What drives consumer returning behavior for resource recovery? Evidence from South Korea’s circular scheme

The transition to a circular economy requires not only waste reduction (reduce) by producers but also participation in reuse and recycling by various stakeholders. The international community is striving to establish systems that encourage recycling participation from various stakeholders, including consumers, for the purpose of resource circulation. The European Union (EU) enacted the Packaging and Packaging Waste Regulation 2025/40 in February 2025, strengthening recycling obligations for packaging materials and regulations on the use of recycled content. Accordingly, all EU member states must establish a Deposit Return System (DRS) by January 1, 2029, to ensure the recovery of disposable plastic beverage bottles and metal beverage containers (EU, 2024). This aims to secure high-quality resources in large quantities by inducing consumer return behavior through the economic incentive of a deposit, ultimately supporting the increase in producers’ recycling rates.

According to the Regional Plastics Outlook for Southeast and East Asia (OECD, 2025a), Korea was evaluated as having a very well-established Extended Producer Responsibility (EPR) system, which imposes a certain amount of recycling obligation on producers for product and packaging waste. Korea’s EPR has the strength of clearly defining the roles of producers, consumers, local governments and the central government in the entire process from collection to recycling. Korea’s EPR is recognized for achieving a 95% collection rate for municipal solid waste, supported by a robust monitoring system and the active participation of consumers in meticulous source separation (OECD, 2025b). Despite such quantitative achievements, Korea’s incineration rate remains among the highest in the Asia-Pacific region, with the incineration and landfill rate for waste synthetic resins exceeding 60% (Korea Resource Recirculation Maru, 2025). This suggests that while quantitative collection is high, there is a clear limitation in realizing high-quality recycling.

Wrong separate discharge practices by consumers, such as failure to remove contents and contamination by foreign substances, and the technical limitations of separating composite material packaging are major causes that hinder the recycling process and degrade the quality of recycled content (Geueke et al., 2018; Ragaert et al., 2017). Korea faces chronic limitations within its current EPR system, characterized by low-quality recycled materials and a consequently poor actual recycling rate (The Minjoo Party, 2024). Therefore, securing high-quality recycled resources necessitates a system that encourages consumers to directly return resources of the same material, moving away from the existing mixed discharge method. This suggests the growing need for a resource circulation system based on consumers’ active participation.

As of 2022, packaging accounts for approximately 30% of plastic usage in Korea (OECD, 2024). Food and beverage packaging, in particular, is often produced with composite materials that are difficult to recycle, and contamination with food residue further complicates recycling. Nevertheless, the domestic confectionery and beverage market, which involves the use of difficult-to-recycle packaging, is continuously expanding, with national confectionery and beverage expenditure reaching approximately US$5.7m in 2024, an increase of about 40% compared to 2018 (approximately US$3.6m) (Korea Tourism Organization, 2025).

Consequently, based on the Act on the Promotion of Saving and Recycling of Resources, the Korean government has been implementing the “disposable cup deposit system,” a consumer-targeted resource circulation policy that imposes a resource circulation deposit on disposable cups when ordering beverages at confectionery and beverage stores and refunds the deposit upon cup return, since December 2022. However, contrary to the original plan for nationwide implementation, the system is currently being operated on a reduced scale in Jeju-do (Jeju-si and Seogwipo-si) and Sejong cities and is managed under the discretion of local governments (Kim and Park, 2023). Due to the passive operation, store participation rates have significantly decreased from 94.6% at the initial stage of the system to 44.8% in Jeju-do in 2024, and from a maximum of 64.9% to 31.3% in Sejong (Kim, 2024a, 2024b). Furthermore, it was found that only about half of the cups sold at participating stores are returned, indicating an urgent need for additional measures to improve the resource recovery rate from the consumer perspective.

Resource recovery from the consumer perspective is influenced by multifaceted factors such as infrastructure factors representing accessibility and convenience of return, operating methods of participating stores and regional environment. Previous studies have confirmed that the installation and accessibility of unmanned return machines are key variables affecting consumers’ return rate of reusable waste (Reshetnikova et al., 2025a; Konstantoglou et al., 2023). Specifically, factors such as the presence of cup return machines inside stores participating in the deposit system, the regional distribution and number of collection points outside of stores, operating hours of collection points outside of stores and the possibility of cross-brand cup returns can directly affect consumers’ return behavior (Yang et al., 2026). Therefore, policymakers must closely consider these influencing factors when designing consumer-centric resource circulation systems, thereby enabling the efficient allocation of limited resource circulation funds and the establishment of differentiated policies tailored to regional characteristics.

