Evaluating Public Service Experience in Ecological Institutions: A Structural Pathway Approach to Visitor Satisfaction at the Sejong National Arboretum

Service quality and expectation disconfirmation

Service quality refers to the customer’s overall perception of a service and is considered as a key factor in evaluating public service performance (Parasuraman et al., 1988; Zeithaml et al., 1996). In the public sector, where intangibility is particularly emphasized, citizens’ cognitive evaluations serve as the primary criterion for determining satisfaction, rather than physical performance. The concept of service quality originated from the disconfirmation paradigm, which was initially developed through research on product quality and customer satisfaction (Oliver, 1980). Grönroos (1982, 1984) proposed a theoretical framework suggesting that service quality is shaped by the comparison between expected and perceived service. He distinguished between “technical quality” (what is delivered) and “functional quality” (how it is delivered), emphasizing that the latter is more closely associated with customers’ subjective evaluations of their service experience.

This theoretical foundation was later systematized into the SERVQUAL model by Parasuraman et al. (1985, 1988), which identified five key dimensions for assessing customer perceptions of service quality: reliability, responsiveness, assurance, empathy, and tangibles. Since its development, the model has been widely applied across various service industries as a determinant of subsequent customer behaviors, including satisfaction, repeat usage, and recommendation (Cronin and Taylor, 1992; Brady and Cronin, 2001). In particular, Rust and Oliver (1994) proposed a three-dimensional framework that added the service environment to the existing dimensions of technical and functional quality, highlighting the complex cognitive components that shape the overall service experience.

Service quality is not merely a momentary evaluation; it plays a crucial role in fostering customer satisfaction and serves as the foundation for subsequent behaviors such as repeat usage/repurchase, positive word-of-mouth (WOM), and policy adoption. In the context of public services, users often place greater importance on publicness and reliability than on efficiency alone. Accordingly, “expectation disconfirmation”—a cognitive evaluation based on the gap between pre-use expectations and perceived service performance—emerges as a key mediating factor (Oliver and Bearden, 1983). Notably, research has shown that customers with stronger confidence in their expectations are more susceptible to the effects of negative disconfirmation (i.e., when performance falls short of expectations; Yi and Bak, 2020). Therefore, there is a growing recognition of the need to move beyond simple performance measurement and instead interpret customer satisfaction in terms of the formation and fulfillment of expectations as perceived by the customer.

Public customer satisfaction index (PCSI)

Today, the evaluation of public institution services is evolving beyond a focus on operational efficiency or satisfaction with civil complaint handling, toward a more comprehensive approach that encompasses the public’s perceived experience with policy implementation processes and their outcomes. In particular, the multi-layered nature of public services underscores the need to understand the process of customer satisfaction formation as a distinct structural framework from that of the private sector. In response to this need, the Ministry of Economy and Finance (2015) developed the Public Customer Satisfaction Index 2.0 (PCSI 2.0), a measurement framework tailored specifically for public institutions. The PCSI 2.0 quantifies the entire public service process through five core dimensions: service quality, social responsibility, expectation-disconfirmation (degree of expectation congruence), customer satisfaction, and perceived performance. This model was designed to enable a precise assessment of public institutions’ service performance from the customer’s perspective.

PCSI 2.0 seeks to move beyond simple evaluations at the service encounter level by structuring the cognitive and emotional judgment processes involved in public service experiences—from the formation of customer expectations to the evaluation of experiences and final perceptions. Whereas the existing CSI and PCSI 1.0 primarily focused on measuring static satisfaction levels based on service quality, PCSI 2.0 incorporates a more complex framework that accounts for the contextual nature of public service delivery and the potential for realizing public value. Customer satisfaction and service performance serve as outcome variables that holistically capture the entire evaluative process. Going beyond mere customer satisfaction, these variables help explain the diffusion pathways of policy effects—such as policy acceptance, institutional trust, and the formation of public beliefs. In conclusion, PCSI 2.0 is an integrated evaluation framework for analyzing the causal structure of public services—from expected service levels and perceived quality to value recognition and acceptance of policy outcomes. It offers a valuable foundation not only for building practical feedback systems, but also for advancing theoretical understanding of public services and improving policy performance management.

