Public Perspective on Increasing Renewable Energy Use Ratio in Public Buildings in South Korea

Abstract

The South Korean government plans to increase the share of renewable energy (RE) used in public buildings by 10% from the current 30% to 40% by 2030. This article seeks to estimate the public willingness to pay (WTP) for this increase. To this end, a contingent valuation was applied, with 1000 households randomly selected and surveyed through one-on-one interviews. The payment vehicle and WTP elicitation method were determined to be income tax per household and the one-and-one-half-bound model, respectively. The annual WTP per household was estimated to be KRW 2712 (USD 2.04) with statistical significance. When expanded to the population, this produces an annual value of KRW 60.15 billion (USD 45.23 million). The increase in the RE use share can not only reduce greenhouse gas emissions but also result in savings on electricity bills. The sum of these two can be considered as benefits, and the sum of the construction and maintenance costs incurred due to the increase can be considered as costs. The cost–benefit analysis indicates that the present value of net benefits and the benefit-to-cost ratio were estimated to be KRW 667.3 billion (USD 501.7 million) and 1.48, respectively. Consequently, the increase is socially desirable and should be implemented immediately.

1. Introduction

In December 2021, the South Korean government submitted its 2030 nationally determined contribution (NDC) to the United Nations Framework Convention on Climate Change Secretariat, which aims to reduce greenhouse gas (GHG) emissions by 40% in 2030 compared to 2018. In 2018, South Korea’s GHG emissions were approximately 730 million tons, of which the building sector accounted for 7.2%, a relatively small share [1]. However, in order to achieve the reduction target in the 2030 NDC, the government stated in the 2030 NDC that GHG emissions in the building sector should be reduced by 32.8%, from 52.1 million tons in 2018 to 35.0 million tons in 2030.

There are two main ways to abate GHG emissions in the building sector [1]. The first way is to reduce energy use in buildings by expanding the supply of high-efficiency devices and using smart energy management systems. Using high-performance insulation materials and high-efficiency appliances can improve the energy efficiency of buildings and reduce energy use. The second way is to install renewable energy (RE) facilities such as solar power and geothermal power in buildings to directly use the electricity produced from RE [2,3,4,5,6,7,8,9]. The government is implementing a policy to support part of the cost of installing solar power facilities for households.

In addition, when constructing, expanding, or remodeling public buildings with a floor area of 1000 m² or more, it has been legally mandatory to self-produce electricity by installing RE facilities for at least 10% of the building’s total energy consumption since 2011. This figure has increased from 10% to 30% since 2020, and is expected to increase to 40% by 2030. According to the Korean Law Information Center, public buildings refer to buildings such as city halls, police stations, and health centers that are owned and occupied by the central government, local governments, and government-affiliated organizations for the purpose of performing their duties. In order to increase the use of RE power in public buildings, the utilization rate of existing RE facilities must be increased or new RE facilities must be installed in the buildings. Ultimately, the expansion of the use of RE power will increase the cost of constructing public buildings, and the additional cost will be covered by the people’s taxes.

When the costs and benefits of expanding the use of electricity generated from RE in public buildings are compared, the latter must be greater than the former in order for investing the additional cost to be socially justified. The sum of construction costs and maintenance costs corresponds to the costs, and the sum of internal benefits and external benefits corresponds to the benefits. Internal benefits, which can be considered market value, can be defined as the reduction in operating costs, such as the reduction in electricity bills due to improved energy efficiency in public buildings. External benefits refer to non-market environmental benefits resulting from the reduction in GHG and/or air pollution emissions.

Since environmental goods that generate external benefits are not usually traded in the market, the estimation of external benefits is somewhat complicated. Therefore, the application of a specially designed economic technique is required to estimate the external benefits. The most widely used economic method for this purpose is the stated preference approach. In this study, the contingent valuation (CV) method is adopted to estimate the external benefits resulting from the increased use of RE in public buildings. When CV is applied, information on the willingness to pay (WTP), which is the amount that consumers are willing to pay to consume a specific non-market good while reducing their consumption of other goods, is derived from the analysis of the data collected through the CV survey. From an economic perspective, the WTP is the external benefit mentioned above.

The fundamental hypothesis explored in this study is that an increase in the share of RE use in public buildings in South Korea is economically justified. To test this hypothesis, the economic feasibility of such an expansion is examined within the rigorous framework of cost–benefit analysis (CBA). The economic benefits are conceptually divided into two primary categories: internal benefits and external benefits. Internal benefits are represented by the reduction in electricity expenditure resulting from the self-generation of electricity through RE facilities, which offsets the need to purchase electricity from the grid. On the other hand, estimating external benefits poses a more complex challenge, as these encompass non-market environmental gains such as reductions in GHG emissions and air pollutants. To address this complexity, the CV method is employed. CV involves collecting data via carefully designed surveys, and applying economic theory coupled with econometric modeling to rigorously estimate individuals’ WTP for non-market goods. Consequently, the ultimate conclusion of this research hinges on whether the planned expansion of RE usage in public buildings can be deemed economically viable based on the combined evaluation of internal and external benefits relative to associated costs.

In order to estimate the external benefits arising from increasing the share of electricity generated from RE being used in public buildings, this article aims to estimate the public WTP for such an increase by applying CV. Furthermore, the article aims to examine the economic feasibility of the increase by using not only the estimated WTP but also other related quantitative information. To the best of the authors’ knowledge, this is the first such attempt in the literature. This can increase the utility of the article. The remainder of the article is organized as follows. Section 2 deals with the research methodology, and specifically examines CV, the preparation of the questionnaire, the conduct of the survey, and the analysis of the CV data, in that order. Section 3 presents and discusses the results from analyzing the data obtained from the CV survey. Section 4 contains the conclusions.

