at 5% levels suggesting the farmers are inefficient. In other words, the cost incurred on the production is more

Causes of Inefficiency When farmers were found inefficient, the farms specific inefficiency was estimated. The causes of inefficiencies were searched out i.e. the socio economic or demographic characteristics of the individual farmers and developed the following model and the results are presented in the Table VI.

TIi = constant+α1FSZi+α2AGEi+α3EDUi+α4OFFi+α5ASSETi+α6WEALTHi+ α7ANIi+α8CREDITi +α9FRAGi+α10HOLDi+α11SUBi+α12EXTi+ ε…(7) Where TI = Technical Inefficiency FSZ = Farm Size AGE = Age of the respondent EDU = Education of the respondent OFF = Off Farm Work ASSET = Farm Assets WEALT= Non Farm Assets ANI = Working Animals CREDIT = Credit FRAG = Fragmentation HOLD = Landholding SUB = Subsistence Need EXT = Extension Services The results presented in Table III indicate that the greater the family size, the younger (age) the household, high education of the household head and the greater wealth, assets or credit contribute to reduce inefficiency. Fragmented land and large size of holding both contribute positively to inefficiency while the extension visits expose farmers to better techniques, which contribute to

improvement of efficiency. As farmers’ subsistence needs are satisfied by their own production of food crops, it tends to contribute positively to inefficiency as these prevent farmers to minimize their costs and restrict them to achieving cost efficiency. The OLS and ML estimates on per acre basis are presented in Table IV. All the coefficients have positive signs as expected. The constant term is higher in ML than in OLS and λ is 0.7451 which is not significantly different from zero. When λ = 0 the ML is equivalent to OLS and farmers are technically efficient in the use of bundle of inputs and all errors are symmetric. Following Jondrow (1982) farm specific cost efficiency for individual farm was calculated. Frequency distribution of cost efficiency of the individual farms are presented in Table V which shows a wide range of inefficiency. Technically efficient farms 18.09 % 16.19 %and 14.29 % are in the range of 50-55, 15-20 and 5-10 respectively.

Muhammad Saeed and Naseer Ali Khan.. Measurement of economic efficiency in agriculture in Peshawar….

The mean efficiency is 11.5 per cent or in other words the minimum efficiency is 3 per cent. The results show that there is considerable scope to improve the efficiency in agriculture of the Peshawar district. This inefficiency is caused by lack of finance fragmented land, larger holdings, subsistence needs, older and uneducated farmers, and lack of contact with extension workers. The farmers can be made more efficient if rural education, agricultural extension services in the area are strengthened to guide the farmers to exploit the available farm technology. A higher family size increases efficiency because at the time of peak season, there is shortage of labour. Education has a positive and significant impact on technical efficiency. Credit improves farmers' liquidity and facilitates the purchase of inputs and encourages farmers to introduce HYV to improve the yield per acre. Fragmented land reduces the efficiency index. Extension services will improve efficiency as better management and information utilization should lead to greater benefits to farmers. CONCLUSIONS An investigation into technical efficiency issue using Cobb-Douglas production, cost and profit functions can be estimated. It is, however, difficult to state which of the approaches is most recommended for an empirical study on the measurement of efficiency. One of the important conclusions of a conceptual debate is probably the differences between the original pure measure of technical efficiency defined by Farrell (1957) and a slightly different measure of efficiency called “cost inefficiency” or “profit inefficiency” through

