Does Adaptation to Saltwater Intrusion Improve the Livelihoods of Farmers? Evidence for the Central Coastal Region of Vietnam

Abstract

Saltwater intrusion poses severe threats to rice farming in Vietnam. Farmers can adapt by producing other crops or switching to other production models. This study evaluates the impact of implementing different saltwater-intrusion-adaptation strategies on farmers’ livelihoods by applying propensity score matching to cross-sectional survey data for 414 farmers in the Central Coastal region of Vietnam. We consider both economic and social indicators and find that there is considerable heterogeneity in the outcomes. With the exception of switching to new rice varieties, all adaptation strategies considered in the study significantly increase saline-land productivity, mainly as a result of higher revenues. Moreover, for these strategies, food security is found to be significantly higher, whereas life satisfaction is only higher for those farm households that cultivate vegetables, shrimp, or lotus-fish. Adopting new rice varieties is found to significantly decrease saline-land productivity, whereas the social impacts are not significant.

1. Introduction

In Southeast Asia, Vietnam is amongst the countries that are expected to be most heavily affected by climate change due to its large coastline, its agriculture-dependent economy, and its poorly developed rural areas [1,2,3]. Vietnam’s coastal regions are particularly regarded as being highly vulnerable because of pressure from population increases and from sea-level rise [4]. Anticipated temperature increases of 0.8–2.7 °C by 2060 [5] go hand in hand with an expected sea level rise of up to 78–95 cm by 2100 [6], changing river flow discharges, increasing groundwater extraction rates [7,8] and increasing levels of salinization of land surface [3,9]. Moreover, natural disasters, such as typhoons, floods, and droughts, are increasingly threatening rice cultivation in Vietnam [2,3,10]. It is estimated that out of 650,000 ha of rice-production surface in the coastal areas of Vietnam, 100,000 ha is at risk of saltwater intrusion due to sea level rise [11]. The stakes are high as Vietnam is the second biggest rice exporter in the world and climate change impacts are estimated to reduce productivity levels in this sector by 2050 by 3.4–6.7% compared to 1990 levels [12]. This productivity drop might not only strongly impact the livelihoods of the affected farmers, but also the stability of the Vietnamese economy as a whole. Furthermore, Narloch and Bangalore (2018) report that Vietnam’s poorer households are disproportionally exposed to environmental and climate risks [13]. Rice production is a crucial source of income for 75% of the poor households in the country and about half of non-poor households in its rural areas [3]. Implementing suitable adaptation strategies will be important to try and reduce the negative impacts of climate change on farmers’ livelihoods.

In this study, we will focus on the impact of adopting different adaptation methods to address problems of saltwater intrusion (SWI) in the Central Coastal region of Vietnam. This region has been identified as being considerably vulnerable to climate change [14], in particular due to its high exposure to salinity problems, droughts, and sea level rise [2]. The two main causes of saltwater intrusion—sea level rise and higher evaporation rates due to temperature increases [14]—are increasing the impacts of SWI in the region, both today and in the future. Over the past couple of years, local authorities have been promoting the uptake of adaptation methods by informing farmers about the impacts to be anticipated in the (near) future and by recommending specific adaptation methods that are suitable given the natural, social, and economic circumstances in the area. As a result, rice farmers in the Central Coastal region of Vietnam have recently started to apply a variety of adaptation methods, including growing new varieties of rice that are more tolerant to salt, switching to the cultivation of other crops (papyrus or vegetables), or implementing other production models (shrimp or lotus-fish cultivation). Local authorities typically provide technical assistance for adopting the recommended adaptation methods, yet there is no financial support. This implies that a large majority of the farmers in the region have to borrow capital to implement the adaptation methods.

The current situation in the Central Coastal region of Vietnam—where some farmers have implemented SWI adaptation methods, while others have not—presents an opportunity to evaluate the impacts of the different adaptation methods on the livelihoods of farmers. This will help policymakers to understand the effectiveness of the different SWI adaptation strategies and develop recommendations related to future actions, both within the study area and beyond. Most studies in the field of climate change adaptation in agriculture focus either on the determinants of the farmers’ choice for a specific adaptation method or on constraints and barriers related to the implementation of such strategies [10,15,16]. Only a limited number of studies have investigated the impact of adopting climate change adaptation methods on the livelihoods of farmers given the complexity involved in impact evaluation. These studies tend to focus mostly on the economic impacts, neglecting the other pillars of sustainable development [17,18]. Moreover, research that evaluates the impacts of CC adaptation on the livelihoods of farmers is limited both for Vietnam in general, and for the Central Coastal region in particular. Therefore, in this study, we will investigate the impacts of SWI adaptation strategies on farmers’ livelihoods in the study area in a comprehensive way by focusing both on economic indicators (saline-land productivity and net farm income) and social indicators (rate of children attending school, food security, and life satisfaction). Based on key informant interviews, the FGDs, and field trip observations, we noted that some of the current SWI adaptation methods are impacting the environment—i.e., draining water sources and deteriorating soil quality. As we rely on self-reported data from farm households to investigate the impact of adaptation methods on farmers’ livelihoods, it is difficult to take the environmental impacts into account. We apply propensity score matching to overcome problems related to self-selection in treatment. By considering both the economic and social outcomes of adaptation choices, we can investigate whether there is a trade-off between both. Local authorities can draw on our results when designing policy interventions in order to tackle problems related to climate change in general and to saltwater intrusion in particular.

The remainder of this paper is structured as follows: Section 2 presents a literature review that focuses on the different economic and social outcome variables considered in our study. Section 3 describes the study area, the data collection process, the different SWI adaptation methods under consideration, and the econometric approach. The results are presented in Section 4, while Section 5 looks into the policy implications of our research.

2. Literature Review

In this study, we evaluate farmers’ livelihoods based on a range of social and economic indicators. At the start of the research project, we organized focus group discussions (FGDs) in which representatives of the local authorities of both provinces, researchers in the climate change field as well as the representatives of cooperatives and farmers participated. The purpose of these FGDs was to assess the current status of rice production and the impacts of SWI (see Appendix A) and to identify the most useful social and economic indicators to evaluate adaptation practices (see Appendix B). Social and economic indicators are typically also complemented by environmental indicators in sustainability assessments—see [18] for a literature review on measuring agricultural sustainability in developing countries. However, we decided not to include environmental indicators in this study, as ecological assessments require data and expertise that is often missing at the farm-household level. As we rely on self-reported data from farm households to investigate the impact of adaptation methods on farmers’ livelihoods, it is difficult to take the environmental impacts into account. These impacts are, however, important for local governments when designing adaptation policies, and can be investigated through other research methods—e.g., key informant interviews and expert consultations [19].

2.1. Economic Indicators

The economic indicators used to investigate the impact of farmers’ choices on their livelihoods can reflect productivity and profitability [18]. As we focus on the evaluation of different SWI adaptation strategies, we focus on productivity indicators following [17,20]. The first economic indicator in our study is saline-land productivity, which is measured as the value of agricultural output (expressed in monetary terms) per hectare of salted land. Land productivity indicators often feature in sustainability assessments of agricultural production in general [17] and of rice production in particular [21]. The second economic indicator we include in our analysis is net farm income as measured by the ratio of total net income from the agricultural activities of the household divided by the total land used to generate this income [20,22]. This indicator thus looks beyond the salted plots of the farm households and focuses on all agricultural activities and all farm land. It thus provides a complement to the first indicator in a similar way as conducted by [23,24]. Facing the same SWI situation, farmers who apply SWI adaptation methods can be expected to have higher saline-land productivity than those who did not adapt, and, as a result, we would expect to also find a higher net farm income amongst the adapting farm households. In this study, we did not select savings as an economic indicator because survey questions related to savings are often perceived as sensitive and farmers in the study area try to avoid answering them.

We assume that all adaptation methods positively impact the economic indicators selected in our study as these methods are expected to increase both (saline) land productivity and farm incomes.

2.2. Social Indicators

There is a lot of literature on social indicators to be included in sustainability assessments (see [18] for a review). Food security is included in most social impact analyses related to agricultural activities [24,25,26,27,28,29], whereas distributional impacts are also being taken into consideration on many occasions [17,28,30,31]. A second widely used social indicator is the level of education within the households as measured by either the education level of the head of the household or the rate of school-aged children in the family that go to school [24,27,28,29]. Less widely used social indicators include housing facilities [32] or working and living conditions [29], land tenure as a means to reducing poverty and increasing food security [24,33], and knowledge and awareness of resource conservation [28]. More recently there is a growing interest in quality-of-life assessments based on subjective well-being (SWB) data—see, for instance, [34,35,36].

In this study, we focus on three social indicators—food self-sufficiency (referred to as “food security” in this study), the rate of children attending school, and life satisfaction—as these indicators were identified by focus group discussions (FGDs) as the most appropriate for our study area—rural areas in Vietnam.

Food security is one of the main priorities in the agricultural development policies of the Vietnamese government to try and help the 2.3 million poor households in rural Vietnam to escape poverty and ensure their subsistence. In our survey, we use the short form of the Household Food Security Scale (HFSS) developed by [37] to measure food security, as this set of questions has been validated in developing countries and/or amongst low-income households on numerous occasions—see, for instance [38,39]. Next, as the short form of the HFSS focuses on constraints at the household level and ignores broader determinants such as “the community wide unviability of sufficient quantities of nutritious food or religious beliefs” [37], we believe this scale better corresponds to the aim of our study—evaluating the impact of SWI adaptation strategies at the household level. However, we must acknowledge that the HFSS strongly focuses on the quantitative dimension of food security—e.g., having enough (calories) and not so much on its qualitative dimension (variety and quality of food products). Appendix C presents the six questions in the short form of the HFSS.

