Toward Resilience: Assessing Retail Location’s Complex Impact Mechanism Using PLS-SEM Aided by Machine Learning

Comments 1: [Fortalecer o arcabouço teórico que sustenta os constructos e esclarecer o conceito de “resiliência” em contextos de varejo;]

Response 1: Thank you for pointing this out. In response, we have revised the Introduction to better situate our concept of RSR(retail space resilience) within the broader academic discourse.

In response, we have revised the Introduction to better situate our concept of RSR within the broader academic discourse. We clarified that while the term "retail resilience" initially referred to the operational sustainability of retailing as an urban function, recent studies have shifted focus toward the resilience of physical retail spaces, typically measured through spatial vitality and viability. The revised content is located from line 31-71.

We adopt this dual definition and explicitly define RSR as “the ability of urban retail spaces to continuously function as containers for commercial activities, encompassing both vitality (the ability to attract footfall) and viability (the ability to operate over time).” These two components are measured through customer footfall and lifespan, respectively, as detailed in the assessment metrics in Section 2. The revised content is located from line 88-92.

Comments 2: [Atualizar e expandir a revisão da literatura, especialmente com contribuições pós-2020 e debates críticos em geografia do varejo, resiliência e estratégias omnicanal;]

Response 2: Your comment is highly constructive and has greatly helped us refine the conceptualization of retail space resilience by incorporating urban planning research related to retail resilience published after 2020. In particular, we have drawn upon studies that examine the impact of geo-spatial location attributes on resilience in retail spaces. These studies have identified several location attributes that have been empirically linked to retail space resilience (RSR), including: accessibility[1,2], amenity[3], agglomeration[4] and socio-demographic patterns[5].

By reviewing recent empirical research on retail space resilience, we found that most studies have focused on a limited set of two to three geo-spatial indicators. However, retail locations encompass multiple dimensions, all of which may influence retail space resilience. This study aims to fill this research gap by investigating the complex impact mechanisms of diverse geo-spatial location attributes on resilience in urban retail spaces.

The revised content is located at line 58-71.

1. Formánek, T.; Sokol, O. Location Effects: Geo-Spatial and Socio-Demographic Determinants of Sales Dynamics in Brick-and-Mortar Retail Stores. J. Retail. Consum. Serv. 2022, 66, 102902, doi:10.1016/j.jretconser.2021.102902.

2. Kickert, C.; Vom Hofe, R.; Haas, T.; Zhang, W.; Mahato, B. Spatial Dynamics of Long-Term Urban Retail Decline in Three Transatlantic Cities. Cities 2020, 107, 102918, doi:10.1016/j.cities.2020.102918.

3. Sung, H. Estimating the Spatial Impact of Neighboring Physical Environments on Retail Sales. Cities 2022, 123, 103579, doi:10.1016/j.cities.2022.103579.

4. Colaço, R.; De Abreu E Silva, J. Commercial Classification and Location Modelling: Integrating Different Perspectives on Commercial Location and Structure. Land 2021, 10, 567, doi:10.3390/land10060567.

5. Parker, C. Multifunctional Centres: A Sustainable Role for Town and City Centres. Institute of Place Management 2015.

Comments 3: [É altamente recomendável incorporar estudos recentes e tematicamente alinhados, como:

Alves, FF, Veloso, CM, Félix, EGS, Sousa, BB, & Valeri, M. (2024). Empreendedorismo e tecnologias de autoatendimento como impulsionadores da fidelização do consumidor ao varejista durante a pandemia de COVID-19 . EuroMed Journal of Business , 20(5), 166–189.

Veloso, CM, & Sousa, BB (2022). Drivers das intenções comportamentais do consumidor e a relação com a qualidade do serviço em contextos específicos do setor . Revista Internacional de Varejo, Distribuição e Pesquisa do Consumidor , 32(1), 43–58.

Pinto, JP, Veloso, CM, Sousa, BB, Walter, CE, & Lopes, E. (2022). Práticas de gestão e consumo (pós) pandémico de marcas próprias: Perspetiva do retalho online e offline num contexto português . Sustentabilidade , 14(17), 10813.

Esses artigos, juntamente com outros estudos relevantes de 2024 e 2025, podem melhorar a base conceitual e empírica do seu manuscrito;]

Response 3: The articles you provided are highly insightful. Collectively, they offer valuable input on how enhancing customer experience through private label strategies and self-service technologies can significantly strengthen consumer loyalty and satisfaction, which in turn improves retail resilience by increasing adaptability and competitiveness during disruptions such as the COVID-19 pandemic.

Inspired by the central themes of these three articles, this study adopts consumer behavior theory as a unifying thread throughout the paper. During the development of the hypothesis pathway framework, we chose to use the three theoretical constructs, which are trip chaining behavior, agglomeration effect, and spatial interaction model, as the central theoretical threads throughout the manuscript. These theories were incorporated into the formulation of hypotheses by drawing from both retail location and consumer behavior literature, forming three key impact pathways. In the Discussion section, we used these same pathways as an analytical framework to interpret the findings, comparing them with assumptions from classical theories to extract the study’s theoretical contributions. Ultimately, the comparison between pre-established hypotheses and empirical results suggests that resilience in urban retail spaces depends more on the demand side, constituted by consumer behaviors and preferences, than on the supply side, such as the inherent physical attributes of retail locations.

Therefore, in discussion 4.2, we cited [Veloso, C.M.; Sousa, B.B. Drivers of Customer Behavioral Intentions and the Relationship with Service Quality in Specific Industry Contexts. Int. Rev. Retail Distrib. Consum. Res. 2022, 32, 43–58, doi:10.1080/09593969.2021.2007977.], the revised manuscript is as follows:

“The underlying reason is that nowadays the retail development gradually transitioned from producer-oriented to consumer-centered[105]. Early retail location theories were developed during the formative stages of physical retail, when retail activity was largely production-driven. However, in today’s context, consumers have access to multiple consumption channels, including e-commerce, which has fundamentally shifted the power dynamic in favor of consumers[106].”

  The revised content is located at line 658-664.

Comments 4: [Esclarecer o papel e a contribuição de cada método, especialmente como Random Forest complementa ou aprimora os resultados de SEM e OLS;]

Response 4: We carefully considered this suggestion and decided to revise the overall research methodology framework by simplifying it to focus on the integration of Random Forest and PLS-SEM, removing the previously included OLS-based pairwise correlation analysis.

Based on a review of existing studies that combine machine learning with PLS-SEM, we used the Random Forest model for preliminary variable selection to identify location attributes with strong explanatory power for RSR. We then constructed a theoretical impact pathway framework by integrating theories from both retail locations and consumer behavior, into which the selected variables were embedded as measurement variables. PLS-SEM was subsequently used to validate these hypothesized pathways. This machine learning-assisted variable selection approach was chosen to ensure a more comprehensive identification of relevant location attributes influencing RSR, allowing us to develop a more robust and explanatory impact pathway framework.

