## Introduction

Since its detected inception in late December 2019, COVID-19 has been ravaging the humanity and is affecting all countries in the world. While the medical profession has made vaccines available in 2021 and have recently developed cures for the disease, non-clinical approaches, such as mobility restriction, facial mask wearing, contact tracing and social distancing, have been adopted, with some success, to stem the spread of the virus, especially at its early stage. Tian et al. (2020) show that lockdowns, travel restrictions, closure of entertainment venues, and national emergency measures delay the growth and the scale of COVID-19. Chinazzi et al. (2020) provide evidence that while the Wuhan travel ban had a relatively limited effect in containing infections in mainland China as many cities had already received infected people, its international effect is more pronounced. Cyranoski (2020) reports that China’s mammoth response to the virus, such as strict lockdowns and travel restrictions, has been effective in stemming the virus. Schlosser et al. (2020) show that COVID-19 lockdown reduced mobility in Germany and Mitze et al. (2020) show that face masks considerably reduce infections in Germany. These studies seem to suggest that limiting people’s physical mobility and enhancing self-protection awareness can pay dividend in preventing a wider spread of COVID-19.

However, there are also ample criticisms of government-initiated travel restrictions. Meier et al. (2020) argue that lockdowns, travel restrictions and limiting freedom of movement violate international law, which can lead to economic isolation and human rights violations. Many governments indeed struggle to maintain the subtle balance between executive power and human rights in the fight against COVID-19 (Gostin and Wiley, 2020).

We examine how a culture of individualism affects the fight against COVID-19. Individualism captures the extent to which people in a society are mentally and habitually empowered to make their own choices (Hofstede 1980, 2001). An individualistic culture attaches a greater reward to individual achievements, promoting innovation and economic growth (Gorodnichenko and Roland 2011, 2017). However, limitations of individualism were also recognized long ago. De Tocqueville (1835) expresses the concern that individualistic people tend to prioritize self-interest over public interest. Patterson (1993) points out that individualism can devolve into selfishness despite its commitment to human potentials and accomplishments. During a pandemic, at the policy-making level, the government of a country with an individualistic culture can be more hesitant in initiating restrictions which may compromise personal freedom; at the social level, citizens of country with an individualistic culture are less likely to follow government orders, rendering social distancing, mobility restrictions and other government-initiated policies more difficult to enforce. In sum, while an individualistic culture has its indisputable merits, they can potentially thwart an effective fight against COVID-19 due to the difficulty in both policy coordination and social uncooperativeness, resulting in a less effective containment of the pandemic.

We start with an international analysis examining virus-combatting efficacy leveraging on cross-country variation in individualism. We control for national characteristics which can underline emergency handling styles, economic conditions, medical capacity, virus testing capacity, past epidemic experience which may directly affect the reported number of contagion and casualty as well as the extent of vaccination. We cover the entire period of years 2020 and 2021. We find that COVID-19 is more severe in countries scoring high on individualism. In addition, individualism appears to thwart the effectiveness of social distancing and mobility restriction policies in fighting COVID-19, aggravating the virus situation. To alleviate the endogeneity concern and strengthen causal inference, we adopt an instrumental variable approach by using a country’s number of Nobel Prize winners as an instrument, which affects a country’s virus outcome through affecting individualism. Overall, our international evidence suggests that individualism leads to a more severe COVID-19 situation in terms of infection and death throughout the prolonged pandemic trajectory from 2020 to 2021.

We next conduct a within-country study focusing on one country, Germany, where we can largely assume identical national-level virus-fighting policies, testing capacity and reporting transparency. The German setting offers us a unique opportunity to leverage on the disparity in the individualism-collectivism paradigm between former East and West German districts, inherited from their divergent political trajectories prior to the German reunification in 1990. Formal socialist regime and economic structure, prior to the reunification of the East and West Germany, have caused the social culture of the former East Germany to be more collectivist-inclined, as a result of decades of institutional learning and political socialization (Jennings and Niemi 1968; Niemi and Sobieszek 1977; Healy and Malhotra 2013; Sack 2017). As collectivists are characterized with social norm-guided behaviours while individualists are characterized with attitude-guided behaviours (Triandis 1996), we expect that people in former East German districts are more likely than those from former West German districts to accept government virus-fighting policies that are potentially intrusive to personal freedom, mitigating the virus situation in former East German districts. We find that without considering differences in district characteristics, former East German districts had a less severe COVID-19 situation in 2020 but a more severe situation in 2021. However, when we incorporate district characteristics that critically affect the health outcome during the pandemic, former East German districts had a less severe COVID-19 situation throughout 2020 and 2021.

Overall, we demonstrate that an individualistic culture is less than conducive to fighting COVID-19. This inefficacy can potentially be attributed to a lack of stringent policy responses as well as social uncooperativeness. People in countries with an individualistic culture are less likely to heed their governments’ restrictive policies and coordinate their efforts in containing the virus. While we do not have strong evidence that individualism systematically affects the issuance of virus-fighting policies by the government, we show that individualism reduces the effectiveness of these policies in an international setting. In the presence of an unprecedented pandemic, individualism, through more assertively and unconditionally demanding personal freedom, limits the effect of executive powers and potentially imposes a negative health externality on the general public and the society.

