River Ecosystem Resilience: Applying the Contingent Valuation Method in Vietnam

Abstract

Mine water drainage interferes with ecosystems of the discharging river, whilst river ecosystem services in return affect the livelihoods and well-being of adjacent communities. In Ha Long, Vietnam, surface coal mining will be closed after 2025 following the national and provincial strategy toward sustainable development. This opens up an opportunity to rehabilitate the associated rivers to increase resilience in the surrounding social-ecological system (SES) heavily affected by water pollution from coal mines. Increasing resilience through rehabilitation is costly and policy makers often lack precise enough information on public benefits to make good decisions. In this study, we apply the concept of SESs to analyze the interrelationship between river ecosystems and human society with its institutions and local governance surrounding the Ha Long coal mining area. Applying a contingent valuation survey with 410 households living near to the mining operations, we assess the public benefits resulting from three different but partly combined projects to increase resilience. Results show that households are significantly willing to contribute to all proposed resilience increasing projects. Factors explaining willingness to pay (WTP) are diverse and are found to be related to common institutions and local governance. Through a comparison of the results for the three projects, we find the presence of embedding effects and identify factors leading to upward biased estimates of WTP. Our study contributes to a better understanding and valuation of public benefits in SES necessary for public policy towards increasing resilience in developing countries.

1. Introduction

The role of fossil fuel coal has been of such vital importance in human life since the first industrial revolution that it has been widely used as a fundamental energy source for production and living, typically for generating electricity. Coal mining, like all energy sectors, is water-intensive. On the one hand, water usages perform various important functions in operating activities such as coal selection, as a means for machine and truck cleaning, as a medium to mitigate coal dust, for recultivation of coal dumps, and for domestic uses in the coal mining workshops. On the other hand, a huge amount of mine drainage water with high metal content or acidity from mining activities is discharged into the environment and poses a threat to surface and groundwater as well as to the surrounding ecosystems (Tiwary, 2001) [1]. In this context, the heterogeneity of water usage purposes and quality of wastewater opens the potential to collect and treat the discharged mine water by different but adequate processes, then reuse it internally and externally. In many mining areas in the world, water availability is limited due to extensive mining, either permanently or seasonally depending on the regional or local climatic conditions or the demand of users. This exemplifies that mining companies and local governments have to consider water use not only from a pure business perspective, or the financial perspective, but also from a more general societal perspective, or the economic perspective, since through extensive water use and wastewater generation their operations are closely linked to the quality of the nearby environment. Additionally, the quality of the natural environment and river ecosystem services in return affect the livelihoods and well-being of adjacent communities (Adams et al., 2020 [2]; Costanza et al., 2017 [3]). These mutual interactions between human society and ecosystems refer to the social-ecological systems (SESs) of which the social (human) part and the ecological (environment) part are equally important and mutually dependent (Berkes, 2017) [4]. In recent years, SES resilience has become key to sustainability andinvolves complex and mutual relationships between the environment, society, economy, and adaptive governance (Berkes, 2017 [4]; Li et al., 2020 [5]).

Nevertheless, available literature concerning economic concepts for the solutions of these mine water-related issues is mainly about an innovative technological application within the mining workshops (Haibin & Zhenling 2010 [6]; Dharmappa et al., 2000 [7]); yet the potential impacts on external societies are rarely considered thoroughly. Technology is commonly regarded as the essence of environmental projects, which offer technical solutions for preventing pollution. This perspective, however, enormously disregards the importance of measuring the loss from environmental pollution or the gain from environmental protection, particularly in monetary terms. For better environmental policies and generally sustainable development goals, not only financial profits but also the performance towards SES resilience and sustainability from a mining project should be measured on an equal basis (Damigos, 2006) [8].

The concept of social-ecological systems (SESs) is used as a basic unit of analysis under corresponding contexts of institutions and governance as integrated parts (Berkes, 2017 [4]). Due to the dynamic, complex, and multi-level interactions in SESs, a holistic and consistent concept or framework for the analysis of SESs is still missing despite a flourishing quantity of publications concerning SESs over time (Gain et al., 2020 [9]; Ostrom 2009 [10]; Colding and Barthel 2019 [11]; Berkes and Folke 1998 [12]). We adopt the concept of SESs in our study simply and narrowly to understand the mutual interactions of humans and the environment and how common institutions and local governance interact in the coal mining region corresponding to a specific environmental plan and river rehabilitation project. A comprehensive understanding of complex linkages in SESs of the case study is essential to plan environmental initiatives and decide on the implementation of plausible measures as well as on which stakeholders are appropriate to contribute to the financial funding of such projects. After that, we employ the contingent valuation method (CVM) to estimate the social benefits coming from improvements in river rehabilitation and ecosystem services in terms of money.

Contributing the largest ratio—nearly 50 percent in 2011—to the industrial sector of the economic structure of Quang Ninh province, the hard coal mining industry has proven to be an important sector (Vietnam Institute for Urban and Rural Planning, 2013 [13]) that brings substantial employment, tax revenues, and investment in environmental projects and regional development. Despite these benefits and being a national energy source, coal mining is a great source of environmental pollution. Located closely to the coast of Ha Long Bay, which is recognized for its emerald water and impressive limestone islands and has been a UNESCO world natural heritage site twice, coal mining companies in recent years have constantly taken serious responsibilities by exploiting this fossil fuel coal and operating mining activities in a sustainable way. From the local perspective, the main objectives of environmentally friendly management of mine water and river rehabilitation are largely improving household livelihoods of nearby communities but may also be improving tourism development from anational or provincial perspective. On that basis, institutions and governance play unseparated roles in environmental initiatives in general and the river rehabilitation project in our particular case study.

This study, therefore, does not attempt to develop a comprehensive concept of SESs, but rather to properly analyze the dynamic concept of complex SESs under the lens of economic valuation to a specific case study in practice related to mine water under the local context of informal institutions and governance in a low-income area of a developing country. Applying the CVM, this study attempts to identify the positive welfare effects to society from efficient mine water treatment and management to improve the river aquatic environment, hence a river rehabilitation, in the vicinity of the hard coal mining region in Ha Long city, Vietnam. Out of the existing literature so far, CVM has only rarely been applied to evaluate the non-market benefits of river rehabilitation from a polluted one caused by coal mining activities (Damigos, 2006 [8]). Additionally, this study is the first one considering mine water-related issues to the environment from the economic perspective in the context of Vietnam. Through this, we aim to promote more attention to the essential economic perspective regarding social welfare in any environmental project and policy planning to harmonize the economic development targets with the pursuit of steady SES resilience and long-term sustainability, especially in developing economies. Furthermore, challenges found in general literature and guidelines whilst applying a stated preference method in developing countries such as Vietnam are described in this study.