This study empirically examined factors affecting the consumer’s resource recovery rate from the consumer perspective for Korea’s disposable cup deposit system, which have not been sufficiently analyzed. Previous domestic studies on the disposable cup deposit system in Korea addressed system acceptance (Cho and Nam, 2025), consumer return intention (Lee and Cha, 2024) and changes in café visit demand (Kim and Lee, 2023), but they had the limitation of not adequately considering physical environmental factors such as actual infrastructure and store operation characteristics that could affect the return rate. To bridge this gap, this study used actual operational data from the Container Deposit System Management Organization (COSMO) and empirically analyzed how regional infrastructure, participating store operational characteristics and regional environmental factors influenced consumer resource recovery rates across 48 administrative districts (eup, myeon and dong) in Jeju-si, Seogwipo-si and Sejong-si from January 2024 to August 2025. The significance of this study lies in proposing effective operational measures and specific policy implications for a consumer-participatory resource circulation system based on empirical results to enhance the recovery rate of high-quality recycled resources.

Various social psychological theories have been used to identify the motivations and determinants that induce consumers’ resource circulation behavior. The theory of reasoned action (TRA) is a psychological theory explaining that attitude and subjective norm form behavioral intention, and this intention leads to actual behavior (Yadav et al., 2022). Here, subjective norm refers to the social pressure perceived from a reference group, such as family or local community, regarding the performance of a specific behavior. In other words, TRA suggests that an individual’s positive attitude combined with social norms strengthens behavioral intention, which then leads to behavior (Hoque et al., 2025).

TRA has been used as an important theoretical framework for explaining the motivation for consumers’ recycling and resource circulation behavior. Multiple previous studies have empirically shown through surveys that an individual’s positive attitude and surrounding social pressure significantly increase recycling intention (Valentin et al., 2024; Khan et al., 2020; Strydom, 2018). However, TRA has a limitation in that it is difficult to fully explain behavior when non-volitional factors, such as external environment or personal capacity constraints, intervene, as it assumes that an individual can fully control their behavior by their own will (Arias and Trujillo, 2020). For example, even if an individual has the will to protect the environment and feels social pressure, it is difficult for this to lead to actual recycling behavior if physical constraints, such as the absence or low accessibility of recycling collection bins, exist.

The theory of planned behavior (TPB) was proposed to complement the limitations of TRA, explaining that behavioral intention is determined by the combination of attitude, subjective norm and perceived behavioral control (Zhang et al., 2017). Perceived behavioral control refers to the comprehensive perception of various factors, including external environmental factors, individual ability and resources, that affect the performance of a behavior (Strydom, 2018).

In recycling behavior, perceived behavioral control functions as a key determinant shaped by the complex interaction between individual will and external environmental conditions, such as facility accessibility and the time or effort required for the process (Arias and Trujillo, 2020). Within the specific framework of a DRS, this concept is not just about a person’s psychological state; it depends heavily on physical factors like the density of return points and their operating hours (Reshetnikova et al., 2025b). Specifically, the density of return points and the availability of unmanned return machines directly reduce the logistical friction consumers face, thereby strengthening perceived behavioral control at the aggregate level (Hsiao and Hung, 2025). This aligns with recent cross-national evidence identifying physical accessibility as the primary driver of DRS performance across multiple jurisdictions (Reloop, 2025).

Beyond spatial convenience, the role of participating stores includes critical social influence mechanisms. Integrating descriptive social norm theory (Cialdini, 2003), this study posited that voluntary participation by non-obligated stores functions as a visible and community-level signal. When consumers see local stores voluntarily participating, it sends a clear signal that returning cups is a common and socially responsible practice in their neighborhood. This visible community support functions as a descriptive norm, naturally nudging people to follow suit as they perceive it as the standard way to behave. This norm has been shown to significantly predict actual performance (Lee and Cha, 2024). Consequently, this study built its empirical model on TPB by using store participation and infrastructure variables as objective proxies for perceived behavioral control and community-level social norms.

