Residential Consumers’ Willingness to Pay for Sustainable Grid Resilience Against Climate-Induced Large-Scale Outages of Long-Duration: Evidence from South Korea

Abstract

South Korea faces escalating climate change threats that increase the risk of large-scale outages of long duration. However, efforts to expand the grid are often limited by low consumer acceptance of higher tariffs. This study employs a nationally representative contingent valuation survey of 1000 households to quantify residential consumers’ willingness to pay (WTP) for grid expansion to avoid a hypothetical 24 h nationwide blackout caused by extreme weather. The findings reveal an average monthly WTP of KRW 2226 (USD 1.54), equivalent to about KRW 0.60 trillion (USD 0.4 billion) annually—approximately 12% of planned grid investment needs. Among the socioeconomic variables, the negative coefficient on generation suggests younger cohorts exhibit higher WTP, consistent with—though not conclusive evidence of—their expectation of greater exposure to future climate risks. Similarly, the presence of children is positively associated with WTP, indicating family protection motives that encompass both immediate household needs and intergenerational considerations related to the distribution of climate-related burdens. These findings provide policy-relevant insights for designing equitable and acceptable tariff schemes that support critical investments to strengthen sustainable grid resilience amidst escalating climate risks.

1. Introduction

Climate change is subjecting power grids worldwide to unprecedented weather extremes, leading to a heightened risk of large-scale outages of long duration (LSLD) as demonstrated by recent severe blackouts in Ecuador, the United States, Spain, and Portugal [1,2,3]. Achieving sustainable grid resilience—defined as the ability of a system to anticipate, absorb, recover from, and adapt to high-impact shocks—requires substantial infrastructure investment as extreme weather events intensify, with estimates suggesting that approximately USD 600 billion annually will be required globally by 2030 to meet climate goals [4,5]. Notably, the transition to renewable energy, although essential for climate mitigation, also increases grid vulnerability and the urgency of investment is underscored by recent disruptions observed during energy transitions in multiple countries.

South Korea faces similar challenges; the grid remains too constrained to prevent LSLD in the face of extreme weather, and curtailments in power generation are already frequent. Although no nationwide blackout has occurred since 2011, the risk is growing in the absence of timely grid expansion, yet financial constraints—stemming from residential tariffs set below production costs—undermine the feasibility of the required investment. The 11th Long-Term Transmission and Substation Facilities Plan mandates KRW 72.8 trillion (USD 50.5 billion) in spending, but funding obstacles persist due to the Korea Electric Power Corporation (KEPCO)’s substantial debts and losses. Ultimately, grid expansion depends on residential consumers’ willingness to pay (WTP) higher tariffs, a challenge relevant not only in South Korea but also in many other countries facing significant grid investment needs (e.g., ref. [6]).

South Korea’s power grid is highly vulnerable due to both structural aging and escalating climate risks. Approximately 40% of transmission and distribution facilities exceed 25 years of service life, constraining capacity amid rising electricity demand driven by the Green New Deal and the digital economy. Renewable integration—projected to reach 30% by 2030, including offshore wind and rooftop PV—further strains the isolated grid through intermittency and reverse power flows, necessitating upgrades to intelligent high-voltage switchgear systems. The government acknowledges these challenges in the “2030 Smart Grid Development Plan,” committing KRW 5 trillion over the decade, with switchgear procurement accounting for 18% of the budget. However, residential tariffs remain below cost-recovery levels for social equity, limiting KEPCO’s investment capacity amid a KRW 205 trillion debt burden and a KRW 72.8 trillion investment requirement under the 11th Long-Term Plan. This study estimates residential WTP at approximately KRW 0.60 trillion annually—about 12% of the required annual investment—highlighting the need to address persistently low tariff acceptance that has hindered grid modernization.

A considerable body of literature has examined residential consumers’ WTP to avoid outages [7,8,9], though most studies focus on short outages lasting one to two hours. More recent research has begun to explore outages of longer durations, exceeding 24 h, yet such studies remain limited, and virtually none have been conducted for South Korea [10,11,12]. This study estimates South Korean residential consumers’ WTP specifically for grid expansion to avoid a hypothetical 24 h nationwide outage induced by extreme weather. The good being valued is defined as the certain avoidance of this specific outage through targeted infrastructure investment, with particular emphasis on grid expansion as the primary means of achieving such avoidance. In this context, the analysis also evaluates whether the estimated WTP could provide a meaningful basis for financing the substantial investments required for grid infrastructure.

Since extreme weather events induced by climate change unfold over long time horizons and South Korea has not yet experienced LSLD, uncertainty about the future is expected to shape South Korean consumers’ WTP responses. In other words, their expectations about how the climate will evolve and perceptions of whether these changes will personally affect them are likely to significantly influence their WTP. In particular, younger generations with greater expected exposure to future climate risks and households with children may perceive these risks more acutely, thereby leading to higher WTP. In contrast, factors with limited relevance to expectations about future climate risks—such as past outage experience, political ideology, and gender—are expected to exert little influence on WTP.

Accordingly, this study examines how generation, presence of children, outage experience, political ideology, and gender affect WTP, offering insights into the socioeconomic factors that shape tariff acceptance. The results make two key contributions to the existing literature. First, in a context where research on this issue remains limited globally and is virtually absent for South Korea, this study provides estimates of residential consumers’ WTP for grid investment to avoid a 24 h nationwide outage and derives associated policy implications. Second, by identifying socioeconomic factors that significantly influence WTP—particularly those related to expectations about future climate risks and perceptions of their direct personal relevance—this study advances the understanding of the acceptance of residential tariff increases in South Korea. To the best of the authors’ knowledge, this represents the first application of this conceptual framework in the existing literature.

In summary, the principal aim of this study is to quantify the WTP of South Korean residential electricity consumers for expanding the power grid to avoid potential LSLD induced by climate change, to examine the socioeconomic factors influencing this valuation, and to derive policy-relevant insights. The structure of this paper is as follows: Section 2 provides a comprehensive review of the relevant literature; Section 3 details the method used, including the contingent valuation (CV) approach; Section 4 presents the empirical findings alongside their policy implications; and Section 5 concludes with a summary and suggestions for future research.

2. Short Literature Review

2.1. A Review of the Related Literature

A substantial body of literature has estimated residential electricity consumers’ WTP to avoid power outages [7,8,9], with most recent studies, as shown in

Table 1. Overview of recent studies on residential consumers’ willingness to pay (WTP) to avoid power outages.

Sources	Methods a	Countries	Outage Durations b	Mean WTPs
Cohen et al. [13]	CE	19 EU nations	1 h	EUR 0.32 to 1.86
Baik et al. [10]	CV	United States	24 h	USD 0.35 to 1.2 per kWh
Morrissey et al. [14]	CE	England	20 m, 1 h, 4 h, and 8 h	GBP 5.29 to 31.37
Kim et al. [15]	CV	South Korea	1 h	KRW 1522 per month
Baik et al. [11]	CV	United States	10 d	USD 43 to 60 per day and USD 19 to 29 per day
Carlsson et al. [16]	CV	Sweden	3 m, 1 h, 4 h, and 12 h	SEK 7.1 to 234.3
Vennemo et al. [12]	CV	Norway	24 h	NOK 678 to 979
Matsubara et al. [17]	CV	Japan	2 h	JPY 501.1 to 559.9 per kWh

Notes: ^a CE and CV indicate choice experiment and contingent valuation, respectively. ^b d, h and m imply days, hours, and minutes, respectively.

, relying on the stated preference (SP) method, which employs survey instruments that present hypothetical outage scenarios and elicit the respondents’ WTP, thereby allowing researchers to adjust the parameters of the outage (e.g., duration and timing) to match specific research objectives. Consequently, it has been widely applied, with CV and choice experiment (CE) representing the most prominent applications.

CV-based studies typically assume a single outage scenario (i.e., single attribute) and directly elicit the WTP, whereas CE-based studies assess trade-offs across multiple combinations of outage scenarios and their associated costs to avoid them (i.e., multiple attributes). For instance, Carlsson et al. [16] applied CV and directly asked consumers about their WTPs to avoid outages of three minutes, one hour, four hours, and twelve hours, yielding mean values of SEK 7.1, SEK 25.5, SEK 90.4, and SEK 234.3, respectively. Alternatively, Morrissey et al. [14] employed CE to infer WTPs by having consumers compare multiple combinations of outage duration, timing, and avoidance cost and choose among them, and reported mean marginal values per minute of GBP 1.17, 0.40, 0.10, and 0.05 for outage durations of twenty minutes, one hour, four hours, and eight hours, respectively. Despite methodological differences, both studies revealed a consistent insight: as the duration of the outage increases, the mean WTP to avoid it rises correspondingly.

