Structural Changes in Human Mobility Under the Zero-COVID Strategy in China

Abstract

Over the past two years, China has wrested domestic control of the COVID-19 pandemic through non-pharmaceutical interventions. However, the extent to which the pandemic has changed people’s travel behavior in the new normal under the zero-COVID policy is not yet clear. This study investigates the profound effects of the zero-COVID strategy on human mobility in 365 Chinese cities over time. Our results suggest the following: (1) Even between city pairs with no local cases, intercity mobility decreased by an average of 16%, whereas intra-city mobility increased by 9% compared with the pre-pandemic baseline. Long-distance travel decreased substantially, while trips below 100 km increased slightly. These new trends indicate a localized pattern which is presumably caused by the wide adoption of teleworking and virtual classes, as well as concerns about the safety and availability of public transportation. (2) Containment measures caused a considerably acute decline in intercity short-distance trips but exerted a markedly lasting effect on long-distance trips. (3) Cities with lower levels of urbanization, smaller population sizes, less labor force, and lower GDP and GDP per capita experienced greater reductions in mobility, which may be due to their insufficient management capabilities. (4) The Chinese government has adopted adaptive measures to contain outbreaks. Stricter and more timely responses led to faster recoveries in the second half of 2021 compared with the first half. Inter- and intra-city mobility decreased by 41% and 26%, respectively, within the first 2 weeks of an outbreak, and it took 6-7 weeks to return to pre-outbreak levels.

Keywords

COVID-19 human mobility Zero-COVID Chinese city

Introduction

The novel coronavirus disease (COVID-19) has caused millions of deaths globally and profoundly affected many aspects of society. Since the initial outbreak, China has committed to a zero-COVID strategy, using multiple strict measures, such as temporary lockdowns, mass testing, contact tracking, isolating close contacts, and closed-off management, to keep the pandemic within a limited spatial and temporal scale (see Supplementary Materials for more information). A zero-COVID strategy was also implemented earlier in several countries, such as South Korea, Singapore, Vietnam, and New Zealand in 2020–2021 (Llupià et al., 2020). However, it was gradually abandoned as the virus evolved. China was one of the few countries that had continued to implement a zero-COVID strategy for over 2 years of the pandemic (the policy was phased out in December 2022).

The COVID-19 pandemic has spurred various related studies in different disciplines. Through a systematic review of related papers published from January 2020 to December 2021, scholars classified the COVID-19 geographical research into four main themes: the spread of the pandemic, social management, public behavior, and the effects of the pandemic (Xi et al., 2022). These themes were not independent of each other. Many studies covered more than one theme. Several theoretical models were widely used to predict epidemic evolution and evaluate the effectiveness of interventions (Chang et al., 2021; Lai et al., 2020). One of the most well-known models is the SIR model, which stands for Susceptible, Infected, and Recovered. This model is commonly used to simulate the spread of infectious diseases and analyze how different factors, such as human mobility and social distancing policies, can affect disease transmission. Other important variant models, such as the SEIR model, which stands for Susceptible, Exposed, Infected, and Recovered have also been used. The SEIR model is similar to the SIR model but includes an additional exposed stage, representing individuals who have been infected but are not yet contagious. In addition to these models, various regression models can be used to analyze the relationship between socio-economic status and health outcomes. For example, negative binomial and multivariate regression models were used to analyze the correlation between various factors and COVID-19 transmission (Mu et al., 2020; Niu et al., 2020).

Human mobility and COVID-19 have complex interplay in their relationships. Mobility reduction during the COVID-19 outbreak has been observed worldwide (McKenzie and Adams, 2020). For example, non-pharmaceutical interventions in the early stage of the COVID-19 outbreak, including suspending inter-city public traffic, closing schools, scenic attractions, and non-essential services, canceling mass events, stay-at-home order, and encouraging working from home, reduced approximately 70% inter-city mobility and 40% intra-city mobility in China (Mu et al., 2020). Early lockdowns in Germany (Schlosser et al., 2020), France (Pepe et al., 2020), and the US (Xiong et al., 2020) caused a decrease of 35%–50% in mobility and took about 3 months to recover. The reduction of mobility reduced human-to-human transmission and has been proven effective in mitigating the spread of the disease, delaying the epidemic progression, and lowering the infection peak (Chinazzi et al., 2020; Kraemer et al., 2020; Lai et al., 2020; Tian et al., 2020).

Mobility change is considered a proxy for the effects of containment measures. Different regions, population groups, and social-distancing policies can all affect the degree of mobility change that occurs. For example, a city in the border region will likely take stricter measures because it faces higher importation risks and thus has a more significant reduction in mobility. Certain population groups, such as people in high-income neighborhoods, could reduce their mobility more than those in low-income neighborhoods because high-income groups were more likely to be allowed to work from home (Jay et al., 2020). People’s support for public health policies also plays a role in determining the level of mobility change. For example, higher levels of national identification made people more willing to follow the guidelines and thus predicted lower mobility during the early stage of the pandemic (Van Bavel et al., 2022).

COVID-19 has changed mobility patterns in many aspects. Scholars found evidence that Zurich’s micro-mobility connections (e.g., short-distance ride trips) remained similar during the lockdown period. Still, the number of trips along connections reduced significantly (Li et al., 2021a). Additionally, trips in different distance ranges did not change proportionally. Short-term lockdowns could cause fewer people to travel long distances during the pandemic (Galeazzi et al., 2021; Pullano et al., 2020; Schlosser et al., 2020).

China has been on a steady path to socio-economic recovery during 2020–2021 under the zero-COVID strategy. The GDP expanded by 2.3% and 8.1% year-on-year in 2020 and 2021, respectively, making China a leading major economy during the pandemic (International Monetary Fund, 2022). However, the Chinese government’s aggressive containment measures also caused many concerns, for example, economic disruption, financial instability, decrease in productivity, and negative social and psychological effects. Hence, which strategy, mitigation or elimination, will have heavier effects on the cites is debatable. Most countries employed a mitigation strategy, while China implemented an elimination strategy.

