The Impact Of Alternative Wetting And Drying Technique Adoption On Technical Efficiency.doc

Dịch vụ viết thuê đề tài – KB Zalo/Tele 0917.193.864 – luanvantrust.com
Kham thảo miễn phí – Kết bạn Zalo/Tele mình 0917.193.864
UNIVERSITY OF ECONOMICS ERASMUS UNVERSITY ROTTERDAM
HO CHI MINH CITY
VIETNAM
INSTITUTE OF SOCIAL STUDIES
THE NETHERLANDS
VIETNAM – THE NETHERLANDS
PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS
THE IMPACT OF ALTERNATIVE WETTING AND DRYING
TECHNIQUE ADOPTION ON TECHNICAL EFFICIENCY:
EMPIRICAL EVIDENCE FROM RICE PRODUCTION IN MEKONG
RIVER DELTA, VIETNAM
BY
HUYNH NGOC SONG MINH
MASTER OF ARTS IN DEVELOPMENT ECONOMICS
HO CHI MINH CITY, DECEMBER 2017

UNIVERSITY OF ECONOMICS INSTITUTE OF SOCIAL STUDIES
HO CHI MINH CITY THE HAGUE
VIETNAM THE NETHERLANDS
VIETNAM - NETHERLANDS
PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS
THE IMPACT OF ALTERNATIVE WETTING AND DRYING
TECHNIQUE ADOPTION ON TECHNICAL EFFICIENCY:
EMPIRICAL EVIDENCE FROM RICE PRODUCTION IN MEKONG
RIVER DELTA, VIETNAM
A thesis submitted in partial fulfilment of the requirements for the degree of
MASTER OF ARTS IN DEVELOPMENT ECONOMICS
By
HUYNH NGOC SONG MINH
Academic Supervisor:
DR. LE THANH LOAN
HO CHI MINH CITY, DECEMBER 2017

DECLARATION
I hereby declare that this thesis entitled “The impact of alternative wetting and
drying technique adoption on technical efficiency: empirical evidence from rice
production in Mekong River Delta, Vietnam” has been completely written by
myself. The study is the result of my own work combined with supervision and
guidance from Dr. Le Thanh Loan of University of Economics, Ho Chi Minh
city, Vietnam. I guarantee that the results with all suggestions in this study are
fully based on my personal work and knowledge which are strictly followed the
disciplines of Vietnam Netherlands Programme. This study, or any related
documents of this dissertation, has certainly not been submitted for any previous
qualifications or any other institutions and resources. I am also responsible for all
the contents in this research.
Date: 07 December 2017
Signature: _______________ Full name: Huynh Ngoc Song Minh
i

ACKNOWLEDGEMENT
The past two year with Vietnam – the Netherlands programme has been such a
memorable and special experience in my life. I feel truly thankful for all of the
knowledge and skills that I have the chance to learn which are extremely
important for me to complete this thesis successfully.
First and foremost, I would like to express the deep gratitude to my supervisor,
Dr. Le Thanh Loan. It has been an honor to be her only master student in
Vietnam and the Netherlands programme 22nd
course. She has been sharing with
me the integrant researching experience from collecting data to completing
thesis. I appreciated all of her contributions of time, ideas, dedicated guidance
and support during my thesis process. The enthusiasm that she has for this project
was extremely motivational for me, even during tough times in this master
journey. I am also thankful for the excellent example she has provided as a
successful woman economist and professor.
Secondly, I would like to thank the funding from FAO and CGIAR for the
project titled "Documenting Adoption of the AWD Water Management
Technique in Vietnam" in the MRD, Vietnam in 2016.
Furthermore, I also want express my appreciation to Prof. Dr. Nguyen Trong
Hoai, Dr. Pham Khanh Nam and all the lecturers as well as the entire associates
of Vietnam – the Netherlands Program for their dedication and willingness to
support all students in my class. Especially, I would like to thank Dr. Truong
Dang Thuy and Dr. Le Van Chon for their valuable suggestions which help me to
complete my thesis. In addition, I am extremely appreciative the valuable time
with my classmates in course 22, particularly, all of the members in my study
group, for their encouragement and cooperation during the course.
After all, I want to express how valuable it was to me for receiving the strongest
encouragement and support from my beloved family, especially my mom.
Because all of their sacrifices which generate the best conditions for me to finish
this program and this thesis.
ii

TABLE OF CONTENTS
DECLARATION .................................................................................................... i
ACKNOWLEDGEMENT ..................................................................................... ii
TABLE OF CONTENTS ...................................................................................... iii
LIST OF FIGURES ................................................................................................ v
LIST OF TABLES ................................................................................................ vi
ABSTRACT ......................................................................................................... vii
ABBREVIATION ............................................................................................... viii
CHAPTER 1: INTRODUCTION .......................................................................... 1
1.1 Problem Statements ...................................................................................... 1
1.2Research Objectives ...................................................................................... 8
1.3 Scope of the study ......................................................................................... 9
1.4 Structure of the thesis ................................................................................. 10
CHAPTER 2: LITERATURE REVIEW ............................................................. 11
2.1 Overview about the AWD technique .......................................................... 11
2.1.1 AWD definition .................................................................................... 11
2.1.2 AWD guideline in Viet Nam ................................................................ 13
2.1.3 The AWD score .................................................................................... 15
2.1.4 The impact of adopting AWD .............................................................. 16
2.2 Overview about technical efficiency of production function ..................... 19
2.2.1 Theory of frontiers production technical efficiency ............................. 19
2.2.2 Empirical Technical Efficiency Review .............................................. 21
2.2.3 Review of Determinants on Technical Efficiency ............................... 22
2.3 Summary ..................................................................................................... 23
CHAPTER 3: DATA AND METHODOLOGY ................................................. 25
3.1 Methodology ............................................................................................... 25
iii

3.1.1 Conducting the AWD Score ................................................................. 25
3.1.2 Analytical Framework .......................................................................... 28
3.1.3 Econometrics Model ............................................................................. 31
3.2 Data ............................................................................................................. 34
CHAPTER 4: RESULTS AND DISCUSSION ................................................... 37
4.1 Descriptive Statistics .................................................................................. 37
4.1.1 Data Description ................................................................................... 37
4.1.2 Correlation Matrix ................................................................................ 40
4.2 Empirical Results ........................................................................................ 43
4.2.1 The AWD adoption degree and challenges for AWD adoption .......... 44
4.2.2 Results of average technical efficiency of rice production in the MRD
region ............................................................................................................. 47
4.2.3 Results of determinants on the technical inefficiency .......................... 48
4.3 Discussion ................................................................................................... 51
CHAPTER 5: CONCLUSION ............................................................................. 53
5.1 Main findings .............................................................................................. 53
5.2 Policy implications ..................................................................................... 54
5.3 Limitations .................................................................................................. 54
REFERENCES ..................................................................................................... 56
APPENDIX .......................................................................................................... 63
iv

LIST OF FIGURES
Figure 2.1: The isoquant for technical efficiency estimation from the input-
oriented.................................................................................................................20
Figure 2.2: The frontier for technical efficiency estimation from the output-
oriented.................................................................................................................20
v

LIST OF TABLES
Table 3.1: The synthesis of signals that farmers use to observed water level on
the field during irrigation process ........................................................................26
Table 4.1: Descriptive Statistics...........................................................................38
Table 4.2: Correlation Matrix ..............................................................................41
Table 4.3: Variance inflation factor .....................................................................42
Table 4.4: Correlation Matrix ..............................................................................43
Table 4.5: Percentage of different AWD adoption level .....................................44
Table 4.6: AWD adoption score by provinces.....................................................45
Table 4.7: Challenges of AWD adoption in MRD province................................46
Table 4.8: Estimated Average Technical Efficiency ...........................................47
Table 4.9: The technical inefficiency determinants model ..................................48
Table 4.10: Akaike's information criterion and Bayesian information criterion . 50
Table A1: Comparison between estimated results between half-normal and
truncated distribution efficiency model ...............................................................63
vi

ABSTRACT
One of the most serious issues that potentially lead to total rice yield losses
is climate change and its consequence, water scarcity. To counteract with this
problem, the International Rice Research Institute has developed and promoted
the alternate wetting and drying (AWD) water saving technique among rice
growing countries to save irrigation water as well as enhance productive
cropping. However, after widely adopted, farmers have adjusted the technique
differently in term of irrigating schedule and practice. These realities lead to a
problem in measuring the degree of AWD technique adoption at farm level and
investigating its impact on rice production. From the original AWD score, this
study suggests a modified AWD score including water drainage practice to
represent for the adoption degree of each farm, based on that AWD application
impact on the rice production technical efficiency is also evaluated. Using the
sample of 250 farms surveyed in Mekong River Delta provinces, the adjusted
AWD score is calculated for each farm. Subsequently, a Stochastic Frontiers
Cobb-Douglas production function is regressed using maximum log likelihood
method to measure the technical inefficiency, after which, a function of technical
inefficiency determinants is investigated, where AWD score was included as a
main factor. Results indicate that higher AWD application degree can improve
technical efficiency of the production. Thus, AWD technique should be
continually promoted on large scale adoption and strictly followed IRRI
instructions to improve rice production technical efficiency.
Key words: Alternative wetting and drying technique (AWD), Technical
efficiency, Mekong River Delta, Vietnam.
JEL: Q12, Q15
vii

ABBREVIATION
PH – Power of hydrogen
AWD – The Alternate Wetting and Drying
Technique IRRI – The International Rice Research
Institute CH4 – Methane
MRD – Mekong River Delta
1M5R – One must do and five reduction campaign
KPA – Kilopascal
KG – Kilogram
DAS – Days after sowing
FGDS – Focus groups discussions
KIIS – Key informant interviews
VIF – Variance inflation factor
BIC – Bayesian information criterion
AIC – Akaike information criterion
CM – Centimeters
viii

CHAPTER 1: INTRODUCTION
In this chapter, firstly, a brief overview of the problem setting is provided
and based on that, the research problem of this study is given. Also, the research
questions and main objectives are described together with a short introduction of
the data and methodology used for this study. Finally, the structure of the
research is included.
1.1 Problem Statements
According to the Food and Agriculture Organization of the United Nation,
nature provides us more than 50,000 edible plants, however, only three of them,
which are rice, maize and wheat are considered as the world leading staple food.
The main reason is because these three directly provide over 60% of energy and
42% of calories intake for the entire human population. Out of these three, rice is
of the most important role. Rice feeds almost half of the human being, especially
in low and middle – income countries. People depend mostly on rice in their
daily meals. As the world population is growing rapidly, it would lead to the
increasing food demand in the near future (Easter, Rosegrant et al. 1998). The
major supply for food comes from agricultural products, especially rice as
proved, and the total rice consumption, in the coming year, is also expected to
increase.
In fact, the Food and Agricultural Policy Research Institute has projected
that the global demand for rice consumption will arise from 439 million tons in
2010 to 496 million tons in 2020 and reach 555 million tons by 2035. The
predicted upward trend in the global rice consumption can be observed through
actual data around the world. Firstly, rice is mainly consumed in Asia. This
region accounted for 90% of the total world rice consumption. Although per
capita consumption in China and India declines continuously because of
increased income and a rapid urbanization, Asia also contributes 67% of the total
increase. The Asia rice consumption of 388 million tons in 2010 will level up to
465 million tons in 2035. Secondly, outside Asia, where rice has not become a
1

staple food yet, per capita consumption shows the same increasing trend.
Particularly, in Africa, rice is the fastest growing food, both urban and rural
residents here used to eat rice only in their special occasion, but recently, rice has
become their daily food. As a consequence, an arising demand of 30 million tons
more will be needed by Africa. Rice consumption will surge 130% from 2010
and remain growing onwards. There is a gap between demand and supply of rice
in Africa, moreover, this continent accounted for 32% of global rice trade in
2015, importing 14.3 million tons of the total 44.6 million tons traded worldwide.
In the Americas, total rice consumption is also projected to rise by 33% over the
next 25 years as a result of steadily increasing incomes, as well as continued
population growth. Even in the Middle East and developed European Countries, a
significant increase of rice consumption was observed. This partly causes by
migrants from countries where rice is more often consumed, along with wider
globalization of food availability and tastes.
Generally, the demand for rice continues to rise, and for every one billion
people added to the world’s population, 100 million more tons of rice is needed
to be produced annually. While rice consumption is increasing around the world,
most of its production only centers in Asian countries. According to Food and
Agriculture Organization of the United Nation, the top 10 rice producing
countries in the world today are India, China, Indonesia, Bangladesh, Thailand,
Vietnam, Burma, the Philippines, Cambodia, and Pakistan. To meet the
increasing demand, global rice yields now must rise faster than the past to keep
the world market prices stable at affordable levels for the billions of rice
consumers. However, with the current state of slow productivity growth,
inefficiency production and unsustainable management of natural resources,
expansion of the rice production would be limited. Furthermore, the International
Food Policy Research Institute forecasts that by 2050 rice prices will increase
about 32% to 37% and the yield losses in rice could be 10% to 15% as a result of
climate change. Phenomenon as sea-level rise causing flooding, salinity, and
water scarcity will be negatively affected rice production.
2

