INTRODUCTION

Cambodia is located in Southeast Asia and the total land area is about 181,035 km². About 80% of the population lives in rural area and 65% depends primarily on agricultural production for food supply, family income, and livelihoods (USAID, 2018). Cambodia is a typical agriculture country, with one-third of the total land area allocated to agricultural production. In 2016, the economy was attributed significantly to agriculture accounted for 26%, industry for 31% and service for 43% of GDP (Gross Domestic Product), in which agriculture employed about 45% of the labor force in 2014. Agricultural resources consist primarily of 4.6 million hectares of cultivated land, with 68% of cultivated land was devoted to rice and 32% to other crops (MAFF, 2017). Rice is the main staple crop and the most important commodities in terms of food security and livelihoods. It provides about 65%-75% of the population’s energy needs as it is the staple of the traditional Cambodian diet. The Royal Government of Cambodia regards rice as “white gold”, which signifies how important the crop is to the country. It is the major agricultural export commodity as well as the most important source of crop value added and the key driver of agricultural growth (Bingxin and Shenggen, 2009).

Cambodian farmers have experience in rain-fed rice cultivation for at least 2,000 years. However, the average rice yield was slowly increased from 3 tons per hectare in 2010 to 3.2 tons per hectare in 2016 (MAFF, 2017). Moreover, the yield of rice in Cambodia still has been low compared to the other rice producing countries such as China, Vietnam, Lao, Philippine and Myanmar where rice productivity in 2016 was recorded 6.93 tons/ha, 5.58 tons/ha, 4.26 tons/ha, 3.86 tons/ha and 3.81 tons/ha respectively (FAOSTAT, 2016).

In Cambodia, rice is the main diet and source of income and rice production is constrained by poor farming practices (farming is still in subsistence stage with traditional way), unavailable good seeds, improper insect and pest management, inefficient use of available technologies, poor soil fertility, insufficient infrastructure (irrigation systems) and high cost of inputs. Low agricultural productivity and production inefficiency in farming have aggravated food insecurity and have grave consequence for food security in the country. In contrast, production efficient would give more export revenue and income for the country. In this context, enhancement in the efficiency level is the main factors of productivity growth for sustainable food production and ensuring food security in developing agriculture economics. The key requirement to improving productivity is that farmers need to use production inputs more efficiently. Efficiency studies can help the country to determine the extent that farmer can increase their rice productivity by improving the neglected source and without improving the resource base or developing new technology and help to determine the under and over-utilization of input. More importantly, it would in turn allow Cambodia to make a crucial step for construction of the developed country and give a direction for the adjustments required in long run to attain food sustainability for ensuring food security and reducing the poverty for all farmers in the country. In order to enhance rice productivity in Cambodia, it is very important to identify the core factors influencing it. The purposes of the study are twofold. First, Data envelopment analysis (DEA) model is used to measure technical efficiency of rice production as an explicit function of discretionary variables. In addition, socioeconomic and farm characteristic variables which are assumed to affect the efficiency of rice production are used in a Tobit model which is used to identify the factors influencing the efficiency levels of rice production in Cambodia.

DATA AND MODEL

Data and Variables

The cross-sectional data were collected during May, 2017 in Tram Kak and Kiri Vong districts, Takeo province. The primary data collection was done by structured interview with rice producers gathered from a total of 80 randomly selected rice farm households in 2016 production year. For the first stage of efficiency analysis by using DEA model, one output and five inputs were included. The output was defined as the rice yield and expressed in kilograms per hectare (kg/ha). The five inputs were rice cultivation size (ha), seed (kg/ha), chemical fertilizer (kg/ha), pesticides and herbicides cost (riel/ha), mechanization cost (riel/ha). For the second stage of efficiency analysis by using Tobit regression model, six independent variables measured upon the DEA efficiency level which obtained from the first stage analysis. Independent variables used in the analysis including sex of family head, age of family head, education level of family head, the main occupation of family head, number of rice growing time and type of seeds planted by farmers.