A second social indicator we include in our study is the rate of children attending school. The FGDs pointed to the fact that most farmers in the study area are poorly educated and that child labor is common among many farm households. Therefore, not all children have equal opportunities to attend school in the study area. The rate of children attending school better reflects the current educational situation as well as the financial situation of farm households. We observed that the household with a higher income tends to send all kids to school and the one with a lower income prefers to send the kids to farming work rather than to school. Hence, we believe that focusing on this indicator yields more useful insights given that the SWI adaptation choices impact both incomes and demand for labor.

A third social indicator that we include in our study is life satisfaction, as measured by the Satisfaction With Life Scale (SWLS) by [40]—see Appendix D for the full set of questions. By using the SWLS, we focus on life satisfaction instead of affect or eudaimonia, two related concepts within the literature on subjective well-being [41] as we are primarily interested in the farmers’ evaluation of their lives in the broadest and most cognitive sense.

The social impacts of implementing different adaptation methods can relate to income effects, differences in the agricultural product(s) being cultivated, and labor intensity effects. First, the income effects might influence all three social indicators as higher incomes from farming activities will increase the odds of farmers being able to let their children go to school and will most likely lead to higher self-reported levels of both food security and life satisfaction. Second, the agricultural product effect relates to the characteristics of the agricultural outputs involved in the adaptation method—e.g., their potential for consumption within the household (importance in daily meals) and the ability to store the outputs in order to depend less on the market conditions at the time of harvesting. In this regard, rice production is expected to positively influence food security given that is an important part of daily meals in rural Vietnam and farmers can produce and stock rice for self-consumption. Third, through our survey, we find that labor accounts for the majority of the costs of production in the crop production group: labor costs make up 41.5%, 85.8%, and 55.7% of the total costs for the cultivation of new rice varieties, papyrus, and vegetables, respectively. In the aquaculture production group, labor accounts for only 10.9% and 27.4% of the total costs of shrimp and lotus-fish cultivation, respectively. Therefore, labor-intensive adaptation methods are expected to negatively impact the rate of children attending school because children need to stay at home to help their parents in farming activities.

3. Material and Methods

3.1. Study Area

The Central Coastal region of Vietnam is dominated by a subtropical humid climate with two main seasons: dry and rainy. This region has nine provinces with 12.6 million inhabitants and accounts for 19% of the total surface of Vietnam in which 706,000 ha of paddy fields are located (10% of all paddy fields in the country). In our study, we selected the Thua Thien Hue and Quang Nam provinces because these two provinces are leaders in both the rice production surface and the proportion of SWI-affected areas. In the Thua Thien Hue province, 6% of the 84,400 hectares of paddy land is affected by SWI. Within this province, the district of Quang Dien (Thua Thien Hue Province) was selected as the first study area given its high proportion of paddy fields affected by SWI (13%) [42]. The province of Quang Nam has 87,396 ha of paddy land, of which 7816 ha is saline (9%) [42]. Here, we chose the Duy Xuyen district (Quang Nam Province) as the second study area since it is the district in which SWI is most problematic–affecting 2258 hectares of paddy field, or 29.1% of the total paddy land surface. Based on annual reports on salt intrusion of local authorities in each district, three different communes were selected as representatives of three different levels of SWI impacts (high, moderate, and mild) [42]. In both districts, the majority of affected paddy fields are located alongside the coast or a tidal river. The risk of SWI is particularly high in the summer season, with large negative impacts on rice production. In some areas, the water sources used for irrigation have turned saline due to high salinity levels of the soil in combination with seawater intrusion, and this is sometimes even in the rainy season. To cope with SWI, about half of the farmers in the study area have started to apply adaptation methods over the last five years, and many more plan to apply adaptation methods over the coming years.

3.2. Data Collection

The data for our study is collected through a survey at the farm household level. The survey was designed based on a literature review, field observations, and inputs from different focus group discussions (FGDs) in the Quang Dien and Duy Xuyen districts. Field observations allowed us to identify the adaptation methods that are currently in place in the study area, whereas the FGDs provided input for both the initial design of the questionnaire and its finalization after a final round of feedback as well as the selection of the most appropriate economic and social livelihood indicators for the study area. The FGDs were organized with representatives of the Agricultural and Rural Development Department (Economic division), local cooperatives, Farmer’s Unions, and farmers. A pilot survey was conducted to pre-test the questionnaire with a sample size of 30 respondents in each district.

We used a multi-stage stratified random sampling strategy for our research based on the percentage of households in each district who did or did not adapt to SWI, taking into account the current implementation rates of the different adaptation methods to calculate the number of each household type to be included in the survey. The sample of each adaptation method as well as of the non-adapters is divided respecting the three different SWI levels (high, moderate, and mild) and also takes into account the percentage of poor households. There are various ways to determine the level of saltwater intrusion—e.g., testing for soil quality or checking salinity levels in irrigation water—but due to the lack of official information on SWI levels in the study area, we make use of the expertise of the FGD participants who indicated which plots belong to which SWI levels. In the final stage, individual farm households were chosen randomly from the official household lists of the communes. To collect the data, face-to-face interviews were conducted in 2020 by 12 interviewers who had been intensively trained. The eligible interviewees are heads of households or their spouses. For the non-adaptation cases, we made sure that the farmers were still growing rice in their salted paddy fields. These farmers could have previously applied adaptation methods, but not for the last four years. For the adaptation cases, we only included farmers who had been implementing an adaptation method for the last two to three years before being interviewed in order to be able to elicit all data required for analysis. We asked for production costs and yields for the crops on the farm and for net farm income in the year 2019. The total sample for each district was set to be 200, but we visited 220 farmers to have a 10% back-up. In the end, we had 205 valid and completed questionnaires for the Duy Xuyen district and 209 for the Quang Dien district.

3.3. Overview of SWI Adaptation Methods in the Study Area

Given the substantial SWI impacts in the study area, some farmers started applying adaptation methods ten years prior to our study. At the start of our research, we compiled a list of the different adaptation methods in the study area based on two conditions: (1) the adaptation methods are still being implemented today, and (2) the selected methods have been applied for at least two years prior to the interviewing period. There are five different adaptation methods that are taken into consideration in our study together with the frequencies that were found in our sample (414 observations). We found that 55% of the farm households were still cultivating traditional rice varieties (non-adaptation), while the adapting households mostly converted to new rice varieties (NR–21%), shrimp production (SR—11%), or papyrus cultivation (PP—7%). The lotus-fish model (LF—3%) is a relatively new adaptation method that is gaining interest, while vegetable production (VG—2%) is not often selected as this adaptation method is both labor and water-intensive and can mostly be employed on mildly salted land making it less future proof. (The detailed description of each adaptation method is presented in Appendix E). The advantages and disadvantages of each adaptation method have been listed below:

• Switching to new salt-tolerant varieties of rice: Farmers continue to produce rice but switch to new varieties that are more salt-tolerant. The local cooperatives sell new rice varieties to farmers at a price below that of the market as a result of government subsidies. The main disadvantage of this method is that the yield is about 25% lower than that of traditional rice.

• Moving to planting papyrus: This method implies that farmers convert their paddy plots into papyrus fields. This crop is neither labor, nor capital, nor time-intensive. The main limitation of this adaptation method is that the consumption market tends to be unstable: papyrus is the main input of the mat industry, which faces a narrowing demand as consumer preferences change.

• Moving to shrimp cultivation: Farmers convert their land to build ponds for shrimp cultivation. In the process, the land is sometimes combined with that of the farmer’s neighbors. This adaptation method requires high initial investments and technical support from the staff of the agricultural extension office. Shrimp production can bring about higher profits but farmers face many risks ranging from diseases to fluctuating selling prices.

• Moving to vegetable cultivation: Farmers convert their paddy plots into vegetable fields. Farmers can produce three seasons per year while the type of vegetables depends on the season. For example, spring onion, cabbage, and peas will be cultivated from March until May; cucumber, okra, and courgettes will be cultivated from June to September; and sweet potatoes, pumpkins, and potatoes from July to February. However, in our study, we do not distinguish between different types of vegetables being produced. Revenues for farmers typically increase, yet profits may not as this method is labor-intensive (planting, caring, and harvesting). Moreover, cultivating vegetables requires substantial amounts of irrigation and farmers may face water shortages in the dry season.

• Moving to lotus-fish cultivation: Farmers convert their paddy fields into fish ponds in which lotus plants are grown and fish are kept at the same time. Farmers get revenues from both fish and lotus products. This method requires lower initial investments than shrimp cultivation, and also the risks involved are lower.

In addition, it is noted that the value chain of vegetable production is dominated by wholesalers who keep prices paid to farmers low. As a result, we observe that farmers who switched to vegetable production in the past are increasingly moving towards other adaptation methods. Amongst the five adaptation methods, papyrus cultivation and vegetable production are considered to be the riskiest as selling prices for farmers tend to fluctuate, whereas shrimp production is the most difficult adaptation method to implement due to the high initial investments that are required and the intensive techniques involved.

3.4. Conceptual Framework and Econometric Approach

The aim of this study is to estimate the causal effect on a number of outcome variables of applying a number of adaptation methods to cope with SWI. We will estimate this effect in two ways: once for the whole sample and once for each of the five adaptation methods. This section starts with the econometric setup and then discusses the result.

