The Moderating Role of ESG Administration on the Relationship between Tourism Activities and Carbon Emissions: A Case Study of Basic Local Governments in South Korea

Abstract

This study examines the impact of tourism activities (TA) on carbon emissions (CE) in South Korea and investigates how ESG (environmental, social, and governance) administration power moderates these relationships. To explore four research areas—the relationship between TA and CE; variations across three years (2019–2021); the moderating effect of ESG policies; and the influence of control variables—MIMIC models were employed using secondary data from credible national institutions. The main results across the years and ESG groups (high vs. low) are as follows: (1) The positive influence of individual TA on CE ranked as food and beverage > shopping > recreation > accommodation, with no differences across the years or ESG groups. (2) ESG administration alleviated the carbon-emitting effects of TA, with significant moderation in 2019 and 2021, but not in 2020 (particularly, the effect sizes of TA were greater in low ESG groups). (3) Although tourism development stimulates TA more strongly in the high ESG groups, the CE induced by TA is significantly mitigated by ESG administrative support, resulting in smaller effects than those observed in the low groups. (4) The latter part covers diverse discussions on the influence of control variables—such as infection safety, atmospheric pollution, tourism development, income levels, green space, and local population.

1. Introduction

Currently, climate change, induced by global warming, has emerged as one of the foremost issues recognized worldwide [1]. A more specific concern pertaining to this crisis is the emission of greenhouse gases, identified as the predominant driver of global warming [2,3]. Investigations into greenhouse gas (GHG) have been conducted extensively across a diverse range of social science fields (e.g., economics [4], sociology [5], political Science [6], environmental policy [7], and ecology [8]). Such research unequivocally emphasizes the critical importance of adopting diverse approaches to adapt to climate change. The tourism industry is equally subject to the policy challenges of environmental changes. Tourism development impacts sustainable landscapes, while evolving environmental systems reciprocally alter the structure of the tourism industry [9]. From this perspective of non-recursive effects, the potential impact of the interplay between the tourism sector and climate change warrants particular attention and thorough examination [10].

Undeniably, developing an environment that promotes tourism consumption substantially contributes to regional economic growth. It is manifested in job creation, the promotion of cultural exchanges, and infrastructure enhancements in local communities [11]. Reviving tourism is especially recognized as a rapid and reliable method to expand employment after the economic downturn caused by COVID-19 [12]. However, it also remains an incontrovertible fact that these activities have led to notable detriments—such as the destruction of natural environments and ecosystems, irreversible harm to cultural heritage, and the excessive consumption of resources like water and energy [13]. In this way, numerous studies on environmental degradation have been conducted, primarily assessing carbon emissions, which are major contributors to the GHG increase.

Upon examination of research trends, recent global studies have pointed out the adverse effects of tourism activities on carbon emissions [14,15,16,17,18]. However, not all outcomes are consistent, and this variability is attributed to the temporal and geographic conditions specific to each region under study. For example, some cases were found where tourism activities reduce carbon emissions (e.g., [19,20,21]), and other cases were identified where they exhibit no considerable effects (e.g., [22,23,24]). The point is that research into developing tourism policies for carbon emissions focuses on the unique conditions of each region. Notably, an intriguing finding was the variation in results within the same country—for instance, in India, both positive and negative effects of the tourism industry on carbon emissions have been observed, along with some studies showing no significant effect (e.g., [25,26,27]). These conflicting results highlight a crucial policy issue, as they may contribute to increased confusion in tourism–environment policy decision making.

This study seeks to draw attention to the fact that the trends remain inconsistent and fragmented, due to the unique characteristics and research conditions of each region. The advantage of specific case analysis lies in its ability to offer a critical opportunity to understand regional specificity and diversification, providing a thorough examination of complex phenomena beyond simple generalizations. Given this framework, the present research plans to investigate the South Korean context in depth. The rationale for selecting this case as a research subject is as follows: First, conflicting results have similarly been reported in this country. One study [18] identified positive effects, while another [21] revealed negative impacts, and the other [22] found no significant influence. Second, substantial evidence indicates that South Korea is actively pursuing harmony between tourism and environmental policies, making it a representative country with a significant interest in environmental policy [28]. Third, South Korea has undergone relatively rapid growth in the tourism industry, accompanied by subsequent negative environmental impacts [29]. Fourth, and most crucially, there is a lack of foundational research in this area. The findings discussed herein are based on comprehensive searches conducted through academic databases such as Web of Science, Scopus, ScienceDirect, Google Scholar, and others, as of March 2024. The search terms used included ‘tourism activity/activities’, ‘carbon emissions’, ‘greenhouse gases’, and related terms. Previous research related to carbon emissions predominantly exhibits a bias toward environmental and transportation sectors (e.g., [30,31]). Several preceding works (e.g., [32,33]) have examined the link between the tourism industry and the environment; however, almost all cases focused on the influence of global warming on tourism activities.

Therefore, this study focuses on the South Korean scenario in order to address the impact of tourism activities on carbon emissions, which significantly contribute to greenhouse gases. Ultimately, it aims to produce empirical evidence that supports broader generalizations. Moreover, in this research, it has been hypothesized that the tourism industry’s impact on carbon emissions would vary across different sub-sectors. Thus, in acknowledging the distinct traits of tourism sectors such as shopping, accommodation, food and beverage, and recreation, the current study avoids the oversimplification of treating tourism as a monolithic category and reveals sector-specific implications.

At this juncture, one aspect that brings about a critical issue is that, while the global precedent literature has successfully demonstrated the damage from various tourism factors, the exploration of mitigation strategies remains persistently necessary [34,35]. Accordingly, exploring effective countermeasures is imperative not only for sustainable tourism practices, but also for fostering harmony between humans and the environment [36,37]. Another key purpose of this study is to discover strategies for alleviating the negative impacts of tourism activities on carbon emissions. Regarding the potential regulatory tool, the current research proposes leveraging an administrative keynote, grounded in environmental, social, and governance principles (ESG). The philosophy of ESG administration is fundamentally based on the principles of environmental protection, community engagement, and ethical management [38]. By applying the general concept of ESG, it is expected that policies, if guided by an ESG-oriented approach to sustainable tourism, could mitigate the environmental pollution stemming from tourism activities—i.e., the greater the ESG administrative power of local governments, the more effectively carbon emissions induced by tourism activities will be mitigated. Evidence supporting the effectiveness of ESG activities is emerging in various domains. Public governance has been shown to stimulate ESG participation and enhance companies’ environmental, social, and governance performance [39]. Additionally, the efficacy of ESG strategies has been validated in the public sector [40]. For instance, in New Zealand, the implementation of strict environmental regulations for sustainable tourism has led to a significant reduction in the environmental footprint of the tourism industry [41]. In South Korea, ESG has emerged as a focal issue receiving considerable attention in academia, business management, and public policy [42,43]. Above all, South Korea possesses diverse tourism resources, rendering the tourism industry a vital economic asset [44]. However, the environmental pollution and ecosystem destruction resulting from increased tourism pose serious challenges [45]. Consequently, the systematic approach of addressing these issues through ESG administrative measures is highly suitable for South Korea’s context. It can serve as a key strategy to promote sustainable tourism development. Thus, within this context, the South Korean case is deemed appropriate for verifying the effects of ESG initiatives.

Thus, this study develops an additional objective to verify whether the detrimental influences of tourism activities on carbon emissions are appropriately moderated by ESG policies. It is pivotal in the post-COVID-19 era to discern the specific sectors of tourism activities in terms of carbon emissions, alongside assessing the potential of an ESG-oriented administrative power to relieve the detrimental effects of such activities. If tourism expansion leads to increased carbon emissions, it necessitates a reevaluation of the revenue models for numerous tourism destinations [46]. This approach is anticipated to play a crucial role in crafting responses to climate change and devising sustainable tourism policies, thereby underpinning its significance through the generation of valuable insights.

To achieve its objectives, this study initially reviews the theoretical background to formulate research questions and develop an analytical model. Then, secondary data will be collected from nationally accredited statistical portals for empirical tests. Following data cleansing, the study will apply the Multiple Indicators Multiple Causes (MIMIC) model to data from the following three distinct periods: 2019 (pre-COVID-19), 2020 (the year COVID-19 emerged), and 2021 (post-COVID-19). Finally, a comparative analysis will be conducted to evaluate the impact of varying levels of ESG administration across these timelines.

2. Literature Review

2.1. Tourism Activities and Carbon Emissions

This research embarked on a thorough review to establish a theoretical foundation. In developing a research model that deals with the interactions among tourism activities, carbon emissions, and ESG policies, the relevant concepts were initially explored. Upon scrutinizing the universal definitions, tourism activities can be defined as the overall use and consumption of goods and services provided to tourists within the industry, such as lodging, dining, retail, entertainment, etc. [47,48]. Similarly, carbon emissions are defined as carbon dioxide and other GHG released into the atmosphere by human activities, such as the combustion of fossil fuels and industrial processes [49,50]. Adhering to the generally recognized definitions of tourism activities, this study delves into the links between tourism activities—including shopping, accommodation, food and beverage, and recreation—and carbon emissions.

Initially, the mainstream impact of tourism activities on carbon emissions was examined. As a result, a relatively straightforward positive relationship emerged between the two phenomena—i.e., as tourism activities become more active, carbon emissions are expected to increase. Anser et al. [14] examined the influence of tourism revenue on CO₂ emissions from 1995 to 2015 in G-7 countries. The findings revealed that tourism revenue contributes to an increase in CO₂ emissions, and an inverted U-shaped relationship between economic growth and CO₂ emissions was also confirmed. Balsalobre-Lorente et al. [15] investigated the nexus between economic growth, international tourism, energy consumption, and carbon emissions for OECD countries from 1994 to 2014. This study found that energy consumption, tourism, and economic growth exacerbate climate change. In addition, an inverted U-shaped relationship between international tourism and carbon emissions was discovered.