Based on TPB, this study intended to empirically demonstrate, using Korea’s disposable cup deposit system, that appropriate environmental establishment, such as consumer convenience and accessibility, is essential for enhancing the resource recovery rate from the consumer perspective. Previously, research on waste management in Korea has primarily focused on producer-centered management systems, such as the effectiveness of EPR (Choi and Rhee, 2020; Kang et al., 2023). However, since these studies often analyze material improvements or recycling effectiveness under the assumption that all sold products are recovered, they have failed to adequately account for consumer behavior at the actual recovery stage and regional differences in recovery rates (Ko et al., 2016). Recently, there has been growing discussion that the performance of resource circulation systems is determined not only by the production stage but also by the operational efficiency of the collection, recovery and sorting stages (Tian and Sarkis, 2026); accordingly, interest in recovery systems premised on consumer participation is also increasing.

Research on DRS in Korea is relatively limited and has primarily focused on qualitative reviews of system operations or analyses of the effects of deposit levels (Koh and Park, 2013; Kim, 2024a, 2024b; Kim and Im, 2019). While these studies have demonstrated the importance of economic incentives, they have paid relatively little attention to the physical and operational environmental factors that influence consumers’ actual return behavior. In particular, since the deposit amount remained fixed during the analysis period of this study, this research focused on environmental factors − such as accessibility, operating hours and return infrastructure − rather than price incentives.

Internationally, there has been a growing body of research using factors such as attitudes, social norms and perceived behavioral control to explain recycling and resource circulation behaviors (Strydom, 2018; Khan et al., 2020; Valentin et al., 2024). Furthermore, there has recently been an increase in studies that go beyond survey-based analyses of behavioral intentions and use actual behavioral data to verify the impact of changes in the physical environment and policy design on recycling performance (Akbulut-Yuksel and Boulatoff, 2021). However, existing DRS research is still dominated by comparative case studies (Yi, 2022) or survey-based studies on consumer intentions (Kim and Lee, 2023; Cho and Nam, 2025), and there is a relative lack of research that links behavioral theory to the operational performance of the system using actual operational data.

In particular, while studies on consumer behavior have repeatedly confirmed that consumers prefer stores with return facilities or that social norms increase behavioral intentions (Lee and Cha, 2024), most are based on self-reported intention data, which limits their ability to directly explain observed behavioral outcomes such as actual return rates. Conversely, studies on system operations emphasize institutional factors such as return infrastructure, operational structures and deposit levels but have the limitation of not sufficiently reflecting the mechanisms underlying consumer behavior. In other words, the existing literature has developed with behavioral and operational approaches treated separately, and empirical research linking these two streams remains scarce.

Accordingly, this study aims to extend individual-level behavioral theory to regional-level operational performance to analyze how consumer return behavior manifests as aggregate return rates. Specifically, we used the accessibility of the collection infrastructure, operational characteristics and the level of store participation as proxy variables for perceived behavioral control to verify the relationship between theoretical concepts and actual system performance. Furthermore, by using actual operational data from COSMO, a DRS operator in South Korea, this study sought to conduct an integrated analysis of behavioral and operational factors, which had previously been addressed separately in existing research.

This study analyzed factors affecting the cup return rate by region, using a total of 20 months of data from January 2024 to August 2025, with 14 eup, myeon and dong in Sejong city and 34 in Jeju-do (Jeju-si and Seogwipo-si) as the analysis unit. Under the current system, while return volume tracking is possible through the labels attached to the cups for recycling, there are technical limitations in matching and tracking the sales location and the return location (in-store/out-of-store) of individual cups one-to-one. That is, it is impossible to distinguish whether a cup sold at a specific store was returned to that store, another store or an unmanned return machine, making it difficult to calculate the accurate return rate at the individual store level. Accordingly, this study set the analysis unit as the administrative districts where the stores and return points are located, rather than individual stores. The dependent variable, the regional cup return rate, was calculated as the ratio of the total returned volume at all return points (stores and collection points outside of stores) within the administrative unit to the cup sales volume of all participating stores within that jurisdiction, provided by COSMO.