## Theoretical background

### Individualism/collectivism as cultural/societal phenomena

According to Hofstede’s (1980) influential work on culture, individualism is defined as a focus on rights above duties, a concern for oneself and immediate family, an emphasis on personal autonomy and self-fulfilment, and the establishment of one’s identity on one’s personal accomplishments. Hofstede also argues that individualism is in part a reflection of social-structural conditions and assumes that individualism and collectivism form a single continuum, with low individualism corresponding to high collectivism. At a societal level, Schwartz (1990) defines individualistic societies as fundamentally contractual with specific obligations and expectations focusing on achieving status; while collectivist societies are communal societies characterized by diffuse and mutual obligations and expectations based on ascribed status. These definitions of individualism/collectivism as societal cultures reveal a dilemma between personal freedom and social obligations. Therefore, the societal value orientation of individualism/collectivism can have profound implications on the way a state deals with the inherent tension between collective/public goods and individual freedom (Conway et al. 2006). As manifested in state policies and legislation decisions, the government of an individualistic culture can be more hesitant in implementing policies that compromise citizens’ demands for personal rights and freedom of choices.

Other scholars, based on individual psychological constructs, conceptualize individualism and collectivism as worldviews that differ on the issues they make salient (Triandis 1996; Kagitcibasi 1997). Triandis (1996) illustrates individualism and collectivism as syndromes of constructs that tend to co-occur, with the collectivist syndrome defined as viewing self as interdependent, giving precedence to in-group goals, expecting norms rather than attitudes to guide behaviour, and engaging in communal rather exchange relationships; and individualism defined as the opposite set of symptoms. The characteristics of norm-guided behaviours within collectivism and attitude-guided behaviours within individualism suggest that such psychological constructs can be manifested in contrasting attitudes toward prescribed status and hierarchy, such as the state authority. Collectivist legislators can view government restrictions as an acceptable form of normative influence on behaviours and collectivistic oriented citizens would display more willing acceptance of the legitimate authority and higher adherence to in-group norms, for example, state policies and government-initiated restrictions.

### Individualism/collectivism and government-initiated restrictions

Prior literature has illustrated the association between individualism/collectivism and the acceptance of authority, including government-initiated restrictions, applying different measures of the construct and in different societal contexts. For example, Hofstede (1980) finds a pronounced negative association between individualism and power distance at the aggregate level. Gelfand et al. (1996) show that the concepts of individualism and authoritarianism are perceived to be semantic opposites. Kemmelmeier et al. (2003) demonstrate, both at the individual level and the societal level, that authoritarianism is associated with those aspects of collectivism referring to hierarchy and submission to in-group authority. In two post-communist societies, Bulgaria and Poland, a high level of authoritarianism is significantly related to a low level of individualism. Conway et al. (2006) provide evidence that an index of government-initiated restrictions is positively associated with measures relevant to cultural collectivism, even when controlling for an economic indicator strongly associated with collectivism.

### Prediction for COVID-19

COVID-19 offers a great opportunity to study how individualism/collectivism influence the trajectory of a pandemic. While an individualistic culture has its indisputable merits, they can potentially thwart an effective fight against COVID-19. Our argument is straightforward. In countries where individualism is widely valued and respected in civic and economic life, governments may be more reluctant to impose mobility restrictions and these restrictions, when imposed, are less likely to be heeded by people, preventing their enforcement and thus an effective fight against the virus. In other words, the value associated with individualism can aggravate the spread and thus the severity of COVID-19 through policy inaction and/or social non-cooperativeness. We therefore expect a positive association between individualism and the severity of COVID-19.

## An international analysis of individualism and COVID-19

### Variables and statistics

To test our prediction, we start with an international analysis. We consider three metrics to gauge a country’s severity of COVID-19, Confirmed Cases, Deaths and Death Ratio. This information is obtained from Our World in Data. Specifically, We define Confirmed Cases as the logarithm transformation of one plus the cumulative daily number of confirmed cases in every 10,000 people of a country’s population. We define Deaths as the logarithm transformation of one plus the cumulative daily number of deaths in every 10,000 people of a country’s population. We also define a death ratio variable Death Ratio which is the number of deaths divided by the number of confirmed cases (in percentage format). Definitions of these variables and all subsequent variables are consolidated in Supplemental Table 1.

Values for the variable Individualism are obtained from the World Values Survey. World Values Survey’s multi-item index measures the extent to which people in a society are mentally and habitually empowered to make their own choices and to pursue them in their actions. The index covers the domains of motivational empowerment (emancipative values), intellectual empowerment (formal education) and behavioural empowerment (social movement activity).

We control for general national characteristics such as geographic size, Land_area; the percent of population over age 65, Population_age65; annual GDP growth rate, GDP_growth; the percent of exports of goods and services over GDP, Export; final consumption expenditure as a percentage of GDP, Consumption; total alcohol consumption per capita (liters of pure alcohol, projected estimates, 15+ years of age), Alcohol; international tourism, number of arrivals, divided by population, Arrival; refugee population by country or territory of asylum, divided by population, Refugee; individuals using the Internet as a percentage of population, Internet; fixed broadband subscription per 100 people, Broadband; and essential healthcare capacity, UHC service coverage index. Values for these variables are obtained from the World Bank. We also control for countries’ health security ratings, Healthsecurity. It is the 2019 Global Health Security (GHS) Index that rates countries’ health security conditions. This information is obtained from the Economist Intelligence Unit and the Johns Hopkins Center for Health Security. We form an indicator for the presence of direct flights to Wuhan, the epicentre of the initial outbreak of the virus, Directflight_WH, obtained from the Wuhan International Airport. We control for countries’ past experience with SARS (Severe Acute Respiratory Syndrome) in 2003, which can enhance the alertness of a nation when encountering future pandemics. SARS_experience is an indicator that equals 1 for countries with at least one confirmed SARS case in 2003, and 0 otherwise.