Additionally, we introduced three new location attribute variables, point density, parking facilities, and social media data (from Weibo), and renamed the original social media variable from Dianping as “consumer review”. This was done to maintain sufficient number of variables after Random Forest filtering, thereby supporting a meaningful PLS-SEM analysis.

The results validated this approach: the newly added variables were all identified by Random Forest as having notable contributions to RSR, and they replaced previously lower-contributing LA variables such as global integration, consumer review (Dianping), and KDE. These modifications confirm the validity of the revised assessment metrics and ensure that all variables included in the final PLS-SEM model meaningfully contribute to RSR, thereby enhancing the overall reliability of the study.

The revised content is located at line 314-355.

Comments 5: [Aprofundar a interpretação dos resultados, conectando as descobertas com insights conceituais e não apenas resultados estatísticos;]

Response 5: Thank you for your in-depth suggestion. In response, we have made substantial revisions to the discussion of our results to ensure that the findings offer meaningful theoretical contributions to the understanding of retail space resilience. Below are the specific changes made in accordance with your feedback:

（1）We have downplayed the negative effect of scale in the discussion, as it was not the strongest impact pathways in our results. This finding has been repositioned in Section 4.3 of the results discussion. Additionally, we situated this result within more recent academic debates on urban retail resilience. Specifically, we referenced several representative studies that highlight key challenges associated with large-scale retail spaces, including limited pedestrian accessibility, overly standardized management practices, and the lack of personal bonding with consumers, and therefore together it provides a theoretical basis for interpreting our finding.

To strengthen the credibility of this finding, we referenced two highly cited empirical studies on retail resilience. A 2011 report by the UK retail location data provider Local Data Company (LDC) [6] found that larger retail centers exhibited higher vacancy rates and lower sales performance compared to small and medium-sized retail spaces. Additionally, Enoch M. et al. (2020)[7] studied the post-lockdown vitality of six large-scale shopping centers on UK high streets and found that smaller centers experienced less drop in vitality, while larger centers saw footfall levels decline by 57–75%.

Finally, to clarify the boundary of this study’s findings, the discussion reiterates that the path coefficient of scale on RSR is relatively weak, indicating that scale is not necessarily negative in all contexts. Drawing on another empirical study, we also acknowledge that large-scale retail centers can attain resilience when successfully integrated with diverse lifestyle-oriented experiences.

The revised content is located at line 679-722.

（2） As mentioned in Response 4, to avoid data redundancy and confusion while improving the model’s explanatory power, we removed the moderating effect of lifespan and instead incorporated it as a second measurement variable of RSR.

Additionally, we acknowledge that the original discussion lacked a thorough explanation of the full pathway from Sociodemography → Amenity → RSR. To address this, we significantly revised the theoretical hypothesis section, shifting from a purely data-driven approach to one that’s with theoretical grounding. After conducting preliminary variable selection using Random Forest, we proposed hypothetical impact pathways based on three key theoretical perspectives: agglomeration effects, spatial interaction models, and trip chaining behavior. These revisions are reflected in Section 3.2.2, Theoretical Hypothesis.

Trip Chaining Behavior serves as the theoretical basis that connects H7 and H8, specifically linking the pathway from Sociodemography → Amenity → RSR. Trip Chaining behavior is a part of consumer behavior theoretical disciplines that focus on the rationale behind how consumers commit in multi-purpose shopping trips. As mentioned in the assessment metric system of section 2.3.2, consumptions and socializing tend to happen between a chain of trips. Therefore, consumers tend to patronize the locations that are able to provide opportunities for a diverse range of activities, which in return would contribute in the resilience of retail spaces. Also, mentioned in the trip chaining behavior theories, whether the initiative is a household or work-related trips is a decisive factor that whether there will be a chance to a chain of other triggered events, including consumption. Residential household tend to trigger a chain of interdependent trips. While work-related trips are relatively low in chances to be happen in a chain of trips, because work is mostly the primary and sole purpose of trips. Based on this theoretical rationale, we propose: H7 Amenity positively influence Retail Space Resilience, and H8 Socio-demographic attributes around a retail location influences the urban amenity attributes.

The revised content is located at line 425-512.

(3) Due to the expansion of the assessment metrics, we introduced a new variable, parking facilities, which demonstrated stronger explanatory power for RSR and, based on Random Forest results, replaced the previously included but low-performing variable, global integration. As a result, the path coefficient from accessibility to RSR in the revised PLS-SEM model is 0.291, indicating a moderate effect, and its p-value confirms the validity of this relationship. 

Furthermore, in the discussion section “4.1 Empirical Validation of Classical Location Attributes and the Emerging Primacy of Amenity in Retail Space Resilience,” we have clarified that although accessibility is a direct influencing variable, it should not be interpreted as a primary driving factor.

These findings suggest that while the assumptions regarding accessibility in classical retail location theories are supported by our empirical results, the target variable in this study, RSR, reflects a more contemporary understanding of retail landscapes, which have evolved significantly since the emergence of retail location theory in the 1930s. Consequently, Amenity and Trip Chaining Behavior, both rooted in consumer behavior patterns, emerge as the location attributes and impact pathways of primary importance in shaping resilience outcomes in today’s urban retail environments.

The revised content is located at line 585-678.

Comments 6: [Articular implicações políticas e gerenciais claras e acionáveis;]

Response 6: Thank you for pointing this out. In response, we substantially revised the Conclusion section to ensure that the final policy recommendations are closely aligned with the empirical findings.

Drawing on the theoretical contributions highlighted in the Discussion, we proposed three targeted and actionable policy suggestions that directly respond to the key impact pathways identified in the analysis. This study recommends reorienting retail planning around consumers’ daily mobility patterns—such as commuting or leisure-driven trip chains—by creating mixed-use, walkable retail nodes within residential and employment hubs. Rather than treating malls as isolated destinations, planning should integrate them into everyday urban flows to increase spontaneous consumption opportunities. Additionally, to strengthen the positive effects of retail agglomeration while mitigating homogenization, the study suggests establishing a “retail cluster quality assessment mechanism” to evaluate functional diversity and consumer differentiation, along with incentive tools like tax breaks or rent subsidies to encourage complementary and differentiated retail mixes.

The revised content is located at line 742-766.

Comments 7: [Revise o manuscrito para maior clareza e consistência, especialmente em relação à terminologia técnica e à estrutura das frases.]

Response 7: Thank you for your feedback. Through substantial revisions across the manuscript, we aimed to more clearly articulate our research goal. The specific modifications made throughout the paper are as follows:

(1) In the Introduction section, we added a clear articulation of the research gap to strengthen the rationale behind our research objective:

"While the existing literature on retail resilience is extensive, only a limited number of studies have approached the topic from a spatial perspective, contextualizing retail resilience within the framework of urban geography. Among those, most have examined only two or three geo-spatial variables. However, retail location is inherently multi-dimensional, and various spatial attributes may influence the resilience of brick-and-mortar retail spaces. This study aims to fill this research gap by investigating the complex impact mechanisms of multiple geo-spatial location attributes on urban retail space resilience."