On the one hand, a few studies pointed out that the duration of a lockdown matters more than its strictness (Guan et al., 2020). Cities have generally been robust to short-term shocks (Glaeser, 2021) but may be negatively affected by long-term restrictions (Chen et al., 2022; Fezzi and Fanghella, 2021). Scholars argued that disruptions caused by the pandemic would have a longer-term effect on mobility than the duration of the emergency (Christidis et al., 2021). For example, a significant increase has been found in people’s expectations for teleworking since the pandemic (Barrero et al., 2021). A recent US survey showed that only 8% of Manhattan office workers were in the workplace full time as of late April 2022 (Goldberg, 2022). Over 90% of the job hunters in China expected to work remotely at least one day per week, and 50% expected at least 3 days per week (Zhilian Zhaopin, 2022).

On the other hand, evidence suggested the virus itself, other than restrictions, should be responsible for the majority of the economic damage (Sheridan et al., 2020). People’s behavior was more influenced by the fear of infection and death than social distancing policies (Alexander and Karger, 2021; Goolsbee and Syverson, 2021). A longitudinal analysis across 15 countries indicated that the average containment policies in countries pursuing elimination strategies were less stringent than those pursuing mitigation strategies because of the lower pandemic intensity (Aknin et al., 2022).

Knowledge about mobility changes during the pandemic has been accumulating rapidly. However, most studies focused on the short-term effects of intervention measures in the early stage of the pandemic. The observation often lasted for several weeks or months only. Whether new trends will emerge in human mobility in the long recovery period and the extent to which the zero-COVID strategy will change people’s mobility behaviors in the pandemic’s new normal have yet to be revealed.

In China, over 75% of the cities did not experience any local outbreak in 2021. Life in most Chinese cities was COVID-19-free, entering an era of the long recovery and the pandemic’s new normal. Hence, the main questions in this study include the following: (1) How have people adapted to new mobility patterns in the pandemic’s new normal? (2) What types of mobility and cities are most affected by containment measures? and (3) How does mobility change with policy variations and virus evolution in the context of China? Our empirical analysis uses a nationwide dataset provided by a large location-based service provider in China, along with epidemic, demographic, and socio-economic data collected from various sources. We quantify the magnitude of effects of counter-COVID measures on human mobility in 365 Chinese cities under different policy scenarios over time by employing difference-in-difference (DID) and event study methods. Our results shed light on the profound effects of the zero-COVID strategy on human travel behaviors, providing valuable insights for policymakers worldwide to evaluate responses to current and future challenges.

Data and methods

Study units

This study covers 365 administrative units in mainland China, including 4 centrally administered cities (zhixiashi), 332 prefecture-level units (diji danyuan), and 29 county-level units (xianji danyuan) which are directly under the administration of provinces (see Figure S1 in the Supplementary Materials). These units are independent of each other. We refer to these units as “cities” in this study. A detailed introduction to the administrative units in China can be found in Ma (2005).

Data

Human mobility data. Mobility data of Chinese cities were released by Baidu Mobility (https://qianxi.baidu.com/), which tracked daily trips in and out of cities and intra-city mobility (see Supplementary Materials for more information). Baidu is one of China’s largest providers of location-based service (LBS). Baidu data defined a valid intercity trip as a single mobile device transferring from departure city $i$ to destination city $j$ and staying in destination city $j$ for at least 4 hours. The departure city was typically identified as a place of usual residence or overnight stay (Niu et al., 2020). We should note that one person might take several intercity trips within a day under such definitions if they stayed at each destination for more than 4 hours. Different definitions of the duration may produce a different number of trips. Moreover, Baidu data were based on anonymized, aggregated data, making it difficult to identify and analyze individual behavior and patterns. Additionally, Baidu data could be biased in favor of users who are more likely to use location-based services, such as younger people. The uneven use of mobile Internet between regions could also exaggerate the number of trips in developed areas (e.g., people used more than one smartphone) or underestimate the number of trips in undeveloped areas (e.g., people did not have access to the Internet). In this study, all mobility data were aggregated at the administrative city level. A comprehensive dataset for 365 cities over 365 days in 2021 served as the basis for much of our empirical analyses. Data in 2020 were also included for comparison (dates from May 7 to 21 September 2020 were excluded because of availability).

Socio-economic and epidemic data. In this study, socio-economic data covered five variables for each city: urbanization level, population, population in the labor force, GDP, and GDP per capita. Data were collected and compiled from (1) Tabulations on the 2020 China population census by county (National Bureau of Statistics of China, 2022) and data for prefecture-level and centrally administered units were aggregated from county-level units, (2) The 2020 statistical bulletin on the national economic and social development of provinces and cities, such as Shanghai (Shanghai Municipal Bureau of Statistics, 2021), and (3) Statistical yearbooks of provinces and cities, such as Guangdong province (Guangdong Provincial Bureau of Statistics, 2021). Urbanization level was calculated by dividing the number of people living in the urban area by the total population. The population in the labor force was derived from the population by industry category in the long table census data. It was adjusted by the sampling ratio accordingly (see Supplementary Materials for more information). GDP per capita was calculated by dividing the total GDP of a city by its population in 2020. We classified cities into high (H) and low (L) groups using each variable. For example, if a city’s urbanization level was above average, it was classified as the H group (see Figure S2 for the distribution of cities in H and L groups). The daily epidemic data were obtained from the National Health Commission of China (http://www.nhc.gov.cn) and the Health Commission of each province.

