INTRODUCTION
Ho Chi Minh City (HCMC) is a major economic hub in Vietnam, with a population of 7.8 million and a density of 3,721 people per square kilometer The city's rapid urbanization has outpaced infrastructure development, leading to significant challenges, including an overwhelmed sewer and drainage system that contributes to severe urban flooding Additionally, unstable weather patterns, characterized by heavy rainfall, exacerbate these flooding issues.
To tackle the flooding issue in Ho Chi Minh City (HCMC), the Government has approved two significant plans: the urban drainage improvement and sewer development plan (Decision 752, 2002) and the MARD 1 construction plan (Decision 1547, 2008) Both initiatives aim to alleviate flooding but employ different strategies; the first focuses on enhancing the inner urban drainage system, while the second involves constructing large-scale hydraulic structures around the city Given the varied causes of flooding in HCMC, a combined approach utilizing both projects is essential for a comprehensive solution.
To address urban flooding in Ho Chi Minh City, two major projects have been approved: the urban drainage improvement and sewer development (Project 752) and the hydraulic construction plan (Project 1547) However, Project 1547 is encountering challenges with attracting investment and underestimating costs, potentially delaying its completion until 2025.
The Ministry of Agriculture and Rural Development (MARD) oversees projects aimed at mitigating flood damage Although the outcomes of two specific projects remain undisclosed, their cost-benefit analysis suggests that they are designed to address all tangible flood-related damages over the next 50 years, indicating an expectation of no significant flooding during this period (Steering Centre for Urban Flood Control Program, 2013).
The effectiveness of the implemented measures will be determined through future inspections of completed projects Although a cost-benefit analysis (CBA) for project 1547 has been conducted, it primarily relied on a ‘flood risk approach’ that did not account for the demand and willingness to pay (WTP) of residents in Ho Chi Minh City (HCMC) The findings of this study could guide decision-making for urban flooding projects and fee collection in HCMC and beyond Additionally, future flood control projects in HCMC should consider WTP as a crucial metric for estimating project benefits.
Evaluate inhabitants’ level of awareness about flood risks in HCMC
Determine the aggregate WTP for a hypothetical anti-flooding project
Find out the factors governing the WTP of HCMC inhabitants.
The Contingent Valuation Method (CVM) will be utilized alongside direct interview surveys to assess the willingness to pay for urban flood elimination in Ho Chi Minh City (HCMC) The findings from this study will inform policymakers about the feasibility of future projects and aid in evaluating the effectiveness of projects 1547 and 752.
LITERATURE REVIEW
Willingness to pay and willingness to accept
Willingness to pay (WTP) is defined as the monetary value an individual assigns to a good or service, as noted by Pearce (1997) This valuation is influenced by the individual's income, meaning those with higher incomes tend to value goods or services more than those with lower incomes In contrast, willingness to accept (WTA), as highlighted by Field (1997) and Pearce (1992), is not limited by income Consequently, individuals typically express a higher willingness to accept than their willingness to pay for the same item.
Bateman et al (2002) illustrate the concepts of Willingness to Pay (WTP) and Willingness to Accept (WTA) through an indifference curve graph, where the vertical axis indicates an individual's monetary expenditure on a private good (y) and the horizontal axis denotes the quantity of a public good (x) The indifference curves I and I’ depict combinations that yield the same utility level for two different individuals, with curve I representing a lower utility level compared to curve I’.
Figure 1 Measure of change in human welfare
When the quantity of a public good increases, its value can be assessed using various measures, including Willingness to Pay (WTP) and Willingness to Accept (WTA) For instance, if the quantity of public goods rises from x0 to x1, an individual's initial consumption point shifts from A to C, resulting in a decrease in private consumption represented by BC This reduction, BC, signifies the WTP for the enhanced public good, essentially reflecting the compensating variation associated with this increase.
If an individual's initial consumption is at point B and there is a reduction in public goods from x1 to x0, the individual may still experience an increase in private consumption, moving their consumption point from B to D.
The concept of DA represents the necessary private consumption that must be compensated to offset the loss of a public good In this context, the willingness to accept (WTA) for the reduction of the public good is quantified as DA, reflecting the compensating variation required for this loss.
The third and fourth measures represent the equivalent variation for the increase and decrease of public goods, termed as equivalent gain and equivalent loss, respectively Although theoretically equal to willingness to accept (WTA) and willingness to pay (WTP) in value, they are distinct concepts derived from different types of inquiries Equivalent variations are determined by asking how much would be considered good or bad in terms of gaining or losing a public good, rather than how much one would pay or receive as compensation for that gain or loss.
In theory, willingness to pay (WTP) and willingness to accept (WTA) should be equal, with only minor discrepancies allowed; however, numerous studies have demonstrated significant differences between the two Brown (2000) highlighted that disparities persist unless specific conditions are met, including the absence of income effects, transaction costs, perfect information about goods and prices, and a market that provides accurate references In practice, the gap between WTP and WTA is inevitable due to the difficulty of fulfilling all these conditions in a single study Researchers such as Nas (1996), Hanley (1997), and Hanemann (1984) argue that the divergence arises from the substitution between private and public goods, suggesting that WTP and WTA can only be equal when environmental goods and services are perfectly substitutable.
Figure 2 WTP and WTA in the case of perfect substitutability
Figure 3 WTP and WTA in the case of imperfect substitutability
Total Economic Value
The total economic value (TEV) of any change in welfare resulting from a policy or project encompasses the net sum of all relevant willingness to accept (WTA) and willingness to pay (WTP) TEV is categorized into use and non-use values, where use value is further divided into actual value and option value Non-use value includes existence value, bequest value, and altruism value.
The use value of an environmental good comprises two key components: actual use and option value Actual use pertains to the direct benefits derived from the good or service, such as visiting a park or mitigating urban flooding through prevention projects In contrast, option value reflects the willingness to pay for the preservation of the good for future use To assess the use value of environmental goods, two prevalent techniques are employed: the Travel Cost Method and the Hedonic Pricing Method, both of which fall under stated and revealed preference approaches.
