Commercial dynamics in urban China during the COVID-19 recession: Vulnerability and short-term adaptation of commercial centers in Shanghai

Abstract

Studying the commercial dynamics during the COVID-19 recession could help deepen our understanding of how the pandemic damages the commercial economy and how to against the pandemic. This study aims to explore the vulnerability and adaptation of commercial centers using a weekly consumption data of UnionPay cards in Shanghai. A vulnerability index and multiscale geographically weighted regressions (MGWR) are employed. Our results suggest that retail, leisure, and entertainment sectors are less vulnerable to the pandemic at the early stage, when catering, life service, and wholesale sectors are more influenced. Catering, life service, and wholesale sectors were better adapted to the second wave of the pandemic, while the retail and entertainment sectors were even more vulnerable. Further analysis using MGWR models suggests that the commercial centers with higher consumption volume are better adapted to the shock. The diversity of commercial sectors mainly reduces low-level commercial centers' vulnerability to the pandemic. The commercial centers targeting high-end consumers with wider hinterland were less adapted to the pandemic. These research outcomes reveal the disparities in commercial centers' vulnerability against COVID-19 and highlight adaptation's role during the pandemic.

1. Introduction

The prevalence of COVID-19 has shocked the global economy with economic downturns, relocation of companies, shrinking consumption, and a sharp increase in unemployment (Nicola et al., 2020, 2020; van Ham et al., 2021). The pandemic has significantly affected people's economic activities in commercial centers, representing urban vitality (Florida et al., 2021). Thus, studying the changes in consumption activities during COVID-19 helps deepen our understanding of how the pandemic damages the commercial economy and how to against the pandemic.

Economic geographers often use the concept of vulnerability to analyze the sensitivity of economic phenomena to risk scenarios (Martin & Sunley, 2015). A comparative analysis before and during the economic recession is commonly used to identify the vulnerability, describing how much the economic phenomena have been damaged (Abbasi-Shavazi, 2021; Papatheodorou & Pappas, 2017). Adaptation describes another dimension of how the economic phenomenon is influenced by the risk scenarios (Hu & Hassink, 2017; Marquet & Miralles-Guasch, 2018). So far, most of the frameworks regarding vulnerability and adaptation during the economic recession were developed based on the financial crisis in 2008. The prevalence of COVID-19 provides a good opportunity to understand the underlying mechanism of vulnerability and adaptation of economic phenomena against a deadly pandemic.

So far, there are some empirical studies available on how the pandemic changed the economic system. Bartik et al. (2020) found that small businesses are highly vulnerable to COVID-19 and do not tend to recover after the pandemic because of limited access to supporting programs. Business in the central urban regions of Miami declined during the pandemic, contributing to commercial suburbanization (Li & Stoler, 2022). The role of state power on economic vulnerability was revealed by a case study in north-eastern China that goes beyond the structure of economic factors (Hu et al., 2022). Other studies also provide a more detailed vision of the vulnerability of a specific sector, such as international trade, international logistics, etc. (Ivanov & Dolgui, 2020; Klassen & Murphy, 2020; Yu et al., 2021; Zong et al., 2021). These studies focus on how COVID-19 damage the regional economy; however, little attention has been paid to how economic phenomena adapt to the pandemic. Also, the underlying mechanism still needs further exploration.

In this research, we examine the performance of commercial centers during the COVID-19 recession from the perspectives of vulnerability and short-term adaptation. More specifically, this study employs a UnionPay card dataset in Shanghai to explore the changing pattern of urban consumption during the prevalence of COVID-19. A vulnerability index is employed as a quantitative tool, and the MGWR model helps reveal the underlying mechanism. The research outcomes would address the following research questions: (1) What is the spatial-temporal pattern of commercial centers' vulnerability and adaptation against COVID-19? (2) What are the determinants of vulnerability and adaptation of the commercial centers against COVID-19? (3) What useful policy implications, if any, can enhance the performance of urban commercial centers against COVID-19 and crises of that kind in the future? The research outcomes are expected to provide a new perspective for the analytical framework of the commercial dynamics of global cities during the COVID-19 recession.

2. Literature review and hypotheses

2.1. Vulnerability, short-term adaptation, and determinants

Since the 21st century, the world has increasingly suffered from shocks of various risks and crises, and the dynamics of regional economy aftershocks have aroused great attention from scholars (Giannakis & Bruggeman, 2017a; Martin, 2012). In urban and economic geography, there was limited attention to vulnerability before the prevalence of COVID-19. Martin and Sunley (2015) provided a comprehensive framework for economic resilience against risk scenarios. The vulnerability was highlighted as a critical component, which was defined as the sensitivity of the economic phenomenon to the risk scenario at the early stage of the disaster. So far, many indicators can help capture economic vulnerability, such as employment, per capita incomes, foreign trade, air pollution by economic activities, urban amenities, etc. (Cheng et al., 2022; Martin, 2012; Martin & Sunley, 2015; Martin et al., 2016; Zhai & Yue, 2022), for the shocks such as successive economic crises, climate change and COVID-19 at the regional and national scales (Brakman et al., 2015; Cellini & Torrisi, 2014; Courvisanos et al., 2016; Dubé & Polèse, 2016; Eraydin, 2016; Sensier et al., 2016; Van Bergeijk et al., 2017; Wang et al., 2022; Xiao et al., 2022).

In Martin & Sunley's framework (2015), another notion was mentioned, namely "robustness", referring to how economic entities adapt to risk scenarios. However, this notion is less mentioned in urban and geography research. We found that the definition of "robustness" is similar to adaptation, which is popular in urban ecology and climate research (Carter et al., 2015; Georgescu et al., 2014; Santini et al., 2019). During COVID-19, adaptation has been widely discussed as the pandemic is relieving the issues such as online learning strategies (Xhelili et al., 2021) and pediatric emergencies (Tan et al., 2020). However, economic adaptation to the pandemic is less discussed in current literature, which needs further exploration.