Methods

DID model. We used DID models (Xu, 2021) to evaluate the effects of counter-COVID-19 measures on intercity mobility in the pre- and post-treatment periods across cities

Y_{i t} = {T r e a t}_{i t} \times α + ρ_{i} + μ_{t} + ε_{i t}

(1)

where

Y_{i t}

represents mobility level in city

i

on date

t

. We used the percentage change of mobility volume relative to the benchmark mobility.

{T r e a t}_{i t}

denotes the treatment status of city

i

on date

t

. Given the carryover effect from treatment to future outcomes, treatment status of city

i

takes a value 1 if there is any local case in the previous 14 days and 0 otherwise (see Figures S3–S4). Coefficient

α

estimates the difference in intercity mobility between the treated and control cities before and after treatment. A negative and significant

α

suggests that containment measures reduce mobility level. The regression model also includes city-fixed effects

ρ_{i}

and date-fixed effects

μ_{t}

to control for the time-invariant attributes of cities (e.g., geographical location, short-term economic development, and population size) and common time trends (e.g., national holidays and seasonal fluctuations that affect all cities similarly).

ε_{i t}

is the error term. Standard errors are clustered at the city level.

Dynamic DID (event study). We used the dynamic DID models and included leads and lags of treatment in the regressions to further investigate how mobility trends evolved

Y_{i t} = \sum_{k = T_{0}, k \neq - 1}^{T 1} {T r e a t}_{i k} {\times α}_{k} + ρ_{i} + μ_{t} + ε_{i t}

(2)

where

{T r e a t}_{i k}

is a dummy variable equal to 1 if the observation’s period relative to the first treated period is the same value as k and 0 otherwise (also 0 for control groups). In this regression, treatment denotes the resurgence of local cases in city

i

, often after over 4 weeks of zero cases. Most cities experienced no more than one outbreak throughout 2021. Even in cities that experienced multiple waves, sufficient time intervals between successive treatments were observed (See Figure S3). If a city experienced more than one outbreak within 2 months, then we only considered the first period a city receives the treatment. If there were no more than two local cases in a city for 14 consecutive days (which means no local transmission and low risk, see Supplementary Materials for more information), this city was taken as the control city. We also included December 2020 in the analysis to better reflect the leads and lags of the outbreak in January 2021 (See Figure S5). A visualization of outbreaks and timeline in the dynamic DID model was also presented in Figure S5. Coefficient

α_{k}

estimates the effect of containment measures in the k^th week before or after local cases appear. In this case, we used week instead of day to avoid the interference of fluctuations in daily mobility within the week. The dummy

k = - 1

was omitted in the regression. Thus, the effects of counter-virus measures in different periods are relative to the week before the report date of local cases.

T_{0}

and

T_{1}

are the lowest and highest number of leads and lags, respectively, during the treatment period. In our analysis, at least 5 weeks before the local case report (

k \leq - 5)

was regarded as one endpoint. In addition, at least 8 weeks after the local case report (

k \geq 8)

was regarded as another endpoint. Thus, endpoint estimates would be less precise and contain greater variance.

Density and intensity of intercity mobility. We calculated the density and intensity change for different distance ranges to identify the structural change of the mobility network. If mobility flows existed from city a to city b, we counted a directed edge between a and b. The density of intercity flows was defined as the ratio of extant edges to potential edges between cities. It indicated potential traffic connections between cities. The intensity of intercity flows was defined as the average volume of extant intercity trips. Given that data were aggregated at the city level, we used air distance (i.e., great circle distance) between city centroids as a proxy of trip distance. Density values range from 0 to 1, that is, no connection to fully connected within a certain distance scale.

Results

Localized mobility pattern

Mobility change in the renewed waves. Five major epidemic waves occurred in 2021 (Figure 1a. A summary of the five renewed waves is listed in Table S1. On 18 March 2020, China reported no new local cases for the first time since the initial COVID-19 outbreak in January 2020. After that, only sporadic local cases appeared in a few cities. A strong correspondence was observed between inter and intra-city mobility reductions and city/region-specific outbreaks (Figure 1b). The most remarkable reduction in mobility happened during the first wave before the Chinese New Year. The Chinese government announced a “stay in place” policy, encouraging people to stay in their working city to reduce the risk of virus spread. Widespread preventative measures were adopted in nearly all provinces in January 2021 (Zha et al., 2022). Most outbreaks in 2021 were well-controlled. The average duration of outbreaks was 16 days (95% CI: 14–19) and the average size was 92 cases (95% CI: 52–122). Stringent regulations were implemented to reduce importation risks, such as limiting international flight frequency and seat capacity and mandatory quarantine for all international entrants. Importation risk was limited when local transmission was under control (Han et al., 2021). We chose 1 week (i.e., January 4–10) in 2020, unaffected by COVID-19 or any public holiday event, as the reference period. Compared to benchmark mobility, substantially reduced intercity and slightly increased intra-city mobility were observed in most cities during the 2021 renewed waves (Figure 1b).

Figure 1.

COVID-19 cases and mobility change during the major renewed waves. (a) Number of daily new local and imported cases from January 17, 2020, to December 31, 2021. Colors represent different regions that reported local cases. (b) Distribution of percentage changes in intercity outflow and intra-city mobility of cities in the major outbreaks in 2021 compared with the benchmark mobility.

Decreased intercity mobility. Prompt and efficient containment measures were adopted in a city once a local case was detected. We divided cities into two types: City I, which refers to cities with local cases in the past 14 days, and City II, which refers to cities without local cases. Our analysis shows intercity mobility from/to City I became significantly reduced after a local case was reported (Figure 2a and Figure S6). On average, the daily volume of intercity mobility in City I reduced by 70% compared with the benchmark level.

Figure 2.