The Travel Cost Method (TCM) is a widely used approach for assessing the recreational value of sites such as parks and hunting areas by estimating expenses related to travel, including gasoline, opportunity costs, entry fees, and service charges This method is significantly influenced by the distance to the site, with demand curves for site visits derived from survey data that considers visiting rates from various zones, local populations, and travel costs By analyzing the demand curve, one can infer the site's value through the calculation of consumer surplus, reflecting individuals' revealed preferences for recreational experiences However, it's important to note that TCM primarily measures the actual use value within the Total Economic Value (TEV) framework, emphasizing that the overall welfare of a site encompasses more than just its recreational value.
The Hedonic Pricing Method (HPM) is a widely used approach to assess the use value of environmental goods This method operates on the principle that the market price of a good reflects its associated amenities For example, the value of a property may be influenced by nearby environmental features, illustrating how HPM captures the economic worth of these attributes.
Use value Non-use value
Actual use Option value For others Existence
Hedonic Pricing Model (HPM) analyzes how the price of a house reflects its surrounding amenities, allowing for the inference of the amenity's marginal value or implicit price This model helps construct a demand curve for specific amenities, determining their economic value However, HPM does not account for the total economic value of environmental goods, as it overlooks non-use values, such as existence value, which pertains to the worth of preserving an environment without direct benefits Consequently, both HPM and Travel Cost Method (TCM) are primarily effective for assessing use value only.
Non-use value encompasses the willingness to pay (WTP) to preserve certain goods, even when there is no intention of using them This value includes existence, altruism, and bequest components Existence value reflects the WTP for a good to simply exist, driven by a sense of responsibility or concern for its presence, such as protecting a rare species Altruistic value pertains to the WTP to maintain goods for the benefit of others in the current generation, while bequest value arises from the desire to ensure that these goods remain available for future generations.
The Contingent Valuation Method (CVM) addresses the limitations of Traditional Cost-Benefit Analysis (TCM) and Hedonic Pricing Method (HPM) in measuring non-use value As noted by Mitchell and Carson (1989), CVM excels in capturing existence value by presenting respondents with hypothetical scenarios to express their willingness to pay (WTP) for environmental goods, such as river cleansing or electricity access in remote areas While CVM effectively measures non-use value, it struggles to capture actual use value, as respondents base their WTP on personal preferences tied to the hypothetical scenarios rather than real usage.
According to Steering Centre for Urban Flood Control Program (2013), the total economic value of the project 1547 could be summarized as following.
Table 1 Total economic value of the project 1547
Actual use Flood damage prevention (the damage could be cleaning cost, infrastructure damage, loss of productive land, crops, livestock, human lives, ecological goods…)
Option value Potential future uses
Utility Theory and the Utility Difference Approach
Following economic theory, Bateman et al (2002) defined the general form of indirect utility function as:
The utility function V(Y, P, S, Q) illustrates how a household's satisfaction is influenced by income (Y), the prices of goods (P), the availability of non-market goods (Q), and various socioeconomic factors (S) Higher prices can limit the quantity of goods a household can acquire, thereby reducing utility, while lower prices enhance access to goods and increase satisfaction Additionally, an increase in non-market goods signifies an improvement in utility, while a decrease indicates a decline Overall, the interplay of these elements shapes the overall utility experienced by households.
Assuming there is a provision of non-market good that raise the level of non-market good from initial Q0 to Q1 The household’s utility will be pushed up to a higher level:
Households seeking to enhance their well-being through non-market goods must invest a certain amount; however, as this investment increases, their disposable income decreases, leading to diminished overall utility Consequently, the maximum price a household is willing to pay for such improvements is equivalent to the amount needed to restore their utility to its original level prior to the enhancement.
The equation V(Y, P, S, Q0) = V(Y − C, P, S, Q1) illustrates the relationship between a household's well-being and its willingness to pay (WTP) for changes in that well-being Here, C represents the household's maximum WTP, which is also referred to as the compensating variation Importantly, C is influenced by various parameters within the model and is limited by the household's income, ensuring that the maximum WTP does not surpass the available income.
The household indirect utility function is represented as C = C(Q0, Q1, Y, P, S) = WTP ≤ Y, assuming a linear form This model considers fixed prices of market goods and the socioeconomic factors of the household The utility function prior to any changes in the non-market good is expressed as v0 = α0 + βY + γP + γS + γQ0 + μ0.
� 0 represents the unobservable factors that may influence the utility of the household.
With the provision of the non-market good, the new indirect utility function will be: v 1 = α 1 + βy + γp + γs + γq 0 + μ 1
The change in utility resulting from the variation in quantity (q) will be reflected in the term \( \delta_1 \) With enhanced welfare, households are anticipated to allocate a portion of their income to achieve the equation \( v_0 = v_1 \), where \( \beta y + \alpha_0 + \mu_0 = \beta(y - C) + \alpha_1 + \mu_1 \) This leads to the relationship \( \alpha_1 - \alpha_0 + \mu_1 - \mu_0 \).
The utility difference attributed to the provision of non-market goods can be expressed as α 1 − α 0 + μ 1 − μ 0 This observable utility difference can be modeled as a linear function of various household socioeconomic characteristics (X1, X2,…, Xk), represented by the equation: α 1 − α 0 = β 0 + β 1 X 1 + β 2 X 2 + ⋯ + β k X k.
Contingent Valuation Method
The contingent valuation method (CVM), as defined by Mitchell and Carson (1989), is a survey-based approach that assesses individuals' willingness to pay (WTP) for improvements in environmental amenities The first application of CVM was conducted by Davis (1963) to estimate the benefits of outdoor recreation in Maine's backwoods This method involves complex surveys that create hypothetical scenarios to gauge respondents' preferences regarding changes in goods or services Participants are asked how much they would be willing to pay for improvements or how much compensation they would require to forgo those goods or services The collected data from a representative sample are then analyzed to determine the economic value of the goods or services in question.