Khanal et al. [16] employed an autoregressive distributed lag bounds testing approach, using data from 1976 to 2019, to verify the long-term cointegration relationship between tourism and environmental degradation. The estimated results indicated that tourism, energy consumption, and economic growth positively influence Australia’s CO₂ emissions, both in the short and long term. They highlighted the importance of implementing sustainable tourism policies, noting that the tourism industry in Australia poses an obstacle to achieving zero carbon emissions. Udemba et al. [17] analyzed the relationship between pollution emissions, energy consumption, tourism activities, and economic growth using data from China spanning from 1995 to 2016. The results confirmed a positive causal relationship between CO₂ emissions and all of these factors. Zhang and Liu [18] examined the link between CO₂ emissions, real GDP, non-renewable and renewable energy, and tourism from 1995 to 2014 for 10 Asian countries. According to the results, non-renewable energy is the leading cause of carbon emissions, and tourism development catalyzes environmental degradation. Notably, the environmental Kuznets curve (EKC) hypothesis was rejected in both the overall and partial samples.

Consequently, there was ample evidence that various activities within the tourism industry were causing carbon emissions. However, to diagnose each trigger effect in more detail, it was determined that the tourism industry should be analyzed based on sub-sectors. In the field of tourism, sectors impacting carbon emissions typically include lodging, dining, retail, entertainment, and transportation. Several studies particularly highlight the transportation sector as the predominant contributor to carbon emissions. However, the conclusions regarding the other sectors remain provisional and heavily depend on the specific conditions of each study. The carbon footprint of tourism sectors can vary based on the classification of tourism activity. Specifically, while transportation significantly influences carbon emissions, it serves not as a distinct tourism activity, but as a support mechanism integral to tourism activities. As a result, given the scope of this research, it was not feasible to address transportation-related activities as separate from tourism activities. Instead, these were treated as being embedded within the actual tourism activities. Consequently, the subcategories of tourism activities in this work were designated as shopping, accommodation, food and beverage, and recreation before commencing the in-depth investigation. The key findings are presented as follows:

First, numerous substantial impacts of the shopping sector on carbon emissions were discovered. Kitamura et al. [51] evaluated the carbon footprint of the industry by utilizing Japan’s input–output tables with tourism consumption data. Their findings showed that transportation was the predominant source of carbon emissions, accounting for 56.3%, followed by souvenirs at 23.2%, accommodation at 9.8%, food and beverage at 7.5%, and tourism activities at 3.0%. Zhong et al. [52] conducted a top-down analysis of carbon emissions using economic and environmental metrics—including the Tourism Satellite Account. They estimated the direct carbon emissions in millions of tons as follows: transportation at 50.14, shopping at 8.14, tourism at 1.33, accommodation at 4.19, food and beverage at 4.73, and entertainment at 0.67. Liu et al. [53] analyzed tourism consumption survey data from approximately 50,000 individuals in Chengdu, China. According to their results, the tourism industry’s energy demand showed the growing significance of transportation and shopping, while the importance of food and entertainment sectors declined.

Second, the effect of the accommodation sector on carbon emissions also proved to be significant. Tang and Ge [54] developed a carbon input–output table to quantify the CO₂ emissions from various tourism-related industrial sectors. Their comparative analysis identified aviation transport (22%), accommodation (11%), clothing (10%), food services (10%), and business services (7%) as the primary contributors to CO₂ emissions. Filimonau et al. [55] addressed the shortcomings of the traditional methods of estimating tourism’s carbon footprint. They holistically considered the indirect carbon requirements from the non-use phases of tourism products or services. According to the results, the carbon intensity factors, which include transportation, accommodation, tourism activities, food and beverage, and shopping, highlighted the significant contribution of indirect GHG emissions to the total carbon footprint of holiday packages. Chen et al. [56] undertook a detailed examination of the trends and influential factors regarding tourism carbon footprints. Their findings underscored significant contributions to carbon footprints from accommodation and transportation. They also posited that insights into tourists’ behavioral preferences and characteristics are crucial for guiding future empirical research.

Third, the food and beverage sector also demonstrated an apparent effect on carbon emissions. Xiong et al. [57] utilized quarterly time-series data from 2005 to 2019 to investigate the impact of four sub-tourism sectors on GHG emissions. The analysis reported that restaurants and beverage establishments exerted a significantly greater impact compared to other sectors. Sun [58] developed a uniform carbon measurement framework to evaluate the impact of tourism on GHG emissions. An empirical analysis of the carbon footprint from tourism expenditure showed that aviation (38%) and land transportation (21%) are the primary sources, followed by significant contributions from shopping (17%) and food services (13%). Lenzen et al. [59] performed a quantitative analysis of the global carbon footprint of tourism across 160 countries, utilizing origin and destination data. Between 2009 and 2013, tourism’s global carbon footprint was determined to constitute about 8% of global greenhouse gas emissions, with transportation, shopping, and the food sector emerging as the main contributors.

Lastly, the carbon-emission-inducing effect of the recreation sector was confirmed to be significant as well. Moutinho et al. [60] analyzed the core factors influencing CO₂ emission fluctuations across five sub-sectors of the Portuguese tourism industry from 2000 to 2012. Their analysis revealed that crucial elements, including accommodation, food and beverage services, transportation, and leisure sports, significantly affect tourism intensity. Tang et al. [61] devised a factor decomposition model to explore the carbon emissions linked to energy consumption in the tourism industry. Using the case study of Wulingyuan, they integrated various sectors (such as transportation, accommodation, activities, and catering) to assess the direct carbon dioxide emissions and observed significant increases in accommodation and tourism activities. Cao et al. [62] integrated the carbon footprint concept with the life-cycle theory to examine the tourist carbon footprint in Guilin. Sector wise, the carbon footprint of tourist transportation showed a marked decreasing trend, whereas the carbon footprints of tourist accommodation, activities, and restaurants exhibited a significant expansion trend.

The current study also reviewed previous research with conflicting results. This was deemed essential for promoting a comprehensive understanding of the research topic and recognizing the complexity and variability of the phenomenon of interest. In this way, identifying potential influencing factors and conditions affecting carbon emissions helped to prevent confirmation bias regarding causality. It contributed to solidifying the theoretical framework by enhancing the reliability of the research model. The results of research cases that are inconsistent with the mainstream are as follows:

Aziz et al. [19] applied the moments quantile regression method to analyze the correlation between tourism, renewable energy, economic growth, and carbon emissions using annual data from 1995 to 2018 for BRICS countries. The results showed that tourism hurt CO₂ emissions from the 10th to the 40th quantiles but was insignificant for the remaining quantiles. Dogan and Aslan [20] conducted a panel data analysis on the EU and candidate countries from 1995 to 2011. The results revealed that energy consumption increases emission levels, while real income and tourism mitigate CO₂ emissions. In Fethi and Senyucel’s [21] study, dynamic causal relationships were tested using annual panel data from 1996 to 2016 for 50 major tourist destinations. As a result, it was found that tourism development had a positive impact on CO₂ emission levels in some countries, while, in others, it had a negative impact.

Mishra et al. [24] analyzed the dynamic link between tourism, transportation, economic growth, and carbon emissions using data from the United States. By applying the wavelet coherence technique to monthly data from 2001 to 2017, they revealed strong but inconsistent associations between the variables and confirmed the presence of significant co-movement across different time scales. Liu et al. (2019) [23] examined the dynamic link between international tourism revenue, economic growth, energy use, and carbon dioxide (CO₂) emissions in Pakistan from 1980 to 2016. The key results showed that tourism revenue did not significantly impact environmental quality, while economic growth and energy consumption were the primary determinants of CO₂ emissions. Dogru et al. [22] investigated the relationship between tourism development, economic growth, renewable energy consumption, and carbon dioxide (CO₂) emissions in OECD countries and reported results that contradict existing studies. For instance, tourism development had a negative impact on CO₂ emissions in Canada, the Czech Republic, and Turkey, while it had a positive and significant impact in Italy, Luxembourg, and Slovakia. Additionally, Belgium, France, New Zealand, and Slovakia demonstrated a transition to sustainable tourism practices.

The information so far has been organized in the following ways: Conflicting results persist regarding the impact of the tourism industry on carbon emissions. These disparities imply a lack of a standardized method to confirm causality between these two factors. The point is that the inducing effect of tourism activities on carbon emissions is a global phenomenon. Accordingly, the approach to reducing the tourism industry’s influence may be premature.

2.2. The Moderating Role of ESG Administration

The current study concentrates on ESG administrative efforts as a mitigation measure for the negative effects of tourism activities on carbon emissions. ESG, an acronym for environmental, social, and governance factors, can be defined as a set of criteria for evaluating an organization’s operations and business model concerning its sustainability and ethical impact [63]. Accordingly, ESG administration can denote the systematic and strategic approach that an organization or entity employs to integrate ESG criteria into its policies and practices, aiming to ensure responsible governance at all operational levels [64,65]. Recently, increasing scrutiny of how corporations impact social welfare has led to a surge in analyzing business activities across all sectors through the ESG framework [66]. Numerous nations are steadfastly enhancing their efforts to promote development in harmony with the core values of sustainability, rooted in ESG principles [67].

Baratta et al. [68] conducted a bibliometric analysis to examine the impact of ESG practices on carbon emissions reduction across major industrial sectors. The key findings emphasize the importance of collective efforts in adopting ESG measures and advocate for the standardization of ESG strategic variables. Cong et al. [69] investigated the correlation between ESG investment and carbon dioxide emissions. The results showed that a 1% increase in environmental investment results in a 0.24% reduction in carbon dioxide emissions and a 0.56% decrease in carbon emission intensity. Furthermore, they suggested that the effects of ESG investment differ significantly between developed and less-developed regions. Alandejani and Al-Shaer [70] evaluated the impact of political uncertainties on firms’ ESG performance and efforts to reduce carbon emissions. Utilizing data from 2013 to 2020 in the United States, China, and the United Kingdom, they found that higher levels of uncertainty lead to increased corporate engagement in ESG initiatives and the setting of emission reduction targets.