Data Envelopment Analysis (DEA) Model

Data envelopment analysis (DEA) is a linear programming method to create a non-parametric piece-wise surface or frontier over the data. Efficiency measurement are calculated relative to this surface and it was defined as the weighted sum of output over weight of input in constant return to scale assumption (Charnes et al., 1978). DEA models can be measured by an input-orientation or output-orientation. Input-orientation minimizes inputs while maintaining the same outputs level. In contrast, output-orientation means to increases outputs with the same input level (Malana and Malano, 2006). Nevertheless, the two measures provide the same TE scores when constant return to scale technology applied but are not equal when variable return to scale was applied (Mailena and Shamsudin et al., 2014).

The choice of input-orientation or output-orientation of DEA model was according to the quantities of inputs or outputs which manager of a farm has more control over (Coelli, et al, 2002). Under the condition of Cambodia, farmers have more control over inputs than output, so the input-orientated model would be more appropriate and used in this study. A model that had input-orientation under constant return to scale (CRS) and variable return to scale (VRS) assumption was used to measure the total technical efficiency of the rice farmers which presented by Charnes and Cooper (1978) that based on works of Farrell (1957).

In the first stage, an input-oriented variable return to scale (VRS) DEA model for calculation of pure technical efficiency is estimated as:

Where, θ is a scalar and represents the pure technical efficiency of i_th farm and it will satisfy θ≤1 with the value of 1 implying a point on the frontier and hence technically efficient farm. N1/λ=1 is representative of a convexity constraint which guarantees that an inefficient firm is only benchmarked against firms of a similar size. Y is representative of the output matrix and X represents the input matrix. This linear programming problem can be solved N times one for each firm in the sample.

According to Coelli et al. (2005), scale efficiency (SE) was measured as the proportion of technical efficiency on constant return to scale to technical efficiency on variable return to scale (pure technical efficiency). It can be used to calculate scale efficiency by:

When SE = 1 it is indicated that scale efficiency or constant return to scale; When SE < 1 indicates scale inefficiency. Moreover, scale inefficiency happens because of the presence of decreasing return to scale (DRS) or increasing returns to scale (IRS).

Tobit model

In order to examine the influence factors of technical efficiency in rice cultivating practice of farm households that obtained from DEA model, a Tobit model was used to explain variation in efficiency across farms. According to James Tobin (1958), Tobit model is censored regression models where an expected error does not equal to zero. However, the reasons of choosing a Tobit regression model to test the hypotheses in this study were the value of technical efficiency score (dependent variable) ranges between zero and one that handles the feature of distribution of censored efficiency level and it has been widely used in many previous studies around the world. Following the Introduction to Econometrics (2nd edition) written by Maddala, G. S. (1992) was specifying the Tobit model in this study by the following function:

Where,

Tobit regression parameters are estimated in Stata software.

ESTIMATION AND RESULTS

The summary statistics of inputs, output, and socioeconomic and farm characteristic variables used in the study are shown in Table 1. Under ideal weather conditions and management practices, the average yield of rice production of the total household survey was about 3,136 kilograms per hectare (kg/ha) per year with the minimum of the production at 117 kg/ha and the maximum at 7,538 kg/ha. The average of seed used on rice farms was about 228 kg/ha per year which range from 29 kg/ha to 769 kg/ha. On average, the fertilizer used on paddy farms was about 260 kg/ ha per year and there were some farmers used it up to 967 kg/ha. The variable on input used among farms was found on the utilization of pesticide and herbicide was 134,308 Riel/ha (US$ 32.9/ha) per year with the maximum approximately 1,150,000 Riel/ha (US$ 281.8/ha). The average expenditure for land preparation and harvesting by using mechanization was about 600,498 Riel/ha (US$ 147.1/ha) per year and there were some farmers spent up to 1,880,000 Riel/ha (US$ 460.7/ha).

With reference to the farm and farmer specific variables presented on table 1, it was indicated that about 80% of the family heads were male while 20% were female. On average, age of family head was 49.48 years old with the youngest being 26 years and the oldest being 83 years. The average education level is 2.42 (No education, primary school, secondary school and high school were used 1, 2, 3, 4 to represent respectively). The main occupation of family head was a farming which about 85% while none farming occupation was 25%. About 75% of farmers were growing rice 1 time per year while 25% of farmers were growing rice 2 times per year. Moreover, the average value of type of seed lanted/variety is 2.10.