The Contingent Valuation Method (CVM) is one of the four primary techniques for assessing the environmental benefits of proposed regulations, noted for its ability to quantify non-use and passive-use benefits, making it superior to the Travel Cost Method (TCM) and Hedonic Pricing Method (HPM) (Bateman and Willis, 1999) Officially recognized by the US Water Resources Council as a valid valuation technique by the late 1970s, CVM has been widely utilized in various countries since 1980, particularly in developing nations (Do and Bennett, 2009; Hoa and Ly, 2009) However, the method has faced significant debates regarding its validity over the decades, with key discussions highlighted in the works of Cummings et al (1986) and Mitchell and Carson (1989).
In a landmark study, Carson et al (1992) assessed the economic losses from the Exxon Valdez oil spill in Prince William Sound, Alaska Despite initial skepticism from Exxon, the company responsible for the disaster, the study underwent several adjustments and ultimately gained acceptance from both Exxon and the U.S Government The final estimate of the damages caused by the spill was determined to be $2.8 billion, derived from a median willingness to pay (WTP) of $31 per household.
Some relevant CVM case studies
In 1988, Thunberg conducted a pivotal study on willingness to pay (WTP) for reducing flood risk, aiming to assess the monetary value of avoiding trauma or psychological damage from floods The study involved 142 telephone respondents in Virginia, who were interviewed about their WTP for a 20% reduction in flooding probability for their homes compared to a 40% risk Participants were presented with two payment options: a lump-sum payment and annual payments, along with follow-up questions to clarify the proposal and reduce bias The findings revealed a WTP of $314 for the lump-sum payment, highlighting the study's significant contribution in developing a WTP model specifically for flood risk reduction.
Shabman and Stephenson (1996) conducted studies building on Thunberg's (1988) findings to compare estimates from Hedonic Pricing Method (HPM), Contingent Valuation Method (CVM), and the Properties Damage Avoided (PDA) method with actual willingness to pay (WTP) choices from their own CVM survey Utilizing a payment card elicitation question and a follow-up confirmation to minimize hypothetical bias, the study revealed that HPM yielded a higher mean estimate of $1,333, compared to CVM's $314, while the PDA method produced an intermediate estimate of $597 Additionally, the research highlighted a significant discrepancy between hypothetical WTP ($124) and actual WTP when faced with real choices ($93 annually).
Zhai et al (2007) utilized the Contingent Valuation Method (CVM) to assess willingness to pay (WTP) for flood reduction in Japan, surveying 962 households to gauge interest in various flood control measures The study implemented a careful survey procedure to minimize bias, including the distribution of gifts and personalized letters from the Project Director, which aimed to reassure respondents that their input would influence policy decisions The findings revealed that WTP for anti-flooding measures, such as early warning systems and structural interventions, ranged from ¥2,887 to ¥4,861 Additionally, factors such as income, preparedness, risk perception, and access to environmental information were found to affect WTP for flood control initiatives.
METHODOLOGY
Current flooding and flooding control state in HCMC
Flood control has become a pressing issue for authorities in Ho Chi Minh City, affecting not only low-lying areas but also central districts during the rainy season Key factors contributing to flooding include tidal influences, high water discharge from the Saigon and Dong Nai rivers, flooding from the Mekong Delta, and heavy rainfall, often in combination The situation is exacerbated by an inefficient drainage system, urban elevation changes, land subsidence, poor spatial planning, and a lack of public awareness and engagement.
The ongoing flooding crisis in Ho Chi Minh City has led to substantial economic losses, with annual damages estimated between 6,000 billion VND and 22 billion VND, according to the Steering Centre for Urban Flood Control Program (2013) This worsening situation is exacerbated by three key factors: rising sea levels and erratic rainfall linked to climate change, land subsidence due to extensive high-rise construction and depletion of underground water, and inadequate sewer rehabilitation and flood protection measures that lag behind the city's rapid urbanization (Truong, 2010).
Figure 5 Monthly mean rainfall in HCMC and mean sea water level at Vung Tau
Source: Steering Centre for Urban Flood Control Program (2013)
The average rainfall in Ho Chi Minh City (HCMC) remains elevated from May to October, while the sea level at Vung Tau peaks in October and November Flooding typically occurs during the high rainfall months, particularly intensifying in September and October when sea levels are also at their highest This indicates that while heavy rainfall is the primary cause of flooding, the elevated sea levels exacerbate the persistent flooding issues (Phi, 2009).
In 2001, the Government approved a comprehensive urban drainage improvement and sewer development plan for Ho Chi Minh City, based on JICA's 1999 proposal This extensive project encompasses enhancements to 186 drainage routes, with 132 routes receiving dedicated support.
2 Due to the lack of sea level tracking stations in HCMC Measurements from the Vung Tau station were used.
The project, now referred to as Project 752 following Decision 752, is funded through Japanese ODA sources and the city budget, with a total investment estimated at 40,380 billion VND (Decision 752/QĐ-T Tg, 2001) Spanning a minimum of 20 years, the project aims for completion by 2020 Once finished, it will significantly reduce rain-induced flooding in the city by improving water circulation from urban areas to canals and enhancing the hydraulic capacity of the urban drainage system, thereby alleviating inundation during heavy rainfall events.
In addition to sewer rehabilitation and enhancing the urban drainage system, crucial measures for urban flood control and development, various hydraulic plans have been proposed to address the flooding issues in Ho Chi Minh City (HCMC) caused by heavy rainfall and tidal levels The Steering Center of the Urban Flood Control Program has recommended three key plans to the HCMC People's Committee and the Government: the Soai Rap barrier, the MARD plan, and a modified version of the original MARD plan.
In 2008, the Government issued Decision 1547 to plan and implement the MARD project Given the project's vast scale, known as Project 1547, it has been divided into multiple hydraulic construction phases.
The project aims to construct 12 sluice gates and 172 kilometers of dykes to enclose Ho Chi Minh City and parts of Long An Province, along with 8 bridges and the rehabilitation of 27 small canals, with a total estimated investment of 11,531 billion VND While the primary function of this infrastructure is to control tidal effects and improve water discharge from city canals to the Saigon River, it is crucial to note that effective urban flooding management, particularly during heavy rains, also depends on the efficiency of the city's drainage and sewer systems Since 2001, ongoing enhancements to these systems have been implemented under project 752, highlighting the importance of a comprehensive approach to urban flood control.