The remaining structure of the paper is organized as follows. Section 2 reviews the literature on SESs, ecosystem services, and CVM to value environmental goods. Section 3 describes the method comprising the case study and how we conducted our survey. Section 4 is devoted to analyzing the data and results of empirical experiments, which are then discussed and concluded in Section 5.

2. Literature Review

The concept of social-ecological systems (SESs) has recently emerged after Berkes and Folke (1998) [12] first introduced a concept matching the interlink or two-way feedback interaction among humans and ecosystems for mutual social-ecological resilience more than 20 years ago. The term emphasizes that the social (human) subsystem and the ecological (nature) subsystem are equally important, mutually dependent, and co-evolutionary. In addition, ecosystem services refer to the relation between human well-being and the natural environment, which is defined by the benefits humans directly or indirectly obtain from nature (De Groot et al., 2002 [14]; Millennium Ecosystem Assessment 2005 [15]; Fisher et al., 2009 [16]). Attempts to translate benefits derived as ecosystem services into financial incentives to conserve or provide them and monetary units for their measurement reflect such an economic aspect of the linkages between the environment (ecological) and humans (social) in terms of welfare (Adams et al., 2020 [2]) that would be utilized for the management of SES resilience and sustainability as well as making corresponding policies. This approach may, however, oversimplify the dynamic and complex interactions in SESs regarding the limited understanding of social diversity and power (belonging to the social subsystem) (Fabinyi et al., 2014 [17]) as well as a lack of attention to the uncertainties of ecosystem services (belonging to the ecological subsystem) (Norgaard, 2010 [18]). Despite a lack of holistic scientific understanding, learning about dynamic, complex, and multi-level SESs is of crucial importance to prevent severe environmental problems, foster human well-being, and promote SES resilience and sustainability (Gain et al., 2020 [9]; Ostrom, 2009 [10]).

Amongst various approaches, we apply the basic concept of SESs to analyze the complex interrelationship between ecosystems and humans in which institutions and governance are considered integral parts of these integrated systems (Berkes, 2017 [4]; Ostrom, 2009 [10]). In our specific case study, we define that (i) the social part includes all human activities inside the coal mining area such as economic activities (coal mining, discharging mine water, recultivation of coal dump, etc.) and outside the mining workshop, i.e., livelihoods of the surrounding community (e.g., individuals who make diverse use of the river for recreation, discharging wastewater, flood control); (ii) the ecological part comprises all river ecosystems, e.g., water, plants, and aquatic animals; (iii) institutions are the social norms, cultures, or informal rules of society that determine attitudes and behaviors of people and how they interact amongst the community as well as with the environment (Berkes, 2017 [4]; Ostrom, 2010 [19]). Likewise, broader linkages among people as well as between people and the environment are also motivated by management and policies at the level of local governance (Berkes, 2017 [4]; Duit et al., 2010 [20]). Institutions and governance, therefore, are very much different across regions and countries, especially between the developing countries and the developed counterparts where the majority of scientific studies take place, and thus on those the common guidelines are based. “One size fits all” results are hence inadequate (Ostrom, 2010 [19]).

How to assess changes in ecosystem services in economic terms recently and increasingly gains the attention of environmental researchers and policymakers (Gain et al., 2020 [9]). Since being provided free of charge by nature, ecosystem services represent non-tradable goods that are usually impossible to assign property rights to, thus are non-excludable in the market mechanism. By nonrivalry and non-excludability, ecosystem services are categorized as public goods, i.e., the availability of a public good is not reduced by using it, and preventing other individuals from using it is impossible, respectively. Before the 1980s, ecological research and environmental and resource economics research worked independently with limited collaboration. After the establishment of the new field “ecological economics”, the gap between the two disciplines was bridged whilst the academic theories were connecting with practice (Costanza et al., 2017 [3]). Due to the nature of uncertainties and being a newly established field, policymakers and stakeholders have been more likely to underweight the contribution of ecosystem services to human welfare (Costanza et al., 1997, 2017 [3,21]). This phenomenon is rather common in developing and less-developed economies due to the public budget constraint and the weak functioning government (Carlsson et al., 2015) [22].

Over the years, a variety of research has taken place in attempts to estimate the value of environmental goods, nevertheless, the majority of them were under the circumstances of developed countries. Since 1990, out of over 4000 publications, only 146 studies have their roots in developing countries (Durand-Morat, Wailes, and Nayga 2016) [23]. Possible reasons for this enormous imbalance are the lack of supporting factors such as financial funding, government policies, related infrastructure, and trained researchers for primary research. Due to differences in wealth, social and cultural milieu, as well as spiritual values, research carried out in developing countries faces more challenges and therefore should be carried out under serious caution. Before applying standard practice guidelines of the traditional valuation methods, tight coordination with the local government and local experts of relevance is essential to understand the environmental good and the research milieu (Börger et al., 2021 [24]; Carlsson et al., 2015 [22]). In recent years, scholars worldwide have continuously promoted this research theme to the less developed countries (Christie et al., 2012 [25]; Amuakwa-mensah et al., 2018 [26]).

Besides the revealed preference methods, stated preference methods such as the CVM have become widely applied to estimate economic values of non-market goods, thus having a unique role in welfare analysis, particularly in the context of environmental goods. Additionally, stated preference methods are the only ones able to measure non-use or passive-use values or changes that are independent from market choices so that no market prices exist nor are available to observe. The contingent valuation method (CVM) is a standard valuation tool in environmental economics to assess the value of changes in non-market goods such asmost environmental goods in monetary terms (Bergstrom & Loomis, 2017 [27]; Carson, 2012 [28]). CVM studies are typically conducted in the form of a representative survey in which households affected by an expected change in public goods stemming from an environmental project presented to them are asked about their willingness to pay (WTP) for getting these changes (Ostermiller et al., 2015 [29]; Loomis et al., 2000 [30]). From these survey data, households’ average WTP can be computed which is then interpreted as the welfare change due to the benefits received from the described project.