Drawing upon these backgrounds, we construct a framework to simplify the complex commercial dynamics during the COVID-19 recession, making all the processes easier to be expressed quantitatively. We measure vulnerability as the decline of the economic performance of the objects before and during the shocks. Adaptation is usually classified into short-term and long-term adaptations, which are usually influenced by a different mechanisms (Keeler, McNamara, & Irish, 2018; Xiao et al., 2022). In this research, we only focus on short-term adaptation, describing the disparities in economic entities' performance during two different waves of the pandemic.

Although there are already some discussions about the conceptions of vulnerability and adaptation, the underlying mechanism of the disparities are still unclear regarding economic geography. The economic structure is suggested as a critical determinant of the differentiated performance of the regional economies to shock and disturbance (Holm & Østergaard, 2015; Breathnach et al., 2015; Martin et al., 2016; Brown & Greenbaum, 2017). It is believed that diversification of the industrial structure is conducive to resisting economic recession because a diversified industrial structure can disperse the impact of shocks and act as a "shock absorber" when shock disturbances occur, compensating for the impact of shocks (Angulo et al., 2018; Cainelli et al., 2019; Martin & Sunley, 2015). Conversely, regions with a single industry are vulnerable to regional lock-in, as it is challenging to find new substitutes in the short term once the dominant industry has declined significantly (Frenken & Boschma., 2007; Crescenzi et al., 2016). More specifically, the regions dominated by heavy industry are more vulnerable, and financial services or enterprises tend to be less influenced (Martin, 2012; Navarro- Espigares et al., 2012; Simmie & Martin, 2010). Additional attention has been paid to human capital and neighborhood environment factors, as more urbanized regions with more creative classes and skilled labor force tend to be less economically vulnerable (Brakman et al., 2015). Also, Martin et al. (2013) suggested that the agglomeration of firms weakens the vulnerability.

Although plenty of studies have tried to explore the internal determinants of vulnerability, insufficient attention has been paid to the impact of regional external linkages. Regions with high external economic linkages tend to be economically advanced. The linkages contribute to the mobility of the resources, technologies, and global assets for regional development with high efficiency, which can improve the ability to combat the crisis (Eraydin, 2016; Giannakis & Bruggeman, 2017b; Martin & Sunley, 2015). However, the prevalence of COVID-19 shut down inter-regional and global linkages, and global resistance against the pandemic tells a different story.

2.2. Impact of COVID-19 on urban commercial economy

Consumer behavior is people's primary daily economic activity, which is also a critical indicator of the regional economy. Before the pandemic, people's consumption preferences has been significantly changed as online shops are replacing bricks and mortar stores. Therefore, there are plenty of commercial areas undergoing structural changes in shopping real estate even before the spring of 2020 (Dolega & Lord, 2020; Singleton et al., 2016). The prevalence of COVID-19 would accelerate this process and has dramatically changed people's consumption behavior (Donthu & Gustafsson, 2020; Eger et al., 2021). Social-distancing rules and the fear of infection due to close contact limited the appetite for face-to-face consumption. People have to change their regular consumption habits, and online consumption has become increasingly popular (Sheth, 2020). Therefore, recently published studies (e.g., Pantano et al., 2020) show a shift towards the online channel, potentially fulfilling Cairncross's (2001) prediction of geography's death in retail activities.

Even though people still choose to use multiple consumption modes depending on the types of products and services, purchase volumes, and the accessibility to the commercial center. Since noticeable differences exist in consumers' demand for different commercial formats, the degree of a recession of different commercial formats varies regarding the impact of a pandemic. For example, the community-level physical outlets of local commercial centers do continue to play a fundamental role in the local provision of goods and often serve as a local social hub (Clarke & Banga, 2010). Regarding the low cost of delivery and transportation fee, these local commercial centers would be preferred by low-income people and elders (Moon et al., 2021; Taha et al., 2021).

Face-to-face contact would be avoided unnecessary consumption during a pandemic (Dannenberg et al., 2020; Pantano et al., 2020; Truong & Truong, 2022). For example, shopping centers and restaurants have been stroke by COVID-19 because people's mobility is limited, and the social distance prevents people from gathering (Beckers et al., 2021). The high-end commercial centers that provide high-quality services or products are less likely to be replaced by online shops, which are more likely to survive during the pandemic (Eger et al., 2021). However, the operation of these shopping centers requires better accessibility to high-quality consumers, while the pandemic still influences people's preference for long travel (Beckers et al., 2021).

Furthermore, commercial managers would change their strategies to better adapt to people's consuming behavior to prevent the decline of physical commercial formats during the pandemic (Verma & Gustafsson, 2020). This condition employs a profound integration between channels (Verhoef et al., 2015). Such strategies provide more synergies than cannibalization and offer competitive merchant advantages over pure online players (Herhausen et al., 2015). Compared to large companies such as Amazon, Alibaba, or Zalando, small businesses are usually disadvantaged and react slower (Quinn et al., 2013). Small businesses have to transform from physical forms to online channels to prevent further losses during the pandemic (Bartik et al., 2020).

2.3. Research framework and hypotheses

The discussions above suggest that people's consuming behavior is influenced by factors such as the potential consumers nearby, the industries' diversity, and the size of the shopping centers. Also, the adaptation of the commercial centers would manifest heterogeneity regarding scale, space, and time. Drawing upon these backgrounds, three research hypotheses will be tested concerning the vulnerability and short-term adaptation of commercial centers against COVID-19 (see Fig. 1 ).

