The COVID-19 curtain: Can past communist regimes explain the vaccination divide in Europe?

Abstract

As of December 2021, all former Communist countries from Central and Eastern Europe were still lagging behind in terms of COVID-19 vaccination rates in Europe. Can institutional legacy explain, at least in part, this heterogeneity in vaccination decisions across Europe? To study this question we exploit novel data from the second wave of the SHARE (Survey of Health, Ageing and Retirement in Europe) COVID-19 Survey fielded in Summer (2021) that covers older individuals in 27 European countries. First, we document lower COVID-19 vaccine take-up amongst those who were born under Communism in Europe. Next, we turn to reunified Germany to get closer to a causal effect of having lived behind the Iron Curtain. We find that exposure to the Communist regime in East Germany decreased one's probability to get vaccinated against COVID-19 by 8 percentage points and increased that of refusing the vaccine by 4 percentage points. Both effects are large and statistically significant, and they hold when controlling for individual socio-economic and demographic characteristics. We explore several possible mechanisms. The East-West Germany gap does not seem to be explained by differences in the impact of the first wave of the pandemic or in general exposure to vaccines. We find that East Germans have lower social capital than West Germans and that social capital correlates negatively with Covid-10 vaccine uptake, but only a small fraction of the East-West Germany Covid-19 vaccination gap can be explained by our measures of social capital.

1. Introduction

As of today, vaccination is still the most efficient protection from the COVID-19, and vaccine hesitancy the main barrier to achieving the sought-after herd immunization (WHO, 2020). Already in pre-COVID-19 time, reluctance to take vaccines was listed as one of the Top Threats to Global Health by the World Health Organization (WHO, 2019). Although the COVID-19 vaccine has become available in all European Union (EU) countries at no cost for residents, some countries have been lagging behind in vaccination rates. To this day, all former Communist countries from Central and Eastern Europe exhibit lower vaccination rates than Western European countries, occupying the last ten positions in terms of vaccination rates (all ages) in Europe (ECDC Vaccine Tracker on December 10th, 2021a, ECDC, 2021b). Can institutional inheritance explain, at least in part, the heterogeneity in vaccination decisions across Europe?

We use novel data from wave 2 COVID-19 of the Survey of Health, Ageing and Retirement in Europe (SHARE) that was conducted in Summer (2021) across 27 European countries, along with other features of the SHARE data (including retrospective life histories and early-life circumstances) to answer this question. According to our estimates from SHARE, the nine countries with the lowest vaccination rates against COVID-19 are so-called “Post-Communist” countries. In Romania and Bulgaria only 29 and 21% of the 50+ population had received their first jab by mid-July 2021 (see Fig. 1 ). Besides, and this will be key to our empirical strategy, former German Democratic Republic (East Germany) and Federal Republic of Germany (West Germany) exhibit important differences in terms of vaccination against Covid-19, East Germans being more similar to Post-Communist countries, West Germans to Western countries.

Fig. 1.

Uptake of Vaccine against COVID-19 across Country of Residence among 50+ in Summer 2021. Note: Wave 2 COVID-19 of SHARE. Authors' own calculations (weighted data).

To our knowledge, this is the first paper to uncover a causal relationship between exposure to past Communist regimes and vaccination decision against COVID-19 (instead of intentions or attitudes towards vaccination, which might differ from decisions, as put forward by Dai et al., 2021; Becher et al., 2022). Assessing causality is a challenge, as countries separated by the Iron Curtain differed before they were separated after the end of the Second World War. Exploiting the richness of our data, we can discard many potential drivers of vaccine (non-)uptake, such as lower education, better (or worse) health, lower socioeconomic status, or unfavorable early-life conditions. To help us overcome the causality challenge, we exploit the quasi-natural experiment provided by the separation and later reunification of East and West Germany.