In this study, the independent variables that can represent consumer accessibility and convenience were established by dividing them into regional infrastructure factors of the deposit system, participating store operation characteristics, whether the participating stores are using standard cups and regional characteristics of the area where the stores are located:

• Based on previous research (Reshetnikova et al., 2025a) indicating that infrastructure such as unmanned return machines significantly affects recycling participation, this study set the number of stores with unmanned return machines installed and the number of collection points outside of stores as variables representing consumer return point accessibility. The number of both in-store and collection points outside of stores was expressed as density by dividing by the area of the respective administrative units. And the average operating hours of collection points outside of stores, representing consumer return convenience, was selected as the regional infrastructure factors variable for the disposable cup deposit system.

• Participating store operation characteristics also represent consumer return point accessibility and return convenience. Konstantoglou et al. (2023) pointed out that the saturation of collection bins or lack of retail store accessibility are major factors hindering returns. Accordingly, this study included the average operating hours of participating stores, the frequency of cup collection by carriers at collection points and the cross-brand return rate as variables. Furthermore, the number of voluntarily participating stores, which join the system despite not being legally obligated, was considered a variable representing the regional community’s willingness to participate in the resource circulation system. To empirically verify the effect of this community-level voluntary participation on the consumer return rate, the number of voluntarily participating stores was included in the model.

• This study aims to verify whether a higher number of stores using standard cups (plain, mandated) with high recycling efficiency for disposable cups in the area leads to differences in the return rate. However, the number of stores using standard cups is a variable that reflects not only demand-side factors such as consumer recycling awareness and brand preference but also the influence of supply/operation and institutional factors, such as the store operator’s willingness to adopt and the provision of incentives by COSMO.

• Finally, regional characteristics were set as control variables. It included the tourism expenditure in the confectionery and beverage sector and the floating population (number of local, domestic, foreign visitors) per eup, myeon and dong, which can structurally influence the total volume of beverage cup sales and returns, using data from the Korea Tourism Data Lab (Korea Tourism Organization, 2025). The definitions of the main variables in this study are as shown in Table 1.

In this study, a time fixed effects model including time dummies, as shown in equation (1), was applied to conduct a panel regression analysis. Furthermore, municipality dummy variables were added as explanatory variables to control for the city-specific characteristics of the three municipalities (Jeju-si, Seogwipo-si and Sejong-si) to which the analysis units (eup, myeon and dong) belong. This was done to secure the reliability of the analysis by controlling for unobserved city-specific characteristics and time-specific bias:

where i and t denoted zone and month, respectively. γ_r and δ_t denoted coefficients of the municipality dummy variables (Jeju-si, Seogwipo-si and Sejong-si) and time fixed effects, respectively.

Table 2 presents the descriptive statistics for the main variables of the 48 administrative units (zones) under analysis. As a result of analyzing the distributional characteristics of the data, this study identified three distinct features:

1. spatial mismatch between consumption and return;

2. concentration of infrastructure and limitations in public accessibility; and

3. the generalization of cross-brand returns.

First is the “spatial mismatch” and concentration phenomenon of cup return behavior. The mean of the dependent variable, return rate (Return, 1.53), is approximately 2.6 times higher than the median (0.58), and the skewness reached 5.27, showing a distribution with a long tail to the right. The fact that the median is less than 1 (0.58) signifies that the majority of regions are “net-consumption zones” where only about half of the sold cups are recovered. In contrast, the maximum value reaching 32.02 shows that in certain areas, more than 32 times the sales volume is being returned. This suggests a clear “waste mobility,” where the locations where consumers purchase cups and where they return them do not coincide. In other words, consumers carry their cups and return them intensively at specific “return hubs,” such as major transportation nodes, government offices or dense office districts (Jejusori, 2023). This non-normality of the data supports that log transformation of the dependent variable is statistically essential for the regression analysis.

Second is the polarization of return infrastructure and constraints on operating hours. The kurtosis for the variables number of collection points outside of stores (Out) and number of stores with unmanned return machines installed (Instore) was very high at 14.73 and 12.43, respectively. This proves that return infrastructure is highly concentrated in only a few commercial and administrative centers, while the vast majority of other regions have no or very insufficient return facilities. It is also noteworthy that the average operating hours of collection points outside of stores (OutHr, 13.67 h), which are public in nature, barely differ from the average operating hours of participating stores (StoreHr, 13.58 h). While unmanned return machines are generally expected to promote consumer convenience through 24-h operation, in reality, most machines are installed inside government offices or buildings, showing hours similar to store business hours. This suggests that public return machines are not sufficiently performing an alternative function to in-store returns.