Research objectives and hypotheses

This study aimed to empirically analyze the causal pathways shaping customer satisfaction through the experience of using public institutions, utilizing the PCSI 2.0 model. Specifically, we focused on the structural relationship between service quality and perceptions of social responsibility, both of which influence customer expectation-disconfirmation and satisfaction. Accordingly, we proposed a set of research hypotheses and developed the corresponding conceptual model (see Fig. 1).

Fig. 1

Research model.

In particular, based on the expectation-disconfirmation theory (Oliver, 1980), which has been widely discussed in the literature, we conceptually positioned expectation-disconfirmation as a key mediating variable affecting customer satisfaction. Furthermore, rather than positing a single a priori hypothesis regarding whether expectation-disconfirmation serves as a full or partial mediator, we sought to empirically identify the structural form of the mediating effect by statistically testing the significance of both direct and indirect effects through structural equation modeling (SEM). This approach represents an analytical strategy grounded in prior theoretical research and empirically uncovered the actual functioning of the mediating pathway.

Site description

The Sejong National Arboretum is the first urban national arboretum in South Korea. Situated within Sejong Central Park, it encompasses an area of 65 hectares (649,997 m²) and was inaugurated in October 2020. The arboretum functions as a sustainable green infrastructure within the urban environment, with primary objectives including the systematic conservation of both domestic and international plant species, ecological education, research, and the promotion of a garden culture grounded in citizen participation.

Key facilities consist of a four-season greenhouse, a traditional Korean garden, and various themed gardens. Beyond exhibiting seasonal vegetation and landscape displays, the arboretum provides experiential learning opportunities through plant education programs, ecological interpretation, and community engagement initiatives. These multifaceted functions establish the arboretum as a public ecological and cultural space that integrates environmental education with cultural enrichment, rather than serving solely as a passive viewing site.

From a policy perspective, the Sejong National Arboretum functions as key infrastructure of both balanced national development and sustainable urban creation. Sejong City, conceived as a multifunctional administrative capital, has emphasized ecological values and the preservation of green spaces as core principles since its inception. Within this context, the arboretum has been positioned as key infrastructure for conserving urban ecosystems and addressing climate change.

Its operations are managed by the Korea Arboreta and Garden Institute (KOAGI), under the jurisdiction of the Korea Forest Service, which employs a public institution management model that prioritizes the public good over profitability. Consequently, the arboretum attracts a diverse range of visitors, including the general public, educators, researchers, students, and ecological civic organizations. These visitor demographics provide a valuable foundation for empirically analyzing visitor satisfaction and assessing the effectiveness of related policies based on user purposes and expectations.

Survey design

This study conducted an online survey over a period of approximately five months, from June 1 to October 31, 2024, to empirically analyze visitor satisfaction with public services at the Sejong National Arboretum. The target population consisted of general visitors to the arboretum. To ensure accessibility and ease of participation, the survey was administered in a non-face-to-face format using QR codes. Informational banners were strategically placed at key points along the exhibition route to encourage voluntary participation. Duplicate responses based on IP addresses and insincere or incomplete responses were excluded, resulting in a valid sample of 1,060 cases.

The survey items were developed based on the PCSI 2.0 model proposed by the Ministry of Economy and Finance (MOEF, formerly the Ministry of Strategy and Finance) in 2015. Items were formulated according to the operational definitions of key concepts, and responses were measured using an 11-point Likert scale (0 = Strongly Disagree, 10 = Strongly Agree). The measurement items were categorized into four variables, as summarized in Table 1.

Table 1

Survey items for visitor satisfaction at the National Sejong Arboretum

Service quality is a concept perceived through the gap between customer expectations and the actual services received. It plays an even more crucial role in the context of public services, which are inherently intangible and consumed as they are delivered (Yoo et al., 2019). In this study, service quality is defined as “the extent to which an institution’s customer satisfaction management activities influence overall satisfaction.” It was measured through multi-dimensional items, including satisfaction with exhibition content, clarity of information delivery, comfort of the physical environment, and the convenience of institutional processes.

Social responsibility can be defined as “an institution’s responsibility for the impact of its decision-making and activities on society and the environment through transparent and ethical behavior” (Kim and Lee, 2012). It is considered a key criterion for evaluating the publicness and sustainability of public institutions. This construct was used as an indicator to measure the extent to which the Sejong National Arboretum fulfills its role in delivering public benefits and consists of items such as user interest in the arboretum, raising awareness of biological resources and the perceived public benefits.