202

duality approach which is used in this study. Technical inefficiency at farm level was estimated using the translog production function and correlation analysis between production and cost based measures of inefficiency was conducted in this empirical analysis. The assumption of output being exogenous and independent of residual, inefficiency in the dual cost function approach does not permit us empirically to estimate the pure Farrell measure of technical inefficiency. Cost based measure is some kind of combination of technical and allocative inefficiency or deviations of actual cost from minimum cost which may be regarded as a measure of cost inefficiency. The measure of inefficiency using stochastic production frontier approach yields a plausible measure of technical inefficiency. It is, however, comforting to know that the expected results based on the cost inefficiency were borne out by the data from 210 farms of the Peshawar District of North-West Frontier Province of Pakistan. The subsequent regression analysis using various socio-economic, demographic and environmental factors to explain inefficiency index provided us with the results to confirm our a priori views that the causes of inefficiency are lack of finance (credit, farm assets, non farm assets), fragmented land, larger holdings, subsistence needs, older and uneducated farmers, and lack of contacts with extension workers. Attempts to improve efficiency can be made through land reform programs, agricultural credit banks and farm education.

203

Sarhad J. Agric. Vol. 23, No. 1, 2007

Table III Names

Descriptive Statistics of variables Measuring unit

Mean

S.D

Family Size

Numbers

6.01

2.60

Age of H.head

Years

47.02

11.22

Education

Years of School

1.05

2.37

Off Farmwork

Hours/Month

24.51

20.70

Farm Assets

Rs/Acre

1020.65

4620.88

Non Farm Assets

Rs/Acre

7480.66

8240.60

Working Animals

Number/Acre

0.31

0.42

Credit

Rs/Acre

744.52

3821.63

Fragmentation

Number/Acre

0.62

0.60

Extension Contacts

Numbers

3.34

2.24

Farm Size

Acres

Subsistence Need

Rs/Acre

Fertilizer Price

6.20

5.33

2742.54

2670.15

Rs/Kg

2.69

0.50

Human Labour

Rs/Day

24.43

5.05

Animal Labour

Rs/Day

50.66

8.66

Tractor

Rs/Hour

55.62

10.20

Output-Price

Rs/40 kg

35.52

40.76

Value of Output

Rs/Acre

3833.00

940.58

Fertilizer

Kgs/Acre

125.27

40.73

Human Labour

Days/Acre

20.24

5.31

Animal Labour

Days/Acre

5.55

6.39

Tractor

Hours/Acre

2.30

1.10

Manure

Maund/Acre

70.02

35.20

Table IV Parameters Α β1 β2 β3 β4 β5 R2

Estimates of Cobb-Douglas production function OLS estimates Coefficients T-values 5.597 70.95* 0.0280 3.520* 0.0322 2.076* 0.030 2.139* 0.0254 6.286* 0.040 5.177* 0.55

λ=

σu σv

σ = 2 σ2u + σ2 v Likelihood Value = Significant at 5% level

*

(OLS and Frontier approach) Composed error Estimates Coefficients T-values 5.985 60.90* 0.0279 3.30* 0.0227 1.71* 0.0350 2.10* 0.0554 6.21* 0.040 4.37*

0.74515

3.84

0.92144

2.44

534.58

204

Muhammad Saeed and Naseer Ali Khan.. Measurement of economic efficiency in agriculture in Peshawar….

Table V Farm specific technical inefficiencies in stochastic Cobb-Douglas production frontiers Inefficiency Index No. of Farms (%) 1-5 16 7.62 5-10 30 14.29 10-15 20 9.52 15-20 34 16.19 20-25 11 5.24 25-30 4 1.90 30-35 5 2.38 35-40 8 3.81 40-45 11 5.24 45-50 6 2.86 50-55 38 18.10 55-60 9 4.29 60-65 7 3.33 65-70 0 0.00 70-75 8 3.81 75-80 2 0.95 80-85 1 0.48 85-90 0 0.00 90-95 0 0.00 All 210.00 100.00 Mean 11.5 Std 6.6 Min 3.0 Max 81.2 Table VI