Third is the generalization of cross-brand returns. The average cross-brand return rate (Xrate) is 0.34, meaning more than one-third of the total return volume occurs at brands other than the one purchased. In particular, the existence of regions where the maximum value reaches 1.0 (100%) empirically demonstrates that consumers prioritize physical convenience at their current location over brand loyalty when choosing a return point. It can be seen that this is the core mechanism sustaining the return rate.

Finally, in the visitor statistics, the fact that the number of local visitors (VstLoc, average approx. 780,000) significantly exceeds the number of domestic tourists (VstDom, average approx. 340,000) suggests that the current deposit system operates based on the daily lifestyle habits of local residents. To ensure statistical reliability, VstDom was excluded from the final model due to a VIF exceeding 10 (10.81). Notably, even after applying log transformations to the main variables, including the return rate, regional tourism expenditure and number of visitors, a re-examination confirmed that all remaining independent variables maintained stable VIF values below 5, ensuring the model’s robustness against multicollinearity.

To evaluate the factors influencing the consumer cup return rate, this study examined four model specifications, ranging from basic pooled OLS to sophisticated panel data models (Table 3). To account for potential heteroskedasticity and serial correlation, robust standard errors are clustered at the regional level. Critically, all four models (M1 through M4) were found to be statistically significant at the 1% level, confirming the overall validity of the model specifications.

This study first introduced regional dummy variables (Jeju-si, Seogwipo-si and Sejong-si) from M2 onwards to effectively control for city-specific characteristics and unobserved spatial variances that are otherwise difficult to capture at the aggregate level. Building upon this, this study applied a log transformation to both the dependent variable and selected independent variables (e.g. regional tourism expenditure and number of visitors) to address large unit differences and extreme variance in the raw data. This refinement not only facilitated a more robust comparison across variables with different scales but also significantly enhanced the model’s explanatory power, as evidenced by the increase in the adjusted R² from approximately 0.25 in the initial models (M1, M2) to 0.46 in the panel specification (M3).

To determine the most appropriate panel specification among these refined models, this study performed a Hausman test comparing the random-effects model (M3) and the fixed-effects model (M4). The result (chi-square statistic = 3.149, p = 0.997) indicated no systematic difference between the two models, statistically justifying the use of the random effects model (M3) as the primary specification. Although the adjusted R² (0.459) indicates that about 46% of the variance is explained, the remaining unexplained variance likely reflects measurement constraints inherent in aggregate-level behavioral data and unmeasured individual-level and community-level factors, such as personal environmental values or neighborhood-specific social norms. Notably, the high consistency in both the direction and significance of coefficients across all specifications, especially between M3 and M4, demonstrates the structural robustness of the findings. This confirms that the empirical results remain stable and reliable, even when accounting for the potential influence of unobserved variables that are inherently difficult to capture in the data.

The most notable finding in the analysis results is that the cross-brand return rate (Xrate) has the strongest positive effect (β = 2.783) on increasing the return rate. Given the log-linear model specification, a 10 percentage-point increase in the cross-brand return rate is associated with an approximately 32% increase in the district-level cup return rate. This finding aligns with the international comparative research by Reloop (2025), which analyzed data from 32 jurisdictions across Europe, North America and Oceania. This global data set identifies systemic accessibility and convenience as the primary structural drivers of DRS performance. According to the TPB, an individual’s behavior is reinforced by “perceived behavioral control,” the belief that they can easily perform that behavior (Arias and Trujillo, 2020). A high cross-brand return rate signifies that an environment has been established where consumers can return cups anywhere without the constraint of having to return to the original place of purchase, suggesting that this may have served as a key factor in reducing the psychological and physical burden of returning cups as perceived by consumers (Konstantoglou et al., 2023). Therefore, rather than considering the installation of additional collection points outside stores to increase return rates, encouraging participation from all brand stores and strengthening the institutional framework that allows consumers to return deposit cups anywhere may contribute to improving return performance.