Lee and Cho [71] investigated the relationship between carbon emissions, carbon disclosures, and corporate value through the environmental component of ESG scores. An analysis of data from 841 Korean firms showed that companies with superior environmental performance tend to voluntarily disclose their carbon emissions, stressing the importance of historical context in analyzing environmental policy. Ionescu et al. [72] examined the association between the ESG factors of tourism companies and their market value, assessing their predictive power for performance. The analysis indicated that governance factors play a crucial role in influencing market value, a trend observed consistently across various geographic regions. Kumar [73] addressed the influence of ESG sustainability practices within the tourism industry. The study revealed that, faced with increasing economic and political uncertainties at the national level, tourism companies actively seek to build positive relationships with stakeholders, diligently focusing on their societal and environmental contributions and the resulting public perception.

In addition, Bae et al. [74] analyzed the ESG management attributes of food and beverage firms and found their significant impact on customer brand loyalty and word-of-mouth endorsements. Buallay et al. [75] examined the correlation between ESG scores and performance metrics—operational, financial, and market—within the tourism sector. Utilizing data from 1375 instances across 37 countries from 2008 to 2017, they demonstrated significant causality between ESG scores and both operational and market performance. Yoon et al. [76] employed ESG scores to evaluate CSR performance and its influence on corporate value assessment. The findings revealed that CSR practices have a positive effect on a company’s market presence. Additionally, they noted that, in environmentally sensitive industries, the impact of CSR on value creation is less pronounced compared to other industries.

From the perspective of public local governments responsible for managing tourist destinations, the adoption of ESG administrative policies—emphasizing responsibility, equity, environmental friendliness, and sustainability—is anticipated to mitigate the environmental degradation caused by tourism activities. In South Korea, this policy direction has become increasingly pronounced in recent years, suggesting a potential uptrend in its effectiveness [42,43].

2.3. Control Variables

This study also contemplated the key control variables that must be managed to prevent their influence on the primary effects within the research model. Accordingly, the model structure was expanded to comprehensively account for exogenous variables (which are external to the independent variable and affect the dependent variable; instances where variables influence both the independent and dependent variables; and antecedent variables that exert a significant impact preceding the independent variables) [77].

Initially, the infection safety level was considered an exogenous variable influencing tourism activities. The local infection safety level refers to the degree to which local governments protect their communities from the risk of infection within a particular environment (such as medical institutions, public spaces, etc.) [78]. Numerous studies have consistently demonstrated that COVID-19, in particular, exerts a negative impact on local tourism initiatives (e.g., [79,80]). This decline was primarily due to reduced tourist arrivals and plummeting demand, driven by social anxieties and restrictions on gatherings [78,81].

A substantive linkage exists between the atmospheric environment and tourism demand. Specifically, air pollution can be defined as the presence of harmful substances in the atmosphere, such as chemicals, particulate matter, or biological materials, which can negatively affect human health, environmental quality, and the climate [82]. Climatic fluctuations, acting as a deterrent to outdoor activities, limit local tourism demand; furthermore, air pollutants (such as fine particulate matter, ultrafine dust, and yellow dust) are inversely related to the attractiveness of outdoor tourist sites and the propensity for domestic travel [83,84]. Consequently, atmospheric pollution levels significantly influence tourists’ decisions regarding engagement in local tourism activities [85,86].

The local tourism development level can be defined as the quality of the infrastructure, services, and activities available to facilitate the tourist experience within a specific region [87]. A significant influence on tourism demand is exerted by the extent of local tourism development, including the distribution of tourism resources, infrastructure, local engagement in tourism, policy frameworks, human capital, and promotional efforts [88,89]. Regions endowed with abundant tourism resources and advanced infrastructure are more likely to attract a higher number of tourists [90,91].

The local income level can be defined as the average economic capacity or earnings of individuals or households within a specified local area [92]. Kuznets [93] hypothesized that income inequality initially intensifies during the early stages of economic growth but begins to decrease after surpassing a certain income threshold, a phenomenon described as the Kuznets curve. With the rise of ecological concerns, a similar relationship between economic growth and environmental impact was conceptualized, leading to the development of the environmental Kuznets curve (EKC) [94]. The EKC hypothesis suggests that the degradation of the natural environment initially worsens with economic growth but improves after achieving a certain income level [95].

In light of this perspective, it is anticipated that an increase in income at the early stages correlates with a rise in carbon emissions; however, beyond a certain point, pollution levels decrease, exhibiting an inverted U-shaped curve [96]. This study plans to further investigate the EKC hypothesis as a foundation to enhance the validity of the research model. Additionally, it is a well-known fact that tourism activities are influenced by disposable income [97], therefore, the present study has treated the income level as an exogenous variable that influences both tourism activities and carbon emissions.

Additionally, green spaces play a key role in mitigating the urban heat island effect. Local green spaces can be defined as areas within a regional setting that are designated for natural elements, intended to enhance environmental quality and provide aesthetic benefits to the community [98]. Biological sequestration by plants and soil serves as a carbon sink, contributing to reducing surface temperatures and enhancing urban resilience [99]. Various studies support the role of urban green infrastructure in maintaining carbon balance (e.g., [98,100]).

The correlation between population density and carbon emissions elucidates the profound impact of human activities on the global climate system [101]. The local population size refers to the total number of individuals residing within a specific geographical or administratively defined area at a given time [102]. The basis for this relationship lies in the increase in energy consumption, waste production, and transportation needs associated with higher population densities [103]. Elevated population densities lead to increased vehicular usage and greater electricity consumption for heating and cooling in residential and commercial buildings [104]. Collectively, these factors result in an increased reliance on fossil fuels, thereby contributing to higher carbon emissions [105].

As a result of reviewing previous research, it has been decided to address a total of six control variables in this work. Incorporating additional exogenous variables into the research model could also be beneficial. However, given the conditions of the theoretical background, the data availability, and analytical methods, the current selection has been deemed optimal.

Based upon the above theoretical considerations, meaningful and realistic analysis results are expected to be derived only when the influence of each of the four sub-tourism industry sectors—such as shopping, accommodation, F&B, and recreation—is considered. Thus, an essential structural model is created where the four lower tourism sectors are subordinated to the upper sector, tourism activities, which in turn influence carbon emissions through electricity and gas use. Here, ESG administrative power is anticipated to control carbon emissions from tourism activities effectively. The model is also expanded to a structure in which the exogenous variables (including the six control variables mentioned above) that influence the central causal relationship are controlled. Additionally, a final model can be set up to compare the periods before and after the outbreak of COVID-19. From an empirical perspective, in-depth consideration of a model embodying the structure between these variables is necessary. Additionally, the feasibility of collecting the required data is another crucial factor in applying the methodology.

3. Methodology

3.1. Research Questions

Empirical evidence from prior research indicates that carbon emissions are expected to be positively influenced by tourism activities [14,15,16,17,18], such as shopping [51,52,53], accommodation [54,55,56], food and beverage [57,58,59], and recreation [60,61,62]. It has also been proposed that the ESG administrative framework could intervene in the relationship between tourism activities and carbon emissions, suggesting that the greater the ESG administrative power of local governments, the more effectively carbon emissions induced by tourism activities will be mitigated [42,43,68,69,70,71,72]. Furthermore, as for improving the precision of the findings, this study dealt with essential control variables, such as infection safety [78,79,80,81], atmospheric pollution [83,84,85,86], tourism development [88,89,90,91], income level [95,96,97], green space [98,99,100], and local population [101,103,104,105]. Consequently, the research questions (RQ) that the current study has to address are as follows:

RQ1. 

How do tourism activities (i.e., shopping, accommodation, food and beverage, and recreation) correlate with carbon emissions in South Korea?

RQ2. 

Does the impact of tourism activities on carbon emissions vary across the years 2019, 2020, and 2021—reflecting the period before and after COVID-19?

RQ3. 

How are the influences of tourism activities on carbon emissions moderated by the implementation of ESG policies?

RQ4. 

What control variables (including infection safety, atmospheric pollution, tourism development, income level, green space, and local population) are effective in clarifying the relationship between tourism activities, carbon emissions, and ESG administration?

3.2. Data Collection

Tourism consumption expenditure is utilized as a proxy indicator for tourism activities [106]. The data were obtained from credit card spending amounts officially provided by the Korea Tourism Organization (KTO). These figures consolidate transactions from two major banks and are compiled on a monthly and district basis across various sectors—including accommodation, shopping, food and beverage, and recreation. The accommodation sector comprises hotels, guesthouses, and other lodging facilities. The shopping sector includes purchases of tourism-based souvenirs, transactions at large shopping malls, and similar establishments. The food and beverage sector encompasses general restaurants, cafes, and bars. The recreation sector broadly covers tourist amusement facilities, such as water sports, ski resorts, fishing spots, etc.

Carbon emissions have been selected as the core outcome variable for this study, since greenhouse gases comprise six types (including carbon dioxide, methane, nitrous oxide, hydrofluorocarbons, and perfluorocarbons), among which carbon dioxide accounts for over 80% of total greenhouse gas emissions [107,108]. In this paper, the annual building energy consumption data for electricity and city gas were measured in tonnes of oil equivalent (toe). Consumption data for oil and renewable energy sources, excluding electricity and city gas, were not available by administrative region and thus could not be utilized in this study. The data were collected under the auspices of the Korea Real Estate Board and published by the National Statistical Office (NSO). Accordingly, the energy consumption data were converted into carbon dioxide emissions (tCO₂) for analysis.

The data relevant to ESG administration power were obtained from the Korea ESG Evaluation Institute—a pioneering initiative that extends the focus to local governments in South Korea. The dataset originates from a custom evaluation model developed by the institute, aligned with the Korean Sustainable Development Goals. The assessment framework involves a mix of indicators across environmental (14 items), social (23 items), and governance (4 items) dimensions. This model employs a combination of quantitative indicator assessments (70% weight) and qualitative policy evaluations (30% weight). This unique index, measured only once around 2021, accounts for the cyclical nature of local government planning in Korea and the potential influence of prior activities on subsequent ESG evaluations.

In particular, the basic planning years for South Korea’s local governments are set in four-year cycles (2018–2022). This suggests that performances prior to the evaluation period could impact the assessment for that period (i.e., policies, projects executed, or outcomes in 2019 could be reflected in the ESG evaluation indicators for 2020 and 2021). Therefore, in order to ensure the continuity and comprehensiveness of the analysis, the temporal range was set from 2019 to 2021, a three-year period, treating the distinctive characteristics of each region within this time as unique variables.