We incorporate COVID-19 pandemic-related variables that can directly affect the reported infection and death cases. We control for virus testing, as confirmed cases are critically reliant on testing. Without controlling for testing, a bias can emerge in estimation if testing is systematically associated with individualism. For a subset of countries in our sample, we are able to obtain their virus testing data from Our World in Data. We define Test as the logarithm transformation of one plus the cumulative daily number of tests performed in every 10,000 people of a country’s population. We form an indicator Test_missing that equals 1 for observations missing testing information on a day for a country, and 0 otherwise. When Test_missing equals 1, Test is set to 0. We also control for vaccination that potentially protects citizens from infections, Vaccination, which is the logarithm transformation of one plus the cumulative number of COVID-19 vaccinations over its population (in every 10,000 people). This item is from Our World in Data. For observations with missing vaccination information, Vaccination is set to 0. We also form an indicator Vaccination_missing that equals 1 for missing vaccination information on a day for a country, and 0 otherwise. We account for country-level time trend by controlling for the relative timing of countries’ first confirmed COVID-19 case. Relativeday is the number of days from the day when the first COVID-19 case was confirmed in a country to the day of observation. We also include its squared term, Relativeday2. Supplementary Table 1 lists all variable definitions and data sources for the international analysis. We include continent indicators. Supplemental Table 2 shows descriptive statistics for all variables.

### Instrumental variable estimation

To alleviate the concern for the omitted variable bias and strengthen causal inference, we adopt an instrumental variable approach. We use a country’s cumulative number of Nobel Prize winners as of 2019 as an instrument for individualism. Nobel Prizes are awarded to people based on intellectual and academic achievements. The cumulative number of Nobel Prize winners therefore should be exogenous to the pandemic that started in 2020. Before the outbreak of a pandemic, it would be difficult to predict the outcome and severity of the pandemic based on the cumulative number of Nobel Prize winners. This instrument likely satisfies the exclusion condition.

We estimate the following second stage instrument variable estimation model focusing on Individualism as follows:

$$\begin{array}{l}{\rm{Confirmed}}\,{\rm{Cases}}_{i,t}/{\rm{Deaths}}_{i,t}/{\rm{Death}}\,Ratio_{i,t} \\ = \beta _0 + \beta _1{\rm{Individualism}}_i+\, \beta _2{\rm{Vaccination}}_{i,t} + \beta _3{\rm{Vaccination}}\_{\rm{missing}}_{i,t}\\ \quad\,\,+\, \beta _4{\rm{Test}}_{i,t} + \beta _5{\rm{Test}}\_{\rm{missing}}_{i,t}\\ \quad\,\,+\, \beta _6{\rm{Relativeday}}_{i,t} + \beta _7{\rm{Relativeday}}_{i,t}^2\\ \quad\,\,+\, {\rm{National}}\_{\rm{controls}}_i + {\rm{Continent}}\,{\rm{Fixed}}\,{\rm{Effects}} + \varepsilon _{i,t},\end{array}$$
(1)

where i indexes countries and t indexes calendar days for our test. t-statistics are computed using heteroscedasticity-consistent standard errors.

Table 1 presents the instrumental variable estimation for confirmed cases, deaths and death ratio. In the first stage, which incorporates all the controls in the second stage (Panel A), the number of Nobel Prizes Nobel_Prize is significantly and positively associated with Individualism on all four dates (April 15, 2020; October 15, 2020; April 15, 2021 and October 15, 2021). Therefore, this instrumental variable satisfies the relevance condition. It also marginally passes the instrument strength test.

In the second stage (Panel B), on all four dates (April 15, 2020; October 15, 2020, April 15, 2021 and October 15 2021), the instrumented Individualism is positively and significantly associated with confirmed cases. Panel C presents instrumental variable estimation for deaths. Again, on all four dates (April 15, 2020; October 15, 2020; April 15 2021 and October 15 2021), the instrumented Individualism is positively and significantly associated with deaths. The results for death ratio are presented in Panel D. The coefficients on instrumented Individualism are insignificant though still positive.

Results on individualism from daily instrumental variable regressions from January 1, 2020 to December 31, 2021 are depicted in Fig. 1. For confirmed cases, the daily coefficients on instrumented Individualism are positive and significant almost throughout. The daily coefficients are also positive and significant for deaths. As for the death ratio, the daily coefficients on Individualism are largely insignificant, though still positive. Based on these results, individualism appears to be associated with a more severe COVID-19 situation.