Under the assumption that any systematic difference between vaccination preferences across the old East-West German divide are related to different political institutions, we are able to measure how 45 years of Communism (and the transition period that followed) impacted individuals' decision to get vaccinated in pandemic times. How could the Communist era affect vaccination in East Germany compared with West Germany? On the one hand, one could expect individuals who experienced a collectivist culture, which put greater emphasis on common goals than on individual pursuits, to prioritize the health and survival of others, over potential personal safety concerns related to the new vaccine. Communist regimes had also implemented a longstanding tradition of mandatory vaccination, which could lead East Germans to accept vaccination more easily, as they were used to it. On the other hand, there could be a backlash effect of the past Communist regime on individuals’ preferences in favor of not getting the COVID-19 vaccine, if those who were exposed to an authoritarian government are more averse to any kind of State control or public recommendations that might remind them of past regime propaganda, including national and supranational injunctions to get vaccinated.

We find strong evidence in favor of the latter hypothesis. We find lower COVID-19 vaccine take-up amongst individuals born under a Communist regime in Europe, which holds controlling for individual pre-pandemic health, socioeconomic characteristics, and current country of residence. In the reunified Germany setting, we consistently find that individuals who were born in East Germany were less likely to get vaccinated against COVID-19 and more likely to refuse the vaccine compared to their Western counterparts. Likewise, when looking at other preventive measures such as social distancing, we also document that Eastern Germans complied less than Western Germans. Nonetheless, the backlash effect is not apparent for the vaccine against influenza, which has been routinely inoculated to Germans for decades; its prevalence was higher in East Germany.

Using external data sources about vaccination campaigns and the day of the interview of the SHARE Corona Survey, we are able to discard vaccine supply as a potential driver of our results. After ruling out several other mechanisms such as a differential impact of the pandemic or general vaccine skepticism, in East compared to West Germany, we show that East Germans exhibit lower levels of social capital, measured as voluntary or charity work, political engagement, and trust in people (following Putnam, 1993; Glaeser et al., 2000; Conzo and Salustri, 2019). We find that some aspects of social capital, in turn, are positively associated with COVID-19 vaccination. Although we find supportive evidence of a role for our measure of social capital in lower vaccination rates in East Germany, most of the East-West Germany COVID-19 vaccination gap remains unexplained.

Our results are subject to the following caveat regarding their interpretation: when finding significant differences between East and West Germany, we cannot distinguish whether they are the result of individuals' exposure to decades of a Communist regime or to the fall of Communism. We therefore explore the consequences of both the rise and fall of Communism on individuals’ vaccination outcomes.

Our paper is related to a large literature about the long-lasting causal impact of institutions on individuals' preferences and choices. In particular, we relate to a strand of the literature about the role of past Communist institutions on individuals' preferences. Starting with a seminal paper by Alesina and Fuchs-Schündeln (2007), many looked at the German division as a natural experiment (see Becker et al. 2020 for a review of that literature). However, Becker et al. (2020) argue that the comparison between East and West Germany should not always be interpreted as a natural experiment because of differences which predate Communism, in terms of political preferences, women's labor force participation, and religion. Still, we believe that the nature of our research question (focused on vaccination outcomes) and the richness of our set of covariates allow us to alleviate these concerns. We contribute to this literature by investigating the long-lasting impact of the past Communist regime on health-related decision-making.

Our paper is closer –in terms of the research question– to Costa-Font et al. (2021), who study the association between Communism and vaccination. Using a pre-COVID-19 Survey from 2018, they show that individuals from post-Communist countries have lower vaccination trust primarily due to lower governmental trust (rather than interpersonal trust). We build on this paper by going beyond associations: after providing suggestive evidence supporting a role of the Communist regimes in lower vaccination take-up in Post-Communist countries, we dive into the German setting to get closer to causality in estimating the impact of the exposure to Communism on COVID-19 vaccination. Moreover, the richness of our data allows us to rule out the role of other socio-demographic factors.