In South Korea, the connections between the COVID-19 outbreak, vaccine distribution, and the post-COVID-19 policies and local tourism spending mean that it is prudent to control for annual levels of infection safety [109]. The dataset for this analysis was collected by the Ministry of Public Administration and Security and officially released by the NSO. It includes annual district-level regional safety indices (such as traffic accidents, fire, crime, living safety, suicide, and infectious diseases), and the infectious disease safety ratings were chosen as a control variable. In this work, reverse coding was utilized on a scale from 1 to 5, where a score of 5 represents the highest level.

The foundational data for atmospheric pollution were sourced from the official annual and municipal air quality indices provided by the Korea Environment Corporation (KEC). These indices cover pollutants such as particulate matter (PM10 and PM2.5), nitrogen dioxide (NO₂), ozone (O₃), sulfur dioxide (SO₂), and carbon monoxide (CO). Utilizing these six types of pollutants, the comprehensive air quality index (CAI) was calculated using the method published by the KEC. The CAI is an air quality assessment indicator developed to reflect the human health impact and perceived pollution level. It integrates the concentrations of six types of air pollutants to provide a comprehensive measure of air quality. The CAI is scored from 0 to 500, with higher scores indicating poorer air quality.

The Tourism Activation Index (TAI) is an official dataset provided by the KTO’s Korea Tourism Data Lab. This index is designed to comprehensively assess the efficiency and outcomes of the tourism industry at the local government level. It serves as a composite indicator of tourism vitality, scored from 0 to 100, in which higher scores indicate greater activation of local tourism. In light of the study’s purpose, the TAI was considered an appropriate measure for gauging regional tourism development. Thus, the average annual values of each municipal area, provided monthly, were utilized for analysis. There was theoretical clarity regarding the exogenous effects of income on tourism activities and carbon emissions. Income data were collected from the National Tax Service (NTS). The current study utilized the total income amount (in KRW million) reported in the annual, municipal-level comprehensive income tax filing status data.

The green space ratio was derived from official urban planning status data. This dataset, updated annually and broken down by municipal levels, was systematically gathered by the Korea Land and Geospatial Informatix (LX) Corporation. It quantifies the percentage of green spaces relative to the total urban areas. Regarding population data, this study adopted figures from the population census conducted by the NSO, incorporating annual and municipal resident numbers. This dataset includes all domestic and foreign residents living in each region. Due to its reliability, the data serves as a sampling frame for various household surveys. Hence, its inclusion in this analytical framework was considered appropriate.

Table 1. Summary of input variables.

Variables	Data Contents	Unit	Sources
Electricity	Carbon emissions induced by electricity usage	toe	Korea Real Estate Board
Gas	Carbon emissions induced by city gas usage
Shopping	Spending amount on shopping activities	KRW	Korea Tourism Organization
Accommodation	Spending amount on accommodation facilities
Food and Beverage	Spending amount on food and beverage services
Recreation	Spending amount on recreation activities
ESG Administration	ESG activity evaluation for local governments	Ranking	Korea ESG Evaluation Institute
Infection Safety	Local infection safety evaluation	5-point rating	Ministry of Public Admin. and Security
Atmospheric Pollution	Local air pollution measurements	Max. 500	Korea Environment Corporation
Tourism Development	Local tourism activation assessment	%	Korea Tourism Organization
Income Level	Local income level aggregation	KRW	National Tax Service
Green Space	Green area ratio to urban area	%	Land and Geospatial Informatix Corp.
Local Population	Local population estimates	People	National Statistical Office

below organizes the operational definitions of the variables mentioned above and their sources.

3.3. Data Transformation

Prior to the comprehensive analysis, the consistency of the entire dataset was verified. Except for a few inevitable missing values in the atmospheric pollution data—due to observational conditions—no missing data were identified. These missing values were imputed using a Bayesian spatial linear regression model. To reduce noise, outliers were adjusted to their upper and lower limits based on the ‘Q3 + 1.5 × IQR’ and ‘Q1 − 1.5 × IQR’ formulas, respectively. Accordingly, standardized values for variables—including tourism consumption activities, infection safety, atmospheric pollution, tourism development, income level, green space, and local population—were computed in advance.

Due to concerns over regional political conflicts and uneven development, the ESG administrative level, scored out of 100, remains undisclosed; therefore, only ordinal data were accessible. This means that detailed scores for each were not available; therefore, the information available consists solely of a ranked list of municipalities. Utilizing the most feasible method for clear categorization, this study distinguished between high-level and low-level ESG groups (henceforth, groups with varying ESG compliance levels are termed ‘high ESG group’ and ‘low ESG group’ for brevity) based on median split [110,111], assigning binary variables of 0 (low) and 1 (high) to them.

The carbon dioxide emissions (tCO₂) for electricity and city gas consumption were calculated according to the following procedure: Converting electricity and city gas consumption into tCO₂ necessitates understanding the concepts of tonne of oil equivalent (toe), tC, and tCO₂. The International Energy Agency defines toe as a unit to standardize the calorific value of different energy sources to that of crude oil, which is also referred to as energy consumption. To calculate tC, the toe of the consumed fuel is multiplied by the carbon emission factors specific to each type of fuel. The carbon emission factor for city gas is set at 0.637, as proposed by the IPCC [112], and at 0.459 for electricity, as provided by the Korea Electric Power Corporation [113].

Lastly, the calculation of tCO₂ involves multiplying tC by the ratio of the molecular weight of carbon dioxide (44) to the atomic weight of carbon (12). Thus, the data of tCO₂ for electricity and city gas consumption were derived following the aforementioned three-step process.

3.4. Analytical Modeling

In addressing the research questions posed, this study rigorously evaluated various methodologies to analyze the interrelationships among key identified variables. Consideration was given as to whether the control variables could be adequately incorporated, year-over-year comparative analyses could be executed, and sub-variables of the main factors could be effectively managed. Consequently, the MIMIC model was selected as the most suitable approach, due to its comprehensive structure that aptly accommodates the complexity of these relationships [114,115]. Figure 1 below illustrates the analytical framework of the MIMIC model constructed for this study.

The MIMIC model’s mathematical formulation can be described as follows.

$${\mathrm{Y}}_{\mathrm{i},\mathrm{j}}={\mathsf{\lambda }}_{\mathrm{j}}{\mathsf{\eta }}_{\mathrm{i}}+{\mathsf{ϵ}}_{\mathrm{i}}$$

(1)

$${\mathsf{\eta }}_{\mathrm{i}}={\mathsf{{\rm Y}}}_{\mathrm{k}}{\mathrm{X}}_{\mathrm{i},\mathrm{k}}+{\mathsf{\zeta }}_{\mathrm{i}}$$

(2)

In these equations, ${\mathrm{Y}}_{\mathrm{i},\mathrm{j}}$ represents an observed indicator j for construct i, serving as measurement items of the construct; ${\mathrm{X}}_{\mathrm{i},\mathrm{k}}$ indicates the observation variable representing covariate k, which influences construct i as a precursor; ${\mathsf{\eta }}_{\mathrm{i}}$ means the carbon emissions symbolized by construct i; $\mathsf{\lambda }$ and $\mathsf{{\rm Y}}$ the coefficient vectors to be estimated; and $\mathsf{ϵ}$ and $\mathsf{\zeta }$ denote the error terms.

In Equation (2), ${\mathsf{\xi }}_{\mathrm{i}}$ (i = 2) encompasses tourism activities and carbon emissions. Equation (1) specifies ${\mathrm{x}}_{\mathrm{i},\mathrm{k}}$ (k = 6) to include factors such as infection safety, air pollution, tourism development, income level, green space, and local population. Using the foundational MIMIC model established in Equations (1) and (2), this study delineates the structural connections between the two constructs, with respect to six covariates.

Given the analytical structures outlined in the equations, the latent variables ${\mathsf{\eta }}_1$ and ${\mathsf{\eta }}_2$ in the MIMIC model can be described with Equations (3) and (4). Here, ${\mathsf{\beta }}_{21}$ represents the influence of ${\mathsf{\eta }}_1$ on ${\mathsf{\eta }}_2$.

$${\mathsf{\eta }}_1={\mathsf{\gamma }}_1{\mathrm{X}}_1+{\mathsf{\gamma }}_2{\mathrm{X}}_2+{\mathsf{\gamma }}_3{\mathrm{X}}_3+{\mathsf{\gamma }}_4{\mathrm{X}}_4+{\mathsf{\zeta }}_1$$

(3)

$${\mathsf{\eta }}_2={\mathsf{\beta }}_{21}{\mathsf{\eta }}_1+{\mathsf{\gamma }}_5{\mathrm{X}}_4+{\mathsf{\gamma }}_6{\mathrm{X}}_5+{\mathsf{\gamma }}_7{\mathrm{X}}_6+{\mathsf{\gamma }}_8{\mathrm{X}}_7+{\mathsf{\zeta }}_2$$

(4)

In addition, the observation variables Y₁ to Y₆ would be modeled according to Equation (5), with each Y serving as an effect indicator for its corresponding $\mathsf{\eta }$, as follows:

$$\begin{gathered} {\mathrm{Y}}_1={\mathsf{\lambda }}_1{\mathsf{\eta }}_1+{\mathsf{ϵ}}_1 \\ {\mathrm{Y}}_2={\mathsf{\lambda }}_2{\mathsf{\eta }}_1+{\mathsf{ϵ}}_2 \\ {\mathrm{Y}}_3={\mathsf{\lambda }}_3{\mathsf{\eta }}_1+{\mathsf{ϵ}}_3 \\ {\mathrm{Y}}_4={\mathsf{\lambda }}_4{\mathsf{\eta }}_1+{\mathsf{ϵ}}_4 \\ {\mathrm{Y}}_5={\mathsf{\lambda }}_5{\mathsf{\eta }}_2+{\mathsf{ϵ}}_5 \\ {\mathrm{Y}}_6={\mathsf{\lambda }}_6{\mathsf{\eta }}_2+{\mathsf{ϵ}}_6 \end{gathered}$$

(5)

3.5. Analysis Procedure

The primary methods and processes for analysis are as follows: Initially, the present research will examine the basic statistical measures and correlations for each variable along with data cleansing. Subsequently, the validity and reliability of the constructs will be assessed through factor analysis and internal consistency checks. Next, the influence of the key variables and the overall suitability of the model will be appraised based on MIMIC (multiple indicators and multiple causes) models. Finally, this study will explore the moderation effects across different years (2019, 2020, and 2021) and varying levels of ESG (environmental, social, and governance) policy stances (high vs. low).