2. Materials and Methods

2.1. Method: CV

This research applies CV to estimate the external benefits ensuing from increasing the use of RE-generated electricity in public buildings. CV has been widely used to evaluate the economic value arising from the consumption of a good that is not traded in the market [10,11,12,13]. The CV method was selected for this study as it offers a well-established and rigorous framework for eliciting individuals’ WTP for non-market environmental goods, such as increased RE usage in public buildings. CV is particularly suited for valuing goods and services that lack market prices, enabling the direct assessment of consumers’ monetary preferences through hypothetical scenarios. This approach facilitates the estimation of both use and non-use values, capturing broad societal benefits beyond immediate market transactions. Furthermore, CV’s flexibility allows for careful survey design tailored to the specific context of RE policy, enhancing relevance and informational content for decision-making. The method’s widespread application in environmental and energy economics attests to its robustness and acceptance, supporting its use to generate reliable and policy-relevant benefit estimates in this study.

While alternative non-market valuation techniques, such as revealed preference (RP) methods and choice experiments (CE), offer valuable insights, they present limitations which motivated the preference for CV in this context. RP methods rely on observed market behaviors, which can be restrictive when valuing new or hypothetical policy scenarios like the proposed increase in RE usage in public buildings, where market data are unavailable or insufficient. CE methods, though capable of capturing trade-offs among multiple attributes and offering richer preference information, often require more complex survey designs and sophisticated econometric analyses, which can increase respondent burden and complicate interpretation. Additionally, CE typically estimates relative preferences rather than a direct monetary value, necessitating further modeling steps. Given the study’s goal to obtain straightforward and interpretable monetary measures of willingness to pay for an incremental policy change, CV provided the most appropriate balance of practicality, methodological rigor, and policy relevance.

CV has also been commonly adopted in studies related to energy use in the building sector. For example, Zalejska-Jonsson [14] derived the WTP of residents of eco-friendly and general apartments for green-certified buildings in Sweden. Robinson et al. [15] found that tenants of green buildings in the United States were willing to pay a 9.3% premium for green features that provided a pleasant living environment.

Kim et al. [16] investigated consumers’ WTP for zero-energy apartments that minimize energy use in buildings and produce electricity through RE facilities, thereby eliminating the use of fossil fuel-generated electricity. Encinas et al. [17] found that new homebuyers in Santiago, Chile, had a WTP of EUR 3900 to 8400 for energy-efficient homes. Ji et al. [13] used CV to investigate the public’s WTP for green roofs constructed to increase energy efficiency and reduce noise in buildings. Takuh et al. [18] adopted CV to analyze the WTP of middle-income households for the construction of green residential buildings in Makurdi, Nigeria.

Therefore, the application of CV in this article to collect data on WTP for increasing the utilization rate of RE in public buildings targeting the South Korean public is consistent with previous studies. In addition, no previous studies directly related to this research were found. Since CV uses a survey to collect data, whether their reliability and validity are satisfied or not is an important issue [19,20]. This is because the CV technique using a survey must satisfy the conditions of reliability and validity. The main conclusion from several previous studies is that both reliability and validity are secured in CV [21,22,23,24,25,26,27,28].

The most important starting point in applying CV is to clearly define the status quo state and the target state. In other words, what is evaluated in CV is not the target state, but the change from the status quo state to the target state. The status quo state means that the mandatory proportion of electricity produced by self-generation using RE in the total electricity used in public buildings is 30% as of 2030. On the other hand, the target state means that this proportion should be 40% as of 2030. In other words, the evaluation target of this study is to increase the RE electricity usage ratio in public buildings by 10%p. The procedures for applying CV are largely composed of survey design, sampling and conducting the survey, and a statistical analysis of the CV data. Each procedure will be described in detail below.

2.2. CV Survey Design

To obtain reliable survey data, the CV questionnaire must be carefully designed. The authors developed the questionnaire following several guidelines for CV surveys suggested in previous studies [29,30,31,32]. The final version of the CV questionnaire consists of three parts. The first part details the government’s plan to increase the use of RE in public buildings. It also asks respondents about their opinions about this plan. Through this part, the respondents are gradually immersed in the hypothetical market that the authors wanted to create.

The second part elicits WTP responses from the respondents. For this, five points must be decided in advance. The first point is that the unit of the survey must be either the individual or the household. In this regard, the use of household units was decided. In addition, respondents were limited to the head of the household or the spouse of the head of the household. The second point is the decision about the choice of respondents. In this study, people aged 20 to 65, which corresponds to the economically active population, were selected as respondents. This was to elicit responsible answers since the question about WTP was given to them.

The third point is the determination of the payment vehicle. It should be related to the object being evaluated and familiar to the respondents. A well-chosen payment vehicle can prevent strategic behavior of respondents while overcoming the hypothetical bias that may arise from the setting of a hypothetical market. In this regard, the annual income tax per household was adopted. As mentioned above, since the survey unit for this study was the household, not the individual, the income tax per household was adopted. Income tax is one of the representative national taxes in South Korea and will actually be invested in the construction of RE facilities in public buildings. The fourth point is the determination of the payment period. It was set to be the next ten years.

