Water Use Efficiency in Rice Production: Implications for Climate Change Adaptation in the Vietnamese Mekong Delta

  • Vo Hong Tu
  • Nguyen Duy Can
  • Yoshifumi Takahashi
  • Mitsuyasu Yabe
Original Research Paper
  • 45 Downloads

Abstract

Water scarcity and competition in the Vietnamese Mekong Delta becomes a great matter of concern under the contexts of climate change and tributary dam construction in upstream countries. Moreover, water productivity or water use efficiency in rice production was quite low in the region due to free-rider problem. Therefore, the study was aimed at measuring water use efficiency at farm level and investigating the factors affecting water use efficiency. The study applied stochastic frontier analysis to estimate water use efficiency for 159 rice farmers in the upstream of the Vietnamese Mekong Delta. The study shows that the average water use efficiency was 18.81%, indicating that rice farmers did consume water very inefficiently. On the average, the rice farmers could reduce their water consumption about 81.19%, keeping their output level and other inputs constant. To improve water use efficiency, the study found that land accumulation or large-scaled production is essential.

Keywords

Climate change Stochastic frontier analysis Tobit regression Water use efficiency 

Introduction

Water is one of the most important components to maintain our life and nature. However, water scarcity becomes a great matter of concern under the increasingly adverse impacts of climate change (Gosling and Arnell 2016; Mekonnen and Hoekstra 2016). In addition, agricultural intensification and rising irrigated area to meet the demands of increasingly additional population is amplifying water stress (Assouline et al. 2015; Lobell et al. 2008). Such stress becomes more serious as total global water consumption in agricultural sector is accountable for 85% (Shiklomanov and Rodda 2004).

The Vietnamese Mekong Delta (VMD) is one of the biggest rice producers in the world with the total cultivated area of 4.2 million ha, producing more than 25 million tons of rice in 2015 (GSO 2016). The VMD is suffering from water scarcity due to the largely unregulated and free access of water among upstream countries and the impacts of climate change as the VMD locates in the lowest basin of the Mekong River. The Mekong River, which passes through six countries (China and then go through Myanmar, Lao PDR, Thailand, Cambodia and Vietnam), journeys over 2700 miles from the Tibetan Plateau, and then flows into the East Sea at the VMD (Biggs et al. 2009; White 2002). Therefore, the VMD is considered as the last beneficiary in terms of water and natural resources from the Mekong River. In addition, to respond to the increasing needs on economic development and electricity consumption, many upstream countries along the Mekong River have recently constructed various hydropower plants, which leads to biodiversity losses and changing patterns of water availability in the downstream (Baran and Myschowoda 2009; Dugan et al. 2010). Moreover, the numbers of definitely planned dams were predicted to increase seriously from 41 in 2015 to 78 tributary dams in 2030 (MRC 2009; Ziv et al. 2012). Such scenario poses more stressful pressure for water management to safeguard water availability for agricultural production in the VMD. To address this problem, we have to consider interventions at both regional scale and individual water user. The former seems to be more difficult than the latter because of trans-boundary conflicts of interest. Thus, this current study focuses on how to consume water wisely by individual rice farmers to adapt to climate change and mitigate water conflicts among multiple users.

Agricultural production in Asia in general and the VMD in particular are facing two major challenges: increasing total production to meet an additional 1.3 million of rice consumers by 2025 while safeguarding enough natural resources, including water for future generations (Bouman 2007; Cantrell 2004). Previous studies show that water resource is becoming increasingly scarce due to high intensification, climate change (increased temperature and changing rainfall patterns), and high competition among water users, which is a threat to agricultural sustainability as water is an indispensable input in agricultural production process (Nhan et al. 2007).

Rice production is a main crop and consumes a huge volume of water in the VMD (e.g., 7,920 m3 ha−1 in spring-winter season and 3520 m3 ha−1 in summer-autumn season) (Nhan et al. 2007). However, water productivity in the VMD was much lower as compared to water-saving irrigation techniques in other Asian tropical countries, 0.8–1.2 versus 1.6–1.9 kg rice per one cubic meter of water (Bouman and Tuong 2001; Tuong et al. 2005). Thus, efficient water consumption is very important amid the increasing claims on water by residents, industries, and other intensified agricultural activities as well as climate change. Moreover, irrigation water is freely accessible, which leads to free-rider problems. Under this situation, it raises conflicts among water users such as agriculture-industry conflict and upstream-downstream conflict. The former explicitly presents the conflicts among sectors while the latter indicates the conflicts among locations. The upstream-downstream conflicts seem to be more important in terms of policy interventions because the majority of regional population earns their livelihoods through agricultural production. The main reason for such conflicts originates from a situation that rice farming in the upstream consumes a lot of water in dry season, leading to lower water table and salinity intrusion in the downstream. Under the contexts of serious water scarcity and serious water competition, efficient water use is one of the best solutions.

Reviews from literature show that two main approaches widely employed to measure water use efficiency (abbreviated as WE hereafter) in agricultural production are engineering/hydrological method and econometric approach (stochastic frontier analysis or data envelopment analysis) (Scheierling et al. 2014; Sharma et al. 2015). The former (hydrological method) defines WE as the ratio between consumed water quantity and output yield. Such calculation do not take the relationship between output and other inputs (fertilizers, pesticides, machinery cost, etc.) into account, leading to biased estimations (Billi et al. 2007; Sharma et al. 2015). The latter takes the relationship between output and all inputs into account when measuring WE (Dhehibi et al. 2007; Hong and Yabe 2017). It cannot be denied that output level is affected by multiple inputs rather than a certain single input, i.e., water. The current study therefore applied the econometric approach to estimate WE. With regard to stochastic frontier analysis (SFA) and data envelopment analysis (DEA), the well-known merit of SFA is a parametric approach that can separate noise effects apart from deterministic frontier while DEA is non-parametric approach that bases on mathematic programming to estimate production frontier and do not consider factors beyond control of producers. Therefore, the study estimated WE by using stochastic frontier analysis. Applications of this approach (SFA) can be found in the studies of Dhehibi et al. (2007), Karagiannis et al. (2003), McGuckin et al. (1992), and Tang et al. (2014). So far, to our best of knowledge, there have been no studies applying SFA to estimate WE can be found for the case of rice production in the VMD.

