**1. Introduction**

Global climate change has intensified precipitation irregularity, lake and river surface decline, and water quality deterioration [1–3], which in turn has hampered water management efficiency [4,5]. In South Korea, drought and flood damage keeps recurring [6–8]. Hence, countermeasures against climate change are being promoted. One of the highlighted issues is dam operation [9,10]. Recently, since water supply and demand management in response to climate change has become a national problem, a variety of measures have been proposed for effective water resource management. This includes restructuring the main role of hydroelectric dams to supply water during drought, and flood defenses [11]. Efficient dam operation plans are urgently required to manage drought, flood stress, and water quality. Multi-purpose dams benefit local people, directly and indirectly, by providing domestic and industrial water, electricity generation, and ecotourism as well as drought relief and flood prevention [12,13]. Furthermore, the fact that the reservoir water condition is highly relevant to drinking water quality and recreational value for local people has added significance [14,15].

The point of interest here is that water supply and distribution should be government controlled, since the benefits of using dams are characterized by public goods more than private goods [16]. Therefore, it has become a major concern to confirm the input cost validity (The feasibility of the dam project is determined by a cost and benefit economic analysis, and the result of comparing these two figures affects investment decisions [17].) when implementing dam operational improvement projects for public use. Such procedural

**Citation:** Oh, H.; Yun, S.; Lee, H. Willingness to Pay for Public Benefit Functions of Daecheong Dam Operation: Moderating Effects of Climate Change Perceptions. *Sustainability* **2021**, *13*, 14060. https://doi.org/10.3390/su132414060

Academic Editors: Alban Kuriqi and Luis Garrote

Received: 9 November 2021 Accepted: 16 December 2021 Published: 20 December 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

justification can be ensured in case the benefit exceeds the cost [17,18]. At this point, as the operational benefits (including drought prevention, flood protection, and water quality management) [19–21] are services for unspecified individuals, public interest valuations are eventually considered for judging government project performance. First, given that time or cost constraints are unavoidable, examining core factors of value inducement and identifying influencer priorities might be regarded as important to foster business efficiency.

Several relevant studies have emphasized the importance of identifying climate change impacts on water management. Vital research problems about the economic value of drought stress alleviation, flood risk management, and water quality improvement have been globally discussed, thus, contributing to the awareness of the economic value of the public benefits provided by dam functions. However, the results of these studies did not examine the direct value. Furthermore, it is difficult to immediately compare results from different analysis environments due to different spaces and timeslots.

Therefore, this study primarily investigates the economic value of the role of dams in coping with climate change and benefiting the public through drought management (DM), flood control (FC), and water quality monitoring (WM). Thus, it ultimately provides foundational information through which to highlight the public benefits of establishing water-resource countermeasures against climate change. Considering the severe damage caused by floods, droughts, and water pollution by South Korea's changing climate, dealing with the three functions is desirable. In addition, the study investigates the moderating effect of climate change risk perceptions on the economic value of each public function; the perceived public value can vary significantly according to the climate change awareness level. Confirming whether there are discriminative values is considered necessary for highly acceptable policy drives. Some studies (e.g., [22–24]) have shown that climatechange awareness affects the acceptance of dam operation policies. Thus, generalizing the results of the study without additional verification regarding the subdivided value might distort the value judgment.

