Culture and effectiveness of distance restriction policies: evidence from the COVID-19 pandemic

Abstract

Natural disasters bring indelible negative impacts to human beings, and people usually adopt some post hoc strategies to alleviate such impacts. However, the same strategies may have different effects in different countries (or regions), which is rarely paid attention by the academic community. In the context of COVID-19, we examine the effect of distance restriction policies (DRP) on reducing human mobility and thus inhibiting the spread of the virus. By establishing a multi-period difference-in-differences model to analyse the unique panel dataset constructed by 44 countries, we show that DRP does significantly reduce mobility, but the effectiveness varies from country to country. We built a moderating effect model to explain the differences from the cultural perspective and found that DRP can be more effective in reducing human mobility in countries with a lower indulgence index. The results remain robust when different sensitivity analyses are performed. Our conclusions call for governments to adapt their policies to the impact of disasters rather than copy each other.

1. Introduction

COVID-19 is a global pandemic with the fastest transmission speed, the widest infection range and the greatest difficulty in prevention and control that human beings have encountered in the past century [1]. Coronavirus is rapidly spreading around the world. As of 5 March 2021, one year after the outbreak began, the number of confirmed cases worldwide had surpassed 110 million, with more than 2.5 million deaths. In the face of this major health event, governments in most countries have implemented distance restriction policies (DRP) and advised people to wash their hands and wear masks in order to prevent spreading the virus from person to person. Currently, in addition to physical measures, countries are also trying to contain the spread of the virus by using medical means like hospitalization and developing specific drugs and vaccines [2]. However, the medical effect is often limited by the variation of the virus, and DRP is still one of the most effective methods for the prevention and control of the disease [3].

Policymakers and agencies worldwide have quickly introduced various COVID-19-specific DRP, such as closing schools, restricting mass gatherings and advising people to quarantine at home [4], which have had an unprecedented impact on the global economy, especially in the tourism and entertainment industries [5,6]. Recognizing the economic damages of the DRP, governments and policymakers are eager to know whether these policies can effectively prevent human movement and the spread of COVID-19. In academia, there have already been some attempts to quantify the influence of DRP on human movement. This evidence suggests that, in general, the implementation of these DRP have successfully limited human mobility to some extent. Nouvellet et al. [7] pointed out that there was a clear relationship between mobility and transmission before and after the relaxation of strict control policies, and sustained social distance is effective to prevent infection. For example, Morita et al. [8] claimed that policies, such as workplace closures, restrictions on public transport and domestic travel, lead to a significant increase in social distancing, which further prevents the spread of COVID-19. Similarly, Armstrong et al. [9] investigated the effect of COVID-19 policies on human beings’ mobility patterns in 75 cities in Canada and the United States and found that these policies can effectively reduce human mobility after implementation. Praharaj et al. [10] indicated that although communities in the four cities responded to the policy change differently, mobility drop-off began during the same period after the policy was implemented and the declining trend of trip requests did not reverse after the lifting of lockdown restrictions. Hsiang et al. [4] showed that a series of DRP deployed by governments worldwide to reduce transmission rates achieved measurable and effective results. Praharaj and Han [11] indicated that authorities should focus on implementing strict distance restriction policies in hotspot locations that are most vulnerable.

However, although the DRP proved to be successful in some countries, they failed in some others. Although most countries have more or less adopted similar DRP to reduce human contact, the effectiveness of these measures varies significantly between countries. In some countries or areas, the DRP not only cannot effectively prevent the movement of the population, but also backfires by arousing anti-social-distancing protests. During the COVID-19 prevention and control period, it has been reported that people in the United Kingdom, Germany, France, Italy, Austria and some other countries erupted in large demonstrations to oppose the DRP. A substantial number of people gathered in these protests, which further increased human mobility and facilitated the spread of COVID-19. For example, in New Zealand, several anti-blockade protests were held, with people opposing the government’s COVID-19 social restrictions [12–14]. In Europe, there have been a number of anti-blockade protests in the UK related to the COVID-19 pandemic since the first blockade began in April 2020. On 16 May, 50 anti-blockade and anti-violence protesters in Hyde Park defied the social distance rule. Smaller protests took place in Manchester, Glasgow, Belfast and elsewhere on the same day. Protests eased over the summer as the blockade was lifted. However, protests began again when local lockdowns were reintroduced in response to the second wave of the pandemic. On 30 May, about 200 citizens from different parts of Italy held an unauthorized demonstration with the slogan ‘Give us back our freedom’ [15]. On 29 August, rallies protesting coronavirus restrictions broke out in several cities, including Paris, London, Vienna and Zurich. In Paris, some protesters said the government was exaggerating the outbreak and undermining their right to freedom. In addition, thousands of anti-lockdown protesters travelled from across the country to attend central London’s fourth mass protest against the government’s coronavirus measures on 24 October [16].

DRP often failed along with the rise of anti-social-distancing protests, a proportion of residents were less willing to comply with the DRP in some regions and countries. Protestors tied their protests into arguments about liberty and an anti-government ideology that opposed the extended period of social distancing and stay-at-home orders, calling for an end to government restrictions [17]. Therefore, the effectiveness of DRP may be influenced by certain factors. There are likely to be public opinion, religious, spiritual and social dimensions of individual responses to policies that may diminish the effectiveness of policy implementation [18,19]; the main reason may come from different cultural backgrounds among countries [20,21]. People with different cultures respond to policies differently, so governments and policymakers should take cultural factors into account in the enacting and implementation of policies [22].

Culture is a collective phenomenon that is at least partially shared by people living in the same social environment [23]. Culture guides human behaviour, and it has a wide and important influence on people’s collective behaviour in social settings [24–27]. Human behaviours, such as the daily movement of the population, are considered to influence the occurrence of disease events [28] and to influence pathogen adaptation, host susceptibility, spatial distribution and pathogen exposure and contact rates [22]. Previous literature confirms that culture has a certain influence on the transmission of the virus. Ngwa et al. [29] showed that cultural factors play an important role in cholera exposure and transmission in the far north of Cameroon. Fairhead [30], in his study of Ebola, argued that ignoring cultural factors at the beginning of the outbreak led to inefficient public health measures in the forest areas of the Republic of Guinea. The direct impacts of cultural factors on COVID-19 prevention have been examined in the existing literature. For instance, Muurlink and Taylor-Robinson [31] found that cultural factors may affect the gender balance reported in COVID-19 infection rates in a systematic manner. Huynh [32] argued that the proportion of the population gathered in public places is relatively low in a country with high uncertainty avoidance, so there is more social distance. Frey et al. [33] showed that collectivist and democratic countries respond effectively in terms of reducing human mobility.