To date, most SP-based studies have estimated WTP to avoid short-term outages lasting less than 24 h, as shown in Table 1. In particular, a considerable number of studies have focused on outages of one to two hours, such as Cohen et al. [13], Kim et al. [15], and Matsubara et al. [17]. Their findings indicate that WTPs vary substantially across countries, even for comparable outage durations, reflecting differences in socioeconomic conditions such as consumer awareness and income levels. For example, Kim et al. [15] conducted a CV study involving 1000 South Korean households, estimating a mean monthly WTP of KRW 1522 (USD 1.41) to avoid a one-hour outage occurring monthly. As climate change has increased the frequency and intensity of extreme weather events, the estimation of WTP to avoid LSLD (lasting 24 h or more) has attracted growing scholarly attention.

So far as the authors are aware, however, only three such studies exist, and none focus on South Korea [10,11,12]. Baik et al. [10] employed CV to estimate the WTP of residents in Allegheny County, the United States, for back-up services during a 24 h summer outage. Results revealed higher WTP for high-priority (HP) loads (e.g., lighting, air-conditioning) than for low-priority (LP) loads (e.g., entertainment devices), at USD 0.75 per kWh versus USD 0.51 per kWh. When additional outage-related information was provided, the disparity widened (HP: USD 1.2 per kWh vs. LP: USD 0.35 per kWh), enabling respondents to articulate their WTPs more clearly, although considerable uncertainty regarding the precise amounts persisted throughout the responses.

This outcome reflects the rarity of LSLD, which limits respondents’ familiarity with the potential damages from such events. Consequently, the provision of information has helped respondents refine their preferences, thereby accentuating the divergence between WTPs for HP and LP services. These findings suggest that, in the context of LSLD, (i) partial backup services meeting consumers’ bare necessities may be of particular importance, and (ii) respondents’ perceptions of potential damages can represent a significant factor shaping their WTPs.

Baik et al. [11] extended this analysis to a ten-day winter outage in the northeastern United States and discovered that residential consumers had a WTP of USD 43 to 60 per day for household-level backup services (e.g., domestic appliances) and USD 19 to 29 per day for community-level backup services (e.g., water and sewage). Similarly, Vennemo et al. [12] employed CV to study Norwegian households’ WTP to avoid a 24 h outage and reported that the mean WTP increased by nearly 50% (from NOK 678 to NOK 979) when the outage affected not only households but also infrastructure services such as public transport. These two findings underscore that, when LSLD occur, residential consumers’ WTP extends beyond individual households to encompass broader community services.

The foregoing literature review yields four significant insights that collectively advance the understanding of the research domain.

• Firstly, only a limited number of studies have estimated WTP to avoid LSLD, and none have focused on residential consumers in South Korea. This study, therefore, provides novel insights and policy relevance by estimating their WTP for avoiding a 24 h nationwide outage.
• Secondly, the rarity of LSLD creates considerable uncertainty regarding potential damages from such events, making perceptions of these impacts a critical determinant of WTP. Building on this reasoning, this paper further tests the hypothesis that expectations about the occurrence of such outages induced by extreme weather events, together with perceptions of the direct personal damages they would entail, significantly influence WTP. In particular, socioeconomic characteristics such as generation, presence of children, and educational level are expected to be closely associated with these expectations and/or perceptions and to exert significant effects on WTPs, whereas gender, political ideology, and experience of past outages are not. As far as the authors are aware, this study constitutes the first such attempt reported in the literature.
• Thirdly, although distributed energy resources (DERs) may provide a limited electricity supply during LSLD, they can satisfy consumers’ bare necessities, highlighting their critical value. This suggests a potential trade-off between consumers’ WTP for DERs and for the grid. Accordingly, this paper investigates whether the adoption status of home solar—either already installed or intended for future installation—exerts a significant effect on WTP for grid investment, with particular attention to the direction of this relationship.
• Finally, prior research shows that residential consumers exhibit a WTP not only for household-level but also for community-level services, and that the provision of information on the potential damage produced by an outage reduces the uncertainty in the WTP responses. Building on this evidence, this study ensures that the respondents receive sufficient information on both the household- and community-level impacts.

2.2. Further Literature Review on Resilience Under Multiple Hazards

Recent resilience research increasingly recognizes that infrastructure systems, including power grids, operate under recurrent and interacting hazards rather than isolated events, with profound implications for modeling outage risks and restoration strategies. Wei et al. [18] model dynamically interdependent hazards where prior events influence the frequency and severity of subsequent disruptions, demonstrating that proactive maintenance reduces cumulative downtime by mitigating compounding effects in path-dependent systems. Extending this, Wei et al. [19] introduce multimodal performance trajectories under multiple recurrent hazards, challenging traditional single-disruption resilience curves and emphasizing temporal patterns in recovery phases.

These frameworks align with power grid contexts, where cascading failures from extreme weather amplify vulnerabilities. For instance, Kong et al. [20] optimize resilience-based restoration for dependent networks against cascading failures, showing that prioritizing interdependencies minimizes performance loss. Panteli et al. [21] quantify multi-hazard resilience in transmission networks, integrating weather-induced outages with probabilistic cascading models.

These studies yield three key insights with direct implications for this research. First, LSLD should be conceptualized as the cumulative outcome of recurrent, interdependent hazards—such as typhoons followed by grid overloads—thereby lending plausibility to the 24 h nationwide blackout scenario adopted in this study, particularly in the context of South Korea’s isolated grid and intensifying climate extremes. Second, the nonlinear, path-dependent nature of resilience trajectories underscores consumer uncertainty in valuing outage damages, reinforcing this study’s conceptual framework that expectations about future climate risks and related socioeconomic factors (e.g., generational effects, presence of children) play a central role in shaping WTP.

Lastly, this paper addresses a critical research gap at the intersection of engineering resilience modeling and economic policy analysis: while technical literature advances hazard-dependent restoration and multimodal performance metrics, empirical quantification of residential consumers’ WTP to finance such resilience investments remains scarce, particularly in non-interconnected grids exposed to typhoon risks, such as South Korea. By estimating South Korean households’ WTP for grid expansion, this study bridges this gap and provides policymakers with evidence-based support for designing equitable tariff mechanisms that align public preferences with engineering imperatives for climate-resilient infrastructure.

3. Methodology

3.1. Method: CV

Building on the literature review, this study employs CV to quantify the WTP of residential electricity consumers in South Korea for grid expansion aimed at avoiding a 24 h nationwide outage, as the method allows for the adjustment of outage parameters (e.g., duration and scale) and is well suited for a single-attribute scenario [22]. Reviewing the CV literature, Venkatachalam [23] noted that the validity and reliability of a CV are influenced by several factors, including the disparity between the WTP and the willingness to accept (WTA), embedding effects, the quality of information provided, the choice of elicitation technique, and hypothetical bias. At the same time, the literature highlights that many of these errors and biases can be mitigated through the careful design and implementation of the survey.

Accordingly, guidelines have been suggested to ensure the rigorous application of CV [24,25,26]. Key recommendations include: (i) using WTP rather than WTA; (ii) selecting goods familiar to the respondents; (iii) presenting clear and credible scenarios with limited uncertainty; (iv) providing sufficient (not excessive) and unbiased information; (v) employing realistic and credible payment vehicles; (vi) pretesting the questionnaires; (vii) adopting sampling strategies that accurately reflect the target population; (viii) favoring in-person interviews over mail or telephone surveys; and (ix) carefully choosing the elicitation technique. This study adheres closely to these guidelines, with details provided in the following subsections.

3.2. Procedure for Applying CV

The implementation of a CV study typically involves five stages: (i) defining the good, scenario, and payment vehicle; (ii) designing the questioning format; (iii) implementing the field survey; (iv) adopting the WTP model; and (v) analyzing the WTP data.