Usually, funders of public goods comprise the local government, private organizations, and society. The contribution of society to environmental goods is described in various articles employing stated preference methods in the context of developed countries (Costanza et al., 2014 [31]; de Groot et al., 2012 [32]; Loomis et al., 2000 [30]; Atkins and Burdon 2006 [33])and developing countries. In terms of coal mining-related problems, the literature is yet limited, therefore, we provide a general overview of studies regarding the acceptance of society towards environmental projects in developing countries elicited by stated preference methods. On average, households in a rural and poor area in Bangladesh are willing to pay USD 4.3 or USD 6.0 per year (depending on the calculation method) for a flood protection scheme via a surrounding embankment (Brouwer et al., 2009) [34]. People living in Beijing are willing to pay RMB 16.50 (EUR 2.1) per month for an environmental improvement in the Tarim River Basin through sustainable water and land management (Ahlheim et al., 2013) [35]. In Indonesia, the urban river rehabilitation in Jakarta attracts concerns of households within the catchment that total contribution per year is over USD 4 million for park space and about USD 6 million for forest conservation (Vollmer et al., 2016) [36]. The dominant pollution of the River Ganga in India stemming from raw sewage discharge leads to the households’ WTP at INR 8.36 (EUR 0.14) per month for an upgrade in the treatment plant capacity and quality (Birol & Das, 2010) [37]. In addition, Malaysian people are statistically willing to pay for the improvement in the Matang Mangrove Wetlands through five wetland management attributes (Othman et al., 2004) [38].

Vietnam, meanwhile, is not an exception. Do and Bennett (2009) [39] investigate the value of biodiversity protection in Tram Chim National Park—a part of the Mekong Delta wetland ecosystem, leading to the estimated benefit of USD 3.9 million which outweighs the real cost. Another article reports that the mean WTP of Mekong Delta urban households for wetland conservation in the U Minh Thuong National Park of Vietnam is VND 16,150 (USD 0.78) per month and VND 31,520 (USD 1.49) monthly when protest respondents are excluded. Accordingly, the amount of annual contribution reaches about USD 10.97 million (Khai & Yabe, 2014) [40]. Using the contingent valuation approach, the demand of Vietnamese for rhino conservation in Cat Tien National Park is captured through the mean WTP at USD 2.5 per household (Truong, 2007) [41]. Besides the findings of significant WTP for marine conservation in Nha Trang Bay, Börger et al. (2021) [24] notably suggest that a portion of respondents’ WTP is potentially directed by the sense of duty toward a public project due to social and cultural factors in low-income countries. Briefly, through empirical research about the informal payment across ten developing countries, Olken and Singhal (2011) [42] show that in the previous 12 months, 62 percent of total households contributed to public goods in terms of money and the ratio of the amount over total taxes was 16 percent in the rural areas of Vietnam.

3. Methods

3.1. The Case Study

In Vietnam, the coal mining industry started about 180 years ago. In 1994, Vietnam’s national coal and mineral industries holding corporation limited (Vinacomin) was established by the Prime Minister’s enactment with one hundred percent of the charter capital coming from the state. A majority of coal mines are found in the northern regions of the country and are exploited as either open surface mining or underground mining. Specifically, about 95% of total hard coal in Vietnam comes from Quang Ninh province. Since then, coal mining has had the most extreme impact, in comparison to other industries, on the nearby environment and surrounding communities in these regions despite being the national energy source, bringing enormous employment and generating high profits. In recent years, coal mining companies have steadily shown great responsibility by exploiting and operating coal mining activities in a sustainable way.

Following the national and provincial development strategy, Vinacomin companies were one of the first and leading followers that have progressively pursued environmentally benign production and sustainable development called “brown” to “green” transmission in economic development. On that basis, mining companies have invested in a variety of environmental projects to reduce negative impacts and protect the environment during their mining operation. That includes planting tree projects to rehabilitate mine waste dumps; installing automatic and online monitoring systems to frequently monitor the amount of dust and gas emissions as well as the quality of discharging mine water; upgrading capacity and installing new technology in the mine water treatment plants. In parallel, not only has the exploiting capacity steadily decreased over time but the gradual transformation of so-called broad to in-depth exploitation has also occurred, meaning a shift from open surface mining to an increasing share of underground mining, and leading to entirely underground mining in the future. The former mining method has been gradually shrinking and will close in 2025, whereas the latter mining method has gradually expanded and will continue to expand beyond the year 2030 (GoV 2016) [43]. Mine water from underground mining is less likely to depend on the season and is thus more stable in volume than in open-pit mining; however, it is expected to be different in terms of both quantity and quality, which requires an appropriate adjustment in treatment processes and capacity of treatment plants.

Notwithstanding the continuous enhancement in equipment and technology to collect and purify mine water, the vulnerability remains in current systems. Data from Vinacomin in 2014 implied that 74% of all mine water was treated, meaning that the other untreated part ran directly into the environment. At the same time, quality standards from the National technical regulation for industrial wastewater in Vietnam, with which mining companies have to comply, have also become stricter. The National technical regulation for industrial wastewater QCVN 40:2011/BTNMT is currently in effect but is expected to be replaced from the beginning of 2025 by QCVN 40:2021/BTNMT (being a draft at the current time). On account of the scarcity of this natural resource as well as the sharp rise in urban and touristic development in that area, Ha Long is forecast to be in a shortage of fresh water at 28,000 cubic meters per day by 2030. These developments have urged mining companies to handle water resources and wastewater in a much more environmentally friendly and efficient way. For the above reasons, the WaterMiner project, funded by the German Federal Ministry of Education and Research (BMBF), has focused on improving the efficiency and effectiveness of mine water management within Hon Gai peninsula—the eastern part of Ha Long city, through material flow analysis (Greassidis et al., 2020) [44], the technical concepts about surface run-off and mine water treatment (Ulbricht et al., 2018) [45], and the economic concepts to consider changes in non-market ecosystem services presented in this paper. Figure 1 illustrates the project area.