The objective of this study is to estimate WE in rice production in the upstream region of the VMD and investigate the determinants of gaps in WE.

The remaining structure of the paper is as follows. The next section presents analytical framework applied to measure WE by using SFA. Section 3 describes about data collection. The next provides results and discussions from the study, followed by conclusions in Sect. 4.

Methodology

Analytical Framework

WE in this study is defined as the ratio of feasible minimum water input to observed level of water, given other inputs and output constant. Such efficiency can be estimated by using stochastic frontier analysis devised by Aigner et al. (1977) and Meeusen and Van den Broeck (1977).

Assume that a farm uses a vector of inputs, denoted by X (X ∈ R+), to produce a single output, denoted by Y (Y ∈ R+) (see Table 1 for more details of inputs and output). Therefore, the stochastic production frontier presenting the relationship1 between inputs and output is given in compact form as follows:
$$ {Y}_i=f\left({X}_i,\beta \right){e}^{\varepsilon \mathrm{i}}\kern1em \mathrm{i}=1,2,\dots \dots \dots \mathrm{N} $$
(1)
where
Table 1

Descriptive statistics of variables used for production function

Variable description

Mean

Min

Max

SD

Y

Output quantity (kg ha−1)

8862.84

3400.00

13,000.00

1633.52

X 1

Water cost (1000 VND ha−1)

560.32

117.46

5000.00

563.74

X 2

Fertilizers (kg ha−1)

296.91

51.18

965.00

132.54

X 3

Pesticide costs (1000 VND ha−1)

4476.45

441.18

15,000.00

3038.52

X 4

Seed quantity (kg ha−1)

162.30

27.06

400.00

63.15

X 5

Machinery cost (1000 VND ha−1)

4020.40

2000.00

17,000.00

1384.44

Source: Estimated from the farmer survey in 2016, n = 159. Please see Appendix 2 for more details on how we measured these indicators

VND Vietnamese currency (US$1 ≈ 22.000 VND), SD standard deviation

Subscript i indexes all observed farms.

f(X i , β) presents the deterministic production frontier.

β is a vector of parameters to be estimated. These parameters (β) are estimated by using maximum likelihood estimation approach. According to Coelli et al. (2005), the joint probability density function which expresses the likelihood of observing the sampled farmers is a function of the unknown parameters β and σ2(variance of composed error term). Therefore, the maximum likelihood estimator of β is derived by maximizing the likelihood function with respect to β. The likelihood function is \( \mathrm{Ln}\ L=\raisebox{1ex}{$-I$}\!\left/ \!\raisebox{-1ex}{$2$}\right.\ln \left(2\pi \right)-\raisebox{1ex}{$I$}\!\left/ \!\raisebox{-1ex}{$2$}\right.\ln \left({\sigma}^2\right)-\raisebox{1ex}{$\kern0.15em I$}\!\left/ \!\raisebox{-1ex}{$2{\sigma}^2$}\right.{\sum}_{i=1}^I{\left({y}_i-{X}_i\beta \right)}^2 \). It seems complicated as we have to deal with many mathematical formulas. However, in reality, it is quite easy to obtain these parameters by using canned routines in Stata® or Limdep software (see the books of Greene William (2012) and Kumbhakar et al. (2015) for detailed instructions on Limdep and Stata, respectively). After obtaining parameters β, it is easy to estimate the composed error term by entering these parameters and observed inputs into Eq. 1. However, so far, it is unable to separate noise effect (v i ) and technical inefficiency effect (u i ). To solve this problem, Jondrow et al. (1982) developed a formula, which is presented in Eq. 2.

ε i is asymmetric composed error term, indicating the difference between v i and u i (ε i  = v i  − u i ). v i which is a random error component follows iid assumption (independent and identical distribution) and normal distribution (\( {v}_i\sim N\left[0,{\sigma}_v^2\right] \)). v i presents noise effects beyond the control of farmers such as weather, luckiness,… u i is half-normal error term, distributed as \( {u}_i\sim {N}^{+}\left(0,{\sigma}_u^2\right) \), presents output-oriented technical inefficiency for the ith farm.

Output-oriented technical efficiency (abbreviated as OTEi hereafter) is estimated by the shortfall from observed output level to possible maximum level of output, keeping inputs constant (Coelli et al. 2005; Kumbhakar and Lovell 2003). OTEi is estimated by using the formula of Jondrow et al. (1982). This formula is widely applied in literature to estimate farm-specific technical efficiency also. According to these authors, u i is estimated as the conditional expectation of u i , given conditionally on composed error ε i . The formula is given as follows:

$$ {\mathrm{OTE}}_i={\mathrm{e}}^{-{u}_i}=\frac{y_i}{f\left({X}_i,{\beta}^{\ast}\right){e}^{v_i}}=E\left[{e}^{\left(-{u}_i|{\varepsilon}_i\right)}\right] $$
(2)
where y i is the observed output level and \( f\left({X}_i,{\beta}^{\ast}\right){e}^{v_i} \) is the deterministic frontier adjusted by noise effect (Bravo-Ureta and Pinheiro 1997). As was mentioned, the noise effects being two-sided or normal distributed represent disturbances known as acts of god in rice production. Like the classical linear regression model, the error terms exist because the variation of dependent variable could not be explained entirely by only observed independent variables in the model but also unobservable variables. Therefore, the noise effects or disturbance terms beyond the control of farmers need to be adjusted or separated from the stochastic frontier by using the formula proposed by Jondrow et al. (1982). If the noise effects are not separated apart from the deterministic frontier, the estimation of technical efficiency will be biased.
For better understanding on the relationship between inputs and output as well as the way to estimate OTEi, Fig. 1 shows the 3D graphical illustration of production frontier. For simplicity, we consider a simple production function in which two inputs (X1 and X2) are used to produce a single output (Y).
Fig. 1

Graphical illustration of production frontier and technical efficiency measures

Regarding the common properties of the production function, Fig. 1 represents the deterministic frontier by the non-decreasing, quasi-concave surface OX1RRFXiR. Point R on the surface is the observed farmer R that has output level Y R which is produced by using X 1R and X iR . The surface ABCR is the identical output quantity, Y R , of farmer R. From this surface, Fig. 2 can be derived as cross-sectional production frontier in normal input (X i ) and water input (X 1 ) space, keeping output level constant at Y R . From Fig. 1, it is also easy to obtain a measure of OTE, which is provided by the ratio of |OY R |/|OY F |. This ratio is consistent with Eq. 2.
Fig. 2

Graphical illustration of measurement of WE

Reviews from literature, two main production functions applied in agriculture are Cobb-Douglas and translog forms (Coelli et al. 2005; Kumbhakar and Lovell 2003). In order to estimate WE in this study, the translog form is more preferable because of its flexible property (Reinhard et al. 2000; Tu 2017).