Fig. 1.

Analytical framework for commercial dynamics during the COVID-19 recession.

At the early stage of the pandemic, most activities with physical gatherings are restricted to prevent further spread. This has inevitably led to disruptions and a decline in the commercial economy's supply and demand. However, heterogeneity exists due to the differentiated scales and structures of the commercial centers. Following the literature above suggests that diverse industrial structures and high volume would decrease vulnerability and better adaptation, the first hypothesis is proposed.

H1

Commercial centers with a huge consumption volume and high consumption diversity are less vulnerable and better adapted to COVID-19.

On the one hand, the pandemic has restricted people's mobility, especially inter-city travel. Therefore, distant consumers would be prevented, and the commercial center with a wider consumption hinterland might be influenced. On the other hand, the economic recession significantly affects the residents' income and economic status, leading to a shrinking consumption (Nicola et al., 2020). In this condition, the socioeconomically disadvantaged population groups would be more influenced, while the high-income population would be less vulnerable (Consolazio et al., 2021).

H2

Commercial centers with a wider hinterland and fewer high-end consumers are more vulnerable to COVID-19.

Since COVID-19 was completely new and highly contagious, the commercial sectors were unable to respond quickly and effectively to the shock at the very beginning. Subsequently, the merchants introduced countermeasures and continued exploring strategies to restore their operating capacity. Each commercial format has already accumulated experience, and the transformation of the operation strategy of merchants can promote commercial centers to cope with the COVID-19 shock. After recovering from the first wave of the pandemic, commercial centers tend to be less vulnerable to the pandemic.

H3

The commercial centers' location, size, and consumption structure would influence adaptation to the pandemic.

3. Study area, data and methodology

3.1. Study area

This study is implemented in Shanghai, a critical hub in the global city network and a gateway to Asia. Shanghai has the most active economic activities in China and the most diverse commercial centers. By 2019, the GDP of Shanghai was 3815.53 billion yuan. In recent years, the commercial centers in Shanghai have become the foundation of the transformation from monocentric to polycentric urban structures. A hierarchical structure of commercial urban centers has been created during the process of suburbanization, and structural flattening is emerging.

By May 31, 2022, over 649,341 confirmed cases of COVID-19 had been reported in Shanghai since the pandemic outbreak. As a commercial capital, Shanghai has the most vibrant commercial system with a nationwide hinterland, including cities abroad. Shanghai is also a high-density city, with over 24.28 million permanent population and a high rate of urbanization. The high human mobility of Shanghai makes it riskier for the wide spread of the pandemic. Regarding the epidemic prevention policy in China, Shanghai implemented the strictest control policy to stop the spread of the virus in the early stage of the pandemic. The first case was found in Shanghai on January 20, 2020 (Fig. 2 ). Shanghai implemented the travel control policy on January 23. Later, the spread of COVID-19 was defined as a Critical Incident Level 1 Response on January 24. On February 8, Shanghai was partially lockdown in most areas, and non-essential workers were encouraged to work from home while the public transit system continued to operate but with reduced schedules. As the pandemic evolved, the government shifted its focus from the most strict control policy in February 2020 to the flexible control policy in late March to maintain its financial and economic function since its urban commercial economy has been severely impacted. Even though this partial lockdown was effective in mitigating the pandemic, it was disruptive to economic activities. While the GDP of the first two quarters of 2020 in Shanghai was down 2.6% compared with the same period in 2019, the wholesale and retail trade, business services, and accommodation and catering sectors have dropped by 9.4%, 11.6%, and 31.3%, respectively.

Fig. 2.

Changes in the number of existing confirmed cases in Shanghai.

We select Shanghai for the following reasons. First, Shanghai has a flexible control policy compared to other Chinese global cities like Beijing, which is significantly influenced by political power. Thus, the research outcomes in Shanghai could be compared with other global cities such as New York and London, contributing to broader implications. Second, the loosened control policy provides a good opportunity for us to study how the pandemic changed the market and people's consumption preferences. So far, the drop in commercial centers has been detected, and the research can improve the understanding of commercial vulnerability and adaptation during the COVID-19 recession. Furthermore, Shanghai is a typical emerging global city in developing countries, and the case study provides good empirical experiences for the sustainable commercial development of other cities. Thus, how the commercial centers in Shanghai respond to the pandemic is worthy of exploration.

This study focuses on 205 commercial centers in 16 administrative districts (Fig. 3 ). We classify the commercial centers into four categories, following He et al.'s methodology (2020). The first-level commercial centers contain all the 21 municipal-level commercial centers in Shanghai. The second-level commercial centers contain all 48 district-level commercial centers. The third-level commercial centers contain 70 large- and middle-scale community-level commercial centers. The fourth-level commercial centers contain the 66 small-scale community-level commercial centers in Shanghai. The urban regions of Shanghai have been classified into three for describing the spatial heterogeneity in the following sections, including the central urban regions within the inner ring road, the ring road regions between the inner ring road and outer ring road, and the outer urban regions out of the outer ring road.

Fig. 3.

The distribution of commercial centers in Shanghai.

As documented in China, there are several waves of COVID-19, such as the Delta and Omicron virus. In 2020, confirmed cases in Shanghai peaked in the 7th week, and the second wave came in the 14th week (Fig. 2). In this case, we try to identify commercial centers' vulnerability and short-term adaptation to the virus, which could reflect commercial dynamics during the COVID-19 recession. We selected week 7 and week 14 as the typical time sample of the first two round outbreak periods in 2020 to analyze the changes in the vulnerability of commercial centers in Shanghai as the pandemic progresses.