4. Results

4.1. Descriptive Statistics and Difference Tests for Input Variables

Table 2. Descriptive statistics based on year and ESG division.

		Total	2019(a)		t-Test’s p-Value	2020(b)		t-Test’s p-Value	2021(c)		t-Test’s p-Value	F-Test
			High	Low		High	Low		High	Low
			Mean	Mean		Mean	Mean		Mean	Mean		High		Low
			(S.D.)	(S.D.)		(S.D.)	(S.D.)		(S.D.)	(S.D.)
CE	ELC	108.921	100.710	118.493	0.084	97.645	116.139	0.063	101.222	119.315	0.080	0.067	n.s.	0.057	n.s.
		(80.426)	(84.572)	(77.185)		(80.754)	(75.571)		(84.482)	(77.985)
	GAS	124.861	105.765	143.368	0.014	104.110	141.926	0.011	107.567	146.429	0.011	0.027	n.s.	0.045	n.s.
		(120.114)	(118.438)	(121.835)		(114.675)	(118.844)		(118.335)	(121.602)
TA	SHP	53.456	55.399	57.515	0.786	55.196	55.349	0.984	49.577	47.698	0.784	0.355	n.s.	1.082	n.s.
		(58.711)	(63.726)	(59.103)		(63.614)	(57.330)		(58.498)	(49.353)
	ACM	3.246	3.503	4.112	0.247	3.488	4.110	0.274	2.066	2.198	0.593	6.683	***	10.375	***
		(3.784)	(3.931)	(4.357)		(4.281)	(4.682)		(2.116)	(1.783)		c < b = a		c < b = a
	FNB	95.637	97.020	122.798	0.027	95.437	120.459	0.027	61.535	76.572	0.035	8.131	***	12.509	***
		(83.151)	(89.227)	(94.230)		(86.582)	(90.783)		(55.442)	(56.620)		c < b = a		c < b = a
	RCN	3.799	4.600	4.977	0.484	3.998	4.318	0.490	2.263	2.639	0.139	14.555	***	16.501	***
		(3.572)	(4.445)	(4.045)		(3.780)	(3.539)		(1.970)	(2.035)		c < b = a		c < b = a
CV	SAF	3.067	3.304	2.848	0.001	3.184	2.952	0.101	3.280	2.832	0.001	0.426	n.s.	0.414	n.s.
		(1.124)	(1.109)	(1.129)		(1.081)	(1.149)		(1.075)	(1.120)
	POL	86.982	110.574	106.409	0.087	82.594	79.292	0.044	71.651	71.374	0.758	232.760	***	243.174	***
		(21.036)	(19.996)	(18.349)		(13.682)	(12.049)		(7.882)	(6.224)		c < b < a		c < b < a
	DEV	46.846	43.508	49.566	0.058	43.229	50.243	0.033	43.858	50.670	0.046	0.018	n.s.	0.058	n.s.
		(26.101)	(25.259)	(25.066)		(25.791)	(25.929)		(27.229)	(26.506)
	INC	649.522	683.421	542.845	0.106	699.502	549.900	0.093	794.159	627.302	0.101	0.598	n.s.	0.860	n.s.
		(733.237)	(814.112)	(525.106)		(829.968)	(541.197)		(944.368)	(623.435)
	GRN	65.002	69.293	60.930	0.001	69.152	60.892	0.001	68.969	60.775	0.002	0.011	n.s.	0.002	n.s.
		(20.573)	(17.497)	(22.620)		(17.471)	(22.633)		(17.425)	(22.673)
	POP	199.915	176.797	223.387	0.016	177.891	222.264	0.022	178.045	221.109	0.027	0.003	n.s.	0.007	n.s.
		(153.576)	(146.102)	(157.497)		(147.893)	(157.049)		(148.918)	(156.774)

Note (1) CE = Carbon Emissions, TA = Tourism Activities, CV = Control Variables, ELC = Electricity, GAS = City Gas, SHP = Shopping, ACM = Accommodation, FNB = Food and Beverage, RCN = Recreation, SAF = Infection Safety, POL = Atmospheric Pollution, DEV = Tourism Development, INC = Income Level, GRN = Green Space, POP = Local Population; Note (2) Units for CA: 1000 tCO₂, Units for TA: KRW 100 M, Units for IL: KRW 1 M, Units for LP: 1 K persons, USD 1 ≈ KRW 1200; Note (3) *** p < 0.01; n.s. = not significant.

presents the descriptive statistics for the analysis data over three years and for both high- and low-level ESG groups, along with the mean difference test results. Firstly, significant differences in carbon emissions between the high and low ESG groups were observed in terms of electricity usage (p < 0.10) and gas usage (p < 0.05), a trend that remained consistent over the three-year period. Regarding tourism consumption, statistical significance was found only in the food and beverage activities (p < 0.05), while no differences were revealed in the other attributes (shopping, accommodation, and recreation). However, according to Duncan’s post hoc tests ($\mathsf{\alpha }$ = 10%), differences across the years remained in accommodation, food and beverage, and recreation, with both high and low groups. It showed relatively higher consumption in 2019 and 2020 compared to 2021. Evidently, these results indicate the prolonged economic downturn in South Korea due to COVID-19 [79,80,81]. As for the control variables, the infection safety levels indicated that regions with high ESG scores maintained more stable conditions as opposed to those with low scores (p < 0.01), although the disparity slightly diminished during the COVID-19 outbreak in 2020.

The atmospheric pollution showed improvement from 2019 to 2021 in both groups. It mirrored a global trend of environmental quality enhancement due to COVID-19’s impacts [116]. Variations in tourism development levels between the high and low ESG groups were notable each year in 2019 (p < 0.10), 2020 (p < 0.05), and 2021 (p < 0.05), with the low ESG group exhibiting consistently higher development. Inequality in income levels was marginally significant at a 10% level in 2020 but remained generally stable across the study period. Regarding the green space, the distribution was initially higher in the low ESG group in 2019 (p < 0.01); however, this pattern shifted in the subsequent years (p < 0.01), with the high ESG group showing greater ratios. Additionally, the population size was consistently larger in the low ESG group (p < 0.05).

4.2. Correlations for High- and Low-Level ESG Groups

Table 3. Correlations for ESG division.

	Carbon Emissions		Tourism Activities				Control Variables
	[1]	[2]	[3]	[4]	[5]	[6]	[7]	[8]	[9]	[10]	[11]	[12]
	ELC	GAS	SHP	ACM	FNB	RCN	SAF	POL	DEV	INC	GRN	POP
[1]		0.841 ***	0.705 ***	0.293 ***	0.809 ***	0.646 ***	0.242 ***	−0.180 ***	0.833 ***	0.726 ***	−0.191 ***	0.851 ***
[2]	0.821 ***		0.631 ***	0.231 ***	0.781 ***	0.522 ***	0.214 ***	−0.192 ***	0.786 ***	0.685 ***	−0.330 ***	0.899 ***
[3]	0.778 ***	0.721 ***		0.258 ***	0.756 ***	0.566 ***	0.184 ***	−0.198 ***	0.675 ***	0.643 ***	−0.260 ***	0.690 ***
[4]	0.462 ***	0.492 ***	0.490 ***		0.460 ***	0.412 ***	0.013 n.s.	−0.040 n.s.	0.410 ***	0.159 ***	−0.115 **	0.221 ***
[5]	0.813 ***	0.798 ***	0.810 ***	0.560 ***		0.698 ***	0.203 ***	−0.298 ***	0.811 ***	0.648 ***	−0.272 ***	0.773 ***
[6]	0.648 ***	0.654 ***	0.617 ***	0.458 ***	0.706 ***		0.160 ***	−0.315 ***	0.575 ***	0.488 ***	−0.045 n.s.	0.581 ***
[7]	0.269 ***	0.259 ***	0.320 ***	0.183 ***	0.307 ***	0.198 ***		−0.092 *	0.148 ***	0.323 ***	0.119 **	0.293 ***
[8]	−0.195 ***	−0.230 ***	−0.238 ***	−0.185 ***	−0.338 ***	−0.315 ***	−0.171 ***		−0.164 ***	−0.153 ***	0.045 ***	−0.191 ***
[9]	0.845 ***	0.825 ***	0.796 ***	0.521 ***	0.777 ***	0.670 ***	0.321 ***	−0.184 ***		0.697 ***	−0.324 ***	0.775 ***
[10]	0.577 ***	0.611 ***	0.588 ***	0.250 ***	0.594 ***	0.422 ***	0.340 ***	−0.145 ***	0.639 ***		−0.082 n.s.	0.768 ***
[11]	−0.281 ***	−0.376 ***	−0.263 ***	−0.155 ***	−0.277 ***	−0.134 ***	−0.084 n.s.	0.067 n.s.	−0.281 ***	−0.179 ***		−0.213 ***
[12]	0.875 ***	0.854 ***	0.776 ***	0.450 ***	0.842 ***	0.643 ***	0.298 ***	−0.205 ***	0.858 ***	0.658 ***	−0.305 ***

Note (1) The variable names are as listed in Table 2; Note (2) The figures in the upper triangle and lower triangle indicate low and high ESG groups, respectively; Note (3) *** p < 0.01, ** p < 0.05, * p < 0.10; n.s. = not significant.

provides the results of testing the correlations among the variables for both high and low ESG groups. The results for the aggregated three-year period showed no differences in the general trend of yearly correlations. As a result, almost all relationships were significant at the 1% level, except for a few isolated cases. However, the correlation between infection safety and atmospheric pollution—specifically in relation to accommodation in the low ESG group—demonstrated inconsistency over the years. The directionality of the theoretical signs was consistent across the board, and no indicators of multicollinearity were detected among the explanatory variables. Consequently, it was determined that there are no critical issues that would hinder the execution of causality analysis.