In general, ${∆}_i$, the effect of adaptation by farmer i on the outcome variable is expressed by the following equation:

(1)${∆}_i=Y_i^A-Y_i^N$

where $Y_i^A$ and $Y_i^N$ represent the outcome variable of farmer i in the case it adapts (A) and in the case it does not adapt (N), so ${∆}_i$ is i’s causal treatment effect. A fundamental problem of Equation (1) is that either $Y_i^A$ or $Y_i^N$ is unobservable: a household either adapts or it does not. The observed outcome can be written as follows:

(2)$Y_i=D_i{Y}_i^A+\left(1-D_i\right)Y_i^N\hspace{1em}{D}_i=\mathrm{0,1}$

In Equation (2), the treatment dummy $D_i$ is 0 in the case of non-adaptation and 1 in the case of adaptation. In this paper, we only focus on estimating the average treatment effect on the treated (ATT), which is the following:

(3)$\widehat{{∆}_i}=E\left\{Y_i^A-Y_i^N|D_i=1\right\}$

$Y_i^N$ is clearly unobserved. It is the counterfactual outcome in case the adapter had not adapted. If adaptation was random, $\widehat{{∆}_i}$ could be estimated by replacing $Y_i^N$ by the outcomes of non-adapting farmers j. The problem is that adaptation to cope with SWI is not random but may depend on both the farmers’ own characteristics and their farms’ characteristics. As farmers select themselves for treatment, studying the impact of an adaptation method on economic and social outcomes is not easy, as non-adapting farmers differ in characteristics from those adapting. Imagine that better-educated farmers are both more productive and more inclined to adapt. Comparing their saline-land productivity after adapting with the productivity of a less educated non-adapting farmer might overestimate the impact of adapting. A part of the effect might be due to education and not due to the adaptation technique. Hence, OLS estimation might cause biased estimates. An instrument-variable (IV) approach can be applied by using one variable in the treatment equation as an instrument for the specification of the outcome equation [43]. However, finding a valid instrument for empirical studies is a huge challenge [44]. Another widely applied method to solve the problem of self-selection bias is propensity score matching (PSM), which will be applied in this paper. The mechanism of the PSM technique is to make the treatment group (adapters) and the control group (non-adapters) comparable in terms of observable variables [45]. Each adapter is matched with a group of non-adapters that are similar in terms of observable characteristics. Unlike OLS or IV, PSM is a non-parametric technique, it doesn’t require the assumption of functional form or normal distribution of unobserved covariates. This technique only requires a set of observable covariates [44]. PSM can be performed if three data conditions are fulfilled: (1) treatment and control groups are produced from the same data source, (2) control and treatment groups are subject to the same economic incentives, and (3) a sufficient number of variables are required that can explain both outcomes and participation probability [44,46,47]. These conditions are fulfilled for our dataset.

There are two identifying assumptions for estimating the ATT [48,49]. The first assumption is the unconfoundedness or conditional independence assumption: selection into treatment is random, conditional on a number of observed variables $X_i$. More formally, $\left.Y^A,Y^N\perp D_i\right|X_i$. The second assumption concerns the overlap or common support condition: for all treated observations a control observation is needed, or the probability of choosing adaptation given $X_i$ should be lower than 100%. More formally, this can be written as $p\left(D_i=1|X_i\right)<1$.

PSM does not match adapters and non-adapters directly on $X_i$, but matches them on the propensity score and the probability of choosing to adapt. Reference [50] have shown that if assignment to treatment is unconfounded given $X_i$ (so $Y^A,Y^N\left.\perp D_i\right|X_i$), then it is unconfounded also given the propensity score $\left(Y^A,Y^N\perp D_i|p\left(X_i\right)\right)$, with $Y^A$ and $Y^N$ being the outcome of interest in the case of adaptation and in the case of non-adaptation, respectively. The common support or overlap condition requires a necessary overlap between treatment and control groups in the distribution of the propensity score [51]. Note that common support is based on the propensity score and no common support of X is required. Hence, we only used the observations in the common support areas (e.g., where the PS of treated observations is not lower than the PS of control observations and the PS of control observations is not larger than the PS of treated observations).

The probability of assignment to treatment, conditional on pre-treatment variables is expressed by the following:

(4)$p\left(X_i\right)=\mathrm{E}\left\{D_i=1|X_i\right\};p\left(X_i\right)=F\left\{h\left(X_i\right)\right\}$

where F{.} is the normal or logistic cumulative distribution.

Consequently, the PSM procedure consists of five steps. First, the probability of taking the treatment is estimated based on the pre-treatment characteristics $X_i$. Second, the region of common support is determined. Then the balancing property is tested. Fourth, for each subject, the propensity score is calculated. Finally, the outcome of interest of an adapting farmer (e.g., income) is compared with the outcome of a matched non-adapting farmer with the same propensity score [50].

The ATT estimated by PSM is the difference between the observed outcome of the adapters and the expected outcome of a group of non-adapters with the same propensity score. The second part of Equation (5) is the counterfactual that has been collected among the non-adapting farmers indexed by j.

(5)$\widehat{{∆}_i}=E\left\{E\left(Y_i|p\left(X\right),D_i\right)=1-E\left(Y_j|p\left(X\right),D_j=0\right)\right\}$

where ${∆}_i$ indicates the ATT estimator and p(X) refers to the propensity scores as introduced in Equation (4).

In this study, the adapters and non-adapters are matched by the kernel method with the default Gaussian kernel if both have similar PS. The big advantage of this method is the lower variance because more information from control groups and weighting regression is used to produce the counterfactual outcome [52]. During the estimation process, the PSM method has to take into account the variance attributable to the derivation of the propensity score, the finding of common support, and the way in which treated and non-treated individuals are matched [52]. Hence, the standard errors might be incorrectly estimated [44]. We applied bootstrapping to deal with this problem. This method is a common way to calculate the standard errors in the context that analytical estimations are “biased and unavailable” [52]. Technically, each bootstrap conducts the re-estimation of the results of the whole estimation process (propensity score, common support, …). We repeated the bootstrapping 50 times, which resulted in 50 bootstrap samples and 50 estimates of the ATT [52].

The ATT is considered as the average effect on the household that finally adopted one method to cope with SWI [47,49]. Reference [49] favored the ATT over ATE, arguing that in the case of many potential barriers to the participation and completion of one program, it might be unrealistic to estimate the effects of the program if it were applied to all observations. In our case, due to potentially high requirements to apply some adaptation methods, such as shrimp or lotus-fish cultivation (higher initial investments), it might be inexact to estimate the effects of one adaptation method if all farmers apply it. We use the Rosenbaum bound test to check the sensitivity of ATT by calculating the p-value of the Wilcoxon signed rank. The smaller the p-value is (less than 0.05), the more significant the treatment effect is [50,53]. Finally, we compare mean standardized differences before and after matching to assess the comparability of the treated and control groups in matched samples [50]. The standardized difference that has been referred to as Cohen’s Effect Size Index of 0.2, 0.5, and 0.8 can be used to indicate small, medium, and large effect sizes, respectively [54]. Reference [55] suggests that this value should be compared to 0.2 to evaluate whether the two groups are comparable.

In this study, we conduct PSM twice: once for adaptation in general and once for adaptation to each of the five specific methods. The first stage of PSM is estimating propensity score (PS) by using a probit model to calculate the probability of each household adapting (adaptation in general and adaptation in each method). In the probit model, we only include variables that are unaffected by adaptation [52] (to satisfy the conditional independence assumption): labor force, age, education of household head, number of dependents, land certificate, SWI situation, and percentage of salted land. There are other potential variables such as education, member of organization, etc. but the propensity score was not found when we added these variables into the probit model.

4. Results

4.1. Farm Households’ Characteristics

Summary statistics of farm households‘ characteristics are presented for adapters and non-adapters (see also Appendix F). These characteristics are the potential variables to enter a probit model that allows one to calculate the probability of each household of adapting—first adaptation in general, and second adopting each adaptation method individually). The average age of the household head in our sample is 51 years, with 41% of the household heads having completed secondary school. We observe significant differences in the education level amongst the adapters. The average level of education of farmers who apply NR, SR, and LF is lower than that of the non-adapters. These results reflect that education might not be an important factor in the adaptation decision of households in the case of the adaptation methods mentioned above. The number of family members working on the farm is relatively similar amongst adapters and non-adapters as we only find significant differences for farm households that apply NR (negative) and SR (positive). This finding can be explained by the fact that SR production is more labor-intensive than other adaptation methods. Finally, the percentage of dependents in the household is the proportion of dependents—i.e., family members younger than 12 or older than 70—to the total number of family members in the households. We find that the percentage of dependents in the household of farmers who apply LF is significantly lower than that of non-adapters. This result can be explained by the fact that LF is a labor-intensive adaptation method.

We find that farmers who apply SR or LF are, on average, more likely to have short-term land certificates compared to farmers who do not apply adaptation methods. The reason might be that farmers more easily convert or accumulate land if this land is rented for a shorter period of time.

Regarding the social capital variables, we find that 88% of the household heads are members of the Farmers’ Union and that 89% of their wives are members of the Vietnamese Women’s Union. No significant differences were found between non-adapters and adapters. The Vietnamese Women’s Union and Farmers’ Union are both political organizations that operate under the control of the Vietnamese government. The duties of the Farmers’ Union include protecting and enhancing the economic well-being and quality of life of farmers, fishers, ranchers, and rural communities. The Vietnamese Women’s Union defends the legal and legitimate rights and interests of women and focuses, amongst others, on mainstreaming the development and empowerment of women and assisting women’s education and economic development. All farmers above 18 years old can become members of the Farmers’ Union, whereas all women above 18 years old have the opportunity to join the Vietnamese Women’s Union.