Hydraulic construction projects refer to the development of large-scale structures such as dams and dykes, which are distinct from irrigation or sewer and drainage rehabilitation projects These projects serve critical functions, including anti-tidal flooding measures However, funding limitations have restricted progress, resulting in the completion of only one sluice and 33 kilometers of dykes, as outlined in Decision 1547/QĐ-Ttg from 2008.
In 2012, the city initiated a project to rehabilitate the Tham Luong – Ben Cat – Rach Nuoc Len canal route, a vital waterway in Ho Chi Minh City This effort is part of Project 1547, which aims to enhance drainage conditions in the Go Vap, Binh Thanh, Tan Binh, Tan Phu, Binh Tan, and Binh Chanh districts The project involves comprehensive dredging of the canal and the construction of sluices, sewers, and water processing facilities along the route.
Figure 6 Number of flooding locations in HCMC from 2008 to 2014
Source: Steering Centre for Urban Flood Control Program (2013)
From 2008 to 2014, HCMC made significant strides in reducing urban flooding, as illustrated in Figure 6 The city successfully decreased the number of flooding locations from 126, showcasing a dedicated effort to improve urban infrastructure and manage water-related challenges effectively.
From 2008 to 2013, the central area of Ho Chi Minh City (HCMC) saw significant progress in flood alleviation, with improvements made in 11 locations However, 10 areas still experience severe flooding due to rainfall, highlighting the urgent need for a comprehensive upgrade of the drainage system in these regions.
The procedure of conducting a CVM study
According to Bateman et al (2002) A typical CVM work procedure follows the steps:
Testing the questionnaire and conducting the main survey
First, the research question must be specified and a hypothetical market or scenario is established The plausibility of the hypothetical market or scenario have to be ensured.
When selecting a survey method, options include face-to-face, telephone, or mail surveys Among these, face-to-face surveys are often favored for their efficiency, although they can be more expensive and time-consuming (Bateman et al., 2002).
To ensure a quality sample, it is essential to identify the necessary number of observations, the optimal locations and timing for data collection, and the appropriate methods to be employed.
The design of the questionnaire is a crucial step in the procedure, focusing on eliciting the scenario for the valuation question It involves selecting the type of valuation question—such as single-bounded dichotomous choice, double-bounded dichotomous choice, or open-ended choice—determining the payment vehicle for the hypothetical market, and deciding how to present the questions to respondents The table below outlines three common types of valuation questions and their specific details.
Table 2 Three types of commonly used elicitation questions
Description Respondents are asked to state a single amount of WTP for the scenario
Single-bounded dichotomous choice question
Respondents are asked to give a discrete choice (yes or no) to a pre- specified amount of WTP
Double-bounded dichotomous choice question
Respondents are asked to give a discrete choice to a specified amount, then another higher (lower) amount will be given if the respondent answer yes (no) to the initial question
WTP obtained Continuous WTP data Discrete choice WTP data
Advantages The exact value from respondent, accurate if the respondents know their WTP well.
Disadvantages Too difficult for most respondents (NOAA, 1993)
High non-response rate High possibility of
WTP understating due to the lack of knowledge about
Easier for respondent to answer since only a single yes-no question involved.
Possibility of yea-saying and nay-saying
Easier for respondents since it requires respondents to give the yes-no answer twice. Higher statistical efficiency than single- bounded question Incentives compatible (Carson & Groves,
2007) First bid question may impact the answer on the following question. More complicated econometric techniques required.
In addition to traditional elicitation methods, alternative approaches like the bidding game format and the payment card method can be utilized The bidding game involves respondents facing multiple rounds of discrete choice questions, with an optional open-ended willingness to pay (WTP) question included at the end However, this method has notable drawbacks, such as starting bias, where initial bid values can skew subsequent responses, and the potential for outliers Furthermore, it is not suitable for mail surveys or those lacking interviewer assistance Conversely, the payment card method presents respondents with a range of monetary amounts, starting from the lowest, and asks them to evaluate each incrementally increasing amount This method enhances the bidding game by significantly reducing outliers and yielding more accurate estimates Nonetheless, it is time-consuming and requires careful consideration from respondents, which can be challenging, especially for those with limited knowledge about the proposed amounts and their actual WTP, potentially leading to invalid estimates.
The choice of payment vehicle is crucial in contingent valuation method (CVM) studies, as it determines how funds are collected from respondents When a compulsory payment vehicle is used, inaccuracies may arise if respondents' actual willingness to pay (WTP) falls between the bid and their expected costs Conversely, voluntary payment vehicles, such as donations, may lead to inflated WTP estimates since respondents can choose any amount to pay Ultimately, there is no universally ideal payment vehicle; the selection should be tailored to the study's context and the appropriateness of the vehicle for the specific case.
The testing of the questionnaire is crucial and should involve a focus group interview, which allows for unstructured discussions with a small group of respondents Focus groups are a key qualitative research method (Malhotra, 1996) Additionally, a pilot survey should be conducted with a limited number of respondents to refine the questionnaire and train interviewers effectively (Bateman et al., 2002) Following these steps, the actual survey can be implemented.
After collecting the data, econometric techniques are employed to estimate the willingness to pay (WTP) in relation to various independent variables The complexity of this estimation process varies based on the type of data used; for example, interval data derived from double-bounded dichotomous valuation questions require more sophisticated techniques compared to continuous data.
The subsequent phase involves assessing the validity and reliability of the results It is essential to determine if the survey is comprehensible to respondents with diverse backgrounds and educational levels Additionally, it is crucial to evaluate whether the variables are appropriate for measurement and if alternative valuation methods yield consistent results (Bateman et al., 2002).