The first problem is the sea level. When the sea level rises as predicted, a
large area of low – lying lands, deltas and coastal areas in Asia will be
submerged, leading to salinity throughout the region and making rice production
become vulnerable. For instances, in all of the hydrology system of the Mekong
River Delta (MRD), one of the primary rice growing area in Vietnam will be
damaged, sediment discharge and shoreline gradient will change. Flooding is
also caused by rising sea-level, rice cannot survive if they are submerged under
water, and flooding makes it difficult for harvesting. Currently, about 20 million
hectares of the world’s rice growing areas is at risk of occasionally being
flooded, particularly in major rice growing area as India and Bangladesh.
Generally, risen in sea-level would mainly reduce quality, size of cultivated land,
and the amount of irrigated water.
The second major problem in rice production is water scarcity. Water is one
of the most important inputs for rice production, in fact, without water rice
cannot grow. Rice systems depend on their ecological resilience largely from
intensive water use in order to control weed, soil salinity, pH, and to avoid heat.
Water for agriculture uses around the world are becoming increasingly scarce
(Rijsberman 2006). Scarcity irrigated water source is mainly caused by reduction
of water resources and quality, malfunctioning of irrigation systems, and
increased water use competition from other sectors such as urban and industrial
users. In conclusion, one of the most important issues in rice production
nowadays is water scarcity.
Asia, the world biggest rice production region, has been experienced long
developing history of rice production, and for more three millennia, their
irrigation system presents sustainable condition. Nevertheless, recent rapid
population growth that leads to a declining share of land, water, labor, energy
resources and overuse of production inputs made more than 23 million hectares
of rice production areas in South and Southeast Asia facing water scarcity. Tuong
and Bouman (2003) estimated that by 2025, 2 million hectares of Asia's irrigated
dry season rice and 13 million hectares of its irrigated wetland rice may
experience “physical water scarcity” and the rest of the approximately 22 million
3

hectares of irrigated dry season rice in South and Southeast Asia may suffer from
“economic water scarcity”, which results from competing water uses and climate
change. Also, in northwestern India, declining groundwater levels will pose a
serious threat to one of the world’s largest grain basket. These challenges rise in
rice production together with the increase demand in rice consumption require
agriculturists to rethink about the current management paradigms and finding
new solutions to prevent and address water scarcity condition.
In order to alert this future scenario, numerous efforts are made to develop
new water saving technologies for rice production. Different new technology
suggestions are the alternate wetting and drying technique (AWD) (Bouman and
Tuong 2001), continuous soil saturation (Borrell, Garside et al. 1997), irrigation
at fixed soil moisture tensions varying from zero to 40 kPa (Sharma, Bhushan et
al. 2002, Singh, Choudhury et al. 2002), or irrigation at an interval of one to five
days after disappearance of standing water (Chaudhary 1997). These water
management practices are called partial aerobic rice systems. These techniques
not only bring hope for rice farmers suffered from water scarcity but also save
water from rice production for other economic or environmental purposes. The
Integrated Rice Research Consortium has recapitulated, researched, and
completed these water management practices, and launched these worldwide for
all of the rice producing countries.
Actually, keeping the farm non – continuously flooding practices have been
used for several decades as a water saving method, but in many cases, farmers
were following an uncontrolled or unplanned watering practice. After the
intervention of IRRI, from then, among all of the water saving techniques, the
alternate wetting and drying technique (AWD) is one of the most significantly
and widely adopted technique. The definition of AWD is introduced by IRRI as
follow, “Alternate Wetting and Drying (AWD) is a water-saving technology that
farmers can apply to reduce their irrigation water for rice fields without
decreasing its yield. In AWD, irrigation water is applied a few days after the
disappearance of the ponded water. Hence, the field gets alternately flooded and
non – flooded. The number of days of non-flooded soil between irrigations can
4

vary from one to more than 10 days depending on the number of factors such as
soil type, weather, and crop growth stage.”
At first, farmers practiced ‘forced’ AWD early in 2006 among the region of
Angat Maasim River Irrigation System. After that, some practices that keep non-
flooded conditions in the rice field for short interval of growing days become
commonly for about 40% of rice farmers in China and more than 80% of rice
farmers in North Western India and Japan (Richards and Sander 2014). However,
nowadays farmers follow a ‘safe’ AWD in which they maintain the threshold of
15 centimeters subsurface water level for the next time they pump water and
flooding the field. (Lampayan, Palis et al. 2009). This method has also been
recommended method for rice areas which faced scared irrigation water status in
South and Southeast Asia. In Philippines, safe AWD is firstly adopted at the
Tarlac Province since 2002, with farmers who use deep – well pump irrigation
systems (Lampayan, Palis et al. 2009). Nowadays, the International Rice
Research Institute (IRRI) has been promoting alternate wetting and drying as a
smart water saving technology for rice cultivation through national agricultural
research and extension its adoption mainly in Bangladesh, the Philippines, and
Vietnam.
After years of adoption, IRRI has worked in partnership with national
research institutions to conduct researches studied about the impact of AWD in
order to develop this technique. Generally, in economical dimension, a vast
majority of study showed that, certainly, AWD can reduce the amount of water
input or water cost. However, the impacts of AWD on the yield across regions
and countries are inconsistent. In some cases, researches stated that AWD
technique adoption does not reduce the amount of total yield, while others
suggested that this technique increases the amount of total yield and the
remaining concluded that AWD technique can result in total yield losses. The
reason behind those vague impacts of adopting AWD technique on the yield is
mainly because the famers apply AWD differently. Nevertheless, under the new
and simple practices of “safe AWD” suggested by IRRI, researchers hope that in
general AWD could bring positive economic impacts for rice production.
5

Another influence of AWD technique in cultivation is on environmental
dimension, scientists reported that applying AWD could also reduce the
greenhouse gases emissions and saving the water for the environment. Overall,
AWD and safe AWD has been field tested and validated by rice farmers in
Bangladesh, Indonesia, Laos, Philippines, Myanmar and other countries. The
technique has also been proved to offer potential reducing yield gaps, increasing
rice production, protecting the environment which finally can generate positive
benefits for both rice farmers and society at large. Consequently, AWD technique
is now centered in many extension efforts by formal institutes and non –
governmental organizations across number of Southeast Asia countries. Materials
for training and extending purposes on AWD are also being widely added in
numerous agricultural colleges, universities and opened certification plans.
In Vietnam, Agriculture has always appeared as one of the key sector of the
economy. Over the past decade, Vietnam agriculture has significantly developed
which helps Vietnam became the world top exporters in rice, rubber, coffee,
pepper, cashew nuts and other agricultural products. According to the Vietnam
General Statistics Office, among all the agriculture products, paddy is the most
important crop in Vietnam, occupying for almost 40% of gross output of
agriculture sector. Vietnam is also one of the top rice productions, the second
biggest rice exporters worldwide only after Thailand. Although Vietnam’s
climate was said to be suitable for cultivation activities, the nation’s agricultural
products are judged as low quality in comparison with other countries, especially
Thailand. The main reason is because Vietnam agricultural development is
fundamentally based on exploiting the natural resources rather than based on
technology. The future perspective of rice production in Vietnam is not bright. In
the latest estimation of General Statistics Office Vietnam, the country’s total rice
yield in 2016 is 43,6 million milled tons, reduce 4% compared with the amount
of 2015, which are mainly due to water scarcity, high level of salinity, serious
storm and flooding. If agriculturalist in Vietnam and the government does not
making the best effort to improve this current state, Vietnam rice production will
face many obstacles in the near future.
6

The Mekong River Delta is one of the major rice production areas of
Vietnam. This area accounts for more than 50% of the total products during main
cultivation season (Winter – Spring). However, this area is also suffering under
the general threat of salinity and late, uncertain rainy season, which leads to
water scarcity and reduces productivity. In 2016, the water source cannot supply
enough for irrigation system and high level of salinity pose a significant 10%
reduction of the total rice output, making the productivity in main season falling
back to 6.4 tons per hectares and the trend is projected to remain downwards in
the following year. Reduction in total yield also leads to high level of rice price.
In order to address these problems in rice production for the whole country
including MRD area, the Vietnam Plant Protection Department of the Ministry of
Agriculture and Rural Development has partnered with the Irrigated Rice
Research Consortium to adopt new water saving technology and orient the new
strategy for rice production. Through this partnership, alternate wetting and
drying technique was introduced and implemented in Vietnam incorporated with
various campaigns. One of the most popular campaigns is “One Must Do, Five
Reductions” (1M5R) program launched since 2009. After years of applying, each
of these contents belonged to the program has been successfully promoted and
applied nationally to improve rice production including AWD water saving
technique. Recognizing the benefits that can be derived when AWD is widely
adopted, in 2011, Vietnam’s Ministry of Agriculture and Rural Development
again highlighted AWD as one of the improved cultivation techniques for rice
production to be implemented broadly through 3.2 million hectares of rice
cultivation areas by 2020. With this policy support, the adoption of AWD
continues to be mainstreamed in different programs of Vietnam’s Ministry of
Agriculture and Rural Development. However, the current state of AWD
adoption still contains three particular problems:
Firstly, there are no clear evidences about the impact of widely adopted
AWD technique on rice production in Vietnam to convince the famers and policy
makers in implementing this technique more resolutely. In fact, although, it has
been broadly applied and suggested from agricultural policy for years, Vietnam’s
7

rice production is still facing challenges of salinity and water scarcity, which
leads to declined and erratic trend in total output.
Secondly, there are many researches investigate the impact of AWD
technique around the world, but the major was experimental field tests or
descriptive studies instead of econometrical technique or model. In fact, both
experimental fields test and descriptive studies has some limitations. For
experimental field tests, scientists only study about the AWD impact under
strictly adopted conditions. While AWD in widely implement does not exactly
follow the instruction, Yamaguchi, Luu et al. (2016) observed that famers made
some modifications which indicate that they have adapted AWD for their local
farming conditions. In addition, as analyses from descriptive studies cannot be
that of persuasion compared to econometrical studies, then a regression model for
determining AWD technique impact on rice production is required.
Finally, AWD diversified adoption practices create the challenge in
determining the adoption degree in each farm compared to original AWD
instruction. IRRI has documented a measurement to address this issue which is
the AWD score suggest by Moya et al., 2004. Nevertheless, this measurement
has not considered water drainage which is an important cultivated practice
related to specified topography, hydrology characteristics in Vietnam.
Therefore, in the attempt to fill these gaps, initially, this thesis suggested a
modified score including water drainage effect to measure degree of AWD
technique adoption at farm level in Vietnam. After that, a regression model with
AWD score is suggested to determine the impact of AWD technique and provide
a clear evident to support agricultural policy that enhance AWD adoption on a
larger scale.
1.2 Research Objectives
This research would use the data of 2016 main crop season from 250
interviewed farms in MRD, Vietnam. The investigation center in answering the
question of how changing from traditional continuous flooding into alternative
wetting and drying irrigation would make contribution for rice producers in
8

MRD provinces economically. In details, this paper aims to address three main
objectives:
- Firstly, to evaluates the degree of AWD adoption in each farm using the
modified AWD score and analyze the challenges for AWD adoption in
different provinces of the MRD.
- Secondly, to measures the average technical efficiency of the rice
production in MRD provinces during main crop season in 2016 by using
the production function approach.
- Finally, to estimate the impact of AWD adoption and other determinants
on technical efficiency of rice production in MRD provinces.
1.3 Scope of the study
Under the study topic of: “The impact of alternative wetting and drying
technique adoption on technical efficiency: Empirical evidence from rice
production in Mekong River Delta, Vietnam.” A cross-sectional survey dataset
from a group of 250 rice producers from four main provinces in the MRD,
Vietnam is analyzed to address the research problems. The dataset includes
information about the household characteristics, farming conditions, watering
practices, inputs and outputs of each farms during main seasons (Winter-Spring)
of 2016. Initially, an adjusted AWD score is conducted base on the original
AWD score by Moya et al., 2004, combined with AWD definition introduced by
IRRI and water withdrawing practice from the famer. This score reflects the
extent of adoption for each farmer in MRD, Vietnam. Subsequently, the Frontier
Production Functions approach is applied to measure the Technical Efficiency of
these farms. Finally, AWD score and other key factors are examined to see their
impact on rice farm’s technical inefficiency.
Overall, this study is the first research that determines the impact of
adopting AWD technique widely in Vietnam by using a econometrical model.
Moreover, an adjusted measurement for AWD technique which including water
withdrawing practice is newly contributed for the literature review. Finally, while
numerous studies about AWD impacts are generated, this is also the first time
that AWD adoption effect on rice production’s technical efficiency is evaluated.
9

The findings of this study states that adopting AWD technique in large scale can
improve the technique efficiency of rice production. In specific, this means after
applying the AWD technique, producing rice become more effective and with
each level of input, famers can generate higher level of yield as the output.
1.4 Structure of the thesis
The study is organized as following manners. Chapter 1 is the introduction
which provides general view of the study and suggests the main research
problem, research objectives, the importance and structure of the research.
Chapter 2 presents general knowledge about AWD technique, from its definition
review, to standard AWD practice guideline, and its impacts studies from
previous researches. Continuously, the theory of technical efficiency and some
empirical studies measures technical efficiency in agriculture sector are
introduced. After that, review of technical efficiency determinants is given with a
summary to suggest our study gaps. Chapter 3 briefly describes the dataset, data
collection process, the analytical framework and the final empirical estimation
model. Chapter 4 exhibits the estimation results and discussions. The last chapter
is the conclusion.
10