The fifth point is the determination of the method of eliciting WTP responses. The method falls into one of two categories: open-ended question and closed-ended question methods. The open-ended question method causes respondents to report WTP as a continuous value. This method is not often used because it causes a non-response bias or causes them to report unreasonable protest WTPs [21]. In addition, it can cause strategic behavior of respondents. This study used the closed-end question method, which allows respondents to choose one of several options. More specifically, among the various models corresponding to the closed-ended question method, the dichotomous choice (DC) model was applied.

The DC model presents a certain amount to the respondent and asks her or him to choose whether to pay or not, that is, “yes” or “no.” Since the respondent responds “yes” if her or his WTP is greater than or equal to the amount and “no” otherwise, the DC model is incentive-compatible. In particular, it is easier for the respondent to answer than the open-ended question method and can reduce the protest WTPs [29]. Among the various DC models, this study applies the one-and-one-half-bound (1.5B) model given by Cooper et al. [33]. This model has the advantage of improving on the statistical inefficiency of the single-bound (1.0B)-model that asks only one DC question while reducing the response effect of the multiple-bound-model that asks two or more DC questions.

The third part of the questionnaire includes questions about the respondent’s socioeconomic characteristics and mobile phone number. For example, data on whether the person is the head of the household, the number of family members, age, gender, education level, and average monthly household income were collected. The collected data were used in the covariate model described later. The written mobile phone number was used by a supervisor from a professional survey company to contact the respondent and check whether the interviewer conducted the survey properly. Furthermore, a blank space was provided at the end, and respondents were asked to freely write down any comments they had about the survey.

2.3. Sampling and Conduction of the CV Survey

Random sampling was conducted in sixteen provinces across the country, excluding Jeju Island, based on the census data provided by the National Statistical Office in 2020 for data collection. In a CV survey, the sample must represent the population adequately. Therefore, the random sampling was carried out by a professional survey company with extensive experience in CV surveys to ensure the accuracy of the survey and reflect the characteristics shown in the census data. The questionnaire was initially reviewed by a focus group of 30 individuals and was subsequently revised with the help of the company’s supervisors. Through this process, ambiguous expressions in the questionnaire were refined, and its readability was improved. The sample size was set to be 1000 based on the recommendations of Arrow et al. [29] and the Korea Development Institute [34].

This survey employed the one-on-one individual interview method among the survey methods of telephone, mail, and Internet. This method, while costly, yields reliable results as well-trained interviewers provide respondents with sufficient information. Supervisors from the survey company contacted the respondents to ensure that the interviews were conducted appropriately. During this process, some questionnaires were deemed to have incomplete responses and were excluded, leading to additional surveys being conducted. The interviewers reported that most respondents answered the WTP questions consistently and without difficulty. As a result, this research was able to gather data from a CV survey conducted in March 2024, targeting 1000 households nationwide.

2.4. Statistical Model of the CV Data

To apply the 1.5B model, two values, consisting of a high amount and a low amount, must be determined first. The entirety of the respondents were divided into two groups, one confronted with the high amount and the other with the low amount. If a respondent presented with the high amount answers “yes,” the WTP question ends. However, if she/he answer “no,” she/he is asked about her/his intention of paying the low amount. At this time, one who answers “no” is asked once more whether their WTP is 0. This allows for a distinction between a small positive WTP and a WTP of 0. Conversely, when the low amount is presented first, if the respondent answers “yes,” they will be asked about their intention of paying the high amount. If they answer “no,” they will be asked whether their WTP is 0, allowing for the identification of a WTP of 0.

When analyzing the collected data, it is necessary to apply a model that can allow for a WTP of 0. This research used the spike model, which is a widely used model that can reflect a WTP of 0 in dealing with DC CV data [35,36]. Thus, the spike model and the 1.5B model are combined in this paper. The spike model treats the probability of having a WTP of 0 as a spike in the cumulative distribution function (cdf) of the WTP. The response to the DC question is either “yes” or “no.” Let the presented amount and WTP be denoted by $R$ and $F$, respectively. The probability of responding “yes” to the presented amount can be expressed as:

$$\left\{\begin{array}{c}\mathrm{Pr}\left(\text{“}\mathrm{yes}\text{”}\right) = \mathrm{Pr}\left(F\ge R\right) = 1-G_F\left(R;s_0,s_1\right)\\ \mathrm{Pr}\left(\text{“}\mathrm{no}\text{”}\right) = \mathrm{Pr}\left(F<R\right) = G_F\left(R;s_0,s_1\right) \end{array}\right.$$

(1)

where $G_F\left(·\right)$ is the cdf of $F$, and $s_0$ and $s_1$ are the parameters of the cdf. In general, the logistic distribution is used when dealing with the distribution of WTP, and the specific form of $G_F\left(·\right)$ is:

$$G_F\left(R;s_0,s_1\right) = \left\{\begin{array}{c}{[1+\mathrm{e}\mathrm{x}\mathrm{p}\left(s_0-s_{1}R\right)]}^{-1} \mathrm{if} R > 0\\ {[1+\exp \left(s_0\right)]}^{-1} \mathrm{if} R = 0 \\ 0 \mathrm{if} R < 0 \end{array}\right.$$

(2)

where the cdf is defined over the entire real number domain, but the functional form changes depending on the interval of $R$. Spike is defined as ${[1+\exp \left(s_0\right)]}^{-1}$. Moreover, the average WTP formula is calculated as $\left(1/s_1\right)\ln [1+\exp \left(s_0\right)]$, using Equation (2) and the average formula using the cdf. The model presented so far does not include covariates. The process of examining the effect of covariates on the likelihood of answering “yes” is useful to verify internal consistency or theoretical validity. Accordingly, the authors intend to conduct a welfare analysis based on the estimation results of the model that does not include covariates to investigate how covariates affect the probability of responding “yes” to the proposed amount. For this, $s_0$ should be replaced with $s_0+x\prime \beta$ where $x$ is a covariate vector reflecting the socioeconomic characteristics of the respondents and $\beta$ is a vector of parameters to be estimated.