An input-oriented DEA model was used to measure technical efficiency from a constant return to scale (TE_CRS), pure technical efficiency/technical efficiency from a variable return to scale (TE_VRS) and scale efficiency (SE) of rice farmers. On average, the estimated technical efficiency scores of rice farmers in study area under TE_CRS assumption and TE_VRS assumption were about 0.53 (ranging from 0.07 to 1.00) and 0.67 (ranging from 0.14 to 1.00) respectively. On average, farmers are only producing 53% (CRS) and 67% (VRS) of the output of the best-practices farmers at the same level inputs. This indicated that farms should improve about 47% (CRS) and 33% (VRS) of the efficiency in the utilization of input at the same production level. However, the sample farms with average technical efficiency score implied that farmers may reduce their use of input by 47% and 33% while holding the same level of rice production (Table 2).

A frequency distribution of farm-specific on technical efficiency of rice production is presented on table 3 for rice producers in the study area. Among a total 80 sample of farmers who reached fully efficiency, there was about 16% (13 farmers) under TE_CRS and 26% (21 farmers) under TE_VRS. Furthermore, about 18% of the sample farmers achieved technical efficiency score of 0.60 to 0.99 or closer to the frontier output. This indicates that the opportunities to increase rice production in Tram Kak and Kiri Vong districts of Takeo province is still quite large because there are still more than 60% of the sample farmers in which their output achievement is far below the frontier output. However the highest technical efficiency scored was at 1.00, it's implied that there is considerable room to increase technical efficiency with the current rice production resources.

Scale efficiency provides useful information on the current status of farm operation which helps farmers know whether or not change their production scale for improving efficiency. The average of scale efficiency of the studies rice farms was 0.80 ranging from 0.08 to 1 where 15 farmers are scale efficiency. This indicates that they are operative near to the optimal size and farms inefficiency is due to inefficiency in input use and may be because of inappropriate scale or mis-allocation of production resources. However, the decomposition of the total technical inefficiency measure produced the estimation of 33% pure technical inefficiency (TE_VRS) and 20% scale inefficiency. To eliminate the scale inefficiency, farmers can increase their average technical efficiency level from 53% to 67% by reducing their scale inefficiency. Figure 1 shows that, the majority of them (46%) achieve scale efficiency above 0.90 indicating that more than 40% of these farms were operating quite close to the optimal size. Moreover 15% of farms have scale efficiency between 0.80-0.89, 11.3% of farms have scale efficiency between 0.70-0.79, 7.5% of farms have scale efficiency between 0.60-0.69, 6.3% of farms have scale efficiency between 0.50-0.59, 12.5% of farms have scale efficiency between 0.40-0.49, and only 1.3% of farms have scale efficiency between 0.00-0.09. This shows that some of these farms are not close to the efficiency level.

The result of Tobit regression model between socioeconomic and farm characteristic variables and efficiency are displayed on table 4. It presented that, four variables were found to significantly affect the level of technical efficiency including sex of family head, main occupation, the number of rice growing time and type of seeds planted/ variety. Among the four above mentioned significant factors, it is the only type of seeds planted/variety that had a significantly negative relationship with technical efficiency.

Sex of family head was positively and significantly affecting technical efficiency at 10% level. A comparison between male and female households head indicated that farmers who are a male family head were likely to have higher technical efficiency than those who female family head. This can be suggested by the fact that, male farmers have more opportunity to obtain knowledge, better access to technology, information, and inputs (seed, fertilizer, pesticide, herbicide…) than female farmers. In addition, most female farmers are physically weak and farming practices such as land preparation, watering, applying pesticides and herbicides, and harvesting is tedious and need the active involvement of farmers which lead women farmers to be less efficient. It is consistent with the finding result of Sokvibal (2017), who found that sex of family head was significantly and negatively affecting the technical efficiency of rice production in Cambodia. A study by Betty (2005) implied those male households were more technically efficient than the female family head.

Main occupation of family head was found to be positive and significant at 10% level. The coefficient indicated that the people who are engaged in farming activities (rice cultivation) only were more technically efficient than those who have many activities. This indicates that the people who consider farming as a primary occupation mainly focus on their crop, spend most of their time, resources and effort on farming which increase efficiency compared to their counterpart.

The study results revealed that number of rice growing time had a positive and significant relationship with technical efficiency at 1% level in the study areas. This implies that farmers who grow rice one time per year were likely to be more technically efficient than farmers who grow rice two times per year. It is concluded that farmers who grow rice one time per year had a good crop management, especially proper use of input resources and have relatively adequate resources and time for farming.