To summarize, it is important to adopt a cultural perspective to understand the impacts of DRP on human mobility in culturally diverse countries during the COVID-19 pandemic. A cultural perspective can be useful for governments and policymakers to understand why similar DRP taken by different countries result in different outcomes, that is, the effectiveness of DRP. As mentioned above, many of the anti-social-distancing demonstrators took no protection during the gathering and demanded their freedom back. Such mass demonstrations may undermine the effectiveness of DRP. According to Pedersen and Favero [34], social distancing is an effective means of containing the outbreak, but only if the public participates. Dai et al. [35] also noted that full cooperation between the government and the public is essential to alleviate this crisis.

In this study, we argue that cultural factors have an important influence on the effectiveness of DRP in limiting human mobility. Drawing upon existing cultural studies, namely, studies on Hofstede’s cultural dimensions framework [23], there is reason to believe that people in countries with different cultural structures hold different attitudes toward DRP. We consider the indulgence dimension of Hofstede’s cultural dimension in this study. The indulgence culture may influence the executive power of the government and public cooperation, thus influencing residents’ tendency to comply with the DRP. Specifically, we postulate that people in countries with lower levels of indulgence are more likely to adhere to government-imposed DRP. As a result, in low-indulgence countries, DRP are more effective in limiting human mobility than those in high-indulgence countries.

To test our proposed hypothesis, we use a panel dataset of human mobility, DRP implementation, number of confirmed cases and temperature across 44 countries during the period 13 January to 30 August 2020, with indulgence cultural factors added. After controlling for country- and time-specific confounders, we find strong evidence that the decline in mobility is more pronounced in countries with lower levels of indulgence culture after the implementation of DRP.

We make the following important contributions to the relevant literature. First, we supplement the literature on the effectiveness of DRP in containing the COVID-19 pandemic [36,37]. Our results confirm that, in general, governments’ DRP lead to a decline in human mobility. Second, we propose that national culture has a certain impact on the effectiveness of DRP in preventing COVID-19. The majority of studies have investigated how social distancing policies implemented by government or cultural factors directly affect human mobility and transmission of COVID-19. However, these studies often do not adequately consider the suppressive effects of policy implementation, meaning that under certain circumstances, policies may produce unexpected and opposite outcomes to those intended. Additionally, there is a lack of analysis into the underlying reasons for this counter-effect phenomenon, particularly regarding the impact of national culture on the effectiveness of policy implementation. We further find that the indulgence culture plays an important moderating role in strengthening or attenuating the effect of DRP on human mobility. This will help governments to develop contextualized and country-specific pandemic control policies in accordance with their own cultures.

2. Methods

2.1. Hypothesis development

2.1.1. The impact of distance restriction policies on human mobility

During the COVID-19 pandemic, countries worldwide instituted various DRP to keep social distance to minimize human contact and halt the virus from spreading. For example, people turn their daily work into online work. They communicate and hold work meetings through video-conferencing tools, such as Zoom, Alibaba-owned DingTalk, Tencent’s WeChat Work and other video software [38]. Schools around the world also make full use of software for online teaching, like organizing tests and so on [39]. In order to further reduce the mass gathering of people, major shopping malls, restaurants, cinemas, tourist attractions and other entertainment venues were closed in the early stages of the pandemic and then restricted the flow of people during the mitigation period [40]. In addition, holiday celebrations, rallies and some major sporting events, such as the Union of European Football Associations Euro 2020 football championship, the Formula 1 Grand Prix in China and Olympic boxing qualifying events, were cancelled or postponed during the outbreak [41].

DRP, including both mandatory DRP and non-mandatory DRP, can have an impact on human mobility. Undoubtedly, formal institutions, such as laws and treaties, can influence public behaviour by imposing punishment on wrongdoers who do not comply with the policies. In some countries, the DRP are mandatory, such that people who do not comply with mandatory DRP will get punished. Meanwhile, informal institutions can support changes in social norms to help overcome global problems [42]. Non-mandatory DRP can limit human movement and alleviate the global COVID-19 pandemic by establishing social norms. Social norms are defined as external influences that are conditional and expectation-based [43], or ‘shared expectations’ are achieved through external sanctions [44]. Reynolds [45] summarized three main theories to explain how norms affect human behaviour. The information account views other people’s behaviour as a heuristic that provides quick and useful information. By observing others, we learn and gain social approval for our actions. That is, social proof, which is that ‘we use information about the behaviour of others to help us determine the correct behaviour’ [46, p. 149]. People may follow the behaviour of others because they believe in the wisdom of the crowd or because they get some kind of external approval or internal comfort from conformity under peer pressure to avoid being excluded [47]. The social sanction account emphasizes rewards and punishments. People who comply with norms may be based on the deterrence of policies, because those who violate norms will be suppressed and punished, while those who obey norms will gain social recognition. The self-categorization and internalization account includes the cognitive abilities to expand the self to include others as ‘similar-to-self’. As a social identity, the self can regard others as part of an ingroup, and when this social identity plays a psychological role, the influence between ingroup members increases [45, p. 14]. Summarizing the above argument, in terms of our research context, if a country implements some non-mandatory DRP, a social norm against human movement will be established. People will follow others to avoid interpersonal human contact during the COVID-19 pandemic to gain social recognition and social approval. As a result, DRP will exert a negative impact on human mobility. Therefore, we hypothesize:

H1: The implementation of policies associated with distance restriction can reduce human mobility.