3. Results and Discussion

3.1. Results

3.1.1. Data

To reduce biases commonly associated with CV methods, this study implemented several measures. First, to minimize strategic bias, a DC elicitation format was employed. Since the DC approach is incentive-compatible, it substantially curtails strategic misreporting by respondents. Second, to address information bias, visual aids were utilized to enhance respondents’ understanding of the valuation good, while cost information was deliberately withheld to avoid cost-related bias. Third, to prevent ordering bias, the sets of bid amounts were randomly assigned to respondents. Fourth, to mitigate potential response effects between initial and follow-up answers, the 1.5B model was applied instead of a traditional double-bound framework. These methodological precautions collectively contribute to enhancing the reliability and validity of the WTP estimates reported in this study.

The presented amounts and the results of the responses to them are summarized in

Table 1. Number of answers for each set of bids in the sample.

Bids a		Number of Answers
First	Second	“yes-yes”	“yes-no”	“no-yes”	“no-no”	Totals
1000	3000	19	12	7	34	72
2000	4000	6	12	7	47	72
3000	6000	6	12	6	47	71
4000	8000	5	11	9	46	71
6000	10,000	4	10	8	49	71
8000	12,000	5	8	3	55	71
10,000	15,000	6	6	18	42	72
	Totals	51	71	58	320	500
First	Second	“yes”	“no-yes”	“no-no-yes”	“no-no-no”	Totals
3000	1000	17	9	5	40	71
4000	2000	14	4	9	45	72
6000	3000	12	2	8	50	72
8000	4000	11	9	5	46	71
10,000	6000	8	6	8	49	71
12,000	8000	6	5	12	48	71
15,000	10,000	10	2	14	46	72
	Totals	78	37	61	324	500

^a The unit is Korean won (USD 1.0 = KRW 1330 at the time of the survey).

. The main contents of Table 1 are also presented in Figure 1 to aid visual understanding. The presented amounts range from KRW 1000 to 15,000. There were classified into a total of seven sets. The upper part of Table 1 presents the survey results from the questionnaire where lower amounts were offered first, while the lower part of the table shows the response results where higher amounts were presented first. The survey results indicate that out of a total of 1000 households, 644 households reported a WTP of 0, meaning that more than half, or 64.4%, of the respondents indicated they have no intention of paying for the increased use of electricity produced by RE in public buildings. Additionally, it can be observed that as the presented amounts increase, the percentage of “yes” responses decreases.

3.1.2. Estimation Results of the Model

Responses obtained from the 1.5B model may violate procedural invariance, necessitating further analysis [37]. The cause of the violation is the response effect. When applying the 1.5B model, the response effect occurs when the first and second responses come from different WTP distributions, even though they should come from the same WTP distribution. This research aims to analyze the 1.0B model using only the first response obtained from the 1.5B question and to check the procedural invariance of the WTP responses by comparing the estimation results of the two models. The estimation results of the two models using the maximum likelihood method are presented in

Table 2. Results from estimating the models.

Variables	One-and-One-Half-Bound Model d	Single-Bound Model d
Constant	−0.6006 (−9.14) #	−0.6036 (−9.17) #
Bid amount a	−0.1612 (−16.16) #	−0.1322 (−13.13) #
Spike	0.6458 (42.95) #	0.6465 (42.98) #
Average of yearly
household willingness to pay	KRW 2712 (USD 2.04)	KRW 3299 (USD 2.48)
t-values	13.72 #	11.81 #
95% CI b	KRW 2355 to 3153	KRW 2822 to 3913
	(USD 1.77 to 2.37)	(USD 2.12 to 2.94)
90% CI b	KRW 2406 to 3075	KRW 2893 to 3810
	(USD 1.81 to 2.31)	(USD 2.18 to 2.86)
Log-likelihood	−1054.65	−888.92
Wald statistics (p-values) c	188.18 (0.000)	139.56 (0.000)
Sample size	1000	1000

^a The unit is KRW 1000 (USD 0.75). ^b It means confidence interval computed using the method given in Krinsky and Robb [38]. ^c The null hypothesis is that the model is mis-specified. ^d The values shown in the parentheses next to the coefficient estimates are t-values. ^# implies that the estimate holds statistical significance at the 5% level.

.

The estimated coefficient for the presented amount is shown with a negative sign, which indicates that as the presented amount increases, the response rate of “yes” decreases. The spike values estimated from the 1.5B model and the 1.0B model are 0.6458 and 0.6465, respectively. These two values are close to the proportion of respondents with a WTP of 0, which is 64.4%, suggesting that the use of the spike model is appropriate. According to the estimation results of the Wald statistic, the null hypothesis stating that “The estimated model is meaningless” is rejected at the 5% level. Additionally, all estimated coefficients achieve statistical significance at the 5% level.