As for other determinants of COVID-19 vaccination that are not limited to institutions, using the same SHARE Corona Survey, Bergmann, Hannemann, Bethmann, and Schumacher (2021) show the role of socio-demographic variables for vaccine uptake among the elderly, while the other studies we know of explore attitudes about vaccines (see Galasso et al., 2020; Lazarus et al., 2021; Karlsson et al., 2021). We add to this literature by looking at vaccine uptake instead of vaccination attitudes, i.e. at revealed rather than stated preferences, which might differ for COVID-19 vaccination (see Dai et al., 2021; Becher et al., 2022). Besides, we are able to look separately at two different yet commonly confounded concepts –vaccine hesitancy and vaccine refusal–which are of particular importance for policy makers to design vaccination campaigns. Finally, we compare individuals’ vaccination decisions between the newly existing COVID-19 vaccine and two other long-standing vaccines, against influenza or pneumonia, which most studies cannot do.

Last, our paper is part of a series of studies on to what extent individuals adopt “good practice” behavior to protect each other from the virus, or comply with the set of preventive measures imposed upon them by public health authorities. Durante et al. (2021) in Italy, Bargain and Aminjonov (2020) in Europe, Barrios, Benmelech, Hochberg, Sapienza, and Zingales (2021) in the US, find that places with higher social capital are more likely to voluntarily comply with social distancing measures. We contribute to the literature on social capital by showing its role as a driver for COVID-19 vaccination.

The paper proceeds as follows. Section 2 describes the context behind COVID-19 vaccination in Europe. Section 3 details the dataset and discusses the identification strategy. Sections 4 and 5 show the main findings and discussion. Section 6 concludes. External statistics and robustness checks are in the Appendix.

2. COVID-19 vaccine

This Section presents the context surrounding COVID-19 vaccination in Europe in Summer (2021). First, we argue that because the COVID-19 vaccine was available to our target population, in sufficient supply, and free, we can discard the supply channel as a potential driver of cross-country heterogeneity in vaccine uptake. In addition, as the COVID-19 vaccine was not subject to any vaccine mandate at the time of our study, it allows us to treat vaccination outcomes as individuals’ decisions. Likewise, we end this Section with a discussion about the vaccination campaign in Germany. In Germany, the supply to regions (the Länder) was organized at the federal level and the vaccination campaign was not systematically different between East and West German regions.

2.1. Availability and eligibility

As stated by the European Center for Disease Control, “By January 2021, all 30 EU/EEA countries had started COVID-19 vaccination campaigns, and different COVID-19 vaccine products have been gradually introduced as they became available through the EU Vaccines Strategy” (ECDC, 2021a). Due to the limited supply of vaccine doses at the start of the campaign, countries opted to prioritize those who were most at risk of severe disease, such as the elderly, those with comorbidities, or healthcare workers. Each country established its own calendar, defining phases, from 2 to 16 phases, in order to ultimately reach the full coverage of the adult population. By June 2021, vaccines had been made widely available so that there were very few eligibility restrictions left.

To further check the availability of COVID-19 vaccines, we exploit the Oxford COVID-19 Government Response Tracker (OxCGRT).2 Appendix A.1, Fig. A.1 shows that several countries did not have universal vaccination coverage by mid-June 2021. This is mostly due to age restrictions, which were the last ones to be lifted. Universal coverage of adults above age 50 had been reached for all countries in this study by July 1st.3

Appendix A.1, Fig. A.2 shows the distribution of interview dates of the second SHARE Corona survey by country. Although most interviews took place during July 2021 (the totality for Germany) and the beginning of August, some countries also ran interviews during the last two weeks of June. We address this potential source of cross-country heterogeneity in three ways: first, in all our specifications, we control for vaccine availability (as defined by the OxCGRT), or the number of days since the vaccine was available to the respondent, which vary with individuals’ dates of interview and country of residence; second, we check that our results are robust including the number of days the vaccine was available in a country for each respondent depending on her age and country of residence; finally, we repeat our analysis restricting to respondents who completed the survey after July 1, 2021.4

2.2. One common strategy

The EU adopted a common vaccines strategy with a view to securing supplies and facilitating their distribution. Through advance agreements with vaccine manufacturers, the EU Commission secured the right to buy a specified number of vaccine doses in a given timeframe and at a given price. Part of the objectives of the EU vaccines strategy was to ensure equitable and affordable access to a vaccine as early as possible. As a result, all EU residents have been offered the right to a free vaccine, with no co-payment of any sort.