2.1.2. The moderating role of indulgence culture

Culture is defined as a set of common values, beliefs and norms that distinguish members of a group or a class of people from others [23]. Hofstede’s cultural dimension provides strong theoretical support for the study of cross-cultural differences in the fields of finance [25,27], transportation [26], science and technology [48], climate and environment [49], and consumer behaviour [50]. Human behaviour is partly or entirely determined by cultural practices [22]. Cultural differences tend to become more pronounced when societies in different parts of the world are compared, and these distinctions have a predictable and uniform impact on people’s behaviour. Hofstede has proved that the national attribute of cultural influence does exist, and therefore, his theoretical framework can be used to predict how people act and the reasons for their behaviour [51].

According to the dimension of indulgence and restraint, indulgence is defined as a tendency to allow people relative freedom to satisfy desires related to the enjoyment of life. One can do as one pleases or indulge in leisure activities with friends. Restraint, the opposite of indulgence, reflects a belief that such gratification needs to be constrained and regulated by strict social norms. A person’s behaviour is restricted by various social norms and prohibitions [23, p. 281]. Hofstede et al. [23] argued that this indulgence culture dimension is somewhat like the distinction between loose societies and tight societies in American anthropology. In a loose society, norms can be expressed through a variety of channels and deviant behaviour is easily tolerated, while a tight society holds strong values of group organization, formality, permanence and solidarity [52,53].

DRP may affect human mobility through social norms. Although DRP can effectively reduce human mobility [4], in countries with a higher degree of indulgence, citizens may react more negatively to DRP that restrict free movement, and disobedience to government regulations may occur even if the pandemic is severe. Conversely, countries with lower levels of indulgence, owing to their higher tolerance of social norms, react less strongly and are more likely to adhere to appropriate social distance policies to restrict their travel. If people in low-indulgence countries do not abide by DRP or violate social norms, they may suffer higher social sanctions and more serious criticisms from the public. Based on the above discussion, we hypothesize:

H2. Compared with countries with a high degree of indulgence, the implementation of distance restriction policies reduces human mobility to a greater extent in countries with a low degree of indulgence.

2.2. Data collection and description of variables

To validate our proposed research hypotheses, we collected daily-basis panel data on the following variables: the country-level implementation of DRP, the country-level human mobility data, the indulgence culture index of different countries and regions, the country-level number of confirmed cases data and the country-level temperature data. Because the data are collected from multiple different sources and databases, after integrating all the above relevant variables, there are 44 countries left, whose information on DRP, human mobility, culture index, number of confirmed cases and temperature are all available and complete. The final sample data in our analysis includes 10 055 observations from 44 countries during the observation period between 13 January and 30 August 2020.

2.2.1. Data on the implementation of distance restriction policies

Policy data were obtained from the Coronanet COVID-19 Government Response Events dataset (https://coronanet-project.org). The database provides daily-basis data on the various fine-grained DRP taken by governments across various countries to combat the COVID-19 pandemic since the Chinese government first reported the COVID-19 outbreak on 31 December 2019 [54]. It includes information about the country where the policy was implemented, the type of policy (e.g. external border restrictions, health testing), the start date of the policy, the end date of the policy and so on. There are 16 types of DRP that have been implemented in response to COVID-19. The dataset for our study selected 10 of these policies related to human mobility, which covers a total of 10 055 daily-based observations of the implementation of the 10 DRP across 44 countries from 13 January to 30 August 2020. The start dates of the DRP are recorded as the date of the first implementation in each country. The end date of DRP implementation in each country is also recorded; if a country does not have an end date of DRP, the default end date is the last day of our dataset (i.e. 30 August 2020). Mexico was the first to implement DRP on 25 January. As of 30 August, 42 countries still adhere to the DRP, and 2 countries have terminated the DRP. Table 1 shows the descriptive statistics of the implementation of DRP across 44 counties. In table 1, for each type of the 10 DRP, we calculated the cumulative number of DRP implementation on a daily basis, as well as the number of countries that have ever implemented this type of DRP (table 1).

Table 1.

Descriptive information of the 10 types of DRP.

policy type	number of countries
social distancing	41
closure of schools	41
quarantine	40
lockdown	26
external border restrictions	23
internal border restrictions	29
restrictions of mass gatherings	39
restriction of non-essential businesses	44
restriction of non-essential government services	29
curfew	13

2.2.2. Human mobility data

Human mobility data were collected from the Mobility Trends Reports provided by Apple (https://covid19.apple.com/mobility). These reports contains information about the relative volume of requests in select countries/regions, sub-regions and cities directed through the Apple maps app on a daily basis (the baseline is the volume on 13 January 2020). The mobility information is further calculated as walking mobility and driving mobility [55].

Since there is a high correlation between these two types of mobility data (table 2), walking mobility and driving mobility are analysed separately in order to avoid the impact of collinearity. We used ‘walking mobility’ and ‘driving mobility’ as two separate dependent variables for cross-validation [55]. The estimation results using walking mobility and driving mobility were later proved to be highly consistent with each other (table 2).

Table 2.

Spearman correlation coefficients between DRP, indulgence culture, human mobility, mean temperature and number of confirmed cases.

	DRPit	IVRi	Drive Mobilityit	Walk Mobilityit	Tempit	Confi(t–1)
DRPit	1.000
IVRi	0.022	1.000
Drive_Mobilityit	−0.155*	−0.073*	1.000
Walk_Mobilityit	−0.322*	−0.042*	0.903*	1.000
Tempit	0.356*	−0.031*	−0.006	−0.104*	1.000
Confi(t–1)	0.617*	0.110*	−0.030*	−0.201*	0.3412*	1.0000

*p < 0.05.

IVR, indulgence.

2.2.3. Indulgence index

The country-level indulgence index was compiled by Hofstede et al. [23]. The indulgence index of each country is characterized by the numerical interval of 0–100, that is, the greater number represents the higher degree of indulgence. Indulgence, as the opposite of restraint, refers to the enjoyment of life and pleasure. It shows the extent to which people indulge in the enjoyment of life and seek instant gratification. A country with a high indulgence index means that it is a relaxed society whose citizens have a higher importance of leisure, higher optimism and less moral discipline. Freedom of speech was considered relatively important and maintaining order in this country was not a priority. On the contrary, a country with a low score indicates that it is a nervous society whose citizens are of lower importance of leisure, more pessimism and more moral discipline. Freedom of speech is less important, and maintaining national order is considered a priority [23]. The country’s indulgence index is shown in table 3.