3.2. Data

Big data techniques have been applied in urban and economic research, which have achieved favorable research outcomes (Zhou et al., 2022). In this study, the spatiotemporal dynamics of economic outcomes are expected to be examined at the intra-urban level. China UnionPay, a joint organization of bank cards in China, provides a temporally continued dataset and detailed information on consumption in every commercial center. Through the interbank transaction system, China UnionPay realizes interconnection and resource sharing among commercial banking systems to ensure the use of bank cards across banks, regions, and even borders. The UnionPay Card data applied in this paper contains the interbank transaction data of all UnionPay cards. In 2018, Union pay reached cooperation with "ANT finance" and "Wechat pay" to enhance the central government's control of the financial system in China. Thus, Union pay has covered almost all the transactions in China, including Alipay and Wechat pay. According to this dataset, there was a total of 28.75 million transactions in 2019, and the total consumption of the 205 commercial centers in Shanghai was 1269.34 billion yuan, accounting for 33.27% of the total GDP. The weekly consumption data of UnionPay Card from the 2nd to the 22nd week in 2019 and 2020 contains information on the attribution location and consumption portrait of cardholders and records consumption information such as the volume, frequency, and category in real-time, which can capture the urban commercial economy before and after the pandemic.

3.3. Methodology

3.3.1. Measurement of the vulnerability of commercial centers

Based on our theoretical framework, we take the consumption volume of Unionpay cards as the indicator of the commercial centers under the shock of the COVID-19 pandemic to measure the vulnerability, which can reflect how people's consumption behavior is sensitive to the pandemic. We construct a vulnerability index which can be expressed as follows (Giannakis & Bruggeman, 2017b).

$${Vulnerability}_r=\frac{Y_r^{pt}-Y_r^{bt}}{Y_r^{bt}}$$ (1)

$${Vulnerability}_i=\frac{\left(Y_i^{pt}-Y_i^{bt}\right)/\left(Y_i^{bt}\right)-\left(Y_r^{pt}-Y_r^{bt}\right)/\left(Y_r^{bt}\right)}{\left|\left(Y_r^{pt}-Y_r^{bt}\right)/\left(Y_r^{bt}\right)\right|}$$ (2)

Where, ${Vulnerability}_r$ is the commercial vulnerability of the whole region r; ${Vulnerability}_i$ is the vulnerability of the commercial center i in the region; $Y_i^{bt}$ and $Y_r^{bt}$ are the UnionPay card consumption volume of the commercial center i and the whole region r respectively at the base year week t; $Y_i^{pt}$ and $Y_r^{pt}$ are the UnionPay card consumption volume of the commercial center i and the whole region r respectively at the pandemic outbreak year week t; bt is the week t of the base year 2019; pt is the week t of the pandemic outbreak year 2020; ${Vulnerability}_i>0$ indicates that the impact of the COVID-19 pandemic on the commercial center i is less than that on the whole region average level, which is considered as less vulnerability and vice versa.

3.3.2. Selection of influential factors of vulnerability

Based on our assumptions, we select nine factors to study the vulnerability and short-term adaptation mechanism of Shanghai's urban commercial centers (Table 1 ). Consumption volume can reflect the prosperity of commercial centers. Generally, a larger consumption volume indicates a more prosperous commercial center with a higher level (Lin et al., 2019). Consumption structure can reflect the diversity of products and services the commercial center provides. According to the weekly UnionPay card consumption data, five major commercial formats, including catering, retail,1 wholesale, life services, and leisure and entertainment, were selected for consumption structure and diversity analysis. We use the Shannon diversity index of consumption type to reflect the consumption diversity of a commercial center.

Table 1.

Selection and explanation of variables of influential factors of vulnerability.

Variable	Definition	Unit	Min	Max	Mean	VIF	Tolerance
Consumption volume (X1)	Consumption sum	Million	0.11	6408.48	102.90	1.221	0.819
Consumption diversity (X2)	Shannon diversity index of consumption type		0.20	2.48	1.94	1.084	0.923
Catering proportion (X3)		%	0	100	13.18	1.075	0.930
Retail proportion (X4)		%	0	100	53.01	1.148	0.871
Wholesale proportion (X5)		%	0	96.20	16.04	13.543	0.074
Life services proportion (X6)		%	0	97.98	11.34	15.324	0.065
Leisure and entertainment proportion (X7)		%	0	47.9	6.44	1.218	0.821
Consumption hinterland (X8)	The proportion of cardholders from other cities	%	1.08	78.26	19.24	1.316	0.760
Consumer attribute (X9)	The proportion of high-end consumers	%	0	79.40	9.47	1.289	0.775

Consumption hinterland refers to the catchment area of a commercial center that attracts customers, which is indicated by customers' origin. We use the index of the proportion of UnionPay cardholders from other cities to reflect the consumption hinterland of a commercial center. The higher the level of the commercial center, the wider its hinterland boundary and the stronger its attractiveness to alien consumers (Christaller, 1933).

UnionPay made the consumption portrait of cardholders and divided the cardholder into eight groups.2 Among them, high-end consumers primarily contribute to luxury consumption and account for the largest consumption volume. Generally speaking, the higher level of the commercial center would have a higher proportion of high-end commercial goods and services and target high-end people (Christaller, 1933).