This study investigates the willingness to pay (WTP) of nearby residents for rehabilitating river ecosystem services of the Lo Phong river and the C2 stream in Ha Long city belonging to the northern region of Vietnam. The Lo Phong river is 7.11 km in length, merging with the C2 stream of 2.5 km in length then flowing into the sea—the Ha Long Bay. The river and the stream, thereafter called “the river” for short, are located in the Ha Phong ward with a population of 12,598 inhabitants in 3334 households until the second quarter of 2018, according to the People’s Committee of Ha Phong ward. The population growth rate is 1.26% annually. Out of 24 square kilometers of the ward, about 78 hectares are agricultural land divided into three collective farms, namely Ha Phong, Hong Hai, and Ha Tan, consisting of about 12.5 hectares, 53 hectares, and 12.5 hectares, respectively. Situated on the outskirts of the urban tourism city of Ha Long, this area receives less attention from investment funds but the river undoubtedly influences the environmental quality of the city and the bay. Besides the discharging water and surface run-off from the mines to the river, human interventions from living activities of residents also bring the awareness of environmental protection there into question.

On the one hand, the Lo Phong river and the C2 stream receive water from two large surface mines, namely Ha Tu and Tan Lap, respectively. Before, it had played a role as a transferring channel for mine water pumped from the Ha Tu mining workshop to the treatment plant only until 2017 when a pipeline system was constructed instead. Mine water discharged in the river has not always been fully collected to be treated in the treatment station. Typically in the rainy season, a huge amount of surface water run-off containing an abundance of sediments and impurities such as coal, coal sludge, coal gangue, oil, rock, and soil runs directly from mining workshops into the river.

On the other hand, human intervention poses particular concerns about the pollution of this river. Due to the direct municipal wastewater discharge without any treatment and occasional trash disposal from adjacent residential areas, the river is eventually polluted, which triggers various environmental threats, harms the riparian vegetation, and leads to health-related risks (Hendryx & Ahern, 2009) [46]. Altogether, the concerns regarding polluted river water caused by acid mine water, coal sludge, and municipal wastewater as well as water flow disturbance triggered by rock, soil, and solid trash then the risk of flood in the rainy season have become quite strong.

Over time, with increasing standards from the National technical regulation for industrial wastewater, Standard B according to regulation QCVN 40:2011/BTNMT for industrial wastewater discharging into the environment, mining companies have continuously invested more and more in upgrading their water treatment plants in terms of both technology and capacity. They have also been seeking solutions for collecting and treating surface water run-off from the mine workshop, especially in rainy seasons, before going into rivers. Solutions for surface run-off retention, sedimentation and treatment before running into the Lo Phong river are represented by our research partners in the WaterMiner project. While pursuing sustainability in mining operations, the spill-over effects on adjacent ecosystems of discharging rivers would be positive. This will allow those downstream river ecosystems to provide various better services to households living along or near to the river. Additionally, to achieve the river rehabilitation goals, close coordination between public authorities and Vinacomin is highly necessary to conduct a set of essential measures. These include building facilities to collect and treat all municipal wastewater, constructing installations along river banks to allow access to the river for recreational purposes, and redesigning these embankments so that excess water in the river bed can better infiltrate into the soil and adjacent aquifers to protect residential areas from floods. All of these measures contribute to rehabilitating river ecosystem services that benefit the livelihoods of the local communities.

3.2. Survey and Questionnaire

In this survey, the single-bounded dichotomous choice elicitation format in the contingent valuation method was adopted to value hypothetical improvements of river ecosystem services in monetary units. In recent years, the single-bounded dichotomous choice format is increasingly preferred among various CV response formats by virtue of the incentive compatibility. Other formats either violate the incentive compatibility or are only valid under very narrow and strict circumstances (Carson et al., 2014 [47]; Johnston et al., 2017 [48]; Vossler & Holladay, 2018 [49]). To date, most of the best practice guidelines for applying stated preference methods strongly advocate the use of an incentive-compatible response format through a single-binary choice question when valuing public goods, from the NOAA Panel guidelines (Arrow et al., 1993) [50] to the recent contemporary guidance by Johnston et al. (2017) [48]. Since inhabitants living in adjacent areas of the river would all benefit from the river rehabilitation project, albeit at varying degrees, representative members of households were presented with the hypothetical scenario describing the measures to be conducted leading to a certain set of benefits. These benefits were described in form of text, pictures, and oral illustrations to enable interviewees to thoroughly imagine those benefits and assess by themselves how those would change the well-being of their household. Interviewees were then told that public funds would not be sufficient to finance the proposed project and that households would need to contribute in form of a local fee collected directly and monthly for three years to secure the implementation of the measures. If not, the project would be impossible to be realized and no change or benefit would be made. Subsequently, interviewees were given a single dichotomous choice in the form of a specific price and asked if they were willing to pay this price in order to get the project implemented. From the answers of respondents, an average household WTP and then an aggregate WTP of the entire affected population can be calculated.

In this study, two different but subsequent steps of improvement are relevant due to the specific plan of Vinacomin for developing and adapting the mining activities in the Hon Gai peninsula. The elements of that plan are: (1) closing of the open-pit mines after 2025 and continuing the exploitation of the underground mines beyond 2030, (2) renovating the largest open-pit mine (Ha Tu) which discharges mine water into the Lo Phong river after its closing to become an ecological reservoir, and (3) local government building a municipal wastewater treatment plant in the Ha Phong ward until 2025. Based on this, we developed a first scenario (project 1: clean-up) with the benefits of this plan to be obtained until 2025. In addition, further plans envision a broader rehabilitation of the rivers resulting in more comprehensive environmental improvements of river ecosystems and, consequently, in more ecosystem services obtained by the population. This scenario (project 3: rehabilitate) is envisaged until 2030.

However, a direct quantitative comparison of the benefits of the two sizes of the projects in terms of WTP may be problematic for two reasons: (1) for projects stipulating an environmental improvement and, therefore, asking for respondents’ WTP, the so-called Hicksian Compensating Variation (HCV) represents the underlying theoretical welfare measure. Strictly speaking, the HCV does not fulfill the ranking condition, i.e., a direct comparison of monetary measures of two projects with two different situations is not possible if the price levels in these different situations are not equal. (2) As a stated preference method, CVM is prone to various psychological biases, i.e., WTP being affected not only by the changes in perceived benefits but by other factors unrelated to those benefits. Reason (1) can be discarded since it can reasonably be assumed that respondents do not expect the price levels of consumption goods to change in different ways between the two situations (clean-up vs. rehabilitate). We, therefore, assume that in the context of our survey the assessed WTP interpreted as HCV fulfills the ranking condition. However, reason (2) is relevant and in order to better compare the benefits of the two potential situations in monetary terms we introduce an interim scenario (project 2) that assesses the WTP for a change from situation “clean-up” to situation “rehabilitate”. If no bias as described under (2) exists, we would expect that WTP₃ = WTP₁ + WTP₂. Consequently, our CVM survey will contain three separate sub-samples of respondents as summarized in

Table 1. Description of scenarios used in contingent valuation questionnaires.