(2) This study aims to clearly convey the narrative of retail space resilience through a restructuring and enrichment of the Introduction. This introduction starts with an academic narrative by connecting a real-world issue, which is the decline of physical retail spaces in urban areas, to a focused research question on retail resilience. Based on the REPLACIS academic project in Europe and other relevant literature, we established the definition of Retail Space Resilience, which are constituted by vitality and viability. We highlight a research gap in current studies that the resilience in terms of urban physical retail spaces are overlooked and the complexity arises from multiple location attributes’ impact towards RSR is not yet discovered. By gradually narrowing the focus and referencing relevant theories, it sets up the rationale for examining multiple geo-spatial attributes. The proposed mixed-method approach is introduced modestly to support this aim. The revised content is located at line31-108.

At the same time, we recognize that articulating a clear and coherent research "story" requires a solid theoretical foundation. The original data-driven framework of pathway construction lacked the theoretical depth needed to align with the study’s objectives. Therefore, in the revised Theoretical Hypothesis section, we used the Random Forest model as a tool for preliminary variable selection, and subsequently grounded the proposed hypothetical impact pathways in two academic disciplines: retail location theory and consumer behavior theory. This approach ensures that the analytical framework is both data-informed and theoretically coherent. The revised content is located at line425-512. 

Based on the Spatial Interaction Model, we propose that both the scale and accessibility of retail centers can foster consumers’ social interactions, forming the theoretical basis for Hypotheses H1 to H3. Drawing on the concept of Agglomeration Effects, we suggest that accumulated social network resources and transportation infrastructure at retail locations act as key drivers of retail agglomeration, which in turn influences Retail Space Resilience (RSR), leading to Hypotheses H4 to H6. Finally, grounded in consumer behavior theory, specifically Trip Chaining Behavior, we argue that whether consumers are classified as workers or residents affects their travel patterns and demand for multi-purpose trips, which shapes the supply of surrounding amenities and ultimately impacts RSR. This rationale supports Hypotheses H7 and H8. The revised content is located at line376-522.

(3) In the discussion of our experimental results, we acknowledge that the original version was indeed superficial. Therefore, we have substantially rewritten the Discussion section, placing greater emphasis on the theoretical contributions revealed by the findings. The revised discussion highlights how the results diverge from classical retail location theories and articulates the underlying rationale behind the newly identified impact pathways, thereby strengthening the study’s theoretical significance and relevance.

The first part of our discussion is retitled to be “Empirical Validation of Classical Location Attributes and the Emerging Primacy of Amenity in Retail Space Resilience”, which is mainly discussing the direct influencing location attributes, Agglomeration, Accessibility and Amenity. The main finding is that while classical retail location theories emphasize immediate customer attraction, this result shows that long-term resilience depends more on creating diverse, experience-oriented environments for consumers. Additionally, the study reveals a shift from traditional centrality toward infrastructure-based accessibility, reflecting evolving urban mobility and consumer expectations.

The second part of the discussion is”The constraints of Agglomeration effects and the Growing Influence of Consumer Trip Chaining”. in this section we compare and examine multiple impact pathways in greater detail, adding theoretical depth and complexity to the discussion.

While agglomeration effects do influence urban retail in Shanghai, consumer trip chaining behavior exerts a stronger impact on RSR. This reflects a shift from production-oriented to consumer-centered retail dynamics, where mobility patterns and multi-purpose urban trips increasingly shape resilience outcomes. Additionally, the study expands traditional trip chaining theory by showing that both residential and employment-based consumer flows contribute to retail vitality, especially in dense, amenity-rich business districts.

The third part of the discussion is “The Limited Influence of Scale: Rethinking Spatial Interaction Models in Modern Retail Environments”. This study finds the scale has a limited and negative mediating effect on RSR, challenging assumptions in classical spatial interaction models that larger size imposes larger attraction towards consumers. Therefore, in order to enhance the rigor of the test result, we included recent literature as well as empirical evidence and suggested that low tenant flexibility and weak personal bonding with consumers in large malls may undermine their resilience. 

(4) Finally, the value of this study is the reframing of retail location planning by emphasizing consumers’ mobility patterns and spatial perceptions as key drivers of retail space resilience, shifting the focus from destination-based design to integrated, everyday urban pathways. It contributes a machine-learning aided methodological approach that combines Random Forest model and PLS-SEM to uncover the complex, multi-pathway impacts of geo-spatial location attributes on retail space resilience.

4. Response to Comments on the Quality of English Language

Response 1: A professional academic editing service was engaged to improve both the clarity and overall quality of the manuscript, ensuring that the language rigorously and accurately conveys the research objectives.

Comments 1: [Your research topic is about the complex influence mechanism of retail location on the resilience of retail space. However, some fundamental concepts have not been clearly explained yet, and there may even be errors. Retail space resilience (RSR) is defined as the ability to maintain vitality and viability, without clearly distinguishing it from existing concepts such as retail vitality and sustainability. In addition, location attributes (LAs) contain 12 indicators, but the classification logic and the boundary between them and sociodemographic are not explained in the theoretical framework, resulting in a lack of rigor in the variable system.]

Response 1: Thank you for pointing this out. You are absolutely right that the original manuscript lacked a clear connection between our definition of Retail Space Resilience (RSR) and the existing literature on retail resilience, which is essential for clarifying the research subject.

In response, we have revised the Introduction to better situate our concept of RSR within the broader academic discourse. We clarified that while the term "retail resilience" initially referred to the operational sustainability of retailing as an urban function, recent studies have shifted focus toward the resilience of physical retail spaces, typically measured through spatial vitality and viability. The revised content is located from line 31-71.

We adopt this dual definition and explicitly define RSR as “the ability of urban retail spaces to continuously function as containers for commercial activities, encompassing both vitality (the ability to attract footfall) and viability (the ability to operate over time).” These two components are measured through customer footfall and lifespan, respectively, as detailed in the assessment metrics in Section 2. The revised content is located from line 88-92. 

Additionally, your suggestion to embed the classification logic of location attributes (LAs) within a theoretical framework is well taken. The original version lacked sufficient theoretical grounding for the assessment metrics. In the revised Section 2.3.2, we now provide a detailed explanation of each LA, including its conceptual basis and current measurement approaches. These are drawn from two academic perspectives: retail location theory (supply-side) and consumer behavior theory (demand-side), allowing us to build a comprehensive and theoretically informed metric system. The revised content is located at line 164-313.

The theoretical structure also sets the stage for later findings, where consumer-related factors are shown to play a dominant role in shaping retail resilience, and thus reinforcing the internal consistency of our analysis. We have included a summary table in the response letter to clearly map each LA to its corresponding theoretical framework.