Nobel Prizes won by a country’s nationals include physics, chemistry, medicine and economics prizes. These prizes indicate a country’s achievements in science and technologies, and thus can potentially directly influence its ability in handling COVID-19. Even though we believe that the total number of Nobel Prize winners is a good instrument as we expect it to influence the severity of COVID-19 only through affecting our measure of individualism, the inclusion of physics, chemistry, medicine and economics prizes can potentially make this instrument a bit endogenous. To alleviate this concern, we exclude physics, chemistry, medicine and economics prize (science Prizes) winners and use a country’s cumulative number of Nobel Prize winners in literature and peace as of 2019, Nobel_Prize_LP, as an instrument. These two categories of Nobel Prizes are likely not related to a country’s achievements in science and technologies and therefore this instrument should be exogenous. The correlation coefficient between science Prizes and the Literature Prize is 0.7308 and that between science Prizes and the Peace Prize is 0.8416. Results (untabulated) are similar to those using the total Nobel Prize winners as an instrument.

Overall, we find a strong and positive individualism-pandemic severity link. With the instrumental variable approach, causality likely goes from individualism to the severity in COVID-19.

### Social distancing and mobility restriction policies

To investigate the channels through which individualism hampers an effective containment of COVID-19 pandemic, we first consider a direct and important means which is government initiated non-pharmaceutical interventions, such as quarantine, social distancing and in general, mandatory mobility restrictions. Recent research has shown that lockdowns and travel restrictions are effective in containing the COVID-19 pandemic (Cyranoski 2020; Chinazzi et al. 2020; Tian et al. 2020). However, these government-initiated restrictive measures compromise personal freedom that is more cherished in an individualistic culture. Indeed, Conway et al. (2006) provide evidence that national cultural collectivism, as opposed to cultural individualism, is positively associated with an index of political restriction across nations. Therefore, governments in countries scoring high on individualism may be more hesitant or slower in imposing mobility restrictions.

We consider eight social distancing and mobility restriction policies: (1) school closing, (2) workplace closing, (3) cancelling public events, (4) restrictions on gatherings, (5) transport closure, (6) stay-at-home requirement, (7) restrictions on internal move, and (8) international travel controls. We perform longitudinal analysis covering the whole period of years 2020 and 2021.

We estimate the following ordered logit regression to determine how individualism affects the extent of adoptions of these restrictive policies across nations:

$$\begin{array}{l}{\rm{Policy}}_{i,t} = \beta _0 + \beta _1{\rm{Individualism}}_i + \beta _2{\rm{Confirmed}}\,{\rm{Cases}}_{i,t} + \beta _3{\rm{Vaccination}}_{i,t}\\ \qquad\qquad\,\,+\, \beta _4{\rm{Vaccination}}\_{\rm{missing}}_{i,t} + \beta _5{\rm{Test}}_{i,t} + \beta _6{\rm{Test}}\_{\rm{missing}}_{i,t}\\ \qquad\qquad\,\,+\, {\rm{National}}\_{\rm{controls}}_i + {\rm{Continent}}\,{\rm{Fixed}}\,{\rm{Effects}} + \varepsilon _{i,t},\end{array}$$
(2)

where Policy refers to one of the eight social distancing and mobility restriction policies (defined in Supplemental Table 1). It is a country-day variable recorded on an ordinal scale that increases in the level of strictness of a particular policy. We control for Confirmed Cases at country-day level to account for its effect on policy adoption that arises from the severity of local pandemic, apart from the individualistic culture. We also control for Vaccination at a country-day level due to complementarity among different policies (for example, governments may opt to ease non-pharmaceutical interventions when a country achieves a high rate of vaccination). Other variables are as defined earlier.

Daily regression coefficients on policy adoption/stringency, from January 1, 2020 to December 31, 2021, on Individualism are visualized in Fig. 2. A general impression is that these daily coefficients are largely insignificant with the “0” line usually within the 90th or 95th percentile confidence intervals. Further, based on the figures, it is hard to make out any particular time trend. Therefore, in terms of policies adoption during the prolonged pandemic, individualism does not appear to systematically affect the stringency and timing of these policies adoptions (except for brief period in early 2020 when individualism appears to slow down the workplace closing policy). Overall, this finding is different from Sebhatu et al. (2020) who analyze the adoption of non-pharmaceutical interventions during the early phase of COVID-19 among OECD countries from February to April 2020 and find that strong democracies (which are more individualistic) are slower to initiate restrictive policies but are more likely to follow/mimic the policies of nearby countries. Our study extends beyond OECD countries covering a more comprehensive period and finds that non-pharmaceutical policies largely diffuse (Shipan and Volden 2008; Simmons and Elkins 2004; Brinks and Coppedge 2006; Katz et al. 1963) and converge between individualistic and collectivistic countries.

#### Policy effectiveness

While both individualistic and collectivistic countries, potentially through policy diffusion, adopt non-pharmaceutical interventions in containing the pandemic, an individualistic culture can however render these policies less effective in fighting COVID-19 through social non-cooperativeness. To test this possibility, we estimate the following model:

$$\begin{array}{l}{\rm{Confirmed}}\,{\rm{Cases}}_{i,t}/{\rm{Deaths}}_{i,t}/{\rm{Death}}{{{\mathrm{ }}}}{\rm{Ratio}}_{i,t} \\= \beta _0 + \beta _1{\rm{LowIndividualism}}_i \cdot {\rm{Policy}}_{i,t}\\ \quad+\, \beta _2{\rm{Policy}}_{i,t} + \beta _3{\rm{Vaccination}}_{i,t} + \beta _4{\rm{Vaccination}}\_{\rm{missing}}_{i,t}\\ \quad+\, \beta _5{\rm{Test}}_{i,t} + \beta _6{\rm{Test}}\_{\rm{missing}}_{i,t}\\ \quad+\, \beta _7{\rm{Relativeday}}_{i,t} + \beta _8{\rm{Relativeday}}^2_{i,t}\\ \quad+\, {\rm{Country}}\,{\rm{Fixed}}\,{\rm{Effects}} + \varepsilon _{i,t},\end{array}$$
(3)

where LowIndividualism equals 1 for a country whose value for Individualism is lower than the sample median, and 0 otherwise. Our analysis covers the period between January 1, 2020 and December 31, 2021.