Table 3.

Descriptive information of 44 countries’ indulgence index.

country	low indulgence index	country	high indulgence index
Latvia	13	Turkey	49
Lithuania	16	Greece	50
Bulgaria	16	Uruguay	53
Estonia	16	Norway	55
Romania	20	Luxembourg	56
India	26	Finland	57
Slovakia	28	Malaysia	57
Czech Republic	29	Brazil	59
Poland	29	Argentina	62
Italy	30	Austria	63
Hungary	31	South Africa	63
Croatia	33	Ireland	65
Portugal	33	Canada	68
Vietnam	35	The Netherlands	68
Indonesia	38	United Kingdom	69
Germany	40	Denmark	70
Philippines	42	Australia	71
Japan	42	New Zealand	75
Spain	44	Sweden	78
Thailand	45	Colombia	83
Singapore	46	Mexico	97
France	48
Slovenia	48

2.2.4. Number of confirmed cases data

People may reduce travel during severe outbreaks. Considering that the severity of COVID-19 may have an impact on human mobility, in the subsequent analysis, we took the number of confirmed cases as a proxy variable of the severity of COVID-19 and added them into the regression model as a control variable. We collected daily country-level number of confirmed cases data from the Our World in Data (OWID) website (https://ourworldindata.org/), which was founded by Max Roser and Esteban Ortiz-Ospina of the University of Oxford in the UK. The website aims to visualize data to show the development of outbreaks in different areas of the world. The site has a wide range of transparent data sources and has been used in a large number of empirical studies [56–58].

2.2.5. Temperature data

Based on previous studies on the relationship between temperature and outdoor activity or mobility behaviour [59–61] and prior research on finding a positive correlation between temperature and human mobility [62], we also controlled for the effect of temperature on human mobility. Temperature data were obtained from the National Oceanic and Atmospheric Administration Centre (www.ncdc.noaa.gov/isd). For each country, we averaged temperature data at different meteorological stations to calculate the average data on a daily basis during the observation period.

2.3. Specification of the multi-period difference-in-differences model

We use a multi-period difference-in-differences (DID) model to estimate the direct impact of DRP on human mobility, as well as the interaction effect between DRP and the indulgence culture. The multi-period DID model was adopted for the following reasons. First, the implementations of DRP in different countries are taken as exogenous intervention events, which are largely determined by the governments and are almost independent of the countries’ temperature, culture and mobility before the implementation of DRP. Second, although most of the countries in our dataset have adopted some of the DRP, the DRP start at different times and last for different time periods. For example, Latvia started implementing DRP on 25 February and did not stop until 30 August. South Africa began implementing the DRP on 18 March and ceased on 11 June. The multi-period DID model allows us to estimate the policy effect when the policy is implemented at different time points [63,64] so is appropriate to be used in this study. The regression equation can be specified as follows:

$${\displaystyle \{\begin{array}{l}\mathrm{W}\mathrm{a}\mathrm{l}\mathrm{k}\mathrm{\_}\mathrm{M}\mathrm{o}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}{\mathrm{y}}_{\mathrm{i}\mathrm{t}}\\ \mathrm{D}\mathrm{r}\mathrm{i}\mathrm{v}\mathrm{e}\mathrm{\_}\mathrm{M}\mathrm{o}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}{\mathrm{y}}_{\mathrm{i}\mathrm{t}}\end{array}=\alpha +\beta \mathrm{D}\mathrm{R}{\mathrm{P}}_{\mathrm{i}\mathrm{t}}+\gamma \mathrm{D}\mathrm{R}{\mathrm{P}}_{\mathrm{i}\mathrm{t}}\ast \mathrm{I}\mathrm{V}{\mathrm{R}}_{\mathrm{i}}+\theta \mathrm{C}\mathrm{o}\mathrm{n}{\mathrm{f}}_{\mathrm{i}\left(\mathrm{t}-1\right)}+\delta \mathrm{T}\mathrm{e}\mathrm{m}{\mathrm{p}}_{\mathrm{i}\mathrm{t}}+{\mu }_{\mathrm{i}}+{\sigma }_{\mathrm{t}}+{\epsilon }_{\mathrm{i}\mathrm{t}}.}$$ (2.1)

The subscript i and t represent the country and date, respectively. $\alpha$ represents the constant term. ${\displaystyle \mathrm{W}\mathrm{a}\mathrm{l}\mathrm{k}\mathrm{\_}\mathrm{M}\mathrm{o}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}{\mathrm{y}}_{\mathrm{i}\mathrm{t}}}$ and ${\displaystyle \mathrm{D}\mathrm{r}\mathrm{i}\mathrm{v}\mathrm{e}\mathrm{\_}\mathrm{M}\mathrm{o}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}{\mathrm{y}}_{\mathrm{i}\mathrm{t}}}$ represent the walking and driving human mobility in country i on day t. ${\displaystyle \mathrm{D}\mathrm{R}{\mathrm{P}}_{\mathrm{i}\mathrm{t}}}$ is the DRP treatment dummy. ${\displaystyle \mathrm{D}\mathrm{R}{\mathrm{P}}_{\mathrm{i}\mathrm{t}}}$ equals 1 if country i implemented any DRP on day t; ${\displaystyle \mathrm{D}\mathrm{R}{\mathrm{P}}_{\mathrm{i}\mathrm{t}}}$ equals 0 if country i did not implement any DRP on day t. Notably, in our dataset, there are 10 subtypes of DRP, but following the approach of Hsiang et al. [4], in equation (2.1), these 10 subtypes of DRP are integrated as one overall DRP treatment indicator. That is, ${\displaystyle \mathrm{D}\mathrm{R}{\mathrm{P}}_{\mathrm{i}\mathrm{t}}}$ equals 1 as long as one or more types of DRP are implemented by country i on day t, and ${\displaystyle \mathrm{D}\mathrm{R}{\mathrm{P}}_{\mathrm{i}\mathrm{t}}}$ equals 0 only if none of the 10 subtypes of DRP are implemented by country i on day t. ${\displaystyle \mathrm{I}\mathrm{V}{\mathrm{R}}_{\mathrm{i}}}$ is a time-invariant dummy variable that represents the country-level indulgence culture. We used the median split of the indulgence index to categorize the 44 countries in our dataset as 23 relatively high-indulgence countries and 21 relatively low-indulgence countries. That is, if the indulgence index in country i equals or is greater than the median (an indulgence index of 48), then ${\displaystyle \mathrm{I}\mathrm{V}{\mathrm{R}}_{\mathrm{i}}}$ is 1. Otherwise, ${\displaystyle \mathrm{I}\mathrm{V}{\mathrm{R}}_{\mathrm{i}}}$ equals 0. The interaction term, ${\displaystyle \mathrm{D}\mathrm{R}{\mathrm{P}}_{\mathrm{i}\mathrm{t}}}$ ∗ ${\displaystyle \mathrm{I}\mathrm{V}{\mathrm{R}}_{\mathrm{i}}}$ , is the variable of major interest, where the estimated value of the coefficient, γ, is the coefficient of the interaction effect between DRP and the indulgence culture. ${Comf}_{i(t-1)}$ represents the number of confirmed cases with a one-day lag. ${\displaystyle \mathrm{T}\mathrm{e}\mathrm{m}{\mathrm{p}}_{\mathrm{i}\mathrm{t}}}$ denotes the average temperature of country i on day t. We controlled the time-invariant countries’ characteristics, such as social and economic conditions (population density and GDP per capita) and the level of medical resources, by including the country fixed effect ${\mu }_i$ in equation (2.1). The time fixed effect, ${\sigma }_t$ , was also included to control for the seasoned effect as well as the global trend of the pandemic. Finally, ${\epsilon }_{it}$ is the error term. In addition, when performing regression, we standardized all variables (Z-score) to eliminate the impact of dimensions and avoid errors in estimates. It is important to note that when analysing regression results, the actual effect of DRP should be calculated by multiplying the regression coefficient by the standard deviation of mobility.