Comments 2: [The authors claim to build a data-driven framework by combining machine learning, OLS and PLS-SEM, but you can't just list various methods, which makes it seem that the connections among the methods lack rationality]

Response 2: We carefully considered this suggestion and decided to revise the overall research methodology framework by simplifying it to focus on the integration of Random Forest and PLS-SEM, removing the previously included OLS-based pairwise correlation analysis and employing RF as a preliminary variable selection tool for subsequent SEM analysis.

Based on a review of existing studies that combine machine learning with PLS-SEM, we used the Random Forest model for preliminary variable selection to identify location attributes with strong explanatory power for RSR. We then constructed a theoretical impact pathway framework by integrating theories from both retail locations and consumer behavior, into which the selected variables were embedded as measurement variables. PLS-SEM was subsequently used to validate these hypothesized pathways. This machine learning-assisted variable selection approach was chosen to ensure a more comprehensive identification of relevant location attributes influencing RSR, allowing us to develop a more robust and explanatory impact pathway framework.

Additionally, we introduced three new location attribute variables, which are point density, parking facilities, and social media data (from Weibo), and renamed the original social media variable from Dianping as “consumer review”. This was done to maintain sufficient number of variables after Random Forest filtering, thereby supporting a meaningful PLS-SEM analysis.

The results validated this approach: the newly added variables were all identified by Random Forest as having notable contributions to RSR, and they replaced previously lower-contributing LA variables such as global integration, consumer review (Dianping), and KDE. These modifications confirm the validity of the revised assessment metrics and ensure that all variables included in the final PLS-SEM model meaningfully contribute to RSR, thereby enhancing the overall reliability of the study.

The revised content is located at line 314-355.

Comments 3: [Symbol error: "β==0.142" (Repeat the equal sign, it should be "β=-0.142")]

Response 3: Thank you very much for your thorough review of the manuscript. We have carefully rechecked the entire text and confirmed that there are no errors.

Comments 4: [There are obvious flaws in the model's validity. The R² of the retail vitality of PLS-SEM is 0.416. Its explanatory power is relatively weak, but it is called strong explanatory power, which violates the basic standard of model fitting.]

Response 4: Thanks for pointing this out. To improve the reliability and validity of the model, we made two key modifications. 

First, in constructing the assessment metrics, we introduced three new LA variables.  Based on their feature importance in the Random Forest analysis, they replaced three previously included variables with lower explanatory power. Second, we removed the moderating effect of lifespan in the PLS-SEM path framework and instead incorporated it as a second measurement indicator of the target variable, Retail Space Resilience (RSR), alongside with footfall.

These adjustments increased the R² value for RSR to 0.485, which, as defined in the manuscript, falls within the moderate explanatory range in PLS-SEM and avoids any misleading interpretation of “strong explanatory power.”

The revised content is located at line 544-548.

Additionally, according to Barbara M. Byrne (2013)[1], W.W. Chin (1998)[2], Jörg Henseler 2009[3], the commonly accepted R² thresholds in PLS-SEM are as follows: <0.19 = very weak, 0.19–0.33 = weak, 0.33–0.67 = moderate, and ≥0.67 = substantial explanatory power. Furthermore, in recent PLS-SEM studies by Amir Zaib Abbasi et al. (2024)[4], and Karoui Sedki et al. 2024[5], the reported R²values were 0.296 and 0.350 respectively, further supporting the conclusion that the explanatory power of our model falls within an acceptable and comparable range.

1. Byrne, B.M. Structural Equation Modeling with Mplus: Basic Concepts, Applications, and Programming; Routledge: New York, 2013; ISBN 978-0-203-80764-4.

2. Chin, W.W. The Partial Least Squares Approach to Structural Equation Modeling. In Modern Methods for Business Research; Psychology Press, 1998 ISBN 978-1-4106-0438-5.

3. Henseler, J.; Ringle, C.M.; Sinkovics, R.R. The Use of Partial Least Squares Path Modeling in International Marketing. In New Challenges to International Marketing; Sinkovics, R.R., Ghauri, P.N., Eds.; Emerald Group Publishing Limited, 2009; Vol. 20, p. 0 ISBN 978-1-84855-468-9.

4. Customer Engagement in Saudi Food Delivery Apps through Social Media Marketing: Examining the Antecedents and Consequences Using PLS-SEM and NCA. J. Retail. Consum. Serv. 2024, 81, 104001, doi:10.1016/j.jretconser.2024.104001.

5. Karoui, S.; Behi, A.T.; Fehri, D.; Belaid, S.; Lacoeuilhe, J. Predicting Label Brand Loyalty: A Comparison of Two Models Using a Partial Least Square-Structural Equation Modeling. J. Retail. Consum. Serv. 2024, 79, 103852, doi:10.1016/j.jretconser.2024.103852.

6. Wrigley, N.; Dolega, L. Resilience, Fragility, and Adaptation: New Evidence on the Performance of UK High Streets during Global Economic Crisis and Its Policy Implications. Environ Plan A 2011, 43, 2337–2363, doi:10.1068/a44270.

7. Enoch, M.; Monsuur, F.; Palaiologou, G.; Quddus, M.A.; Ellis-Chadwick, F.; Morton, C.; Rayner, R. When COVID-19 Came to Town: Measuring the Impact of the Coronavirus Pandemic on Footfall on Six High Streets in England. Environ. Plann. B: Urban Anal. City Sci. 2022, 49, 1091–1111, doi:10.1177/23998083211048497.

Comments 5: [The explanation of the core results remains at the descriptive level without delving into the underlying mechanisms: (1) Location scale has a negative mediating effect (β=-0.016), which contradicts the law of retail gravitation, but it does not explain why the small-scale retail space in Shanghai is more resilient. It is only attributed to the historical texture and lacks empirical support. (2) The retail lifespan reduces the dependence of location convenience on the social population, but does not explain the formation path of this independence. (3) Among the path coefficients, the influence of accessibility → retail vitality (β=0.05) is weak, but it is still regarded as a key driving factor, and the logic is not consistent.]

Response 5: Thank you for your in-depth suggestion. In response, we have made substantial revisions to the discussion of our results to ensure that the findings offer meaningful theoretical contributions to the understanding of retail space resilience. Below are the specific changes made in accordance with your feedback:

（1）First, we have downplayed the negative effect of scale in the discussion, as it was not the strongest impact pathways in our results. This finding has been repositioned in Section 4.3 of the results discussion. Additionally, we situated this result within more recent academic debates on urban retail resilience. Specifically, we referenced several representative studies that highlight key challenges associated with large-scale retail spaces, including limited pedestrian accessibility, overly standardized management practices, and the lack of personal bonding with consumers, and therefore together it provides a theoretical basis for interpreting our finding.