Temporal evolution of percentage change in (a) inter- and (b) intra-city mobility compared with the benchmark. 7DMA refers to 7-day moving averages. City I and City II refer to cities with or without local cases, respectively, in the past 14 days. Ribbon in (a) represents 95% confidence intervals. Vertical color bars indicate different public holidays in 2021. Purple: New Year’s Day (NYD); Yellow: Chinese New Year (CNY); Pink: Qing Ming Festival (QMF); Blue: May Day (MD); Grey: Dragon Boat Festival (DBF), Orange: Mid-autumn Festival (MAF); Green: National Day (ND). Width of bars indicates holiday length.

Similarly, City I–City II and City II–City I mobility were reduced by 57% and 48%, respectively. Even though travel bans were lifted in most cities with no local cases, intercity mobility volume did not fully recover to pre-pandemic levels. On average, the mobility volume in City II was 16% lower than the benchmark level. Mobility fluctuations often appeared around public holidays and school summer holidays (around July) in City II, but not in City I, indicating strong control measures around key dates to prevent the disease from spreading.

Increased intra-city mobility. Intra-city mobility rebounded more rapidly compared to intercity mobility. Cities that reported new local cases also saw immediate reductions in intra-city mobility. However, the decline in intra-city mobility was not as sharp as in intercity mobility. On average, intra-city mobility in City I decreased by about 11%, while intra-city mobility in City II increased by 9% (Figure 2b). Daily mobility can be divided into work-related, household-related, and free-time types (Vilhelmson, 1999). Trips associated with these three types of activities account for about 40%, 10%, and 50% of the total daily distance traveled, respectively (Vilhelmson, 1999). Scholars have suggested that teleworking could alleviate travel time and space constraints and encourage more household-related and free-time activities (Elldér, 2020; He and Hu, 2015; Zhu, 2012). Therefore, changes in work or school schedules, such as working from home or attending virtual classes, may have allowed for more flexible travel within the city. Furthermore, changes in public transportation availability may make it easier to travel within the city than to other cities. Additionally, the desire to engage in outdoor activities to cope with the stress of the pandemic may also contribute to increased intra-city mobility.

Decreased intercity mobility and increased intra-city mobility indicate a considerably localized mobility pattern. Moreover, many other countries experienced more severe outbreaks in 2021 because of more contagious variants. As a result, the average number of daily imported cases in China increased from 14 (95%CI: 12–16) in 2020 to 19 (95%CI: 18–20) in 2021. Chinese cities experienced more renewed waves in 2021 than in 2020 (Figure 1a). Despite this, inter- and intra-city mobility increased slightly in 2021 relative to 2020 (Figure S6 and Figure 2b), reflecting a gradual and steady recovery.

Structural change in mobility distance

Cutoff point of 100 km. Connections between city-pairs become sparser and weaker as distance increases (Figures 3a and b). Many containment measures were aimed at preventing the spread of disease on a large scale, leading to restrictions on long-distance travel, particularly inter-provincial travel. These restrictions resulted in a decrease in the average distance that people traveled. The average distance remained 20% lower than the benchmark level in 2021 (Figure 3c), indicating a fundamental change in mobility patterns, rather than a short-term response to the pandemic. In 2021, the density of short-distance trips below 250 km remained similar relative to pre-pandemic levels (Figure S7(a)). However, direct traffic between many distant cities was cut off. A substantial decrease in the density and intensity of long-distance trips (over 500 km) compared to pre-pandemic levels was also observed (Figures 3b and S7(b)). However, the intensity of trips under 100 km was above the pre-pandemic level in late 2021 (Figures 3b and S7(b)). Distribution of these intensified short-distance flows often centered around major cities, such as provincial capitals in southern China: Guangdong and Hainan provinces (Figure S8). These observations suggest two possibilities. First, people may have taken short-distance weekend trips due to the lack of other options. For example, Hainan Island is naturally isolated from other provinces, making it difficult for people living there to travel to other provinces during the pandemic. As a result, short-distance trips within Hainan Island increased significantly on weekends (Figure 3d).

Figure 3.

Mobility changes relative to travel distance. (a) Density and (b) intensity of daily intercity flow during different periods. Each smoothed line represents a week in Jan 2020, January 2021, June 2021, and December 2021. (c) Temporal evolution of average distance of intercity trips. 7DMA refers to 7-day moving averages. (d) Intensity changes of intercity flows below 100km in Hainan and Guangdong provinces before the pandemic and in the new normal. (e) Effects of containment measures on intercity mobility in different distance ranges. Vertical line indicates the time when the first local case is reported (i.e., treatment). Dummy variable indicating one week before the local case report is omitted from the regression; Thus, difference in mobility one week before the treatment is set to be zero. Error bars indicate 95% confidence intervals.

Second, teleworking became a prevalent option in China during the pandemic (Ono and Mori, 2021), which may have caused out-migration from large cities. Evidence from Oslo and London suggested that many people with teleworking-friendly jobs chose to relocate to neighboring communities near the large city with more affordable housing prices or larger homes, or to live closer to their families (Shepherd et al., 2021; Tønnessen, 2021). Likely, they will not return if teleworking remains an option (Tønnessen, 2021). Note that a 100 km radius may be an effective cutoff for commuting, as evidence from France showed that 95% of work-related trips were within this distance (Pullano et al., 2020). It is highly possible that large cities in Guangdong province experienced similar outflows during the pandemic. As a result, a significant increase in intercity flows below 100 km could be observed in Guangdong province on weekends and weekdays, particularly on Mondays (Figure 3d). This increase was not observed in January 2020 before people became aware of the pandemic. In summary, the increase in short-distance trips below 100 km may be an alternative to medium- or long-distance trips under general prevention measures and a consequence of remote work in the new normal of the pandemic.