The SWI situation between adaptation and non-adaptation groups is significantly different. The farmers that implement papyrus face, on average, more serious impacts from saltwater intrusion than the other farmers, while farm households that switch to new rice varieties or vegetables are typically coping with less serious SWI problems. These findings are coincident with the fact that papyrus is the only SWI adaptation method sticking to traditional agricultural production that can be employed on highly salted land due to the high tolerance of the crop to salinity. On the contrary, vegetable production can mostly be set up on plots with mild SWI levels.

4.2. Outcome Variables

To estimate the effect of the different adaptation methods on households’ livelihoods, we make use of five outcome variables based on a literature review and insights from FGDs focusing on the identification of appropriate livelihood indicators for the study area:

• Saline-land productivity is calculated as the total output produced on salted land per one hectare of the salted land. The total output is expressed in monetary terms based on market prices in the study period.

• Net farm income is calculated as the total net income from on-farm agricultural activities per one hectare of farm size. The total income is expressed in monetary terms based on the market price in the study period.

• Food security is evaluated by the number of affirmative responses to the six questions in the short form of the Household Food Security Scale (see also Appendix C):

Food security situation	Number of affirmative responses
1 = Food secure	0
2 = Food secure—at risk	1
3 = Food insecure without hunger	2–4
4 = Food insecure with moderate hunger	5–6

• The rate of children attending school is calculated as the ratio of school-aged children in the household that go to school to the total number of children in the household in that age range (<18 years).

• Life satisfaction is evaluated using 5 statements taken from the Satisfaction With Life Scale (SWLS) (see also Appendix D).

Table 1 presents the data for the five outcome variables in our study, both for adapters (five adaptation methods) and non-adapters. SR has the highest average level of land productivity thanks to the high revenues involved, followed by LF. The average saline-land productivity of VG is about four times higher than that of NR and triple that of PP. Strikingly, the average saline-land productivity of NR is lower than that of normal rice production due to the lower yields involved. The average net farm income of households who apply PP, VG, SR, and LF is higher than that of non-adapters. However, farmers who cultivate NR have a lower average net farm income compared to non-adapters with the difference being about 142 USD/ha. We also observe from Table 1 that averages of the social indicators of SR and LF methods are higher than those of the other methods as well as the non-adapters.

4.3. Probit Models

A crucial step in estimating the propensity scores is selecting the appropriate specification for the participation equation [56]. As explained above, we estimate two (sets of) models: one to investigate the choice of adapting irrespective of the adaptation method (model 1) and one in which we estimate a set of models for each adaptation method individually (model 2). In both models, we use the same set of independent variables that were (a) relevant to the farmers’ decision with regard to adapting but that were not influenced by their decision, and (b) found to be significant (at 1%, 5% or 10% level) in at least one of the probit models. Table 2 presents the different probit models that will be used to estimate the propensity score on the basis of which the matching between adapters and non-adapters is subsequently conducted. Model 1 reveals that SWI adaptation in general is influenced by all independent variables in the model except for the proportion of salted land and family labor. However, the models for the individual adaptation methods (Model 2) indicate that the selected determinants influence the odds of picking specific adaptation methods in different ways. For instance, the more serious the SWI situation is, the less (more) likely it is that farmers switch to vegetable (papyrus) production. Regarding the proportion of salted land, the higher the percentage of salted land on the farmer’s plot, the higher (lower) the probability of implementing lotus-fish (new rice varieties). Having a long-term land certificate decreases the odds of switching to shrimp or lotus-fish production—a finding that can be explained by the fact that farmers who hold short-term land certificates can more easily convert their plots or accumulate land by negotiating with farming neighbors. Having either a higher number of household members available to work on the farm or a higher dependency ratio reduces the odds of switching to new rice varieties. A higher dependency ratio also reduces the odds of switching to a lotus-fish model. Finally, we find that the older the head of the household is, the more likely the household will switch to shrimp or lotus-fish production.

4.4. Estimated Effects of Adaptation Methods

After calculating the propensity scores, we use the kernel method to match the control and treated groups. The post-matching results assessing the impact of adapting versus non-adapting are shown in Table 3. We find that adapting to SWI significantly impacts all outcome variables except for the rate of children attending school. In general, farmers who adapted to SWI have significantly higher salted-land productivity and net farm income (about 11,428 and 2238 USD/ha) and report higher levels of food security and life satisfaction. The percentage of children in the household that go to school is not found to be significantly different in both groups.

To evaluate the covariate balance, we make use of the two most common diagnostics in the matching literature: the standardized mean difference between the treatment and control groups and the ratio of variances for the two groups [47]. As formal statistical tests of these diagnostics are not valid in propensity score matching, we instead apply guidelines recommended in the literature. In line with [47,57], we require the absolute value of the standardized mean difference to be below 0.1, while a variance ratio between 0.5 and 2 is generally regarded as acceptable [57,58]. As can be seen from Table 4, we do not find any substantial imbalances in our analysis.

The ATT between each adaptation method individually and the control group of non-adapters are presented in Table 5, from which the following conclusions can be drawn. PSM reports the results from propensity score matching using Gaussian kernel matching and probit regression. The significance of the results is confirmed by bootstrapping.

First, switching to new varieties of rice does not lead to significantly higher social outcomes than sticking to traditional rice production. We do, however, find statistically significant negative impacts on both economic indicators. Switching to new-varieties-of-rice methods decreases net farm income by on average 244 USD (about a 15% decrease) and saline-land productivity by 19 USD (about a 1% decrease).

Second, we find that cultivating papyrus leads to a statistically significant increase in livelihoods due to increases in saline-land productivity (+549.67 USD/ha, or 30%), the rate of children attending school (+21%), and food security (+0.36).

Third, the households who decided to replace paddy with vegetables have statistically significantly higher saline-land productivity and net farm income than non-adapters. Saline-land productivity increases substantially as intensive production (3 harvests per year) leads the higher saline-land productivity (+5253.8 USD/ha, or 287%) and increases net farm income by 947.66 USD/ha (about a 58% increase) In addition, household who applied vegetable method report higher levels of food security (+0.39) and life satisfaction (+0.15).

Fourth, switching to either shrimp or lotus-fish production results in positive and statistically significant estimated impacts on all economic and social outcome indicators. These results can be explained by the fact that these two adaptation methods have strong and clear impacts on the livelihoods of farmers. If households apply the shrimp or lotus-fish models, the saline-land productivity significantly increases by about more than 44,848 USD/ha and by around 5601 USD/ha, respectively. Net farm income also rises by about 11,783 USD/ha and approximately 1735 USD/ha if households move to the shrimp and lotus-fish model, respectively. The rate of children attending school rises by around 11% and 10% when farmers move, respectively, to the shrimp and lotus-fish production methods, while the self-reported life satisfaction of farmers increases by 0.61 and 0.98 units when switching to the shrimp or lotus-fish models, respectively. Finally, the results also confirm that both adaptation methods substantially increase the food security of farm households.

We do need to be careful when interpreting these results, as some of the samples for the adaptation methods are small—e.g., we only have observations for 10 households cultivating vegetables, and 11 that produce lotus-fish.

To evaluate the matching process, we again make use of the standardized mean difference between the treatment and control groups, and the ratio of variances for the two groups (see Appendix G). In all but one model we find standardized mean differences below 0.21—the exception being the model for papyrus production where the dependency ratio in the household has a standardized mean difference above 0.1. No problems were found using the variance ratio as all values were found to be between 0.5 and 2.

5. Discussion

In this study, we have analyzed the effects of SWI adaptation methods on the livelihoods of farmers in the Central Coastal Region of Vietnam. Our results point to significantly heterogeneous impacts across adaptation methods. We observe that the households that switch to producing papyrus, vegetables, shrimp, and lotus-fish score better on different economic and social outcomes than their counterparts that stick to traditional rice production. Four out of the five adaptation methods lead to significantly higher levels of saline-land productivity, while the net farm income increases significantly for households that implement either the vegetable, shrimp, or lotus-fish models. Both shrimp and lotus-fish production led to significant improvements in all social outcomes (food security, rate of children attending school, and life satisfaction), while switching to papyrus production increases food security and the rate of children attending school. Switching to vegetable production has only a significant impact on the food security level of the household. We also find that switching to new rice varieties negatively affects both economic indicators in our study (saline-land productivity and net farm income), while the social indicators are not impacted by this choice. These findings can be explained by the fact that the higher costs involved in switching to production of new rice varieties cannot be compensated by higher selling prices—thus leading to lower saline-land productivity. However, rice still plays a crucial role in providing food security to Vietnamese households while it also plays an invaluable cultural role in the rural areas of Vietnam given their long tradition in rice production [59]. Therefore, it is no surprise that a majority of farmers in the study area switched to cultivating new rice varieties. The main reason for the higher positive effects on saline-land productivity of the papyrus plant, vegetable, shrimp, and lotus-fish methods is the higher revenues from these methods. Only farmers who apply the shrimp and lotus-fish method are significantly more satisfied with life than non-adapters. It is noted that saline-land productivity is the only outcome that is significantly affected by all adaptation methods.

We also find that switching to new varieties of rice negatively impacts both economic indicators in our study. The question arises whether farmers have hurried into deciding to apply the new-varieties-of-rice method. Based on the FGD, we noted that when facing SWI problems, farmers have two options: (1) sticking to traditional rice production with lower costs at the beginning, but tolerating a higher proportion of paddy plants that die, or (2) switching to new rice varieties with higher upfront costs but a lower death rate of paddy plants. The factors underlying this trade-off situation (e.g., risk aversion and financial constraints) explain why some farm households continue to produce traditional rice in plots that are seriously affected by saltwater intrusion, while others switch to new varieties of rice in plots with moderate SWI levels.