3. Results

3.1. Descriptive statistics

Table 4 summarizes the descriptive statistics of policy, human mobility, indulgence culture, mean temperature and number of confirmed cases in this study. During the observation period from 13 January to 30 August 2020, the average rate for DRP to be implemented across 44 countries on a daily basis was 0.768. The average daily temperature was 15.852°C. The average number of confirmed cases was 63554.74. The mean values of driving mobility and walking mobility were 1.345 and −8.623, respectively.

Table 4.

Descriptive statistics of DRP, indulgence culture, human mobility, mean temperature and number of confirmed cases.

	mean	s.d.	min	percentile			max
				25th	50th	75th
DRPit	0.768	0.422	0	1	1	1	1
IVRi	46.978	19.957	13	30	47	63	97
Drive_Mobilityit	1.345	55.349	−91.260	–35.730	−1.250	23.400	570.500
Walk_Mobilityit	−8.623	58.734	−94.180	−47.780	−12.055	16.650	788.440
Tempit (°C)	15.852	9.009	–20.749	9.219	15.999	23.767	32.561
Confi(t–1)	63554.74	257965.5	0	23	3124	24 506	38 04 803

Table 2 shows the Spearman correlation coefficient between policy, mobility data in Apple’s reports, indulgence culture index, temperature variables and number of confirmed cases. The implementation of DRP is significantly negatively correlated with driving mobility (r = −0.155, p < 0.05) and walking mobility (r = −0.322, p < 0.05). As we expected, walking mobility and driving mobility are highly correlated with each other.

3.2. Parallel trend test

Before testing our proposed hypotheses, we performed a parallel trend test to verify whether there was a parallel trend in human mobility before the implementation of DRP. The regression equation for the parallel trend test can be specified as follows:

$${\displaystyle {\displaystyle \{\begin{array}{l}\mathrm{W}\mathrm{a}\mathrm{l}\mathrm{k}\mathrm{\_}\mathrm{M}\mathrm{o}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}{\mathrm{y}}_{\mathrm{i}\mathrm{t}}\\ \mathrm{D}\mathrm{r}\mathrm{i}\mathrm{v}\mathrm{e}\mathrm{\_}\mathrm{M}\mathrm{o}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}{\mathrm{y}}_{\mathrm{i}\mathrm{t}}\end{array}=\alpha +{\beta }_T\ast \sum _{T=-8}^{T=12}\mathrm{D}\mathrm{R}{\mathrm{P}}_{\mathrm{i}\mathrm{t}}+\delta \mathrm{T}\mathrm{e}\mathrm{m}{\mathrm{p}}_{\mathrm{i}\mathrm{t}}+\theta \mathrm{C}\mathrm{o}\mathrm{n}{\mathrm{f}}_{\mathrm{i}\left(\mathrm{t}-1\right)}+{\mu }_{\mathrm{i}}+{\sigma }_{\mathrm{t}}+{\epsilon }_{\mathrm{i}\mathrm{t}}.}}$$ (3.1)

In equation (3.1), for each country i, T equals 0 if DRP starts on day T. We observed the differences in human mobility 8 days before the implementation of DRP and 12 days after the implementation of DRP. The estimated coefficients of the DRP treatment effects during the 21 days are further depicted in figures 1 and 2.

Figure 1.

Parallel trend test for the treatment effect of DRP on walking mobility.

Figure 2.

Parallel trend test for the treatment effect of DRP on driving mobility.

According to figures 1 and 2, before the implementation of DRP (i.e. T = −8 to T = −1), the coefficients of ${\displaystyle \mathrm{D}\mathrm{R}{\mathrm{P}}_{\mathrm{i}\mathrm{t}}}$ (i.e. ${\beta }_T$ ) are not significantly different from zero; after the implementation of DRP (i.e. T = 1 to T = 12), the coefficients of ${\displaystyle \mathrm{D}\mathrm{R}{\mathrm{P}}_{\mathrm{i}\mathrm{t}}}$ are significantly below zero. These results serve as strong evidence that there was a parallel trend in human mobility data across 44 countries before the implementation of DRP. However, after some countries started to implement DRP, we can quickly observe a significant decrease in human mobility. We conclude that the different trend in the change of human mobility before the implementation of DRP is not a serious concern. With this observed parallel trend, we can further apply the multi-period DID approach to estimate the treatment effects of DRP.