The main objective of our study is to estimate WE; for clear explanation, X1 is denoted as water cost. Thus, Eq. 1 is re-expressed in translog form as
$$ \ln {Y}_i={\beta}_0+{\beta}_1\ln {X}_1+\frac{1}{2}{\beta}_{11}\ln {X}_1\ln {X}_1+\sum \limits_{n=2}^N{\beta}_{1n}\ln {X}_1\ln {X}_n+\sum \limits_{n=2}^N{\beta}_n\ln {X}_n+\frac{1}{2}\sum \limits_{n=2}^N\sum \limits_{m=1}^N{\beta}_{mn}\ln {X}_n\ln {X}_m+{v}_i-{u}_i $$
(3)

As defined, WE reflects the non-radial proportion which can be reduced without compromising output level, given other inputs constant. For simplicity, denote λ as water use efficiency (λ = WE). Therefore, (1 − λ)X1 reflects the possible proportional reduction of water input. Note that a farm which is considered as full WE is necessary to be technically efficient (u i  = 0) because of weak monotonicity assumption (Reinhard et al. 1999). Therefore, after replacing X1 with λX1 and setting u i  = 0, we get

$$ {LnY}_i={\beta}_0+{\beta}_1\ln \lambda {X}_1+\frac{1}{2}{\beta}_{11}\ln \lambda {X}_1\ln \lambda {X}_1+\sum \limits_{n=2}^N{\beta}_{1n}\ln \lambda {X}_1\ln {X}_n+\sum \limits_{n=2}^N{\beta}_n\ln {X}_n+\frac{1}{2}\sum \limits_{n=2}^N\sum \limits_{m=1}^N{\beta}_{mn}\ln {X}_n\ln {X}_m+{v}_i $$
(4)

Please note that λX1 indicates the minimum amount of water use without compromising the output level while other inputs are constant.

Again, based on the definition of WE (possible contraction of water use without compromising output), the output level in Eq. (3) and that of Eq. (4) are equal. Let them be equal, which yields

$$ \left({\beta}_1\ln \lambda {X}_1-{\beta}_1\ln {X}_1\right)+\left(\frac{1}{2}{\beta}_{11}\ln \lambda {X}_1\ln \lambda {X}_1-\frac{1}{2}{\beta}_{11}\ln {X}_1\ln {X}_1\right)+\left(\sum \limits_{n=2}^N{\beta}_{1n}\ln \lambda {X}_1\ln {X}_n-\sum \limits_{n=2}^N{\beta}_{1n}\ln {X}_1\ln {X}_n\right)+{u}_i=0 $$
(5)

Note that \( \mathit{\ln}{\lambda}_{\mathrm{i}}=\mathit{\ln}{\lambda}_{\mathrm{i}}{X}_1-{\mathrm{lnX}}_1=\ln \left(\frac{\lambda_{\mathrm{i}}{\mathrm{X}}_1}{{\mathrm{X}}_1}\right)={lnWE}_i \).

After manipulating Eq. (5), it yields a quadratic Eq. (6) as follows:

$$ \left[\frac{1}{2}{\beta}_{11}\right]{\left(\ln {WE}_i\right)}^2+\left[{\beta}_1+\frac{1}{2}{\beta}_{11}\left(\ln {X}_1+\ln {X}_1\right)+\sum \limits_{n=2}^N{\beta}_{1n}\ln {X}_n\right]\left(\ln {WE}_i\right)+{u}_i=0 $$
(6)
$$ {\displaystyle \begin{array}{ll}\mathrm{Let}& {a}_i=\frac{1}{2}{\beta}_{11}\kern1em \forall {a}_i\ne 0;\\ {}& {b}_i={\beta}_1+\frac{1}{2}{\beta}_{11}\left(\ln {X}_1+\ln {X}_1\right)+{\sum \limits}_{n=2}^N{\beta}_{1n}\ln {X}_n\end{array}} $$

By using the quadratic formula, the value of WE i is given by

$$ {WE}_i=\exp \left(\frac{-{b}_i\pm \sqrt{b_i^2-4{a}_i{u}_i}}{2{a}_i}\right) $$
(7)
According to Reinhard et al. (2000), the only solution with +√ is chosen due to the assumption that an efficient water use farm is also technically efficient. The WE of the ith farm is given by
$$ {WE}_i=\exp \left(\frac{-{b}_i+\sqrt{b_i^2-4{a}_i{u}_i}}{2{a}_i}\right) $$
(8)

For drawing better policy implications, the output elasticity with respect to water input can be calculated as follows:

$$ \frac{\partial \ln Y}{\partial \ln {X}_1}=\frac{\raisebox{1ex}{$ dY$}\!\left/ \!\raisebox{-1ex}{$Y$}\right.}{\raisebox{1ex}{${dX}_1$}\!\left/ \!\raisebox{-1ex}{${X}_1$}\right.}={\beta}_1+{\beta}_{11}\ln {X}_1+\sum \limits_{n=2}^N{\beta}_{1n}\ln {X}_n $$
(9)

This formula is also applicable for calculating the output elasticity with respect to other specific inputs.

For visible understanding, Fig. 1 illustrates cross-sectional graph of how to estimate WE.

Assume two farmers R and C are on the identical output surface, in which farmer R uses a quantity of water at X 1R and other inputs at X iR while farmer C consumes water at X 1F and other inputs at the same level X iR . Thus, farmer C is considered as efficient water use while farmer R is not. Normally, efficiency index in literature is measured in percentage. Therefore, from Fig. 2, WE of farm R is estimated by |OX1F|/|OX1R|.

Data Collection

The data used in this study were collected in April 2016 in An Giang province and Can Tho city. These two study sites are located in the upstream of the VMD and considered as the major rice producers. From these sites, we have conducted conveniently face-to-face interviews with 202 rice farmers who are practicing both eco-friendly and conventional rice farming. Eco-friendly rice2 in this study includes VietGAP; GlobalGAP; large-scale, ecological engineering; and floating rice. Because the floating rice is a special farming system, which is cultivated during flooding reason without requiring irrigation cost, we did not include these farmers in our estimation. The valid sample size in this study was 159.