Sub-Sample	Status Quo	Future Scenario
Project 1	Existing situation	Clean-up
Project 2	Clean-up	Rehabilitate
Project 3	Existing situation	Rehabilitate

.

We collected three separate samples following the existing situation and two future scenarios.

In the first sample (project 1), the existing situation of the river was used as a status quo and the situation in 2025 was described as cleaning up the river in a future scenario. In the second sample (project 2), the status quo was the situation in 2025 when the river would be cleaned up, thus the future scenario that needs a financial contribution from inhabitants was the situation in 2030 to rehabilitate the river. In the third sample (project 3), the existing situation was status quo and the future scenario of the river rehabilitation was illustrated as the situation in 2030. Within three situations, five key attributes of the river following the described scenarios were identified and presented in detail in

Table 2. Description of the five key attributes in three respective situations.

Attribute	Existing Situation	Situation in 2025 (Clean-Up)	Situation in 2030 (Rehabilitate)
1. Municipal wastewater	Direct into the river	All is collected to the public wastewater treatment plant	All is collected to the public wastewater treatment plant
2. Purity and flows of water	Polluted river Little base flow	Clean water Baseflow	Clean water Environmental flow
3. Flood control	Interrupted flow, weak flood control	Uninterrupted flow, high flood control	Uninterrupted flow, very high flood control
4. Recreation	Rarely fishing	Fishing, relaxing	Fishing, relaxing, playing, doing sport
5. Aesthetics	No	Medium	High

. These include (1) municipal wastewater, (2) purity and flows of water, (3) flood control, (4) recreation, and (5) aesthetics. All three situations were visually illustrated and presented in the questionnaire by pictures in Figure 2.

On this basis, a pilot survey of 30 face-to-face interviews was implemented to investigate the value of these environmental changesfrom households living along and near to the Lo Phong river and C2 stream in November 2017. Before and after this pilot survey, numerous site visits and a total of nine expert interviews as well as many interviews with inhabitants of relevance to understand the locality, current environmental problems, future plans, and appropriate payment amounts had been conducted with the regional authorities from the Department of Natural resources and Environment, the Department of Construction, the clean water company of the province (Quawaco), heads of residential quarters, heads and members of three collective farms. Further, within the WaterMiner project, environmental economists cohesively worked with ecologists, engineers, and specialists from Vinacomincompaniesto outline current issues and assure that future scenarios accurately reflect forward planning and potential solutions.

Then, this field study used a split-sample survey from 410 door-to-door in-person interviews collected in June and July 2018. Three parallel sub-surveys were carried out by four interviewers in the case study area. In detail, there were 120, 114, and 176 participants in projects 1, 2, and 3, respectively. The total response rate of this field survey was about 84%. A single household was used as the sampling unit and whoever was available at that time and able to represent the decision of the entire household was interviewed with the required minimum age of 20. To increase the chance of meeting working laborers at home, we worked on both weekdays and weekends as well as in late afternoons so that participants in the interview could be diversified. This ensures a representative sample of different occupations and ages.

Based on the CVM and the NOAA Panel guidelines (Arrow et al., 1993) [50], the questionnaire was designed with four parts: (i) an introduction of the WaterMiner project and the survey, (ii) an explanation of terminologies, e.g., ecosystem services, and warm-up questions regarding the connection/influence of the river and its ecosystem services with participants’ livelihood, (iii) description of current situations and proposed scenarios, the WTP elicitation question and follow-up questions, and (iv) socio-demographic questions to investigate correlations between participants’ socioeconomic characteristics and their WTP. Both ex-ante and ex-post approaches were used in part 3 of this survey to mitigate the hypothetical bias when stating WTP in the CVM (Whitehead and Cherry 2007 [51]; Loomis 2014 [52]).

Following the ex-ante approach, a cautious reminder about budget constraints was directly presented prior to the WTP question so that respondents answered this question in careful consideration of their financial ability. To avoid overstating the WTP, we also emphasized the obligation of households to pay as much as stated in this survey since a local tax/fee would be collected monthly by all households in the area for a period of three years if this referendum passed. This coercive payment mechanism is suggested to be a crucial aspect of the ex-ante approach in a binary choice WTP question to enhance the consistency between behavior on paper and in reality (Carson & Groves, 2007) [53]. Since participants of the survey were only households living in the vicinity of the river, information regarding the ongoing environmental planning of the local government and the mining companies in this area, which was the construction of a municipal wastewater treatment plant and an upgrading of mine water treatment process, respectively, was mentioned to highlight the importance of this survey in the process of further environmental planning in the region. Thence, we emphasized that the implementation of the river rehabilitation project depended tremendously on the results of this survey, i.e.,whether the majority of households supported it. Otherwise, the project would not be conducted in the future and the river would remain as in its current situation. In addition, a “cheap talk” script was used to explicitly encourage the honesty of respondents while answering and to avoid interviewer bias by trying to save the image of themselves in the eyes of the interviewer nor to please the interviewers. The writing style of the script was neutral and explicitly highlighted that there should be no right or wrong answer, but the inhabitants’ true opinions upon the proposed scenario of the river were highly appreciated.

Following the ex-post approach, participants answered follow-up questions by rating their attitude towards a series of attitudinal questions on a 5-point Likert scale after going through the WTP question. These attitudinal questions are of importance to additionally understand the influence of respondents’ mindsets on their paying decision.

We used the single bounded dichotomous choice elicitation format (Loomis et al., 2000) [30] where an exact amount to be paid by each household (the bid) was proposed and respondents could then accept or decline this offer based on how much they value the benefits stemming from this project (The WTP question in the questionaire was presented as Supplementary Materials). In our survey, six WTP levels/bids were used in six separate and randomly chosen groups in each of the three sub-samples (projects) and asked separately, including VND 20, 40, 60, 90, 130, and 200 thousand per month (equivalent to the USD 0.88, 1.76, 2.64, 3.95, 5.71, and 8.79 at the time of the conducted survey). The payment period would last for three years. One of the particular concerns was the mechanism for collecting the monetary contribution of households. Though annual payments are commonly used in the stated preference method literature, a monthly payment was considered more appropriate for our study since households in our research pay monthly fees and bills for the provision of such as electricity, water, and internet. This is in line with other studies conducted in developing countries as well as in Vietnam with similar characteristics of the research context, for example in the literature section of this paper. For instance, Börger et al. (2021) [24] make a change from an annual payment at the beginning to a monthly payment for five years that was found to work better for the coastal and marine conservation project in Vietnam.