The annual average WTP per household estimated through the 1.5B model for expanding the usage rate of electricity produced by RE in public buildings is KRW 2712 (USD 2.04). Through the 1.0B model the result is KRW 3299 (USD 2.48). The t-values for the two WTP estimates calculated through the 1.5B and the 1.0B models are 13.72 and 11.81, respectively. Both results achieve statistical significance at the 5% level, but there is a difference of about KRW 533 (USD 0.40). The significance of this difference can be assessed using the confidence interval estimation technique presented by Krinsky and Robb [38]. Although there is a difference in the average WTP between the two models, it is difficult to conclude that the estimation results are different because the 95% confidence intervals overlap. As a result, it can be determined that the response effect obtained by applying the 1.5B model is not significant at the 5% level.

3.1.3. Estimation Results of the Model with Covariates

The definitions and explanations of the variables used in the covariate model for this study are presented in

Table 3. Description of variables used in the model.

Variables	Definitions	Mean	Standard Deviation
Education	The respondent’s education level in years	14.24	2.13
Knowledge	Dummy for the respondent knowing about increasing renewable energy consumption in government buildings before the survey (0 = no; 1 = yes)	0.16	0.36
Gender	The respondent’s gender (0 = male; 1 = female)	0.50	0.50
Head	Whether the respondent is the head of household (1 = head; 0 = otherwise)	0.54	0.50
Age	The respondent’s age (0 = younger than or equal to forty-three; 1 = older than forty-three)	0.66	0.47
Income	The respondent household’s monthly income (unit: million Korean won)	4.88	1.98

. The estimation results of those models are presented in

Table 4. Results from estimating the models with covariates.

Variables a	One-and-One-Half-Bound Model e	Single-Bound Model e
Constant	−0.1919 (−0.20)	−0.1276 (−0.13)
Bid amount b	−0.1680 (−16.60) #	−0.1389 (−13.83) #
Education	0.0448 (1.19)	0.0479 (1.31)
Knowledge	−0.8519 (−4.90) #	−0.8394 (−4.80) #
Gender	0.7919 (3.15) #	0.7769 (2.00) #
Head	0.8443 (3.32) #	0.8149 (2.09) #
Age	−0.5379 (−3.61) #	−0.5189 (−3.37) #
Income	0.1099 (2.88) #	0.1073 (3.00) #
Spike	0.6527 (42.27) #	0.6532 (42.29) #
Average of yearly household willingness to pay	KRW 2532 (USD 1.91)	KRW 3054 (USD 2.30)
t-values	13.60 #	11.78 #
95% CI c	KRW 2210 to 2950	KRW 2604 to 3646
	(USD 1.66 to 2.22)	(USD 1.96 to 2.74)
90% CI c	KRW 2258 to 2874	KRW 2674 to 3539
	(USD 1.70 to 2.16)	(USD 2.01 to 2.66)
Wald statistics (p-values) d	184.89 (0.000)	138.82 (0.000)
Log-likelihood	−1023.44	−858.67
Sample size	1000	1000

^a They are described in Table 3. ^b The unit is KRW 1000 (USD 0.75). ^c It means confidence interval computed using the method given in Krinsky and Robb [38]. ^d The null hypothesis is that the model is mis-specified. ^e The values shown in the parentheses next to the coefficient estimates are t-values. ^# implies that the estimate holds statistical significance at the 5% level.

. The estimated coefficients of all variables, except for the Education variable, are statistically significant at the 5% level. If the estimated coefficient has a positive sign, the value of the corresponding covariate increases, and the probability of responding “yes” to the presented amount also increases. If the sign is negative, it can be concluded that as the value of the corresponding covariate increases, that probability decreases. In particular, the coefficient for the presented amount was estimated as negative, indicating that as the presented amount increases, the probability of the respondents’ saying “yes” decreases.

More specifically, among the six covariates, the estimated coefficients for Gender, Head, and Income have positive signs. This indicates that these three variables have a positive effect on the probability that respondents will answer “yes” to the presented amount. For example, respondents with higher income are more likely to answer “yes” to the given amount. These three coefficient estimates achieve statistical significance at the 5% level. However, the estimated model that includes covariates may yield different average WTP estimates depending on which covariates are selected. Therefore, the following analysis will be based on the estimation results from the 1.5B model without considering covariates.

3.2. Discussion of the Results

The results of this study will be discussed from three aspects: the determination of how to handle the responses indicating a WTP of zero, the calculation of the total external benefits, and the implementation of a CBA based on total benefits that takes into account internal benefits.

3.2.1. Determination of How to Handle the Responses Indicating a WTP of Zero

It is necessary to derive implications from the analysis of why many respondents indicated a WTP of zero. Out of 1000 respondents, a total of 644 reported a WTP of zero. Among those 644, 39.6% responded that “The increased use of RE in public buildings should be pursued with taxes already paid.” Next, 14.4% of the 644 stated that “The increased use of RE in public buildings holds no value for me.” The former can be seen as a form of protest bid, while the latter can be considered a true WTP of zero. It is reasonable to handle true zero WTP data with a spike model. However, how to address protest bids is an important issue in applied CV research.

In other words, one must choose between analyzing the remaining data by discarding the protest bids and analyzing the entire data while treating them as having a WTP of zero. This study chose the latter for two reasons. The first reason is that since protest bids are an important expression of intent, it is deemed preferable to include them in an appropriate manner in the dataset rather than discard them. The second reason is that protest bids can be regarded as a WTP of zero since they indicate a refusal to pay. If protest bids are excluded, the estimated average WTP would be greater than the value obtained in this study. Therefore, the approach adopted in this study reflects a more conservative perspective. Maintaining a conservative viewpoint in the CBA conducted in this study can help prevent inefficient investments of public financial resources.

Nevertheless, the authors conduct an additional robustness check by examining the potential implications of excluding protest bids.