Reviews from literature showed that the common variables (both independent and dependent variables) used in estimating a production function for agricultural sector are pure content of fertilizers, pesticide cost, seed quantity, labor, machinery cost, and water use (Coelli et al. 2005; Reinhard et al. 2000). The total amount of fertilizers is the sum of pure contents of potassium, nitrogen, and phosphorous. The measure of pure content is based on the percentage of active nutrients for each types of fertilizer. For instance, nitrogen 46% indicates that 100 kg of this fertilizer contains only 46% of active nitrogen. Labor is considered as an important input in production function, but it was difficult to measure and separate it from machinery cost as mechanization was considerably introduced to almost all stages of rice production in the Mekong Delta. Therefore, we aggregated both labor and machinery cost together and labeled as machinery cost. The dataset describing inputs and output used in estimating production frontier is summarized in Table 1.

In the study sites, water use is measured in total cost for irrigation because of existing various pumping methods in the study sites such as electricity-based pumping, fuel-based pumping, and combination of these two. The output level was measured at farm gate, indicating that the moisture content was high. We adopted such measurement because the majority of farmers were selling their product right after harvesting.

For drawing proper policy implications, the study also consider the factors affecting the WE. From the existing literature, the influencing factors considered in this study are presented in Table 2.
Table 2

Definitions of farm-specific characteristics

Variables

Notation

Description

Crops year−1

CROP

Dummy, 1 if three crops and 0 otherwise

Farm size

FARMSIZE

Continuous variable, ha

Rice experiences

EXPER

Continuous variable, years

Membership in organization

NETWORK

Dummy, 1 if participate, 0 otherwise

Female attention level

FEM

Continuous variable, percentage

Education

EDU

Continuous variable, years

Double eco practices a

ECO

Dummy, 1 if yes, 0 otherwise

Perception of water scarcity

SCARCITY

Dummy, 1 if yes, 0 otherwise

Perception of internal water conflict b

INTERCON

Five-point scale, 1 = low to 5 = high

Perception of external water conflict b

EXTERCON

Five-point scale, 1 = low to 5 = high

Perception of risks c

RISK

Continuous variable, in range [16–80]

Perception of environment d

ENVIR

Dummy, 1 if yes, 0 otherwise

Perception of water conflict trend

TREND

Five-point scale, 1 = low to 5 = high

Paddy plots

PLOT

Continuous, number of paddy plots

aSingle eco practice indicates that rice farmers are applying one of the following eco practices: VietGAP, GlobalGAP, ecological engineering, and large-scale rice while double eco practices presents those who are applying two of them

bInternal water conflict and external water conflict represents the perception of farmers towards water use conflicts between downstream and upstream in the VMD and between the VMD with upstream countries, respectively

cRisk is a comprehensive index of 16 five-point scaled indicators (weather variability, water scarcity, market fluctuation, rice-related diseases, management and technology, etc.)

dPerception of environment is considered as a dummy variable if farmers perceived both polluted irrigation water and biodiversity losses

Farm size is assumed to have a positive relationship with the WE. As rice farmers have bigger farm size, they of course need lesser times to ferry pumping facilities to the paddy fields, which results in lower transportation cost. Normally, the farmers with small farm size or many paddy plots have to incur more cost on pumping and transportation because the seasonal calendar of different plots differs from each other, which leads to more times of transportation and pumping. The study also assumes that rice experiences are positively correlated with the WE. The more experiences farmers have, the more knowledge and technique they accumulate. Therefore, they can manage their rice farming better in terms of water use. Membership in organization brings farmers more opportunities to learn and share new knowledge from each other. It is also assumed that those who are members of an organization will achieve better WE. Female participation level in rice production and educational level of the household head are also important factors that help farmers make use of water more efficiently. In the study site, eco-friendly techniques were recently introduced in rice farming with the hope of lowering the negative impacts of cultivation on environment. Therefore, application of eco-friendly technique or practices is assumed to be positively associated with the WE. The Mekong Delta has been suffering from severe impacts of climate change, especially water scarcity due to water use competition in the upstream countries. Normally, flooding season happens annually in the Mekong Delta due to two main sources: one is the water flow from the Mekong River and the second originates from precipitation. In the years of 2015 and 2016, the flooding levels were quite low compared to the previous years, resulting in water scarcity and saline intrusion in the downstream region. Therefore, the perception of farmers towards water scarcity is an important factor affecting their consumption behavior. Perception of internal water use conflict is defined as indicator presenting the conflict among upstream and downstream regions within the Mekong Delta. This variable is assumed to be positively correlated with the WE. Perception of external water use conflict represents the competition of water use among the Mekong Delta and upstream regions (i.e., Thailand, Cambodia, etc.). This variable is also assumed to be positively correlated with the WE. It is also assumed that as farmers perceive more serious water competition, they will tend to consume water more efficiently.

Results and Discussions

Estimation of Water Use Efficiency

As explained in the methodology, the WE is measured based on the estimated parameters and error term u i (the term represents the technical inefficiency) from production frontier (see Appendix 1 for more details). Therefore, we have to estimate the technical inefficiency (u i ) or technical efficiency (\( {\mathrm{OTE}}_{\mathrm{i}}={\mathrm{e}}^{-{u}_i}\Big) \). Prior to estimating the technical efficiency for each farmer, it is essential to test the presence of technical inefficiency. The null hypothesis in this case is absence of technical efficiency (i.e., \( {\mathrm{H}}_0:{\sigma}_u^2=0 \) and \( {\mathrm{H}}_1:{\sigma}_u^2>0 \)). From the estimates in Appendix 1, we can test the presence of technical inefficiency by using z test. The calculated z value was 154.29 (i.e., \( z=\overline{\uplambda}/ se\left(\overline{\uplambda}\right)=5.9249/0.0384 \)), which is greater that that of the critical value of 2.57 at 1% of significance, suggesting that the null hypothesis (absence of technical efficiency) was rejected. Accepting the alternative hypothesis (H1) means that there exist technical inefficiency effects among surveyed farmers. So, the inefficiency effects (u i ) can be derived for each individual farmer by using the formula proposed by Jondrow et al. (1982). We now turn to estimate the technical efficiency and WE by using Eqs. (2) and (8), respectively. The technical and water use efficiency scores are all summarized in Table 3.
Table 3