4.3. Estimating MIMIC Model for Pooled Data

Table 4. Estimates of MIMIC model for the entire group.

Var1	Var2	Std. Est.	Est.	S.E.		AVE	CR	α
Carbon Emissions (CE)	Electricity	0.974	1.000			0.814	0.897	0.908
	Gas	0.823	0.844	0.023	***
Tourism Activities (TA)	Shopping	0.827	1.000			0.597	0.851	0.828
	Accommodation	0.523	0.600	0.040	***
	Food and Beverage	0.934	1.130	0.034	***
	Recreation	0.715	0.865	0.039	***
TA	CE	0.404	0.476	0.039	***
Infection Safety	TC	0.026	0.021	0.013	*
Atmospheric Pollution		−0.040	−0.033	0.013	***
Tourism Development		0.696	0.575	0.032	***
Income Level		0.238	0.196	0.030	***
Income Level	CE	0.216	0.210	0.037	***
Income Level2		−0.144	−0.140	0.012	***
Green Space		−0.030	−0.030	0.008	***
Local Population		0.493	0.480	0.028	***
Model Fit	χ2 = 253.155 (df = 41, p = 0.000), RMR = 0.025, SRMR = 0.026,
	GFI = 0.952, AGFI = 0.893, CFI = 0.977, RMSEA = 0.083

Note (1) Pooled data N = 750; Note (2) *** p < 0.01, * p < 0.10.

presents the results of the MIMIC model analysis for the pooled sample. Basically, the model was estimated based on the maximum likelihood (ML) method; moreover, due to difficulties in fully achieving multivariate normality, as indicated by the Kolmogorov-Smirnov test results, the Bollen-Stine bootstrapping method was employed for robust ML estimation [117]. Additionally, confirmatory factor analysis (CFA) of the model structure revealed no estimation violations, such as standard errors exceeding 2.50, Heywood cases, or standardized coefficients above 1.00.

Furthermore, the average variance extracted (AVE) and composite reliability (CR) for carbon emissions and tourism activities, along with Cronbach’s alpha, met the thresholds of 0.50, 0.70, and 0.70, respectively, ensuring both validity and reliability. Regarding model fit assessment, the χ² value was reported as 253.155 (df = 41, p < 0.001); however, given the sensitivity of the χ² statistic to sample size and the number of observation variables, its significance does not solely determine model fit. Therefore, the current study diagnosed its fitness considering absolute, incremental, and parsimonious fit indices collectively, with SRMR = 0.026, GFI = 0.952, AGFI = 0.893, CFI = 0.977, and RMSEA = 0.083, deeming the model acceptable. The RMSEA, along with the CFI and SRMR, is recognized as one of the most favored fit indices in path analysis, CFA, and structural equation modeling (SEM) [118,119]. The RMSEA value can primarily be affected by the model’s complexity and the division of the sample, which reduces degrees of freedom and potentially enlarges the RMSEA. This challenge is frequently encountered in path analyses and simple CFA models, where even well-specified models may exhibit an RMSEA exceeding the standard upper limit of 0.10. Although no values exceeded 0.10 in this work, since they were greater than the optimal level of 0.05, a more cautious approach was needed.

The key findings demonstrated that only the infection safety was marginally significant at the 10% level, while all other variables achieved significance at the 1% level, aligning with the anticipated directionality of theoretical signs. As expected, tourism activities were significantly influenced by shopping ($\mathsf{\lambda }$ = 0.827), accommodation ($\mathsf{\lambda }$ = 0.523), food and beverage ($\mathsf{\lambda }$ = 0.934), and recreation ($\mathsf{\lambda }$ = 0.715) at the 1% significance level. Similarly, carbon emissions were completely explained by electricity use ($\mathsf{\lambda }$ = 0.974) and gas use ($\mathsf{\lambda }$ = 0.823). Furthermore, the analysis revealed that tourism activities contributed to an increase in carbon emissions ($\mathsf{\beta }$ = 0.404, p < 0.01), and the tourism activities are also influenced by higher levels of tourism development ($\mathsf{\gamma }$ = 0.696), stable infection safety ($\mathsf{\gamma }$ = 0.026), and higher income ($\mathsf{\gamma }$ = 0.238). Conversely, poor atmospheric quality was associated with a reduction in tourism activities ($\mathsf{\gamma }$ = −0.040). Carbon emissions were observed to decrease as green space ratios increased ($\mathsf{\gamma }$ = −0.030), in contrast to the increase associated with larger populations ($\mathsf{\gamma }$ = 0.493).

4.4. Testing Measurement Inavariance

To conduct comparisons between groups, it is often required to test whether the measurement variables reflect identical attributes across groups—a process known as measurement invariance testing.

Table 5. Test results of measurement invariance.

	Between Years									Between Yearly High- and Low-Level ESG Groups
	2019 vs. 2020			2020 vs. 2021			2019 vs. 2021			2019			2020			2021
Model	χ2	d.f.		χ2	d.f.		χ2	d.f.		χ2	d.f.		χ2	d.f.		χ2	d.f.
I	2.295	4	n.s.	9.025	4	n.s.	3.277	3	n.s.	3.743	3	n.s.	6.029	3	n.s.	1.845	2	n.s.
II	2.306	10	n.s.	9.418	10	n.s.	3.837	9	n.s.	20.662	9	**	23.513	9	***	20.339	8	***
III	15.839	62	n.s.	64.438	62	n.s.	85.228	61	n.s.	265.250	61	***	256.970	61	***	277.313	60	***
IV	22.920	68	n.s.	69.755	68	n.s.	108.896	67	n.s.	320.248	67	***	280.633	67	***	303.766	66	***
	Between Years for High-level ESG Groups									Between Years for Low-level ESG Groups
	2019 vs. 2020			2020 vs. 2021			2019 vs. 2021			2019 vs. 2020			2020 vs. 2021			2019 vs. 2021
Model	χ2	d.f.		χ2	d.f.		χ2	d.f.		χ2	d.f.		χ2	d.f.		χ2	d.f.
I	1.506	4	n.s.	6.275	4	n.s.	7.591	4	n.s.	2.442	4	n.s.	3.611	4	n.s.	7.070	4	n.s.
II	1.604	10	n.s.	6.755	10	n.s.	8.328	10	n.s.	2.604	10	n.s.	4.304	10	n.s.	8.183	10	n.s.
III	13.966	62	n.s.	44.477	62	n.s.	68.494	62	n.s.	12.248	62	n.s.	46.116	62	n.s.	59.672	62	n.s.
IV	22.638	68	n.s.	47.731	68	n.s.	87.265	68	*	17.024	68	n.s.	49.941	68	n.s.	73.861	68	n.s.

Note (1) I: Measurement weights, II: Measurement intercepts, III: Structural covariances, IV: Measurement residuals; Note (2) *** p < 0.01, ** p < 0.05, * p < 0.10; n.s. = not significant.

summarizes the test results conducted over three years and between two ESG level groups. The tests were based upon confirming metric invariance, which assumes the weak equivalence [120]. As a result of the χ² comparison test with the baseline model, there were mostly no significant differences, indicating that equivalence at the 5% significance level was maintained.

However, complete measurement invariance was not established between the 2019 and 2021 models, nor between the high and low ESG groups. In this case, the following three alternatives are considered viable [121]: First, the removal of non-invariant items from the model can enhance its utility for a specific study; however, this approach becomes impractical in the absence of alternative variables for critical factors. Second, the application of partial factorial invariance permits the factor loadings of identical items to be equal across groups, while allowing for differences in those of non-identical items. Steenkamp and Baumgartner [122] suggest that, due to the challenges of achieving complete equivalence, researchers should aim for, and be confident in, achieving partial measurement equivalence. Third, verified non-invariance can be interpreted as an indication of meaningful differences between groups. This perspective encourages researchers to interpret non-uniformity in responses as reflecting actual differences between groups.

This led to a sequential lift of restrictions on paths with significant differences in factor loadings (for the ESG groups, the constraints were removed for accommodation or recreation, and, for the yearly comparison, the constraints were released for food and beverage and/or gas usage measurement variables), thereby ensuring partial measurement invariance. Encountering non-invariance across groups is not uncommon in research. The literature cites numerous instances where scales do not exhibit comparability across different groups (e.g., [123,124,125]). Robitzsch and Lüdtke [126] have posited that meaningful and valid comparisons can still be conducted even in the absence of strict measurement homogeneity. This indicates that the lack of measurement homogeneity does not inherently prevent meaningful group comparisons and should be evaluated within the specific context and objectives of the research. Given that the primary goal of this study is to explore the structural differences between ESG groups rather than to verify a theory, it was determined that the research could accommodate a degree of partial measurement invariance and accept the premise regarding group differences.

4.5. Estimating MIMIC Models across Years

Table 6. Estimates of MIMIC models across years.