To strengthen the credibility of this finding, we referenced two highly cited empirical studies on retail resilience. A 2011 report by the UK retail location data provider Local Data Company (LDC) [6] found that larger retail centers exhibited higher vacancy rates and lower sales performance compared to small and medium-sized retail spaces. Additionally, Enoch M. et al. (2020)[7] studied the post-lockdown vitality of six large-scale shopping centers on UK high streets and found that smaller centers experienced less drop in vitality, while larger centers saw footfall levels decline by 57–75%.

Finally, to clarify the boundary of this study’s findings, the discussion reiterates that the path coefficient of scale on RSR is relatively weak, indicating that scale is not necessarily negative in all contexts. Drawing on another empirical study, we also acknowledge that large-scale retail centers can attain resilience when successfully integrated with diverse lifestyle-oriented experiences. 

The revised content is located at line 679-722.

（2） First, as mentioned in Response 4, to avoid data redundancy and confusion while improving the model’s explanatory power, we removed the moderating effect of lifespan and instead incorporated it as a second measurement variable of RSR. The revised content is located at line 514-522.

Additionally, we acknowledge that the original discussion lacked a thorough explanation of the full pathway from Sociodemography → Amenity → RSR. To address this, we significantly revised the theoretical hypothesis section, shifting from a purely data-driven approach to one that’s with theoretical grounding. After conducting preliminary variable selection using Random Forest, we proposed hypothetical impact pathways based on three key theoretical models: agglomeration effects, spatial interaction models, and trip chaining behavior. These revisions are reflected in Section 3.2.2, Theoretical Hypothesis(line 425-512)

Trip Chaining Behavior serves as the theoretical basis that connects H7 and H8, specifically linking the pathway from Sociodemography → Amenity → RSR. Trip Chaining behavior is a part of consumer behavior theoretical disciplines that focus on the rationale behind how consumers commit in multi-purpose shopping trips. As mentioned in the assessment metric system of section 2.3.2, consumptions and socializing tend to happen between a chain of trips. Therefore, consumers tend to patronize the locations that are able to provide opportunities for a diverse range of activities, which in return would contribute in the resilience of retail spaces. Also, mentioned in the trip chaining behavior theories, whether the initiative is a household or work-related trips is a decisive factor that whether there will be a chance to a chain of other triggered events, including consumption. Residential household tend to trigger a chain of interdependent trips. While work-related trips are relatively low in chances to be happen in a chain of trips, because work is mostly the primary and sole purpose of trips. Based on this theoretical rationale, we propose: H7 Amenity positively influence Retail Space Resilience, and H8 Socio-demographic attributes around a retail location influences the urban amenity attributes.

(3) Due to the expansion of the assessment metrics, we introduced a new variable, parking facilities, which demonstrated stronger explanatory power for RSR and, based on Random Forest results, replaced the previously included but low-performing variable, global integration. As a result, the path coefficient from accessibility to RSR in the revised PLS-SEM model is 0.291, indicating a moderate effect, and its p-value confirms the validity of this relationship.

Furthermore, in the discussion section “4.1 Empirical Validation of Classical Location Attributes and the Emerging Primacy of Amenity in Retail Space Resilience,” we have clarified that although accessibility is a direct influencing variable, it should not be interpreted as a primary driving factor.

These findings suggest that while the assumptions regarding accessibility in classical retail location theories are supported by our empirical results, the target variable in this study, RSR, reflects a more contemporary understanding of retail landscapes, which have evolved significantly since the emergence of retail location theory in the 1930s. Consequently, Amenity and Trip Chaining Behavior, both rooted in consumer behavior patterns, emerge as the location attributes and impact pathways of primary importance in shaping resilience outcomes in today’s urban retail environments.

The revised content is located at line 585-678.

Comments 6: [The language of this article needs to be polished by researchers from English-speaking countries. Your logic is not clear and your research goals are rather vague. It was just an experiment conducted, but it lacked an interesting story, a reasonable explanation, and self-consistent logic. Finally, the value of your research still needs to be further refined.]

Response 6: Thank you for your feedback. Through substantial revisions across the manuscript, we aimed to more clearly articulate our research goal. The specific modifications made throughout the paper are as follows:

(1) In the Introduction section(line72-79), we added a clear articulation of the research gap to strengthen the rationale behind our research objective:

"While the existing literature on retail resilience is extensive, only a limited number of studies have approached the topic from a spatial perspective, contextualizing retail resilience within the framework of urban geography. Among those, most have examined only two or three geo-spatial variables. However, retail location is inherently multi-dimensional, and various spatial attributes may influence the resilience of brick-and-mortar retail spaces. This study aims to fill this research gap by investigating the complex impact mechanisms of multiple geo-spatial location attributes on urban retail space resilience."

(2) This study aims to clearly convey the narrative of retail space resilience through a restructuring and enrichment of the Introduction. This introduction starts with an academic narrative by connecting a real-world issue, which is the decline of physical retail spaces in urban areas, to a focused research question on retail resilience. Based on the REPLACIS academic project in Europe and other relevant literature, we established the definition of Retail Space Resilience, which are constituted by vitality and viability. We highlight a research gap in current studies that the resilience in terms of urban physical retail spaces are overlooked and the complexity arises from multiple location attributes’ impact towards RSR is not yet discovered. By gradually narrowing the focus and referencing relevant theories, it sets up the rationale for examining multiple geo-spatial attributes. The proposed mixed-method approach is introduced modestly to support this aim. The revised content is located at line 31-108.

At the same time, we recognize that articulating a clear and coherent research "story" requires a solid theoretical foundation. The original data-driven framework of pathway construction lacked the theoretical depth needed to align with the study’s objectives. Therefore, in the revised Theoretical Hypothesis section, we used the Random Forest model as a tool for preliminary variable selection, and subsequently grounded the proposed hypothetical impact pathways in two academic disciplines: retail location theory and consumer behavior theory. This approach ensures that the analytical framework is both data-informed and theoretically coherent. The revised content is located at line425-512.

Based on the Spatial Interaction Model, we propose that both the scale and accessibility of retail centers can foster consumers’ social interactions, forming the theoretical basis for Hypotheses H1 to H3. Drawing on the concept of Agglomeration Effects, we suggest that accumulated social network resources and transportation infrastructure at retail locations act as key drivers of retail agglomeration, which in turn influences Retail Space Resilience (RSR), leading to Hypotheses H4 to H6. Finally, grounded in consumer behavior theory, specifically Trip Chaining Behavior, we argue that whether consumers are classified as workers or residents affects their travel patterns and demand for multi-purpose trips, which shapes the supply of surrounding amenities and ultimately impacts RSR. This rationale supports Hypotheses H7 and H8. The revised content is located at line376-522.

(3) In the discussion of our experimental results, we acknowledge that the original version was indeed superficial. Therefore, we have substantially rewritten the Discussion section, placing greater emphasis on the theoretical contributions revealed by the findings. The revised discussion highlights how the results diverge from classical retail location theories and articulates the underlying rationale behind the newly identified impact pathways, thereby strengthening the study’s theoretical significance and relevance.