The number of voluntarily participating stores (Vol) also had a significant positive effect (β = 0.201) on the return rate. This connects to the broader circular economy literature’s emphasis on stakeholder cooperation as a systemic prerequisite. According to the ICC framework (ICC, 2025), consumers must be provided with consistent and easy-to-use entry points to participate in the circular economy. The participation of stores that are not subject to legal obligations, even in a situation where institutional binding force has weakened, proves that an eco-friendly social norm has been formed within the local community. Voluntarily participating stores provide a psychological nudge effect that encourages local residents to join the system. While the participation rate of legally mandatory brand stores is actually decreasing (Kim, 2024a, 2024b), the findings of this study suggest that such a community-based, voluntary (“bottom-up”) ecosystem may be more closely linked to return outcomes than administrative enforcement (“top-down”).

An interesting discovery is that the number of stores with unmanned return machines installed (Instore) actually had a significant negative effect (β = −0.073) on the return rate. This result contradicts the common notion that automated equipment would increase convenience. This result suggests that proximity alone is insufficient if the user experience is friction-generating, a nuance that complements recent findings where “lack of nearby collection points” was a primary barrier (Reloop, 2025). The empirical evidence supports this finding through a descriptive analysis of districts with the highest machine density. In Jeju-si, districts such as Nohyeong-dong and Yeon-dong maintain near-universal coverage, with 77% to 90% of participating stores equipped with machines. However, return rates in these areas remained stagnant near the overall median of 44%, failing to outperform other regions despite the higher infrastructure density.

This alignment between the statistical output and data-driven observations can be interpreted through several factors, first of which is technical difficulty. Unmanned return machines can cause consumers to give up on returning due to issues such as lack of storage space or can generate unnecessary waiting times (Jejusori, 2023). Second is the rigidity of service. In stores where machines are installed, there may be a tendency for staff to force the use of the machine or refuse in-person returns. Furthermore, as the installation of a dedicated application and authentication procedures are mandated when using unmanned return machines, the perceived hassle of using the system increased for consumers. These physical and digital constraints likely acted as factors that hindered consumers’ willingness to return (Kim and Park, 2023; Kang, 2018).

The positive associations observed for the number of local visitors (VstLoc) and the regional dummy variables for Seogwipo-si and Sejong-si suggest that resource circulation behaviors may be positively associated with repeated exposure to the system and routine experiences in everyday settings. Specifically, a 1% increase in the local visitor count is associated with a 0.27% increase in the return rate, reflecting how daily routines play a crucial role in sustaining return behavior. Furthermore, tourism statistics compiled during the study period (January 2024 to August 2025) provide objective evidence of these regional characteristics. Compared to Jeju-si, the total number of visitors was only 32% in Seogwipo-si and 24% in Sejong-si. Similarly, tourism expenditure in the confectionery and beverage sector was just 26% in Seogwipo-si and 15% in Sejong-si relative to Jeju-si. These figures confirm that unlike Jeju-si, which has an overwhelmingly high dependence on the tourism industry and external visitors, Seogwipo-si and Sejong-si possess a much stronger character as residential and administrative living areas. These results indicate that system performance is maximized when return behavior is seamlessly integrated into daily life routines. Exposure to the system within a repetitive daily living area provides a favorable environment for consumers to learn and routinize return behavior. This habituation, which functions as a learning effect, ultimately led to a higher return rate.

This study is significant in that it empirically identified, based on actual operational data, the factors affecting the consumer’s resource recovery rate for Korea’s disposable cup deposit system, which have not been sufficiently analyzed empirically. While existing domestic disposable cup deposit system studies focused on system acceptance, consumer return intention and changes in café visit demand (Cho and Nam, 2025; Lee and Cha, 2024; Kim and Lee, 2023), this study used actual disposable cup deposit system operational data to empirically analyze, through panel regression analysis, the impact of real-world physical and operational conditions on the cup return rate, including equipment operation characteristics, operating features and regional environmental factors that represent consumer accessibility and convenience to return locations.