Table 5. Distribution of reasons for unwillingness to pay.

Reasons for Unwillingness to Pay	Number of Responses
1. This should be covered by taxes that have already been paid	255
2. Sufficient information has not been provided to make a judgment	68
3. This issue is not important enough to warrant prioritization	84
4. The government has already spent too much money on the plan	17
5. This issue is not a matter of concern for our household	87
6. The mandatory increase policy has little value to me	93
7. Our household does not have the financial capacity to pay	9
8. No additional tax revenue will be allocated to this plan	17
9. Others	14
Totals	644

reports the distribution of the stated reasons for unwillingness to pay. Among these, reasons 5 and 7 can be interpreted as representing genuine zero WTP, whereas all other reasons are regarded as protest bids responses. Based on this classification, of the 644 zero WTP responses, 548 are categorized as protest bids. When the analysis is restricted to the subsample consisting only of the 356 respondents who reported a positive WTP and the 96 respondents classified as having a true zero WTP, the estimated mean WTP derived from the 1.5B model increases substantially to KRW 5562, which also holds statistical significance. This figure is nearly twice as large as the mean WTP of KRW 2712 obtained from the same model when estimated using the full set of 1000 observations. In the interest of conserving space, the detailed estimation results of the model are not reported here. These findings indicate that the exclusion of protest bid responses leads to a significant upward bias in the estimated mean WTP. However, as will be discussed later, even when protest bids are included, the CBA results still suggest that the expansion of RE is socially desirable. Thus, the exclusion of protest bids does not materially alter the overall policy implications derived from the CBA.

3.2.2. Calculation of the Total External Benefits

It is necessary to estimate the external benefits, a major component of the total benefits arising from the implementation of policies to expand the use of RE. Here, external benefits do not refer to tangible benefits that are directly visible, such as accounting cost savings, but rather to intangible benefits that are not traded in the market. External benefits can be estimated by multiplying the estimate of the average WTP by the number of households in the population. An important issue to be discussed at this point is whether the sample well represents the population. The conclusion of this study regarding this matter is affirmative. Information about the characteristics of the sample and population used in this study is presented in

Table 6. Comparison of characteristics between sample and population.

Characteristics		Sample a	Population b
Household income c		KRW 4.68 million	KRW 4.96 million
Gender
	Female	50.0%	50.4%
	Male	50.0%	49.6%
Region
	Seoul	22.2%	19.0%
	Pusan	7.4%	6.7%
	Daegu	4.9%	4.7%
	Incheon	5.9%	5.5%
	Gwangju	3.3%	2.9%
	Daejeon	3.3%	3.0%
	Ulsan	2.0%	2.1%
	Sejong	0.5%	0.7%
	Gyunggi	25.3%	24.4%
	Gangwon	2.5%	3.2%
	Chungbuk	2.6%	3.2%
	Chungnam	3.6%	4.3%
	Jeonbuk	3.4%	3.6%
	Jeonnam	2.2%	3.6%
	Gyungbuk	4.8%	5.4%
	Gyungnam	6.1%	6.5%

^a The number of respondents is 1000. ^b comes 2020 Population Census implemented by Statistics Korea and is available from Statistics Korea [39]. ^c denotes the average.

. The population values are available from the Population Census implemented by Statistics Korea [39].

At first glance, there does not appear to be a significant gap between the two characteristics. Of course, the sample value is KRW 0.28 million less than the population value in average monthly household income. However, considering that survey respondents tend to exclude financial income when responding, the difference is considered negligible. In addition, in this study, the authors did not conduct random sampling but commissioned a public opinion survey agency to conduct the survey and requested that the sample be consistent with the Population Census data from Statistics Korea. Therefore, the sample used in this paper is considered to be representative of the population.

According to Statistics Korea [39], South Korea had 22,179,969 households in 2024, the year in which the survey was conducted. By multiplying the total number of households by the estimated average WTP of KRW 2712 (USD 2.04) per household per year, derived from the 1.5B model presented in Table 2, the total external benefits amount to KRW 60.15 billion (USD 45.23 million) annually. Given that the payment horizon is set at ten years, these benefits are expected to be realized over the period 2024–2033. An important issue that follows is whether to apply constant prices or nominal prices. In this study, benefits and costs are evaluated using constant prices as of the survey year, thereby excluding the effects of inflation.

3.2.3. A Cost–Benefit Analysis

A CBA will be carried out for the policy that aims to increase the proportion of RE usage in public buildings from the current 30% to 40% by 2030, representing a 10%p increase. For this, the total benefits must be estimated. Since the external benefits have been calculated by year, the internal benefits need to be determined. The internal benefits can be seen as the savings on electricity bills resulting from the increased use of RE in public buildings. The main content of the increase involves the installation of 900 additional solar power systems with a capacity of 100 kW each in public buildings annually over the ten-year payment period from 2024 to 2033.

The lifespan of the additionally installed solar power systems is assumed to be 20 years. Therefore, the cumulative installed capacity will increase by 90,000 kW annually from 2024 to 2033, and then decrease by 90,000 kW annually from 2044 to 2053. The average utilization rate of solar power is assumed to be 15%. This means that electricity from solar power will be generated for 1314 h, which is 15% of the total 8760 h in a year. Since public buildings typically do not operate on holidays, the number of working days in a public building is assumed to be 260 days per year. As of 2024, the price of electricity used in public buildings was KRW 164 per kWh. The duration of the internal benefits is set for 29 years, starting from 2024, when the installation of solar power systems begins, until 2052, when the lifespan of the systems installed in 2033 expires. The estimates of external benefits, the estimation process and estimates of internal benefits, and the total benefits estimate are included in

Table 7. Results from estimating the economic benefits.