One important issue worth noting when conducting a CVM study is the problem of
Hypothetical bias refers to the discrepancy between a respondent's stated willingness to pay (WTP) in a survey and their actual WTP due to the hypothetical nature of the scenario (Loomis, 2011) While the theoretical underpinnings of this bias remain unclear, Loomis (2011) identified five potential solutions to mitigate it First, respondents can be convinced that survey results will inform policy decisions, which enhances the incentive for accurate responses (Carson & Groves, 2007) Second, the use of "cheap talk" encourages respondents to provide truthful WTP as if making a real financial decision (Cummings and Taylor, 1999), although its effectiveness is debated Third, eliminating uncertainties regarding payment and the provision of goods can help, as clarified by Mitani and Flores (2010), though this may require researchers to misrepresent the certainty of the goods offered Fourth, Champ et al (1997) suggested employing follow-up questions after binary choice queries to gauge respondents' certainty, which has shown effectiveness in some studies These strategies collectively aim to reduce the impact of hypothetical bias in willingness to pay assessments.
In 1994, the National Oceanic and Atmospheric Administration (NOAA) suggested a method to reduce the final Water Treatment Plant (WTP) costs by 50%, although the rationale behind this percentage remains unclear (NOAA, 1994).
To aggregate the results effectively, one can multiply the average or median willingness to pay (WTP) by the total population Alternatively, welfare changes for the entire population resulting from a project or policy change can be determined by regressing WTP responses against independent variables, allowing for the extraction of parameters that facilitate the calculation of an aggregate measure of welfare change (Freeman, 1993).
Survey procedure of the study
The Contingent Valuation Method (CVM) is particularly effective for assessing non-marketed goods and services, making it ideal for this study (Bateman et al., 2002) Additionally, CVM proves to be more practical in developing countries, where it typically achieves higher response rates than in developed nations (Grosh & Glewwe).
The Contingent Valuation Method (CVM) offers several advantages over traditional assessment methods, particularly its capacity to capture existence value, which reflects an individual's willingness to pay for the preservation of goods or services (Mitchell & Carson, 1989) Notably, there has been a lack of research measuring willingness to pay (WTP) for urban flooding prevention in Vietnam Consequently, this study will employ CVM to evaluate the WTP for flood prevention measures in Ho Chi Minh City.
The survey will be conducted through face-to-face interviews, focusing on the flooding situation in Ho Chi Minh City (HCMC) Respondents will provide information about their household characteristics and expectations regarding flooding, followed by an introduction to a hypothetical project—Project 1547, approved by the HCMC People’s Committee in 2008, along with a brief overview of Project 752, which aims to rehabilitate the urban sewer and drainage system Although these projects have been initiated, their slow progress necessitates the use of a hypothetical scenario where the projects are expected to fully alleviate urban flooding upon completion After detailing the proposed project, respondents will have the chance to ask questions before being presented with a key inquiry about their willingness to contribute financially to the project.
The study utilized direct in-home face-to-face interviews as the survey mode, highlighting several advantages identified by Bateman et al (2002) These advantages include ensuring accurate responses from household members, controlling the flow of information, providing interviewer assistance, correcting misunderstandings, enhancing respondent motivation, capturing verbatim comments, and monitoring respondent confidence, particularly regarding attitudes towards specific disasters Additionally, the implementation of computer-assisted personal interviewing (CAPI) involved using electronic tablets with built-in questionnaires, streamlining data collection and reducing errors associated with traditional paper surveys.
The study employs a double-bounded dichotomous contingent valuation method to determine respondents' willingness to pay (WTP) for a hypothetical flooding prevention project in Ho Chi Minh City (HCMC) This approach is favored for its high statistical efficiency compared to open-ended or single-bounded methods, as noted by Bateman et al (1999) Each participant is presented with two pre-specified amounts that define the limits of their maximum WTP: an initial bid followed by a higher or lower bid based on their response By analyzing the combinations of responses—yes-no, yes-yes, no-yes, and no-no—the study infers the WTP interval for each individual, providing valuable insights into public valuation of the project.
The payment vehicle of choice is also crucial to a CVM study (Morrison, Blamey,
In this study, two payment vehicles were considered for collecting hypothetical payments from respondents: a one-time voluntary payment and a one-time mandatory increase in electricity bills While the mandatory approach might seem more convenient for urban households, research by Morrison, Blamey, and Bennett (2000) highlights that such a method can lead to vehicle bias due to respondents' skepticism about the one-time nature of the charge Additionally, Ryan (2006) noted that collective payment mechanisms often yield higher willingness to pay (WTP) estimates compared to voluntary options To mitigate potential bias, this study opted for the one-time voluntary payment vehicle.
To mitigate hypothetical bias in the study, three key strategies were implemented First, respondents were assured that the hypothetical projects, specifically project 1547 and project 752, had been actualized, with details provided about their initiation and scale, while intentionally withholding current progress to prevent protest bids Second, after completing the bidding question, respondents received a vocal confirmation prompt to affirm their certainty regarding their bids, encouraging them to reconsider and adjust their responses if needed Lastly, interviewers consistently emphasized the importance of honest answers throughout the survey to enhance response integrity.
At the onset of the study, a focus group discussion aims to gather insights on urban flooding, informing the questionnaire design and establishing basic bid range suggestions This focus group comprises six individuals who have lived in Ho Chi Minh City for over a decade Participants evaluate the clarity of the proposed scenario and assess the appropriateness of specific questions.
Following the construction and revision of the questionnaire by the focus group, a pilot survey was conducted to pre-test its effectiveness The primary aim of this pilot survey was to identify appropriate bid levels and validate the questionnaire's questions A total of 20 randomly selected households participated, providing specific open bids in response to the scenarios presented, rather than merely answering with a yes or no The results from the open bid question in the pilot survey are detailed below.
Figure 7 Open-bid responses from the pilot survey
According to the collected pilot-survey data, 18 sets of bid were designed, these sets of bids would be used in the final survey.
Table 3 Details of 18 sets of bid used in the survey
The sample for the study was selected using a two-stage sampling design, beginning with the random selection of four districts: District 6, Thu Duc District, Binh Tan District, and Binh Thanh District Within each district, one area was designated for the survey, which was conducted with the assistance of a designated area leader to ensure the integrity of the responses Households were informed in advance via telephone, and participants were restricted to those of working age, specifically 18 years or older The total sample size comprised 180 individuals, excluding pilot observations, distributed across the selected areas.