Mobility reduction across distance scales. We used the dynamic DID models to investigate how the effects of containment measures vary across distance scales over time (see Methods). D1–D4 ranges represent different distance ranges: D1: 0–250 km, D2: 250–500 km, D3: 500–1000 km, and D4: ≥1000 km. Interventions have varying effects on distance groups (Figure 3e). Detailed numbers of coefficients are listed in Table S2. The results show that long-distance trips were restored more slowly than short-distance trips. For example, the estimated coefficients rebound notably after 2–3 weeks for D1, but 3–4 weeks for D2-D4. It takes about 6–7 weeks for D1 to return to the pre-outbreak level, 7–8 weeks for D2 and D3, and more than 8 weeks for D4. Furthermore, the results show a greater short-term reduction in short-distance trips after reporting local cases. For example, the largest mobility changes in four distance ranges after treatment are −44%, −34%, −30%, and −27% relative to 1 week before treatment. Therefore, containment measures have a more acute effect on short-distance trips in the short term but exert a more lasting effect on the long-distance trips in the long term.

Heterogeneous effects across cities and over time

Mobility reduction and socio-economic disparity. Variations of mobility reduction across cities are associated with socio-economic disparities. Cities were classified into “high (H)” or “low (L)” groups according to five variables. Cities with outbreaks are listed in Table S3, and the summary statistics are listed in Table S4. Most outbreaks in 2021 occurred in cities in the H group, which is consistent with the human-to-human transmission dynamics of COVID-19. Contact possibilities are higher between individuals in more urbanized, populated, and prosperous cities. In particular, cities with high urbanization level have more local cases (Wilcoxon test, p values <0.05). However, inter- and intra-city mobility reduction tends to be stronger for cities with lower urbanization levels, less population, less labor force, lower GDP, and lower GDP per capita (Figure 4). Detailed numbers of coefficients are listed in Table S5. The recovery in the two groups of cities follows similar patterns (Figure S10). This result reflects stricter restrictions in the L group and could be interpreted by various management abilities across cities. Capable local leadership is crucial for controlling the pandemic (Li et al., 2021b). Cities in the L group are relatively vulnerable from the economic, social, and clinical aspects. Technology support, medical resources, human resources, and the capability for massive testing in the L group are not comparable with those in the H group. Consequently, “one-size-fits-all” plans may be adopted for the entire city or a large area. Additionally, cities in the L group may rely more on tourism and other industries that are heavily affected by restrictions on travel and gatherings. Therefore, the effects of the zero-COVID policy may be more apparent in the L group than in the H group.

Figure 4.

Heterogenous effects of containment measures on intercity outflows and intra-city mobility in different groups of cities. Regressions (Equation 1) are used separately for each variable in each group of cities. Estimated coefficient shows the change in inter or intra-city mobility (see methods). Error bars indicate 95% confidence intervals.

Spatial heterogeneity of policy effects. The outbreaks of COVID-19 in different cities may lead to the implementation of varying policies and, as a result, different changes in mobility patterns. Cities are affected by the virus differently, depending on factors such as their risk for disease spread. For instance, some border cities in the southwestern and northeastern regions have implemented strict policies to reduce the risk of importation and prevent local spread. Dehong, a border city in Yunnan province, saw a decrease in outward mobility of at least 50% on 242 out of 365 days in 2021 compared to the benchmark level, and intra-city mobility was below the benchmark level on 253 out of 365 days. Socioeconomic disparities may also explain some differences in mobility patterns between cities. Even though cities in the middle, southwestern, or northwestern regions often experienced milder outbreaks than those in the eastern region, their mobility reductions were often more substantial and lasted longer. This may be because cities in less developed areas have fewer resources, such as technical support and medical help, and cannot implement flexible and tailored intervention measures. Additionally, some cities may be more dependent on industries such as tourism, which can be heavily impacted by travel restrictions. As a result, the effect of such policies on human mobility may vary from city to city.

Speedy recovery under adaptive policies. The zero-COVID strategy requires sustained effort and adaptive measures to contain outbreaks. Previous analysis has revealed that general mobility in 2021 has already increased compared with 2020, indicating a decrease in the negative effect on people’s travels and effort to balance disease control and socio-economic development. To further investigate the dynamic effects of phased policies in 2021, we compared the inter- and intra-city mobility between the treatment and control groups of cities before and after the local cases report in the first and second half year of 2021 (Periods One and Two), respectively (see Methods). To better reflect mobility changes before the outbreak in January 2021, we also included December 2020 in Period One. Results are shown in Figure 5 and Table S6. If the estimated coefficients are statistically insignificant (p ≥ 0.05), then no systematic difference can be observed between the treatment and control groups of cities as compared with their pre-outbreak week. In Periods One and Two, the estimated coefficients are statistically insignificant (p ≥ 0.05) when k ≤ −2. Coefficients decrease immediately after local cases are detected (k ≥ 0), indicating prompt mobility restrictions after treatment. Intercity mobility does not return to the pre-outbreak level until at least 8 weeks later in Period One, but only 6–7 weeks in Period Two. The largest decreases in intercity outflow in Periods One and Two are −33% and −41%, respectively, suggesting stricter control in Period Two. This result is consistent with the finding that the average strictness of containment policies in the second half of 2021 was slightly higher than in the first half (Zha et al., 2022). More stringent and timely interventions often bring more rapid recovery. Intercity mobility does not start to rebound until 6-7 weeks later in Period One, but only 2–3 weeks later in Period Two. Similar conclusions can be observed in intra-city mobility. Moreover, intercity mobility is more severely affected by containment measures than intracity mobility. For example, inter- and intra-mobility could decrease by 41% and 26%, respectively, within 2 weeks after treatment. These findings show that the Chinese government has been improving epidemic prevention strategies. Even though the more infectious Delta variant caused more outbreaks in the second half year, Chinese cities have managed to detect the infection early, improve the precision of geographic control, shorten the inference period, and minimize the negative effects of restrictions on people’s daily lives.

Figure 5.