CHAPTER 2: LITERATURE REVIEW
In this chapter, initially, the overview about the AWD technique including
the AWD definition, AWD standard guideline practices, AWD adoption
measurement and AWD impact is provided. The following part offered the
literature review of technical efficiency, the indicator that is used to analyze
AWD impact. Firstly, a review about the theory of the technical efficiency, is
mentioned and secondly, the estimation method, empirical studies and the
determinants of technical efficiency are described.
2.1 Overview about the AWD technique
2.1.1 AWD definition
Analyses of IRRI suggested that, rice required flooding condition to grow since
this type of plant absorbs more than twice of the water input amount compared to
other crops at field level. However, 60% to 80% of total water input for rice
production is actually unproductive, because through seepage and percolation,
the irrigated water can rejoin the groundwater or water downstream. Therefore,
reducing irrigation water to rice fields is reduction of unproductive seepage and
percolation water losses (Saleh and Bhuiyan 1995, Bouman and Tuong 2001, Li
and Li 2001, Tabbal, Bouman et al. 2002). Based on this insight, several water
management practices were created including the alternate wetting and drying
technique.
The alternate wetting and drying (AWD) technique for rice production is
one of the water management techniques that have been developed by IRRI over
years. The general idea behinds this technique is instead of keeping the field
continuously flooded as traditional cultivation, famers could wait for the soil to
dry out for one to several days after the disappearance of ponded water before it
is flooded again. In fact, this idea was originally developed from an Indian term
called “intermittent irrigation” (Sandhu, Khera et al. 1980). However, Stoop,
Uphoff et al. (2002) and Uphoff, Felske et al. (2001) suggested that a similar
11

discontinuous flooding irrigation method was included in the system of Rice
Intensification, an integrated farm management technique developed by the Jesuit
priest Father Henri de Laulanie in Madagascar since 1980. In detailed of the
system of Rice Intensification guideline, during the growing period, the field
should not be continually flooding, a moderate amount of water should be
regularly provided to maintain an integrated condition of aerobic and anaerobic
cultivation soil and before harvesting, lower level of pond water should be kept
on field surface.
Even though, initial concept of AWD appeared in other regions, recently,
AWD development has been centered in East and Southeast Asia. In fact, a
particular AWD guideline is conducted in each area, with a specific instruction
about schedule, duration, and frequency of non-flooded periods. For instance, Li
and Barker (2004) mentioned that since irrigation water became increasingly
scared in China as an effect of highly demand water for other uses, a form of
AWD is widely practice in some area very early. In this practice, water is refilled
in about a week for heavy soils and in about five days for light soils, the level of
irrigated water is about five to six centimeters above the field surface, during the
interval days between irrigations, water ponds disappear from field surface and
naturally dries through seepage and percolation. The implement instructions
might be slightly different between sites in term of the cultivation conditions,
habits, type of input (seed, water, land), and also the in term of expressions.
However, generally, these are usually very inflexible and complex, which is hard
for farmers to strictly adopt the technology, especially when they think of
potential yield loss. In order to simplified the recommendations, some
agriculturist express AWD definition as keeping the field non-flooded from one
to ten of days (Bouman, Humphreys et al. 2007). Other scientists suggest famers
to irrigate when the tension of the soil water in the root zone reached a threshold
value of 10 kPa (the index measured soil moisture level for the plant). Again, this
instruction was inconsequential for farmers as they do not have appropriate
equipment to soil measure level. In conclusion, obsession about potential yield
loss, unpractical and complicated instruction while the impact of water scarcity is
not physically or economically visible are obstacles for AWD broadly adoption.
12

Understand the limited of adopting and promoting AWD broadly. In 2002,
IRRI developed simpler instruction of AWD include with a practical tool to
allow farmers to reduce irrigation water input while maintaining yield. The
definition of safe AWD suggested by IRRI is:
“Alternate Wetting and Drying (AWD) is an irrigation technique in which
water is applied to the field a number of days after the disappearance of ponded
water. This is in contrast to the traditional irrigation practice of continuous
flooding (meaning never let the ponded water disappear). This means that rice
fields are not kept continuously submerged but are allowed to dry intermittently
during the rice growing stage. The number of days in which the field is allowed
to be “non-flooded” before irrigation is applied can vary from 1 day to more than
10 days.”
The mentioned practical tool is a 30 centimeters length water pipe. Under
safe AWD adoption a maximum irrigated water level of five centimeters above
the field surface and a threshold of 15 centimeters water dropped level below the
surface is empathized.
2.1.2 AWD guideline in Viet Nam
In Vietnam the standard AWD technique, is introduced in a guide book
published by the Sub – Department of Plant Protection in 2011. Actually, this
book provide instruction for all components in “one must do and five reduction
(1M5R)” national agriculture campaign. In 1M5R program, famers practice “one
must” is to use certificated seeds instead of poor quality seeds and “five
reductions” are to reduce the amount of sowed seed, agrichemicals, fertilizers,
irrigation water, and postharvest loss. According to Dinh, Chung et al. (2013),
this program was an effort of policy makers to increase net returns by removing
inefficiencies in rice production. 1M5R campaign suggests famers applying
AWD technique, specific instructions were provided to explain how to
implement AWD in Vietnam. AWD standard guideline in Vietnam stated that
farmers control their irrigation water level at a height of five centimeters above
the field surface and wait until the water drop to maximum level of 15
13

centimeters below the field surface for the next irrigation. In details, the
irrigation schedule to apply safe AWD for rice production is described as follow.
Firstly, farmers are recommended to use a water tube to observe the water
level and irrigate adequately. This water tube can be made of cheap and popular
material in Vietnam, for instance plastic water tube or bamboo. The water tube is
about 30 centimeters height with the diameter of around 10 to 15 centimeters so
that water level can be easily noticed from inside the tube. On the side of the
water tube, there are some 0.5 centimeters holes and scale marks. The farmers
can place more than four pipes over their fields to observe the water level.
Placement position should be careful while choosing to help the farmers
conveniently observed water level and control the irrigation. A recommendation
is that the position should be representative of the average water depth in the
field. Famer should put the pipe at the depth of 20 centimeters under the ground
then remove the soil inside it so water level can be observed. The water level
inside the pipe is the same as the level of water on the field.
Secondly, AWD technique’s implement schedule is instructed base on the
growing process of the rice. During the first seven days, the soil needs to be
moistened after sowing but also needs to avoid flooding. Fertilization should be
conducted at seven to ten days after sowing (DAS), the paddy should be flooded
to a depth of one to three centimeters of water level about the field in this period.
Water height should be continuously maintained at three to five centimeter level
during 10 to 20 DAS because flooded irrigation is necessary for rice growth
during this period and it also controls weeds. The second fertilization should be
conducted in 18 to 20 DAS. Rice is in vegetative growth during 25 to 40 DAS,
about 60% soil moisture is sufficient for prosperous growth. Thus, AWD should
be conducted during this period as in drying condition, the roots would go further
under the field surface to find water sources and the rice plants can grow stronger
with longer root which also help to avoid harvest losses. Additionally, rice is
prone to sheath blight disease during this period, shortening of the flooded
conditions by AWD restricts spreading of the pathogenic fungus (Rhizoctonia
solani). Third fertilization should be conducted during forty to forty – five DAS
14

and the water depth should be maintained at one to three centimeters this time.
Nothing significant is noted during forty – five to sixty DAS. The period of sixty
to seventy – five DAS corresponds to the flowering stage; rice requires large
volumes of water in this stage, thus, water depth should be continuously kept at
five centimeters. Farmers should drain the water from the paddy fields ten to
fifteen days before harvest. Draining the water before harvesting promotes rice
ripening and facilitates the machine harvesting operation.
2.1.3 The AWD score
As the definition of AWD is relatively general while the practices of AWD
are differently conducted among rice producing areas, people could hardly define
AWD technique adoption. Furthermore, when AWD is widely adopted, local
famers also do not strictly follow the guidelines but adjust the practice differently
in their own ways. For instance, Satyanarayana, Thiyagarajan et al. (2007)
indicated that some farmers found it possible to applied alternate wetting and
drying irrigation for the whole cultivation season and this practice could even
generate some benefit. Consequently, a measurement of AWD adoption is
required to determine the degree of AWD adoption for each farm.
Among the previous studies, there are two approaches to proxy AWD
adoption. The first one is using a dummy variable which equals to zero when the
farmers apply continuous flooding irrigation and equals to one when the famer
use AWD irrigation (Rejesus et al., 2011). Nevertheless, it is hard to distinguish
between AWD non-adopters and AWD adopters. Additionally, a dummy variable
could not be able to reflect the diversified adoption level among famers. The
second appropriate is the AWD score which is originally introduced by Moya,
Hong et al. (2004). This score formulates AWD adoption mainly based on the
irrigation times and irrigation water level. AWD score has been documented by
IRRI as an approach to measure AWD adoption at farm level and also has been
used by several researchers to investigate the AWD impacts (Moya, Hong et al.
2004, Mushtaq, Dawe et al. 2006, Li and Li 2010).
However, the degree of AWD adoption might not only depend on water
irrigation but also on water drainage. First of all, drainage helps to save water
15

because the amount of water drained could be reused downstream which is
similar to the case of water losses due to seepage and percolation in the study of
Hafeez et al. (2007). Moreover, drainage is strongly affected AWD adoption
level in the rainy season as the field is usually flooded due to rainwater while
AWD technique encourages famers to use less water and allow the disappearance
of ponded water on paddy surface a few days between irrigation times
(Yamaguchi et al., 2016). Withdrawing water is also useful in reducing CH4
emissions of rice production (Leon et al., 2015). Overall, these argument shows
that water drainage should also be added in estimation of the AWD adoption
degree.
Consequently, this study suggests an adjusted AWD score which also
includes the effect of water drainage practice. After that the modified AWD score
is calculated to represent the adoption level of AWD technique for each farm.
2.1.4 The impact of adopting AWD
From the original definition and suggestion of IRRI, AWD technique is
suggested to reduce an amount of irrigation water without affecting the yield. The
explanation for this impact is that rice still be able to absorb enough water
through its roots in some growing period due to the previous flooding stage.
However, other researcher reported that AWD technique can even increase grain
yield because this held the root to grow stronger (Tuong, BAM et al. 2005, Yang,
Liu et al. 2007, Zhang, Zhang et al. 2008) and AWD also help the soil become
drier and more stable, which enhanced harvesting if famer uses harvesting
machines (Richards and Sander 2014).
Studies about the AWD impacts have been conducted in many different rice
growing countries, but mostly these are field experimental analyses. In India,
Singh, Aujla et al. (1996) had conduct a field experimental research and reported
that AWD technique reduced the water uses by 40 to 70 percent compared to
traditional continuously flooding associated with no yield loss. In China, Rui-
Zhong, Wen-Chao et al. (2002) reported a reduction of irrigation water for 7 to
25 percent also with field experimental method. In Bangladesh, Palis, Lampayan
et al. (2012) reported that AWD reduces the number of irrigations by 28%,
16

irrigation costs by 20%, and increase in yield by 0.4 to 0.5 ton per hectare. In
China, (Pan, Liu et al. 2017) analyzed through field experiment method and
suggest that compared to traditional irrigation, the AWD adoption reduces 24%–
71% of irrigation water without resulting in yield loss. In Philippines, AWD can
reduce the amount of water input, up to 38% of the total initial water irrigated
without adversely affecting rice yields (Rejesus, Palis et al. 2011). Other regions
report that, this method increases water productivity by 16.9% compared with
continuously flood irrigation (Tan, Shao et al. 2013). High – yielding rice
varieties developed for continuously flood irrigation rice system still produce
high yield under safe AWD (Yao, Huang et al. 2012). This method can even
increase grain yield because of enhancement in grain – filling rate, root growth
and remobilization of carbon reserves from vegetative tissues to grains (Tuong,
BAM et al. 2005, Yang, Liu et al. 2007, Zhang, Zhang et al. 2008). AWD can
reduce the cost of irrigation by reducing pumping costs and fuel consumption
(Lampayan, Rejesus et al. 2015). According to the practice brief report of Food
and Agriculture Organization of the United Nation about the AWD technique,
this method can also reduce the labor costs by improving field conditions at
harvest, allowing mechanical harvest. AWD leads to firmer soil conditions at
harvest, which is suitable to operate machines in the field (Richards and Sander
2014). Therefore, AWD increases net return for farmers.
Since the definition of AWD adoption was not clearly defined in an official
way and AWD technique is only broadly adopted for a few recent years, number
of studies about AWD impact in large sample is relatively less than field
experiment method. Noticeably, in 2004, Moya et al. conducted the AWD score
to measure the general adoption of each farm. He also found that AWD can
reduce the water irrigation and imposed no impact on production profit for
famers in China. Rejesus, Palis et al. (2011) conducted a broadly study about
AWD in large scale, however, this paper is not clearly determined the AWD
adopter and the regression results is not able to control potential bias from the
unobservable selection. Findings of this research also suggest that AWD leads to
reduction in irrigation hours associated with no significant yield loss.
17