2.3. Not mandatory at the time of the interview

At the time of the survey, vaccination was not mandatory in any country, except in some cases for healthcare workers. Vaccine mandates such as the French “pass sanitaire” or the Italian “green pass” were not implemented until August 2021.5 Moreover, we expect that for the more sedentary elderly the vaccination requirements play less of a role as an incentive for vaccination than to more mobile younger individuals.

2.4. Vaccination campaign in Germany

Since this study focuses in more details on Germany, we end this Section with a brief overview of the COVID-19 vaccination campaign there. Vaccination has officially started on December 27, 2020. During the first months, the federal government was responsible for financing and distributiog vaccines to the vaccination centers. In turn, state governments (at the Länder level) ensured storage and distribution of doses in vaccination centers and to mobile vaccination teams associated with vaccination centers (see e.g. Lorenz-Dant,K and Stefanie Ettelt, 2021; Campaign, 2021). The supply of vaccines was proportional to population in each Land (Campaign, 2021).

The federal vaccination strategy consisted of three phases depending on the targeted population. During the first phase, vulnerable individuals, elderly above 80 or health workers, were eligible to get COVID-19 vaccine. The process of making appointments was organized at the Land level. In particular, individuals needed to make an appointment online or by phone (in Baden-Wurttemberg, Bavaria, Brandenburg, Hesse, North Rhine-Westphalia, Rhineland-Palatinate, Saarland, Saxony, Saxony-Anhalt, Schleswig-Holstein, Thuringia) or were contacted directly with an invitation for vaccination by a letter (in Berlin, Mecklenburg Western-Pomerania, Lower Saxony). Logistics problems and overwhelmed hotlines at the beginning of the campaign were present but not in a systematically different way across East and West Germany in the first phase (Campaign, 2021).

After April 6, 2021, medical practices were involved in the vaccination campaign, and starting on June 7^th, vaccination priority was lifted in Germany (Campaign, 2021).

Finally, regarding the representativeness of the data used in this study across German Länder, the share of vaccinated respondents in our data is, on average, higher than the one available in the official statistics on vaccination rates among adult individuals in Germany. However, these differences are not systematically different across Eastern and Western German regions. If anything, vaccination rates are slightly more overestimated in East Germany in our survey data compared with official statistics (see Table A1 in Appendix A.2).

3. Data and empirical strategy

In this Section, we first describe the SHARE data used in the study. Next, we discuss the identification strategy.

3.1. SHARE data

We use longitudinal data from the Survey of Health, Ageing, and Retirement in Europe (SHARE), SHARELIFE, and the SHARE Corona Survey. SHARE is a multidisciplinary and cross-national panel database of individuals aged 50 or older.6 For that reason, all individuals in our sample were born before the fall of Communism in Eastern Europe.

3.1.1. SHARE Corona Survey

Since the beginning of the COVID-19 pandemic, two special SHARE Corona Surveys have been conducted: the first between June and August 2020; and the second between June and August 2021. Each respondent in the second Corona Survey was asked “Have you been vaccinated against COVID-19?",7 8 Among those who did not get vaccinated the survey asks if they have already made an appointment.9 Finally, we can distinguish between respondents who do not want to get vaccinated or are undecided.

Additionally, the data includes vaccination choices against the flu (before and during the pandemic) and against pneumonia (during the pandemic).

Age, gender, an employment dummy, and the day-month-year of the interview are taken from the second SHARE Corona Survey.