The results of the study coincide with the outcomes of the analytical hierarchy process study [19] in which farmers were asked to prioritize the adaptation methods pushed by local authorities over the next five years. Unsurprisingly, shrimp and lotus-fish production ranked highest on the list of potential adaptation methods. We focus in this study on the current impacts of the different adaptation methods, ignoring some of the potential risks involved in the implementation. The FGDs that were organized over the course of our research project allow us to elaborate on some of these risks. Shrimp production requires constant monitoring of growing conditions and a rapid response in case of an eventual outbreak of diseases in the pond in order to yield the rewards of the high revenues involved in this adaptation method. The lotus-fish production is a relatively new adaptation method that involves low risks related to production and market conditions. Fish are less likely to catch diseases than shrimp and the demand for lotus products is high.

Papyrus and vegetable production also have positive impacts on farmers’ livelihood but the fact shows that local farmers do not favor these two methods. For the production of papyrus, the outlook might be less positive as the consumption market is changing due to the changing customers’ preferences. Papyrus is, at the moment, mainly sold as an input to the mat production industry, yet it is increasingly being replaced by other products (plastics) that are cheaper. Next, the production of vegetables involves a higher risk than the other adaptation methods as the selling price of vegetables depends heavily on the wholesalers involved in the value chain. In addition, due to the high risks involved in vegetable production (damages to crops) and the high labor-intensity involved, the popularity of vegetable production as an adaptation method is decreasing. If policymakers want to promote a diverse set of adaptation methods, they need to install policies such as providing insurance and avoiding interfering with market mechanisms in order to reduce the risks mentioned above [9]. These policies would motivate farmers to implement a wider range of adaptation methods.

Successfully implementing SWI adaptation methods does not only depend on the efforts of individual farmers as local authorities also play a crucial role [9]. The FGDs that we organized indicated three important roles for local authorities. First, financial support is identified as the most important factor in this regard as almost all farmers in the study area have to borrow money from credit services. However, according to our survey, only around 19% of the farmers in the study area have made use of the services of the Vietnam Bank of Social Policies—a state bank that offers low interest rates on loans to farmers. If local authorities would increase the lending capacity of this bank, more farmers could apply SWI adaptation methods. Next, technical support also plays an important role in the local adaptation processes. The production of shrimp and lotus-fish requires technical skills to select the best seeds and to take care of and feed the animals. Finally, better forecasts of future SWI developments would also help farmers to choose the best available adaptation method with respect to their situation. Until now, only in the Mekong River Delta have farmers been trained to install applications on smartphones that automatically monitor water sources as well as receive information about water levels in the river, pH level, water temperature, and salinity. This new technique will help farmers in our study area proactively and effectively respond to salinity intrusion.

Farmers are aware of the benefits of other adaptation methods, especially for food security but they still hesitate to switch due to the following: (1) Farmers have a long tradition of rice production, therefore, it’s not easy to change their mind-set. (2) There are barriers related to finance, techniques, and information that prevent farmers from adapting. (3) Rice is still the main food in Vietnamese families. Farmers keep producing rice to at least maintain food security if farmers can’t switch to other crops, shrimp farming, or lotus-fish farming (4). For policy, it matters that, even if rice production is the traditional food, climate change and SWI render rice cultivation more complicated in the longer run. Policymakers wanting to face this problem should not shy away from looking at other crops.

6. Conclusions

We find that, except for switching to new rice varieties, all SWI adaptation methods that farmers in the study area are currently applying have positive impacts on their livelihoods. For most adaptation methods, we find positive impacts on both economic and social indicators. The implementation of the currently available adaptation methods can be further enhanced with risk reduction mechanisms and suitable financial and technical support from local authorities. However, as the problems related to saltwater intrusion are becoming ever more serious—mainly driven by more frequent and more severe periods of droughts—and consumption markets for certain agricultural products tend to fluctuate, farmers have to carefully decide which adaptation method to implement in the future. Shrimp and lotus-fish production are worth considering in this regard thanks to their positive effects on farmers’ livelihoods. Due to SWI, food security is undermined gradually, so actions are needed from both farmers and policymakers. The article shows that these changes are also beneficial for productivity. A policymaker could inform farmers, and change their mindset. To the extent that policymakers have control over the number of plots of rice production, they can foster these changes by allocating more plots to other crops.

One of the limitations of our study is that we rely on limited sample sizes for a number of adaptation methods to explore the impact of each of these methods individually. This might explain our failure to calculate propensity scores when adding more variables to the probit model—e.g., location, off-farm income, and membership in different organizations. Next, we only focus on evaluating the impacts of adaptation methods on the economic and social dimensions, as assessing ecological impacts requires abundant technical data and expertise that is not found at the household level to report, nor within the capacity of the interviewers to assess. As a result, our results do not present a full sustainability assessment of SWI adaptation methods. Future research can try and overcome this and also take environmental impacts into account. Yet, given that the environmental impacts mostly relate to externalities and cannot be uniquely studied at the farm level, such an analysis would require using a different evaluation framework.

Overall, our study confirms that farm-level adaptation improves the livelihoods of farmers through increased levels of income and improved social conditions. However, the fact that a majority of farmers in the study area are not adapting to SWI could imply that farmers face certain constraints when choosing SWI adaptation methods—e.g., financial, technical, or information barriers. Therefore, further research can explore which determinants impact the farmers’ decision-making to expand one’s knowledge of SWI adaptation in the study area. In addition, the sample size of some of the adaptation methods under consideration is low, especially for VG and LF, so future research is recommended to confirm our results.

Author Contributions

T.D.L.N., B.B. and S.S. designed and drafted outline. T.D.L.N. collected data and drafted the manuscript. B.B. and B.D. revised manuscripts. T.D.L.N., B.D., S.S. and B.B. discussed the results. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Informed consent was obtained from all subjects involved in the study.

Data Availability Statement

All relevant study data is contained within the article.

Acknowledgments

Special thanks to the local authorities of two districts Duyen Xuyen and Quang Dien.

Conflicts of Interest

The authors declare no conflicts of interest.

Footnotes

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Tables

Table 1

Outcome indicators.

Variables(Unit)	Total Sample	Non-Adapter	Adapter
			NR	PP	VG	SR	LF
Saline-land productivity (USD/ha)	7187.14(1473.68)	1829.52(24.27)	1764.28(162.23)	2348.57(25.48)	7126.19(356.3)	47,236.19(535.56)	7910.95(652.2)
Net farm income (USD/ha)	2710.32(326.05)	1633.87(94.12)	1491.58(181.44)	2010.88(26.83)	2266.19(455.25)	10,665.71(552.68)	3951.55(434.26)
Rate of children attending school (%)	54.58(1.84)	48.01(2.43)	47.95(4.49)	53.44(7.29)	53.3(11.92)	90.21(5.64)	100(11.14)
Food secure(1 = Food secureto 4 = Food insecure with moderate hunger)	2.74(0.05)	2.95(0.06)	3.04(0.11)	2.68(0.19)	2.6(0.32)	1.45(0.15)	1.72(0.31)
Life satisfaction (1 = Totally dissatisfied to5 = totally satisfied)	2.75(0.03)	2.67(0.43)	2.71(0.76)	2.48(0.12)	2.8(0.21)	3.2(0.1)	3.45(0.2)
N° of observations	414	229	89	29	10	46	11

(Source: Authors’ calculations, 2022). Note: Robust standard errors in parentheses.

Table 2

Probit model results explaining the estimated effects on the probability of choosing (specific) SWI adaptation methods.

	Model 1	Model 2
Variables	Adapter	NR	PP	VG	SR	LF
Family labor (no.)	−0.14(0.11)	−0.31 ***(0.08)	0.01(0.14)	−0.11 *(0.18)	0.63 ***(0.17)	0.19(0.20)
Dependents in the household (%)	−0.008 **(0.004)	−0.008 ***(0.003)	−0.005(0.004)	−0.01(0.007)	0.01 **(0.006)	−0.008(0.008)
Age of HH head	0.08 ***(0.02)	0.01(0.01)	0.03(0.02)	0.01(0.03)	0.07 **(0.02)	0.06 *(0.03)
SWI situation	−0.12 **(0.03)	−0.12(0.09)	0.41 **(0.15)	−0.46 **(0.22)	−0.13(0.16)	0.09(0.23)
Proportion of salted land	−0.006(0.006)	−0.02 ***(0.005)	0.02 ***(0.007)	0.007(0.01)	0.04 ***(0.009)	0.06 ***(0.02)
Land certificate	−1.05 ***(0.25)	0.09(0.21)	0.19(0.30)	0.03(0.4)	−2.54 ***(0.38)	−1.17 ***(0.43)
Number of observations	414	318	258	239	275	240
Wald Chi 2	27.37 ***	47.63 ***	16.47 ***	8.88 *	119.64 ***	24.35 ***
Pseudo R2	0.14	0.12	0.09	0.10	0.48	0.27
Log Likelihood	−270.12	−164.68	−82.45	−37.08	−64.35	−32.47

(Source: Authors’ calculations, 2022). Note: Robust standard errors in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

Table 3

Estimated effects of SWI adaptation (general).

	Saline-Land Productivity (USD/ha)	Net Farm Income(USD/ha)	Rate of Children Attending School (%)	Food Security(1–4)	Life Satisfaction(1–5)
Number of treated (Adapters)	185	185	185	185	185
Number of control (Non-Adapters)	229	229	229	229	229
ATT	11428.57 ***(303.76)	2238.09 ***(480.01)	16.20	−0.49 ***	0.18 **
Wilcoxon signed rank (WSR) p-value	0.00	0.00	0.41	0.06	0.01

(Source: Authors’ calculations, 2022). robust standard errors in parentheses; ** p < 0.05, *** p < 0.01.

Table 4

Covariate balance with and without matching.