In Vietnam, studies about AWD have also not been widely conducted.
Overall, Bouman and Tuong (2001) suggested that AWD can reduce water use
for irrigation in Vietnam. Another study from Yamaguchi, Luu et al. (2016)
provided a general description of AWD adoption in An Giang Province, Vietnam
and reported that the practice in general is differently adjusted from standardized
AWD.
There are also other studies which investigate the environment impact of
AWD technique. For instance, AWD can reduce CH4 emissions (山口哲由, 南
川和則 et al. 2016). CH4 is produced by the anaerobic decomposition of the
organic material in the flooded paddy field. Since the water level is allowed to
drop below the soil surface, it removes the anaerobic condition for some time
until the next irrigation and stops the production of CH4 from the rice field. In
general, AWD reduces the total amount of CH4 released during the rice growing
season. This method has been assumed to reduce CH4 emissions by an average
of 48% compared to continuous flooding by the 2006 Intergovernmental Panel on
Climate Change methodology. Alternate wetting and moderate soil drying lessen
cadmium accumulation in rice grains (Yang, Liu et al. 2007) and AWD technique
can dramatically reduce the concentration of arsenic in harvested rice (Price,
Norton et al. 2013). All of these results can generate positive impact on rice
consumer health. This method can also reduce insect pests and diseases (Palis,
Hossain et al. 2005). Beside these advantages, IRRI also suggests that as water is
important for growing rice, uncontrolled practices of non-trained farmers can
reduce productivity. AWD might also enhance more weed growth in the crop
field.
However, in this study we mainly focus on the economic effect of AWD
technique adoption. Following the previous studies, AWD can affect the input
(reduce the water irrigation), because as long as the famers adopted AWD
technique, they reduce the irrigation time compare to the traditional continuous
flooding practices. As a result, studies about the impact of AWD on the amount
of irrigation water show a consistent reduction among different regions and study
18

methods applied. However, the impacts of adopting AWD technique on yield or
other output measurements (profit, income, net return) are relatively vague. The
reason is because when each famer applies this technique differently the result on
yield would be different. In this research, an adjusted AWD score is conducted to
determine degree of adoption of each famer, and the impact of AWD technique
on the technical efficiency of rice production after widely adopted is also
investigated for rice producers in the MRD province of Vietnam.
2.2 Overview about technical efficiency of production function
2.2.1 Theory of frontiers production technical efficiency
There are different ways to measure the performance of a production, which
represents the ability of a producer to covert inputs into outputs. A commonly
apporoach is “productivity” which represents the ratio between the amount of
outputs over inputs. Another concepts which is also widely used in production
analysises is “efficiency”. As these two concepts is closely related, if a
productive unit increase their productivity its production is more efficient,
confusion between these two concepts has always existed, however, productivity
and efficiency are not precisely the same.
The fundemental idea behind efficiency production, as explained by Farrell
(1957), is approached in two ways: Input-orientated and Output-orientated.
The input perspective measures the minimum amount of inputs that each
productive unit requires to produce a certain set of outputs. Figure 2.1
demonstrates a productive unit (A) using two inputs (X1 and X2) to produce
output (Y). The II’ line is called the isoquant which represents all of the
minimum set of inputs that is used to generate a given level of output. This
means on II’ the amount of inputs (X1, X2) is most efficient used to produce
output (Y). Farrell (1957) determined that the deviation of a practical production
point (A) from the isoquant relates to technical inefficiency of the production. As
a consequence, the ratio OB/OA is defined as technical efficiency of the firm
with the efficiency input per unit of output values determined at point B.
19

Figure 2.1: The isoquant for technical efficiency estimation from the input-
oriented
Source: Author
From the output perspective, Farrell presents the production frontier which
determines the possible maximum output level that is achievable from a certain
level of inputs, with a given existing technology. Figure 1.2 illustrates a
productive unit (A) produces an amount of output (Y) from vector of inputs (X).
The production frontiers f(X) curve shows all the maximum level of output (Y*)
that can be obtained by using each vector of inputs X. Farrell also suggested that
a productive unit is considered as technical efficiency when it operates on this
frontier and technical efficiency associated with each level of input is measured
by the ratio of Y/Y*.
Figure 2.2: The frontier for technical efficiency estimation from the output-
oriented
Source: Author
Generally, the output-oriented and input oriented provides the same
production frontier and determines the same set of efficient firms, only measured
20

efficiencies of the inefficient firms are may be different between two approaches
(Coelli, Rao et al. 1998). In some industries, where firms have a requested level
of output demand to supply, researchers often apply the input-oriented to
estimate the minimum level of input. By contrast, in those industries that have a
fixed level of resources, for instance, in agriculture, each farm only has a certain
area of cultivation land, the output-oriented is commonly adopted.
Base on the concept of technical efficiency, a vast body of literature review
has been conducted to evaluate agricultural production efficiency with different
approaches. The result of technical efficiency measurement shows the ability of
each farm site to produce the maximum amount of output from a give level of
input and resources. In addition, another relevant aspect which researchers are
highly interested about is to define the determinants of the inefficiency. When
these determinants are learned, these factors also show their impacts on
efficiency and their roles in the production.
2.2.2 Empirical Technical Efficiency Review
The Farrell efficiency theory has provided a strong foundation for
researchers to develop different empirical frontiers model and practical ways to
measure production efficiency. Generally, classification of efficiency can be
approached by different angles. For instance, based on the functional form
criteria, there are two types: Parametric and non-parametric estimation. The
parametric method applies a standard production function form such as Cobb-
Douglas, Translog, Quadratic and Normalized quadratic function. Among these
type, Cobb-Douglas (Hannesson 1983), and Translog (Squires 1987, Pascoe and
Robinson 1998) has been most widely used. In fact, Translog is a general form of
Cobb-Douglas. The remaining non-parametric method does not depend on any
specific functional form.
Form another perspective technical efficiency generally can also be
distinguished into deterministic estimation or stochastic frontiers estimation.
Specifically, deterministic model assumes all the errors term in the function or
the deviation from the frontier visually presented technical inefficiency, while
stochastic frontiers assumed the errors term is a sum of two component: an
21

idiosyncratic error term which is normally distributed and a technical inefficiency
with different distribution assumptions. The empirical review of studies using the
mentioned approach to measure technical efficiency in agriculture will be
specified in each following part.
Firstly, the deterministic production frontiers method can be divided into
two groups, studies using non-parametric estimations called the data
envelopment analysis (Wadud and White 2000, McDonald 2009) and parametric
estimation (Ali and Chaudhry 1990, Huang and Kalirajan 1997).
Secondly, another approach is the stochastic frontier. This method, recently,
compared to the deterministic production approach, is more widely applied.
However, particular empirical model for each study is depended on the type of
data. In this study, the dataset is cross-sectional type and our review in this part is
mainly about studies measured technical efficiency for cross-sectional data type.
In order to measure technical efficiency using stochastic frontier approach a type
of parametric production function should be appropriately chosen. The two most
common functional types are Cobb-Douglas (Kalirajan 1981, Ekanayake Taylor
and Shonkwiler, Phillips and Marble) and Translog Cost function (Huang and
Bagi, Kalirajan 1984)
2.2.3 Review of Determinants on Technical Efficiency
Another dimension of technical efficiency analysis is determined factors
that affect the technical efficiency. According to Bravo-Ureta and Pinheiro
(1993), a meta-analysis about a set of 30 studies estimate the technical efficiency
and the impact of determinants on technical efficiency suggested that the group
of these determinants can be demographic, geographic and other social-
economics characteristics of the farming household. For instance, age, gender,
education, experience and other information of the famer can affect their
production efficiency. Furthermore, variable relates to operating the production
such as management policy, information, modernization or technique adoption is
also considered as determinants of technical efficiency. Specifically, under the
topic about impact of new technique adoption on technical efficiency, Kalirajan
(1984) had already conducted the study about how adopting new technology
22

could affect the paddy production. However, as the results showed insignificant
influence of the new technique on technical efficiency, Kalirajan concluded that
the famers is not adequately adopted the technology. In 1990, Kalirajan again
studied about the impact of different established cropping methods on technical
efficiency and found a significantly positive relation. In fact, to generate an
overview on deciding the determinants of technique efficiency, Belotti, Daidone
et al. (2012) indicated that these determinants of technical efficiency are factors
that can affect the production, but are not output or input.
In considering about the AWD score, firstly, AWD technique adoption has
showed its impacts on the production as reducing water input and potentially
affecting the output but AWD score is neither input nor output. Secondly, AWD
score measure the degree of new technique adoption which is similar to other
technique adoption variables in the literature review about factors that influence
production technique efficiency. Consequently, according to these arguments, the
AWD score represents for AWD technique degree of adoption is examined as a
determinant on technical efficiency in this study.
2.3 Summary
Following the literature review pathway, in this study, firstly, an adjusted
AWD score, based on the original score formula, combined with the AWD
technique definition of IRRI and drainage practices of famers in the MRD
provinces, is conducted and measured for each farm. This adjusted AWD score
presents the degree of AWD technique adoption for each rice production.
Continually, the stochastic frontiers Cobb-Douglas production function is applied
to measure technical efficiency. Finally, the AWD score is examined as a
determinant of technical efficiency on the rice production to determine the impact
of AWD technique after widely adopted in the MRD provinces of Vietnam. This
study is expected to provide another contribution in the vast empirical work of
investigation about the impact of AWD technique after widely adopted,
specifically, in the MRD provinces, Vietnam. This is also the initial paper
examines the impact of AWD adoption on technical efficiency of rice production.
Under results of this study, another evident is provided as foundation
23

for agricultural policy which promotes widely adoption of AWD technique
throughout rice production in Vietnam.
24

CHAPTER 3: DATA AND METHODOLOGY
This chapter describes about the data and methodology applied in this study. The
former part provides the analytical framework and the empirical model of the
research. The latter part illustrates the data collecting process and the dataset.
3.1 Methodology
3.1.1 Conducting the AWD Score
As previously discussed, when AWD technique is adopted on a large scale,
practices among farmers are various in term of watering schedules and
observation methods. Therefore, a general approach to distinguish between
adopters and non-adopters or a measurement of AWD technique’s adoption
degree is required to go further in analyzing AWD impacts. However, to create
this approach a clear definition about AWD adoption should be determined. Base
on the mentioned IRRI definition of AWD technique, for this project, AWD
adoption is defined as “An irrigated rice production using non-continuous
flooding when water is readily available.” Referred from this AWD adoption
definition, whenever the field is watered at the non-flooded, AWD technique is
adopted. As the consequences, an new adjusted AWD score is developed base on
this definition and the original AWD score formula for China suggested by
Moya, Hong et al. (2004). Since the AWD practices is differently adopted from
site to site, including Vietnam, the formula of AWD score is modified based on
Vietnam farmers cultivated practices. The adjusted AWD score formula for
measure degree of AWD adoption at farm level is suggested as follow:
= 1×X +1×f(D )+0.5×Y +0×Z
(1)
X +f(D )+Y +Z
i=1,2,…,N
With f(D ) is a conditional function of D as:
f(D ) = {
f(D ) = 0 D = 0 (2)
f(D ) = (0.01 + D ) D > 0 (3)
25

Where proxies for the AWD technique‘s degree of adoption score in the i – th
farming household, Z , Y , X are the total irrigation times that the i – th farming
household irrigates when the paddy surface was in the status one (dry or broken
field), two (wet or saturated field) and three (flooded with standing water field),
respectively. D is the number of draining times, which means the total times that the
i – th farming household withdraw the water out of the paddy.
Table 3.1: The synthesis of signals that farmers use to observed water level on
the field during irrigation process
Source: Author synthetic
In order to explain for this modified score, firstly, AWD adoption is defined
as when there are alternate appearances of flooded and non-flooded conditions
on the field. This means when water is refilled in non-flooded second and third
paddy surface statuses in Table 3.1, AWD technique is adopted. Thus, if the
farmer irrigates when they can make foot-print and see bird crack (or the field is
too dry to leave foot-print), they are adopting AWD. Nevertheless, as IRRI
instructed that for safe AWD practices, famer can leave the paddy non-flooded for
a number of days until the water level reaches the threshold of 15 centimeters
2
6
iv. Deeper than
10 cm
d. Small cracks on the
paddy surface
iii. From 6 to
10 cm
c. Cannot make a
footprint on
3. Dry or broken (c and d,
the paddy
surface
or iii and iv)