3.3. Model estimation and hypothesis testing

Table 5 reports the main empirical results. We observed a significant negative correlation between policy and human mobility (including both walking mobility and driving mobility), and hence H1 was supported. The results indicate that if any DRP are implemented by a certain country on a certain day, for countries with lower levels of indulgence, on average it can lead to a 13.51 (0.23 ∗ 58.734 ≈ 13.51) (t = −9.040) decrease in walking mobility and a 3.49 (0.063 ∗ 55.349 ≈ 3.49) (t = −2.936) decrease in driving mobility. It can be seen that the impact of DRP on walking mobility is much greater than that of driving mobility.

Table 5.

Model estimation results.

variables	Walk_Mobilityit	Drive_Mobilityit	Walk_Mobilityit	Drive_Mobilityit
DRPit			−0.230***	−0.063***
			[−9.040]	[–2.936]
DRPit*IVRi			0.181***	0.106***
			[7.028]	[4.856]
Tempit	0.397***	0.484***	0.409***	0.491***
	[37.34]	[50.61]	[39.75]	[51.12]
Confi(t–1)	–0.0884***	–0.0910***	−0.0877***	–0.0894***
	[−13.72]	[−13.09]	[−13.56]	[−12.94]
country fixed effect	yes	yes	yes	yes
time fixed effect	yes	yes	yes	yes
constant	−1.51e−08	−1.68e−08	0.111***	0.0106
observations	10 055	10 055	10 055	10 055
R 2	0.641	0.756	0.643	0.756

Note: This table reports equation (2.1) estimated coefficients and t values.

***p < 0.01, **p < 0.05, *p < 0.1.

Moreover, the interaction term, ${\displaystyle \mathrm{D}\mathrm{R}{\mathrm{P}}_{\mathrm{i}\mathrm{t}}}$ ∗ ${\displaystyle \mathrm{I}\mathrm{V}{\mathrm{R}}_{\mathrm{i}}}$ , is significantly and positively related to human mobility, indicating that in countries with a high degree of indulgence, the negative effect of policy on walking mobility is weaker. This result confirms that indulgence culture moderates the relationship between policy and human mobility. Finally, we also found that temperature and the number of confirmed cases had an impact on human mobility. When the temperature is low or the number of confirmed cases is large, human mobility is inhibited.

3.4. Robustness checks

We performed several robustness tests. First, we show the results that the countries are not grouped by the indulgence cultural index, that is, IVR is measured using standardized real data (table 6). We found that the results were still robust for walking mobility, but the impact of DRP disappeared for driving mobility, meaning that the impact of DRP on countries with lower levels of indulgence index was neutralized. Therefore, it is necessary to group countries by indulgence index in the basic model. On the other hand, it also indirectly indicates that the impact of DRP on driving mobility is indeed much weaker than that of walking mobility.

Table 6.

Robustness checks—estimation results under when IVR is measured using standardized real data.

variables	Walk_Mobilityit	Drive_Mobilityit
DRPit	−0.130***	−0.00574
	[−7.251]	[−0.345]
DRPit*IVRi	0.0645***	0.0157
	[5.513]	[1.469]
Tempit	0.407***	0.486***
	[39.17]	[50.65]
Confi(t–1)	−0.0885***	−0.0906***
	[−13.59]	[−12.99]
country fixed effect	yes	yes
time fixed effect	yes	yes
constant	0.0981***	0.00415
observations	10 055	10 055
R 2	0.643	0.756

Note: This table reports equation (2.1) estimated coefficients and t values.

***p < 0.01, **p < 0.05, *p < 0.1.

Second, considering that a country has different policy combinations at different times, and different countries have different policy combinations at the same time, we add policy fixation to control differences in the effects of different subtypes of DRP over the same time period. As shown in table 7, the interaction term ( ${\displaystyle \mathrm{D}\mathrm{R}{\mathrm{P}}_{\mathrm{i}\mathrm{t}}}$ ∗ ${\displaystyle \mathrm{I}\mathrm{V}{\mathrm{R}}_{\mathrm{i}}}$ ) is still positive and significant, and our conclusion is still consistent. The regression equations corresponding to table 7 can be specified as follows:

Table 7.

Robustness checks—estimation results with policy fixation.

variables	Walk_Mobilityit	Drive_Mobilityit
DRPit	−0.553***	−0.476***
	[−10.60]	[−10.07]
DRPit*IVRi	0.253***	0.175***
	[10.35]	[8.282]
Tempit	0.319***	0.370***
	[24.67]	[31.37]
Confi(t–1)	−0.0673***	−0.0531***
	[−9.144]	[−7.087]
observations	10 055	10 055
country fixed effect	yes	yes
time fixed effect	yes	yes
policy-type fixed effect	yes	yes
constant	0.0163	−0.0909***
R 2	0.784	0.858

Note: This table reports equation (3.2) estimated coefficients and t values.

***p < 0.01, **p < 0.05, *p < 0.1.

$${\displaystyle \{\begin{array}{l}\mathrm{W}\mathrm{a}\mathrm{l}\mathrm{k}\mathrm{\_}\mathrm{M}\mathrm{o}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}{\mathrm{y}}_{\mathrm{i}\mathrm{t}}\\ \mathrm{D}\mathrm{r}\mathrm{i}\mathrm{v}\mathrm{e}\mathrm{\_}\mathrm{M}\mathrm{o}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}{\mathrm{y}}_{\mathrm{i}\mathrm{t}}\end{array}=\alpha +\beta \mathrm{D}\mathrm{R}{\mathrm{P}}_{\mathrm{i}\mathrm{t}}+\gamma \mathrm{D}\mathrm{R}{\mathrm{P}}_{\mathrm{i}\mathrm{t}}\ast \mathrm{I}\mathrm{V}{\mathrm{R}}_{\mathrm{i}}+\theta \mathrm{C}\mathrm{o}\mathrm{n}{\mathrm{f}}_{\mathrm{i}\left(\mathrm{t}-1\right)}+\delta \mathrm{T}\mathrm{e}\mathrm{m}{\mathrm{p}}_{\mathrm{i}\mathrm{t}}+{\mu }_{\mathrm{i}}+{\sigma }_{\mathrm{t}}+{\mathrm{p}}_{\mathrm{i}\mathrm{t}}+{\epsilon }_{\mathrm{i}\mathrm{t}}.}$$ (3.2)

The policy fixation, $p_{it}$ , was included to control for the impact of different policy combinations on human mobility. In our sample, there are 184 types of policy combinations.