The first part of our discussion is retitled to be “Empirical Validation of Classical Location Attributes and the Emerging Primacy of Amenity in Retail Space Resilience”, which is mainly discussing the direct influencing location attributes, Agglomeration, Accessibility and Amenity. The main finding is that while classical retail location theories emphasize immediate customer attraction, this result shows that long-term resilience depends more on creating diverse, experience-oriented environments for consumers. Additionally, the study reveals a shift from traditional centrality toward infrastructure-based accessibility, reflecting evolving urban mobility and consumer expectations.

The second part of the discussion is”The constraints of Agglomeration effects and the Growing Influence of Consumer Trip Chaining”. in this section we compare and examine multiple impact pathways in greater detail, adding theoretical depth and complexity to the discussion.

While agglomeration effects do influence urban retail in Shanghai, consumer trip chaining behavior exerts a stronger impact on RSR. This reflects a shift from production-oriented to consumer-centered retail dynamics, where mobility patterns and multi-purpose urban trips increasingly shape resilience outcomes. Additionally, the study expands traditional trip chaining theory by showing that both residential and employment-based consumer flows contribute to retail vitality, especially in dense, amenity-rich business districts.

The third part of the discussion is “The Limited Influence of Scale: Rethinking Spatial Interaction Models in Modern Retail Environments”. This study finds the scale has a limited and negative mediating effect on RSR, challenging assumptions in classical spatial interaction models that larger size imposes larger attraction towards consumers. Therefore, in order to enhance the rigor of the test result, we included recent literature as well as empirical evidence and suggested that low tenant flexibility and weak personal bonding with consumers in large malls may undermine their resilience. 

(4) To enhance the self-consistency of the paper, we chose to use the three theoretical constructs, which are trip chaining behavior, agglomeration effect, and spatial interaction model, as the central theoretical threads throughout the manuscript. These theories were incorporated into the formulation of hypotheses by drawing from both retail location and consumer behavior literature, forming three key impact pathways. In the Discussion section, we used these same pathways as an analytical framework to interpret the findings, comparing them with assumptions from classical theories to extract the study’s theoretical contributions. Ultimately, the comparison between pre-established hypotheses and empirical results suggests that resilience in urban retail spaces depends more on the demand side, constituted by consumer behaviors and preferences, than on the supply side, such as the inherent physical attributes of retail locations.

Lastly, we substantially revised the Conclusion section to ensure that the final policy recommendations are closely aligned with the empirical findings. Drawing on the theoretical contributions highlighted in the Discussion, we proposed three targeted and actionable policy suggestions that directly respond to the key impact pathways identified in the analysis. This study recommends reorienting retail planning around consumers’ daily mobility patterns—such as commuting or leisure-driven trip chains—by creating mixed-use, walkable retail nodes within residential and employment hubs. Rather than treating malls as isolated destinations, planning should integrate them into everyday urban flows to increase spontaneous consumption opportunities. Additionally, to strengthen the positive effects of retail agglomeration while mitigating homogenization, the study suggests establishing a “retail cluster quality assessment mechanism” to evaluate functional diversity and consumer differentiation, along with incentive tools like tax breaks or rent subsidies to encourage complementary and differentiated retail mixes.

Finally, the value of this study is the reframing of retail location planning by emphasizing consumers’ mobility patterns and spatial perceptions as key drivers of retail space resilience, shifting the focus from destination-based design to integrated, everyday urban pathways. It contributes a machine-learning aided methodological approach that combines Random Forest model and PLS-SEM to uncover the complex, multi-pathway impacts of geo-spatial location attributes on retail space resilience.

4. Response to Comments on the Quality of English Language

Response 1: A professional academic editing service was engaged to improve both the clarity and overall quality of the manuscript, ensuring that the language rigorously and accurately conveys the research objectives.

Comments 1: [The theoretical tension of the research question is not sufficiently embedded in the existing theoretical framework of “urban retail decline-resilience”. Although the concept of “retail space resilience” is introduced, the article does not systematically review the core literature in the field, nor does it distinguish the theoretical boundary between ‘resilience’ and “vitality”.]

Response 1: Thank you for pointing this out. You are absolutely right that the original manuscript lacked a clear connection between our definition of Retail Space Resilience (RSR) and the existing literature on retail resilience, which is essential for clarifying the research subject.

In response, we have revised the Introduction to better situate our concept of RSR within the broader academic discourse. We clarified that while the term "retail resilience" initially referred to the operational sustainability of retailing as an urban function, recent studies have shifted focus toward the resilience of physical retail spaces, typically measured through spatial vitality and viability. The revised content is located from line 31-71.

We adopt this dual definition and explicitly define RSR as “the ability of urban retail spaces to continuously function as containers for commercial activities, encompassing both vitality (the ability to attract footfall) and viability (the ability to operate over time).” These two components are measured through customer footfall and lifespan, respectively, as detailed in the assessment metrics in Section 2. The revised content is located from line 88-92. 

Comments 2: [The basis of variable selection is vague and lacks clear theoretical attribution logic. Although 12 indicators of location attributes are used, the interpretation of their categorization and causal structure is mainly based on empirical speculation, without solid theoretical support, and the distinction and connection between variables are not clear.]

Response 2: Your suggestion to embed the classification logic of location attributes (LAs) within a theoretical framework is well taken. The original version lacked sufficient theoretical grounding for the assessment metrics. In the revised Section 2.3.2, we now provide a detailed explanation of each LA, including its conceptual basis and current measurement approaches. These are drawn from two academic perspectives: retail location theory (supply-side) and consumer behavior theory (demand-side), allowing us to build a comprehensive and theoretically informed metric system. The revised content is located at line 164-313.

The theoretical structure also sets the stage for later findings, where consumer-related factors are shown to play a dominant role in shaping retail resilience, and thus reinforcing the internal consistency of our analysis. We have included a summary table in the response letter to clearly map each LA to its corresponding theoretical framework.

Comments 3: [The combination of methods shows a tendency of “technology stacking” and lacks reasonable logic of causal modeling. Random Forest, OLS, PLS-SEM and other methods are used at the same time, but the positioning, use and logical relationship between the results of each method are not clearly explained, which poses the risk of “model plattering”.]

Response 3: We carefully considered this suggestion and decided to revise the overall research methodology framework by simplifying it to focus on the integration of Random Forest and PLS-SEM, removing the previously included OLS-based pairwise correlation analysis and employing RF as a preliminary variable selection tool for subsequent SEM analysis.