3.2.1. Defining the Good, Scenario, and Payment Vehicle

The design of a CV questionnaire necessitates a precise definition of the good being valued, which is the transition from the baseline to the target state rather than the target state itself. In this study, the baseline state is defined as a certain 24 h nationwide outage occurring once a year during the summer peak due to extreme weather (e.g., a super typhoon). The target state—and thus the good being valued—is grid expansion that avoids this specific outage entirely. This study values the deterministic transition from outage occurrence to outage avoidance via grid investment, rather than probabilistic risk reduction. Thus, the good to be valued is the reinforcement of the grid to prevent such an outage, with the benefit defined as the avoidance of the associated damages.

To empirically justify the plausibility of this baseline scenario—a 24 h nationwide blackout occurring annually during summer peaks due to extreme weather in the absence of sufficient grid investment—we consider South Korea’s unique vulnerabilities. Geographically positioned in the primary typhoon pathway of the Western North Pacific basin, the country experiences typhoon landfalls nearly every year, with an average of 3–4 typhoons making direct hits annually between 1951–2020 [27]. These events routinely cause localized power outages; for instance, Typhoon Hinnamnor (2022) and Typhoon Maemi (2003) resulted in outages affecting over 1 million households, with durations exceeding 24 h in affected regions due to cascading failures in transmission infrastructure.

Critically, South Korea operates an isolated power grid, lacking interconnections with neighboring countries, rendering it a “power island” unable to import or export electricity during crises—unlike interconnected European or North American systems [4]. Under severe typhoons, localized outages risk escalating into widespread blackouts, as evidenced by near-misses during Typhoon Kong-rey (2018), where grid overloads threatened systemic collapse. Moreover, rapid renewable energy penetration—reaching 20% by 2025 and projected at 30% by 2030—exacerbates grid instability through intermittency of inverter-based generation, heightening blackout risks without synchronous reserves [28]. Probabilistic assessments indicate a non-negligible annual risk of LSLD under Representative Concentration Pathway (RCP) 8.5 scenarios, with exceedance probabilities rising 15–25% by mid-century due to intensified typhoon extremes [3,28].

This study followed best-practice guidelines in defining the scenario. First, expert interviews were conducted to identify the most plausible outage scenario in South Korea. Second, respondents were provided with sufficient information on expected damages, since no 24 h nationwide outage has occurred in South Korea and respondents’ baseline knowledge was expected to be limited and asymmetric; as Bergstrom et al. [29] showed, such information can improve the validity of WTP estimates. Third, information covered both household- and community-level damages, as LSLD affect both household services (e.g., appliances) and community services (e.g., transportation and communication infrastructure); prior studies confirm that consumers express WTP for both [11,12]. Fourth, respondents received time-structured information on expected damages occurring immediately after the outage, and after one, three, seven, ten, and twenty-four hours. This was intended to mitigate their limited understanding of the realistic implications of numerical risk changes—that is, increases in outage duration—since embedding effects can be reduced when information highlights the realistic scale of such changes [30].

The selection of an appropriate payment vehicle is critical to minimizing hypothetical bias. This study adopted the monthly household electricity bill, which satisfies the four key criteria outlined by Sajise et al. [31]: being credible and minimizing strategic bias, acceptable to respondents, directly linked to the good being valued (investment in the grid), and easily understood and widely applicable. In South Korea, this choice is particularly suitable because residential consumers pay monthly household-based charges that directly finance investment in the grid.

3.2.2. Designing the Format of the Questionnaire

Consistent with established guidelines, this study chose to elicit the WTP rather than the WTA, as numerous studies have shown that the magnitude of the WTA usually exceeds that of the WTP due to income and substitution effects [32,33,34]. It also employed the referendum (or closed-ended) dichotomous choice (DC) format, which is characterized by higher incentive compatibility and a lower susceptibility to strategic bias than open-ended formats [24,35]. Three main DC formats are commonly found in the literature: single-bound (SB), double-bound (DB), and triple-bound [36]. While additional bounds increase the efficiency by reducing the variance in the estimates of the WTP, they also raise the risk of response effects, where subsequent answers are inconsistent with earlier responses. To balance these trade-offs, Cooper et al. [37] developed the one-and-one-half-bound (OB) format, which combines the low response effects of SB with the efficiency gains of DB. This study accordingly employed the OB format.

As illustrated in Figure 1, the OB DC format employs two bid amounts: a lower bid ($A^L$) and a higher bid ($A^H$). Respondents are assigned to two groups, with one group offered $A^L$ first and the other offered $A^H$ first. If a respondent presented with $A^L$ first answers ‘yes,’ the respondent is subsequently offered $A^H$; if the initial answer is ‘no,’ no further question is asked. Conversely, if a respondent presented with $A^H$ first answers ‘yes,’ no additional question is asked; if the answer is ‘no,’ $A^L$ is subsequently offered.

This study implements a comprehensive suite of ex ante and ex post strategies to minimize hypothetical bias inherent to CV methods, following the National Oceanic and Atmospheric Administration guidelines (Arrow et al. [24]) and best practices in the environmental valuation literature. Ex ante, the questionnaire incorporated cheap talk scripts explicitly warning respondents about potential overstatement of WTP in hypothetical scenarios and emphasizing budget constraints; used the realistic, familiar payment vehicle—monthly household electricity bills directly linked to grid investment; and delivered detailed, time-structured information on outage damages via trained in-person interviewers to enhance scenario credibility and respondent comprehension.

Ex post, the OB DC format with randomized bid vectors reduces incentives for strategic responses, while the spike model accommodates the high 51.5% zero-WTP rate through maximum likelihood estimation that jointly models response patterns and point estimates. These measures align with empirical evidence showing significant bias reduction—collectively lowering hypothetical WTP by 20–50% relative to open-ended formats—while maintaining estimate reliability for policy analysis.

3.2.3. Implementing the Field Survey

Reliable data collection requires careful attention to the administration of the survey, the selection of the respondents, and the mode of the survey. Following established guidelines, the field survey was conducted in March 2025. The unit of analysis was the household, with respondents restricted to the head of household or spouse, aged 20–65, representing the economically active population. This approach reflects the fact that electricity bills—the chosen payment vehicle—are paid at the household level, and ensures that WTP responses are informed by household income and budget constraints.

A sample size of 1000 households was selected to balance cost and representativeness. Arrow et al. [24] noted that, with random sampling, a DC format yields a sampling error of approximately ±3% at this size. Accordingly, this study employed stratified random sampling based on the most recent national census [38] to achieve an adequate representation of the population’s socioeconomic attributes. The sampling was conducted by a professional polling firm with substantial experience in CV studies.

Face-to-face interviews were chosen, despite the higher costs involved, over telephone, mail, or online surveys, as qualified interviewers can effectively deliver sufficient information to the participants and thereby improve the reliability of the data [39]. The questionnaire was pretested with a focus group of 30 participants and refined based on their input to revise ambiguous wording and improve readability. Interviews were subsequently verified through follow-up contacts, and incomplete or unreliable responses were discarded and replaced with additional surveys, yielding 1000 valid household responses.

3.2.4. Adopting the WTP Model

This study applies Hanemann’s [40] utility difference model, grounded in McFadden’s [41] random utility framework, to derive Hicksian compensating surplus from DC CV data [35,42]. Let $y$ denote the respondent’s income, $s$ a vector of their observable socioeconomic characteristics affecting their preferences, $q_0$ the state in which a 24 h nationwide outage occurs, and $q_1$ the state in which such an outage is avoided. The utilities $u(q_0,y;s)$ and $u(q_1,y;s)$ are modeled as random variables with expectations $v(q_0,y;s)$ and $v(q_1,y;s)$, respectively. Letting ${ϵ}_0$ and ${ϵ}_1$ be independent and identically distributed random terms with zero mean, the respondent’s utility function is

$$u\left(q_j,y;s\right)=v(q_j,y;s)+{ϵ}_j, (j=0,1)$$

(1)

If the respondent answers “yes” to paying amount A to avoid the outage, then $u\left(q_1, y-A;s\right) \ge u\left(q_0,y;s\right)$. Using Equation (1), this condition can be reformulated in terms of the utility difference function $∆v\left(A;s\right)$, such that a “yes” response implies

$$∆v\left(A;s\right)=v\left(q_1, y-A;s\right)-v\left(q_0, y;s\right) \ge {ϵ}_0-{ϵ}_1$$

(2)