Variety or type of rice seeds planted by farmers was found to be negatively and significantly affecting technical efficiency of rice production at 5% level. A comparison between improve seed and local seed suggested that farmers who used improve seed were likely to be more technically efficient than those who used only local seed or both. This implied that improved seeds are high yielding, tolerant to drought and flooding, resistant to pests and diseases, and early maturing, hence increased output and improved efficiency. It is consistent with the results from another study by Kibirige, (2008) on the technical efficiency of maize farmers in Masindi district concluded that farmers who used improved seeds were more efficient than who those did not use them. A study by Willy (2015) on technical efficiency among small-holder sweet potato farmers in Zambia highlighted that farmer who used hybrid seeds were more efficient than those housed local seeds. A study by Betty (2005) on technical efficiency in Kenyan’s maize production implied that farmers who used of certified purchased hybrid seeds had higher technical efficiency.

CONCLUSIONS

The analysis of the technical efficiency levels of rice farmers revealed that they could benefit from the adoption of the best practice methods of rice production because the result of the study indicated a range of difference in efficiency across the farms. The average technical efficiency of rice production in the study area was found about 0.67 whereas rice producers could reduce their input costs by 33% while holding the same production level. Four factors including sex of households head, main occupation of households head, number of rice growing time and the type of seeds planted/variety were found as the factors influencing the efficiency levels of rice production in Tram Kak and Kiri Vong districts of Takeo province. The scale efficiency of rice producers was also determined and it was relatively high (0.80) which indicated that farmers are operating near to the optimal size. This implied that the causes of farm inefficiency are due to misallocation or improper use of input resources.

Furthermore, about 18% of the sample farmers achieved technical efficiency score of 0.60 to 0.99 or closer to the frontier output. This indicates that the opportunities to increase rice production in Tram Kak and Kiri Vong districts of Takeo province is still quite large because there are still more than 60% of the sample farmers in which their output achievement is far below the frontier output. However the highest technical efficiency scored at 1.00 implies that there is considerable room to increase the technical efficiency with the current rice production resources. These results suggest that the policy-makers should pay more attention to technical training, providing financial support and strengthening the research institutions responsible for seed production and multiplication for farmers to increase their technical efficiency in rice production.

적 요

본 연구의 목적은 농가의 사회·경제적 특성을 조사하여 대 상 지역 쌀 생산의 기술적 효율성 수준을 분석하는 것이다. 또한 이 연구에서는 캄보디아의 쌀 생산성 수준에 영향을 미 치는 요인들을 확인하고자 하였다.

이 연구에 사용된 정보는 2016년에 캄보디아 남부에 위치 하는 다케오주의 트람꺽군과 키리웡군에서 무작위로 선택된 80개의 농가를 면접 조사한 자료이다. 농가의 사회·경제적 특 성을 조사하기 위하여 기술 관련 통계 자료가 활용되었다. 또 한 inputorientation data envelopment analysis는 기술적 효율 성 점수를 추정하는데 사용되었다. 더불어 Tobit regression은 쌀 생산의 효율성 수준에 영향을 미치는 요인을 확인하는데 사용되었다.

연구 결과에 따르면 연구 결과에 따르면 연구 분야 농가의 평균 기술 효율은 0.67인 반면, 쌀 생산업자들은 같은 생산 수준을 유지하면서도 투입 비용을 33% 절감할 수 있는 것으 로 나타났다. 다케오주의 뜨람꺽군 및 키리웡군의 쌀 생산의 효율성 수준에 영향을 미치는 네 가지 요인이 발견되었다. 토 빗 회귀분석 결과, 가장의 성별, 가장의 주요 직업, 쌀의 재배 횟수를 포함한 요인들이 기술적 효율성에 긍정적인 영향을 미 치는 반면 심어진 종자와 품종은 기술적 효율성에 부정적인 영향을 미치는 것으로 추정되었다. 쌀 생산 규모의 효율성은 0.80로 분석되었으며, 이는 농민들이 적정 규모에 가깝게 운영 하고 있음을 나타낸다. 이는 생산관련 비효율의 원인이 투입 자원의 잘못된 배분이나 부적절한 사용으로 인한 것임을 의미 한다.