Technical efficiency and water use efficiency

Efficiency score

Technical efficiency

Water use efficiency

Count

%

Cumulative

Count

%

Cumulative

< 5%

0

0

0

28

17.61

17.61

5–10%

0

0

0

28

17.61

35.22

10–20%

0

0

0

53

33.33

68.55

20–40%

0

0

0

31

19.50

88.05

40–60%

0

0

0

14

8.81

96.86

60–80%

71

44.65

44.65

5

3.14

100.00

> 80%

88

55.35

100

0

0.00

100.00

Mean

83.14

18.81

Max

98.02

73.01

Min

73.78

0.70

SD

8.47

16.34

Source: Estimated from the farmer survey in 2016, n = 159

The study found that the average output-oriented technical efficiency was about 83%, indicating that rice farmers could expand their output level about 17% at their current inputs. This result also implies that technically inefficient farmers could follow the “modern” production technologies used by the highest technically efficient farmers to improve their output level. Note that the highest technically efficient farmers who are on the stochastic frontier are derived from maximum likelihood estimation approach. By using Eq. (8), we can perform the measurement of WE. As mentioned, the common method used previously to measure WE is hydrological approach, which attempted to calculate the index so-called water productivity (the amount or value of product over water volume or value). This index only considered the two factors: output and water use, so water productivity would be reduced if output level was low and water input increased. Moreover, this index cannot be compared among regions or locations due to large variations in both rice yields and water use. In addition, not all water is consumed by rice, which leads to the measurement of water productivity biased. The current study applied SFA approach, which considered the relations between not only water use and output but also other inputs in a function; it explicitly indicates that the WE in this study reflects the total possibility of reducing surplus water use due to poor management and others. Therefore, the estimated results would be more reliable and empirical for policy implications under the context of increasingly water scarcity. Table 3 shows that the average WE was quite low (only 18.81%), suggesting that the rice farmers have used water seriously inefficiently. The study also found that the rice farmers could reduce their water consumption by about 81.19% (i.e., 100–18.81%), keeping their output level and other inputs constant. For better understanding, consider an example that the total water cost that farmer A and farmer B used to produce the same output of 8000 kg ha−1 are, respectively, 400,000 VND and 600,000 VND while keeping other inputs constant. If farmer A is considered as the maximum likelihood estimate of farmer B, the WE of farmer B is measured by the ratio of 400,000/600,000. This ratio means that farmer B could reduce its water cost by 200,000 VMD ha−1 or 33.33% (i.e., 1-(400,000/600,000)).

With the average water cost of 560,320 VND ha−1 (see Table 1), the study suggests that the rice farmers could reduce about 454,923 VND ha−1 without compromising the output level, holding other inputs constant. These results partially reflect the advantages of SFA approach in estimating WE and amplifying the seriously overused situation of water resource in rice production in the Mekong Delta.

Like many other regions/countries, water resource in the Mekong Delta is considered as a public good (non-rival and non-excludable), which leads to the free-rider problem. More than 10 years ago, in order to promote rice production, the government has removed irrigation water fee collection on rice farmers. The seriously inefficient consumption of water was probably due to poor water management and the situation that the majority of rice farmers did not apply any scientific methods to monitor or measure water consumption in the paddy fields. The common way that the rice farmers used to check water demand of the paddy is by eye observation. The pumping and volume of water into the paddy were also decided by their own experiences. Water consumption varies greatly among farmers. The highest WE was about 73% while the lowest was only 0.7%.

In conclusion, the WE was quite low and it varies greatly among farmers. On the average, the rice farmers could reduce their water consumption by about 81.19% without compromising the output level while keeping other inputs constant. For drawing better policy implications, the next section will focus on the factors affecting WE or how to improve WE in the upstream of the Mekong Delta.

The Determinants of Water Use Efficiency Gaps

Prior to estimating the determinants of WE gaps among rice farmers, we have to specify the underlying production technology in order to understand the relationship between inputs and output. As mentioned above, the production function in translog form was employed to determine the production technology. The estimates of the production function for rice farmers by using SFA are presented in Appendix 1. Now, we turn to estimate the output elasticity with respect to inputs by using Eq. (9), which is presented in Table 4.
Table 4

Elasticity of output with respect to specific inputs

Indicators

Mean

Lower quartile

Median

Upper quartile

Water cost

0.0116

− 0.0277

0.0146

0.0531

Fertilizers

0.0062

− 0.0263

0.0116

0.0445

Pesticide costs

− 0.0130

− 0.0298

− 0.0118

0.0019

Seed quantity

− 0.0632

− 0.1379

− 0.0727

− 0.0125

Machinery cost

0.1292

0.0728

0.1334

0.1781

Total

0.0707

− 0.1488

0.0751

0.2651

Source: Estimated from the farmer survey in 2016, n = 159

Table 4 shows that the output elasticity with respect to water cost was quite small, suggesting that an increase in water cost results in very small additional production of output. The output elasticity with respect to machinery cost was the highest, indicating that this variable has played an important role to the improvement of rice productivity in the Mekong Delta. In fact, mechanization in agricultural production has been promoted recently, particularly in rice production in order to reduce harvest and post-harvest losses and lower production cost. Table 4 also shows that the returns to scale for rice farmers are decreasing (the sum of all elasticities was 0.07). This finding was in line with that of Tu (2015). The decreasing returns to scale implicitly indicate that the input-oriented technical efficiency was smaller than output-oriented technical efficiency.