Letting $\eta = {ϵ}_0-{ϵ}_1$ and denoting its cumulative distribution function (cdf) by $F_{\eta }(·)$, the probability of obtaining a “yes” answer is

$$\mathrm{Pr}\left(“\mathrm{yes}”\right)=\mathrm{Pr}\left\{∆v\left(A;s\right)\ge \eta \right\}=F_{\eta }\left\{∆v\left(A; s\right)\right\}$$

(3)

Alternatively, if the respondent chooses “yes” to the same question, this implies that their WTP is greater than or equal to $A$. Letting $G_{WTP}(·)$ denote the cdf of the random variable WTP, the probability of a “yes” response can be formulated as

$$\mathrm{Pr}\left(“\mathrm{yes}”\right)=\mathrm{Pr}\left(WTP\ge A\right)=1-G_{WTP}\left(A;s\right)$$

(4)

Equations (3) and (4) jointly imply

$$\begin{gathered} \mathrm{Pr}\left(“\mathrm{yes}”\right)=1-G_{WTP}\left(A;s\right)={ F}_{\eta }\left\{∆v\left(A;s\right)\right\} \\ \mathrm{Pr}\left(“\mathrm{no}”\right)=G_{WTP}\left(A;s\right)=1-F_{\eta }\left\{∆v\left(A;s\right)\right\} \end{gathered}$$

(5)

As this study employs the OB format, each respondent can fall into one of six possible response patterns, depending on whether the higher bid ($A^H$) or the lower bid ($A^L$) is offered first. Let $I_{ik}=1$ if respondent $i$ exhibits response pattern $k$ and 0 otherwise, and let $P_{ik}(A^H, A^L;s_i)$ denote the corresponding probability defined in terms of $G_{WTP}(A;s_i)$ as in Equation (5). The log-likelihood function for the 1000 sampled households is then

$$\ln L={\sum }_{i=1}^{1000}{\sum }_{k=1}^6\left[I_{ik}·\ln P_{ik}\left(A^H,A^L;s_i\right)\right]$$

(6)

3.2.5. Analyzing the WTP Data

Since no 24 h nationwide outage induced by extreme weather has yet occurred in South Korea, some respondents may state a zero WTP, reflecting scepticism about the likelihood of such events. Consistent with prior research on the avoidance of outages, this study assumes that the WTP is non-negative, and accordingly added a follow-up question to those in Figure 1. Respondents who rejected the lower bid ($A^L$) were asked whether they were unwilling to pay at all, thereby identifying whether they had a positive WTP below $A^L$ or a zero WTP. Survey results showed that 51.5% of the sample (515 out of 1000 households) reported a zero WTP.

Thus, the WTP distribution is split between zero and positive values, requiring a model that can distinguish the two and mitigate inconsistencies that may arise if the questionnaire’s design and distributional assumptions about the WTP are misaligned [43,44]. In studies such as this, even though negative WTPs are excluded during the survey implementation, allowing the cdf of WTP, $G_{WTP}(A;s)$, to span from $-\infty$ to $+\infty$ can result in a negative mean WTP estimate when the acceptability of bid amounts is low, and a substantial number of respondents state that their WTP is zero.

The literature offers differing views on how to treat zero WTP responses. Arrow et al. [24] recommend identifying protest bids through follow-up (debriefing) questions and excluding them from the estimation of the WTP. Strazzera et al. [45], however, caution that overlooking the selection bias caused by protest responses may lead to biased estimates. Halstead et al. [46] propose reporting a range of WTP estimates, with the lower-bound estimate obtained by treating protest bids as legitimate zeros. The purpose of this research is to measure the actual willingness of South Korean residential electricity consumers to accept the tariff increases required for grid expansion. Since residential electricity prices have historically been set well below their production costs for political reasons, the conservative lower-bound estimates of WTP—treating all zero bids as legitimate—is the option best aligned with this study’s purpose. Accordingly, following Halstead et al. [46], all zero bids are treated as legitimate zeros.

Kriström [47] proposed the spike model to account for WTP distributions that include both zero and positive values. This study adopts that approach. Following established practice, $\eta$—introduced in Equation (3) as the random variable representing the distribution of utility differences $∆v(A;s)$—is assumed to follow a logistic distribution, and $∆v(A;s)$ is specified as a linear-additive function [41]. Letting $a$ denote the constant term and $b$ the coefficient on the bid amount ($A$), the utility difference function is

$$∆v\left(A;s\right)=a-b·A$$

(7)

Accordingly, the cdf of WTP takes the following form:

$$G_{WTP}\left(A;s\right) = \left\{\begin{array}{c}{[1+\exp (a-b·A)]}^{-1} \mathrm{if} A > 0\\ {[1+\exp \left(a\right)]}^{-1} \mathrm{if} A = 0\\ 0 \mathrm{if} A < 0\end{array}\right.$$

(8)

The parameters in Equation (8) are obtained through maximum likelihood estimation (MLE) based on Equation (6). The spike is defined as ${\left[1+\exp \left(a\right)\right]}^{-1}$. Additionally, the mean WTP is derived to be $\left(1/b\right)\ln \left[1+\exp \left(a\right)\right]$. To allow for the effects of socioeconomic characteristics in the model, some covariates can be incorporated. Let $X$ denote the covariate vector and B the associated coefficients. In this specification, $a$ in Equation (7) through (9) is replaced by $(a+X\prime B)$. $B$ is estimated by MLE, and the mean WTP conditional on the covariates is then derived. WTP is the bid amount ($A$) that makes respondents indifferent between the outage state ($q_0$) and the avoided-outage state ($q_1$) [48]. Letting $X=\left\{x_i\right\}$ and $B=\left\{{\beta }_i\right\}$, the marginal effect of the $i$-th covariate on the mean WTP is $\partial MWTP/\partial x_i={\beta }_i/b$.

4. Results and Discussion

4.1. Data

This study explores how respondents’ socioeconomic characteristics—including generation, presence of children, home solar adoption, education, outage experience, political ideology, and gender—influence their WTP for grid expansion to avoid LSLD as climate change intensifies. Prior research on these determinants yields mixed findings; while some studies emphasize the effects of age, gender, and family structure, others find political orientation or educational attainment more relevant, and past outage experience less so [6,10,11,15,49]. Given the absence of major LSLD in South Korea, this study focuses on how future climate expectations and perceived personal risks shape WTP, identifying four influential factors: plausibility of extreme events, anticipated damages, perceived direct impact, and availability of mitigation options, as illustrated in Figure 2.

To operationalize the conceptual framework, this study selects key variables that address critical policy questions: whether residential consumers’ acceptance of tariff increases for grid expansion rises with the increasing risk of LSLD due to climate change; whether older generations with children are more willing to bear climate adaptation costs, thus potentially reflecting family motives; and whether DERs, such as home solar installations, are viewed as substitutes for or complements to grid investments.

As summarized in

Table 2. Summary statistics and hypothesized impacts of socioeconomic variables.

Variables	Units	Mean	Standard Deviations	Minimum	Maximum	Anticipated Significance	Expected Sign
Generation	Likert	1.42	0.67	0.00	2.00	Significant	(−)
Children	(=1)	0.84	0.37	0.00	1.00	Significant	(+)
Solar_installed	(=1)	0.02	0.13	0.00	1.00	Significant	—
Solar_intended	(=1)	0.21	0.41	0.00	1.00	Significant	—
Education	Years	14.60	2.1	6.00	20.00	Significant	(+)
Outage	Numbers	0.12	0.42	0.00	4.00	Not significant	—
Political ideology	Likert	3.08	1.22	0.00	6.00	Not significant	—
Female	(=1)	0.50	0.50	0.00	1.00	Not significant	—
Salience	Likert	2.62	0.91	0.00	4.00	Significant	(+)
Household income	Million Korean won	6.04	4.12	0.80	100.00	Significant	(+)
Family size	Persons	2.97	1.12	1.00	7.00	Significant	(−)

Notes: Variables are indirect proxies for conceptual channels. Generation → future climate risk expectations; Children → family protection motives; Solar adoption → DER substitution/complementarity perceptions; Education → climate risk awareness. Associations should be interpreted cautiously; (+)/(−) indicate expected positive/negative effects on WTP, while — denotes no clear prior expectation or insignificance.