3.1.2. Retrospective information

We retrieve early-life information from the SHARELIFE survey conducted in wave 3 (2007) or in wave 7 (2017). SHARELIFE collects information from birth to the moment of the survey, which allows us to build individuals’ residential history and to identify the region of residence each year. This will make it possible to differentiate life experiences that happened before the end of the Communist regime (before 1989), during the transition period (1990–2000), and currently (which we do as robustness check in Appendix L).

First, we create an indicator for being born in a post-Communist country. It is equal to one for all respondents born in the Czech Republic, Poland, Hungary, Slovenia, Estonia, Croatia, Lithuania, Bulgaria, Latvia, Romania, and Slovakia. It is coded as 0 for Sweden, the Netherlands, Denmark, Finland, Belgium, Austria, France, Switzerland, Luxembourg, Spain, Italy, Greece, Portugal, Cyprus, and Malta. We leave out Germany in this exercise to make sure it does not drive our results.

In the case of Germany, due to the country division for almost 40 years, we consider a regional variable to define exposure to Communism. The “East Germany” indicator takes value one for all individuals born in Brandenburg, Mecklenburg-Western Pomerania, Saxonia, Saxonia-Anhalt, and Thuringia. It is equal to zero if respondents were born in Baden-Wuerttemberg, Bavaria, Bremen, Hamburg, Hesse, Lower Saxony, North Rhine-Westphalia, Rhineland-Palatinate, Saarland and Schleswig-Holstein. Using the region of birth can potentially lead to a treatment misassignment problem; for example, the “East Germany” dummy is equal to one for a respondent who was born in Saxonia in 1939 and then moved in 1942 to Bavaria. However, few respondents moved between the East-West border before 1950, the date Germany was divided. Yet, as a robustness check, we re-estimate the model defining an “East Germany” dummy based on the region of residence in 1950, 1989 or 2007. The results remain unchanged. We leave out Berlin as we do not know if a respondent lived in the Eastern or Western part of the city. We also check that defining an “East Germany” dummy from the special question in SHARE for the German sample about the region of residence in 1989 (hence including Berlin) does not change the results.

Next, we construct proxies for early-life socioeconomic status: being vaccinated at age 16; being born in a rural area; a dummy for having more than 25 books in the house at age 10; and the number of home equipments (e.g., hot running water supply inside the house and others).

3.1.3. Longitudinal information

Finally, we match individuals from the second SHARE Corona Survey with their answers in previous SHARE waves, over 8 waves from 2004 to 2019. We define seven educational ISCED-1997 categories, income and wealth quartiles at the country level before the pandemic. Regarding predetermined health, we include four categories for self-perceived health, the EURO-D depression scale (the sum of 12 symptoms of depression) and the number of chronic diseases (out of 13). To control for fertility and marriage history, we include two indicators: having a child and living with a partner. We also control for living in a big city, a smaller town or a rural area.

Our final sample includes 34,173 respondents who were born between 1916 and 1964 when we pool all European countries. When restricting to Germany, we study 1,518 respondents. Tables B.1 and B.2 in Appendix B report descriptive statistics of the variables used in the analysis of East versus West Europe and East versus West Germany.

3.2. Empirical strategy

We run two analyses. The sample of countries and individuals differ, but the main model is similar:

$$y_i=\alpha Eas{t}_i+\beta X_i+{\epsilon }_i\text{,}$$ [1]

where $y_i$ captures the vaccination decisions of an individual i. Baseline controls, $X_i$, include a constant, age, a female indicator, and week of interview (specifications including only those main controls are estimated and displayed in “Panel I″ of our regression tables). In the models with additional controls (corresponding to “Panel II”), we add to ${\mathit{X}}_{\mathit{i}}$ a set of current and early-life characteristics: educational ISCED-1997 categories, an employment indicator, income and wealth quartiles, four categories of self-perceived health, the EURO-D depression scale, number of chronic diseases, an indicator of cognitive functioning proxied by the low number of recalled words in the word recall cognition test, an indicator of the quality (network satisfaction) and of the quantity of one's social network (network size), an indicator for having a child, living with a partner, living in a big city, a town, or a rural area, being vaccinated at age 16, being born in a rural area, having more than 25 books and the number of home equipment in the house at age 10.