4. Results

4.1. Data Description

Table 3. Description of variables used in the empirical analysis.

Variables	Description	Mean			Values or Measurement Levels
		Project 1	Project 2	Project 3
		N = 120	N = 114	N = 176
bid	The monthly payment				20, 40, 60, 90, 130, 200 (thousand VND)
income	Total household income per month	8062.5	8379.3	8941.9	Numeric variables (thousand VND)
value_result	Respondents’ evaluation of the result of river rehabilitation to their livelihood	3.72	4.0	3.83	1 = non-valuable; 2 = a little valuable; 3 = fairly valuable; 4 = valuable; 5 = very valuable
other_ppl	“Do you think that other people in this area also support the project?”	3.36	3.61	3.48	1 = strongly disagree; 2 = disagree; 3 = maybe; 4 = agree; 5 = strongly agree
government	“The local authority have good actions to protect the environment”	3.57	3.48	3.42	1 = strongly disagree; 2 = disagree; 3 = maybe; 4 = agree; 5 = strongly agree
age	The age of respondents	50.68	52.88	51.17	Numeric variables
gender	Gender of respondents	0.59	0.63	0.60	0 = male; 1 = female
education	Education level of respondents	1.84	1.81	1.73	1 = secondary or lower; 2 = high school; 3 = college/vocational training; 4 = university or higher
live	The number of years living in the current house?	26.27	26.21	24.30	Numeric variables
hhsize	Total number of individuals living in the house	3.68	3.67	3.82	Numeric variables

USD 1 is equivalent to VND 22,765 (at the time of the conducted survey).

summarizes the characteristics of the variables used in the empirical analysis for three projects to understand households’ WTP for river rehabilitation. The socio-demographic variables include income, age, gender, education level, the total number of persons living in a house, and the number of years living there. The attitudinal variables of interest including the perceived value of river rehabilitation, the perception of other people’s reactions to learn about informal institutions, and the perception of the local government about environmental protection to learn about local governance are expected to be partial explanations. On average, the participants’ ages in three projects are 50.68, 52.88, and 51.17 of which 59%, 63%, and 60% are female, respectively. A majority of them have been living there for a long period; the mean is around 25 years, which implies that they understand the river characteristics quite well and have witnessed considerable changes in the river and its ecosystem services over time. The education level is categorized into four groups and high school is the average education level of respondents, the mean education index is around 1.8. Perhaps due to the low education level and under the circumstances of living on the outskirts of Ha Long city near to mining workshops, the mean income of households there is quite low as at about VND 8.46 million per month (equivalent to about USD 372 per month at the time of the survey). On average, a household consists of four individuals.

A five-level Likert scale is used for responses to all attitudinal questions. Two questions regarding variables value_result and other_ppl are presented after respondents answer their WTP while the variable government appears in part (ii) of the questionnaire before the WTP question. In all three projects, the average grade of variable value_result is nearly 4 meaning that the result of the river rehabilitation is valuable to their livelihoods, though the mean in project 1 is a bit lower than in projects 2 and 3—3.72, 4.0, and 3.83, respectively. This observation goes well with the future scenario of projects (see Table 1). Project 1 consists of the clean-up scenario whereas the rehabilitate scenario is described as the final result in both projects 2 and 3, of which the rehabilitate scenario signifies a larger environmental improvement and, thus, more ecosystem services provided. Next, we consider the informal common institutions in the community through respondents’ thoughts on the reaction of other inhabitants (other_ppl) towards the river rehabilitation project. Respondents feel more than fairly confident that other inhabitants would be willing to pay at the given bid level. Again, we observe the same trend as in the previous attitudinal variable value_result so that the same possible reason is expected to explain the lower rating of variable other_ppl in project 1 (3.36) than in projects 2 and 3 (3.61 and 3.48). Respondents rate their satisfaction towards the local government regarding environmental protection in their living area (government), yielding 3.57, 3.48, and 3.42 in three projects, respectively.

Looking at the WTP response, the distribution of acceptance ratios by bid level follows a smooth decreasing trend for all three projects (see Figures S1–S3 in Supplementary Materials), meaning that the higher the bid level, the fewer number of respondents support the project at that given price. At the same bid, the corresponding acceptance ratios are generally lower in project 1 than in project 3. Note that the proposed future scenario in project 3 signifies a larger environmental improvement as compared to project 1 with the same starting point—status quo, reflecting a higher benefit generated by the former project. The acceptance ratios are 85.71% in project 1 and 91.67% in project 3 for the lowest bid level of VND 20 thousand, whereas they are 11.11% in project 1 and 29.17% in project 3 for the highest bid level of VND 200 thousand. These figures imply evidence that respondents in the survey were sufficiently sensitive to the cost as well as the trade-off between costs and benefits. We will take a closer look at the results of projects 1 and 3 in the coming section.

4.2. Regression Results

Table 4. Logistic regression results.