, Generation serves as an indirect proxy for future climate risk expectations, measured on a Likert scale across three age groups (20–34, 35–49, 50–65). While younger cohorts may anticipate greater lifetime exposure to future climate-induced outages—implying an expected negative coefficient—this variable imperfectly captures such expectations. Education, measured by years of formal schooling, serves as a proxy for climate risk awareness (though not a direct measure). Higher education levels may be associated with greater understanding of climate impacts on grid reliability, suggesting a positive association with WTP, although this relationship remains indirect. Solar adoption status serves as an indicator of DER familiarity, which may be associated with perceptions of substitution or complementarity with respect to grid investments, captured by two dummy variables: Solar_installed (1 if home solar panels are currently installed) and Solar_intended (1 if planning to install home solar in the future). Both variables are expected to exhibit mixed effects on WTP: they may reduce WTP if DERs are perceived as substitutes for sustainable grid-based resilience under budget constraints, or increase WTP if they are viewed as complementary to grid expansion. Children (1 if household includes children under 18) serves as a proxy for family protection motives, reflecting potential parental concerns about outage impacts on dependents, and is expected to increase WTP.

Other variables considered are past outage experience, political ideology, and gender, but these are expected to have limited influence on WTP given the uncertainty surrounding LSLD and their limited relevance compared with short-term outages. To control for individual heterogeneity, the study also includes salience of outages (importance assigned to LSLD), monthly household income, and family size. These factors capture respondents’ perceptions and financial capacity, enabling a comprehensive understanding of determinants driving acceptance of tariff increases for enhancing grid resilience.

4.2. Results

This study employs the OB DC spike model to estimate South Korean residential electricity consumers’ WTP for grid expansion aimed at avoiding a potential 24 h nationwide outage caused by extreme weather events (e.g., a super typhoon). Another objective is to test hypotheses regarding the effects of socioeconomic characteristics—such as generation, presence of children, and educational level—on WTP responses. Accordingly, both a model without covariates and a model with covariates were estimated. The results are summarized in

Table 3. Estimation results of the model.

Variables	Model Without Covariates	Model with Covariates a
Constant	−0.1070 (0.084) *	−1.8526 (0.002) ***
Bid amount b	−0.2880 (0.000) ***	−0.3025 (0.000) ***
Generation		−0.2232 (0.045) **
Children		0.4545 (0.052) *
Solar_installed		0.5808 (0.210)
Solar_intended		0.7817 (0.000) ***
Education		0.0665 (0.049) **
Outage		0.0968 (0.492)
Political ideology		−0.0605 (0.231)
Female		−0.1710 (0.171)
Salience		0.2860 (0.000) ***
Household income		0.0765 (0.002) ***
Family size		−0.1392 (0.059) *
Monthly household average willingness to pay	KRW 2226 (USD 1.54)	KRW 2128 (USD 1.48)
p-values	0.000 ***	0.000 ***
95% confidence intervals c	KRW 1990 to 2490 (USD 1.38 to 1.73)	KRW 1910 to 2379 (USD 1.32 to 1.65)
Spike	0.5267 (0.000) ***	0.5252 (0.000) ***
Wald statistics (p-values)	501.04 (0.000) ***	534.44 (0.000) ***
McFadden’s pseudo-R2		0.033
Log-likelihood	−1193.12	−1153.79
Number of observations	1000	1000

Notes: ^a p-values are enclosed in parentheses alongside each coefficient estimate. ^b The unit is 1000 Korean won (USD 1.0 = KRW 1442). ^c These are estimated by the method proposed by Krinsky and Robb [50]. ***, **, and * indicate that the estimate secures statistical significance at the 1%, 5%, and 10% levels, respectively.

. The estimated mean monthly household WTP is KRW 2226 (USD 1.54) for the model without covariates and KRW 2128 (USD 1.48) for the model with covariates. Their 95% confidence intervals (CIs) are KRW 1990 to 2490 (USD 1.38 to 1.73) and KRW 1910 to 2379 (USD 1.32 to 1.65), respectively. These CIs, computed using the method proposed by Krinsky and Robb [50], overlap, indicating no substantial difference between the two estimates.

The model that includes covariates reveals that generation, presence of children, intention to adopt home solar, and educational level significantly affect WTP at the 5%, 10%, 1%, and 5% levels, respectively, while actual adoption status of home solar, past outage experiences, political ideology, and gender do not. These results align with the hypotheses summarized in Table 2, with the exception that the actual adoption status of home solar is found to be statistically insignificant, contrary to the hypothesis, suggesting that past adoption of home solar may have been driven more by other motivations—such as electricity cost savings—rather than by a desire to avoid LSLD induced by climate change. More specifically, generation, presence of children, and educational level affect WTP negatively, positively, and positively, respectively, as hypothesized. On the other hand, no a priori hypothesis was formulated for intention to adopt home solar; the results show a positive effect. In addition, the control variables—salience of outages, household income, and family size—are significant at the 1%, 1%, and 10% levels, respectively, with their signs (positive, positive, and negative, respectively) consistent with the hypotheses and general economic theory.

The theoretical validity or internal consistency of the models are further supported by the following considerations: (i) in the model without covariates, the coefficient for the bid amount is significant at the 1% level and the constant at the 10% level, while in the model with covariates, both coefficients are significant at the 1% level; (ii) the Wald statistics testing the null hypothesis that all coefficients equal zero indicate that this hypothesis is rejected at the 1% level in both models, confirming their overall statistical significance; and (iii) the spike estimates for the two models are 0.5267 and 0.5252, both significant at the 1% level, closely matching the observed share of respondents with zero WTP (51.5%).

According to Bateman et al. [51], the OB DC model remains susceptible to response effects, defined as inconsistencies between responses to earlier and later bids. To test whether such effects exist in the survey data, this study estimated SB DC models using only responses to the first bid and compared the resulting CIs for mean WTP with those from the OB DC models. As shown in

Table 4. Estimation results of the single-bound model.

Variables	Model Without Covariates	Model with Covariates
Monthly household average willingness to pay	KRW 2557 (USD 1.77)	KRW 2430 (USD 1.69)
p-values	0.000 #	0.000 #
95% confidence intervals a	KRW 2273 to 2907 (USD 1.58 to 2.02)	KRW 2164 to 2761 (USD 1.50 to 1.91)

Notes: ^a They are estimated by the method proposed by Krinsky and Robb [50]. ^# indicates that the estimate is statistically significant at the 1% level.

, the 95% CIs from the SB DC models—KRW 2273 to 2907 (USD 1.58 to 2.02) without covariates and KRW 2164 to 2761 (USD 1.50 to 1.91) with covariates—overlap with those from the OB DC models—KRW 1990 to 2490 (USD 1.38 to 1.73) and KRW 1910 to 2379 (USD 1.32 to 1.65), respectively. The absence of substantial differences between the two approaches suggests that response effects are not present in the survey data, further supporting the validity of the OB DC estimates.

To quantify the effects of the socioeconomic variables on the WTP, estimates of the mean marginal WTP (MMWTP)—representing average marginal impacts—were calculated and are summarized in

Table 5. Mean marginal willingness to pay (WTP) estimates of the socioeconomic variables.

Variables	Mean Marginal WTP a	p-Values
Generation	KRW −738 (USD −0.51)	0.045 **
Children	KRW 1502 (USD 1.04)	0.053 *
Solar_installed	KRW 1920 (USD 1.33)	0.210
Solar_intended	KRW 2584 (USD 1.79)	0.000 #
Education	KRW 220 (USD 0.15)	0.049 **
Outage	KRW 320 (USD 0.22)	0.492
Political ideology	KRW −200 (USD −0.14)	0.231
Female	KRW −565 (USD −0.39)	0.171
Salience	KRW 945 (USD 0.66)	0.000 #
Household income	KRW 253 (USD 0.18)	0.002 #
Family size	KRW −460 (USD −0.32)	0.059 *

Notes: ^a The unit is Korean won (USD 1.0 = KRW 1442). ^#, **, and * indicate that the estimate is statistically significant at the 1%, 5%, and 10% level, respectively.

. While the primary spike model estimates average monthly WTP for complete outage avoidance, policymakers require granular insights into consumers’ marginal valuation of incremental improvements in sustainable grid resilience. We formalize MMWTP as the partial derivative of the mean WTP (MWTP), where $MWTP=\left(1/b\right)\ln \left[1+\exp \left(a+X\prime B\right)\right]$. Thus, ${MMWTP}_{x_i}=\partial MWTP/\partial x_i={\beta }_i/b$ can be derived. Table 5 operationalizes this by estimating covariate-specific MMWTPs using this formula. These marginal estimates bridge the gap between DC point estimates and continuous policy trade-offs, enabling cost–benefit analysis of targeted tariff designs that align demographic-specific valuations with optimal grid investment levels.