Variable	Standardized Difference		Variance Ratio
	Raw	Matched	Raw	Matched
Family labor (no.)	−0.057	−0.039	0.719	0.789
Dependents in the household (%)	−0.126	0.011	0.656	0.674
Age of HH head	0.179	0.020	0.762	0.908
SWI situation	−0.106	0.009	0.923	0.932
Proportion of salted land	−0.122	0.071	1.677	1.578
Land certificate	−0.250	0.019	1.172	0.995

(Source: Authors’ calculation, 2022).

Table 5

Estimated effects of adaptation methods.

Outcome Variables		NR	PP	VG	SR	LF
Number of treated (Adapters)		89	29	10	46	11
Number of control (Non-Adapters)		229	229	229	229	229
Economic indicator	Saline-land productivity	−19.36 **(14.81)	549.67 **(62.97)	5253.8 ***(962.71)	44848.09 ***(6993.49)	4601.9 ***(846.91)
	Net farm income	−244.12 *(180.20)	−11.26(322.24)	947.66 *(1542.97)	11783.8 ***(1297.16)	1735.87 *(763.17)
Social indicator	Rate of children attending school	3.40(5.57)	21.77 **(7.38)	3.92(11.18)	11.54 *(9.54)	10.54 *(0.04)
	Food security	−0.006(0.12)	−0.36 **(0.21)	−0.39 **(0.27)	−1.34 **(0.29)	−0.94 *** (0.43)
	Life satisfaction	0.14(0.08)	−0.13(0.11)	0.15 *(0.14)	0.61 **(0.19)	0.98 ***(0.24)

(Source: Authors’ calculation, 2022). robust standard errors in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

Appendix A

Table A1

Result of FGD meeting 1: Current status of rice production and impacts of SWI.

No	Current Status of Rice Production	Impact of SWI
1	Rice plays an important role	Rice production output decline
2	The rice planted area has been fluctuating by the conversion of agricultural land into industrial or urban land and by farmers moving away from rice production to other crops or even to aquaculture	Rice production surface decline
3	Rice quantities have slightly decreased	Fresh water was inadequate for irrigation
4	Changes in the sowing date for winter-spring rice
5	Changes in crops that are being cultivated—e.g., Moving to drought-tolerant crops
6	Rice-fish rotation systems

Appendix B

Table A2

Result of FGD meeting 2.

No	Social Indicators	Economic Indicators
1	Food security	Land productivity
2	Housing facilities	Profitability
3	Social equity in income and welfare distribution	Yield per hectare (crop productivity)
4	Farmers knowledge	Net farm income
5	Awareness of resource conservation	Benefit and cost ratio
6	Education level of household members.
7	Calories supply, micronutrient supply
8	Child nutritional status
9	Number of migrated household member

Appendix C

• In the last 12 months, did you or any other adults in your household ever have to cut the size of your meals or skip meals entirely because there wasn’t enough money for food?

  • □

    Yes

  • □

    No

  • □

    Don’t know

  • □

    Refused

• How often did this happen?

  • □

    Almost every month

  • □

    Sometimes, but not every month

  • □

    Only one or two months

  • □

    Don’t know

  • □

    Refused

• In the last 12 months, did you ever eat less than you felt you should because there wasn’t enough money to buy food?

  • □

    Yes

  • □

    No

  • □

    Don’t know

  • □

    Refused

• In the last 12 months, were you ever hungry but didn’t eat because you could not afford enough food?

  • □

    Yes

  • □

    No

  • □

    Don’t know

  • □

    Refused

• The food that I/we bought just didn’t last, and I/we didn’t have money to get more. Was this often, sometimes, or never true for you or the other members of your household in the last 12 months?

  • □

    Often

  • □

    Sometimes

  • □

    Never true

  • □

    Don’t know

  • □

    Refused

• I/we couldn’t afford to eat balanced meals. Was this often, sometimes, or never true for you or the other members of your household in the last 12 months?

  • □

    Often

  • □

    Sometimes

  • □

    Never true

  • □

    Don’t know

  • □

    Refused

Source: Blumberg et al. (1999) [37].

Appendix D. Questions in the Satisfaction With Life Scale (SWLS)

Please imagine a ladder with steps numbered from 0 at the bottom to 10 at the top. The top of the ladder represents the best possible life for you and the bottom of the ladder represents the worst possible life for you.

B1. On which step of the ladder would you say you personally feel you stand at this time? [0–10].

The following question asks how happy you feel, on a scale from 0 to 10. Zero means you feel “not at all happy” and 10 means “completely happy”.

B2. Taking all things together, how happy would you say you are? [0–10]

The following questions ask how satisfied you feel, on a scale from 0 to 10. Zero means you feel “not at all satisfied” and 10 means “completely satisfied”.

B3. Overall, how satisfied with your life were you 5 years ago? [0–10]

B4. As your best guess, overall, how satisfied with your life do you expect to feel in 5 years’ time? [0–10]

Below are five statements with which you may agree or disagree. Using the 1–7 scale below, indicate your agreement with each item. Please be open and honest in your responding. The 7-point scale is as follows:

• Strongly disagree.

• Disagree.

• Slightly agree.

• Neither agree nor disagree.

• Slightly agree.

• Agree.

• Strongly agree.

B5. In most ways, my life is close to my ideal. [1,2,3,4,5,6,7]

B6. The conditions of my life are excellent. [1,2,3,4,5,6,7]

B7. I am satisfied with my life. [1,2,3,4,5,6,7]

B8. So far, I have gotten the important things I want in life. [1,2,3,4,5,6,7]

B9. If I could live my life over, I would change almost nothing. [1,2,3,4,5,6,7]

Source: Diener et al. (1985) [40].

Appendix E

Table A3

Frequency table of SWI adaptation methods in the study area.

Method	Frequency	Description
Non-adaptation	229	Farmers do not apply any adaptation methods and keep producing traditional rice varieties.
NR	89	Farmers continue to produce rice but switch to new varieties that are more salt-tolerant. The local cooperatives sell new rice varieties to farmers at a price below that of the market (Agricultural materials are subsidized by the local governments). The main disadvantage of this method is that the yield is about 25% lower than that of traditional rice (Estimates from key informant interviews).
PP	29	Farmers convert their paddy plots into papyrus fields. This crop is neither labor, nor capital, nor time-intensive. The main limitation of this adaptation method is that the consumption market tends to be unstable: papyrus is the main input of the mat industry, which faces a narrowing demand as consumer preferences change.
SR	46	Farmers convert their land to build ponds for shrimp cultivation. In the process, the land is sometimes combined with that of the farmer’s neighbors. This adaptation method requires high initial investments and technical support from the local agricultural extension office. Shrimp production can bring about higher profits but farmers face many risks ranging from diseases to fluctuating selling prices.
VG	10	Farmers convert their paddy plots into vegetable fields. Income from vegetable production of farmers typically increase, yet profits may not as this method is labor-intensive (planting, caring, and harvesting). Moreover, cultivating vegetables requires substantial amounts of irrigation and farmers may face water shortages in the dry season.
LF	11	Farmers convert their paddy fields into fish ponds in which lotus plants are grown and fish are kept at the same time. Farmers get revenues from both fish and lotus products. This method requires lower initial investments than shrimp cultivation, and also the risks involved are lower.
Total	414

NR: New rice varieties, PP: Papyrus, VG: Vegetables, SR: Shrimp, LF: Lotus-Fish. Source: Authors’ analysis, 2022.

Appendix F

Table A4

Descriptive statistics of the farm households—full sample (414 observations) and split-up between adapters and non-adapters.

Variables(Unit)	Total Sample	Non-Adapter	Adapters
			NR	PP	VG	SR	LF
Household characteristics
Family size(No.)	6.27(0.06)	6.33(0.89)	5.80 ***(0.16)	6.13(0.26)	6.0(0.43)	7.08 ***(0.20)	5.81(0.41)
Dependents in the household (%)	41.33(1.32)	42.86(1.93)	40.07(3.52)	37.72(5.69)	28.3(9.3)	44.97(4.52)	25.54 **(8.9)
Family labor(no.)	4.54(0.05)	4.56(0.76)	4.25 **(0.14)	4.51(0.22)	4.7(0.36)	4.95 **(0.17)	4.63(0.35)
Age of household head (years)	50.73(0.28)	50.27(0.39)	51.69(0.76)	52.82 **(1.17)	50.6(1.93)	50(0.91)	50(1.82)
Education level of household head(1 = secondary or above, 0 otherwise)	0.4	0.4	0.29 **	0.79 ***	0.6	0.21 **	0.3
Farm characteristics
Proportion of salted land (%)	78.03(0.82)	78.42(0.98)	67.52(1.96)	80.98(2.96)	82.89(4.82)	90.18 *(2.34)	92.2 ***(0.14)
Land certificate(1 = long-term owner, 0 otherwise)	0.57	0.62	0.67	0.79 *	0.6	0.72 **	0.27 **
SWI situation(1 = Mild2 = Moderate3 = High)	2.01(0.04)	2.05(0.5)	1.87 *(0.10)	2.51 **(0.15)	1.5 **(0.25)	1.86(0.12)	2.09(0.24)
Social capital
Member of Vietnamese Women’s Union(1 = member, 0 otherwise)	0.88	0.87	0.91	0.75	0.9	0.93	1
Member of Farmer’s Union(1 = member, 0 otherwise)	0.89	0.89	0.87	0.86	0.9	0.93	0.91
N° of Observations	414	229	89	29	10	46	11

(Source: Authors’ calculations, 2022). Note: comparisons are made between adapters and non-adapters for five categories of method using one-sided t-tests (* p < 0.1, ** p < 0.05, *** p < 0.01).