		2019		2020		2021
Var1	Var2	Std. Est.		Std. Est.		Std. Est.
Carbon Emissions (CE)	ELC	0.977	***	0.975	***	0.972	***
	GAS	0.798	***	0.832	***	0.838	***
Tourism Activities (TA)	SHP	0.839	***	0.832	***	0.812	***
	ACM	0.523 ab	***	0.449 b	***	0.591 a	***
	FNB	0.938	***	0.932	***	0.931	***
	RCN	0.808 a	***	0.734 a	***	0.605 b	***
TA	CE	0.417	***	0.392	***	0.397	***
SAF	TA	0.018	n.s.	0.046	*	0.013	n.s.
POL		−0.046	*	−0.043	n.s.	−0.032	n.s.
DEV		0.677	***	0.718	***	0.699	***
INC		0.255	***	0.206	***	0.244	***
INC	CE	0.230	***	0.241	***	0.224	***
INC2		−0.163 a	***	−0.152 ab	***	−0.136 b	***
GRN		−0.026	*	−0.030	**	−0.036	**
POP		0.481	***	0.486	***	0.484	***
AVE (CR, α)	CE	0.796 (0.886, 0.914)		0.821 (0.901, 0.905)		0.825 (0.904, 0.905)
	TA	0.646 (0.875, 0.852)		0.594 (0.847, 0.818)		0.561 (0.831, 0.814)
Model Fit		χ2 = 90.886 (df = 41, p < 0.01), RMR = 0.027, SRMR = 0.029, GFI = 0.949, AGFI = 0.886, CFI = 0.984, RMSEA = 0.070		χ2 = 117.309 (df = 41, p < 0.01), RMR = 0.027, SRMR = 0.028, GFI = 0.935, AGFI = 0.856, CFI = 0.976, RMSEA = 0.086		χ2 = 86.076 (df = 41, p < 0.01), RMR = 0.030, SRMR = 0.029, GFI = 0.950, AGFI = 0.890, CFI = 0.985, RMSEA = 0.066

Note (1) The variable names are as listed in Table 2; Note (2) χ² difference: ^c < ^b < ^a, ^ab belongs to both a and b; Note (3) Each year’s N = 250; Note (4) *** p < 0.01, ** p < 0.05, * p < 0.10; n.s. = not significant.

depicts the results of estimating the MIMIC model for each year. Similar to the pooled model (as referenced in Table 4), the fit for all three yearly models was satisfactory, with no issues detected in the CFA model fit, AVE, CR, Heywood cases, Cronbach’s alpha, and abnormal standardized coefficients. The main findings highlight that the observation variables for carbon emissions and tourism activities, along with most control variables, were effective. However, the significance of infection safety exclusively emerged in 2020, while the relevance of atmospheric pollution was only observed in 2019. In addition, no discrepancies were found with theoretical signs, supporting the EKC hypothesis. Then, slope tests were conducted to compare coefficient differences across the yearly models for each variable. The results revealed year-over-year differences in the coefficients for accommodation (greater in 2021 than in 2020), recreation (greater in 2019 than in 2021), and the square of income (greater in 2019 than in 2021).

4.6. Estimating MIMIC Models across Years and ESG Groups

Table 7. Comparison of MIMIC models across divisions.

		2019 High		2019 Low		2020 High		2020 Low		2021 High		2021 Low
Var1	Var2	Std. Est.		Std. Est.		Std. Est.		Std. Est.		Std. Est.		Std. Est.
CE	ELC	0.985	***	0.966	***	0.978	***	0.970	***	0.974	***	0.969	***
	GAS	0.723	***	0.895	***	0.777 ‡	***	0.892 ‡	***	0.787 ‡	***	0.896 ‡	***
TA	SHP	0.878	***	0.804	***	0.876	***	0.800	***	0.866	***	0.776	***
	ACM	0.632	***	0.406 ab	***	0.516	***	0.371 b	***	0.616	***	0.543 a	***
	FNB	0.945 ‡	***	0.949 ‡	***	0.924 †	***	0.956 †	***	0.904 ‡	***	0.972 ‡	***
	RCN	0.822 a	***	0.787 a	***	0.767 a	***	0.689 a	***	0.620 b	***	0.586 b	***
TA	CE	0.397 †	***	0.464 †	***	0.365	***	0.439	***	0.355 ‡	***	0.531 ‡	***
SAF	TA	0.020	n.s.	0.038	n.s.	0.015	n.s.	0.071	**	−0.031	n.s.	0.064	**
POL		−0.053	n.s.	−0.020	n.s.	−0.059	n.s.	−0.026	n.s.	−0.036	n.s.	−0.018	n.s.
DEV		0.695 †	***	0.610 †	***	0.747	***	0.662	***	0.813 ‡	***	0.531 ‡	***
INC		0.224	***	0.324	***	0.173	*	0.258	***	0.152	*	0.375	***
INC	CE	0.281	***	0.137	n.s.	0.300	***	0.168	n.s.	0.345	***	0.075	n.s.
INC2		−0.187	***	−0.152	***	−0.166	***	−0.149	***	−0.171	***	−0.128	***
GRN		0.004	n.s.	−0.051	**	−0.010	n.s.	−0.048	**	−0.016	n.s.	−0.052	**
POP		0.479	***	0.517	***	0.471	***	0.511	***	0.430	***	0.511	***
AVE (CR, α)	CE	0.713 (0.829, 0.912)		0.893 (0.944, 0.914)		0.773 (0.870, 0.895)		0.879 (0.935, 0.912)		0.779 (0.874, 0.894)		0.881 (0.936, 0.914)
	TA	0.690 (0.897, 0.888)		0.617 (0.857, 0.808)		0.624 (0.865, 0.847)		0.579 (0.834, 0.783)		0.568 (0.836, 0.825)		0.572 (0.835, 0.802)
Model Fit		χ2 = 84.793 (df = 41, p < 0.01), RMR = 0.039, SRMR = 0.041, GFI = 0.912, AGFI = 0.806, CFI = 0.974, RMSEA = 0.093		χ2 = 82.745 (df = 41, p < 0.01), RMR = 0.032, SRMR = 0.041, GFI = 0.913, AGFI = 0.807, CFI = 0.973, RMSEA = 0.091		χ2 = 83.255 (df = 41, p < 0.01), RMR = 0.034, SRMR = 0.036, GFI = 0.913, AGFI = 0.806, CFI = 0.975, RMSEA = 0.091		χ2 = 83.590 (df = 41, p < 0.01), RMR = 0.032, SRMR = 0.036, GFI = 0.912, AGFI = 0.804, CFI = 0.972, RMSEA = 0.092		χ2 = 73.909 (df = 41, p < 0.01), RMR = 0.034, SRMR = 0.035, GFI = 0.922, AGFI = 0.828, CFI = 0.980, RMSEA = 0.080		χ2 = 75.185 (df = 41, p < 0.01), RMR = 0.041, SRMR = 0.035, GFI = 0.917, AGFI = 0.816, CFI = 0.978, RMSEA = 0.082

Note (1) The variable names are as listed in Table 2; Note (2) χ² difference: ^c < ^b < ^a, ^ab belongs to both a and b; Note (3) ^‡ and ^† indicate 5% and 10% χ² difference, respectively; Note (4) Each division’s N = 125; Note (5) *** p < 0.01, ** p < 0.05, * p < 0.10; n.s. = not significant.

shows the analysis of variations in the MIMIC model across years and between high and low ESG groups. In alignment with prior results, all six sub-models exhibited satisfactory model diagnostic figures. The primary model comparisons revealed the following: In 2019—prior to the COVID-19 outbreak—a marginal difference was observed in the impact of tourism activities on carbon emissions between low ESG groups ($\mathsf{\beta }$ = 0.464) and high ESG groups ($\mathsf{\beta }$ = 0.397) at the 10% significance level (p < 0.10). This distinction became temporarily non-significant in 2020, the year COVID-19 emerged, but regained significance in 2021, in the post-COVID-19 period, at the 5% significance level, with high ESG groups showing a $\mathsf{\beta }$ coefficient of 0.355 compared to low ESG groups at 0.531.

Significant distinctions between high and low ESG groups were primarily attributed to the following drivers: Particularly, the differences pertained to the gas usage explaining carbon emissions, which displayed variations between the groups in 2020 ($\mathsf{\lambda }$ = 0.777 for the high group < $\mathsf{\lambda }$ = 0.892 for the low group) and 2021 ($\mathsf{\lambda }$ = 0.787 for the high group < $\mathsf{\lambda }$ = 0.896 for the low group). Additionally, food and beverage consumption was consistently observed to be higher in the low ESG group across all years—i.e., the coefficients of the low ESG group (λ = 0.949; 0.956; 0.972) are larger compared to their counterparts in the high group (λ = 0.945; 0.924; 0.904), respectively. While control variables largely did not show differences between the groups, tourism development emerged as a factor that partially moderates these relationships—with high ESG groups demonstrating superior levels in comparison to their low ESG counterparts.

Importantly, the results indicate that regions with higher ESG levels, despite having more active tourism development, emit less carbon than those with lower ESG levels. Furthermore, within low ESG groups, the significance of infection safety for tourism activities and the impact of green space on carbon emissions were notable. This implies that infection safety positively affects tourism activities, and an increase in green space reduces carbon emissions in low ESG areas. Conversely, the influence of atmospheric pollution was neutralized across the ESG groups, highlighting the role of suppressor variables, which are factors that can make the actual relationship between the two variables appear non-existent. Additionally, variations in recreation across years (2019 = 2020 > 2021) were observed in both high and low groups, with distinct differences in accommodation noted in low groups (2019 = 2020 > 2021), emphasizing the necessity of identifying moderating effects.

5. Conclusions

5.1. Summary and Discussions

The aim of this study was to examine the impact of tourism activities on carbon emissions within the context of South Korea and to investigate how these causal relationships are moderated by the ESG governance framework. To thoroughly explore the four research questions arising from theoretical considerations—namely, the relationship between tourism activities and carbon emissions; variations across three distinct years; the moderating effect of ESG policies; and the influences of control variables—the current study developed MIMIC models. These models incorporated key variables, including carbon emissions attributed to electricity and gas usage, tourism activities (such as shopping, accommodation, food and beverage, and recreation), and ESG administrative orientation, alongside control variables (including infection safety, atmospheric pollution, tourism development, income level, green space, and local population). In this study, secondary data from 2019 to 2021 were rigorously analyzed, with sources compiled from reputable organizations—such as the NSO, KTO, KEC, NTS, LX corporation, and the ESG Economic Research Institute. Before proceeding with the main analysis, a series of preliminary assessments was undertaken to evaluate the descriptive statistics, correlations, construct validity, reliability, model fit, measurement invariance, and multivariate normality. No crucial issues were encountered, paving the way for the subsequent estimation of yearly and high versus low ESG level MIMIC models.