As the value of Policy increases when a certain policy becomes more stringent, it captures the intensity of non-pharmaceutical interventions at a country-day level. We expect the coefficients on Policy to be negative if the adoptions of social distancing and mobility restrictions are effective in fighting COVID-19. While the adoptions of these eight policies are staggered at the country-day level, these adoptions are likely not exogenous to countries in our sample due to policy responses to the severity of the virus situation and policy diffusion among countries (Sebhatu et al. 2020). Nevertheless, the regression result can still shed some light on how the effects of Policy on Confirmed Cases, Deaths and Death Ratio vary as a function of Individualism.

In this test, our focus is on LowIndividualism ∙ Policy. People in countries scoring low on individualism are more likely to adhere to restrictive policies than people in countries scoring high on individualism, rendering these policies more effective in low individualism countries. The coefficient on LowIndividualism ∙ Policy should be negative.

Results are reported in Table 2. The coefficients on Policy alone are positive in most cases in Panels A (confirmed cases) and B (deaths). It appears that social distancing and mobility restriction policies are not effective in tackling COVID-19 in high individualism countries. On the other hand, for confirmed cases (Panel A), the coefficients on LowIndividualism ∙ Policy are negative and significant for all eight policies. For deaths (Panel B), the coefficients on LowIndividualism ∙ Policy are negative and significant for six policies (except for school closing and international travel control). Therefore, government-initiated policies, such as social distancing and mobility restriction, are only effective in reducing infections and deaths in low individualism countries. Askitas et al. (2021) provide evidence that cancelling public events, imposing restrictions on private gatherings and closing schools and workplaces had a significant effect on reducing COVID-19 infections. We, using a much longer sample period, show that while various virus-fighting polices are generally ineffective in reducing COVID-19 infections and deaths in countries scoring high on individualism, it is effective in this regard in countries scoring low on individualism. In sum, we point to individualism hampering the fight against COVID-19 due to less policy adherence from the mass.

The results for death ratio (Panel C) are significant in only two cases (school closing and restrictions on internal move), however, the coefficients on LowIndividualism ∙ Policy are all in the right negative direction. Overall, Table 2 provides evidence that an individualistic culture reduces the effectiveness of social distancing and mobility restriction policies in alleviating the virus situation.

## Former East and West Germany as a natural experiment

### Communism, collectivism, and acceptance of government-initiated restrictions

Different economic systems imply different emphasis on individualist or collectivistic ideologies (Rose 1991). At a conceptual level, communism is a collectivistic approach to state governance (Singelis et al. 1995). The author of The Communist Manifesto, Karl Marx, in his early writing, criticizes the notion of natural rights found in the “Declaration of the Rights of Man” (1791) from the French Revolution as reflecting only the egoistic part of human nature, without acknowledging the community-oriented part of human nature. As a political system, a communist regime can cause a shift towards more collectivistic cultural values from the top down, such as through value inculcation by workplace organizations, by political education and through the control of media by the authorities (Wallace, 1997). Schwartz and Bardi (1997) consider the emergence of communism in Eastern Europe as a natural experiment and demonstrate that Eastern and Western Europe, at the fall of communism in Eastern Europe, differed in their orientations toward values related to individualism/collectivism, in that Eastern Europeans scored higher in collectivistic values (conservatism and hierarchy) while Western Europeans scored higher on individualistic values (affective and intellectual autonomy). They note that, as an important premise of their study that Eastern and Western Europe share a great deal of history prior to the communist revolution. An implication of their finding a higher collectivistic value in the hierarchy is that collectivistic-oriented people can display a higher degree of adherence to political hierarchy and are more willing to accept government-initiated norms and restrictions oriented from in-group welfare. Specifically, on the attitude towards government-initiated economic norms, Corneo and Gruuner (2002) find that in 1992 Eastern Europeans had stronger preferences, than Western Europeans, for redistribution. Alesina and Fuchs-Schündeln (2007) sharpen the causal identification by focusing on one country, German, and find that the forty-five years of communist rule has shaped Eastern German’s preference for government redistribution.

### Tolerance for intrusive government policies in former East and West German districts

Enlightened by the above line of literature, it is meaningful to extend our research to the setting of a communist regime history in Germany as a manipulation on the individualistic/collectivistic value orientation and examine its impact on people’s attitude/tastes for public social policies beyond economic redistribution, such as the acceptance/adherence to government-initiated restrictions in the context of COVID-19.

Starting from 1945 after the end of World War II, Germany was divided into two states with distinct socio-political systems, with West Germany being a Western-style democracy and East Germany being a Marxist–Leninist republic (Fuchs-Schündeln and Schündeln, 2009). While West Germany implemented a capitalist economic system that was closely integrated into the Western world and favoured social and economic freedom, East Germany, under the heavy interference of a powerful government, implemented a socialist economic system that favoured executive orders and collectivism (Alesina and Fuchs-Schündeln, 2007). The two Germany were reunited in 1990.