To provide answers to the great variation of WE and to draw appropriate policy implications, we now turn to find out the factors affecting the WE gaps among farmers by using Tobit regression. The descriptive statistics of dependent variables are summarized in Table 5, and the estimates of Tobit regression are presented in Table 6.
Table 5

Descriptive statistics of farm-specific characteristics

Farm characteristics

Notation

Mean

Min

Max

SD

Crops year−1

CROP

2.71

2.00

3.00

0.45

Farm size

FARMSIZE

2.90

0.10

15.00

2.29

Rice experiences

EXPER

23.58

3.00

50.00

9.61

Social networks

NETWORK

0.41

0.00

1.00

0.49

Female attention level

FEM

22.77

0.00

100.00

24.59

Education

EDU

7.95

0.00

14.00

3.00

Double eco practices

ECO

0.09

0.00

1.00

0.29

Perception of water scarcity

SCARCITY

0.81

0.00

1.00

0.39

Perception of internal water conflict

INTERCON

3.34

1.00

5.00

1.24

Perception of external water conflict

EXTERCON

3.67

1.00

5.00

1.36

Perception of risks

RISK

42.33

20.00

65.00

8.27

Perception of environment

ENVIR

0.53

0.00

1.00

0.50

Perception of water conflict trend

TREND

3.71

1.00

5.00

1.19

Paddy plots

PLOT

1.69

1.00

5.00

1.04

SD standard deviation

Source: Estimated from the farmer survey in 2016, n = 159

Table 6

The Tobit estimates of the determinants of efficiency gaps

Variables

Coefficients

S.E

Variables

Coefficients

S.E

FARMSIZE

1.2224*

0.6887

RISK

− 0.2700*

0.1564

EXPER

− 0.2439*

0.1302

ENVIR

1.3738

2.6295

NETWORK

0.4569

2.6707

PLOT

− 2.2113

1.5405

EDU

− 0.3156

0.4189

Constant

38.8259

8.9778

ECO

− 15.6491**

4.4148

Sigma

15.1961

0.8608

SCARCITY

− 4.0370

3.4445

LR \( {\chi}_{(16)}^2 \)

25.1200

 

INTERCON

− 0.4190

1.3091

Log-likelihood

− 652.8781

EXTERCON

1.3844

1.1885

Observation

159

Source: Estimated from the farmer survey in 2016, n = 159

* and ** indicate the significance levels at 10 and 1%, respectively

The descriptive statistics in Table 5 show that the majority of farmers in the study sites cultivate three crops year−1 with the average of 2.71. The three rice crops year−1 was recently adopted by farmers as flood-protected dykes were built to mitigate the negative impacts of flooding season on local farmers’ livelihoods in the upstream provinces of the Mekong Delta (An Giang, Dong Thap, Can Tho,…). The average farm size was 2.9 ha household−1, but it varies greatly among farmers with the standard deviation of 2.29. Although the 2013 Land Law of Vietnam still regulates the upper limit of landholding up to 3 ha household−1, the land accumulation is happening as an inevitable trend in the Mekong Delta. Recently, a great flow of rural labor force who originated from households with small-scaled production (less than 0.5 ha) out-migrated to cities for non-agricultural jobs. Therefore, the majority of them sold their land to richer farmers.

As the Mekong Delta has a long history of farming activities, rice farmers basically have abundant experiences with the average of 23.58 years. As the membership in organization was recognized to be important for agricultural development in Vietnam in general and the Mekong Delta in particular, the government has recently started to encourage farmers to be involved in organizations/cooperatives. Therefore, the number of farmers joining in organizations/cooperatives has been increasing. In fact, there were 41% of farmers in the study areas who are members in organizations/cooperatives.

Another interesting finding is that the female shares approximately 22% of rice farming activities in terms of participation magnitude. The interviewed rice farmers only finished 7.95 schooling years, which means that they did not complete secondary school. The majority of them perceived water scarcity as a serious problem in the near future. The two main causes leading to water scarcity in the Mekong Delta figured out by farmers are internal and external water use conflicts. Another challenging problem for rice production in the Mekong Delta is fragmented and small-scaled production. The study shows that the number of paddy plots household−1 averaged at 1.69. Now, we turn to discuss about the impacts of these socio-demographic variables on WE.

Table 6 shows that FARMSIZE, EXPER, ECO, and RISK are the significant determinants of the WE gaps. One of the most interesting findings presented in Table 6 is that farm size had a positively significant relationship with the WE, which is consistent with the study of Dhehibi et al. (2007). With the significance level of 10%, if farm size increases 1 ha, the WE accordingly increases 1.22%, keeping others constant. This finding is an importantly empirical evidence for policy makers to promote land accumulation (economies of scale) in rice production. It is easy to explain that farmers with larger farm size could manage their farms better in terms of water pumping. Normally, the rice farmers in the Mekong Delta use small boats to ferry pumping facilities to the paddy, then install them temporarily. For this reason, the transportation cost of small-scaled farmers was indifferent from that of large-farm farmers. So, promoting land accumulation will also contribute to cope with climate change in terms of water scarcity.

Another interesting finding is that those farmers who have more rice experiences consume water less efficiently. The possible explanation is that the farmers with more experiences are of course old people who are conservative, so they tend to refuse new technologies. The main mean that these farmers use in monitoring water demands of rice is eye observation.

ECO had a negatively significant correlation with the WE at 1%. The WE of those farmers who applied both eco-friendly practices simultaneously were 15.6% lower than those who did not. The possible explanation is that those who applied both eco-friendly practices did use more water to plant flowers around the periphery of the paddy fields in case of ecologically engineered rice model. Although the double eco-friendly practice was assumed to be more environmentally friendly in terms of lower use of environmentally detrimental inputs (pesticide and fertilizers), such practice resulted in lower WE. Therefore, an in-depth analysis on the trade-offs between water use efficiency and overall environmental efficiency (i.e., possible reduction of environmentally detrimental inputs) is needed for policy implications.

As a comprehensive indicator, RISK was negatively correlated with the WE. It means that those who perceived more risks in rice production consumed more water. Recently, climate change was recognized to be a serious challenge for agricultural production in the Mekong Delta. High temperature and changed rain patterns have significantly affected rice cultivation. To cope with these situations, different farmers had different adaptation strategies (water management, rice varieties,…), which results in a great variation of water consumption.

In short, in order to improve WE to adapt to increasingly serious climate change, encouraging land accumulation or large-scaled production is one of the possible solutions. The term “large-scaled production” was recently introduced to reorganize rice production in the Mekong Delta. Small-scaled rice farmers that have geographic proximity in terms of land are encouraged to join in a cooperative and to cultivate rice under the same technology package. This strategy is a feasible alternative as land accumulation is under consideration by government to remove upper limit of landholding at 3.0 ha household−1. In addition, although being a member in an organization is not significantly associated with WE in this current study, there are many other benefits like easy transfer of technologies and convenient sharing of experiences among farmers which may indirectly result in higher water use productivity (Aref 2011; Dung 2011). Therefore, as a whole, individual land accumulation can be implemented indirectly via land consolidation in cooperatives. Once farmers join in cooperatives, it is easy to solve the free-rider problem in water consumption by using market-based approach. The market-based approach here refers to collective pumping, which is an indirect system of water fee collection. Instead of separate and individual water pumping, the cooperatives are in charge of pumping water for the entire land area of all members (rice farmers). Then, individual rice farmers have to pay pumping fees for the cooperatives. By doing so, of course, it helps in solving the free-rider problem and improving WE.