This study investigates the benefits of the Daecheong Dam. The Daecheong Dam completed in 1980 as a multi-purpose dam—is 72 meters tall and is 495 meters wide. Its catchment area is 4,134 square kilometers with a capacity of about 1.49 billion cubic meters. The reservoir, formed by dam administrators, is located within Chungcheongnam-do and Chungcheongbuk-do. As a serious water-bloom phenomenon increased after the 2015 dry season, there was an emergency in water quality management regarding Lake Daecheong the source of drinking water for the Chungcheong region. The waterworks authority exercised closer monitoring of harmful algal blooms at Daecheong reservoir [25,26], located in Daejeon metropolitan city, South Korea. The public role of the Daecheong Dam, directly and indirectly, include the benefits of water supply for agricultural, industrial, and residential use, as well as the supervision of water quality for drinking and recreation [27–29]. Drought, flood, and water pollution emphasize how crucial dam operation plans can be. Moreover, this study applied the choice experiment (CE) (The estimation of economic values for public services is carried out in diverse environmental fields (e.g., [30,31]), and CVM and CE have been regarded as typical valuation methods. Among the several econometric methods, the CE designed by Adamowicz et al. [32] has the advantage of subdividing the value of the estimated object into main attributes. Moreover, progressive values can be estimated by phases from the lowest to the highest level. The bundle of alternatives combined by each level of functionality is presented to respondents, after which the most preferred alternative (including a price level) is selected (this can be calculated as the values for each level). In particular, where the effects of water resources development plan vary, CE can be cost-effective by enhancing the feasibility of policy alternatives.) methodology for valuation, which has the advantage of separating the attributes that affect the value of a certain good by level and estimating the value of each level [33].

The study concluded that three dam functions are of high importance through several key pieces of evidence, and verifying them is crucial for South Korea. The role of dams has been specified through three major attributes: DM, FC, and WM. Furthermore, the results of advanced studies (which suggest the perceived seriousness of climate change significantly affects policy support ratings) indicate that the higher the level of consideration in climate change, the greater the likelihood of advocating for the enhancement of dam functions [22–24]. Accordingly, it is expected that people's awareness of climate change may cause meaningful differences in the economic values of the dam's public functions. Hence, the study conducts empirical analyses to achieve specific objectives as follows:

1. Estimate the economic value of Daecheong Dam by the subdivided attributes (DM, FC, and WM).

2. Examine the moderating effect of climate change perceptions on the economic values of the dam's public functions.

This study contributes to the literature by estimating the WTP of Daecheong Dam's functions and how they differ depending on climate change awareness. Taking the two previously mentioned objects into consideration, the following research questions are proposed.

Q1. Do the three attributes (DM, FC, WM) of Daecheong Dam have a significant impact on the increase in the utility of survey respondents?

Q2. Does the MWTP for the public interest function of Daecheong Dam differ depending on the degree of awareness of climate change among survey respondents?

#### **2. Materials and Methods**

#### *2.1. Study Area*

The Daecheong Dam basin (36◦28 33.3 N 127◦28 31.1 E) is 2608 km2, accounts for more than 1/4 of the total area of 9914 km<sup>2</sup> of the Geumgang River basin, and 10 administrative districts. It spans Daedeok-gu, Dong-gu, Yuseong-gu, Daejeon Metropolitan City; Cheongju-si, Boeun-gun, Okcheon-gun, Chungcheongbuk-do; Geumsan-gun, Chungcheongnam-do, Yeongdong-gun; Sangju-si, Gyeongsangbuk-do and Muju-gun, Jeollabuk-do [34] (see Figure 1). In the Daecheong Dam basin, there are 108,852 residents from across 43,140 households, and the population density was analyzed to be 163.08 people/km. The water supply rate in the Daecheong Dam basin was found to be about 86.0%, lower than Korea's total water supply rate, 96.5% (including village water supply, small water supply facility population) [35].

**Figure 1.** Location of the Daecheong Dam basin.

Construction for the Daecheong Dam began in March 1975, and it was completed in December 1980. It is a complex dam composed of a gravity-type concrete dam and a sand dam with a height of 72 m, a length of 495 m, a reservoir area of 72.8, and a volume of 1,234,000 m3. There exists the main dam with a storage capacity of 1.49 billion m3, and three auxiliary dams that prevent water in the reservoir from overflowing to other areas. In addition, there are hydroelectric power plants with a capacity of 90,000 kW and a water channel to supply water to some areas of the Chungcheong region.