Relationship of inefficiency index with farm characteristics TII BASED ON COBB-DOUGLAS COST Name of the variable Coefficient Constant 0.1160 Family Size (FSZ) -0.0001 Age (AGE) 0.0002 Education (EDU) -0.0020 Off Farm Work (hours per month) (OFF) 0.78x10-3 Farm Asset (Rs. Per Acre) (ASSET) -0.2075x10-5 Non Farm Assets (Rs. Per Acre) (WEALTH) -0.514x10-5 Working Animals (No) (ANI) -0.0053 Credit (Rs. Per Acre) (CREDIT) -0.16x10-4 Fragmentation (No Per Acre) (FRAG) 0.0046 Extension Visits (Number) (EXT) -0.0043 Land Size Holding (Acres) (HOLD) 0.0013 Subsistence (Rs/Acre) (SUB) 0.32x10-4

CONCLUSION AND RECOMMENDATIONS Findings of the study confines only to the major crops of one district of the province. Before making any generalization of the findings this sort of studies might be conducted in others districts of the province. Inefficiency was observed in this

T-ratio 5.723 -1.129 1.237 -2.605 0.413 -0.206 -1.526 -2.061 -2.087 1.062 -4.014 2.626 2.057

study, which can be reduced by targeting policies on the factors influencing inefficiency. Agricultural credit, extension services, education and ownership of farm assets are the factors underlying the estimated inefficiency. An improvement in the availability of credit enables

205

Sarhad J. Agric. Vol. 23, No. 1, 2007

the farmers to make timely purchase of the inputs, and extension services provides knowledge to the farmers about the appropriate use of modern inputs such as fertilizers. Education is expected to improve the farmers’ understanding of soil, farming practices, and use of inputs. All these are at work together to enable to become more efficient. It is, therefore, recomended that increased education, extension, services, availability of credit and farm assets all will reduce inefficiency in district as well as in the province. REFERENCES Ali, F., A. Parikh and M.K. Shah. 1994. Measurement of economics efficiency using the behavioral and stochastic frontier approach. Applied Econ. 26: 181-188. Binswanger. 1978. Technology and relative economic Efficiency. Oxford Econ Papers. 30:184-198. Farrell, M.J. 1957. The Measurement of productive efficiency. Royal Statistical Assoc, 120 Series A – General. 253-81. Furguson, C.E. 1969. The Neoclassical theory of production and distribution. Cambridge Univ. Press. Greene, W.H. 1980. Maximum likelihood estimation of econometric frontier functions. J. Economet. 13:27-56. Greene, W., 1992. LIMDEP Computer Program: Version 6.0, Econometric Software. Jondrow, J., C.A.K. Levell and P. Schmidt. 1982. On the estimation of the technical inefficiency in the stochastic frontier production function, J. Economet. 19:233-238.

Jorgenson, D.W. 1961. The development of dual economy. Econ J. 71:309-334. Kuznets, S. 1976. Modern economic growth. New Haven. Yale Univ. Press. Ranis, Gustav and J.C.H. Fei. 1961. A theory of economic development. Amer. Econ. Review. 51:533-565. Ruttan, V.W. 1987. Lectures on technical change in agricultural development. Islamabad, Pakistan Inst of Dev. Studies. Lecture in Econ. Dev. No.6. Schultz,

T.W. 1964. Transforming traditional agriculture. Yale Univ. Press, New Haven.

Shah, M.K and M.N. Khan. 1983. Farmers response to govt economic incentives in Peshawar Tehsil. Instt. of Dev. Studies, Agric Univ. Peshawar, Public. No.175. Shah,

M.K. 1994. Measurement of technical efficiency in agriculture: comparison of deterministic and stochastic frontier approaches. Sarhad J. Agric. 10(2):26-37.

Shah,

M.K. and A. Parikh. 1992a. Various approaches to measurement of technical efficiency in North West Frontier Province of Pakistan. Econ. Res. Centre, UEA, Norwich.

Shah M.K and A. Parikh. 1992b. Measurement of technical efficiency in Pakistan: A stochastic frontier approach. Econ. Res. Centre, UAE, Norwich.

Muhammad Saeed and Naseer Ali Khan.. Measurement of economic efficiency in agriculture in Peshawar….

206