WTP	(1)	(2)	(3)	(4)	(5)	(6)
	Project 1	Project 2	Project 3	Project 1	Project 2	Project 3
bid	−0.0208 ***	−0.0165 ***	−0.0160 ***	−0.0202 ***	−0.0177 ***	−0.0161 ***
	(0.00514)	(0.00471)	(0.00371)	(0.00517)	(0.00489)	(0.00371)
income	0.000239 ***	0.000144 *	0.000205 ***
	(0.0000894)	(0.0000852)	(0.0000732)
lnincome				1.870 ***	1.302 **	1.474 ***
				(0.619)	(0.544)	(0.519)
value_result	0.324	0.781 **	0.528 **	0.406	0.802 **	0.525 **
	(0.310)	(0.356)	(0.219)	(0.314)	(0.363)	(0.219)
other_ppl	0.630 **	0.399	0.213	0.628 **	0.408	0.229
	(0.266)	(0.267)	(0.182)	(0.266)	(0.272)	(0.182)
government	−0.394	0.420	−0.467 **	−0.345	0.415	−0.436 **
	(0.252)	(0.260)	(0.191)	(0.253)	(0.264)	(0.189)
age	−0.0312	−0.0378	0.00177	−0.0317	−0.0370	0.00492
	(0.0217)	(0.0231)	(0.0187)	(0.0220)	(0.0236)	(0.0187)
gender	0.489	−0.360	0.344	0.500	−0.354	0.388
	(0.553)	(0.531)	(0.391)	(0.559)	(0.546)	(0.395)
education	−0.495 *	0.519 *	0.238	−0.539 *	0.467	0.229
	(0.288)	(0.294)	(0.231)	(0.285)	(0.298)	(0.232)
live	0.00440	0.0404 *	−0.0162	0.00564	0.0420 *	−0.0165
	(0.0186)	(0.0228)	(0.0165)	(0.0192)	(0.0235)	(0.0165)
hhsize	−0.112	−0.189	−0.322 *	−0.218	−0.298	−0.323 *
	(0.222)	(0.195)	(0.171)	(0.234)	(0.210)	(0.167)
cons	0.512	−4.849 *	−0.345	−14.19 ***	−14.83 ***	−12.09 ***
	(2.008)	(2.652)	(1.667)	(5.322)	(5.360)	(4.555)
N	120	114	176	120	114	176
Pseudo R2	0.3672	0.2907	0.2522	0.3821	0.3127	0.2537
Log-likelihood	−52.59	−56.05	−90.54	−51.36	−54.31	−90.35

Standard errors in parentheses. *** Significant at 1% level; ** Significant at 5% level; * Significant at 10% level.

shows the regression results of logit models constructed after Haab and McConnell (2002) for the three sub-samples. The dependent variable is the WTP of respondents for the river rehabilitation. WTP is equal to 1 if the answer to the WTP question is Yes and 0 if the answer is No. In all regression models, the bid variables produce significantly negative coefficients meaning that the probability of answering Yes statistically decreases as the bid increases, at the 1% significance level.

We now examine the potential effects of socio-demographic characteristics of respondents on their decision to accept or reject the bid. In the scientific literature, for instance, Pham et al., 2018 [54] for the context of Vietnam, besides using income as a linear variable in models 1, 2, and 3 (income) reported on the left side of Table 4, we also report corresponding models using the natural logarithm of income (lnincome) in models 4, 5 and 6 on the right side of Table 4 to account for the decreasing effect of income as it increases, ceteris paribus, on the probability of respondents’ WTP answer. Comparing the log-likelihoods of models 1 and 4 for project 1, 2 and 5 for project 2, 3 and 6 for project 3, models on the right side of Table 4 produce a slightly higher log-likelihood than models on the left side. Thence, we believe that models using the natural logarithm of income (lnincome) predict the probability of accepting the bid better. As expected, the positive coefficients of income (lnincome) indicate that the probability of accepting any given bid increases as household income rises. From regression results, we see that income (lnincome) has a statistically significant and positive effect on household WTP in models 4 and 6 at the 1% significance level, and in model 5 at the 5% significance level.

In general, age and gender have no significant effect on household WTP. No clear effect of respondents’ education level and the length of time in residence on WTP is found since the variable education and live produce coefficients with mixed signs across three projects and education is significant only in model 4 and live is significant only in model 5 both at the 10% level. These results imply that the probability of households supporting or rejecting the environmental project is independent of their age, gender, education level, and period of residency. Further, in project 3 the number of individuals living in a house (hhsize) has a significant impact on WTP at 10% with a negative sign, suggesting that the probability of accepting the bid decreases with increasing household size. This finding is plausible and in line with other studies since at a given income the budget constraint of a household becomes tighter the more individuals live in a household (Birol et al., 2009 [55]; Borzykowski et al., 2018 [56]; Ostermiller et al., 2015 [29]).

The two attitudinal variables, value_result and other_ppl, produce robust positive coefficients albeit not in all models when comparing the two model sets 1–3 and 4–6. We observe that the individual assessment regarding results of the river rehabilitation (value_result) has a statistically significant effect at a 5% level on the probability of accepting the proposed bid in project 2 and project 3, while the perception of other people’s WTP (other_ppl) has a significant effect at a 5% level only in project 1. Note also that either the variable value_result or other_ppl has a significant effect on the probability of WTP in any single model in Table 4. In other words, individual assessment and individual perception do not have a simultaneous significant effect on the decision-making process of respondents. We interpret these outcomes in the Discussion section.

The attitudinal variable government represents the link between local governance and the river SESs through the respondents’ viewpoint, which has a statistical effect on the probability of WTP for the river rehabilitation found only in project 3 (models 3 and 6) in an adverse direction at a 5% significance level. This means that respondents are more likely to contribute money to improving river ecosystem services if they feel less satisfied with the management of the local governance toward environmental protection so far in that area. From our regression results, both informal common institutions and local governance have significant effects on the likelihood of WTP for the environmental good despite neither in all three projects nor with the same significance level across three projects.

To estimate the mean WTP per month per household, we applied simple logit models including only the bid amount and a constant for all projects (Haab & McConnell, 2002) [57], then summarized the calculation results in

Table 5. Mean WTP per household per month and total WTP per household for each project in a three-year period. Unit: thousand VND.

	Monthly WTP	Annual WTP	Total WTP for the Project
Project 1	86.33	1035.96	3107.88
Project 2	83.85	1006.20	3018.60
Project 3	101.03	1212.36	3637.08

. WTP estimations show that the assumption that WTP₃ = WTP₁ + WTP₂ due to measures of project 3 comprising the sum of measures of projects 1 and 2 is clearly violated. Yet, scope effects in CVM, or more precisely scale effects in our case study, are still reflected when comparing mean monthly WTPs since the mean WTP of project 3 is higher than the mean WTP for project 1 at VND 101.03 and VND 86.33, respectively.

5. Discussion and Conclusions

5.1. Discussion

Our study demonstrates the significant welfare coming from proposed measures to improve river ecosystem services spillover to the community, in this case to households living near Lo Phong river and C2 stream, through the private company’s investment into efficient and effective management of mine water in project 1: clean-up and through the close cooperation between the private and public funding in project 3: clean-up + rehabilitate. These benefits were scrutinized making use of the concept of complex SESs with informal common institutions and local governance as integrated parts. While there is some uncertainty in the exact monetary value of such benefits to be achieved (see the below discussion on comparing the WTPs of such projects and about embeddedness in project 1 to clean up the river), significant benefits from improvements in river-related ecosystem services can be unambiguously expected in the vicinity of mine water rivers due to clean-up and rehabilitation projects.