First, the MMWTP for generation is −738, significant at the 5% level, indicating that individuals aged 35–49 are willing to pay KRW 738 (USD 0.51) less on average than those aged 20–34, and those aged 50–65 are willing to pay KRW 738 (USD 0.51) less than those aged 35–49 to avoid the described 24 h nationwide outage. Second, the MMWTP for the presence of children is 1502, significant at the 10% level, suggesting that respondents with children are willing to pay KRW 1502 (USD 1.04) more on average than those without children. Third, the MMWTP for intention to adopt home solar is 2584, significant at the 1% level, indicating that respondents with intention to adopt home solar are willing to pay KRW 2584 (USD 1.76) more on average than those without. Fourth, the MMWTP for educational level is 220, significant at the 5% level, implying that each additional year of schooling increases the WTP by KRW 220 (USD 0.15).

These effects correspond to 35%, 71%, 121%, and 10% of the estimate of the mean WTP, which is KRW 2128 (USD 1.48), respectively, underscoring their substantial magnitudes and hence the economic significance of these variables. By contrast, the MMWTP estimates for actual adoption status of home solar, past outage experiences, political ideology, and gender are statistically insignificant, consistent with these variables’ lack of significant effects on WTP in the earlier results. Among the control variables, the MMWTP estimates are KRW 945 (USD 0.66) for salience of outages, 253 (USD 0.18) for household income, and −460 (USD −0.32) for family size, significant at the 1%, 1%, and 10% levels, respectively.

The positive and significant coefficient on Children (MMWTP = KRW 1502, p-value = 0.053) aligns with parental protection motives but allows multiple interpretations, as the dummy imperfectly proxies underlying channels. Immediate self-interested drivers—such as heightened household vulnerability during blackouts (e.g., childcare disruptions, remote schooling interruptions, perishable food spoilage, medical device dependency)—may dominate, particularly given the survey’s emphasis on time-structured household damages (Section 3.2.1). Altruistic intergenerational burden-sharing—where parents internalize their children’s future climate risks—remains plausible but cannot be directly observed without measures such as bequest motives or discounted lifetime outage exposure (Grant et al. [52]). Empirical disentanglement is precluded by data limitations: no interaction terms with generation (collinearity concerns) or attitudinal items (e.g., ‘WTP primarily for children’s future vs. current needs’).

4.3. Discussion of the Results

The estimation results are discussed along five dimensions: (i) the level of residential electricity consumers’ acceptance of tariff increases for grid investment to avoid the described 24 h nationwide outage induced by extreme weather, with national WTP calculated and compared to required investments in the grid; (ii) the role of socioeconomic determinants in shaping residential tariff acceptance in South Korea; (iii) further investigation of the zero WTP observations and their underlying drivers; (iv) caveats regarding the extrapolation of CV-based WTP estimates to policy contexts; and (v) the temporal validity and evolution of the framework.

4.3.1. Acceptance of Residential Tariff Increases

Sufficient grid capacity is essential for addressing potential LSLD induced by extreme weather; however, South Korea’s grid remains too constrained for such preparedness: it is already subject to frequent and severe curtailment of generation output. This stems from the concentrated electricity demand in the Seoul Metropolitan Area, while a substantial share of generation facilities is located in other regions, with local community opposition and financial limitations delaying any expansion of the grid. Some major grid construction projects have been delayed for up to 11 years, leading to sharp increases in generation constraints at power plants—603% on the east coast and 62% on the west coast between 2019 and 2023 [53]. Although no nationwide blackout has occurred since the rolling blackout of 2011, the risk of such outages is expected to rise substantially without a timely expansion of the grid.

In addition, the growing share of renewables in South Korea is further exacerbating the instability of the grid. Since the generation of renewable energy is intermittent and connects asynchronously to the grid through inverters, its addition tends to heighten the system’s volatility [54]. A large-scale blackout in Spain and Portugal in April 2025 was traced to voltage surges that triggered cascading protection mechanisms across the grid [55]. Such voltage instability is a well-established risk in power systems with high renewable penetration [56]. Accordingly, a timely expansion of the grid has become increasingly urgent to break the vicious cycle in which climate change increases the frequency and intensity of extreme weather events, while the expansion of renewables—essential for mitigating climate change—simultaneously increases the grid’s vulnerability to such events.

In response, the South Korean government and the KEPCO announced the 11th Long-Term Transmission and Substation Facilities Plan in May 2025. This plan requires investments of KRW 72.8 trillion (USD 50.5 billion) over 15 years from 2024 to 2038, or about KRW 4.9 trillion (USD 3.4 billion) annually [28]. However, KEPCO’s severe financial constraints, with cumulative losses of KRW 31 trillion (USD 21.5 billion) and debt of KRW 205 trillion (USD 142.2 billion) as of the first quarter of 2025, hinder the mobilization of the necessary funding. These difficulties largely stem from political decisions to supply residential electricity at prices far below the cost of production. According to Korea Electric Power Corporation [57], approximately 13.4% of residential tariffs are allocated to grid construction and maintenance. Thus, the survey respondents’ average monthly electricity bill of KRW 60,865 (USD 42.2) translates into a total annual residential contribution of nearly KRW 2.2 trillion (USD 1.5 billion) to grid investment.

The estimation results show that residential consumers in South Korea are willing to pay an additional KRW 2226 (USD 1.54) per household per month for grid expansion to avoid the described 24 h nationwide outage. For policy analysis, we aggregate the estimated household WTP to a national-level figure of approximately KRW 0.60 trillion (USD 0.4 billion) per year. This represents about 27% of the current annual residential contribution to grid construction and maintenance (KRW 2.2 trillion; USD 1.5 billion) and 12% of the annual investment needs under the 11th Long-Term Transmission and Substation Facilities Plan (KRW 4.9 trillion; USD 3.4 billion). While this aggregation illustrates the scale of potential public support as revealed by the CV survey, it does not assess the full feasibility of financing the required investment, which would require consideration of additional revenue sources, KEPCO’s debt burden, regulatory constraints, and implementation costs.

This issue—namely, consumer acceptance of tariff increases to finance grid expansion in response to extreme weather risks—is not unique to South Korea but is relevant for any country requiring large-scale grid investment. According to the International Energy Agency [58], about one-fourth of the world’s electricity network is exposed to severe storms, with more than 10% exposed to tropical cyclones. Moreover, nearly half of global power lines are subject to fire weather exceeding 50 days per year. The report further emphasizes that for countries such as India, Indonesia, and South Korea, the financial health of utilities constitutes a central obstacle to grid investment, underscoring the need for reforms to ensure that utilities can recover costs and thereby enable timely grid expansion.

4.3.2. Socioeconomic Determinants of Tariff Acceptance

This study examined four key socioeconomic variables—generation, presence of children, intention to adopt home solar, and educational level—to derive policy implications for residential consumers’ acceptance of tariff increases. While socioeconomic behavioral patterns align with the conceptual framework proposed in this study (Figure 2), the employed regressors remain indirect proxies for underlying channels through which climate risk expectations influence WTP. Future research that incorporates direct measures of risk expectations (e.g., survey items on perceived outage probability) would strengthen causal identification.

Firstly, the estimates show that generation has a statistically and economically significant negative effect on WTP, suggesting that acceptance of tariff increases for grid investment to avoid the described 24 h nationwide outage will improve over time as younger, more climate-vulnerable generations replace older cohorts. With the intensification of climate change, extreme weather events are likely to become more frequent and severe in South Korea. Kim et al. [59] reported that rising sea surface temperatures near South Korea have amplified the intensity of long-lived tropical cyclones approaching the peninsula, warning that such high-temperature anomalies could become the “new normal” after 2030. Similarly, Kim et al. [27] projected that once-in-a-century extreme rainfall events will increase by more than 20% across South Korea by the end of the century. As the likelihood of LSLD increases with these trends, the urgency of grid expansion will only grow. Thus, this finding is encouraging from a policy perspective, indicating the potential for greater public support for tariff increases in the future.