Based on a review of existing studies that combine machine learning with PLS-SEM, we used the Random Forest model for preliminary variable selection to identify location attributes with strong explanatory power for RSR. We then constructed a theoretical impact pathway framework by integrating theories from both retail locations and consumer behavior, into which the selected variables were embedded as measurement variables. PLS-SEM was subsequently used to validate these hypothesized pathways. This machine learning-assisted variable selection approach was chosen to ensure a more comprehensive identification of relevant location attributes influencing RSR, allowing us to develop a more robust and explanatory impact pathway framework.

Additionally, we introduced three new location attribute variables—point density, parking facilities, and social media data (from Weibo)—and renamed the original social media variable from Dianping as “consumer review.” This was done to maintain sufficient number of variables after Random Forest filtering, thereby supporting a meaningful PLS-SEM analysis.

The results validated this approach: the newly added variables were all identified by Random Forest as having notable contributions to RSR, and they replaced previously lower-contributing LA variables such as global integration, consumer review (Dianping), and KDE. These modifications confirm the validity of the revised assessment metrics and ensure that all variables included in the final PLS-SEM model meaningfully contribute to RSR, thereby enhancing the overall reliability of the study.

The revised content is located at line 314-355.

Comments 4: [The causal assumptions of PLS-SEM model are mostly derived from algorithms, and the theoretical support is weak. The article emphasizes “data-driven framework”, but structural equation modeling is essentially based on the logical construction of causal assumptions, and without clear theoretical and logical support, it is easy to lead to the logical inversion of “fitting data to cause and effect”.]

Response 4: We recognize that articulating a clear and coherent research "story" requires a solid theoretical foundation. The original data-driven framework of pathway construction lacked the theoretical depth needed to align with the study’s objectives. Therefore, in the revised Theoretical Hypothesis section, we used the Random Forest model as a tool for preliminary variable selection, and subsequently grounded the proposed hypothetical impact pathways in three theoretical models: Spatial Interaction Model, Aggomeration effect and Trip Chaining behavior, rooted from retail location and consumer behavior theories. This approach ensures that the analytical framework is both data-informed and theoretically coherent.

Based on the Spatial Interaction Model, we propose that both the scale and accessibility of retail centers can foster consumers’ social interactions, forming the theoretical basis for Hypotheses H1 to H3. Drawing on the concept of Agglomeration Effects, we suggest that accumulated social network resources and transportation infrastructure at retail locations act as key drivers of retail agglomeration, which in turn influences Retail Space Resilience (RSR), leading to Hypotheses H4 to H6. Finally, grounded in consumer behavior theory, specifically Trip Chaining Behavior, we argue that whether consumers are classified as workers or residents affects their travel patterns and demand for multi-purpose trips, which shapes the supply of surrounding amenities and ultimately impacts RSR. This rationale supports Hypotheses H7 and H8.

The revised content is located at line 425-512.

Comments 5: [The representativeness and external validity of the sample are doubtful, and it is difficult to generalize the results of the single-site study. The data only come from six core urban areas in Shanghai, and there is no boundary discussion on the specificity of the sample in this city, which affects the extrapolation of the conclusions.]

Response 5: Thanks for pointing this out. Therefore we added the clarification of the reason why the research area is chosen to be the 6 core urban area. The scope of this empirical study is to capture variations in RSR over a long historical trajectory. The study focuses on six central districts in Shanghai—Huangpu, Xuhui, Changning, Jing’an, Putuo, and Hongkou. Since the 1840s, these areas were Shanghai’s Concession established by foreign powers such as Britain and France, during which time the concept of the department store was introduced to Shanghai, making it one of the earliest urban retail hubs in China. These districts have since witnessed multiple historical transformations, from the concession era to the founding of the People’s Republic and the era of economic reform. Despite these transitions, many retail areas in the city center—such as Nanjing Road, Huaihai Road, and Chenghuangmiao—have continued to operate to this day, demonstrating strong resilience. This long-standing and dynamic retail history makes Shanghai’s central districts a highly valuable context for studying urban retail space resilience.

The revised content is located at line 111-122.

Comments 6: [There is a risk of covariance and redundancy in the definition of variables and the construction of indicators. For example, Footfall and Lifespan are both used to define RSR, and the latter is used as a moderating variable, which is a theoretical and methodological confusion.]

Response 6: Thanks for pointing this out. To improve the reliability and validity of the model, we made modifications to the pathway framework.

In order to avoid covariance and redundancy, we removed the moderating effect of lifespan in the PLS-SEM path framework and instead incorporated it as a second measurement indicator of the target variable, Retail Space Resilience (RSR), alongside with footfall, Shown in Figure 9 in the revised manuscript.

The revised content is located at line 514 - 522.

Comments 7: [The causal direction of the path of key variables is unclear. Some variables, such as “Scale”, are negative mediators in the model, but the explanations in the text are unclear and lack of mechanistic analysis, resulting in a lack of self-consistency in the explanations.]

Response 7: Thank you for your in-depth suggestion. In response, we have made revisions to the discussion of how Scale impacts RSR.

First, we have downplayed the negative effect of scale in the discussion, as it was not the strongest impact pathways in our results. This finding has been repositioned in Section 4.3 of the results discussion. Additionally, we situated this result within more recent academic debates on urban retail resilience. Specifically, we referenced several representative studies that highlight key challenges associated with large-scale retail spaces, including limited pedestrian accessibility, overly standardized management practices, and the lack of personal bonding with consumers, and therefore together it provides a theoretical basis for interpreting our finding.

Second, to strengthen the credibility of this finding, we referenced two highly cited empirical studies on retail resilience. A 2011 report by the UK retail location data provider Local Data Company (LDC) [6] found that larger retail centers exhibited higher vacancy rates and lower sales performance compared to small and medium-sized retail spaces. Additionally, Enoch M. et al. (2020)[7] studied the post-lockdown vitality of six large-scale shopping centers on UK high streets and found that smaller centers experienced less drop in vitality, while larger centers saw footfall levels decline by 57–75%.

Finally, to clarify the boundary of this study’s findings, the discussion reiterates that the path coefficient of scale on RSR is relatively weak, indicating that scale is not necessarily negative in all contexts. Drawing on another empirical study, we also acknowledge that large-scale retail centers can attain resilience when successfully integrated with diverse lifestyle-oriented experiences.

The revised content is located at line 679-722.

Comment 8: [There is an overlapping problem in the identification of mediating and moderating effects, and the path dependency relationship is not fully analyzed. Multiple paths are significant at the same time, but there is no explanation of why these complex interactions are formed, nor is there any control analysis of the conflict of interpretation brought about by multiple parallel paths.]

Response 8: Thank you for this critical suggestion. Indeed, as there’s a coexistence of multiple impact pathways, a control analysis of multiple parallel paths and a comparison among the paths area needed. Therefore, we have substantially rewritten the Discussion section, placing greater emphasis on viewing multiple pathways altogether, and conclude findings based on the comparisons of the coexisting influencing variables and pathways.