One major issue to be discussed is the interpretation of the WTP results presented in Table 5. Since project 3 constitutes the sum of measures included in projects 1 and 2, the expectation for fully rational households valuing the presented projects would have been that WTP₃ = WTP₁ + WTP₂. However, we see that this is clearly not the case. While it is still plausible that WTP₁ and WTP₂ are not significantly different since the measures of those projects may generate more or less equal additional benefits, WTP₃ is much smaller than WTP₁ + WTP₂, consequently violating our assumption. This raises the question of which of the three estimates are valid and which is/are biased. To approach this question, we make use of our regression results shown in Table 4.

Of particular interest are the attitudinal variables value_result and other_ppl in Table 4. From the regression results for these two variables for models 1 and 4, we deduce that the motivation to support and contribute in terms of money for project 1 is less rooted in the benefits that the project yields (value_result being not significant contrary to projects 2 and 3) but in the sense that households want to contribute because they think other households do the same (other_ppl). The combination of value_result being insignificant, i.e., WTP is not determined by respondents’ assessment that the project will be beneficial, and other_ppl being significant and positive, unveils that a considerable part of respondents accepted the bid not for the benefits of that specific project but due a feeling of moral obligation to contribute because they believe other households will do the same. This indicates that WTP is “embedded” in other reasons for accepting the proposed bid, an effect described already by Kahneman and Knetsch (1992) [58] as “purchase of moral satisfaction” or in the context of Vietnam by Börger (2013) [59] as “social desirability” and “social influence” by Carlsson, Johansson-stenman, and Nam (2015) [22]. However, this weakens the interpretation of WTP as a measure of individual households’ benefits from the environmental project and improvement of ecosystem service provision for the clean-up scenario in project 1. In contrast, the level of benefits (value_result) from the improvements in ecosystem services statistically affects the probability of WTP to rehabilitate the river with a positive sign at a 5% significance level in projects 2 and 3 whilst the contribution from other inhabitants (other_ppl) has no significant effect on the likelihood of WTP (in models 5 and 6). These imply that unlike in project 1, the embeddedness is significantly less associated with the probability of respondents’ WTP for projects 2 and 3.

In another respect, the role of common institutions in our study, i.e., other_ppl, is possibly classified between altruism and social pressure explained by Dellavigna et al. (2012) [60]. According to the authors, the altruism (or warm glow) group is mostly motivated in a supply-driven form which maximizes the givers’ utility since they like to give, whilst the social pressure group is mostly motivated in a demand-driven form which reduces the givers’ utility since they hesitate to say no. On that basis, advocates of projects 2 and 3 are more likely to belong to the former group while advocates of project 1 are more likely to belong to the latter group due to the significant embedding effects found only in project 1. Nevertheless, we share the same thoughts with Carlsson et al. (2015) [22] that the field door-to-door experiment is not the main driver of social pressure in our study by three points. Firstly, the total amount of contribution for the project is certainly large relative to the average income. Secondly, river ecosystem resilience directly affects the respondents’ livelihood so this is not a charity donation. Thirdly, embeddedness would be found in all three projects provided social pressure is due to face-to-face interviews. An alternative driver lies in the magnitude of the hypothetical scenarios in our study, i.e., clean-up for project 1 versus rehabilitate for projects 2 and 3 since the hypothetical scenario in projects 2 and 3 is apparently more favorable than the one in project 1. The WTP is significantly lower for the less preferable good than the more preferable one, of which the decrease in giving occurs exclusively amongst small bids (Dellavigna et al., 2012) [60]. Small payments are undoubtedly more likely driven by social pressure than large payments. Similarly, the clean-up scenario is not that appeal to households but the motive behind supporting project 1 is consistent with the social pressure or the sense of duty (Börger et al., 2021) [24]. From these results, we deduce that the “true” benefit of project 1 (WTP 1) is much less than estimated in Table 5, and that the larger part of the total river-related ecosystem services (WTP 3) appreciated by the population in that region will be from more comprehensive rehabilitation of the river ecosystems (WTP 2).

5.2. Conclusions

Due to the various uncertainties involved, measuring the value of ecosystem services would be a challenge that requires a precautionary approach and does not always come to a very precise result (Costanza et al., 2017) [3]. Nonetheless, the process of listing the critical river ecosystem services and the attempt to value them in an apparent way through monetary units in our study contribute to promoting their recognition in public policy and awareness in society and likewise fostering benign environmental management initiatives towards sustainability and resilience, typically in developing economies.

From a methodical point of view, the findings of our study show that in the socio-economic and cultural context of Vietnam, here coastal Quang Ninh province, such state-of-the-art market simulation methods such as the CVM can be conducted yielding valid and plausible results. Regardless of the significant WTPs in terms of money from this study, future research about stated preference methods with new approaches to capture the valid WTP, hence the social welfare, fully exclusive from embedding effects is needed and expected to shed new light on estimating non-market environmental goods, especially in developing countries and low-income regions. In this respect, Börger et al. (2021) [24] suggest that the sense of duty found in their experiment examining WTP for marine conservation measures potentially comes from social factors and the politico-cultural context of low-income economies in general and Vietnam in their particular study. Moreover, the embedding effect is an extreme issue in CVM since embedding may naturally depend on the characteristics of the environmental good itself (Schulze et al., 1998) [61]. Note that our study proposes two different future scenarios, namely clean-up for project 1 and rehabilitate for projects 2 and 3. Given the outcomes of this study and available literature, we recommend future research take serious consideration in getting to know the social and cultural milieu of the specific project, in proposing the magnitude of future scenarios as well as in planning a set of corresponding attributes and measures to achieve the hypothetical environmental good in order to mitigate embeddedness in the contingent valuation method.

On the other hand, the significant effects of common institutions and local governance on the probability of WTP for environmental improvements are found, albeit neither in all projects nor with the same significance level across three projects, suggesting that they are the integrated parts of SESs in our study. Notwithstanding their unseparated roles, the empirical variables in our study as defined in the previous section reflect only a narrow aspect of institutions or governance in the multilevel linkages with SESs. Due to the divergence across regions, cultures, and countries, this study contributes toforming the sound basis for further studies to address institutions and governance in a broader perspective under a certain context with great caution, ideally from local to international levels, whilst interpreting the social-ecological systems and estimating the economic value of ecosystem services in that complex adaptive SES.