Third, we further added country-specific linear time trends to our model (table 8), which can control invisible country-level factors that evolve over time. The regression equations corresponding to table 8 can be specified, respectively, as follows:

Table 8.

Robustness checks—estimation results with country-specific linear time trend.

variables	Walk_Mobilityit	Drive_Mobilityit
DRPit	−0.533***	−0.417***
	[−24.19]	[−22.26]
DRPit*IVR	0.597***	0.467***
	[24.41]	[21.58]
Tempit	0.125***	0.130***
	[8.320]	[9.620]
Confi(t–1)	−0.0526***	−0.0177**
	[−8.053]	[−2.342]
country fixed effect	yes	yes
time fixed effect	yes	yes
country fixed effect*time	yes	yes
constant	0.194***	0.152***
observations	10 055	10 055
R 2	0.833	0.877

Note: This table reports equation (3.3) estimated coefficients and t values.

***p < 0.01, **p < 0.05, *p < 0.1.

$${\displaystyle \{\begin{array}{l}\mathrm{W}\mathrm{a}\mathrm{l}\mathrm{k}\mathrm{\_}\mathrm{M}\mathrm{o}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}{\mathrm{y}}_{\mathrm{i}\mathrm{t}}\\ \mathrm{D}\mathrm{r}\mathrm{i}\mathrm{v}\mathrm{e}\mathrm{\_}\mathrm{M}\mathrm{o}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}{\mathrm{y}}_{\mathrm{i}\mathrm{t}}\end{array}=\alpha +\beta \mathrm{D}\mathrm{R}{\mathrm{P}}_{\mathrm{i}\mathrm{t}}+\gamma \mathrm{D}\mathrm{R}{\mathrm{P}}_{\mathrm{i}\mathrm{t}}\ast \mathrm{I}\mathrm{V}{\mathrm{R}}_{\mathrm{i}}+\theta \mathrm{C}\mathrm{o}\mathrm{n}{\mathrm{f}}_{\mathrm{i}\left(\mathrm{t}-1\right)}+\delta \mathrm{T}\mathrm{e}\mathrm{m}{\mathrm{p}}_{\mathrm{i}\mathrm{t}}+{\mu }_{\mathrm{i}}+{\sigma }_{\mathrm{t}}+{\omega }_{\mathrm{i}\mathrm{t}}+{\epsilon }_{\mathrm{i}\mathrm{t}},}$$ (3.3)

where ω_it is the country-specific linear time trend. It controls factors that change over time and country to some extent. The results are still robust.

Finally, we estimated the effects of the different subtypes of DRP on human mobility. The selected policy group consists of the four most widely adopted DRP out of the 10 subtypes of DRP, with the greatest policy impact (table 9). The regression equations corresponding to table 9 can be specified, respectively, as follows:

Table 9.

Robustness checks—effects of different subtypes of DRP.

variables	Walk_Mobilityit	Drive_Mobilityit	Walk_Mobilityit	Drive_Mobilityit	Walk_Mobilityit	Drive_Mobilityit	Walk_Mobilityit	Drive_Mobilityit
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
main effects of DRP
bu_DRPit	−0.231***	−0.206***	−0.124***	−0.115***	−0.0806***	−0.0973***	−0.115***	−0.117***
go_DRPit	−0.120***	−0.0415**	−0.311***	−0.0448	−0.0909***	−0.0169	−0.0900***	−0.0176
ma_DRPit	−0.149***	−0.0783***	−0.148***	−0.0837***	−0.298***	−0.162***	−0.179***	−0.0991***
lo_DRPit	−0.137***	−0.148***	−0.135***	−0.141***	−0.139***	−0.146***	−0.349***	−0.286***
interaction effect between DRP and indulgence culture
bu_DRPit *IVR	0.269***	0.204***
go_DRPit *IVR			0.393***	0.0591*
ma_DRPit *IVR					0.299***	0.164***
lo_DRPit *IVR							0.404***	0.265***
Tempit	2.842***	3.214***	2.838***	3.188***	2.767***	3.167***	2.785***	3.176***
Confi(t–1)	−0.0929***	−0.0909***	−0.0834***	−0.0843***	−0.0832***	−0.0838***	−0.0767***	−0.0767***
country fixed effect	yes	yes	yes	yes	yes	yes	yes	yes
time fixed effect	yes	yes	yes	yes	yes	yes	yes	yes
constant	0.164***	0.119***	0.166***	0.114***	0.141***	0.105***	0.181***	0.129***
observations	10 055	10 055	10 055	10 055	10 055	10 055	10 055	10 055
R 2	0.651	0.761	0.652	0.760	0.651	0.761	0.652	0.761

Note: ${\displaystyle bu\mathrm{\_}\mathrm{D}\mathrm{R}\mathrm{P}{\mathrm{P}}_{\mathrm{i}\mathrm{t}}}$ , ${\displaystyle go\mathrm{\_}\mathrm{D}\mathrm{R}{\mathrm{P}}_{\mathrm{i}\mathrm{t}}}$ , ${\displaystyle ma\mathrm{\_}\mathrm{D}\mathrm{R}{\mathrm{P}}_{\mathrm{i}\mathrm{t}}}$ , and ${\displaystyle lo\mathrm{\_}\mathrm{D}\mathrm{R}{\mathrm{P}}_{\mathrm{i}\mathrm{t}}}$ represent the policy of restriction of non-essential businesses, the policy of restriction of non-essential government services, the policy of restrictions of mass gatherings and the policy of lockdown, respectively. This table reports equation (3.4) estimated coefficients and t values.