3.3.3. Multiscale geographically weighted regression model

The Geographically Weighted Regression (GWR) was introduced as an extension of global traditional regression models to account for spatially varying relationships between dependent and independent variables caused by neighboring effects (Brunsdon et al., 1996). It adjusts for nonstationarity in relationships by the use of a data-borrowing procedure in order to perform a series of local regressions for each area, which enables us to estimate the model's parameters at any given location in a study area. The bandwidth is a core concept in the GWR, which determines the "influencing spatial scale" for the spatial process (Fotheringham et al., 2017). A constant bandwidth is applied for all the variables in a GWR, assuming that each influential factor operates at the same spatial scale. However, in many cases, a fixed spatial scale is not valid where phenomena involve numerous spatial processes with various spatial scales. For commercial centers and their vulnerability during the COVID-19 recession, the extent of neighboring effects may operate at different scales; when modeling complex spatial processes, it is important to use a multiscale approach. The MGWR is the improved variant of the GWR that removes the constraint of using the same bandwidth and allows the relationship between dependent and independent variables to vary spatially and at different scales (Fotheringham et al., 2017). The MGWR enables the range of data-borrowing to vary across the parameter surfaces so that the scale of the independent and dependent variables will not be inconsistent across the process. In this study, a MGWR model is constructed with the vulnerability of commercial centers as the dependent variable and the nine influential factors as the independent variables.

$$y_i=\sum _{j=1}^m{\beta }_{bwj}\left(u_i,v_i\right)x_{ij}+{\epsilon }_i$$ (3)

Where at the commercial center i, ${\mathrm{y}}_{\mathrm{i}}$ is the vulnerability, $\left({\mathrm{u}}_{\mathrm{i}},{\mathrm{v}}_{\mathrm{i}}\right)$ is the geographic coordinates, ${\mathrm{x}}_{\mathrm{i}\mathrm{j}}$ is the vector of explanatory variables that may affect the vulnerability, bwj is the bandwidth with the bi-square kernel to estimate the j th variable, ${\beta }_{\mathrm{b}\mathrm{w}\mathrm{j}}$ is the coefficient of explanatory variable j, and ${\epsilon }_{\mathrm{i}}$ is a random error term (Fotheringham et al., 2017).

An (adaptive) bi-square kernel, which removes the effect of observations outside the neighborhood specified with the bandwidth and minimizes correct Akaike Information Criterion (AIC), is employed for calculating the optimal bandwidth (Forati & Ghose, 2021). Also, the vulnerability of the commercial centers is sensitive to size, and we construct separate models for the commercial center at each level.

4. The dynamic of commercial centers

4.1. The operation status of commercial centers

In this section, we compare the operation status of commercial centers in Shanghai between 2019 and 2020 from the 2nd to the 22nd week in terms of consumption volume and consumption hinterland to identify the impact of the pandemic on the commercial centers.

4.1.1. Consumption volume

The comparison between the consumption volumes of 2019 and 2020 suggests a decline of above 30%, indicating that the pandemic had a considerable negative impact on the urban commercial economy (Fig. 4 ). At the beginning of the pandemic, in the 2nd and 3rd weeks of 2020, the consumption in Shanghai was less influenced, and the decline of consumption volume was relatively meager. In the 4th and 5th weeks, the drops in consumption and transaction peak because of people's overreaction when pandemics first challenge them. Such an impact lasted about one month (from the 5th to the 9th week), and the consumption and transaction are recovering. Another decline was detected in the 14th week due to the second wave of the pandemic. The correlation coefficients between the consumption and transaction volume of the commercial centers in Shanghai and the total number of existing confirmed infected cases for weeks from 2 to 22 of 2022 were −0.793 and −0.739, respectively, showing a significant negative correlation. This indicates that with the increase in infected cases, the upgrade of governmental restrictions, and the limitation of non-essential consumption travel, the consumption volumes of commercial centers were negatively impacted by the pandemic and showed a sharp decrease.

Fig. 4.

The consumption and transaction volume of the commercial centers in Shanghai from the 2nd to the 22nd week of 2020 compared with the same period of 2019.

4.1.2. Consumption hinterland

The proportion of the consumption and transaction volumes of Unionpay cardholders from other cities decreased by 54.66% and 64.64% from the 2nd to the 22nd week of 2020 compared with the same period in 2019. The proportion of the consumption volume of Unionpay cardholders from other cities had decreased gradually since the 5th week of 2020, which was the Spring Festival holiday in China, with quite a few traveling across cities (Fig. 5 ). Due to the national governmental restrictions and the limitation of non-essential travel between cities, a significant reduction of the consumption hinterland of the commercial centers was exhibited. The proportion of the consumption volume of Unionpay cardholders of other towns has slightly rebounded since the 12th week and then fluctuated again due to the second wave of the pandemic.

Fig. 5.

The proportion of the consumption volume of Unionpay cardholder from other cities from the 2nd to the 22nd week of 2019 and 2020.

4.2. Vulnerability and short-term adaptation of commercial centers to the COVID-19

The vulnerability of commercial centers in Shanghai in weeks 7 and 14 of 2020 against COVID-19 was measured, and higher values indicate less vulnerability (Table 2 ). Different sectors of the commercial economy exhibit differentiated vulnerability. In the 7th week, the vulnerability of different sectors varied significantly, with the wholesale, catering, and life services sectors being significantly influenced, and the retail, leisure, and entertainment sectors were less vulnerable. It can be concluded that people would avoid gathering contact when during the first wave while their awareness of COVID-19 is poor. Thus retail, leisure, and entertainment activities do not decline as much as catering which is more likely to have face-to-face contact and more exposed to COVID-19. In the 14th week, with the adjustment of operation strategy, the vulnerability index of the catering, wholesale, and lift service sectors witnessed a slight increase. However, the vulnerability index of the retail, leisure and entertainment sectors decreased as they were seriously affected by the second wave of the pandemic. Although the pandemic was not as serious as the first wave, the longtime control policy has negatively impacted the economic vitality of these two sectors. Also, the leisure and entertainment sectors are usually small businesses, which adapted to the pandemic slowly. For others, like wholesale, the online shopping mode is a good option during the pandemic to avoid face-to-face contact. Even though catering, wholesale and leisure, and entertainment are still highly vulnerable to COVID-19 because the impacts of the pandemic were significant. Retail and life service targets people's necessary daily activities and are less vulnerable to the pandemic.