We two-way cluster standard errors at age and region of residence level.10 To account for the relatively small number of clusters, we additionally calculate p-values based on the wild-cluster bootstrap-t procedure, based on 1000 replications. Wild bootstrapped p-values appear in squared brackets in the regression tables. Although the two-way wild bootstrap p-values are more conservative than the standard two-way clustering procedure, our conclusions hold. When clustering (using wild bootstrap) at the country or regional level in the case of Germany, which is actually the level of the East treatment, not only do our results hold, but they become more significant, as all estimations yield an East coefficient that is significant at 1 percent level, even when adding all individual controls (see Appendix M).11

In the cross-country analysis (leaving out Germany), $Eas{t}_i$ is equal to one if an individual was born in one of the 11 post-Communist countries regardless of the current place of residence, and 0 in the remaining 15 countries. The set of controls ${\mathit{X}}_{\mathit{i}}$ additionally includes current country of residence's specific characteristics: GDP per capita in 2019, population in January 2020, number of hospital beds per capita in 2019, public and out-of-pocket health expenditures as a share of GDP in 2019, median age of the population, as well as life expectancy, share of the older (55–74 year old) population with high education (both upper secondary and tertiary levels). The cumulative number of COVID-19 infections by June 1, 2020 and by January 1, 2021, and the vaccination policy at the moment of the interview come from the OxCGRT Tracker.12 Alternatively, we replace vaccination availability at the time of the interview with the number of days the vaccine had been available in a country for each age before the interview.

In the analysis based on Germany, $Eas{t}_i$ is equal to one if an individual was born in one of the five East German Länder regardless of current Land of residence, and 0 in the remaining 10 Länder of West Germany. When stated, we include additional controls at the Land level: GDP per capita in 2019, population in January 2020, number of hospital beds per capita in 2019, and cumulative number of COVID-19 infections by June 1, 2020 and by January 1, 2021.

When we broadly compare differences across countries, the estimate $\alpha$ captures associations between exposure to Communist institutions and to the transition and vaccination decisions in 2021. We use these results as first suggestive evidence about the long-run impact of institutions, but we interpret them with caution as we are aware that these countries already differed along cultural and economic lines before the Communism era, were exposed to a variety of experiences under Communism and followed very different paths related to the socio-economic transition after 1989. We therefore abstain from drawing conclusions from the cross-country analysis, and consider it at most as being consistent with our hypothesis. That hypothesis-that we can still find an imprint of the past Communist and Post-Communist experiences on vaccination decisions-will be tested in the German context.

For that purpose, we assume that East and West Germany were similar before the separation according to the relevant outcomes in this study. This setting and identification strategy have been widely used by scholars starting with Alesina and Fuchs-Schündeln (2007).13 In our case, the coefficient, $\alpha \text{,}$ captures the effect of exposure to the Communist regime and to the later reunification on the take-up of vaccines. A recent study by Becker et al. (2020) revisits the key identifying assumption in the East-West Germany comparison, and suggests a careful interpretation of results because of a potential bias depending on the outcome due to differences in the pre-Communist era between the newly assigned country borders. They show no pre-separation differences between East and West in income, GDP per capita, employment in health or domestic sectors. Nevertheless, East and West Germany differed in terms of political preferences, women's labor force participation, and Protestantism. The nature of our research question -studying vaccination outcomes-makes those concerns less pressing than in the case of labor market or gender-related outcomes. Still, some factors, such as church attendance, which differed between East and West Germany before Communism, might be correlated with vaccination outcomes, leading to a bias in our estimates. Church attendance was lower in East than in West Germany, and religiosity has been shown to be associated with refusal of COVID-19 vaccine (Callaghan et al., 2021). Then, our estimate is more likely to be a lower bound of the true effect. As robustness checks, we included controls for individual level frequency of praying and political preferences, as known from previous waves of SHARE, and our findings remain unchanged.

