The wide application of the super-efficiency method included sensitivity testing, identification of outliers, and a method for overcoming the bounded range problem in the second stage of linear regression.Hence, instead of the Tobit regression, we can use standard ordinary least squares regression methods in the next stage.In paddy production, the reduction strategy of inputs will not always be presented in a proportional way like it does in traditional DEA.Hence, if scientists want to evaluate the performance of DMUs using the efficiency scores as the unique index, the radial DEA approaches may mislead their decision.The reason for this problem is because radial DEA ignores the important slacks in presenting the efficiency scores.Meanwhile, the nonradial slack-based measure considers both radial and non-radial slacks.It also considers the overall efficiency of rice farms, segregates the assumption of proportionate changes in inputs, and directly calculates the slacks of each farm.The SBM model, similar to radial DEA model, has three forms in orientation: input, output, and non-oriented.An input-oriented SBM approach sets a target to reduce excessive input usage while keeping the existing output levels stable.The number of SBM and super SBM applications in agriculture, especially in paddy production, is limited.Since the Vietnamese rice sector has limited resources, we applied an input-oriented SBM model to focus on sustainable strategies through efficiency scores and input slacks.The assumption of variable returns to scale will be employed instead of the constant returns to scale because not the cultivators could operate their production at the optimal scales.For the definition of non-radial SBM model, please refer to Appendix 1 and the works of Tone.The following section is a brief description of super-SBM model, cited from Tone.The MKD region can be classified into six agro-ecological zones, including the freshwater alluvial zone, the plain of reeds, the long Xuyen Quadrangle zone, the Trans-Bassac depression zone, the coastal zone, and the Ca Mau Peninsula zone.We conducted a survey2 in the following provinces: An Giang province , Dong Thap province , Can Tho city , and Bac Lieu province.

These four provinces were selected for the household survey because their locations cover over four ecological zones, they have good agricultural extension systems, and effective operations in their paddy cooperatives.In the crop year 2018, the paddy areas of An Giang, Dong Thap, Can Tho, and Bac Lieu provinces in the summer-autumn and autumn-winter were 388, 314.7, 154.8 and 102.9 thousand ha, respectively.Regarding total rice volume, An Giang ranked first with 2.19 million tons,indoor growers followed by Dong Thap.Can Tho and Bac Lieu provinces produced 835.4 and 596.7thousand tons of paddy rice, respectively.Primary data were collected using structural questionnaires.We interviewed 100 paddy producers in each province on the following parameters: households’ demographic information, paddy production activities, application of climate smart agriculture 1M5R,4 3R3G,5 integrated pest management, and AWD practices.SA and AW6 are the two main crop seasons in the production structure in these localities.The DEA model has been conceptualized as an activity analysis model by many scientists.An activity analysis model is initiated by describing the different activities or processes in an organization.In a paddy farm model, products are made up of primary factors, such as labor, seeds, fertilizers, and other capital.In addition, there is the question of how densely to distribute seeds and fertilizers, and how to manage the labor between different production steps.In each situation, the constraints of these inputs reflects the available resources.Thus, DEA models are considered as activity analysis models, in which information about the activities is provided via actual observations.The authors conducted a DEA super– SBM model with one output and ten inputs to calculate the efficiency scores, input slacks, and rank the most efficient farms among the sample units.The output was defined as the total plain rice yield for the crop season.The ten inputs comprised four inputs with physical units, including seed sown density, N, P2O5, and K2O in kilograms per hectare.The remaining six inputs, expressed in monetary units, were ex-penditure on pesticides, herbicides, irrigation, hired labor, land preparation, and harvesting.The family labor for rice farming activities is measured by the number of working hours per hectare in each season.There was no significant correlation between these input variables used in the super-SBM model.After dropping some households that lacked information on production steps and those that had extremely low output, we conducted an analysis on a sample of 380 paddy households from four provinces: An Giang, Dong Thap, Can Tho, and Bac Lieu.Each household was coded with a number based on the province.In the study area, farmers produce three categories of rice varieties: aromatic rice , high quality rice , and medium quality rice.The AR group is good quality rice, characterized by long slender grains, intermediate amylose, intermediate gelatinization temperature, high elongation ratio, and strong aroma.In the SA season, 49.74% of farmers choose to plant this rice group, which includes the Dai Thom 8, Nang Hoa 9, Jasmine 85, and RVT varieties.This figure increases to 65% of the total producers in the AW season.The two AR varieties that were used predominantly in this study were Dai Thom 8 and Nang Hoa 9.The Dai Thom 8 variety has a low amylose content of 16.29%, and the cooked rice has a soft and fragrant taste.This variety is adaptable to many ecological zones, including saline areas and areas with aluminum content in the soils.The Nang Hoa 9 variety was bred from Jasmine 85 and AS996 varieties.It is disease resistant andtolerant to acidic sulfate soil.