Expectation-disconfirmation refers to the cognitive process of evaluating the gap between pre-service expectations and the actual experience, playing a crucial role in shaping customer satisfaction (Oliver, 1997). This variable was conceptualized as “the degree to which the actual experience met prior expectations.” It was designed to encompass both instances where expectations were met and where they were surpassed. Specifically, it includes three items: satisfaction relative to expectations, satisfaction relative to an ideal standard, and satisfaction relative to services offered by other organizations, thereby allowing for a multifaceted measurement of cognitive discrepancies.

Satisfaction is a psychological evaluation resulting from the interplay between customers’ emotional responses and their prior expectations during service consumption. It is closely linked to customer loyalty and subsequent behavioral intentions. The measurement focused on the overall satisfaction experienced by visitors to the Sejong National Arboretum and comprised three items: overall satisfaction, willingness to revisit, and willingness to recommend. This construct not only captures simple emotional responses but also encompasses the potential to induce behavioral intentions, grounded in the rationale behind the independent satisfaction factors proposed by Geva and Goldman (1991).

Analytical strategy

This study employed SEM to examine the structural relationships among factors influencing public service satisfaction among visitors to the Sejong National Arboretum. Data were analyzed using IBM SPSS Statistics 29.0 and Amos 29.0, following a systematic procedure.

First, frequency and descriptive statistical analyses were performed to identify respondents’ demographic characteristics—such as gender, age, and place of residence—and to ensure the representativeness of the sample. Reliability analysis was conducted using Cronbach’s α coefficient to assess the internal consistency of each variable.

Next, confirmatory factor analysis (CFA) was carried out to evaluate the measurement validity of the research model. CFA is a statistical technique used to test whether the theoretically established structure between latent and observed variables are supported by the actual data. In this study, model fit was assessed using a range of goodness-of-fit indices, including the chi-square (χ²) statistic, comparative fit index (CFI), Tucker-Lewis index (TLI), root mean square error of approximation (RMSEA), and standardized root mean square residual (SRMR). In general, a model is considered to have an acceptable fit when the CFI and TLI values are .90 or higher, the RMSEA is below .08, and the SRMR is below .10 (Hu and Bentler, 1999). Based on the results of the CFA, this study comprehensively examined both convergent and discriminant validity. Convergent validity was evaluated using three criteria: standardized factor loadings of .50 or higher, average variance extracted (AVE) of .50 or higher, and construct reliability (CR) of .70 or higher (Bagozzi and Yi, 1988). Discriminant validity was assessed using two approaches. First, following the criteria proposed by Fornell and Larcker (1981), discriminant validity was deemed to be established if the square root of the AVE for each latent variable exceeded the correlation coefficients between that latent variable and all other variables (AVE > ϕ²). Second, if the 95% confidence interval of the correlation coefficient between any two variables did not include 1 (i.e., ϕ ± 2 × S.E.), the two constructs were considered to be statistically distinct. Through this multifaceted assessment, the measurement model in this study demonstrated sufficient discriminant validity among the theoretical constructs.

Finally, the causal relationships among latent variables were comprehensively evaluated through SEM. This study formulated the hypothesis that service quality and social responsibility affect customer satisfaction via the mediating role of expectation-disconfirmation, and subsequently examined the significance of individual paths and the underlying causal structure. The direct, indirect, and total effects among the variables were estimated, and the statistical significance of the mediating effects was specifically tested using 5,000 bootstrap iterations. A mediating effect was considered statistically significant if the corresponding confidence interval did not include zero (Preacher et al., 2007). These analytical procedures were undertaken to comprehensively assess the measurement validity and causal explanatory power of the proposed model, thereby providing both theoretical and practical implications.

Descriptive statistics and correlation analysis

In terms of demographic characteristics, a majority of respondents were female, with individuals in their 30s and 40s comprising the largest age groups. Survey participants were evenly distributed across Sejong, Daejeon, the Chungcheong region, and the Seoul metropolitan area in terms of their place of residence, thereby ensuring regional representativeness (see Table 2).