On the one hand, reunified Germany inherited West Germany’s economic system, political principles and military alliance. National-level policies dealing with COVID-19 are uniform across former East and West German districts, helping us mitigate potential endogeneity associated with latent factors in an international setting, such as heterogeneity in virus-fighting policies, testing capacity and reporting transparency.

On the other hand, the two Germany have persistent differences in ideology and culture, such as gender equality, trust, social behaviour, religion, and preference for redistributive policies, even decades after their reunification (Rohrschneider 1994; Rohrschneider and Schmitt-Beck 2002; Rainer and Siedler 2009; Fisher 2010; Brosig-Koch et al. 2011; Kuhn 2012; Stolz et al. 2020). Alesina and Fuchs-Schündeln (2007) show that people in the former East Germany are more pro-collectivism than those in the former West Germany, as reflected in their more willing acceptance of government-initiated economic norms of state redistribution. Sack (2017) documents that people who grew up in East Germany show a higher affiliation with a socialist model of democracy than people in the West who show higher support for a liberal model, even twenty years after reunification. More relevantly, in a COVID-19 context, Schmelz (2021) surveys respondents towards the end of the first lockdown in Germany and finds that control aversion, or negative responses to government-initiated enforcement to contain the pandemic, is less common among those who were brought up under the socialist regime of East Germany. Eastern German’s collectivistic inclination or norm acceptance arises from forty-five years of socialist education and dependence on government redistribution for the older generations (Gläßel and Paula 2020; Corneo 2004). Intergenerational transmission of political values further serves to prolong and preserve the impact of former socialist rules on the young generations (Jennings and Niemi 1968; Jennings et al. 2009). In other words, even thirty years after the reunification, former East Germans still differ from West Germans in that they are more willing to be guided by norms (if socially approved) and are used to an intrusive public sector or government. Therefore, on the cultural dimension, people from former East German districts are more likely to adhere to government virus-fighting policies than those from former West German districts, thus alleviating the severity of COVID-19 in former East German districts.

Of course, the German setting is not without issues. Distance from German reunification can somehow mitigate East Germans’ penchant for and acceptance of government interventions in social and economic life, as the East and the West gradually converge in attitudes towards government responsibilities (Svallfors 2010). Migration within Germany can also mitigate this difference in attitudes towards the government and its policies (Fuchs-Schündeln and Schündeln 2009). These factors will serve to bias against us finding a significant effect of old ideology on COVID-19. However, it is also important to note that thirty years may not be long enough to erase that penchant. Numerous studies have documented long-lasting effect of socio-political events on local cultures and norms (Acemoglu et al. 2001; Alesina and Glaeser 2004; Nunn and Wantchekon, 2011; Erikson and Stoker, 2011). Ultimately, it is an empirical question whether the pro-collectivism culture among the East Germans can significantly impact the severity of COVID-19.

### Variables and statistics

We track daily virus situations in 399 German districts (districts and district free cities), not including Berlin. Berlin was within former East German territory but divided into East and West Berlins between 1945 and 1990. People’s daily movement within Berlin would make it difficult for us to detect differences in attitudes towards the government and its policies between former East and West Berlin. The number of districts excluding Berlin went from 400 to 399 as two districts merged in 2021.

The numbers of confirmed cases and deaths are collected by Robert Koch Institute (RKI) and can be obtained from Kaggle (COVID-19 Tracking Germany). We define three dependent variables. We define Confirmed Cases as the logarithm transformation of one plus the cumulative daily number of confirmed cases in every 10,000 people of a district’s population. We define Deaths as the logarithm transformation of one plus the cumulative daily number of deaths in every 10,000 people of a district’s population. We also define a death ratio variable Death Ratio which is the number of deaths divided by the number of confirmed cases (in percentage format). Definitions of these variables and subsequent variables are consolidated in Supplemental Table 3.

We define an indicator East that equals one if a district formerly belonged to East Germany, and zero otherwise. We then compute the distance between the geographic centre of a district and the former East and West German border, Distance. This variable is calculated based on information from Germany’s Federal Agency for Cartography and Geodesy. We include Relativeday which is the number of days from the day when the first COVID-19 case was confirmed in a district to the day of observation, and its squared term, Relativeday2. In addition, we obtain district-specific variables. Land_area is the logarithm transformation of total land area (square kilometres) of a district. Population_age65 is the percentage of population above age 65 of a district. GRP_growth is the growth in nominal gross regional product of the state that a district belongs to. Export is the exports of goods and services as a percentage of gross regional product of the state that a district belongs to. Healthcare is the per capita gross expenditure of public assistance towards healthcare of the state that a district belongs to. Migration is the total migration arrivals across the borders of the state that a district belongs to, over its population (in every 100 people). Serviceperson is the number of persons in employment in service activities of the state that a district belongs to, over its population (in every 100 people). Student is the number of students of the state that a district belongs to, over its population (in every 100 people). Educationexp is higher education expenditure of the state that a district belongs to, over its population. Cinema is the number of cinema visits per inhabitant of the state that a district belongs to. Residentialbuilding is the number of residential buildings of the state that a district belongs to, over its population. Passenger in the number of passengers carried by bus, coach and rail of the state that a district belongs to, over its population. Accommodation is the number of arrivals in accommodation establishments of the state that a district belongs to, over its population. Rail is the track length of the rail network (km) of the state that a district belongs to, over its population (in every 10,000 people). Police is the number of police personnel (full-time equivalents) in the core budgets of the state that a district belongs to, over its population (in every 100 people). Inputs for these variables are obtained from Statistisches Bundesamt. Vaccination information is collected by Robert Koch Institute (RKI) and can be obtained from GitHub (COVID-19 Impfungen in Deutschland). Vaccination is the cumulative number of people fully vaccinated against COVID-19 over its population (in every 100 people) in a district. Descriptive statistics related to these variables are presented in Supplemental Table 4.