The first part of our discussion is retitled to be “Empirical Validation of Classical Location Attributes and the Emerging Primacy of Amenity in Retail Space Resilience”, which is mainly discussing the three direct influencing location attributes, agglomeration, accessibility and amenity. The main finding is that while classical retail location theories emphasize immediate customer attraction, this result shows that long-term resilience depends more on creating diverse, experience-oriented environments for consumers. Additionally, the study reveals a shift from traditional centrality toward infrastructure-based accessibility, reflecting evolving urban mobility and consumer expectations.

The second part of the discussion is”The constraints of Agglomeration effects and the Growing Influence of Consumer Trip Chaining”. In this section we compare and examine multiple impact pathways in greater detail, adding theoretical depth and complexity to the discussion.

While agglomeration effects do influence urban retail in Shanghai, consumer trip chaining behavior exerts a stronger impact on RSR. This reflects a shift from production-oriented to consumer-centered retail dynamics, where mobility patterns and multi-purpose urban trips increasingly shape resilience outcomes. Additionally, the study expands traditional trip chaining theory by showing that both residential and employment-based consumer flows contribute to retail vitality, especially in dense, amenity-rich business districts.

The third part of the discussion is “The Limited Influence of Scale: Rethinking Spatial Interaction Models in Modern Retail Environments”. This study finds the scale has a limited and negative mediating effect on RSR, challenging assumptions in classical spatial interaction models that larger size imposes larger attraction towards consumers. Therefore, in order to enhance the rigor of the test result, we included recent literature as well as empirical evidence and suggested that low tenant flexibility and weak personal bonding with consumers in large malls may undermine their resilience.

The revised content is located at line 585-678.

Comments 9: [The policy recommendations are too broad and not precisely aligned with specific empirical findings. For example, although the recommendations of “encouraging TOD” and “integrating open space” are directional, they do not accurately match the findings of this paper's path analysis or the differential impacts.]

Response 9: Thank you for your feedback. Through substantial revisions across the manuscript, we aimed to ensure that the final policy recommendations are closely aligned with the empirical findings. Drawing on the theoretical contributions highlighted in the Discussion, we proposed three targeted and actionable policy suggestions that directly respond to the key impact pathways identified in the analysis. This study recommends reorienting retail planning around consumers’ daily mobility patterns—such as commuting or leisure-driven trip chains—by creating mixed-use, walkable retail nodes within residential and employment hubs. Rather than treating malls as isolated destinations, planning should integrate them into everyday urban flows to increase spontaneous consumption opportunities. Additionally, to strengthen the positive effects of retail agglomeration while mitigating homogenization, the study suggests establishing a “retail cluster quality assessment mechanism” to evaluate functional diversity and consumer differentiation, along with incentive tools like tax breaks or rent subsidies to encourage complementary and differentiated retail mixes.

Finally, the value of this study is the reframing of retail location planning by emphasizing consumers’ mobility patterns and spatial perceptions as key drivers of retail space resilience, shifting the focus from destination-based design to integrated, everyday urban pathways. It contributes a machine-learning aided methodological approach that combines Random Forest model and PLS-SEM to uncover the complex, multi-pathway impacts of geo-spatial location attributes on retail space resilience.

The revised content is located at line 742-766.

Comments 10: [The concept of “resilience” is not sufficiently enhanced, and the theoretical contribution is not sufficiently expressed. Although the article attempts methodological innovation, its contribution to the dimensional expansion, mechanism construction, or measurement innovation of the concept of “retail space resilience” is still weak.]

Response 10: Thank you for your constructive feedback. In response to your comment that the concept of “resilience” is not sufficiently enhanced and that the theoretical contribution requires further clarification, we have undertaken a series of substantial revisions throughout the manuscript to address these concerns directly.

1. Clarification and Strengthening of the Concept of Resilience

To begin with, we acknowledged that our initial definition of “retail space resilience” (RSR) was insufficiently anchored in the broader academic discourse. In the revised Introduction, we have now clearly situated RSR within the evolution of retail resilience literature. We drew a distinction between general “retail resilience” as the operational sustainability in urban retail systems and their the ability to adapt to shocks, and our specific conceptualization of RSR as the resilience in physical retail spaces, encompassing both vitality (customer footfall) and viability (operational lifespan). This distinction helps delineate the scope of our study and situates it within the more spatially grounded tradition of urban geography and retail location theory. The revised content is located at line31-108.

2. Theoretical Contribution and Mechanism Construction

In response to your concern regarding weak theoretical contribution, we have restructured the section on theoretical hypotheses to explicitly ground our framework in three academic disciplines: the Spatial Interaction Model, Agglomeration Effect Theory, and Consumer Behavior Theory, specifically Trip Chaining Behavior. Each of these theories contributes to the construction of distinct impact pathways (H1–H8) linking retail location attributes to RSR. This shift from a purely data-driven pathway construction to a theory-integrated hypothesis framework enhances the theoretical contribution of our study and creates a structured basis for interpreting empirical results. The revised content is located at line376-522.

Furthermore, in the Discussion section, we now directly compare these pathways with assumptions in classical theories. For example, while the Spatial Interaction Model suggests that larger-scale retail centers should foster more interactions and thus greater resilience, our findings show that scale has a limited and sometimes negative effect on resilience in today’s urban context. This contradiction forms the basis of a refined theoretical insight: that consumer-oriented factors now outweigh traditional supply-side advantages, marking a transition in the underlying drivers of retail resilience. The revised content is located at line 584-722. 

3. Dimensional Expansion and Measurement Innovation

To enhance conceptual rigor, we introduced a more nuanced assessment metric system in Section 2.3.2. This assessment metrics system is now derived from the two academic disciplines of retail locations and consumer behavior, to make sure the variable selections are rigorously grounded in theories and not derived from experiences. Then, we employed a hybrid methodological approach combining Random Forest for variable selection and PLS-SEM for validating complex causal pathways. This simplification of methodology fixed the previous problem of confusing relationship between models and the risk of ‘model plattering’. now the relationship between Random Forest model is clearer, which is a preliminary variable selection tool for subsequent PLS-SEM analysis. RF model is executed to recognize the LA variables that are indeed contributing to RSR, and the variables with low contributions(global integration, KDE, and ConsumerReview) are screened out to ensure the validity and significance of the PLS-SEM test results.

The revised content is located at line 314-355.

4. Policy and Practical Relevance

Finally, we restructured the Conclusion to tightly align theoretical contributions with policy recommendations. We emphasize that enhancing retail resilience requires shifting planning logics from mono-functional commercial zones to integrated, trip-chaining-enabled community-oriented retail nodes.

The revised content is located at line 742-765.

We hope these comprehensive revisions strengthen the manuscript’s theoretical foundation and clarify its contribution to the study of retail space resilience.

4. Response to Comments on the Quality of English Language

Response 1: A professional academic editing service was engaged to improve both the clarity and overall quality of the manuscript, ensuring that the language rigorously and accurately conveys the research objectives.