Table 2.

The vulnerability of the commercial centers.

Vulnerability		week 7	week14
Commercial sector	Catering	−0.57	−0.47
	Retail	−0.13	−0.17
	Wholesale	−0.74	−0.41
	Life Services	−0.49	−0.16
	Leisure and entertainment	−0.26	−0.39
Average of different levels	First-level	0.68	0.21
	Second-level	−1.84	−0.54
	Third-level	−1.84	−0.81
	Fourth-level	−1.59	0.29
Average of different circles	Inner-ring road	−0.59	0.13
	Middle-ring road	−1.78	−0.72
	Outer-ring road	−1.96	−0.60
	Suburban-ring road	−1.86	−0.41
	Outside of Suburban-ring road	−1.47	−0.11

The disparity also exists regarding the level of the commercial centers (Table 2). In week 7, the first-level commercial centers were less vulnerable because they are located in the central urban regions with many consumers in the surrounding areas (Fig. 6 ). Also, the function of the first-level commercial centers is more comprehensive, which significantly raises resilience. The lower-level commercial centers did not show a significant difference in resisting the impacts of COVID-19 at the early stage. In week 14, with the experience of living with COVID-19, people would visit large commercial centers less, which is unnecessary to their daily life. However, the fourth-level commercial centers provide the products for people's daily life. Visiting fourth-level commercial centers cannot be avoided because people would not prepare daily supplies as they did during the first wave of COVID-19.

Fig. 6.

The spatial distribution of the vulnerability of commercial centers in Shanghai

Note: high values mean less vulnerable and more resilient.

Fig. 7 presents the clustering results of vulnerability in week 7. The less vulnerable clusters are detected in the central urban regions, and the commercial centers in the ring road regions do not perform well in resisting COVID-19 in week 7. However, Fig. 6 suggests that these vulnerable commercial centers again could gain experience from the strike pandemic quickly, as they become less vulnerable in week 14. Therefore, the experience during the first wave of COVID-19 significant increase the short-term adaptation of the commercial centers and narrowed the gaps between different commercial centers. To this end, the disparities among commercial centers' vulnerability and short-term adaptation patterns are clear; however, there is still no full picture of the mechanism of commercial centers' vulnerability regarding spatial and scale heterogeneity. Thus the MGWR model analysis is expected to provide a comprehensive understanding of the mechanism of commercial centers' vulnerability.

Fig. 7.

The clustering results of the vulnerability of commercial centers in Shanghai.

5. The mechanism of uneven vulnerability and short-term adaptation of commercial centers

The MGWR models were also implemented using the data in week 7 and week 14 to explore the underlying mechanism of the disparities in vulnerability and short-term adaptation against COVID-19. A multicollinearity test of the independent variables was conducted to eliminate the cross-influence of factors. The results show that the variance inflation factors (VIFs) of the other seven independent variables besides life services and wholesale proportion were less than 1.32. The tolerance (T) of the seven independent variables ranged from 0.760 to 0.930, indicating no collinearity and cross-influence of the seven selected indexes. In the results of the MGWR model, the R² of each level commercial center ranged from 0.624 to 0.786, respectively, indicating that the MGWR model has good fitness and that the independent variables had a strong comprehensive explanatory ability.

The results of the MGWR model are represented in Table 3 . The first-level commercial centers are mainly located in the central urban regions, and the spatial heterogeneity will not be discussed in this section. For the consumption volume, in the first wave of the pandemic in week 7, its impacts on vulnerability vary with the level of commercial centers. They are mostly positive for the first- and second-level commercial centers, indicating that the high-level commercial centers with large consumption volumes are robust to resist the strike of COVID-19. However, consumption volume negatively impacts third- and fourth-level commercial centers, usually small- and community-level shopping centers. The commercial centers with low consumption are usually the food market in China supporting people's daily life and whose consumption is less sensitive to pandemics, showing stronger resistance to external shocks. Thus, Hypothesis H1 about the consumption volume is partially supported by our results. The spatial heterogeneity of consumption volume is significant for second-level commercial centers, while the coefficient decrease from central urban to outer urban regions (Fig. 8 ). In the outer regions, the second-level commercial centers are the main commercial center because the first-level ones are in the central urban regions. Thus, the consumption in the second-level commercial centers in outer urban regions would be level sensitive to the pandemic.

Table 3.

Results of the MGWR model.