In Appendix D, we present an alternative specification, replacing the “East Germany” indicator with the number of years a respondent has lived in East Germany between 1949 and 1990.

4. Results

This Section presents our main results, first about all post-Communist countries and then looking at the East-West Germany comparison. In Column 1 in Table 1, Table 2 and 2 , the dependent variable is equal to one for those who got the COVID-19 vaccine or made an appointment. Those who say they want to be vaccinated but have not made any appointment, or those who do not want to be vaccinated or are hesitant about it, are then part of the reference group. We group these individuals together in order to focus on revealed preferences rather than stated preferences. Then in Columns 2 and 3 we focus on the “do not want” and the “undecided” categories. Finally, in Column 4, we analyze the determinants of getting the flu vaccine.

Table 1.

Impact of post-communist institutions and transition on vaccination decisions: Evidence from Europe excluding Germany.

	COVID-19 vaccine			Other
	Got	Do not want	Undecided	Flu
	(1)	(2)	(3)	(4)
Mean dep. var.	0.822	0.083	0.063	0.387
SD dep. var.	0.382	0.276	0.244	0.487
Panel I: main controls
Post-Communist countries	−0.175***	0.079**	0.093***	−0.357***
	(0.003)	(0.026)	(0.002)	(0.000)
	[0.006]	[0.041]	[0.009]	[0.000]
R-squared	0.131	0.067	0.050	0.185
N	34,173	34,173	34,173	34,127
Panel II: controlling for current and early-life characteristics
Post-Communist countries	−0.164***	0.054**	0.084***	−0.421***
	(0.000)	(0.011)	(0.000)	(0.000)
	[0.000]	[0.018]	[0.009]	[0.000]
R-squared	0.081	0.034	0.039	0.18
N	23,383	23,383	23,383	23,352

Note: All models control for the constant, age, gender, week of interview. Other controls are characteristics of the current country of residence: population (January 2020), GDP per capita, number of hospital beds per capita, governmental and out-of-pocket health expenditures as a share of GDP, life expectancy, median age (all in 2019), the share of 55–74 years-old with upper-secondary education, the cumulative number of COVID-19 infections by June 1, 2020 and by January 1, 2021, and vaccination policy at the time of interview. In Panel II, controls for individual current and early-life characteristics are seven educational ISCED-1997 categories; employment dummy; income and wealth quartiles; four levels of self-perceived health; the EURO-D depression scale; the number of chronic diseases; poor word recall; having a child; living with a partner; location (big city, town, rural area); network size; network satisfaction; having been vaccinated at age 16; being born in a rural area; having more than 25 books and the number of home equipment at age 10. P-values of two-way cluster standard errors by age and region of residence are in parentheses. In square brackets, we report wild cluster bootstrap p-values. *p < 0.10, **p < 0.05, ***p < 0.01.

Table 2.

Impact of post-communist institutions and transition on vaccination decisions: Evidence from Germany.

	COVID-19 vaccine			Other
	Got	Do not want	Undecided	Flu
	(1)	(2)	(3)	(4)
Mean dep. var.	0.928	0.040	0.013	0.556
SD dep. var.	0.259	0.195	0.111	0.497
Panel I: main controls
East Germany	−0.082**	0.040*	0.010	0.104**
	(0.014)	(0.052)	(0.154)	(0.011)
	[0.041]	[0.103]	[0.241]	[0.051]
R-squared	0.041	0.015	0.010	0.035
N	1518	1518	1518	1515
Panel II: controlling for current and early-life characteristics
East Germany	−0.075**	0.004*	0.004	0.115***
	(0.018)	(0.074)	(0.633)	(0.010)
	[0.062]	[0.111]	[0.616]	[0.035]
R-squared	0.108	0.065	0.038	0.078
N	1518	1518	1518	1515

Note: All specifications control for the constant, age, gender, week of interview. In Panel II, controls for current and early-life characteristics are seven educational ISCED-1997 categories; employment dummy; income and wealth quartiles; four levels of self-perceived health; the EURO-D depression scale; the number of chronic diseases; poor word recall; having a child; living with a partner; location (big city, town, rural area); network size; network satisfaction; having been vaccinated at age 16; being born in a rural area; having more than 25 books and the number of home equipment at age 10. P-values of two-way cluster standard errors by age and region of residence are in parentheses. In square brackets, we report wild cluster bootstrap p-values. *p < 0.10, **p < 0.05, ***p < 0.01.