Secondly, the presence of children has a statistically and economically significant positive effect on WTP, indicating that older generations in South Korea are willing to pay more than their counterparts without children to protect their offspring. Intergenerational inequality has become a pressing policy concern: while current generations have largely contributed to the climate crisis, future generations are expected to bear a greater share of the costs and damages. Grant et al. [52], for example, projected that the share of individuals born in 2020 experiencing an unprecedented lifetime exposure to extreme weather events will be more than twice that of those born in 1960. Likewise, ICF International [60] estimated that children born in the United States in 2024 will incur nearly USD 0.5 million more in lifetime costs attributable to climate change than prior generations. Accordingly, from a policy standpoint, this is an encouraging result for intergenerational equity in climate burden-sharing, though South Korea’s strong family orientation should be considered when generalizing these findings to other contexts.

Thirdly, intention to adopt home solar has a statistically and economically significant positive effect on WTP, suggesting that South Korean residential consumers do not view DERs as substitutes for grid expansion. Rather, households intending to install home solar exhibit stronger preferences for outage avoidance and hence higher WTP for grid investment. DERs such as home solar and storage are increasingly considered outage backup solutions. Sun et al. [61], for instance, found that among more than 500,000 United States households, 63% could affordably cover approximately half of their critical energy requirements during outages with solar-plus-storage systems. Yet significant limitations remain. Gorman et al. [62] showed that solar-plus-storage systems cannot provide full backup during outages lasting one, three, or seven days, especially for heating and cooling. Thus, this finding is also encouraging from a policy perspective, indicating that support for grid expansion can coexist with growing adoption of DER.

To address potential heterogeneity between actual solar adopters and intenders, we disaggregate the solar variable into two mutually exclusive dummies: Solar_installed (n = 16 households, 1.6%; already installed) and Solar_intended (n = 202 households, 20.2%; planning but not installed, excluding overlaps). Estimating the spike model reveals that Solar_installed remains insignificant (coefficient = 0.5808, p-value = 0.210; MMWTP = KRW 1920), consistent with adopters prioritizing cost savings over outage resilience. By contrast, Solar_intended retains strong significance (coefficient = 0.7817, p-value < 0.001; MMWTP = KRW 2584), suggesting intenders are more likely to view DERs as complements to, rather than substitutes for, grid investments. This distinction implies that pro-renewable attitudes among prospective adopters—possibly driven by environmental motivations or recognition of DER limitations during nationwide blackouts—underlie the positive effect; however, this may not hold for households that have already installed home solar. Accordingly, targeted outreach to intenders may help strengthen support for tariff increases without cannibalizing grid financing.

Fourthly, educational level has a statistically and economically significant positive effect on WTP, implying that respondents with more years of schooling are more willing to pay for grid investment to avoid the described 24 h nationwide outage. From a policy perspective, this finding highlights the role of education and awareness campaigns: by enhancing residential consumers’ understanding of the risks of extreme weather and associated outages, policymakers may improve their acceptance of tariff increases needed to mitigate these risks.

Moreover, the finding that the three variables—generation, presence of children, and educational level—significantly affect WTP lends support to the conceptual framework of this study (Figure 2). This framework posits that South Korean residential consumers’ WTP decisions are shaped by uncertainty about future climate conditions and the associated risk of LSLD. Accordingly, variables more closely related to expectations about how the climate will evolve in the future, and perceptions of whether these changes will personally affect the respondent, significantly affect the WTP, while those less related do not. By the same reasoning, experiences of past outages, political ideology, and gender are not significant.

However, expectations about future climate risks vary across countries. In nations more vulnerable to climate change, residents may experience such outages more frequently, thereby facing lower uncertainty than those in South Korea. Nevertheless, these two issues—consumer acceptance of tariff increases for mitigating climate-change risks and intergenerational inequality in climate burden-sharing—remain important policy concerns globally. Studies in other national contexts or through international collaboration would therefore provide valuable insights.

As climate change progresses, South Korea itself is likely to experience more frequent LSLD. In such a case, the conceptual framework based on uncertainty about future climate impacts (Figure 2) may no longer remain valid. That is, once such outages are no longer uncertain future risks but highly probable present events, variables such as generation, presence of children, and educational level may lose their statistical significance, while the experience of an outage in the past, political ideology, and gender may emerge as significant determinants of WTP, with differences in these variables translating more clearly into preferences for avoiding outages. This highlights the fact that conceptual frameworks need to adapt as policy environments and the associated policy challenges evolve. Depending on the pace at which the impacts of climate change become manifest in South Korea, further research—either refining the current framework or developing alternative ones—will be necessary in the next 5–10 years.

4.3.3. Further Investigation of the Zero WTP Observations

Given that 51.5% of respondents (515 out of 1000 households) reported zero WTP, and that all such responses are treated as true zeros—a pivotal modeling choice underpinning this study’s conservative baseline estimate of KRW 2226 (USD 1.54) per household per month—a more detailed justification is warranted.

Table 6. Distribution of reasons for reporting zero willingness to pay.

Reasons	Number of Observations
1. Such large-scale blackouts are unlikely to occur.	87 (16.9%)
2. A blackout lasting about a day (24 h) is tolerable.	11 (2.1%)
3. Climate change and blackout issues are not important enough to prioritize.	28 (5.4%)
4. Our household cannot afford to pay additional electricity bills.	34 (6.6%)
5. Preventing large-scale blackouts should be covered by the electricity bills already paid.	233 (45.2%)
6. Additional electricity charges will not be used to prevent large-scale blackouts.	66 (12.8%)
7. Installing residential solar panels would be a better alternative.	15 (2.9%)
8. Preventing large-scale blackouts is of little value to me.	19 (3.7%)
9. There is insufficient information in the questionnaire to answer this question.	22 (4.3%)
Totals	515 (100.0%)

provides the distribution of reasons for zero WTP responses, revealing a dominance of protest motives: 45.2% (233 observations) reject additional payments (“Preventing large-scale blackouts should be covered by the electricity bills already paid”), followed by skepticism (16.9%, 87 observations: “Such large-scale blackouts are unlikely to occur”) and distrust (12.8%, 66 observations: “Additional electricity charges will not be used to prevent large-scale blackouts”). True zero responses are typically associated with affordability constraints (6.6%, 34 observations: reason 4) and low personal value (3.7%, 19 observations: reason 8).

We distinguish true zeros from protest bids using a reason-based classification, yielding both narrow and broad sensitivity analyses. The narrow view classifies reasons 4 and 8 (affordability + low value; 53 observations, 10.3% of the zero-WTP responses) as true zeros, excluding the others as protests. The broad view includes reasons 2, 3, 4, and 8 (tolerable + low priority + affordability + low value; 92 observations, 17.8% of the zero-WTP responses) as true zeros.

Table 7. Sensitivity analysis of zero willingness to pay (WTP) treatment (excluding protest bids per Table 6 reasons).

Variables	Baseline	Narrow View	Broad View
Sample size	1000	538	577
True zeros	515	53	92
Protest bids excluded	0	462	423
Mean WTP per household per year	KRW 2226 (USD 1.54)	KRW 4021 (USD 2.78)	KRW 3769 (USD 2.61)
t-values	22.84 #	23.54 #	22.36 #
95% confidence interval a	KRW 2050–2402	KRW 3691–4379	KRW 3455–4116
National WTP per year	KRW 0.60 trillion	KRW 1.08 trillion	KRW 1.02 trillion

Notes: ^a They are estimated by the method proposed by Krinsky and Robb [50]. ^# indicates that the estimate is statistically significant at the 1% level.

presents re-estimated spike model results based on subsamples that exclude protest responses.

The narrow view yields a mean monthly household WTP of KRW 4021, equivalent to 1.8 times the baseline estimate (t-value = 23.54; 95% CI: 3691–4379), while the broad view yields KRW 3769, equivalent to 1.7 times the baseline (t-value = 22.36; 95% CI: 3455–4116). Converted to annual national WTP, the narrow and broad estimates amount to KRW 1.08 trillion and KRW 1.02 trillion, respectively—both exceeding the baseline estimate of KRW 0.60 trillion—thereby supporting the policy viability of tariff-based financing. This confirms the conservative nature of the baseline estimate while providing an upper bound on WTP through the exclusion of protest responses, as informed by the explicit reason distributions reported in Table 6.