### Conflicting pattern in COVID-19 severity between East and West German districts

We have argued that people in East German districts, formerly belonging to a socialist country, are more likely to share a collectivistic culture. Therefore, to be consistent with our prediction, the COVID-19 situation should be less severe in former East German districts.

When we compare COVID-19 situations in East and West German districts without considering districts differences and the extent of vaccination, the results are not consistent with our expectation. Specifically, we regress Confirmed Cases, Deaths and Death Ratio on the indicator East in daily regressions from January 1, 2020 to December 31, 2021 and depict the daily coefficients on East in Fig. 3. The COVID-19 situation was less severe in former East German districts initially. This result for the first wave of the pandemic (before the implementation of the vaccination) is consistent with Schmelz (2021) who demonstrates through surveys that COVID-19 related control aversion is less common among Germans who were brought up under the socialist regime of East Germany. However, after the end of October 2020, the trend started to change with former East German districts having a more severe COVID-19 situation in 2021 (Fig. 4).

The pattern remains the same if we only focus on East and West German districts on the former border or within 50/100/150 kilometres of the former border (Fig. 5). Panel A of Table 3 and the left-hand result of each column in Table 4 also show that, without considering district characteristics, on April 15 and October 15 in 2020, the coefficients on East are almost all negative and significant for Confirmed Cases, Deaths and Death Ratio, consistent with our expectation. However, they all become positive and significant on April 15th and October 15th in 2021. Therefore, without considering district characteristics, we cannot conclude from the results here that former East German districts have a less severe COVID-19 situation. In order to identify a cultural effect on pandemic outcomes, it is necessary to consider various other factors, such as patterns of economic activities, migration, local demography, transportation infrastructure, law enforcement effectiveness that can jointly determine the severity of the pandemic.

### Controlling for district differences

Differences in district characteristics and regional differences in the extent of vaccination can potentially contribute to the above pattern. We estimate the following regression model to determine whether former East German districts have a less severe COVID-19 situation than former West German districts when we control for district characteristics and the extent of vaccination:

$$\begin{array}{l}{\rm{Confirmed}}\,{\rm{Cases}}_{i,t}/{\rm{Deaths}}_{i,t}/{\rm{Death}}\,{\rm{Ratio}}_{i,t}\\ = \beta _0 + \beta _1{\rm{East}}_i + \beta _2{\rm{Vaccination}}_{i,t}\\ \quad+\, \beta _3{\rm{Relativeday}}_{i,t} + \beta _4{\rm{Relativeday}}_{i,t}^2 + {\rm{District}}\_{\rm{controls}}_i + \varepsilon _{i,t},\end{array}$$
(4)

where i indexes districts and t indexes calendar days. t-statistics are computed using heteroscedasticity-consistent standard errors. East is the variable of interest. We control for district-level characteristics in demography, economic development, and healthcare infrastructure. Districts differ in their virus trajectory as some have their first virus infection cases earlier than the others. We thus control for the relative days to the first confirmed COVID-19 case in a district, as well as its squared term.

The introduction of vaccination in combatting the subsequent waves of COVID-19 variants alters the dynamics and complementarity among different mobility restriction policies and non-pharmaceutical interventions, potentially making the stringency of mobility restrictions critically dependent on the vaccination rate/status. Further, as Germany is at the trajectory of election and the change of the governing party, vaccine scepticism was being deliberately instrumentalized by right-wing groups especially in the East. Figure 6 shows that the rate of vaccination appears to be lower in former East German districts in 2021, contradicting Schmelz (2021) survey results prior to the launch of the vaccination, that Germans with prior experience with the former socialist regime report low control-aversive attitude, including vaccination resistance. The low actual vaccination rate for COVID-19 in the eastern district also contradicts the country’s recorded vaccination history with East Germans usually recording a high vaccination rate under enforcement regimes. For this reason, we also consider vaccination for observations in 2021.

The results are reported in Table 4. There are a few interesting and important findings. First, without controlling for district characteristics and local pandemic related variables, while the coefficients on East are negative and significant on April 15th and October 15th in 2020, they are positive and significant on April 15 and October 15 in 2021, appearing to suggest that the trend changed direction in 2021. However, with controls for district characteristics and pandemic-related variables, they are negative and significant on all four dates, continuing to suggest that former East German districts have a less severe COVID-19 situation. This is true for Confirmed Cases (Panel A), Deaths (Panel B) and Death Ratio (Panel C). Second, by comparing the magnitude of the coefficients with and without controls, for examples, on April 15, 2020, we can see that the coefficient on East with controls is much greater than that without controls. Therefore, the coefficient on East is biased upward when district controls are not considered. This would explain why the coefficients on East without control can even be positive in 2021. Overall, when we consider differences in districts characteristics and the extent of vaccination between former East and West German districts, we still arrive at the same conclusion that East German districts have a less severe COVID-19 situation. Figure 7 also shows that with controls, daily regression coefficients are all negative and significant, especially for Confirmed Cases and Deaths.