***p < 0.01, **p < 0.05, *p < 0.1.

$${\displaystyle \{\begin{array}{l}\mathrm{W}\mathrm{a}\mathrm{l}\mathrm{k}\mathrm{\_}\mathrm{M}\mathrm{o}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}{\mathrm{y}}_{\mathrm{i}\mathrm{t}}\\ \mathrm{D}\mathrm{r}\mathrm{i}\mathrm{v}\mathrm{e}\mathrm{\_}\mathrm{M}\mathrm{o}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}{\mathrm{y}}_{\mathrm{i}\mathrm{t}}\end{array}=\alpha +\beta \mathrm{D}\mathrm{R}{\mathrm{P}}_{\mathrm{k}\mathrm{i}\mathrm{t}}+\gamma \mathrm{D}\mathrm{R}{\mathrm{P}}_{\mathrm{k}\mathrm{i}\mathrm{t}}\ast \mathrm{I}\mathrm{V}{\mathrm{R}}_{\mathrm{i}}+\theta \mathrm{C}\mathrm{o}\mathrm{n}{\mathrm{f}}_{\mathrm{i}\left(\mathrm{t}-1\right)}+\delta \mathrm{T}\mathrm{e}\mathrm{m}{\mathrm{p}}_{\mathrm{i}\mathrm{t}}+{\mu }_{\mathrm{i}}+{\sigma }_{\mathrm{t}}+{\epsilon }_{\mathrm{i}\mathrm{t}},}$$ (3.4)

where k indicates the policy type. Our results were verified again when these different subtypes of DRP were treated as separate and independent interventions.

4. Discussion

This research aims to explore the underlying mechanisms of how DRP affect human mobility by assessing the moderating role of indulgence culture. In this study, we observe a negative correlation between policy and human mobility. The moderating analysis showed that countries with lower levels of indulgence had a greater increase in the effect of policies on human mobility than countries with higher levels.

First, for the main effects of DRP on human mobility, our results are consistent with previous studies. In the early days of the outbreak, most countries only imposed international travel restrictions on China with few other preventive measures and little attention paid to the outbreak. With the rapid spread of the pandemic, governments and WHO took similar preventive measures to reduce COVID-19 infection rates. For example, requiring people to wear masks in public places, washing and sanitizing hands frequently, and implementing social distance policies to control the person-to-person transmission of the virus [20]. From the perspective of global prevention and control, strategies can be simply divided into two categories. The first is the blocking strategy adopted by China, South Korea, Iran and Singapore in the early stage of COVID-19, that is, through some more aggressive containment management measures to completely stop the transmission. The second category is the mitigation strategies, which are relatively mild, to increase social distance taken by the United States, Germany, the United Kingdom and Italy. The strong impact of these policies on reducing mobility has been verified in the existing literature [65,66].

Second, and more importantly, we contribute to the literature by analysing the cultural aspect that either strengthens or attenuates the effects of DRP on human mobility. Previous studies have found that indulgence culture has a significant positive effect on the percentage per million of total COVID-19 cases [20]. We moved a step further by investigating the moderation effects of indulgence culture on the relationship between DRP and human mobility. Hofstede et al. [23] pointed out that the percentage of respondents with high optimism was significantly correlated with the score of indulgence. People are more optimistic in a society with higher indulgence and vice versa. Therefore, in a country with higher indulgence, people may have a more optimistic attitude towards COVID-19 and may have a more difficult time coordinating themselves in the face of the pandemic [67]. A high-indulgence society allows for the gratification of basic desires related to the enjoyment of life in a relatively free manner, while a highly restrained society usually constrains citizens’ behaviour through strict social norms [68]. However, a successful response to the COVID-19 crisis will require not only effective government enforcement but also collaborative public participation [69,70]. Li et al. [71, p. 1] believed that governments worldwide should take their own culture into consideration and learn from the early experience of other countries in pandemic prevention, such as ‘timely and reasonable control of local population flow’. Owing to differences in lifestyles and cultural backgrounds, countries should try to strike a balance between culture and policy [67]. In this context, governments should tolerate each other and seek cultural and political consensus so as to jointly overcome this historically tragic pandemic [71].

Our study has practical implications for governments and policymakers. First, we find a negative correlation between DRP and human mobility. Governments need to consider DRP as an effective approach to limiting human movement and preventing the spread of COVID-19. In addition, recognizing the moderating effect of indulgence culture, we suggest governments should develop country-specific pandemic control policies depending on the severity of the outbreak that are in alignment with their own cultures. For low-indulgence countries, they are recommended to stick to their current DRP as these policies are highly compatible with their social norms. However, DRP work less effectively in high-indulgence countries and sometimes even backfire by arousing anti-social-distancing protests, so these countries should consider developing alternative policies to prevent person-to-person transmissions of the COVID-19.

Finally, this paper also has some limitations. First, Apple’s human mobility metrics are generated by the number of navigation requests made to Apple Maps. Although this measure of human mobility has been widely adopted in prior studies, it may still potentially lead to increased estimation bias in a country-level study. Second, since the average temperature of all available meteorological stations is used to represent the temperature of each country, the exposure measurement error is unavoidable. Third, although we controlled for the country fixed effect and time fixed effect in our estimation, there were still a large number of confounding factors that were not controlled, which may have influenced the results.

5. Conclusion

This study provides evidence that in countries with low-indulgence cultures, the implementation of DRP is more effective in reducing human mobility. After controlling for country and time characteristics, the policy has a more obvious result in dampening mobility. Taking cultural considerations into account in the COVID-19 pandemic will help improve the global response to the current crisis, provide implications for governments and relevant policymakers and provide guidance for future implementation of DRP as well.

Ethics

This work did not require ethical approval from a human subject or animal welfare committee.

Data accessibility

The data used in our study are all publicly available from links provided in the main article, or from the supplementary material [72].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors’ contributions

Y.X.: conceptualization, formal analysis, funding acquisition, methodology, project administration, software, validation, writing—review and editing; C.D.: formal analysis, investigation, software, writing—review and editing; W.S.: conceptualization, formal analysis, investigation, software, validation, visualization, writing—original draft, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This work is supported by the National Natural Science Foundation of China (grant nos.: 72201001 and 72101003), Outstanding Youth Research Project for Anhui Universities (grant no.: 2023AH030014) and Anhui Province Higher Education Science Research Project (grant no.: 2022AH050861).