Appendix G. Covariate Balance Summaries with and without Matching for the Different Models of Individual Adaptation Methods

Table A5

Covariate balance with and without matching (new rice varieties model).

Variable	Standardized Difference		Variance Ratio
	Raw	Matched	Raw	Matched
Family labor (no.)	−0.054	0.031	1.028	1.010
Dependents in the household (%)	−0.098	−0.072	0.878	0.966
Age of HH head	−0.076	−0.018	1.202	1.112
SWI situation	0.084	−0.069	1.186	1.106
Proportion of salted land	0.180	−0.032	1.412	1.160
Land certificate	−0.287	−0.029	1.046	1.008

Table A6

Covariate balance with and without matching (papyrus model).

Variable	Standardized Difference		Variance Ratio
	Raw	Matched	Raw	Matched
Family labor (no.)	0.313	0.002	1.119	1.014
Dependents in the household (%)	−0.322	−0.148	0.776	0.951
Age of HH head	−0.283	−0.088	0.756	0.989
SWI situation	−0.358	0.025	0.859	0.904
Proportion of salted land	0.246	−0.002	0.666	0.684
Land certificate	−0.245	−0.043	1.045	1.024

Table A7

Covariate balance with and without matching (shrimp model).

Variable	Standardized Difference		Variance Ratio
	Raw	Matched	Raw	Matched
Family labor (no.)	−0.179	−0.004	1.163	1.120
Dependents in the household (%)	0.081	−0.051	0.772	0.951
Age of HH head	−0.153	0.023	1.206	1.121
SWI situation	−0.139	−0.012	1.134	1.124
Proportion of salted land	−0.087	0.031	1.058	0.992
Land certificate	−0.116	0.033	1.074	0.952

Table A8

Covariate balance with and without matching (lotus-fish model).

Variable	Standardized Difference		Variance Ratio
	Raw	Matched	Raw	Matched
Family labor (no.)	0.055	0.311	1.305	1.223
Dependents in the household (%)	0.066	−0.077	1.468	1.279
Age of HH head	−0.196	−0.068	1.207	0.957
SWI situation	0.119	0.049	1.087	1.049
Proportion of salted land	0.141	−0.069	0.604	0.654
Land certificate	0.193	−0.044	0.949	1.008

Table A9

Covariate balance with and without matching (vegetable model).

Variable	Standardized Difference		Variance Ratio
	Raw	Matched	Raw	Matched
Family labor (no.)	0.060	0.051	1.292	1.176
Dependents in the household (%)	0.082	−0.059	1.418	1.261
Age of HH head	−0.184	−0.057	1.264	0.881
SWI situation	0.044	−0.018	1.158	1.114
Proportion of salted land	0.078	−0.058	0.611	0.768
Land certificate	0.265	−0.039	0.933	1.012

References

1. Shaw, R. Community-based climate change adaptation in Vietnam: Inter-linkages of environment, disaster, and human security. Multiple Dimension of Global Environmental Changes; TERI Publication: New Delhi, India, 2006; pp. 521-547.

2. McElwee, P. The Social Dimensions of Adaptation to Climate Change in Vietnam. 2010; Available online: https://documents1.worldbank.org/curated/en/955101468326176513/pdf/589030NWP0EACC10Box353823B01public1.pdf (accessed on 1 June 2023).

3. Yu, B.; Zhu, T.; Breisinger, C.; Hai, N.M. Impacts of Climate Change on Agriculture and Policy Options for Adaptation The Case of Vietnam. 2010; Available online: https://core.ac.uk/download/pdf/6248317.pdf (accessed on 2 September 2021).

4. Nguyen, T.T.X.; Bonetti, J.; Rogers, K.; Woodroffe, C.D. Indicator-based assessment of climate-change impacts on coasts: A review of concepts, methodological approaches and vulnerability indices. Ocean Coast. Manag.; 2016; 123, pp. 18-43. [DOI: https://dx.doi.org/10.1016/j.ocecoaman.2015.11.022]

5. Irish Aid. Vietnam Climate Action Report For 2016; Irish Aid: Hanoi, Vietnam, 2017.

6. Thuc, T.; Neefjes, K.; Huong, T.T.T.; Van Thang, N.; Nhuan, M.T.T.; Quang, L.; Thanh, L.D.; Huong, H.T.L.; Son, V.T.; Thuan, N.T.H. et al. Viet Nam Special Report on Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaption; United Nations Development Program: Hanoi, Vietnam, 2015.

7. Renaud, F.G.; Le, T.T.H.; Lindener, C.; Guong, V.T.; Sebesvari, Z. Resilience and shifts in agro-ecosystems facing increasing sea-level rise and salinity intrusion in Ben Tre Province, Mekong Delta. Clim. Chang.; 2015; 133, pp. 69-84. [DOI: https://dx.doi.org/10.1007/s10584-014-1113-4]

8. Vu, D.T.; Yamada, T.; Ishidaira, H. Assessing the impact of sea level rise due to climate change on seawater intrusion in Mekong Delta, Vietnam. Water Sci. Technol.; 2018; 77, pp. 1632-1639. [DOI: https://dx.doi.org/10.2166/wst.2018.038]

9. Dam, T.H.T.; Tur-Cardona, J.; Speelman, S.; Amjath-Babu, T.S.; Sam, A.S.; Zander, P. Incremental and transformative adaptation preferences of rice farmers against increasing soil salinity—Evidence from choice experiments in north central Vietnam. Agric. Syst.; 2021; 190, 103090. [DOI: https://dx.doi.org/10.1016/j.agsy.2021.103090]

10. Thoai, T.Q.; Rañola, R.F.; Camacho, L.D.; Simelton, E. Determinants of farmers’ adaptation to climate change in agricultural production in the central region of Vietnam. Land Use Policy; 2018; 70, pp. 224-231. [DOI: https://dx.doi.org/10.1016/j.landusepol.2017.10.023]

11. MARD. Rice Production Evaluation for 2010 and Work Plan for 2011 for Southern Vietnam; Ministry of Agriculture and Rural Development: Hanoi, Vietnam, 2011.

12. GFDRR. Vulnerability, Risk Reduction and Adaption to Climate Change Viet Nam, Climate Change and Adaption Country Profile; The World Bank Group: Hanoi, Vietnam, 2011; Available online: https://www.uncclearn.org/wp-content/uploads/library/wb118.pdf (accessed on 2 March 2023).

13. Narloch, U.; Bangalore, M. Themultifaceted relationship between environmental risks and poverty: New insights from Vietnam. Environ. Dev. Econ.; 2018; 23, pp. 298-327. [DOI: https://dx.doi.org/10.1017/S1355770X18000128]

14. MONRE. Vietnam’s Third National Communication to the United Nations Framework Convention on Climate Change; Vietnam Publishing House of Natural Resources, Environment and Cartography: Hanoi, Vietnam, 2019; Available online: https://unfccc.int/sites/default/files/resource/Viet%20Nam%20-%20NC3%20resubmission%2020%2004%202019_0.pdf (accessed on 8 October 2022).

15. Hassan, R.; Nhemachna, C. Determinants of African farmers’ strategies for adapting to climate change: Multinomial choice analysis. Afr. J. Agric. Resour. Econ.; 2008; 2, pp. 83-104.

16. Deressa, T.T.; Hassan, R.M.; Ringler, C.; Alemu, T.; Yesuf, M. Determinants of farmers’ choice of adaptation methods to climate change in the Nile Basin of Ethiopia. Glob. Environ. Chang.; 2009; 19, pp. 248-255. [DOI: https://dx.doi.org/10.1016/j.gloenvcha.2009.01.002]

17. Christen, O.; O’Halloranetholtz, Z. Indicators for a Sustainable Development in Agriculture; FIL Gesellschaft zur Förderung des Integrierten Landbaus mbH: Bonn, Germany, 2016.

18. Zhen, L.; Routray, J.K. Operational Indicators for Measuring Agricultural Sustainability in Developing Countries. Environ. Manag.; 2003; 32, pp. 34-46. [DOI: https://dx.doi.org/10.1007/s00267-003-2881-1]

19. Nguyen, T.D.L.; Bleys, B. Applying analytic hierarchy process to adaptation to saltwater intrusion in vietnam. Sustainability; 2021; 13, 2311. [DOI: https://dx.doi.org/10.3390/su13042311]

20. Gose, A.K. Farm size and land productivity in Indian agriculture: A reappraisal. J. Dev. Stud.; 1979; 16, pp. 27-49. [DOI: https://dx.doi.org/10.1080/00220387908421742]

21. Chandre Gowda, M.J.; Jayaramaiah, K.M. Comparative evaluation of rice production systems for their sustainability. Agric. Ecosyst. Environ.; 1998; 69, pp. 1-9. [DOI: https://dx.doi.org/10.1016/S0167-8809(98)00089-9]

22. Holden, S.T.; Deininger, K.; Ghebru, H. Impacts of low-cost land certification on investment and productivity. Am. J. Agric. Econ.; 2009; 91, pp. 359-373. [DOI: https://dx.doi.org/10.1111/j.1467-8276.2008.01241.x]

23. Nijkamp, P.; Vreeker, R. Sustainability assessment of development scenarios: Methodology and application to Thailand. Ecol. Econ.; 2000; 33, pp. 7-27. [DOI: https://dx.doi.org/10.1016/S0921-8009(99)00135-4]

24. Praneetvatakul, S.; Janekarnkij, P.; Potchanasin, C.; Prayoonwong, K. Assessing the sustainability of agriculture: A case of Mae Chaem Catchment, northern Thailand. Environ. Int.; 2001; 27, pp. 103-109. [DOI: https://dx.doi.org/10.1016/S0160-4120(01)00068-X]