Water levels
Paddy surface status Signals
(cm under the ground)
1. Flooded or standing a. Water can be observed on the i. From flooded water to 1
water (a or i) paddy surface cm
2. Wet or saturated (b or b. Easy to make a footprint on ii. From 2 to 5 cm
ii) the paddy surface

under the ground (status three) and if the famer keeps longer interval days, the
less irrigation times are applied and more water is saves. Therefore, the degree of
adoption is higher for each time famer re-waters the field in status three rather
than in status two. For this reason, when the famer follows exactly safe AWD
guidance and irrigates in status three, the adoption score for each time they
irrigates is fully given as one point. Also, based on the previous explanation,
even though irrigated in status two is not exactly adequate according to safe
AWD guidance, but there still non-flooded condition on the field, and still save
water compared to traditional continuous flooding practice. Consequently, for
each time farmer irrigates in status two, only 0.5 point is given which records the
presence of AWD adoption but not at the fully degree. Finally, for each time the
famer irrigated in status one with flooded paddy surface, the given score of
adoption would be zero. In general, total AWD adoption score of X times
irrigated in paddy surface status three is given as X multiplies by one score, of Y
times irrigated in status two is estimated as Y multiples by 0.5 score and of Z
times irrigated in status one is Z multiples by 0 score.
Secondly, one special point which the AWD project team had also found in
cultivation practices among rice producers in MRD, Vietnam is that beside the
attempt to irrigate in non-flooded status, they also withdraw water several times
during rice cropping seasons, especially when the crop is unexpectedly irrigated
by the rainwater and suffered from unusual flooded status. While the original
AWD technique encourages famers to allow non-flooded condition on the field
surface, water drainage can enhance AWD technique adoption through
accelerating soil drying process. In addition, this practice can also improve soil
condition by washing alkaline soil. Overall, withdrawing water does not create
the same adoption effective level as irrigation in the second or third paddy
surface status, but it shows the influence on AWD adoption by enhancing soil
drying process. Therefore, the natural logarithm of D is used to measure the score
this effect. As the value of natural logarithm when D equal to one times is zero
meaning one time of water drainage is not recorded, then the modified function
of drainage score f(D ) = (0.01 + D ) is used instead, this value represented the
score for D withdrawing times. However, because drainage is a new typical
27

practices that come from the major group of rice producers in MRD provinces but not all of them,
consequently, for those farmers who do not practice drainage water, their score is estimated as the original
formula and f(D ) = 0.
Finally, the total score is calculated as the sum of all the score for X, Y, Z
times of watering and D times of withdrawing and divided by the sum of total
irrigation times (sum of X, Y, Z) and the efforts of withdrawing Ln(0.01+D) to
estimates the general adoption score for each farming household. The value of
AWD score ranged from zero to one point where one represents the fully degreed
of AWD adoption in each household. Moreover, in the case where farms are
irrigating with exactly the same number of X, Y, Z times, if they withdraw R
times are more they gain higher scores.
In conclusion, the adjusted AWD score can represent the degree of AWD
adoption of each farm in MRD provinces, and when the value of AWD scores is
higher, the farming household is considered as adopting AWD technique more
adequately.
3.1.2 Analytical Framework
A general stochastic frontier model is suggested for production function ( ) as following:
= + +,=1,…, (4)
= − , (5)
~ (0, 2 ) (6)
~ (7)
Where is the logarithm of total output quantity of the − ℎ productive
unit, is a (m+1) row vector of inputs, whose first element is “1” and remaining
values of the vector are the logarithms of m different input quantities which are
used by − ℎ productive unit. = ( 0, 1, … , )′ is a (k+1) column vector
parameters that need to be estimated. As suggested by Aigner et al (1976), the
error term is the sum of two components and . is the normally 28

distributed part ~ (0, 2 ), which represents an idiosyncratic part and shows
measurement and specification error. is one-sided error, which measures the
inefficiency of the model. In order to estimate the model, an assumption about
distribution should be determined. The large body of empirical studies toward
technical efficiency measurement suggests that there are four possible distribution
options for . In specific, these are half-normal distribution
~ +(0, 2 ), which is assumed by Aigner et al (1977), exponential
~ ( ), suggested by Meeusen et al (1977), truncated normal distribution ~ +( , 2 ), applied by Stevenson (1980) and the gamma distributions used by
Greene (1980a,b, 2003).
According to Belotti, Daidone et al. (2012) requirement of identical
inefficiency distribution assumption indicates that maximum likelihood approach is
used more often to estimate this model. Generally, the regression consists of two
stages, firstly the parameter of the production function is estimated through
evaluating the maximum value of log-likelihood function, then in second stage, a
point estimate of the inefficiencies can be obtained using the mean E(u| ̂)(or the
mode M (u| ̂)) of the assumed conditional distribution and the technical efficiency is
defined as:
Effi = exp(− ̂ ) (8)
Where (− ̂) is either E(u| ̂)(or the mode M(u| ̂)). The value of technical efficiency (EFF) is constrained from zero to one. If ̂ is
zero, then TE equals to one, and production is referred as technically efficient.
This study also examined factors that could affect the inefficiency. Belotti,
Daidone et al. (2012) provided a general description that these factors usually are
not the production inputs or outputs, however, it might effect on the production
performance in three main ways: Firstly, the determinants can shift the frontier
function and/or the inefficiency distribution, secondly, the determinants can scale
the frontier function and/or the inefficiency distribution and finally the
determinants can shift and scale the frontier function and/or the inefficiency
distribution. As this study is mainly applied the half-normal distribution
29

assumption, then in this approach the determinants affected on inefficiency by
scaling its distribution, and the function to estimate the factors that impacts
inefficiency suggested by Caudill, Ford et al. (1995) is given as follow:
~ +(0, 2 ) (9)
= exp( ′ ) (10)
Where is the half-normal distribution inefficiency, ′ is a vector of inefficiency
determinants variables and is the vector of estimated parameter. In 1999, Hadri developed the
estimation by allowing estimated variance on the idiosyncratic error as:
~ (0, 2 )
(11)
= exp( ′ ) (12)
Where is the normal distribution term, ′ is a vector of factor variables and is the vector of the estimated parameters.
As mentioned, there are three types of distribution assumption which could
be chosen for technical efficiency distribution. However, exponential distribution
is not widely applied in agriculture and also not suitable for this study, then this
research only provide to model test result between two distribution choices: Half-
normal and truncated. The general likelihood ratio test is given by the following
formula:
= −2[ [ ( 0)] − [ ( 1)]] (13)
This test is used to compare between to models. Half-normal distribution
assumption is a restricted form of the truncated normal with the restriction that
having the mean equal to 0. Thus, the test between the two options is basically to
tes the null hypothesis: Ho : u =0. Firstly, the value of log-likelihood function of
the restricted model (half-normal distribution) ln{L(Ho)} is estimated. Secondly,
the value of log-likelihood function of the unrestricted model (truncated
30

distribution) ln{L(H1)} is also estimated. Continually, the value of is calculated
and the null hypothesis is rejected if >R
2
, whereR
2
is the critical table value
with R restrictions, since this test is one-sided, the tables suggested by Kodde and
Palm is used. Similarly, this test could also be used to test and choose between
two models in two the functional forms of Translog Cost and Cobb-Douglas
function.
Also, in this study, the Bayesian information criterion (BIC) which was
introduced by Gideon E.Schwarz (1978) and the Akaike information criterion
developed by Hirotugu Akaike (1973) are used as criterion for model selection
among two finite estimated models under half-normal and truncated distribution
assumption of technical inefficiency disturbance. The general for these two
criterion is that the model with lowest AIC and BIC is more appropriate for
seclection.
In conclusion, although there is a different approach to measure technical
efficiency in agriculture sector, but this paper will apply the stochastic frontiers
function, using maximum log-likelihood method to estimate the production
function parameters, inefficiency value and examine the determinants that
affected the production inefficiency.
3.1.3 Econometrics Model
This paper applied the Cobb-Douglas functional form for stochastic frontier
estimation. Since this functional form is widely used to estimate technical
efficiency in agricultural sector. Yao and Liu (1998) suggested that Cobb-
Douglas is usually applied at it measures more of the efficient estimation.
Following these arguments, a stochastic frontier model using Cobb-Douglas
functional from is specify in this study as follow:
= ∏ − (14)
=1
The logarithm form of the above function is:
= + ∑ + − (15)
1 =1
31

Where represents the output, are the j-th inputs of the i-th farming
household, and the , … , are the coefficient that must be estimated. In specific, the
production function model in this study is deployed as the following form:
=+1 +2 +3 + 4 + 5 + 6 ℎ +
7 + −
In which is logarithm value of total amount of rice yield that the i-th farm
produced in kg unit, is logarithm value of total cultivation area in the i-th farm in 2 unit,
is the logarithm values of total amount of seed that the i-th farming household used to
produce rice, in kg unit, is the logarithm values of total amount of fertilizer that the i-th
farming household used to produce rice, in g a.i unit, is the logarithm value of irrigation
times of the i-th farm, ℎ is the logarithm value of total machine renting hours of
the i-th farm and is the logarithm of the total working mandays of the i-th farm. The
values of vi represent the idiosyncratic disturbance that cannot be controlled by the
farmer, the values of ui represent the technical inefficiency of each farm and is assumed
to be half-normal distribution ~ +(0, 2 ). The vector of coefficients ( , … , 7) are
coefficients to be estimated.
After that the technical efficiency of i-th (Eff) can be measured under this
formula:
Eff = exp(− ̂ ) (17)
Where (− ̂ ) is either E( | ̂).
The empirical technical inefficiency model
32

The model examines determinants of the technical inefficiency model in
this study is suggested as follows:
= ∑5
(18)
=1
= ∑2
(19)
=1
Formulation (18) estimates impact of vector of determinant variables on
technical inefficiency and is the vector of measured parameters. Formulation (19)
examines vector variable effect on the normally distribution error term and is the
measured coefficients. The specification of the two mentioned models above in
this study is given as:
= 1 + 2 + 3 +
4 + 5
Where represents the mean of the estimated technical inefficiency for the i-
th farm. We include the measured variable of i-th farm to examine the potential
impact of AWD adoption level on technical inefficiency, and determine whether
adopting AWD can improve the efficiency of each rice productive unit. Other
variable of household characteristics which are suggested in the literature review
is also added in the model to examine their impact on technical inefficiency.
These variables including which represents the schooling years of the household
head, the owner of the i-th farm, represents cultivation experience in years of the
household head, the owner of the i-th farm, represent the number of household
member who participates in farming activities. Another variable representing for
the production condition is . This is a dummy variable that represents the type of
soil, equal to one if the soil type is alluvial and equal to zero for others type
because alluvial is suggested as best soil type for growing rice.
=1+2
(21)
33

The second function is to examine variables that affect , which represents
the mean of the normal distribution error term of i-th farm. We include and in this
function as even though these also belong to farming household characteristics,
but the farm cannot control or change their age and gender. The and variables
represent the household head age and gender in the i-th farm, respectively.
3.2 Data
The dataset in this study is provides by AWD Vietnam project, which is
deployed under cooperation between IRRI and Nong Lam University. The
process of collecting data for AWD research consists of three stages: Gathering
AWD literature review and documentation, focus groups discussion (FGDs) and
key informant interview (KIIs), generally farming household surveys.
In the first stage, a general overview of collected literature review and
documentation is conducted to understand about how AWD technique was
adopted globally and specifically in the MRD provinces of Vietnam. The
gathered information comes from different resources including academic
research tools, contacting agricultural government offices in each province that in
charge of promoting AWD adoption and IRRI agency in Vietnam. Base on
searchable information, some typical surveys are designed for the focus groups
and household interviews in the following stage.
Continually, in the second stage, FGDs and KIIs are deployed in the
selected provinces. Initially, due to their hydro systems and presence of historical
AWD campaigns, three provinces including An Giang, Bac Lieu, and Dong Thap
are selected to implemented FGDs and KIIs. However, KIIs also show likelihood
of AWD adoption in Soc Trang provinces and Soc Trang is also added in this
survey. Through this second stage, important insights about AWD current
implemented state are found. The AWD project team detects that after a few
initial trials, most of the famer in MRD provinces do not continually use the pipe
to observe water level on the field. Also, although, specific schedule and
34

frequency for AWD are introduced in 1M5R guide book, famers AWD manage
water irrigation differently depending on experience and cultivation conditions.
As a consequence, updated information about typical AWD practices after
adopting in large scale, for example, incomplete drying or alternative practical
methods to determine field water level for safe AWD are documented. In this
stage, the average of five KIIs and two FGDs is hold for each province. Results
of KIIs also help determining appropriate districts in each province and FGDs
information support to consider suitable areas which strongly represented the
district conditions to perform household surveys in the next stage. Within the
defined district, farming households are randomly sampled from the two best
districts.
In the final stage, AWD survey team collects data from 275 farming
households including information about their rice production inputs and outputs,
household characteristics, farming conditions, and especially specific water
irrigation schedules, frequencies, practices are noted to develop a scoring system
to measure AWD technique degree of adoption.
The collected dataset is questioned from 275 rice producers in MRD
provinces in total, however, only 250 observations fully provide all the required
information that are used to analyze in this study. In specific, among 250 rice
producers, there are 96, 63, 44 and 47 farming household in An Giang, Dong
Thap, Bac Lieu and Soc Trang province, respectively. The AWD survey
primarily focuses in Phu Tan district of An Giang province, Tam Nong district of
Dong Thap province, Phuoc Long, Vinh Loi, Hong Dan district of Bac Lieu
province and Nga Nam town of Soc Trang province. Since AWD is mainly
recommended to apply for dry season, the questionnaire provides general
household characteristics and specific rice production information of 2016 dry
season (Winter – Spring) of the farms in MRD region. In fact, this season is also
the main rice cultivation season in Vietnam.
As mentioned, results of the FGDs and KIIs suggested that most farmers
use different method to replace the AWD pipe for observing field water level and
apply AWD whenever appropriate. The farming household data also presents that
35

approximately 90 percent of households in MRD provinces uses other
approaches to measure water levels rather placing the pani-pipe on the farms. The
most common alternative approach in MRD province is called “bird crack”
method. Under this method, famers only need to observe the soil cracking which
is visually similar to a bird foot and notice that water level reaches 15 centimeters
threshold under the surface. Nevertheless, “bird crack” method cannot be applied
for saline alluvial or partial sandy soil. Furthermore, farmers report that they have
no difficulties in determining irrigation schedule using the “bird crack” or
observational methods.
In order to capture the adequate irrigation schedule together with AWD
adoption level of each farm, AWD team synthesizes all the possible signals for
observational methods which match three particular soil water statuses. Firstly,
when the paddy is in status one, with flooded or has water ponds over the
surface, which can be easily observed, the water level ranges from above to one
centimeters under the ground. Secondly, the field is wet or saturated in status
two, when the farmer can make foot prints by slightly stepping on the soil and
the water level ranges from two to five centimeters under the surface. Finally,
when famers detect cracks or cannot leave foot print on the paddy which is status
three, the soil is dry or broken and water level reaches the threshold of six to
fifteen centimeters under the field surface. Base on this synthesis, each farmer’s
specific irrigation practice can be recorded and degree of AWD adoption can be
measured by using the methodology discussed in the following parts.
36