What are the differences? Numbers in Table 3 offers some clues. Former East and West districts appear to differ significantly on all dimensions. For example, East German districts have more people above 65, lower economic growth, lower export, lower expenditure on healthcare, lower educational expenditure but higher rail road coverage and police presence, etc. Further, most of these district characteristics are significant in Table 4. Absence of considerations for these district characteristics can contribute to biasing the coefficients on East upward.

## Conclusion

We demonstrate that an individualistic culture can impede the containment of COVID-19. Using the number of Nobel-Prize winners to instrument individualism, we show that countries scoring high on individualism generally have a more severe COVID-19 situation. This link between individualism and the severity of the virus situation is likely due to individualism reducing the effectiveness of social distancing and mobility restriction policies aimed at mitigating the virus situation, and not due to these policies being less likely to be implemented in countries scoring high on individualism. Exploiting the setting that people from former socialist East German districts are likely more collectivistic than those from former West German districts, we find that, after controlling for district characteristics and vaccination, former East German districts have a less severely COVID-19 situation than former West German districts.

## Contributions and discussions

We contribute to our understanding of individualism by showing how it potentially impedes the containment of COVID-19. While restrictive non-pharmaceutical interventions largely converge among countries of diverse cultural orientation in the prolonged fight against COVID-19, a greater reluctance among people in more individualistic cultures to heed virus-fighting policies can erode the effectiveness of these policies and impose a negative public health externality during COVID-19.

We add to a growing literature on COVID-19 by examining the role of culture in underlying societies’ efficacy in pandemic containment. Among other social level factors, such as gender (Galasso et al. 2020), age structure (Dowd et al. 2020), trust (Bargain and Aminjonov, 2020) and partisanship (Gollwitzer et al. 2020; Clinton et al. 2021), that impact pandemic related attitudes, behaviours and health outcomes, cultural orientation in the paradigm of individualism-collectivism has received substantial academic attention. While most of the studies have focused on how individualism-collectivism impact behavioural responses, such as mobility reduction, extent of social distancing, and mask usage (Chen et al. 2021; Bazzi et al. 2020; Lu et al. 2021), we demonstrate health outcomes resulting from such a cultural orientation through social uncooperativeness that limits the efficacy of non-pharmaceutical interventions.

As our study covers a most comprehensive pandemic trajectory (entire years 2020 and 2021) as compared with prior studies, we demonstrate the persistency of a cultural effect on pandemic containment throughout the prolonged pandemic trajectory involving multiple virus variants and different policy foci (such as the introduction and the emphasis on vaccination in the later period). We are able to verify or falsify claims made in earlier studies based on data from a limited duration. For example, Maaravi et al. (2021) show a positive association between individualism and the severity of COVID-19 in an international setting, but only using data retrieved in April 2020. Bazzi et al. (2020) show that, in the U.S. setting, frontier “rugged individualism” is associated with less social distancing and mask use as well as weaker local government effort to control COVID-19. However, their analysis is based on data collected by August 2020. We cover data from January 2020 to December 2021 and show that, in an international setting, individualism hampers the fight against COVID-19. Further, Askitas et al. (2021) provide evidence that cancelling public events, imposing restrictions on private gatherings and closing schools and workplaces had significant effects on reducing COVID-19 infections. We, using a much longer sample period, show that while various virus-fighting policies are generally ineffective in reducing COVID-19 infections and deaths in countries scoring high on individualism, it is effective in this regard in countries scoring low on individualism. Sebhatu et al. (2020) who analyze the adoption of non-pharmaceutical interventions during the early phase of COVID-19 from February to April 2020 find that strong democracies (which are more individualistic) are slower to initiate restrictive policies but are more likely to follow/mimic the policies of nearby countries. We show that non-pharmaceutical policies largely diffuse and converge between individualistic and collectivistic countries in a longer term. In sum, we provide evidence that individualism persistently hampers the fight against COVID-19 due to less policy adherence from the mass.

A longer period analysis of the case of Germany is especially important and yield interesting results. Schmelz’s (2021) survey analysis shows that towards the end of the first lockdown in Germany, control aversion, or negative responses to government-initiated enforcement to contain the pandemic is less common among those who were brought up under the socialist regime of East Germany. We show that without considering German district characteristics, COVID-19 was generally less severe in former East German districts in 2020. This pattern changed direction in late 2020 and early 2021, with East German districts having a more severe COVID-19 situation in 2021. However, after we control for districts characteristics and the extent of vaccination, former East German districts had a less severe COVID-19 situation throughout 2020 and 2021. Again, we conclude that individualism affects the equilibrium outcomes of the fight against COVID-19.

In no way are we taunting or belittling the ideas behind individualism with our research. However, in the presence of an unprecedented pandemic, restrictions on personal freedom are often an effective and potentially indispensable weapon against the spread of COVID-19. As people’s personal value for individualism inevitably differs from each other, behaviours of those who oppose mobility restrictions will impose a negative health externality on others and the society as a whole, eroding the effectiveness of policies designed to contain the pandemic.