#### *2.2. Literature Review*

There is a large body of scientific data linking climate change to hydrologic changes such as in precipitation, streamflow and evapotranspiration. Climate change's impact on the hydrologic cycle poses a severe threat to Korea, an area threatened by periodic floods and droughts. Climate change-induced increases in streamflow during the monsoon (a period of significant rainfall, typically May–September) have the potential to exacerbate flood damage, whereas increases in evaporative losses (due to warmer temperatures) during the dry period can exacerbate water scarcity in some areas [36].

Among the measures used to manage water resources in response to climate change, the operation of dams and the improvement of their functions has emerged as a major subject of interest. In the case of multipurpose dams, the importance of dam operation plans for flood control, drought management, and environmental functions is increasing as they are directly or indirectly related to the benefits of water for residents, such as water supply, hydroelectric power generation, and water quality improvement [37,38]. The public interest value of water resources by the dam function makes it difficult to clearly measure benefits, and the absence of a market has acted as a challenge in efficient resource distribution, making government intervention inevitable in water supply and distribution. For this reason, evaluating and proving the validity of non-market value for dam function and water resource use is considered a task that must be preceded in the process of controlling and managing it [39].

Because it is concerned with modeling options ranging over a variety of attributes rather than estimating WTP for a single option, the CE technique presents a potential chance to quantify the economic values of diverse environmental consequences induced by big dam development. The CE methodology, similar to the CV method's referendum model, has its theoretical grounding in the random utility model, which is compatible with economic theory [40–42].

The rationale for estimating the values of dam functions regarding DM, FC, and WM is sourced from various prior studies. First, value estimation studies resulting from drought mitigation are typically conducted in terms of drought relief for watershed protection [43], the willingness to pay (WTP) to avoid drought-water constraints for households and businesses [44], premium payments for agricultural insurance [45], and the value of avoiding drought water-usage restrictions [46].

Furthermore, in studies on flood-risk reduction values, empirical tests on nationwide flood control measures [47], flood risk reduction [48], flood insurance premiums for rural households [49], and the economic value and determinants of flood defenses [50] were explored. Moreover, regarding water quality values, various studies on the value of secure and reliable drinking water [51], the amount of payments to improve in-home water services [52], the value of water quality improvement and determinants that affect the value [53], and the quality improvement value of tap water for urban residents [54] have been carried out. In most of the previous studies mentioned, the contingent valuation method (CVM) and CE were used for measurement. It was also noted that individual characteristics such as gender, age, income level, education level, residential environment, government trust, and perceptions about disasters affected the value determination [45,47–51,53,54].

While a variety of economic valuation cases have been globally executed, it has been confirmed that there are few intermittent studies in South Korea. The precedent studies relevant to the three roles are as follows. Hwang et al. [55] estimated WTP to improve the future status of Korean water scarcity by households using CVM. Thus, Busan residents in Korea perceived water shortage, and about 70% of them were willing to pay. The average payment amount per household was about KRW 3572 (USD 4) per household per month. Choi and Lee [56] calculated home buyer contributions to flood prevention construction through the hedonic price method. According to the results, the buyer's WTP for a 1% reduction in rainfall intensity was KRW 62,101 per square meter, and the WTP for a 1% reduction in annual rainfall was KRW 36,533 per square meter.

Furthermore, Lee et al. [57] used CVM to evaluate the WTP for a future water shortage project in Korea, resulting in about 320 million dollars. They, however, concluded that the project cost was greater than the national utility. Kwak et al. [58] valued WTP for tap-water quality improvement in Busan, Korea, through CVM; the average amount per household was KRW 2124 per month. In addition, Um et al. [59] applied the averting behavior method to estimate WTP to reduce the negative perceptions caused by the discrepancy between the objective pollution level and perceived level. The results highlighted that perceived risk is more effective than objective risk, and the USD range of WTP were [0.07; 1.70] to [4.2; 6.1].