CHAPTER 4: RESULTS AND DISCUSSION
After the overall presentation the research problem, objectives, literature
review, dataset and methodology, in this chapter, all of the findings are exhibited
and discussed. Initially, an overview about the calculated AWD scores and other
variable from the dataset is provided in the descriptive statistical section.
Continually, empirical results of the stochastic frontier production function and
the average technical efficiency are illustrated. Additionally, the main results of
this study examining the impact of AWD adoption on rice production in MRD
provinces, Vietnam is analyzed. Finally, a discussion section is conducted.
4.1 Descriptive Statistics
4.1.1 Data Description
Table 4.1 presents the summary statistic of the research’s main variables.
The former part of this table shows information about the independent input
variables including land, seed, fertilizer, pesticide, water, labor, machine together
with yield is the dependent output variable of the rice production function.
Overall, there are 250 observations in this sample. The information described in
this part of table 4.1 is collected from the 2016 main crop season (Winter –
Spring) of famers in the MRD provinces. The maximum amount of total yield
produced in this season is more than 121 tons of rice, while the minimum value is
less than one tons at farming household level in this sample. Averagely, each
farm can produce 17 tons for the whole season. In fact, the amount of total output
largely depends on the land size of each farm which is one of the main input
variables. In this sample, the cropping area is various from one to 180 hectares
among the observations with the average of approximately 24 hectares for each
rice productive unit. Another important input for rice production is the amount of
total seed used. The amount of seed use in this sample ranges from the minimum
amount of 20 kg to more 2000 kg for each farm, which leads to the average
number of around 400 kg of seed are used in every rice production. The amount
of fertilizer use among farms in the MRD provinces is also dissimilarity
depending on production scales. The maximum amount of fertilizer that a farm
37

could use in this sample is 14.4 thousand kg while the minimum amount is only
44 kg, the mean of fertilizer used amount here is more than one thousand kg.
Table 4.1: Descriptive Statistics
Variable Observations Mean Standard Min Max
Deviation
Yield 250 17964.20 16389.08 769.23 121500
Land 250 23744.74 21896.31 1000 180000
Seed 250 374.33 345.65 20 2380
Fertilizer 250 1160.59 1240.56 44 14400
Pesticide 250 5.73 5.76 0.06 26.62
Water 250 8.72 2.07 2 18
Labor 250 57.81 60.98 0.93 568.88
Machine 250 11.18 10.10 0.77 81.39
AWD score 250 0.6559 0.1912 0.0025 1
Age 250 49.30 11.77 24 84
Gender 250 0.92 0.28 0 1
Education 250 7.31 3.57 0 16
Experience 250 24.36 12.21 3 62
Member 250 1.88 1.03 1 6
Soil 250 0.66 0.48 0 1
Source: Author’s estimation
Beside fertilizer, to avoid harmful pests, farmers also have to use pesticide
as an input. The pesticide amount here is calculated in kg a.i unit which means in
the amount of active ingredient unit of pesticide used. The largest, smallest and
mean amount of total pesticide used is 26.6 kg a.i, 0.057 kg a.i and 5.7 kg a.i
among farms in this dataset, respectively. Water is one of the fundamental input
38

for rice production, this paper proxies water used by number of irrigation times,
during the season, famers irrigated averagely 8.7 times with the maximum of 18
times and minimum of 2 times among all observations in the dataset. The
remaining inputs are labor measured in working man days and machine measure
in hour unit. Among this dataset, each rice production required averagely 58
working man days and 11.2 hours of running machine. In detail, the maximum
man days and machine hour of a farm in the sample are more than 568 days and
80 hours, respectively. By contrast, there is farm required only minimum value of
less than one working man day and one running machine hour.
Additionally, this latter part of table 4.1 also presents the general
information about the estimated AWD score and the demographic, geographic
characteristics of each farming households in the sample. These are actually
factors which are considered as determinants of technical efficiency or dependent
variables in the technical inefficiency regression. AWD score is the main variable
in the set of technical efficiency determinants variables. The score is measured
for each farm using the modified AWD score formula introduced in chapter
three. According to table 4.1, the average AWD adoption score in this dataset is
0.65 which means most of the famers in this area get the score higher than the
average score value of 0.5 on the scale ranged from zero to one. The highest
score achieved from these famers is one which shows the fullest degree of AWD
technique adoption, and the lowest score in this sample is 0.002. Famers who are
also household head in the AWD survey have the average age of 49 years old. In
details, the oldest and youngest famers are 84 and 24 years old, following the
order. Most of the household head is men, accounted for 90 percent of the
sample. Averagely, these famers have more than seven years of schooling,
however, some famers does not have a chance to go to school while other few of
them reached further education with 16 schooling years. Since cropping is one of
the main production activities in the MRD provinces, famers here have a high
mean value of farming experience which is more than 24 years of cultivation.
Specifically, there is famer who grows rice for about 62 years in his life and there
is also new famer who only has three years of farming experience. In addition,
there are averagely two members of each questioned household who participated
39

in cropping activities among all of the observation. In some cases, all of the
family members are famers, reaching the highest number of six members in a
household involve in farming activities. The most important geographic
information that this dataset includes here is the type of soil, soil is a dummy
variable stands for alluvial soil type, according to the dataset, more than 65.6
percent of the farms has fully alluvial soil type.
4.1.2 Correlation Matrix
Table 4.2 illustrates the correlation matrix between all variables in this
research. Firstly, for the production function, most of the input highly correlates
with the amount of output in positive trend, as expected. However, noticeably,
number of watering times is insignificantly associates with output in a negative
relation. Considerably, as noticed, since in rice production, the amount of other
inputs strongly depends on the size of farm land, correlation between them are
significantly high, around 0.9, which could possibly lead to potential
multicollinearity problem. In order to detect multicollinearity, this study
measures the variance inflation factor (VIF) value and shows the results in table
4.3 .Overall, even the VIF value for lnland is high but the mean VIF value for the
model is 8.34, less than 10 implies that multicollinearity is at an acceptable level
for this model. The remaining part of table 4.2 shows results of correlation matrix
between the production function variables and the set of technical efficiency
determinant variables. As expected, since the determinant variables are those that
affect the production but is neither inputs nor output, most of the determinant
variables is positively correlated with yield in a statistically significant way.
Noticeably, the AWD score is significantly associated with yield and labor in a
positive trend, in additional, AWD score is negatively correlated with the
irrigation times, but not significantly.
40

Table 4.2: Correlation Matrix
Lnyield lnland lnseed lnfertilizer lnpesticide lnwater lnlabor lnmachine
lnyield 1.000
lnland 0.988*** 1.000
lnseed 0.925*** 0.927*** 1.000
lnfertilizer 0.896*** 0.901*** 0.874*** 1.000
lnpesticide 0.694*** 0.692*** 0.641*** 0.622*** 1.000
lnwater -0.039 -0.026 0.05 0.021 0.092 1.000
lnlabor 0.858*** 0.863*** 0.812*** 0.783*** 0.708*** 0.086 1.000
lnmachine 0.957*** 0.958*** 0.891*** 0.859*** 0.678*** -0.055 0.814*** 1.000
AWD score 0.109* 0.098 0.102 0.075 0.088 -0.073 0.122* 0.104
age -0.056 -0.068 -0.086 -0.112* -0.153** -0.036 -0.117* -0.064
gender 0.143** 0.133** 0.164*** 0.147** 0.121* -0.054 0.109* 0.149*
education 0.230*** 0.222*** 0.191*** 0.204*** 0.110** -0.075 0.190*** 0.222***
experience 0.016 0.014 -0.041 -0.052 -0.109 -0.043 -0.07 0.021
member 0.218*** 0.222*** 0.205*** 0.183*** 0.203*** -0.046 0.222*** 0.220***
soil 0.049 0.045 0.200*** 0.155** 0.017 0.177*** 0.093 -0.034
*, **, *** respectively denotes significance level of 10%, 5%, 1%
41

Table 4.3: Variance inflation factor
Variable VIF 1/VIF
lnland 24.52 0.040783
lnmachine 12.59 0.079405
lnseed 7.84 0.127628
lnfertilizer 5.65 0.177092
lnlabor 4.51 0.221565
lnpesticide 2.16 0.462124
lnwater 1.11 0.896961
Mean VIF 8.34
The age variable illustrates consistently negative correlations with all input
and output variables. However, these negative correlations are only statistically
significant with fertilizer, pesticide and labor, meaning that older famer might
tend to use less of these inputs in their production. Beside number of irrigation
times, determinant variables such as gender, education of household head and
number of members involved in agriculture activities of each household affect all
other inputs and output in the same positive and statistically significant way.
Finally, experience has no significant correlation with the production function
variables, and soil only positively correlates with seed and water in a significant
way. In general, even though there are significantly correlated relationships, but
these values are less than 0.3 which do not associate with potential
multicollinearity.
The correlation matrix which is showed in table 4.4 is applied between the
variables in the technical efficiency determinants function. Firstly, there are
significantly positive correlations between AWD score and two other variables
which are member and soil. Secondly, age is negatively and significantly
associated with education which means older famers probably got less chance for
higher education.
42

Table 4.4: Correlation Matrix
AWD
age gender education experience member soil
score
AWD
1.000
score
Age -0.015 1.000
gender 0.006 0.040 1.000
education -0.040 -0.282*** 0.132** 1.000
experience -0.051 0.791*** 0.029 -0.171*** 1.000
member 0.168*** 0.079 0.064 0.039 0.117* 1.000
Soil 0.130** -0.037 0.115* -0.031 -0.131** -0.041 1.000
Noticeably, age and experience are highly correlated in a positive way, this
is reasonable since in reality, most of older famer usually has more farming
experience, however, this correlated value is less than 0.8. Male famers are
having more chance to study at school since gender is positively and significantly
correlated with education and soil also. In addition, education is also significantly
correlated with education but in a negative way, this is consistent with the
previous results that older famers might have less chance for education but have
more experience, consequently, people with higher farming experience
negatively associates with their schooling years. Experience also positively and
significantly correlated with the number of members in the household that
involve in agriculture activities but negatively associate with soil type. Overall,
there are different correlation trend among the determinants of technical
efficiency variables, but most of the correlated parameter is less than 0.3 and
could hardly lead to multicollinearity problem.
4.2 Empirical Results
The regression results of this study include three main parts. The first part
describes general current state of the AWD adoption in the MRD and challenges
43

for AWD adoption. The second part is results of Stochastic Frontiers Cobb-
Douglas production function and technical efficiency. The final part is the main
result of AWD adoption’s influence on technical inefficiency. These results are
analyzed for rice production of famers in the MRD provinces, Vietnam.
4.2.1 The AWD adoption degree and challenges for AWD adoption
As illustrated in table 4.5, generally, there are only 12 farms in our dataset
that have the AWD score less than 0.5 on the scale from 0 to 1. Most of the
estimated AWD score for each farm is ranged from 0.5 to 0.7, which also
accounted for 65 percent of the total 250 observations. Moreover, there are 76
farmers whose scores belong to the upper level, larger than 0.7, which is
equivalent to 30 percent of the total sample.
Table 4.5: Percentage of different AWD adoption level
AWD score No. of observation Percentage
Less than 0.5 12 5%
From 0.5 to 0.7 162 65%
Larger than 0.7 76 30%
Table 4.6 provides the level of AWD application in each MRD provinces
from our dataset which are An Giang, Dong Thap, Soc Trang, Bac Lieu. As
mentioned above, the average AWD adoption score in this sample is 0.65. In
specific, Bac Lieu receives higher average AWD score, which is 0.7 while Soc
Trang is the province with lowest mean score value of 0.5. Bac Lieu’s famers
have more similar adoption level as their score range is narrowest only from 0.5
to 1. In fact, there are famers who get the highest score value in every province,
but the famer with lowest score 0.002 is in Soc Trang province. However, the
average scores in each province are not significantly different and also centers at
0.5 to 0.7 level which shows uniformity in AWD adoption of MRD region.
Moreover, the level of adoption between upper region of Mekong delta, including
An Giang and Dong Thap province is more invariable compared to the remaining
provinces in coastal region of Mekong Delta.
44