Effects of containment measures on intercity outflows and intra-city mobility in different periods. Vertical line indicates the time when the first local case is reported (i.e., treatment). Dummy variable indicating one week before the local case report is omitted from the regression. Thus, k = -1 is set to zero (see Methods). Error bars indicate 95% confidence intervals.

Conclusion and discussion

This study uses a nationwide mobile dataset covering 365 Chinese cities to analyze the policy-induced short-term and medium-term disruptions to human mobility in the new normal of the pandemic. We quantify the magnitude of the effects of counter-COVID measures on different types of mobility and determine the evolution of effects under different policy scenarios by employing the dynamic DID models. Our findings suggest a profound effect of the zero-COVID strategy on inter- and intra-city mobility among Chinese cities. First, we found that people have adapted to a more localized mobility pattern in the new normal. In China, only sporadic local cases and few deaths occurred in 2021, and thus, the risk of infection was relatively low. However, even between city pairs with no local cases, mobility volume remained 16% lower on average than the pre-pandemic level. In contrast, intra-city mobility increased by 9% as compensation. This may be due to a combination of factors such as changes in work and school schedules, concerns about the safety and availability of public transportation, leading people to have more short-distance (intra-city) trips in the new normal. The localized mobility pattern also includes a structural change in people’s travel distances. Containment measures resulted in a considerably acute decrease in intercity short-distance trips but a lasting effect on long-distance trips. Since the initial outbreak, average travel distance has not returned to the baseline, indicating a fundamental change in mobility patterns rather than a mere short-term response. Surprisingly, we observed that trips below 100 km nearly returned to the pre-pandemic level in the fourth quarter of 2021 and showed greater intensity. Two reasons may explain this situation. First, leisure travel to a neighboring city during weekends has become popular because of the restrictions and risks involved in inter-provincial trips. Second, the widespread adoption of teleworking has made people reconsider their residence places. People can live within a certain distance of large cities and maintain relatively frequent visits to their workplaces.

Second, reductions in mobility across cities are associated with socio-economic disparities and spatial heterogeneity. Larger reductions are measured in cities with lower urbanization levels, smaller population sizes, smaller labor force, lower GDP, and lower GDP per capita. However, the severity of the epidemic in these cities is often milder than in the high group. The mobility restrictions are determined by the severity of the outbreak and the local government’s ability to manage it. More drastic measures could be implemented in smaller cities with poorer economies under the zero-COVID strategy because of the lack of sufficient technical support, healthcare resources, human resources, and massive testing capabilities. Some border cities and cities in undeveloped regions face greater challenges controlling the outbreak and ensuring people’s wellbeing. Our results highlight the necessity of focusing on severely affected cities using the containment measures. Long-term attention and support are needed to mitigate the harm to socio-economic development and help vulnerable cities recover.

Third, implementing a zero-COVID strategy requires long-term effort and adaptive measures to contain the spread of the virus The Chinese government has been improving measures in its ongoing battle against COVID-19 and its variants. In general, cities adopted more timely, stringent, and tailored intervention measures in the second half of 2021 compared with the first half and recovered from outbreaks faster. We found that inter- and intra-city mobility could decrease by 41% and 26%, respectively, within 2 weeks of an outbreak and take 6–7 weeks to return to pre-outbreak levels. These levels of mobility reduction, along with other social distancing measures, were effective in containing the spread of the virus. The average duration from the first local case to clearance was 15 days, but the recovery process was about three times longer. These findings provide valuable insights that governments can consider when designing their policies for future challenges.

The emergence and global spread of new COVID-19 variants have made managing the pandemic more challenging. This study focuses on China’s situation in 2021, when outbreaks were primarily caused by the original and Delta variants, but not the more transmissible Omicron variant, which has been transmitted locally in China since January 2022. Despite the Chinese government’s efforts to adjust and improve its prevention measures, it has become increasingly difficult to find a balanced approach to managing the pandemic as the virus becomes more transmissible. The zero-COVID policy can disrupt the economy, cause financial instability, reduce productivity and have negative social and psychological effects if outbreaks become more frequent. Recently, China introduced a series of new measures to deal with COVID-19, signaling its transition from the zero-COVID policy. It remains to be seen how human mobility will change as the country continues to reopen.

This research contributes to the literature by addressing the gap in studies on human mobility under the zero-COVID strategy during the long recovery period. Our findings can help researchers and policymakers worldwide understand the complex effects of epidemic control strategies and provide valuable insights for current and future policy designs. Additionally, the dataset used in this study offers a unique and detailed perspective on human mobility, with nationwide coverage and the ability to provide insights over a long period. We also apply interpretable analytic approaches, such as the DID models and event study, to address our research questions. An innovation of this study is that we carefully define the treatment status in the models as the appearance of local cases during a given period and visualize the timeline (e.g., see Figure S5). Many studies do not differentiate between local and imported cases in their analysis. Finally, this study provides a valuable perspective for future researchers to explore the differences in population behavioral changes between countries pursuing elimination and mitigation strategies.

Supplemental Material

Supplemental Material - Structural Changes in Human Mobility Under the Zero-COVID Strategy in China

Supplemental Material for Structural Changes in Human Mobility Under the Zero-COVID Strategy in China by Xiaoyan Mu, Xiaohu Zhang, Anthony Gar-On Yeh, Yang Yu, and Jiejing Wang in Environment and Planning B: Urban Analytics and City Science

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This paper was funded by the Chan To-Haan Endowed Professorship Fund of the University of Hong Kong, Joint Programming Initiative Urban Europe and National Natural Foundation of China (Grant No. 71961137003), Guangdong–Hong Kong-Macau Joint Laboratory Program of the 2020 Guangdong New Innovative Strategic Research Fund (Grant No. 2020B121203000), the National Key R&D Program of China (Grant No. 2020AAA0105402) and the National Natural Science Foundation of China (Grant No. 72274200). We also acknowledge the support from Shanghai Qi Zhi Institute and Xi'an Institute for Interdisciplinary Information Core Technology.