Moreover, these three functions act as major factors of dam operations according to an expert opinion survey that prioritizes the core properties for adapting to climate change. Furthermore, there is much emphasis on paying constant attention to comprehend the managerial importance of these factors [60]. So far, it is clear that CVM, CE, and the hedonic price method were frequently employed as value-estimation methods. CVM, which measured only the single value of the goods, was used most. In addition, the spatial and temporal features of the study site and the demographic characteristics of the study subjects had a significant effect on the estimation results.

#### *2.3. Setting Attributes and Levels*

Concerning the attributes from the previous studies, a content validity examination was further conducted. Thus, those three functions were selected as the final attributes based on carefully reviewed outcomes by experts (professors and senior researchers on environmentology, hydrology, and mineral economics). Focus group interviews (with ten regular people cognizant of Daecheong Dam) were, then, employed to determine specific levels of the attributes. In this respect, interviewees described the image associated with Daecheong Dam's climate change role. Accordingly, functional levels expressible in are cognizable manner were established. Information on techniques relevant to drought mitigation (such as sedimentation reduction and emergency drainage design), flood reduction (such as spillway design and dam raise), and water quality monitoring (such as the installation of devices for reducing non-point sources and sewage treatment facility expansion) was given to the interviewees in advance.

At the end of the discussion, the decision was that it is too restrictive to manifest the diffusion of specific technologies at a certain level. Thus, it was desirable to describe the attribute levels as complementing overall current technologies and creating new crafts beyond the present structure. Subsequently, each attribute level is classified into three phases: low-level (to maintain the status quo), medium-level (to complement existing technologies), and high-level (to develop new technologies along with the complementation). These demonstrate utilities calculated as per the increase in the improvement levels. In addition, a preliminary test for 30 respondents regarding WTP was conducted using open-ended questions to determine appropriate bid levels along with a realistic payment vehicle. Thus, via the focus group interviews and the reviewed attributes [61–63], it was determined that three asking prices of KRW 5000, KRW 10,000, and KRW 20,000 within the range of 15% to 82% of the response distribution [64] should be the annual financial support. The levels are shown in Table 1.


**Table 1.** Description of attribute levels.

#### *2.4. Development of a Measurement Instrument*

The survey questionnaire was composed of demographic items (gender, age, marriage status, education, household income, and resident area), climate change perception index, and CE elements. Choice sets are first structured based on the derived attributes and levels to develop the measurement tool for CE. The procedures are as follows. Since the three attributes of the dam's public benefit function and the annual financial support, respectively include three levels, a total of 81 alternatives exist (3 raised to the 4th power). The study employed a more efficient experimental design using the SAS orthogonal design program because it is an unrealistic field survey that requires responses to all of the alternatives. Thirty-four optimal profiles were extracted, and 18 choice sets were derived from each choice profile involving two optional alternatives along with a reference alternative. Furthermore, the results are confirmed to be statistically significant due to superiority in terms of efficiency and error (D-efficiency = 2.08; D-error = 0.48) [65]. Presenting a set of 18 optional alternatives to one respondent may increase non-sampling errors. In this study, after dividing the entire survey questionnaires into Type A/B/C, six sets of optional alternatives were assigned to each type.

However, if one respondent evaluates all 18 choice sets at once, the response validity might be impacted. Thus, the sample was divided into three blocks to enhance the response validity. Each of the three questionnaire types contained six choice sets. Figure 2 below shows one of the 18 choice sets. The respondents evaluated the choice sets composed of each level of the dam's public benefit function and the annual financial support. Then, they selected the most preferred alternative among two options for further improvement along with one "no-choice" option. Here, the level of each attribute in the "no-choice" option is low (i.e., status quo), and the annual financial support is designated as KRW 0. Respondents choose the most preferred alternative among the three options after reading the contents of the current technology level described in the questionnaire introduction.

Prior to the analysis, Option 3. "Choosing neither option" (see Figure 2) in the questionnaire indicates not selecting any of the two improvement alternatives, implying that the current condition would be maintained. Therefore, the willingness to pay financial support is calculated as KRW 0, but the water expense is still maintained.