Table 4.6: AWD adoption score by provinces
No.of
Min Max Mean
Standard
observation Deviation
Upper region of Mekong Delta
An Giang 96 0.188 1 0.675 0.184
Dong Thap 63 0.215 1 0.678 0.163
Coastal region of Mekong Delta
Bac Lieu 44 0.5 1 0.701 0.196
Soc Trang 47 0.002 1 0.544 0.2
In addition, table 4.7 describes the challenges of AWD technique adoption
among MRD provinces. Firstly, there are approximately 63 percent of famers in
the sample said that they know about AWD technique. Others only practice the
technique spontaneously without adequately understanding about AWD, thus,
they could hardly adopted the technique appropriately. Only 77.08 percent of 96
interviewed famers from An Giang province reports that they know about AWD
technique, while An Giang is the key rice cultivation provinces in MRD region.
Nevertheless, this number is still higher than the remaining provinces, especially
in Soc Trang, less than 40 percent from 47 observations knows about the
technique. Furthermore, only a small number of famers continue to use the
recommended water tube for determining water level on the field while
averagely, about 90 percent of the famer answered that they are using field
observation method. In fact, although using water tube is recommended in safe
AWD by IRRI, the farmers refer the traditional observation method, as it is more
convenient but also allows them to determine the water level correctly. Finally,
irrigation water sources are differently used among MRD provinces. In the upper
region including An Giang and Dong Thap, most the famers uses irrigation
cooperatives or private services which is more costly but easier to control while
in coastal region as water is less available, the major of famers is using water
from power pump which does not cost much but required more efforts.
45

Table 4.7: Challenges of AWD adoption in MRD province
Upper region of Mekong delta Coastal region of Mekong delta
Whole sample
Provinces An Giang Dong Thap Bac Lieu Soc Trang
No. of observation N=96 N=63 N=44 N=47 N=250
Know about AWD 77.08% 60.32% 63.64% 36.17% 62.80%
Determining water depth method
- Water tube 12.50% 4.76% 20.45% 2.13% 10.00%
- Field observation 87.50% 95.23% 79.55% 93.62% 90.00%
Irrigation providers
- Irrigation cooperatives 98.96% 90.48% none 23.40% 65.20%
- Private irrigation services 1.04% 7.94% none 4.26% 3.20%
- Power pump use none 1.58% 100% 72.34% 31.60%
Farmers’ perception of irrigation water
Average perception ranking level score
availability1
Irrigation water availability at the present 9.53 9.29 7.48 7.77 8.78
Irrigation water availability in the future 8.72 7.54 5.82 6.6 7.94
1
Famers are asked to rank irrigation availability level score on a scale from 1 represented the lowest level to 10 represented highest level base on their perception.
46

The later part of table 4.7 measures the perception of famers toward the
availability of irrigation water level on the scale from 1 to 10, with 10 is the
highest available level. Generally, in the present most famer ranking ability to
supply irrigation water for their fields with a high score level at averagely 8.8 for
the whole sample. The lowest level score is 7.5 in Bac Lieu province which
means currently, famers in MRD region do not consider water scarcity as a threat
in their rice production. The ranking score for future circumstance is slightly
lower in all provinces but also relatively high equivalent to 7.9 in average.
4.2.2 Results of average technical efficiency of rice production in the
MRD region
Table 4.8: Estimated Average Technical Efficiency
No.of Mean Technical
observations Efficiency Standard Deviation.
AWD score ≤ 0.5 33 0.926 0.036
AWD score > 0.5 217 0.962 0.027
Total 250 0.957 0.030
After regression of the production function, the average technical efficiency
is also estimated and presented in table 4.8. The estimated average technical for
250 rice productions in the MRD provinces is 95.7 percent which show that the
rice production is slightly fell behind the frontiers. However, the average
technical efficiency estimated in this sample is high more than 90% which
reflects the uniformity in efficiency among famers in the MRD provinces. Most
of the farming household is relatively close to the frontiers and produces
effectively. The average technical efficiency for two groups with different level
of AWD score is also report in table 4.9. Generally, in the group which has the
AWD score equal or less than 0.5, the technical efficiency is 92.6 percent. This
average technical efficiency is less than the average technical efficiency of the
group with AWD score higher than 0.5, which is 96.2%. In fact, the impact of
AWD adoption on technical inefficiency will be discussed more specific in the
following part.
47

4.2.3 Results of determinants on the technical inefficiency
In the second regression, as this study examines which factors affect the
inefficiency or on the reverse, affect the technical efficiency and according to the
literature review, these factors are suggested to be affected on the production but
not exactly inputs or output, thus beside all the suggested determinants as
demographic and geographic characteristics of each household, AWD score is
also included in the regression.
Table 4.9: The technical inefficiency determinants model
Coefficient Standard Error
Determinants of technical
inefficiency model Sigma U
AWD score -3.7450** 1.5419
Education -0.2860* 0.1677
Experience -0.0270*** 0.0104
Member -0.0212 0.6261
Soil -1.3261*** 0.4132
Sigma V
Age -0.0452*** 0.0103
Gender -2.0221*** 0.5820
Number of observations 250
Average sigma V 0.1420
Average sigma U 0.0597
Log likelihood 149.20
According to the results in table 4.9, AWD score is negatively associated
with the technical inefficiency, as expected. The statistically significant
parameter is -3.745 which indicated that when the level of AWD adoption which
represented by AWD score increase by 0.01 point, the technical inefficiency
decreases by 3.745%, ceteris paribus. Consequently, this also increases the
technical efficiency of the production. As the technical efficiency show the
48

potential of achieving the maximum amount of output for a certain level of
inputs, the rice production with higher AWD degree of adoption, can generate
more output from each level of input and the production have higher efficiency.
This impact of AWD adoption on the efficiency can be explained as adoption of
safe AWD is suggested to reduce the amount of input (water) and increase the
output or show no significant reduction on the output, thus AWD can improve
the efficiency of the production. Results of other regressions show that education,
experience together with type of soil also statistically and significantly pose
negative impact on the technical inefficiency. In details, the education’s
coefficient is -0.286 indicating that one higher schooling year of household head
can reduce the inefficiency of the production by 28.6 percent, holding other
things constant. The estimated parameter of experience is -0.027, which means
when the famers have one more years of experience the inefficiency reduce by
2.7 percent in a significant way, ceteris paribus. Moreover, the coefficient of soil
is -1.326 suggesting that when the soil is fully alluvial type, technical
inefficiency decline by 13.26 percent, holding other things unchanged. Parameter
of member variable which represents for the number of household member
involved in agriculture is insignificantly negative, this mean more member
participate in rice production in each household is not necessarily improve the
technical efficiency.
The result also includes the empirical results of variables which belong to
the idiosyncratic error term. This study adds age and gender of household head
into this function because variables affecting the idiosyncratic error term are
factors that affect the production but the famers cannot change these
characteristics. In fact, the result show that both age and gender show negative
and statistically significant impacts on idiosyncratic error term. Meaning that
with both older and male household head, the part include in the idiosyncratic
error term of the production function can be improved. Finally, the average value
of the technical inefficiency part, U sigma is 0.0597 and of the idiosyncratic part,
V sigma is 0.142. The estimated Log Likelihood value is 149.2.
49

As reviewed in the literature review section, there are three distribution
assumptions which are commonly made for the technical inefficiency disturbance
of the production function. These are half-normal, truncated, and exponential
distribution. In order to choose the best option for the analysis, all of these three
assumptions were tested. Firstly, exponential distribution is not appropriate to be
used for this dataset. Results of both regressions under half-normal and truncated
distribution assumption are showed in Table A1 for comparison. Generally, in
both of these models, the estimated parameters of AWD score in technical
inefficiency model are consistently negative, which means higher degree of
AWD technique adoption affects technical efficiency in the positive trend.
However, in half-normal assumption model the effect of AWD score is
statistically significant while in truncated assumption model, it is statistically
insignificant. As to compare which model is more suitable for the dataset, the
Log likelihood Ratio test together with the Bayesian information criterion (BIC)
and Akaike information criterion (AIC) is applied.
Firstly, the half normal is a restricted form of the truncated normal with the
restriction that the average of denotes as u = 0. The value of the generalized
likelihood ratio statistic in this case is (-18.8) calculated from the given
formulation = −2[ [ ( 0)] − [ ( 1)]]. Since the estimated value is negative
(and it is less than the critical c2
value, which is always positive), we cannot
reject the hypothesis that Ho: u = 0 and accept the model which assumes the half
normal distribution. Secondly, Akaike's information criterion and Bayesian
information criterion test in Table 4.10 helps to choose best maximum log
likelihood model, model with the minimum AIC and BIC is the best model which
is half-normal distribution model in this case.
Table 4.10: Akaike's information criterion and Bayesian information
criterion
Model Log likelihood AIC BIC
Truncated Distribution 139.76 -271.52 -257.43
Half-normal Distribution 149.15 -290.31 -276.22
50

4.3 Discussion
Initially, the AWD is applied in a relatively adoption level among farms in
MRD provinces. The average AWD score for the entire dataset is 0.65 and most
of the score centers in the range of 0.5 to 0.7. The score also shows equivalent in
adoption among all of the provinces. Nevertheless, there are challenges that
might influence the AWD technique adoption in large scale in MRD provinces in
general. Firstly, less than 70 percent of the famers in this dataset do not know or
understand AWD technique correctly. Instead, they apply the technique
spontaneously, which consequently leads to inappropriate practices. Secondly,
most of the famers do not continue using water tube to determine water level on
the field, they feel more comfortable with traditional observing methods. Thirdly,
among provinces the water irrigation sources are different, which might cause
difficulty for the farmers to control their irrigation schedule and water level.
Finally, the major of famers in the MRD province is not consider water scarcity
as an important problem in their cultivation activities, which provides less
motivation for them to apply the technique resolutely. As a consequence, these
issues should be noticed and addressed in order to enhance higher degree of
adoption for AWD technique in Vietnam and particularly in MRD region.
Additionally, rice production technical efficiency level is relatively high
which shows uniformity among rice producers in the MRD provinces, Vietnam.
Most of the production is producing closely to the estimated frontier.
The AWD score in the second regression shows that the effort of adopting
this water-saving technique could make the rice production become more
efficient, particularly in the MRD provinces, Vietnam. This means when famer
adopts safe AWD technique at a higher degree which includes keeping their field
in alternate flooded and non-flooded condition, irrigating as the instructed
threshold, practicing withdraw water, they gain higher technical efficiencies. As
the technical efficiencies are higher, then with a certain level of inputs they can
generate more output. Compared to the literature this impacts of AWD on
technical inefficiency can be explained as AWD is suggested to reduce water
input without causing any yield losses or even increasing grain yield. The
51

mechanism behind this impact is that rice is still absorbed enough water through
its roots in some growing period without continuously flooding. Furthermore,
AWD technique can even increase grain yield because this technique hold the
root to grow stronger and help the soil become drier, more stable, which reduce
harvest losses. Consequently, AWD technique should be continually
recommended for farmer to adopt in the larger scale and different location,
especially where water for irrigation are scarce. Also, for each famer, the AWD
technique should be applied more appropriately to achieve higher AWD score
which could directly affect the technical efficiency. Specifically, in the MRD
provinces, famers should apply AWD more adequately in term of irrigation
timing, instructed irrigated water threshold, and should practice withdrawing
water to improve their production. In our sample, 271 out of 250 farms are
adopting AWD with higher level (the score higher than 0.5) and these farms
actually have a higher average technical efficiency. The remaining farms in the
dataset as well as the whole MRD region should be encouraged and supported to
improve their AWD adoption scores in their future crops which consequently
help to save more irrigating water and increase technical efficiency of the general
production.
52

CHAPTER 5: CONCLUSION
5.1 Main findings
Alternative wetting and drying is a water-saving irrigation technique that
has been developed and promoted to adopt widely in many rice growing
countries including Vietnam. As the technique is widely adopted, it has also been
adjusted differently. As a consequence, the degree of adoption for each farm and
its suggested impacts can change in various circumstances. In order to analyze
and determine the level of AWD technique adoption as well as its impact after
widely adopted, this study has conducted the general survey with famers in the
MRD provinces, Vietnam. In this survey, the AWD project team suggested an
modified AWD score which reflects the current AWD technique adoption degree
of each farming household and collected data to measure the score. Furthermore,
the data also contains other information which is used to analyses the impact of
AWD adoption on rice production in this area. Using the technical efficiency
concepts, firstly, a stochastic production function is regress to measure technical
inefficiency of the production, continually, a regression of AWD score and other
determinants is applied to see the impact of AWD on the technical inefficiency of
the rice production. There are three main findings in this study. Firstly, the
average AWD adoption score is 0.65 on the scale from 1 to 10 and most of the
famers estimated score ranged from 0.5 to 0.7 shows a relatively high and
uniformity AWD current adoption state among MRD provinces. Moreover, the
suggested that lack of understanding about AWD, the absence of the water tube
for determining water irrigated level, different in water resource use and farmer
perceptive towards water scarcity are the challenges that should be consider to
improving the level of AWD adoption. The estimated average technical
efficiency in this study is 95.7 percent indicates that most of the farms in the
dataset produce closely to the frontier. Finally, AWD score poses negative impact
on the inefficiency, in specific, an increase of 0.01 in the score can reduce
inefficiency by 3.745 percent, ceteris paribus, which mean higher degree of
AWD adoption can improve the production technical efficiency for famers in the
MRD area and this results is explained base on the impact of AWD on reduction
53

of water input without negatively affect the yield or even increase the yield by
enhancing stronger root and avoid harvest losses.
Overall, this study has contributed an adjusted AWD adoption score based
on a determined definition of AWD adoption and included the effect of water
drainage in Vietnamese AWD practice. This measurement helps to define AWD
after the technique is generally adopted. Furthermore, we also investigate the
impact of AWD in rice production after AWD is widely applied by using a new
approach base on the concept of technical efficiency.
5.2 Policy implications
According to the results, the degree of AWD technique adoption is found to
have negative and statistically significant impacts on the technical inefficiency of
rice production in the MRD provinces, Vietnam. AWD technique then should be
continually promoted to be widely applied in agricultural policies and programs,
specifically for rice production in Vietnam and generally for other rice growing
countries, in order to improve their production technical efficiency.
In addition, compiling with the increase in AWD technique adopting scale,
farmer should also be recommended to apply this technique more appropriately
based on the “safe AWD” practices and achieve higher AWD score. These
mentioned practices include three main key points. Firstly, famer should keep
their field in alternate flooded and non-flooded conditions whenever it is possible
to make sure they are applying AWD. Secondly, famer should irrigating and
applying this technique adequately in term of irrigating schedule and water
threshold. In detailed, AWD technique is highly encouraged during growing and
before harvesting stages, also, a maximum irrigated water level of five
centimeters above the field surface and a threshold of 15 centimeters water
dropped level below the surface is also empathized for safe AWD. Finally,
farmer should consider withdrawing water out of the field to enhance faster
drying process and warrant AWD is correctly applied.
5.3 Limitations
54