ORCID iDs

Xiaoyan Mu

Xiaohu Zhang

Jiejing Wang

Supplemental Material

Supplemental material for this article is available online.

Author Biographies

Xiaoyan Mu is a Senior Research Assistant in the Department of Urban Planning and Design at the University of Hong Kong. She received her Ph.D. degree from the University of Hong Kong, and master’s and bachelor’s degree in Urban and Regional Planning from Peking University and Nanjing University, respectively. Her research interests include population migration, urban mobility, and urban system development in China.

Xiaohu Zhang is an Assistant Professor in the Department of Urban Planning and Design at the University of Hong Kong. Prior to this, he worked in Singapore-MIT Alliance for Research and Technology at MIT Senseable City Laboratory, and Sun Yat-Sen University. His scholarship bridges the information gap in sustainable urban and transportation policy-making with stochastic simulation and big data analytics. Broadly interested in urban data science, his research uses multi-source datasets to advance understanding of pressing urban and transportation issues.

Anthony G.O. Yeh is Chair Professor in the Department of Urban Planning and Design at the University of Hong Kong. His research interests include GIS, urban modelling, transport and big data, smart cities, and urban planning and development in Hong Kong, China, and Southeast Asia.

Yang Yu is an Assistant Professor in Interdisciplinary Information Sciences at Tsinghua University. He received his Ph.D. degree from Stanford University. His research focuses on smart grid, data and algorithm regulation, and market regulation.

Jiejing Wang is an Associate Professor at the School of Public Administration and Policy at Renmin University of China. His research interests focus on urbanization and urban system, spatial mobility, and urban social structure. His recent projects include the social and economic consequences of spatial mobility in China and the relationship between population mobility and urban governance.

References

Aknin

Andretti

Goldszmidt

, et al. (2022) Policy stringency and mental health during the COVID-19 pandemic: a longitudinal analysis of data from 15 countries. The Lancet Public Health 7(5): e417–e426.

Alexander

Karger

(2021) Do stay-at-home orders cause people to stay at home? Effects of stay-at-home orders on consumer behavior. The Review of Economics and Statistics 1–25. Epub ahead of print 22 September 2021. doi: 10.1162/rest_a_01108.

Barrero

Bloom

Davis

(2021) Why Working from home will stick (Working Paper No. 28731). National Bureau of Economic Research: Cambridge, MA, USA. Available at: https://www.nber.org/papers/w28731 (accessed 22 February 2023).

Chang

Pierson

Koh

, et al. (2021) Mobility network models of COVID-19 explain inequities and inform reopening. Nature 589(7840): 82–87.

Chen

Liu

et al. (2022) The economic cost of lockdown in China: Evidence from city-to-city truck flows. Working Paper. Hong Kong: Department of Economics, The Chinese University of Hong Kong. Available at: https://www.econ.cuhk.edu.hk/econ/images/Documents/Truck_Flow_and_COVID19_220315.pdf (accessed 22 February 2023).

Chinazzi

Davis

Ajelli

, et al. (2020) The effect of travel restrictions on the spread of the 2019 novel coronavirus (COVID-19) outbreak. Science 368(6489): 395–400.

Christidis

Christodoulou

Navajas-Cawood

, et al. (2021) The post-pandemic recovery of transport activity: Emerging mobility patterns and repercussions on future evolution. Sustainability 13(11): 6359.

Elldér

(2020) Telework and daily travel: New evidence from Sweden. Journal of Transport Geography 86: 102777.

Fezzi

Fanghella

(2021) Tracking GDP in real-time using electricity market data: Insights from the first wave of COVID-19 across Europe. European Economic Review 139: 103907.

10.

Galeazzi

Cinelli

Bonaccorsi

, et al. (2021) Human mobility in response to COVID-19 in France, Italy and UK. Scientific Reports 11: 13141.

11.

Glaeser

(2021) Urban resilience. Urban Studies 59(1): 3–35.

12.

Goldberg

(10 May 2022) Just 8% of Manhattan Office Workers Are Back Full Time, Survey Shows. The New York Times. Available at: https://www.nytimes.com/2022/05/10/business/return-to-office-manhattan.html (accessed 22 February 2023).

13.

Goolsbee

Syverson

(2021) Fear, lockdown, and diversion: Comparing drivers of pandemic economic decline 2020. Journal of Public Economics 193: 104311.

14.

Guan

Wang

Hallegatte

, et al. (2020) Global supply-chain effects of COVID-19 control measures. Nature Human Behaviour 4(6): 577–587.

15.

Guangdong Provincial Bureau of Statistics (2021) Guangdong Statistical Yearbook 2021. Beijing, China: China Statistics Press.

16.

Han

Fan

, et al. (2021) Quantifying COVID-19 importation risk in a dynamic network of domestic cities and international countries. Proceedings of the National Academy of Sciences of the United States of America 118(31): e2100201118.

17.

(2015) Telecommuting, income, and out-of-home activities. Travel Behaviour and Society 2(3): 131–147.

18.

International Monetary Fund (January 2022) World Economic Outlook Update: Rising Caseloads, a Disrupted Recovery, and Higher Inflation. .. Available at: https://www.imf.org/en/Publications/WEO/Issues/2022/01/25/world-economic-outlook-update-january-2022 (accessed 22 February 2023).

19.

Jay

Bor

Nsoesie

, et al. (2020) Neighbourhood income and physical distancing during the COVID-19 pandemic in the United States. Nature Human Behaviour 4(12): 1294–1302.

20.