**Figure 2.** An example of choice set.

#### *2.5. Sample Collection*

The study population comprised adults aged 20 and older who, directly and indirectly, benefit from various water supply and hydroelectric power generation of Daecheong Dam. For this reason, residents who were aware of Daecheong Dam and living in Daejeon City or its surrounding areas such as Chungbuk and Chungnam were selected as participants. The samples were selected according to gender and age-group properly represented the population (purposive quota sampling). Before the main survey, we conducted a pre-test to check whether the content, arrangement, and phrasing of the items were clear. We completed the final questionnaire by correcting and complementing the questions. A total of 630 questionnaires (210 copies for each type) were distributed under the household unit of analysis by direct face-to-face street-intercept interviews around Geum River, LOHAS Park, near Daecheong Dam. After screening, 603 valid questionnaires were employed for data analyses.

#### *2.6. Analytical Method*

The analytical model of CE is based on the indirect utility function theoretically implied in economics. The function *Uij* in Equation (1) indicates the indirect utility of any individual *i* (= 1, . . . , n), which can be obtained from an alternative *j* (= 1, . . . , *J*) among a choice set *Ci*.

$$\mathcal{U}\_{\vec{ij}} = V\_{\vec{ij}} \left( Z\_{\vec{ij}\prime} S\_i \right) + \varepsilon\_{\vec{ij}} \tag{1}$$

Here, *Vij* accounts for the attribute functions of the alternative (*Zij*) and the individual characteristics (*Si*) of the respondent as the observable elements. In addition, *eij* means unobservable errors, which are relevant to the theoretical foundation for composing the likelihood function.

In the CE analysis, the discrete choice model is applied. If the *j*th alternative of the choice set *Ci* chosen by the respondent *i* generates a greater utility than another alternative *k* [*Uij* > *Uik* (*k* ∈ *Ci*, *k* = *j*)], it is logically clear from the above that the alternative *j* should be chosen. Thus, in Equation (2), the probability of respondent *i* choosing alternative *j* can be written as:

$$\Pr(j \mid \mathbb{C}\_{i}) = \Pr\left(V\_{i\bar{j}} + \varepsilon\_{i\bar{j}} > V\_{i\bar{k}} + \varepsilon\_{i\bar{k}}\right) = \Pr\left(V\_{i\bar{j}} - V\_{i\bar{k}} > \varepsilon\_{i\bar{k}} - \varepsilon\_{i\bar{j}}\right) \tag{2}$$

In the case of estimating the multi-nominal logit model described in Equation (2), if the assumption about the error term independence is satisfied (per the Type I extreme value distribution), then the probability of the respondent *i* selecting alternative *j* is given by Equation (3).

$$P\_{\bar{i}}(j|\mathbb{C}\_{i}) = \frac{\exp\left(V\_{\bar{i}j}\right)}{\sum\_{k \in \mathbb{C}\_{i}} \exp\left(V\_{ik}\right)}\tag{3}$$

The multi-nominal responses derived from the CE questionnaire represent the outcomes where the individuals pursue utility maximization. This is analyzed through the likelihood function in Equation (4).

$$\ln L = \sum\_{i=1}^{n} \sum\_{j=1}^{I} \left\{ \begin{array}{l} \boldsymbol{Y}\_{ij} \cdot \ln[\boldsymbol{Pr}\_i \ (j|\mathbb{C})] \end{array} \right\} \tag{4}$$

In this case, the respondent may or may not select alternative *j*, where the variable *Yij* = 1 indicates that the *i*th respondent has chosen the alternative *j*. Here, 1(·) denotes the indicator function, and "1" is assigned in 1(·) when the *j*th alternative is selected; otherwise, 0 is granted. Hence, the parameters can be calculated by applying the method of maximum likelihood estimation to the log-likelihood function of Equation (4) [66].