Table 2

Demographic characteristics

Descriptive statistical analysis of the key variables (see Table 3) indicated that all measurement items had mean scores above 8.5 and standard deviations below 2.0, suggesting a generally high level of satisfaction and consistency in responses. To assess the normality of the variables, skewness and kurtosis were examined. All values satisfied the criteria proposed by Kline (2023)—absolute skewness less than 3 and absolute kurtosis less than 10—supporting the assumption of normality. Pearson correlation analysis (see Table 4) revealed that the correlation between expectation-disconfirmation and satisfaction was the strongest, with a coefficient of r = .860 (p < .01). Significant positive correlations were also found between satisfaction and social responsibility (r = .738), as well as service quality (r = .713).

Table 3

Descriptive statistics

Table 4

Correlation analysis

The reliability of each construct was confirmed through Cronbach’s α coefficients, all of which exceeded the threshold of .70, indicating acceptable internal consistency. Specifically, product quality (α = .810), delivery quality (α = .955), environmental quality (α = .796), social responsibility (α = .866), expectation-disconfirmation (α = .917), and satisfaction (α = .911) all demonstrated high reliability.

Measurement model evaluation

A confirmatory factor analysis (CFA) was conducted to assess the validity and reliability of the measurement model. The analysis was performed within the framework of a structural equation model, evaluating whether each measurement item adequately represented its corresponding latent variable.

Initial CFA results revealed that certain items exhibited factor loadings below the acceptable threshold or lacked conceptual alignment, leading to a revision of the model. In particular, among the subdimensions of service quality, the three items related to “customer interaction” (SQ4–SQ6) demonstrated a distinct conceptual consistency from the other items and were therefore restructured into a separate subdimension labeled “Employee Service” (see Fig. 2). This revision was grounded in the theoretical and empirical validity that employee service can function as an independent factor determining user satisfaction. Specifically, previous research has consistently highlighted employee behavior as a critical factor in shaping customer perceptions of service (Henning-Thurau, 2004; Wall and Berry, 2007; Baker et al., 2013). This importance is especially pronounced in the service industry, where the nature of employee-customer interactions plays a pivotal role. (Chien-Wen et al., 2013). These prior findings thus offer a strong conceptual and empirical rationale for reorganizing the “customer interaction-related” items into a separate construct (Alhelalat et al., 2017).

Fig. 2

Confirmatory factor analysis (CFA) of the revised measurement model.

The revised measurement model demonstrated excellent overall fit, as evidenced by the following goodness-of-fit indices: χ² = 363.146, df = 48, CMIN/df = 7.566, TLI = .966, CFI = .975, RMSEA = .079, and SRMR = .028. Notably, the TLI and CFI both surpassed the .95 threshold, and the RMSEA and SRMR fell below the recommended cutoffs of .08 and .03, respectively, providing further evidence of a strong model fit (Hu and Bentler, 1999).

Convergent validity of the measurement items was assessed through standardized factor loadings, average variance extracted (AVE), and construct reliability (CR). All factor loadings were .70 or higher, AVE values exceeded .50, and CR values were above .70—meeting the criteria recommended by Bagozzi and Yi (1988), Fornell and Larcker (1981), and Woo (2022). All items demonstrated standardized factor loadings of .70 or higher, with AVE values exceeding .50 and CR values above .70, thus satisfying the criteria recommended by Bagozzi and Yi (1988), Fornell and Larcker (1981), and Woo (2022). These findings provide evidence of consistent item responsiveness and strong measurement reliability within each construct.

Discriminant validity was assessed based on two criteria proposed by Fornell and Larcker (1981). First, discriminant validity is considered to be established when the square root of the AVE for each latent variable exceeds the correlation coefficient between that variable and all other variables (AVE > ϕ²). As shown in Table 5, this criterion was satisfied for most variable pairs. However, for the pair of expectation-disconfirmation and satisfaction, the square root of the AVE was smaller than their correlation co-efficient, indicating a failure to meet this criterion. Second, discriminant validity is also supported when the 95% confidence interval of the correlation coefficient between two latent variables does not include 1 (ϕ ± 2 × S.E.) (see Table 6). In this case, the confidence interval for the correlation coefficient between expectation-disconfirmation and satisfaction (r = .944) was [.780, 1.108], which includes 1, thus failing to meet this second criterion as well.

Table 5

Discriminant validity results based on Fornell and Larcker (1981) criterion (AVE > ϕ²)

Table 6

Discriminant validity results based on correlation confidence intervals (ϕ ± 2 × S.E.)