The study results confirmed that the return rate is determined by openness of return points (cross-brand return rate), the community’s willingness for voluntary participation and daily repetitive performance, rather than the quantitative expansion of physical facilities. Furthermore, the finding that the number of stores with unmanned return machines installed has a negative effect on the return rate is an important implication showing that the expansion of technology-driven infrastructure does not necessarily lead to improved return performance.

Therefore, the results of this study suggest that, rather than prioritizing the expansion of expensive unmanned return machines, policymakers need to consider a strategy that uses all participating stores as return points. However, since this may increase the burden of cup storage and odor management for small-scale stores, win−win support measures, such as differential payment of a handling fee based on cross-brand return performance or shortening the collection cycle to alleviate the stores’ operational burden, must come first.

The result that the number of voluntarily participating stores had a significant positive effect on the return rate suggests that stable return performance is difficult to guarantee solely through system operation centered on legally mandatory businesses, and that the expansion of voluntarily participating stores at the local level may be closely linked to return outcomes. In particular, considering the reality that actual enforcement power over legally mandatory stores has weakened since the system operation method shifted to local government discretion, this result can be interpreted as empirical evidence demonstrating that the return rate depends on the structure of voluntary participation by the local community and stores. Local governments should support the spread of voluntary participation as a community norm by granting these stores “eco-friendly safe store” certifications and providing tangible benefits such as supporting waste bags or reducing local taxes. In addition, there is a need to identify the operational difficulties that small stores actually face during the system participation process through continuous consultation with voluntarily participating stores and to prepare support measures to alleviate their administrative and technical burdens.

Meanwhile, the number of stores with unmanned return machines installed was found to have a negative effect on the return rate, suggesting that the expansion of in-store unmanned return machine installation does not necessarily lead directly to an increase in the return rate. This implies that current returns are not occurring only at stores where in-store unmanned return machines are installed, and that the presence of return machines in stores itself failed to promote cross-brand returns. In light of the negative association between in-store equipment installation and return rates, future infrastructure strategies may benefit from placing greater emphasis on accessibility and user convenience. In this regard, expanding installations beyond individual stores to locations accessible to citizens 24 h a day, such as bus stops and office building lobbies, may offer more flexible and convenient return options. This will reduce the management burden on stores while providing convenience for consumers to access them without business hour constraints.

Moreover, the significant positive effects of the number of local visitors and the regional dummy variables for Seogwipo-si and Sejong-si should be interpreted as interconnected, collectively reflecting the profound influence of daily life routines on return behavior. Considering the study’s limitation in not directly identifying individual consumers’ residency or tourist status, this can be understood as reflecting a spatial characteristic where system participation and return behavior are more stably established due to accumulated exposure and learning about the deposit system in spaces where routine consumption activities are repeated, rather than concluding it as a behavioral difference between local residents and tourists.

These findings of this study suggest that repeated exposure in everyday environments may play a significant role in shaping return behavior. In this context, education and awareness-raising approaches can be considered as complementary measures to reinforce such repeated exposure. For example, integrating information about deposit-refund systems into university and general educational programs, workplace sustainability initiatives or community campaigns can help transform familiarity with the system among diverse groups such as students, workers and local residents into more consistent behavioral practices. Furthermore, policy expansion strategies may benefit from moving beyond a focus on short-term tourists and be designed around community-based spaces where repeated exposure is possible, such as business districts with high concentrations of workers and residents, subway and bus transfer hubs and university campuses. For example, lifestyle-linked strategies such as “lunch-hour intensive collection campaigns” or “double points at return points during commutes” can help strengthen educational and awareness-raising effects while encouraging the transition to actual behavior. This transition may contribute to consumer habit formation, as repeated exposure in everyday settings can help reinforce the long-term behavioral routines associated with a sustainable circular economy.

In interpreting these policy directions, however, it is necessary to consider the inherent limitations of the data used in this study. Since this research relies on aggregate district-level panel data, it cannot fully capture individual-level psychological drivers or the precise cross-district movement of cups, a phenomenon we have defined as “spatial mismatch.” Therefore, the results of this study should be interpreted as reflecting collective behavioral patterns formed under common environmental conditions, rather than as direct evidence of individual consumers’ decision-making mechanisms. Future research using individual tracking data or consumer surveys could further refine the understanding of these unexplained behavioral variances, building upon the structural factors identified in this study.