The indirect utility function *Vij* of this study can be described as the linear function of observable attribute vectors: an alternative specific constant (*ASC*), medium level (*DMMid*) and high level (*DMHigh*) for the DM function improvement, medium level (*FCMid*) and high level (*FCHigh*) for the FC function improvement, medium level (*WMMid*) and high level (*WMHigh*) for the WM function improvement, and financial support (*Bid*) as shown in Equation (5). *β* is an estimated parameter that affects the utility.

$$V\_{i\bar{\eta}} = A\mathcal{S}\mathcal{C} + \beta\_1 \mathcal{D}M\_{\text{Mid},\bar{\eta}} + \beta\_2 \mathcal{D}M\_{\text{Hijgh},\bar{\eta}} + \beta\_3 \mathcal{F}\mathcal{C}\_{\text{Mid},\bar{\eta}} + \beta\_4 \mathcal{F}\mathcal{C}\_{\text{Hijgh},\bar{\eta}} + \beta\_5 \mathcal{W}M\_{\text{Mid},\bar{\eta}} + \dots \tag{5}$$

$$\beta\_6 \mathcal{W}M\_{\text{Hijgh},\bar{\eta}} + \beta\_7 \mathcal{B}\dot{d}\_{\bar{\eta}}$$

Moreover, an extended model into which demographic variables are additionally inserted is estimated for detailed examinations. The model is structured as in Equation (6):

$$\begin{aligned} V\_{\text{ij}} &= A\mathcal{S}\mathcal{C} + \beta\_1 D M\_{\text{Mid},\text{ij}} + \beta\_2 D M\_{\text{H}\text{j}\text{h},\text{ij}} + \beta\_3 F \mathcal{C}\_{\text{Mid},\text{ij}} \\ + \beta\_4 F \mathcal{C}\_{\text{H}\text{j}\text{h},\text{ij}} &+ \beta\_5 W M\_{\text{Mid},\text{ij}} + \beta\_6 W M\_{\text{H}\text{j}\text{h},\text{ij}} + \beta\_7 B \mathcal{i}d\_{\text{ij}} + \sum\_{\text{s=1}}^{\mathcal{S}} \gamma\_{\text{s}} K\_{\text{s}\text{i}} \end{aligned} \tag{6}$$

where *Ksi* is the vector representing the individual characteristics of the *i*th respondent, *s* (= 1, . . . , *S*) is the demographic variable, and *φ* is an estimate of the interaction variables.

Thus, the marginal willingness-to-pay (MWTP) for the attributes can be estimated by Equations (5) and (6), which demonstrate the marginal rate of substitution (The marginal rate of substitution can be defined as the quantity of one good to be discarded to obtain another [67], that is, respondents have to pay more for a higher level of improvement.) between each level of the attributes and the price variable. Therefore, the MWTP, owing to the vector variation of each attribute, can be estimated as the coefficient ratio of the corresponding level to the price variable as shown in Equation (7).

$$\begin{aligned} MWTP\_{DM\_{Mid}} &= \partial B \text{id} / \partial DM\_{Mid} = -\mathfrak{f}\_1 / \mathfrak{f}\_7 \\ &\vdots \\ MWTP\_{WM\_{H\_{H\_{H\_{H\_{H}}}}}} &= \partial B \text{id} / \partial DM\_{H\_{H\_{H}}} = -\mathfrak{f}\_6 / \mathfrak{f}\_7 \end{aligned} \tag{7}$$

This study employed a climate change perceptions index, proposed by the Korea Energy Management Corporation [68], to measure the climate change level cognized by the respondents. An R-type explanatory factor analysis (EFA), based on principal components, corroborated the measurement item validity. In the factor extraction process, only items higher than eigen value 1.0 were factorized with a loading of more than 0.4. To measure the reliability of measurement tools, an internal consistency technique using Cronbach's Alpha Coefficient was applied. If the value of the Cronbach's alpha coefficient is 0.6 or more, the

reliability can be considered valid, and the entire items can be analyzed by synthesizing them on a single scale.