The rice product from Nang Hoa 9 is soft and has a sweet taste with a pineapple leaf aroma.The next rice group is HR, which accounted for 48.42% and 30.78% of the sample in the SA and AW seasons, respectively.This is a group of rice varieties that are selected, bred, and released by the Mekong Delta Rice Research Institute.These OM rice varieties are commonly grown in the MKD, with high yields and good quality.They are also widely adapted to different agro-ecological zones and tolerate acidic and saline soils.OM varieties have long grain rice and a good appearance for export.In our study, OM5451 was the most widely cultivated HR among farmers.The final group of varieties is of medium quality, with sticky rice CK92 and chalky grain IR50404, accounting for only 1.84% and 4.21% of the producers in the two seasons, respectively.The IR50404 variety originated from the International Rice Research Institute and was imported into Vietnam in the early 1990s and released in 1992.This variety is popular among MKD farmers.Due to its high amylose content and chalkiness degree, IR50404 is not suitable to international tastes and thus has a very low commercial value.The main use of this variety is to prepare rice flour.Fig.3 shows the proportion of these varieties cultivated in the four provinces within the study area.In the SA season, farmers in An Giang province mainly produce OM5451 and DT8, while Bac Lieu’s farmers focus on aromatic DT8 and NH9 production.A similar proportion also occurs in households in the Dong Thap province.A total of 91% of farmers in Can Tho city cultivate OM5451 for domestic and export demands because of its high quality and resilience to pests and diseases.In the AW season, when weather is more advantageous, some households in the study area switch their production from HR to AR; in particular, farmers in An Giang and Can Tho changed from OM5451 to DT8.The production in Bac Lieu province does not change significantly since farmers mainly produce aromatics DT8 and NH9.The proportion of varieties in Dong Thap Province appears similar to the SA season.However, this information is only for reference, and the three main groups AR, HR, and MR will be used for analysis in the generation of comparative efficiency scores.The Vietnamese government has recently established a strategy for rice production in the period 2021−2030 by reducing the export volume and focusing on product quality.In this agricultural restructuring policy, by 2030, the proportion of AR, HR, and MR will account for approximately 25%, 40%, and 10% of the total farming area, respectively.The important HR varieties for export, such as OM5451, OM6976, OM7347, and OM4900, should account for 50– 60% of the total cultivated area.In addition, there should be restrictive breeding of sticky rice and MR varieties such as IR50404 or OM576.However, bridging the economic benefits and environmental issues of these rice groups remains an open question.The descriptive statistics of the output and input items used to estimate efficiency scores are presented in Table 2.Households in this study had an average land size of 2.77 ha, with the smallest size being just 0.26 ha and the largest one was 35.1 ha.In general, paddy farms obtained about 6.13 ton/ha and 6.11 ton/ha in SA and AW seasons, respectively.The HR group gave the highest yield at 6.46 ton/ha in the rainy season, while the AR only obtained 5.83 ton/ha.In the AW season, the AR group obtained a yield of 6.14 ton/ha, danish trolley which was 170 kg more than that of the high-quality rice group, and less than the medium-quality rice output at 6.55 ton/ha.

Regarding the inputs used, the seed sown density was relatively high in the two seasons, at more than 160 kg/ha.AR producers applied seeds at 145 kg/ha in SA and 154 kg/ha in the AW season, which was significantly lower than the density of high-quality and medium rice varieties.The amount of fertilizers used, such as nitrogen, phosphorus, and potassium was moderate and similar across the two seasons.In the SA season, farmers used slightly higher amounts of nitrogen fertilizer.The quantities of the other nutrients were approximately 73 kg/ha of phosphorus and 55 kg/ha of potassium.Importantly, when cultivating AR in the wet season, farmers consumed significantly less nitrogen than HR and MR.As mentioned above, GHG increases when more N fertilizer is consumed in paddy fields.Thus, this is a good signal to develop aromatic rice varieties that simultaneously achieve high yield, obtain good selling price, and mitigate GHG emissions owing to a lower consumption of N fertilizer.AR households also used significantly less potassium than the other two groups during the SA season.Among production cost items measured, the purchase of pesticides and herbicides was the costliest with an average of 4.1 and 3.8 million VND/ha in SA and AW seasons, respectively.Being aware of the high risks of pests and diseases, AR farmers have to pay more for pesticides than HR and MR.Paddy farms had much higher irrigation costs,which included pumping and drainage activities in AW season at 1.1 million VND/ha when compared to SA wet season at 761,000 VND/ha.The opposite was true of land preparation costs.Between the two seasons, there was not much difference in hired labor costs and family labor working hours.The results of the input-oriented super-SBM model under variable returns to scale are presented in Table 3.The overall efficiency of paddy farms in the MKD was high, and the AW season showed a more efficient estimate at 0.915 than the SA season at 0.875.The evidence showed that households that grow MR varieties in the SA season achieved the highest SBM score at 1.045 followed by AR and HR.However, the MR score decreased slightly to 0.911 in the AW season.In the AW season, some households switched from HR to AR rice production, and the efficiency scores of these two rice segments increased from 0.914 to 0.917 and from 0.830 to 0.909, respectively.This increase in overall efficiency occurred owing to the advantageous climate conditions with drier weather and less humidity.However, we failed to indicate a statistically significant difference in the overall efficiency scores between the three groups of rice.In Table 4,super-SBM efficiency scores are presented in terms of farm characteristics, including farm size, cooperative membership, 1M5R, and AWD farming practices.In the SA season, paddy farms larger 7than 2 ha achieved significantly higher scores than those that were smaller than 2 ha.Households with cooperative membership also had significantly higher efficiency than non-member households.There are certain advantages for cooperative members, such as technical training support, access to an agricultural extension service, pilot and pioneering CSA programs, and project participation.In local areas, paddy cooperatives also have contracts with input suppliers and machinery services, such as tractors or combine harvesters.Therefore, cooperative members have the opportunity to operate their production efficiently and manage their input use more effectively.In addition, in this SA season, farmers who practiced CSA technology obtained significantly higher efficiency scores.The 1M5R adopters had a super SBM score of 97.8% compared to conventional farmers at 81.6%, while the score of AWD adopters was 1.085 and that of non adopters was 0.835.Thus, it can be concluded that cooperatives and advanced farming packages lead to more efficient field management by members/participants in the disadvantageous weather conditions of the SA season.