The main findings are as follows: First, the investigation revealed that tourism activities significantly impact carbon emissions across four sub-sectors: shopping, accommodation, food and beverage, and recreation. This significant relationship was consistently observed in 2019 (pre-COVID-19), 2020 (the outbreak of COVID-19), and 2021 (post-COVID-19). However, the extent of these impacts varied across the years, with a notable increase in the accommodation sector in 2021 and a decrease in recreation by the same year. Second, despite these variations, the influence of tourism activities on carbon emissions remained significant throughout the three years examined, without any statistical variance in the effect size between the years. While some previous studies have presented contradictory findings (including no significant impact) [19,20,21,22,23,24,26,27], the adverse effect of tourism consumption has been reaffirmed within the South Korean context. Third, the findings further elucidate the role of ESG administrative orientation in alleviating the carbon-emitting effects of tourism activities. According to the results, significant moderation was observed in both 2019 and 2021, but not in 2020. Specifically, the causality magnitude between tourism activities and carbon emissions was found to be greater in low ESG groups compared to high ESG groups for both 2019 and 2021. The distinction between the two ESG governance groups was primarily attributed to the levels of carbon emissions resulting from tourism consumption within the food and beverage sector.

Lastly, the analysis yielded significant insights regarding control variables. Initially, infection safety was identified as a key motivator for tourism activities, notably in regions with lower levels of ESG governance. In addition, the impact of atmospheric pollution on tourism activities, with the intervention of ESG governance, revealed complex dynamics. In particular, ESG administration exerted a suppressor effect, significantly neutralizing the influence of air quality on tourism activities in both high and low ESG groups. Furthermore, the degree of tourism development was highlighted as an influential factor in enhancing tourism activities across all groups, underscoring its importance in both high- and low-level ESG contexts. The contribution of tourism development to stimulating tourism activities consistently showed a stronger presence in the high ESG group compared to the low ESG group, with the difference reaching its peak in 2021. Additionally, the income effects on tourism activities and carbon emissions were significant across all models, except for a few models for the low ESG group. Given that carbon emissions are considered a form of environmental pollution, applying the Kuznets hypothesis was found to be suitable, thus reinforcing the model’s validity. Particularly in the low ESG group, while the income effect on tourism activities was significant, it was only the square of income that showed a significant impact on carbon emissions. The significant roles of green space in mitigating carbon emissions were also confirmed across all models, except within the contexts of high-level ESG governance. Moreover, the influence of population growth on carbon emissions was consistently significant. In particular, this impact was more marked within the low-level ESG governance group, establishing a clear association between carbon emissions, population size, and the level of ESG governance.

To conclude, the impact of tourism activities on carbon emissions showed results consistent with numerous studies. However, the differences lay in the contributions of the sub-sectors. Compared to previous studies, such as those of Xiong et al. [57], Sun [58], and Lenzen et al. [59], the effects of the food and beverage and shopping sectors were significant, showing high similarity in context. On the other hand, there is some divergence from studies by Chen et al. [56] and Tang et al. [61], which identified accommodation and recreation as the most significant contributing factors. The results from different studies can vary based on the tourism destinations, input variables, estimation models, points of analysis, and data used. For instance, an Indian study [25,26,27] demonstrated conflicting results for the same tourism destination, revealing that the inclusion of more historical data led to a more negligible perceived impact of tourism on carbon emissions. Therefore, it is necessary to continue supplementing comparative research by applying more standardized and sophisticated analysis methods.

On the other hand, inconsistent results highlight the need for further investigation. These differences may stem from various factors, such as research methodology, sample characteristics, spatial and temporal attributes, and other unique interactive conditions (e.g., COVID-19). Therefore, it is essential to explore these discrepancies from multiple perspectives. By identifying new variables, establishing improved methodologies, and pinpointing the effects of specific conditions, future efforts can contribute to solidifying the theoretical framework. In conclusion, the study’s key points are as follows: Carbon emissions increase as tourism demand rises, which may become more pronounced as tourism activities intensify post-COVID-19. Tourism development evidently serves as an incentive to stimulate tourism demand. However, if well-established ESG administrative policies are implemented, the carbon emissions induced by tourism can be effectively controlled. Additionally, efforts are required to identify the exogenous factors associated with these ESG administrative policies.

5.2. Implications

The academic contributions of this study are highlighted as follows: First, this study examined the impact of tourism activities within four major sub-industries on carbon emissions by electricity and gas consumption and further assessed the moderating role of ESG administrative power in this relationship. The development of a new tourism–environment model proved the model’s validity, thereby providing theoretical directions and foundations for future research. The findings emphasize the importance of evaluating the relationship between tourism and the environment based on a multi-dimensional aspect rather than a uni-dimensional one. Moreover, verification of ESG administrative power indicates that the ESG notion should be preferentially considered in future tourism–environment research. Second, the tourism impact on carbon emissions still remains inconsistent globally, and such trends are also prevalent in the South Korean case. This study, through an examination of the South Korean context, made a contribution by providing additional empirical evidence on the association between tourism activities and carbon emissions. Furthermore, existing studies present divergent conclusions on which specific industry sectors—whether shopping, lodging, food and beverage, or recreation—contribute most significantly to carbon emissions. Yet, in this paper, the food and beverage and shopping sectors have been identified as having the greatest effect on carbon emissions.

Third, particularly, in research designs with a time-series structure, accident variables, such as COVID-19, must be taken into account. Regarding this issue, the current study substantiated that variations in influence among variables exist through annual model comparisons. Consequently, it implies that the effect of COVID-19 intervention must be rigorously controlled as an exogenous variable in studies utilizing 2020 data. Fourth, this study stands out by focusing on the ESG administrative framework at the level of basic governmental units within South Korea. Existing relevant data predominantly has been focused on evaluation metrics reflecting corporate ESG performance. Recently, however, the concept of ‘ESG normalization’ in daily life has emerged as a focal social issue. There is also a growing preference for governments to adopt these practices in public administration and strive to achieve related goals. This study has, therefore, played a part in shaping future research directions by addressing a relevant and timely research issue.

The practical implications of this study are as follows: First, the analysis revealed that the impact of individual tourism activities on carbon emissions can be ranked as follows: food and beverage > shopping > recreation > accommodation, with no significant variations observed across different years or ESG groups. To achieve environmental objectives, it is essential to employ a cost-effective strategy that allocates resources to carbon reduction initiatives. The accurate identification of each sector’s impact on carbon emissions could contribute to efficient resource allocation when establishing tourism policies. Second, the analysis demonstrated that the impact of tourism activities on carbon emissions was more pronounced in groups with lower ESG levels, whereas the influence of tourism development on tourism activities was stronger in groups with higher ESG levels. These findings underscore the necessity of aligning tourism marketing strategies with ESG levels. All decision-making processes incur trade-offs; however, the current study challenges such a notion, suggesting that, with the proper implementation of ESG-related policies, increased tourism activities can successfully coexist with effective carbon emission control.

Third, the present research derived several implications from various control variables. In groups with low ESG levels, the management of green space and infection safety as strategies to reduce carbon emissions was noted. Conversely, in high-growth ESG administrative regions, effective environmental improvement was challenging to achieve through these measures alone. Additionally, it has been verified that the effect of air pollution on tourism activities is eroded by ESG administrative power. In addition, this study confirmed the anti-environmental tendency of the EKC hypothesis to be rejected in areas with low ESG levels. Furthermore, carbon emissions relative to population were notably higher in areas with lower-level ESG administration. To conclude, it highlights the critical need to develop a comprehensive environmental improvement strategy that incorporates energy, transportation, and industrial policies. Lastly, the analytical findings underscore significant annual variations in the model’s outcomes. This illustrates the diversity in the impacts on carbon emissions, the contributions of sub-sectors linked to tourism activities, the level of ESG administrative intervention, and the effects of various control variables. Consequently, this paper recommends a shift in tourism policy formulation towards strategies focused on short-term goals rather than maintaining a long-term perspective. The current analysis suggests that there is a pressing need for a timely and responsive tourism strategy that adapts to rapidly evolving environmental conditions.

5.3. Limitations

This study has the following limitations: Initially, in this research, domestic tourism and international tourism were not distinguished. Additionally, considering that trends may differ significantly from the end of the COVID-19 pandemic, it is necessary to conduct a comparative analysis with the period of 2022–2023, when tourism activities began to recover rapidly. However, this study faced severe limitations due to the reliance on secondary public data. The National Statistics Portal generally delays data updates. Hence, as of the second quarter of 2024, the energy consumption, regional safety index, income level, and green space rate are only available up to 2022, with some updates published in March 2024. Furthermore, another limitation was that local government data separating domestic and foreign nationals were not provided. Thus, a future analysis comparing the effects before and after the COVID-19 pandemic (e.g., 2019–2021 vs. 2022–2023) is anticipated when data for 2023 become available. Here, it is vital to complement the quantitative analysis with qualitative insights from industry stakeholders, which will help us to understand the impact of COVID-19 on tourism activities and ESG policies from a broader perspective.

Additionally, due to data availability constraints, the selection was limited to four industries. This emphasizes the need for incorporating exogenous factors not currently considered in the model into future research. Consequently, it is envisioned that a more robust analytical framework can be developed by exploring heterogeneous public data sources and employing integration technologies. In addition, the nature of the collected data imposed constraints that necessitated treating ESG administrative power as a fixed characteristic. The ideal analytical framework would utilize ESG evaluation data tailored at every point in time; however, it is acknowledged that achieving this condition is realistically challenging. Therefore, future research might propose methods for securing proxy indicators for ESG administration or survey design for developing direct data.

Moreover—as previously discussed—examining and proving the measurement invariance of a measurement tool is regarded as one of the crucial procedures. There were varied opinions pertaining to the necessity for meeting these standards, the conditions under which they can be alleviated, and the unnecessariness of such conditions. This study posits that meeting these criteria may hold greater significance in terms of academic contribution. Consequently, it is deemed meaningful for subsequent research to further strengthen the measurement invariance across groups and over time. Furthermore, the analytical framework employed in this study is characterized by a cross-sectional dataset spanning three years. Inevitably, there should be limitations in accommodating all variables inherent within a single year. A more refined analysis would result from disaggregating annual data into monthly increments. Accordingly, it is anticipated that future research will introduce a detailed research design capable of overcoming these data constraints.