In addition, as rice farmers in the Mekong Delta did not use any scientific measures to monitor water consumption, it is essential to develop simple and appropriate water management methods and to train them carefully on how to use and social benefits of water conservation through cooperatives’ monthly meetings.

Conclusions

This study investigates the WE for rice production in the Mekong Delta using SFA. By using stochastic frontier analysis approach, we found that rice farmers consume water very inefficiently. The average WE was 18.81, suggesting that rice farmers could reduce 81.19% of current water use, given existing technology. Therefore, further improvement in WE is indispensable under the contexts of climate change and water scarcity.

The factors affecting the WE gaps among farmers are investigated by using Tobit regression. We found that farm size positively affect the WE while experience, double eco-friendly practices, and perception of risk are negatively correlated with the WE. To improve WE and to adapt to increasingly serious climate change, encouraging land accumulation or large-scaled production is one of the possible solutions. In addition, it is also essential for the government and scientists to initiate the development of simple and appropriate water management methods and efficient information transfer.

The WE index derived from stochastic frontier approach is one of promising indicators because it takes the relationship between inputs and output into account. However, it should be noted that this approach is very sensitive to outliers (extreme values of inputs and output) when estimating the frontier. Therefore, data collection and analysis should be conducted and tested carefully.

Footnotes

  1. 1.

    For more details on the relationship between inputs and output in agricultural production, please see the two famous books of Coelli et al. (2005) and Kumbhakar and Lovell (2003). The basic theory or properties of production function and its common forms are also explained in detail in these books. Especially, the application of stochastic frontier analysis is also provided.

  2. 2.

    Ecological engineering is locally called as “paddy field surrounded with flowers”. It means that farmers will cultivate some flowering plants around the periphery of paddy fields to attract more natural enemies. These natural enemies will suppress pest populations under a damage threshold.

    VietGAP is the abbreviation of the Vietnamese Good Agricultural Practices, which means that farmers have to apply and manage wisely fertilizer and pesticide use, irrigation water, waste management, traceability, etc. These indicators are regulated in Decision No. 99/2008/QD-BNN October 15, 2008 of the Minister of Agriculture and Rural Development

    GlobalGAP is the abbreviation of the Global Good Agricultural Practices, which was firstly introduced in 1997 under the term EUREPGAP by Euro-Retailer Produce Working Group. Since 2007, the term EUREPGAP was replaced by GlobalGAP. The standards or requirements of this practice is more stringent than that of VietGAP.

    Floating rice, which is a unique farming practice in the Vietnamese Mekong Delta, is cultivated during flooding season, consumes no pesticide and very little amount of fertilizer. Thus, floating rice is good for not only consumers but also natural environment. Moreover, floating rice also play an important role in biodiversity conservation. In fact, floating rice fields are home to 43 fish species and 54 plant species (Quyen and Vu 2014; Vu and Quyen 2014

    Large-scaled rice is one of the solutions, which aims at reducing production cost (including agro-chemicals) through collective pumping, same seasonal calendar, etc. Moreover, the rice farmers have to join in a farmer group, and to follow a similar knowhow such as “one must do, five reductions” “3 reductions, 3 gains”, etc.

Notes

Acknowledgments

This research (ID number R11616) was funded by RONPAKU, JSPS. We are thankful to the anonymous reviewers for their appreciated recommendations and comments on the manuscript.

Compliance with Ethical Standards

Competing Interests

The authors declare that they have no competing interests.