4.3.4. Caveats on Policy Extrapolation

The national-level aggregation of household WTP—yielding an estimated KRW 0.60 trillion (USD 0.4 billion) annually—serves as an illustrative benchmark for assessing the potential scale of residential contributions under the hypothetical 24 h nationwide outage scenario employed in this CV survey. This figure, derived by extrapolating the sample mean monthly WTP of KRW 2226 (USD 1.54) across South Korea’s approximately 22.7 million households, corresponds to approximately 27% of the current annual residential contribution to grid construction and maintenance (KRW 2.2 trillion; USD 1.5 billion) and 12% of the annual investment requirements outlined in the 11th Long-Term Transmission and Substation Facilities Plan (KRW 4.9 trillion; USD 3.4 billion). However, this extrapolation is subject to limitations inherent to CV methods. It assumes a uniform pass-through of stated preferences into actual tariff adjustments, does not fully capture sampling variability (with a 95% confidence interval spanning KRW 0.48–0.72 trillion), and does not account for potential behavioral responses, such as free-riding in large-scale public goods provision or anchoring effects associated with the DC bid design. Moreover, the estimate treats all zero WTP responses as legitimate (51.5% of the sample), thereby adopting a conservative lower-bound approach that aligns with the study’s policy-conservative orientation but may underestimate true valuations if protest zeros are present.

Beyond methodological constraints, the extrapolation of these CV results to policy contexts warrants caution due to multifaceted real-world financing and implementation challenges. Notably, the Korea Electric Power Corporation (KEPCO) faces accumulated losses exceeding KRW 31 trillion (USD 21.5 billion as of Q1 2025) and total debt surpassing KRW 205 trillion (USD 142.2 billion), which renders residential tariffs alone insufficient to finance grid investment without complementary revenue sources, including industrial and commercial tariffs, government subsidies, debt restructuring, or international financing. Policy support inferred from stated preferences in hypothetical settings may diverge from revealed preferences in practice, particularly in the presence of competing socioeconomic priorities such as inflation mitigation, renewable energy subsidies, and energy equity concerns. In addition, dynamic factors—such as evolving climate risks, technological advancements in distributed energy resources, and regulatory reforms—may further complicate the direct translation of these findings into policy implementation.

Consequently, while this study provides empirical evidence on consumer valuations to inform tariff design and resilience investment prioritization, it does not seek to establish comprehensive financing feasibility or constitute an unqualified endorsement of policy measures. Future research should aim to integrate multi-stakeholder cost–benefit analyses, dynamic modelling of outage probabilities under climate scenarios, and revealed preference studies to more robustly bridge the gap between stated WTP and actionable policy design.

4.3.5. Temporal Validity and Framework Evolution

The conceptual framework (Figure 2) posits that, in the absence of direct experience of LSLD, WTP formation is driven by perceptions of future climate uncertainty, thereby generating distinct effects associated with generation and the presence of children that may not be observed in populations with prior LSLD experience. This premise is inherently time-bound; as climate impacts materialize and South Korea begins to experience LSLD (with a projected 15–25% increase in exceedance probability by mid-century under RCP 8.5), experiential learning may supplant uncertainty as the primary driver of WTP, potentially attenuating the effects of generation and the presence of children while amplifying the influence of outage experience, political ideology, and gender (all currently insignificant, with p-values of 0.492, 0.231, and 0.171, respectively).

A single 24 h blackout could recalibrate public perceptions, thereby shifting tariff acceptance dynamics akin to post-disaster spikes in revealed preference, as observed following Typhoon Maemi in 2003 (localized outages affecting more than 1 million households). Nevertheless, the contribution of this study lies in its ability to capture a pre-experience baseline (March 2025 survey), thereby informing timely policy action within a narrowing window before experiential effects dominate. Post-event replications of the CV survey would enable empirical testing of the evolution of the conceptual framework, while dynamic models incorporating learning trajectories (e.g., Bayesian updating of climate beliefs) could facilitate the projection of WTP dynamics. Policymakers should therefore prioritize grid investments by leveraging current support driven by uncertainty, while anticipating a gradual convergence toward revealed preference equilibria as climate risks increasingly crystallize.

5. Conclusions

As climate change progresses, South Korea faces growing risks of LSLD caused by extreme weather, while insufficient grid capacity further amplifies the likelihood of such events. Although a major grid expansion plan was announced in 2025, securing the necessary investment remains challenging given the low acceptance of residential electricity tariff increases. Against this backdrop, this study estimates South Korean residential consumers’ WTP for grid investment to avoid a hypothetical 24 h nationwide outage. Using CV survey data from 1000 households, the study found that respondents were, on average, willing to pay an extra KRW 2226 (USD 1.54) per month.

This study makes two principal academic contributions. Firstly, in a context where research on this issue remains limited globally and virtually absent for South Korea, it provides an estimate of residential consumers’ WTP for grid investment to avoid LSLD and derives associated policy implications. Secondly, recognizing that South Korea has not yet experienced such outages, the study conceptualizes respondents’ WTP as subject to considerable uncertainty. It therefore identifies the socioeconomic factors that significantly influence WTP—particularly those related to expectations about future climate risks and perceptions of their direct personal relevance—and examines the direction and magnitude of their effects, thereby offering policy-relevant insights for South Korea. To the best of the authors’ knowledge, this represents the first attempt in the literature to apply such a conceptual framework.

From these findings, four key policy implications can be drawn. Firstly, South Korean residential consumers’ WTP for grid investment to avoid LSLD—represented in this study by a 24 h nationwide outage scenario—is substantial. When aggregated across households, the estimated WTP amounts to approximately KRW 0.60 trillion (USD 0.4 billion) annually—equivalent to about 27% of the KRW 2.2 trillion (USD 1.5 billion) in annual residential contributions to grid construction and maintenance, and approximately 12% of the KRW 4.9 trillion (USD 3.4 billion) in annual investment requirements under the grid expansion plan announced in 2025. By demonstrating the potential for greater acceptance of tariff increases, these results provide valuable input for tariff design, allocation, and utilization policies, including the determination of appropriate levels of residential tariff adjustments.

Secondly, among the socioeconomic characteristics considered, generation and the presence of children exhibit significant effects on WTP. Younger generations report higher WTP, indicating that tariff acceptance is likely to increase over time as generational replacement occurs. In addition, households with children exhibit higher WTP, suggesting that older generations are willing to contribute more than their counterparts without children to protect their offspring, who are expected to face greater risks from future outages. These findings provide policy-relevant guidance for determining the appropriate pace of annual tariff adjustments, ensuring both public acceptance of tariff increases and intergenerational equity in cost allocation.

Thirdly, intention to adopt home solar also has a significant effect on WTP. Households intending to install home solar demonstrate stronger preferences for grid expansion, suggesting that DERs are not viewed as substitutes for the grid. While DERs are increasingly promoted as outage backup solutions, their limitations, particularly evident during LSLD, underscore the continued importance of grid investment. This finding is therefore encouraging from a policy perspective, while also highlighting the need for future research and policymaking to determine the socially optimal mix of investments in DERs and grid infrastructure for cost-effective climate adaptation.

Fourthly, in contrast to the variables discussed above, past outage experience, political ideology, and gender do not exert significant effects on WTP. These characteristics are only weakly related to expectations about future climate risks and perceptions of their direct personal relevance. In South Korea, where uncertainty regarding extreme-weather-induced LSLD remains high, differences in these attributes may have been obscured by this uncertainty and therefore were not clearly manifested in the respondents’ SPs. Consequently, tariff acceptance policies and communication strategies may not need to differentiate across these attributes.

Finally, with respect to future research directions, expectations about climate-change risks vary across countries. Residents of climate-vulnerable nations, who already experience LSLD induced by extreme weather, may face lower uncertainty than those in South Korea. Furthermore, as climate change progresses and South Korea begins to experience such events with greater frequency, South Korean consumers’ uncertainty about their occurrence and potential personal impacts—and hence their WTP responses—may likewise diminish. These considerations underscore the importance of conducting comparative studies—both cross-nationally and within South Korea over the next 5–10 years. Depending on both the current status and future trajectory of climate change, the conceptual framework developed in this study may remain valid or require adaptation. Future research should therefore not only replicate this framework across diverse contexts but also refine it to capture heterogeneous national experiences and evolving expectations shaped by the dynamic nature of climate-change risks.