Explanatory variable		Week 7				Week 14
		Absolute average value	Positive value (%)	Negative value(%)	Coefficient of variation	Absolute average value	Positive value (%)	Negative value(%)	Coefficient of variation
first-level commercial center	Consumption volume (X1)	0.697	61.9	38.1	1.102	0.399	57.2	42.8	1.18
	Consumption diversity (X2)	4.207	19	81	1.441	3.441	47.6	52.4	1.444
	Catering proportion (X3)	1.558	42.9	57.1	1.893	1.068	76.2	23.8	1.272
	Retail proportion (X4)	7.04	38.1	61.9	1.438	4.394	19.1	80.9	1.256
	Leisure and entertainment proportion (X7)	1.813	71.4	28.6	2.115	1.591	76.2	23.8	1.495
	Consumption hinterland (X8)	4.351	57.1	42.9	2.092	2.528	57.1	42.9	1.606
	Consumer attribute (X9)	2.471	52.4	47.6	2.331	2.302	28.6	71.4	1.107
second-level commercial center	Consumption volume (X1)	0.472	100	0	0.584	0.197	27.1	72.9	0.757
	Consumption diversity (X2)	0.098	100	0	0.09	1.162	0	100	0.156
	Catering proportion (X3)	0.253	0	100	0.351	3.574	0	100	0.103
	Retail proportion (X4)	0.079	100	0	0.575	1.161	100	0	0.029
	Leisure and entertainment proportion (X7)	2.46	100	0	0.427	5.778	100	0	0.232
	Consumption hinterland (X8)	0.965	0	100	0.026	2.145	25	75	0.667
	Consumer attribute (X9)	0.243	0	100	0.144	1.549	0	100	0.047
third-level commercial center	Consumption volume (X1)	0.299	0	100	0.059	0.336	0	100	0.028
	Consumption diversity (X2)	0.079	60	40	0.981	0.137	94.3	5.7	0.866
	Catering proportion (X3)	0.36	0	100	0.064	0.287	10	90	0.514
	Retail proportion (X4)	0.267	85.7	14.3	0.412	0.248	65.7	34.3	0.512
	Leisure and entertainment proportion (X7)	0.075	74.3	25.7	0.471	0.315	58.6	41.4	0.386
	Consumption hinterland (X8)	0.754	0	100	0.044	0.187	0	100	0.089
	Consumer attribute (X9)	0.494	100	0	0.031	0.139	92.9	7.1	0.416
fourth-level commercial center	Consumption volume (X1)	0.104	0	100	0.117	0.125	0	100	0.128
	Consumption diversity (X2)	0.176	100	0	0.25	0.202	100	0	0.141
	Catering proportion (X3)	0.095	100	0	0.412	0.029	51.5	48.5	0.506
	Retail proportion (X4)	0.028	74.2	25.8	1.121	0.019	34.8	65.2	0.98
	Leisure and entertainment proportion (X7)	0.195	4.5	95.5	0.322	0.117	0	100	0.076
	Consumption hinterland (X8)	0.212	0	100	0.071	0.146	0	100	0.11
	Consumer attribute (X9)	0.091	3.1	96.9	0.975	0.108	0	100	0.157

Fig. 8.

The spatial distribution of the regression coefficient of the MGWR model.

The consumption diversity positively affects the vulnerability of all the second- and fourth-level commercial centers and 60% of third-level commercial centers. These commercial centers provide diverse products and services that can improve their ability to resist the pandemic because of quicker adaptation. However, for the first-level commercial centers, the MGWR model shows that consumption diversity positively affects only 19% of first-level commercial centers' vulnerability. Thus, diversity of service can only reduce small- and medium-sized commercial centers' vulnerability. Thus, Hypothesis H1 about consumption diversity is partially supported by our results. Spatially, there is no significant core-periphery pattern as the positive impact of consumption diversity on the third-level commercial centers is mainly found in the ring road regions (Fig. 8).

We also try to shed light on the vulnerability of different commercial sectors. The catering service makes first-, second-, and third-level commercial centers more vulnerable to pandemics because people would avoid unnecessary face-to-face contact. However, the carting sector positively affects the fourth-level commercial centers with relatively low regression coefficients. It is probable that in the community-level commercial centers, the catering services would serve many fewer people than the large ones, and the out-for-delivery service helps them maintain their consumption.

The retailing sector negatively impacts more than 60% of the first-level commercial centers with high regression coefficients (Table 3), indicating that those commercial centers, with their large proportion of retail rather than essential goods, were severely hit by the pandemic. The retailing sector positively affects all the second-level commercial centers' vulnerability and most of the third- and fourth-level commercial centers. Thus, low-end retailing shows low vulnerability against the pandemic, as we suggest above that the service or the products for people's daily needs are less likely to be influenced by pandemics. In terms of spatial heterogeneity, the negative impacts of retail services' vulnerability are found in the southwestern part of the ring region (Fig. 8).

The leisure and entertainment sector shows a positive effect on all the second-level commercial centers' and most of the first- and third-level commercial centers' vulnerability, while it harms the fourth-level commercial centers. This sector involves various indoor and outdoor commercial formats, and its influences on the vulnerability of commercial centers are complicated. The leisure and entertainment sector in the second-level commercial centers are usually the large ones that can make the strategies to adapt to the pandemics quickly. However, the leisure and entertainment sector in fourth-level commercial centers are usually small businesses likely to be destroyed during a pandemic (Bartik et al., 2020).

Travel limitation in China has been restricted since the pandemic outbreak, and the consumption hinterland negatively affects the value of vulnerability of the second-, third- and fourth-level commercial centers. It is surprising to see the positive impacts on first-level commercial centers, which suggests that the travel limitation less influences high-end people's consumption, and high-end services or product shopping could be done online. Thus, Hypothesis H2 about the consumption hinterland is mostly supported by our results except for the first-level commercial centers. The proportion of high-end consumers positively affects more than half of the first-level commercial centers' vulnerability. Thus, the consumption behavior of the high-end population is less likely to change because of COVID-19, making the commercial centers targeting the high-end population less vulnerable during the pandemic. The proportion of high-end consumers of other levels of commercial centers is quite low, and the coefficients are not considered. Thus, our results do not support Hypothesis H2 about high-end consumers.

The first wave of COVID-19 provides experience to both policymakers and the managers of commercial centers, and commercial centers would be adapted to the pandemic. The temporal variance in the impact of the factors on the vulnerability of commercial centers was observed by comparing the results of weeks 7 and 14. Significant changes can be found in the consumption volume's impact on the second-level commercial centers (Table 3). The results suggest that the second-level commercial centers with smaller consumption volumes better resist the pandemic, indicating that the smaller-scale second-level commercial centers are more likely to adopt flexible measures to resist the second wave of the pandemic. For the consumption diversity, its impact on the short-term adaptation of first- (+28.6%) and third-(+24.3%) level commercial centers is higher. This indicates that in the long run, commercial centers with high consumption diversity tend to have better short-term adaptation.