Kraemer

MUG

Yang

C-H

Gutierrez

Klein

Pigott

, Open COVID-19 Data Working Group du Plessis

Faria

Hanage

Brownstein

Layan

Vespignani

Tian

Dye

Pybus

Scarpino

(2020) The effect of human mobility and control measures on the COVID-19 epidemic in China. Science 368(6490): 493–497.

21.

Lai

Ruktanonchai

Zhou

, et al. (2020) Effect of non-pharmaceutical interventions to contain COVID-19 in China. Nature 585(7825): 410–413.

22.

Zhang

Wang

, et al (2021b) Governing public health emergencies during the coronavirus disease outbreak: Lessons from four Chinese cities in the first wave. Urban Studies: 1-21. Epub ahead of print 26 October 2021. DOI: 10.1177/00420980211049350.

23.

Zhao

Haitao

, et al. (2021a) How did micro-mobility change in response to COVID-19 pandemic? A case study based on spatial-temporal-semantic analytics. Computers, Environment and Urban Systems 90: 101703.

24.

Llupià

Rodríguez-Giralt

Fité

et al. (2020) What is a zero-COVID strategy and how can it help us minimise the impact of the pandemic? ISGlobal. Instalment #26 of the COVID-19 and Response Strategy Series. Available at: https://www.isglobal.org/documents/10179/7943094/26_ISGlobal+COVID19+y+COVIDCero+o+Maxima+Supresion+EN/0a4e83bb-6257-4f5d-8960-16c323b464b2 (accessed 22 February 2023).

25.

(2005) Urban administrative restructuring, changing scale relations and local economic development in China. Political Geography 24(4): 477–497.

26.

McKenzie

Adams

(2020) A country comparison of place-based activity response to COVID-19 policies. Applied Geography 125: 102363.

27.

Yeh

AG-O

Zhang

(2020) The interplay of spatial spread of COVID-19 and human mobility in the urban system of China during the Chinese New Year. Environment and Planning B: Urban Analytics and City Science 48(7): 1955–1971.

28.

National Bureau of Statistics of China (2022) Tabulation on 2020 China Population Census by County. Beijing, China: China Statistics Press.

29.

Niu

Yue

Zhou

, et al. (2020) How urban factors affect the spatiotemporal distribution of infectious diseases in addition to intercity population movement in China. ISPRS International Journal of Geo-Information 9(11): 615.

30.

Ono

Mori

(2021) COVID-19 and telework: An international comparison. Journal of Quantitative Description: Digital Media 1: 1–35.

31.

Pepe

Bajardi

Gauvin

, et al. (2020) COVID-19 outbreak response: a first assessment of mobility changes in Italy following national lockdown. MedRxiv. DOI: 10.1101/2020.03.22.20039933.

32.

Pullano

Valdano

Scarpa

, et al. (2020) Evaluating the effect of demographic factors, socioeconomic factors, and risk aversion on mobility during the COVID-19 epidemic in France under lockdown: a population-based study. The Lancet Digital Health 2: e638–e649.

33.

Schlosser

Maier

Jack

, et al. (2020) COVID-19 lockdown induces disease-mitigating structural changes in mobility networks. Proceedings of the National Academy of Sciences 117(52): 32883–32890.

34.

Shanghai Municipal Bureau of Statistics (2021) Statistical Bulletin on Shanghai's Economic and Social Development in 2020. Available at: https://tjj.sh.gov.cn/tjgb/20210317/234a1637a3974c3db0cc47a37a3c324f.html (accessed 10 July 2022).

35.

Shepherd

HER

Atherden

Chan

HMT

, et al. (2021) Domestic and international mobility trends in the United Kingdom during the COVID-19 pandemic: An analysis of facebook data. International Journal of Health Geographics 20(1): 46.

36.

Sheridan

Andersen

Hansen

, et al. (2020) Social distancing laws cause only small losses of economic activity during the COVID-19 pandemic in Scandinavia. Proceedings of the National Academy of Sciences 117(34): 20468–20473.

37.

Tian

Liu

, et al. (2020) An investigation of transmission control measures during the first 50 days of the COVID-19 epidemic in China. Science 368(6491): 638–642.

38.

Tønnessen

(2021) Movers from the city in the first year of Covid. Nordic Journal of Urban Studies 1: 131–147.

39.

Van Bavel

Cichocka

Capraro

, et al. (2022) National identity predicts public health support during a global pandemic. Nature Communications 13(1): 517.

40.

Vilhelmson

(1999) Daily mobility and the use of time for different activities. The case of Sweden. GeoJournal 48(3): 177–185.

41.

Liu

Wang

, et al. (2022) A Systematic Review of COVID-19 Geographical Research: Machine Learning and Bibliometric Approach. Annals of the American Association of Geographers: 1–18. Epub ahead of print 26 October 2022. DOI: 10.1080/24694452.2022.2130143.

42.

Xiong

Yang

, et al. (2020) Mobile device data reveal the dynamics in a positive relationship between human mobility and COVID-19 infections. Proceedings of the National Academy of Sciences 117(44): 27087–27089.

43.

(2021) Causal inference with time-series cross-sectional data: a reflection. SSRN Electronic Journal. Available at SSRN: DOI: 10.2139/ssrn.3979613 (accessed 10 July 2022).

44.

Zha

Zhang

Zhou

, et al (2022) Chinese provincial government responses to COVID-19. Working Paper Version 2. Blavatnik School of Government. Available at: www.bsg.ox.ac.uk/covidtracker (accessed 10 July 2022).

45.

Zhaopin

Zhilian

(2022) Remote Work during the Pandemic: Reality and Outlook. National School of Development, Peking University. Available at: http://nsd.pku.edu.cn/docs/20220509205116111743.pdf (accessed 22 February 2023).

46.

Zhu

(2012) Are telecommuting and personal travel complements or substitutes? The Annals of Regional Science 48(2): 619–639.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

2.08 MB