References

  1. Aigner D, Lovell CA, Schmidt P (1977) Formulation and estimation of stochastic frontier production function models. J Econ 6:21–37MathSciNetCrossRefMATHGoogle Scholar
  2. Aref F (2011) Agricultural cooperatives for agricultural development in Iran. Life Sci J 8:82–85Google Scholar
  3. Assouline S, Russo D, Silber A, Or D (2015) Balancing water scarcity and quality for sustainable irrigated agriculture. Water Resour Res 51:3419–3436CrossRefGoogle Scholar
  4. Baran E, Myschowoda C (2009) Dams and fisheries in the Mekong Basin. Aquat Ecosyst Health Manag 12:227–234CrossRefGoogle Scholar
  5. Biggs D, Miller F, Hoanh CT, Molle F (2009) The delta machine: water management in the Vietnamese Mekong Delta in historical and contemporary perspectives contested waterscapes in the Mekong region: hydropower, livelihoods and governance:203–225Google Scholar
  6. Billi A, Canitano G, Quarto A (2007) The economics of water efficiency: a review of theories, measurement issues and integrated models water use efficiency and water. Productivity 231Google Scholar
  7. Bouman B (2007) A conceptual framework for the improvement of crop water productivity at different spatial scales. Agric Syst 93:43–60CrossRefGoogle Scholar
  8. Bouman B, Tuong TP (2001) Field water management to save water and increase its productivity in irrigated lowland rice. Agric Water Manag 49:11–30CrossRefGoogle Scholar
  9. Bravo-Ureta BE, Pinheiro AE (1997) Technical, economic, and allocative efficiency in peasant farming: evidence from the Dominican Republic. Dev Econ 35:48–67CrossRefGoogle Scholar
  10. Cantrell R (2004) Challenges and opportunities for rice-based farming in the international year of rice and beyond. Paddy Water Environ 2:1–4CrossRefGoogle Scholar
  11. Coelli TJ, Rao DSP, O'Donnell CJ, Battese GE (2005) An introduction to efficiency and productivity analysis. SpringerGoogle Scholar
  12. Dhehibi B, Lachaal L, Elloumi M, Messaoud A (2007) Measuring irrigation water use efficiency using stochastic production frontier: an application on citrus producing farms in Tunisia. Afr J Agric Resourc Econ 1:1–15Google Scholar
  13. Dugan PJ, Barlow C, Agostinho AA, Baran E, Cada GF, Chen D, Cowx IG, Ferguson JW, Jutagate T, Mallen-Cooper M, Marmulla G, Nestler J, Petrere M, Welcomme RL, Winemiller KO (2010) Fish migration, dams, and loss of ecosystem services in the Mekong basin. Ambio 39:344–348CrossRefGoogle Scholar
  14. Dung NM (2011) Characteristics of the agricultural cooperatives and its service performance in Bac Ninh province, Vietnam International Society for Southeast Asian. Agric Sci 17:68–79Google Scholar
  15. Gosling SN, Arnell NW (2016) A global assessment of the impact of climate change on water scarcity. Clim Chang 134:371–385CrossRefGoogle Scholar
  16. Greene William H (2012) LIMDEP version 10 econometric modeling guide. Econometric Software. Inc, PlainviewGoogle Scholar
  17. GSO (2016) Statistical yearbook of Vietnam Statistical Publishing HouseGoogle Scholar
  18. Hong NB, Yabe M (2017) Improvement in irrigation water use efficiency: a strategy for climate change adaptation and sustainable development of Vietnamese tea production. Environ Dev Sustain 19:1247–1263CrossRefGoogle Scholar
  19. Jondrow J, Knox Lovell C, Materov IS, Schmidt P (1982) On the estimation of technical inefficiency in the stochastic frontier production function model. J Econ 19:233–238MathSciNetCrossRefGoogle Scholar
  20. Karagiannis G, Tzouvelekas V, Xepapadeas A (2003) Measuring irrigation water efficiency with a stochastic production frontier. Environ Resour Econ 26:57–72CrossRefGoogle Scholar
  21. Kumbhakar SC, Lovell CK (2003) Stochastic frontier analysis. Cambridge University Press, CambridgeMATHGoogle Scholar
  22. Kumbhakar SC, Wang H, Horncastle AP (2015) A practitioner’s guide to stochastic frontier analysis using Stata. Cambridge University Press, CambridgeCrossRefGoogle Scholar
  23. Lobell DB, Burke MB, Tebaldi C, Mastrandrea MD, Falcon WP, Naylor RL (2008) Prioritizing climate change adaptation needs for food security in 2030. Science 319:607–610CrossRefGoogle Scholar
  24. McGuckin JT, Gollehon N, Ghosh S (1992) Water conservation in irrigated agriculture: a stochastic production frontier model. Water Resour Res 28:305–312CrossRefGoogle Scholar
  25. Meeusen W, Van den Broeck J (1977) Efficiency estimation from Cobb-Douglas production functions with composed error. Int Econ Rev 18:435–444CrossRefMATHGoogle Scholar
  26. Mekonnen MM, Hoekstra AY (2016) Four billion people facing severe water scarcity. Sci Adv 2:e1500323CrossRefGoogle Scholar
  27. MRC (2009) Economic, environmental and social impact assessment of basin-wide water resources development scenarios. Mekong River Commission, VientianeGoogle Scholar
  28. Nhan DK, Be N, Trung NH (2007) Water use and competition in the Mekong Delta, Vietnam challenges to sustainable development in the Mekong Delta: regional and national policy issues and research needs the sustainable Mekong research. Network:143–188Google Scholar
  29. Quyen LC, Vu TH (2014) Composition of wild fish in floating rice in Vinh Phuoc and Luong an Tra communes, tri ton district, An Giang province in the flood season in 2014 research Center for Rural Development. An Giang University, An GiangGoogle Scholar
  30. Reinhard S, Knox Lovell C, Thijssen GJ (2000) Environmental efficiency with multiple environmentally detrimental variables; estimated with SFA and DEA. Eur J Oper Res 121:287–303CrossRefMATHGoogle Scholar
  31. Reinhard S, Lovell CK, Thijssen G (1999) Econometric estimation of technical and environmental efficiency: an application to Dutch dairy farms. Am J Agric Econ 81:44–60CrossRefGoogle Scholar
  32. Scheierling S, Treguer D, Booker J, Decker E (2014) How to assess agricultural water productivity? Looking for water in the agricultural productivity and efficiency literature (July 1, 2014) World Bank Policy Research Working Paper (6982)Google Scholar
  33. Sharma B, Molden D, Cook S (2015) Water use efficiency in agriculture: measurement, current situation and trends managing water and fertilizer for sustainable agricultural intensification:39Google Scholar
  34. Shiklomanov IA, Rodda JC (2004) World water resources at the beginning of the twenty-first century. Cambridge University Press, CambridgeGoogle Scholar
  35. Tang J, Folmer H, Vlist AJ, Xue J (2014) The impacts of management reform on irrigation water use efficiency in the Guanzhong plain, China. Pap Reg Sci 93:455–475CrossRefGoogle Scholar
  36. Tu VH (2017) Resource use efficiency and economic losses: implications for sustainable rice production in Vietnam. Environ Dev Sustain 19:285–300.  https://doi.org/10.1007/s10668-015-9724-0 CrossRefGoogle Scholar
  37. Tuong P, Bouman B, Mortimer M (2005) More rice, less water—integrated approaches for increasing water productivity in irrigated rice-based systems in Asia. Plant Product Sci 8:231–241CrossRefGoogle Scholar
  38. Vu TH, Quyen LC (2014) Biodiversity of plants in floating rice in Vinh Phuoc and Luong An Tra communes, Tri Ton district, An Giang province in the flood season in 2014 research Center for Rural Development. An Giang University, An GiangGoogle Scholar
  39. White I (2002) Water management in the Mekong Delta: changes, conflicts and opportunities. Unesco Paris, ParisGoogle Scholar
  40. Ziv G, Baran E, Nam S, Rodríguez-Iturbe I, Levin SA (2012) Trading-off fish biodiversity, food security, and hydropower in the Mekong River basin. Proc Natl Acad Sci 109:5609–5614CrossRefGoogle Scholar

Copyright information

© Springer Nature Singapore Pte Ltd. 2018

Authors and Affiliations

  1. 1.Department of Rural Socio-Economics, College of Rural DevelopmentCan Tho UniversityCan ThoVietnam
  2. 2.Laboratory of Environmental Economics, Division of International Agricultural Resource Economics and Business Administration, Department of Agricultural and Resource Economics, Faculty of AgricultureKyushu UniversityFukuokaJapan

Personalised recommendations