25. Tittonell, P.; Muriuki, A.; Shepherd, K.D.; Mugendi, D.; Kaizzi, K.C.; Okeyo, J.; Verchot, L.; Coe, R.; Vanlauwe, B. The diversity of rural livelihoods and their influence on soil fertility in agricultural systems of East Africa—A typology of smallholder farms. Agric. Syst.; 2010; 103, pp. 83-97. [DOI: https://dx.doi.org/10.1016/j.agsy.2009.10.001]

26. Molua, E.L. Climate variability, vulnerability and effectiveness of farm-level adaptation options: The challenges and implications for food security in Southwestern Cameroon. Environ. Dev. Econ.; 2002; 7, pp. 529-545. [DOI: https://dx.doi.org/10.1017/S1355770X02000311]

27. Yunlong, C.; Smit, B. Sustainability in agriculture: A general review. Agric. Ecosyst. Environ.; 1994; 49, pp. 299-307. [DOI: https://dx.doi.org/10.1016/0167-8809(94)90059-0]

28. Hayati, D.; Ranjbar, Z.; Karami, E. Biodiversity, Biofuels, Agroforestry and Conservation Agriculture; Springer: Berlin, Germany, 2011; 390. [DOI: https://dx.doi.org/10.1007/978-90-481-9513-8]

29. Van Cauwenbergh, N.; Biala, K.; Bielders, C.; Brouckaert, V.; Franchois, L.; Garcia Cidad, V.; Hermy, M.; Mathijs, E.; Muys, B.; Reijnders, J. et al. SAFE-A hierarchical framework for assessing the sustainability of agricultural systems. Agric. Ecosyst. Environ.; 2007; 120, pp. 229-242. [DOI: https://dx.doi.org/10.1016/j.agee.2006.09.006]

30. Rigby, D.; Woodhouse, P.; Young, T.; Burton, M. Constructing a farm level indicator of sustainable agricultural practice. Ecol. Econ.; 2001; 39, pp. 463-478. [DOI: https://dx.doi.org/10.1016/S0921-8009(01)00245-2]

31. Rasul, G.; Thapa, G.B. Sustainability of ecological and conventional agricultural systems in Bangladesh: An assessment based on environmental, economic and social perspectives. Agric. Syst.; 2004; 79, pp. 327-351. [DOI: https://dx.doi.org/10.1016/S0308-521X(03)00090-8]

32. Herzog, F.; Gotsch, N.; Gotsch, N. Assessing the sustainability of smallholder tree crop production in the tropics: A methodological outline. J. Sustain. Agric.; 1998; 11, pp. 13-37. [DOI: https://dx.doi.org/10.1300/J064v11n04_04]

33. Dumanski, J.; Terry, E.; Byerlee, D.; Pieri, C. Performance Indicators for Sustainable Agriculture; Rural Development Sector: Washington, DC, USA, 1998; pp. 2-18.

34. Helliwell, J.; Layard, R.; Sachs, J. World Happiness; The Earth Institute, Columbia University: New York, NY, USA, 2012; pp. 1-171. Available online: http://eprints.lse.ac.uk/47487/ (accessed on 6 February 2021).

35. Moro, M.; Brereton, F.; Ferreira, S.; Clinch, J.P. Ranking quality of life using subjective well-being data. Ecol. Econ.; 2008; 65, pp. 448-460. [DOI: https://dx.doi.org/10.1016/j.ecolecon.2008.01.003]

36. Flores, C.C.; Sarandón, S.J. Limitations of Neoclassical Economics for Evaluating Sustainability of Agricultural Systems: Comparing Organic and Conventional Systems. J. Sustain. Agric.; 2004; 24, pp. 77-91. [DOI: https://dx.doi.org/10.1300/J064v24n02_08]

37. Blumberg, S.J.; Bialostosky, K.; Hamilton, W.L.; Briefel, R.R. The effectiveness of a short form of the household food security scale. Am. J. Public Health; 1999; 89, pp. 1231-1234. [DOI: https://dx.doi.org/10.2105/AJPH.89.8.1231] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/10432912]

38. Furness, B.W.; Simon, P.A.; Wold, C.M.; Asarian-Anderson, J. Prevalence and predictors of food insecurity among low-income households in Los Angeles County. Public Health Nutr.; 2004; 7, pp. 791-794. [DOI: https://dx.doi.org/10.1079/PHN2004608]

39. Gulliford, M.C.; Mahabir, D.; Rocke, B. Reliability and validity of a short form household food security scale in a Caribbean community. BMC Public Heath; 2004; 4, 22. [DOI: https://dx.doi.org/10.1186/1471-2458-4-22]

40. Diener, E.; Emmons, R.A.; Larsen, R.J.; Griffin, S. The Satisfaction with Life Scale. J. Personal. Assess.; 1985; 49, pp. 71-75. [DOI: https://dx.doi.org/10.1207/s15327752jpa4901_13] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/16367493]

41. OCED. OECD Guidelines on Measuring Subjective Well-Being; OECD Publishing: Paris, France, 2013; [DOI: https://dx.doi.org/10.1787/9789264191655-en]

42. DARD. Annually Report; Department of Agriculture and Rural Development: Quang Nam, Vietnam, 2019.

43. Abid, M.; Schneider, U.A.; Scheffran, J. Adaptation to climate change and its impacts on food productivity and crop income: Perspectives of farmers in rural Pakistan. J. Rural Stud.; 2016; 47, pp. 254-266. [DOI: https://dx.doi.org/10.1016/j.jrurstud.2016.08.005]

44. Heckman, J.J.; Ichimura, H.; Todd, P.E. Matching as An Econometric Evaluation Estimator: Evidence from Evaluating a Job Training Programme. Rev. Econ. Stud.; 1997; 64, pp. 605-654. [DOI: https://dx.doi.org/10.2307/2971733]

45. Dehejia, R.H.; Wahba, S. Propensity Score-Matching Methods for Nonexperimental Causal Studies. Rev. Econ. Stat.; 2002; 84, pp. 151-161. [DOI: https://dx.doi.org/10.1162/003465302317331982]

46. Bryson, A.; Dorsett, R.; Purdon, S. The Use of Propensity Score Matching in the Evaluation of Labour Market Policies; Working Paper No. 4 Policy Studies Institute and National Centre for Social Research: London, UK, 2002.

47. Austin, P.C. An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivar. Behav. Res.; 2011; 46, pp. 399-424. [DOI: https://dx.doi.org/10.1080/00273171.2011.568786]

48. Weldegebriel, Z.B.; Prowse, M. Climate-change adaptation in ethiopia: To what extent does social protection influence livelihood diversification?. Dev. Policy Rev.; 2013; 31, pp. o35-o56. [DOI: https://dx.doi.org/10.1111/dpr.12038]

49. Imbens, G. Nonparametric Estimatation of Average Treatment Effects under Exogenneity: A review. Rev. Econ. Stat.; 2004; 86, pp. 4-29. [DOI: https://dx.doi.org/10.1162/003465304323023651]

50. Rosenbaum, P.R.; Rubin, D.B. The central role of the propensity score in observational studies for causal effects. Biometrika; 1983; 70, pp. 41-55. [DOI: https://dx.doi.org/10.1093/biomet/70.1.41]

51. Khandker, S.R.; Koolwal, G.B.; Samad, H.A. Handbook on Impact Evaluation: Quantitative Methods and Practices; The World Bank: Washington, DC, USA, 2010.

52. Caliendo, M.; Kopeinig, S. Some practical guidance for the implementation of propensity score matching. J. Econ. Surv.; 2008; 22, pp. 31-72. [DOI: https://dx.doi.org/10.1111/j.1467-6419.2007.00527.x]

53. Lee, W.S. Propensity score matching and variations on the balancing test. Empir. Econ.; 2013; 44, pp. 47-80. [DOI: https://dx.doi.org/10.1007/s00181-011-0481-0]

54. Cohen, J. Statistical Power Analysis for the Behavioral Sciences; Lawrence Erlbaum Associates Publishers: New York, NY, USA, 1998.

55. Leuven, E.; Sianesi, B. PSMATCH2: Stata Module to Perform Full Mahalanobis and Propensity Score Matching, Common Support Graphing, and Covariate Imbalance Testing. 2003; Available online: https://ideas.repec.org/c/boc/bocode/s432001.html (accessed on 11 February 2021).

56. Jalan, J.; Ravallion, M. Estimating the benefit incidence of an antipoverty program by propensity-score matching. J. Bus. Econ. Stat.; 2003; 21, pp. 19-30. [DOI: https://dx.doi.org/10.1198/073500102288618720]

57. Zhang, Z.; Kim, H.J.; Lonjon, G.; Zhu, Y. Balance diagnostics after propensity score matching. Ann. Transl. Med.; 2019; 7, 16. [DOI: https://dx.doi.org/10.21037/atm.2018.12.10]

58. Rubin, D.B. Using propensity scores to help design observational studies: Application to the tobacco litigation. Health Serv. Outcomes Res. Methodol.; 2001; 2, pp. 169-188. [DOI: https://dx.doi.org/10.1023/A:1020363010465]

59. Pedroso, R.; Tran, D.H.; Thi, M.H.N.; Van Le, A.; Ribbe, L.; Dang, K.T.; Le, K.P. Cropping systems in the Vu Gia Thu Bon river basin, Central Vietnam: On farmers’ stubborn persistence in predominantly cultivating rice. NJAS-Wagening. J. Life Sci.; 2017; 80, pp. 1-13. [DOI: https://dx.doi.org/10.1016/j.njas.2016.11.001]