Table 4 Distribution of observations across areas
District Area Number of observations
District 6 An Duong Vuong Street 30
Binh Tan District An Duong Vuong, Kinh Duong Vuong Street 45
Binh Thanh District Nguyen Huu Canh, Nguyen Xi Street 53
Thu Duc District Kha Van Can Street 52
Following the survey, the mean and willingness to pay (WTP) will be calculated using both non-parametric and parametric estimation methods For non-parametric estimation, this study will employ the Turnbull Self Consistency Algorithm (TSCA) as proposed by Bateman et al (2002) to effectively handle double-bounded data In contrast, the parametric estimation will utilize logistic regression to estimate the mean WTP.
Main contents of the questionnaire
The contents of the questionnaire are summarized as follows (for the full questionnaire, see the Appendix).
Table 5 Main contents of the questionnaire
Education: Primary or secondary school, high school, college or university, post-graduate.
Structure: Concrete house or semi-concrete house, floor number, floor square
House location: district and street
Distance from river or canal: (1) less than 100m; (2)
100 m ~ 500 m; (3) 500 m ~ 1 km; (4) 1 km ~ 2 km; (5) more than 2 km
Zone: whether the house is in zone I or not
Past flood experience: Yes or no
Perception of future: Do you think flood situation will eventually get better in the future (yes or no)
Perception of consequences of current flooding (yes or no)
From the scale of 0 to 4 (from not concerned to very concerned):
Environmental risk: Air pollution, noise pollution, water pollution
Urban risk: Robbery, traffic accidents, fire hazards.
Does the house has private anti-flood measures? (yes or no)
Bid question Double-bounded dichotomous question regarding how much the respondent is willing to pay for the proposed project.
Does the respondent think the proposed project will be effective? (yes or no)
Does the respondent think proposed project will be harmful to the environment or to them? (yes or no)
Does the respondent think Government will use their contribution for the project well? (yes or no).
Description of the hypothetical project
The hypothetical project is based mainly on the project 1547 is originally proposed in 2008 First, the figure shows the division of HCMC into 3 zones (see figure 8).
Zone I (yellow zone): contains all of the left-hand side of the Saigon river plus Nha Be District (central HCMC districts and a part of Long An
Zone II (red zone): contains the area around Saigon river – Dong Nai river tri- fork (District 2, 9 and Thu Duc District)
Zone III (blue zone): contains Can Gio district only.
The project emphasizes the protection of zones I and II, while zone III is recognized as a significant ecological area that is vulnerable Consequently, no measures are implemented in zone III as part of the project's initiatives.
Source: Steering Centre for Urban Flood Control Program (2013)
Zone I is safeguarded by 172 kilometers of dykes and 12 sluice gates, providing crucial protection for both urban and rural areas of Ho Chi Minh City against tidal flooding and river discharges This infrastructure not only prevents flooding but also helps regulate water levels in the protected area by improving the outer drainage system Additionally, the enhancement of 186 urban drainage routes will further strengthen the inner urban drainage system.
Source: Steering Centre for Urban Flood Control Program (2013)
Zone II protection emphasizes elevating lowland embankments and constructing smaller dikes to safeguard ecological and tourism zones Additionally, dredging of canals and primary rivers will enhance drainage capacity.
Figure 10 Protection for zone II
Source: Steering Centre for Urban Flood Control Program (2013)
Non-parametric estimation technique
Non-parametric estimation offers a significant advantage over parametric regression models by not requiring any model assumptions Unlike parametric estimation, which depends exclusively on bid level data and household choices, non-parametric methods can effectively handle data variability The double-bounded dichotomous question generates overlapping interval data, which presents unique challenges To address this, Bateman et al (2002) introduced the Turnbull Self-Consistency Algorithm (TSCA) as a solution for analyzing such data The TSCA procedure provides a systematic approach to managing overlapping intervals in estimation.
The study's interval data will be categorized into two main types: basic intervals, representing the smallest units of willingness to pay (WTP) data, and overlapping intervals, which include multiple basic intervals Specifically, the classification will begin with eight primary basic intervals.
Table 6 Interval distribution of WTP responses
Interval code Lower bound Higher bound Number of HHs in interval
TSCA attempts to calculate the survivor function at each of the boundary values Bj
To determine the willingness to pay (WTP) across various thresholds (0; 20,000; 50,000; 250,000; 450,000; 1,000,000; 3,000,000; 5,000,000), we calculate the proportion of households with a WTP exceeding each threshold by dividing the number of such households by the total sample size of 180 However, the accuracy of this calculation may be compromised, as some households' WTP spans multiple intervals, leading to uncertainties in the final counts.
The next step involves calculating the probabilities for each basic interval, determining the likelihood that a household's willingness to pay (WTP) will fall within that range This is achieved by subtracting the probability of the lower bound of one interval from the probability of the lower bound of the subsequent higher interval.
In overlapping intervals, the allocation of households will be distributed among the basic intervals they cover, based on the calculated probability of each household falling within those specific intervals.
In conclusion, the introduction of a new set of households defined by basic intervals allows for the construction of the survivor function By repeatedly iterating through this process, we can achieve convergence of the mean willingness to pay (WTP).
Parametric estimation technique
The utility difference model can be expressed as z = α 1 − α 0 = β 0 + β 1 X 1 + β 2 X 2 + ⋯ + β k X k + βC − (μ 1 − μ 0 ) When the probability of a positive response is modeled using the cumulative distribution function of a logistic covariate and assumes a zero mean error term, it leads to a logit specification This framework allows for the calculation of the likelihood of a 'yes' response.
Whose coefficients will be estimated by maximizing the likelihood function in a logit model.
This study analyzes willingness to pay (WTP) data as discrete choice data, utilizing 180 double bounded responses to generate 360 binary responses Additionally, a bid variable will be incorporated into the model to represent the bid level faced by households during the choice elicitation process.
After estimating the coefficients of the logit model, the next step is to compute the mean and median of willingness to pay (WTP), based on the assumption of a logistic distribution, as outlined by Haab and McConnell (2002).
X̅ is the row vector of sample mean, including 1 for the constant β′ is the column vector of estimated coefficients excluding the coefficient of bid β 0 is the coefficient of ‘bid’ variable.