Year	External Benefits a (A) (Unit: Million Korean Won)	Internal Benefits (B)			Total Benefits e (A + B) (Unit: Million Korean Won)
		Cumulative Additional Installation Capacity b (Unit: MW)	Power Generation c (Unit: GWh)	Electricity Bill Savings d (Unit: Million Korean Won)
2024	60,152	90	84	13,815	73,967
2025	60,152	180	168	27,631	87,783
2026	60,152	270	253	41,446	101,598
2027	60,152	360	337	55,261	115,414
2028	60,152	450	421	69,077	129,229
2029	60,152	540	505	82,892	143,044
2030	60,152	630	590	96,708	156,860
2031	60,152	720	674	110,523	170,675
2032	60,152	810	758	124,338	184,490
2033	60,152	900	842	138,154	198,306
2034		900	842	138,154	138,154
2035		900	842	138,154	138,154
2036		900	842	138,154	138,154
2037		900	842	138,154	138,154
2038		900	842	138,154	138,154
2039		900	842	138,154	138,154
2040		900	842	138,154	138,154
2041		900	842	138,154	138,154
2042		900	842	138,154	138,154
2043		900	842	138,154	138,154
2044		810	758	124,338	124,338
2045		720	674	110,523	110,523
2046		630	590	96,708	96,708
2047		540	505	82,892	82,892
2048		450	421	69,077	69,077
2049		360	337	55,261	55,261
2050		270	253	41,446	41,446
2051		180	168	27,631	27,631
2052		90	84	13,815	13,815

^a They derived by extending the average willingness to pay per household obtained from Table 2 to the entire country. ^b This represents the cumulative capacity of renewable energy (RE) facilities added to public buildings each year over the 10 years from 2024 to 2033. ^c This refers to the increased power generation from RE facilities. ^d They mean the savings on electricity bills, calculated by multiplying the increased power generation from RE facilities by the 2024 public electricity rate. ^e They are calculated as the sum of external benefits and internal benefits.

.

The cost of constructing a solar power plant is KRW 1.20 million per kW, and a 10% contingency cost is additionally considered, yielding a final cost of KRW 1.32 million (USD 992) per kW. The cost of operation and maintenance (O&M) is estimated to be 3% of the cost of construction including the contingency cost. Information on the construction cost, contingency cost, and O&M cost was collected from interviews with industry insiders who sell solar power plants. The estimated results of installing additional capacity, cumulative additional installation capacity, capital costs, O&M costs, and total costs by year from 2024 to 2052 are presented in

Table 8. Results from estimating the costs.

Year	Additional Installation Capacity a (Unit: MW)	Cumulative Additional Installation Capacity b (Unit: MW)	Capital Costs (Unit: Million c Korean Won)	Operation and Maintenance Costs d (Unit: Million Korean Won)	Total Costs e (Unit: Million Korean Won)
2024	90	90	118,800	3564	122,364
2025	90	180	118,800	7128	125,928
2026	90	270	118,800	10,692	129,492
2027	90	360	118,800	14,256	133,056
2028	90	450	118,800	17,820	136,620
2029	90	540	118,800	21,384	140,184
2030	90	630	118,800	24,948	143,748
2031	90	720	118,800	28,512	147,312
2032	90	810	118,800	32,076	150,876
2033	90	900	118,800	35,640	154,440
2034		900		35,640	35,640
2035		900		35,640	35,640
2036		900		35,640	35,640
2037		900		35,640	35,640
2038		900		35,640	35,640
2039		900		35,640	35,640
2040		900		35,640	35,640
2041		900		35,640	35,640
2042		900		35,640	35,640
2043		900		35,640	35,640
2044		810		32,076	32,076
2045		720		28,512	28,512
2046		630		24,948	24,948
2047		540		21,384	21,384
2048		450		17,820	17,820
2049		360		14,256	14,256
2050		270		10,692	10,692
2051		180		7128	7128
2052		90		3564	3564

^a This represents capacity of newly installed renewable energy (RE) facilities in public buildings each year over the 10 years from 2024 to 2033. ^b This represents the cumulative capacity of RE facilities added to public buildings each year over the 10 years from 2024 to 2033. ^c They refer to construction costs of newly installed RE facilities. ^d They indicate operating and maintenance costs of newly installed RE facilities. ^e They are calculated as the sum of capital costs and operating and maintenance costs.

.

Table 9. Results of benefits, costs, and net benefits.