These results appears to reflect the strong theoretical relevance between the two constructs. Previous studies (Oliver, 1980; Bearden and Jesse, 1983) have also emphasized that discrepancies between expectations and actual experiences are key pathways to satisfaction and that expectation-disconfirmation and satisfaction are inherently related concepts with closely aligned cognitive structures. Accordingly, the high correlation and structural similarity observed between these two variables are theoretically justifiable and do not fundamentally undermine the overall validity of the measurement model.

In conclusion, while the measurement model did not fully satisfy the discriminant validity criteria for certain variable pairs—particularly between expectation-disconfirmation and satisfaction—this is considered acceptable in light of the theoretical consistency with existing theoretical backgrounds. Overall, considering the goodness-of-fit indices, convergent validity, reliability, and other indicators, the measurement model in this study is deemed sufficiently reliable and valid for use in structural model analysis.

Structural model and hypothesis testing

An evaluation of the theoretical model established through SEM (see Fig. 3) indicated that the model demonstrated a statistically acceptable fit across the majority of fit indices: χ² = 363.146, df = 48, CMIN/df = 7.566, TLI = .966, CFI = .975, RMSEA = .079, and SRMR = .028. This suggests that the model is consistent with the collected data, thereby providing a solid analytical foundation for hypothesis testing (see Table 7).

Fig. 3

Structural model path diagram.

Table 7

Structural model fit results

To assess the significance of each path, a bootstrapping procedure with 5,000 iterations was conducted (Table 8). The analysis revealed that employee service had no significant effect on either expectation-disconfirmation (B = .010, p > .05) or satisfaction (B = −.004, p > .05). Accordingly, Hypotheses H1 and H2 were not supported. In contrast, social responsibility exhibited a significant positive (+) effect on both expectation-disconfirmation (B = .920, p < .001) and satisfaction (B = .245, p < .001), supporting Hypotheses H3 and H4. Additionally, Hypothesis H5 was supported, indicating that expectation-disconfirmation has a significant effect on customer satisfaction (B = .700, p < .001).

Table 8

Results of structural equation modeling: Path coefficients and effect decomposition

An analysis of the indirect effects revealed that social responsibility significantly influenced satisfaction through the mediation of expectation-disconfirmation (B = .644, 95% CI [.495, .793]), confirming that expectation-disconfirmation plays a key mediating role in the relationship between these variables. In contrast, the path from employee service to satisfaction via expectation-disconfirmation was not statistically significant (B = .007, 95% CI [−.066, .080]), indicating that the indirect effect through expectation-disconfirmation was not supported.

Further analysis of the total effects showed that social responsibility exerted both a significant direct effect (B = .245) and an indirect effect (B = .644) on satisfaction, resulting in a total effect of B = .890 (p < .001). The indirect effect comprised approximately 72.4% of the total, indicating that the formation and fulfillment of expectations constitute a key pathway in the development of satisfaction. Moreover, a closer examination of the items comprising each latent variable revealed that, within the social responsibility construct, “the institution’s provision of public benefits” emerged as the most influential item in determining customer satisfaction. Within the expectation-disconfirmation construct, the item “satisfaction relative to services provided by other institutions” emerged as a key determinant in the formation of expectations. This empirically indicates that visitors assess their experiences not based on the absolute level of service, but rather on the relative performance of public institutions in serving the public good.

These findings can be interpreted as empirical support for the validity of Expectation-Disconfirmation Theory (Oliver, 1980) within the context of public service. Specifically, the path appears to effectively function as follows: when customers recognize the social responsibility initiatives of public institutions, their expectations are adjusted accordingly. If subsequent service experiences align with these expectations, they result in heightened satisfaction. This underscores that the formation and confirmation of expectations serve as fundamental evaluative criteria in public service experiences. In contrast, employee service, as an independent variable, did not have a significant direct or indirect effect on satisfaction. This suggests that customer-facing services may be more strongly influenced by emotional responses or situational conditions than by rational evaluation. Accordingly, employee service in public institutions is more likely to exert its effect through interaction with other factors rather than as an independent factor.

Taken together, the SEM conducted in this study empirically demonstrates that the social responsibility of public institutions significantly influences the formation of public satisfaction, mediated by the processes of expectation formation and fulfillment. This finding suggests that delivering strategic value that reflects public values—beyond merely improving service quality—is a crucial mechanism for enhancing public satisfaction.