Confidence intervals of mean WTP
The estimated willingness to pay (WTP) figures represent only an approximation for the entire population, derived from a sample of households, and different samples may yield varying WTP results Therefore, after calculating the mean WTP using both non-parametric and parametric methods in this study, it is essential to assess the accuracy of these estimates.
Bateman et al (2002) identified two primary approaches to estimation: the analytical approach and the numerical approach, also known as bootstrapping Bootstrapping demonstrates significant robustness and versatility, applicable to various estimation techniques and data types This method involves iteratively estimating willingness to pay (WTP) from numerous randomly generated datasets, with observations drawn from the original dataset The resulting WTP values are then used to infer confidence intervals (Krinsky & Robb, 1986).
This study will utilize bootstrapping techniques to generate a 95% confidence interval for the mean willingness to pay (WTP) The parametric estimation will be conducted using a specific STATA command, while the non-parametric estimation will be performed with Crystal Ball software.
ANALYSIS RESULTS
Descriptive statistics
The final data consists of 180 observations, this part aimed to present the result of the descriptive statistics of some important variables in the collected data.
In the analyzed sample, males constitute 57% (104 out of 180 participants) The age distribution reveals that 35.5% of the sample falls within the 16-24 age range, while 32.2% are aged 25-35, and another 32.2% are between 36-60 years old.
Figure 11 Distribution of age in the sample
Semi-concrete houses Concrete houses
The analysis reveals a significant correlation between household income and housing type The data indicates that the first 26 observations pertain to semi-concrete houses, highlighting a clear distinction in average income levels Notably, households residing in concrete houses exhibit a higher average income compared to those living in semi-concrete structures.
Figure 12 Monthly income distribution and House condition
The average monthly income across the sample is VND 8,725,833, with households in semi-concrete and concrete homes earning VND 4,692,308 and VND 9,406,818, respectively The per capita income stands at VND 3,110,000, which is slightly below the average household income in Ho Chi Minh City This discrepancy can be attributed to a significant proportion of low-income households within the sample, limited access to affluent residential areas, and a tendency among respondents to underreport their income Notably, 28% of the households surveyed reported having an income.
In 2012, the monthly per capita income in Ho Chi Minh City (HCMC) reached VND 3,652,000, exceeding 10 million VND, indicating a significant presence of medium-income households in the city, as reported by the General Statistics Office (GSO) in 2013.
The bid response is a crucial aspect of the data analysis, utilizing a double-bounded dichotomous question format This method presents respondents with two bid questions, allowing for a comprehensive understanding of their preferences The statistics for the initial bid question are detailed below.
Table 7 First bid question response statistics
As the bid level rises, the percentage of affirmative responses also increases Notably, at a bid of VND 50,000, only 18 out of 40 participants responded positively, which is unexpected, as lower bid amounts typically elicit a higher rate of agreement.
‘yes’ Further inspection of these households who said ‘no’ with the initial bid ofVND 50,000 (22 households) is shown below
Figure 13 Income and ownership of households with ‘no’ response at VND 50,000
The average income of the 22 households is 6,800,000, significantly lower than the overall average, with half of these households renting their homes This financial situation may contribute to their apathy and the responses they provided in the survey The statistics from the follow-up questions are detailed in the table below.
Table 8 Second bid question response statistics
The follow-up question shows an increase in 'yes' responses as bid levels rise, with a total 'yes' rate of 45%, significantly higher than the initial question This discrepancy may stem from respondents feeling persuaded by the first question or experiencing shame when answering 'no.' However, a deeper analysis reveals that of the 82 'yes' responses, 35 (42%) come from the lower bid levels of 20,000 and 50,000 Excluding these bids results in a 'yes' response rate of only 34%, suggesting minimal influence bias from the first question in the study.
Non-parametric estimation result
After applying TSCA for the sample data, the resulting survivor function is then constructed as follows:
Figure 14 Survivor function of non-parametric estimation
The average willingness to pay (WTP) is VND 484,654, while the median WTP stands at VND 250,000 A bootstrapped 95% confidence interval for WTP ranges from VND 348,499 to VND 631,757 In the TSCA non-parametric estimation, the median WTP represents the bid value that results in a 50% probability of acceptance, indicating it is determined by boundary values The notable disparity between the mean and median WTP can be attributed to the low probability of higher interval allocations and the variability in interval distances.
Parametric estimation results
This study employs parametric estimation through logistic regression, focusing on a binary dependent variable (yes/no) linked to the bid amount presented to respondents Each participant is surveyed twice using a double-bounded dichotomous valuation question, resulting in a dataset that expands to 360 observations Essentially, each household is analyzed as two distinct entities, maintaining identical characteristics while differing in their bid responses.
The relationship between willingness to pay (WTP) and income is fundamental, as highlighted by Carlsson et al (2002), who noted that WTP is influenced by the marginal utility of income Hong (2001) further elaborated that this relationship is positive for normal goods and negative for inferior goods Additionally, education level impacts WTP by increasing environmental awareness and potentially influencing income Gender also significantly affects WTP, particularly in Vietnam, where women, often not the primary income earners, may be reluctant to pay despite recognizing the value of a proposed project.
The willingness to pay (WTP) for flood reduction measures can be influenced by the type of housing a respondent resides in Specifically, individuals living in rental properties or concrete houses with private anti-flooding measures may be less inclined to pay for such measures Conversely, those residing near canals or in larger homes may be more willing to pay for flood reduction measures, highlighting the importance of housing characteristics in determining WTP.
Zhai et al (2007) established a framework that identifies various factors influencing the willingness to pay (WTP) for flood risk reduction, including resident attributes, house characteristics, flood risk acceptability, and perceptions of both flood and other risks This aligns with previous studies by Freeman (1993) and Kula (1997), which highlighted the connections between WTP, income, socioeconomic variables, and environmental changes.