Beside the attempt for perfections, the study still contains some inevitable
limitations. Firstly, since there are only a small number of conducted researches
that examined the effect of AWD technique after widely adopted, the
measurement options which can proxy for the AWD technique’s adoption degree
for each farm are narrow. Furthermore, the AWD score suggested in this study
has also been modified based on the real practice of famers in MRD provinces, it
has not been previously uses by other studies. Secondly, this study is also the first
research that examines AWD adoption score as a determinant of technical
efficiency, consequently, the explanation for the estimated results are mostly
based on the research analyses and arguments. Thirdly, the water input in the
production function is only measured by irrigation times this is not the most
appropriate proxy for this input. Instead, if further study can capture the exact
amount of total water use in the production from two sources including irrigation
water and rain fed, the regressed function would be improved. Finally, in
agriculture industry, including rice production, each amount of different input use
is usually estimated base on the size of the farms so the correlations between land
and other the input variables are still relatively high, thus when interpreting the
effect of other inputs on the total yield the recommended sufficient amount of
these input uses for one unit of land should be considered. Hence, further studies
that are relevant to this topic can develop these elements in order to complete and
contribute more understanding about the AWD technique as well as enrich the
literature review.
Nevertheless, this study is hopefully developed and contributed new
evidences, findings into the literature of AWD technique adoption’s impact on
rice production in Vietnam, and generally of all rice growing countries. The
results suggest that agricultural policy should put more effort in promoting AWD
technique widely for rice producers. In addition, the technique should also be
applied more adequately with a higher degree adoption score to gain positive
impact on efficiency of rice production. Finally, the water drainage effect is
enhance the drying process of the field and improve the soil condition,
consequently, this practices should be considered to be widely promote in the
attempt of extending AWD technique adoption level.
55

REFERENCES
Ali, M. and M. A. Chaudhry (1990). "Inter-regional farm efficiency in Pakistan's
Punjab: a frontier Production function study." Journal of Agricultural Economics
41(1): 62-74.
Belotti, F., et al. (2012). "Stochastic frontier analysis using Stata."
Borrell, A., et al. (1997). "Improving efficiency of water use for irrigated rice in
a semi-arid tropical environment." Field Crops Research 52(3): 231-248.
Bouman, B., et al. (2007). "Rice and water." Advances in agronomy 92: 187-237.
Bouman, B. and T. P. Tuong (2001). "Field water management to save water and
increase its productivity in irrigated lowland rice." Agricultural water
management 49(1): 11-30.
Bravo-Ureta, B. E. and A. E. Pinheiro (1993). "Efficiency analysis of developing
country agriculture: a review of the frontier function literature." Agricultural and
resource economics Review 22(1): 88-101.
Caudill, S. B., et al. (1995). "Frontier estimation and firm-specific inefficiency
measures in the presence of heteroscedasticity." Journal of Business & Economic
Statistics 13(1): 105-111.
Chaudhary, T. (1997). Water management in rice for efficient production,
Directorate of Water Management Research.
56

Coelli, T., et al. (1998). Additional Topics on Data Envelopment Analysis. An
introduction to efficiency and productivity analysis, Springer: 161-181.
Dinh, N. N., et al. (2013). "Enhancement of performance of organic light
emitting diodes by using Ti-and Mo-oxide nano hybrid layers." Materials
Sciences and Applications 4(04): 275.
Easter, K. W., et al. (1998). Markets for water: Potential and performance,
Springer Science & Business Media.
Farrell, M. J. (1957). "The measurement of productive efficiency." Journal of the
Royal Statistical Society. Series A (General) 120(3): 253-290.
Hannesson, R. (1983). "Bioeconomic production function in fisheries:
Theoretical and empirical analysis." Canadian Journal of Fisheries and Aquatic
Sciences 40(7): 968-982.
Huang, Y. and K. P. Kalirajan (1997). "Potential of China's grain production:
evidence from the household data." Agricultural Economics 17(2-3): 191-199.
Lampayan, R. M., et al. (2009). Adoption and dissemination of “safe alternate
wetting and drying” in pump irrigated rice areas in the Philippines. 60th
International Executive Council Meeting of the International Commission on
Irrigation and Drainage (ICID), 5th Regional Conference.
Lampayan, R. M., et al. (2015). "Adoption and economics of alternate wetting
and drying water management for irrigated lowland rice." Field Crops Research
170: 95-108.
57

Li, H. and M. Li (2010). "Sub-group formation and the adoption of the alternate
wetting and drying irrigation method for rice in China." Agricultural Water
Management 97(5): 700-706.
Li, X. and F. Li (2001). "Study of Au/Au3+-TiO2 photocatalysts toward visible
photooxidation for water and wastewater treatment." Environmental science &
technology 35(11): 2381-2387.
Li, Y. and R. Barker (2004). "Increasing water productivity for paddy irrigation
in China." Paddy and Water Environment 2(4): 187-193.
McDonald, J. H. (2009). Handbook of biological statistics, Sparky House
Publishing Baltimore, MD.
Moya, P., et al. (2004). "The impact of on-farm water saving irrigation
techniques on rice productivity and profitability in Zhanghe Irrigation System,
Hubei, China." Paddy and Water Environment 2(4): 207-215.
Mushtaq, S., et al. (2006). "An assessment of the role of ponds in the adoption of
water-saving irrigation practices in the Zhanghe Irrigation System, China."
Agricultural Water Management 83(1): 100-110.
Palis, F., et al. (2005). "A farmer participatory approach in the adaptation and
adoption of controlled irrigation for saving water: a case study in Canarem,
Victoria, Tarlac, Philippines." Copyright International Rice Research Institute
2005 14: 397.
58

Palis, F., et al. (2012). "Adoption and dissemination of alternate wetting and
drying technology for boro rice cultivation in Bangladesh.”." Manila: Asian
Development Bank.(forthcoming).
Pan, J., et al. (2017). "Grain yield, water productivity and nitrogen use efficiency
of rice under different water management and fertilizer-N inputs in South China."
Agricultural Water Management 184: 191-200.
Pascoe, S. and C. Robinson (1998). "Input controls, input substitution and profit
maximisation in the English Channel beam trawl fishery." Journal of Agricultural
Economics 49(1): 16-33.
Price, A. H., et al. (2013). "Alternate wetting and drying irrigation for rice in
Bangladesh: Is it sustainable and has plant breeding something to offer?" Food
and Energy Security 2(2): 120-129.
Rejesus, R. M., et al. (2011). "Impact of the alternate wetting and drying (AWD)
water-saving irrigation technique: evidence from rice producers in the
Philippines." Food Policy 36(2): 280-288.
Richards, M. and B. O. Sander (2014). "Alternate wetting and drying in irrigated
rice."
Rijsberman, F. R. (2006). "Water scarcity: fact or fiction?" Agricultural water
management 80(1): 5-22.
Rui-Zhong, H., et al. (2002). "Geology and geochemistry of Carlin-type gold
deposits in China." Mineralium Deposita 37(3): 378-392.
59

Saleh, A. and S. Bhuiyan (1995). "Crop and rain water management strategies for
increasing productivity of rainfed lowland rice systems." Agricultural systems
49(3): 259-276.
Sandhu, B., et al. (1980). "Irrigation needs and yield of rice on a sandy-loam soil
as affected by continuous and intermittent submergence." Indian Journal of
Agricultural Sciences 50(6): 492-496.
Satyanarayana, A., et al. (2007). "Opportunities for water saving with higher
yield from the system of rice intensification." Irrigation Science 25(2): 99-115.
Sharma, P., et al. (2002). "Crop-water relations in rice-wheat cropping under
different tillage systems and water-management practices in a marginally sodic,
medium-textured soil." Water-wise rice production. International Rice Research
Institute, Los Baños, Philippines. p: 223-235.
Singh, A., et al. (2002). "Effects of rice establishment methods on crop
performance, water use, and mineral nitrogen." Water-wise rice production. Los
Baños (Philippines): International Rice Research Institute. p: 237-246.
Singh, C., et al. (1996). "Effect of transplanting date and irrigation regime on
growth, yield and water use in rice (Oryza sativa) in northern India." Indian
Journal of Agricultural Research (India).
Squires, D. (1987). "Public regulation and the structure of production in
multiproduct industries: an application to the New England otter trawl industry."
The Rand Journal of Economics: 232-247.
60

Stoop, W. A., et al. (2002). "A review of agricultural research issues raised by
the system of rice intensification (SRI) from Madagascar: opportunities for
improving farming systems for resource-poor farmers." Agricultural systems
71(3): 249-274.
Tabbal, D., et al. (2002). "On-farm strategies for reducing water input in irrigated
rice; case studies in the Philippines." Agricultural water management 56(2): 93-
112.
Tan, X., et al. (2013). "Effects of alternate wetting and drying irrigation on
percolation and nitrogen leaching in paddy fields." Paddy and Water
Environment 11(1-4): 381-395.
Tuong, T. and B. Bouman (2003). "Rice production in water-scarce
environments." Water productivity in agriculture: Limits and opportunities for
improvement 1: 13-42.
Tuong, T. P., et al. (2005). "More rice, less water." Plant Production Science
8(3): 231-241.
Uphoff, H., et al. (2001). "The microbial diversity in picoplankton enrichment
cultures: a molecular screening of marine isolates." FEMS Microbiology Ecology
35(3): 249-258.
Wadud, A. and B. White (2000). "Farm household efficiency in Bangladesh: a
comparison of stochastic frontier and DEA methods." Applied economics 32(13):
1665-1673.
61

Yamaguchi, T., et al. (2016). Diffusion of the water saving technology in the rice
paddies of the Mekong delta. Proceedings of the General Meeting of the
Association of Japanese Geographers Annual Meeting of the Association of
Japanese Geographers, Spring 2016, The Association of Japanese Geographers.
Yang, J., et al. (2007). "Water‐saving and high‐yielding irrigation for lowland
rice by controlling limiting values of soil water potential." Journal of Integrative
Plant Biology 49(10): 1445-1454.
Yao, F., et al. (2012). "Agronomic performance of high-yielding rice variety
grown under alternate wetting and drying irrigation." Field Crops Research 126:
16-22.
Yao, X. and Y. Liu (1998). "Towards designing artificial neural networks by
evolution." Applied Mathematics and Computation 91(1): 83-90.
Zhang, H., et al. (2008). "Postanthesis moderate wetting drying improves both
quality and quantity of rice yield." Agronomy Journal 100(3): 726-734.
山口哲由, et al. (2016). "Alternate Wetting and Drying (AWD) Irrigation
Technology Uptake in Rice Paddies of the Mekong Delta, Vietnam: Relationship
between Local Conditions and the Practiced Technology." アジア・アフリカ地域研
究 15(2): 234-256.
62

APPENDIX
Table A1: Comparison between estimated results between
half-normal and truncated distribution efficiency model
Half Normal Truncated
Stochastic Frontier Model lnyield lnyield
lnland 0.7310*** 0.7380***
lnseed 0.0568* 0.0692
lnfertilizer 0.0403 0.0534
lnpesticide 0.0182*** 0.0160***
lnwater -0.0681* -0.0863**
lnlabor 0.0170** 0.0193**
lnmachine 0.1360* 0.1120**
Constant 1.4660** 1.3252***
Determinants of technical
inefficiency model Sigma U Sigma U
AWD score -3.7450** 0.0956
education -0.2860* -0.2660
experience -0.0270*** -0.0323**
Member -0.0212 -0.0511
Soil -1.3261*** -1.8940
Sigma V Sigma V
Age -0.0452*** -0.0636***
Gender -2.0221*** -1.2130
Mean U
Awd score -1.6930
Number of observations 250 250
Average sigma V 0.1420 0.1300
Average sigma U 0.0597 0.1650
Log likelihood 149.20 139.80
63

The Impact Of Alternative Wetting And Drying Technique Adoption On Technical Efficiency.doc

More Related Content

Similar to The Impact Of Alternative Wetting And Drying Technique Adoption On Technical Efficiency.doc (20)

More from Dịch vụ viết đề tài trọn gói Liên hệ ZALO/TELE: 0973.287.149 (20)

Recently uploaded (20)

The Impact Of Alternative Wetting And Drying Technique Adoption On Technical Efficiency.doc