Year	Benefits (A) (Unit: Million Korean Won)		Costs (B) (Unit: Million Korean Won)		Net Benefits (A−B) (Unit: Million Korean Won)
	Values a	Discounted Values b	Values c	Discounted Values b	Values	Discounted Values b
2024	73,967	73,967	122,364	122,364	−48,397	−48,397
2025	87,783	84,003	125,928	120,505	−38,145	−36,503
2026	101,598	93,036	129,492	118,580	−27,894	−25,543
2027	115,414	101,136	133,056	116,597	−17,642	−15,460
2028	129,229	108,366	136,620	114,564	−7391	−6198
2029	143,044	114,786	140,184	112,491	2860	2295
2030	156,860	120,452	143,748	110,383	13,112	10,068
2031	170,675	125,417	147,312	108,249	23,363	17,168
2032	184,490	129,731	150,876	106,094	33,614	23,637
2033	198,306	133,441	154,440	103,923	43,866	29,517
2034	138,154	88,961	35,640	22,950	102,514	66,011
2035	138,154	85,130	35,640	21,961	102,514	63,169
2036	138,154	81,464	35,640	21,016	102,514	60,449
2037	138,154	77,956	35,640	20,111	102,514	57,846
2038	138,154	74,599	35,640	19,245	102,514	55,355
2039	138,154	71,387	35,640	18,416	102,514	52,971
2040	138,154	68,313	35,640	17,623	102,514	50,690
2041	138,154	65,371	35,640	16,864	102,514	48,507
2042	138,154	62,556	35,640	16,138	102,514	46,418
2043	138,154	59,862	35,640	15,443	102,514	44,419
2044	124,338	51,556	32,076	13,300	92,262	38,256
2045	110,523	43,854	28,512	11,313	82,011	32,541
2046	96,708	36,720	24,948	9473	71,760	27,247
2047	82,892	30,119	21,384	7770	61,508	22,349
2048	69,077	24,018	17,820	6196	51,257	17,822
2049	55,261	18,387	14,256	4743	41,005	13,644
2050	41,446	13,197	10,692	3404	30,754	9792
2051	27,631	8419	7128	2172	20,503	6247
2052	13,815	4028	3564	1039	10,251	2989
Sum	3,364,593	2,050,233	1,900,800	1,382,926	1,463,793	667,306

^a They come from Table 7. ^b A discount rate of 4.5% is used. ^c They come from Table 8.

contains the values and discounted values of the benefits, costs, and net benefits for the period from 2024 to 2052. The values of benefits and costs are extracted from Table 7 and Table 8, respectively, and net benefits are defined as benefits minus costs. A social discount rate of 4.5% was used to derive the discounted values, which is the rate officially adopted by the Korea Ministry of Planning and Finance.

Table 10. Results of the cost–benefit analysis.

Net Present Value	Benefit–Cost Ratio	Internal Rate of Return
KRW 667.3 billion (USD 501.7 million)	1.48	15.3%

presents the results of the CBA. The net present value and benefit–cost ratio are estimated at KRW 667.31 billion and 1.48, respectively. Additionally, the internal rate of return is calculated to be 15.3%. Since they are all greater than 0, 1, and 4.5%, respectively, the policy to increase the use of RE in public buildings passes the CBA. Therefore, this enhancement is socially desirable.

4. Conclusions

The South Korean government aims to increase the proportion of RE used in public buildings from the current 30% to 40% by 2030, an increase of 10%p, to reduce GHG emissions in the building sector. This study analyzed the public WTP for this increase through the application of CV. The estimated average WTP per household reached KRW 2712 (USD 2.04) annually and achieved statistical significance. This value corresponds to the external benefits arising from the increase. Internal benefits could be derived from the reduced electricity costs due to the expanded use of self-produced power. The economic benefits of this increase consist of the sum of these external and internal benefits. A CBA comparing the estimated economic benefits with appropriately estimated related costs showed that the increase is economically viable. Therefore, the authors concluded that this increase would be socially beneficial.

This paper can contribute to the literature in two ways. First, it is the first one to explore the public WTP for the expanded use of RE in public buildings using CV, to the best of the authors’ knowledge. Furthermore, among the various methods of inducing WTP and WTP analysis models that can be used in applied CV studies, those reasonably accepted in the literature were appropriately applied. Second, the paper not only presents the public WTP estimates but also attempts a CBA to assess the economic feasibility of the increase. Particularly, the determination of economic feasibility is a key factor that influences the success of securing related budgets, and this is even more critical in South Korea. Therefore, the results of this study can be usefully employed by the government.

In addition to its relevance for the South Korean context, the findings of this study hold broader implications for other countries that are also seeking to expand RE deployment in public infrastructure. While institutional, policy, and financial conditions naturally differ across national contexts, the methodological approach adopted here—combining CV technique with econometric modeling—provides a transferable framework for evaluating public acceptance and willingness to support RE initiatives elsewhere. The key insights, such as the role of policy credibility, information provision, and perceived co-benefits, are not unique to South Korea and may be adapted to inform decision-making in other economies pursuing the energy transition. Nevertheless, the extent of generalizability should be considered with caution, as contextual factors such as governance structures, cultural values, and levels of technological adoption are likely to shape public responses in different settings.

In acknowledging the limitations of this study, it is important to note that CV methods, while widely applied in environmental and energy economics, are not without potential biases. Issues such as hypothetical bias, strategic bias, and starting-point effects may influence respondents’ stated preferences and lead to over- or underestimation of WTP. Although the survey design incorporated several techniques to mitigate these concerns, the possibility of residual bias cannot be fully ruled out. Similarly, the CBA conducted here is based on a set of simplifying assumptions, including the choice of discount rate, treatment of external benefits, and time horizon of the analysis. These assumptions, though consistent with practices in the literature, necessarily influence the magnitude and interpretation of the results.

Another limitation concerns the representativeness and temporal reliability of the survey data. While the sample was designed to reflect the demographic structure of South Korean households, survey-based results inevitably face challenges of coverage and potential non-response bias. Moreover, WTP for RE deployment is not static; it may evolve over time as public awareness, economic conditions, and policy priorities change. The present study therefore offers a snapshot of preferences at one point in time, rather than a comprehensive prediction of long-term patterns. Future research could extend this work by employing longitudinal surveys, experimental approaches, or cross-country comparisons to strengthen both the external validity and policy relevance of the findings.