Specifically, the impact of the catering sector becomes more positive on the short-term adaptation of the first-level commercial centers (+33.3%) and more negative on that of the fourth-level commercial centers (+48.5%). It still highlights the vulnerability of the small business during pandemics, while the restaurants in large commercial centers can shift from dine-in to take-out. The retailing sector has a more negative impact on the short-term adaptation of the first- (+19%), third- (+20%), and fourth-(+39.4%) level commercial centers. It can be concluded that although the retailing merchants have already optimized their operation strategy, commercial centers with a high proportion of retailing received more negative impacts by the second wave of COVID-19, showing weak and even a lack of resistance. The COVID-19 pandemic has seriously impacted residents' consumption capability, contributing to a shrinking retail market and a recession in the retailing sector in the long run. For the fourth-level commercial centers, the negative impact of the retailing sector on their short-term adaptation is more widespread.

For the consumption hinterland, compared with the results of the MGWR model of week 7, the regression coefficients of the proportion of consumers from other cities decreased in week 14, indicating that the impact of the consumption hinterland on short-term adaptation has been weakened as the pandemic became normalized. Furthermore, the coefficients of the consumer attribute for the first-level commercial centers decrease, indicating that the commercial centers targeting high-end consumers are still vulnerable to the pandemic. Thus, Hypothesis H3 is supported by our results. Synthesizing the result of week 7 and week 14, we find that the low-level commercial centers are less vulnerable at the early stage of pandemics and that the short-term adaptation of high-level commercial centers would be enhanced in the long run.

6. Conclusion and discussion

People's economic activities in commercial centers can represent urban economic vitality, and commercial vulnerability and adaptation to the pandemic are critical to sustainable urban development. In this study, we explore the commercial dynamics in urban China during the prevalence of COVID-19 in Shanghai. The research outcomes shed light on the heterogeneity of commercial centers' vulnerability and short-term adaptation during the COVID-19 recession and the underlying mechanism.

Regarding vulnerability, we find that the high-level commercial centers in the central urban regions are less vulnerable to the pandemic. More specifically, the wholesale and catering sectors are vulnerable; however, they adapted to the pandemic quickly, which might be because of the online and outer-for-delivery service. Leisure and entertainment sectors were less influenced when the COVID-19 breakout; however, they became vulnerable during the pandemic's second wave because of people's fear of face-to-face contact. Retail and life service are the most robust sectors, and the pandemic less influences them during our study period.

The MGWR further reveals the underlying mechanism regarding spatial and scale heterogeneity, focusing on our three research hypotheses. The findings suggest that high-level commercial centers with large consumption volumes can fast adapt to the prevalence of COVID-19. Diversity is critical for small and medium commercial centers to reduce their vulnerability during the COVID-19 recession. Furthermore, the commercial centers targeting high-end consumption and wide hinterland are less vulnerable. The COVID-19 recession has limited impacts on high-end consumers, and distant consumers could shop online to tackle the travel restriction (Consolazio et al., 2021; Li & Stoler, 2022). Finally, we find that the impacts of the variables on commercial centers' vulnerability would vary in different waves of the pandemic, suggesting that the commercial centers would adapt to the COVID-19 recession. However, our research cannot provide insight into how commercial centers adapt to the pandemic.

6.1. Policy implications

This research is an early attempt to provide an overview of the vulnerability and short-term adaptation of commercial centers in a global city during the COVID-19 recession. Several practical insights can be drawn from this study.

First, according to our findings, small commercial centers are vulnerable during the pandemic. Some of them provide essential support for the communities, and their death might intensify social justice in some communities in outer urban regions. Although diversity strengthens their ability to resist the COVID-19 recession, small commercial centers usually have single products. Thus the function of the small community-level commercial centers should be guaranteed during the pandemic, which would influence the resilience of local communities.

Second, the sectors, such as wholesale, that could provide online or out-for-delivery service are less vulnerable and better adapted. Therefore, an online-to-offline (O2O) delivery platform is important for strengthening the resilience of the commercial economy. Currently, China has the most prosperous e-commercial service industry (Zhao et al., 2021), and its positive impact on sustainable urban development has been found in this research. However, there are still some challenging from the Chinese experience, such as the broken logistics network (Liu et al., 2022), which is the foundation of e-commercial service.

6.2. Limitations and future research

Although this paper contributes to the literature on commercial dynamics during the COVID-19 recession, it is not free from limitations, which open avenues for further research. First, our study cannot paint a full picture of commercial resilience against COVID-19 because our data is limited to the 2nd to the 22 nd week in 2020. Our study is limited to the vulnerability and short-term adaptation of commercial centers under the shock of COVID-19. How the urban economy recovers after COVID-19 is expected in future research. In addition, our research did not provide insight into the detailed strategies of the commercial centers and the impacts of the regional control policies. China has experienced the 'Dynamic COVID zero’ policy and now trying the "live with the virus" policy, which would contribute to a different meta of the urban economy. It is a good opportunity to explore how commercial activities adapt to different policy scenarios.

Author statement

L.Z., and W.X. conceived and designed the study. L.Z., Z.Z. and H.Z. collected data and conducted analyses; L.Z., and W.X. drafted the paper and proofreading.

Funding

This work was supported by the National Natural Science Foundation of China (grant number: No. 42071212, No. 42201231, No. 42201455, No. 41701185), the Postdoctoral Science Foundation of China (grant number: No. 2019M651776, No. 2022M713234).