The variables used in the model are described as follows:
The choice of respondents against the bid level
The bid offered to the respondents
Gender Dummy variable, binary data
Number of dependents Continuous data Residence period in HCMC Continuous data Education with four levels 3 dummy variables Households’ characteristics
Structure: concrete house or semi concrete house
Ownership: private or rental Dummy variable Distance to river or canal: four levels
3 dummy variables indicating 4 level of distance from river or canal
Number of floors Count data
Total floor square Continuous data
The street in which the respondents reside in (An Duong Vuong, Kinh Duong Vuong, Nguyen Xi, Kha Van Can, Nguyen Huu Canh)
Past flood experience Perception of future situation Perception of consequences of current situation
Perception about 17 different natural disasters, natural issues and urban issues Each with a scale of 1 to 4
Indicate whether the house has private anti-flooding measures
Effectiveness of the project Harmful environmental impact of the project
1 dummy variables for each question.
1 dummy variable for each question Responses with 3 or 4 on the scale will be coded as 1, 0 otherwise Dummy variable (1 if has private measure, 0 otherwise)
3 dummy variables, one for each question
Initially, an unrestricted logit model was conducted using all 48 variables; however, it did not yield significant results for 40 of them at the 5% level Subsequently, a stepwise procedure was applied, resulting in a restricted logit model with 8 variables that clustered at the 5% significance level A likelihood ratio contrast test between the two models indicated that the null hypothesis regarding the necessity of variable inclusion could not be rejected The results of the restricted model are detailed in the Appendix, which includes the full findings for both models.
The analysis of the dependent variable, choice, reveals significant findings related to various factors A notable negative coefficient is observed for bid_in_million (-3.057), indicating a strong inverse relationship with choice, supported by a p-value of 0.00 Gender shows no significant impact, with a coefficient of 0.098 and a p-value of 0.80 Educational attainment appears to influence choice, particularly for those with a college education (-2.394, p = 0.01) and high school education (-0.926, p = 0.32), while vocational and university education show no significant effects Age also plays a role, with a coefficient of -0.029 and a p-value of 0.03 suggesting a slight negative impact Geographic factors such as street_anduongvuong (-2.124, p = 0.00) and street_kinhduongvuong (-1.273, p = 0.00) significantly affect choice Income positively correlates with choice (0.347, p = 0.00), while distance from the river shows varied effects, with only the distance of 500m to 1km yielding a coefficient of 0.779 and a p-value of 0.19, indicating no significant impact.
The 6 Stepwise procedure involves the gradual elimination of unsatisfactory variables through individual LR tests, where the null hypothesis posits that a variable equals zero Clustering is the process of grouping certain variables, allowing them to either be included or excluded together in a restricted model, often utilizing dummy variables within a category Additionally, bids and income are quantified in millions of VND, with dummy variables being italicized, such as gender_male.
4 education dummies are in unrestricted model only Other variables are in the restricted model
The distance to the river and canal category, while not significant in the restricted model, remains in the model due to its predictive power The negative coefficient for distances between 100m to 500m suggests that respondents are less willing to pay as the distance decreases, while other distances show a positive relationship Additionally, other variables in the model are statistically significant at the 5% level.
The analysis reveals that bid and income variables exhibit negative and positive correlations, respectively, indicating that an increase in bid leads to a decreased likelihood of voting for the project, while higher income raises this probability This aligns with both common sense and economic theory Notably, the impact of each VND increase in bid is eight times more significant than that of income, highlighting greater sensitivity of respondents' willingness to pay (WTP) to bid changes compared to income variations Additionally, the age variable shows a negative correlation, suggesting that older respondents are less likely to support the project, whereas younger individuals demonstrate a stronger inclination to vote in favor.
The voting tendencies in Ho Chi Minh City reveal significant disparities, particularly among residents of An Duong Vuong Street and Kinh Duong Vuong Street, who show a reluctance to invest in a proposed project compared to those living on three other streets This hesitance may stem from the frequent flooding experienced in these areas, which raises uncertainty about the project's potential outcomes According to the Urban Flooding Control Centre (2013), Ward 6, encompassing both An Duong Vuong and Kinh Duong Vuong Streets, has faced persistent flooding issues since 2008.
The mean and median willingness to pay (WTP) derived from the restricted model with eight variables is VND 380,000, falling within a bootstrapped 95% confidence interval of VND 268,822 to VND 480,079 A comparison between the unrestricted and restricted models is presented below.
Table 11 Unrestricted model and Restricted model comparison
Number of significant variables at 5% 8 5
Number of variables with VIF >10 6 0
Likelihood Ratio contrast test Not preferred Preferred
The restricted model demonstrates a comparable predictive power with a count R² of 0.83, closely aligning with the unrestricted model's 0.85, making it effective for estimating willingness to pay (WTP) Additionally, the unrestricted model suffers from significant collinearity issues, and there is no notable difference in WTP between the two models Consequently, a mean WTP of 380,000 is established as a reliable parametric estimate for the study's valuation.
The parametric technique yields a more conservative estimate (VND 484,654) compared to the non-parametric method, likely due to the significant negative coefficients in the restricted model While the non-parametric approach focuses solely on willingness to pay (WTP) choices and bid amounts, it overlooks other important factors, such as income Although there is no definitive evidence favoring either method, the conservative principle suggests that the approach yielding the lower result is preferable (Daniel & Jouni, 2002).
The willingness to pay (WTP) differs between males and females, with recalculated estimates showing VND 360,000 for males and VND 400,000 for females This finding challenges the common belief that males, as primary earners, generally pay more However, the gap is minimal and can be attributed to housewives' heightened concern for their homes.
The anticipated impact of education on willingness to pay (WTP) in this study may not be substantiated Specifically, the average WTP for 99 households with respondents holding a college degree or higher is VND 250,000, whereas the average WTP for the remaining 81 households is VND 490,000 This indicates that individuals with higher education do not necessarily exhibit a greater WTP, and the disparity in WTP between different education levels may be attributed to income variations Consequently, the income channel through which education influences WTP appears to be absent in this study.
Figure 15 Monthly income classified by education groups
Five education level groups from left to right are high school, vocational training, college, university and post-graduate The average income for 2 first groups and last
3 groups are VND 8,884,000 and 8,590,000 respectively This may partially explain the neutrality of the education level toward WTP.