4.1. Evidence from Europe

We first look into how post-Communist institutions are related to vaccination outcomes in all European countries excluding Germany. Table 1 shows that the probability of getting the COVID-19 vaccine drops by 17 percentage points (21 percent) for those individuals who were born in a Communist country in the 20th century (Column 1). The magnitude of this estimate decreases but remains large and statistically significant, when controlling for individual socio-demographic and early-life characteristics and the evolution of the pandemic in a country (see Panel II).14 The dramatic drop in the vaccination rate goes with an increase in vaccine refusal and hesitancy (Columns 2 and 3).

The flu vaccine is an interesting outcome to look at, as it has been routinely given for decades, and therefore is subject to less skepticism among the population, in both Eastern and Western countries. We find that respondents born in post-Communist countries are less likely to get the vaccine against flu than their counterparts from non-post-Communist countries. However, since the access to flu vaccine is not homogeneous across countries under analysis, the estimate might capture supply effects: national vaccination campaigns, economic situation, climate conditions; we run two additional specifications. First, when we control for country of residence, meaning that the estimates are identified thanks to the sample of migrants from/to post-Communist countries, then the negative coefficient on the flu vaccine disappears (see Panel a, Table C1, Appendix C).15 When we further restrict the sample to non-post Communist countries (still including country-of-residence fixed effects), meaning that the estimates are identified only due to movers from post-Communist countries to non post-Communist countries, the negative coefficient on the flu vaccine also disappears. In contrast, the negative estimate on the COVID- 19 vaccine remains significant in the two specifications (see Panel b, Table C1, Appendix C). Although correlational, these results lend support to Communism affecting the uptake of COVID vaccine but not of the flu vaccine, as these specifications allow us to rule out potential unobserved country-of-residence supply factors.

4.2. Evidence from Germany

When estimating Equation (1) in Germany, it appears the legacy of the rise and fall of Communist institutions decreases the probability of getting the COVID-19 vaccine by 8 percentage points for those born in East Germany (Column 1 in Panel I, Table 2). This is actually more than the difference in vaccination rates between East and West Germany (88% in East Germany, 93% in West Germany in our data), as we compare individuals who are more similar in terms of observable characteristics. The point estimate of this effect barely decreases when including more confounders like socio-demographic and early-life determinants. When controlling for more life experiences, such as having a child, living with a partner, being disabled or sick, years of working experience, and unemployment experience, and when considering these events separately depending on which of the three key periods they have occurred, (the Communist era (until 1989), the transition period (1990–2000) and present times (post 2000)), our coefficient of interest remains remarkably stable (see Appendix L, Table L.1). In terms of magnitude, our point estimate for being born in East Germany vs West Germany is quite large, as it more than doubles the probability of not being vaccinated against COVID-19, which is (unconditionally) 7% in West Germany. We use relevant variables such as age, health (the youngest and the more healthy are less likely to get the vaccine), and a proxy for socio-economic background (having few books at home when growing up is associated with a lower probability of getting the vaccine), which seem to predict vaccination, to benchmark our main effect. We find it to be comparable to the (non causal) effects of being 18 years old younger, having an excellent or very good health (compared to a poor health), or more than twice that of having grown up in a home with few books.16

Neither regional characteristics before the outbreak of Corona, nor the intensity of the pandemic affect our main findings (Appendix H, Table H.1). We comment more on this in Section 5.2.

We also find that exposure to the rise and fall of Communism in Germany increases the probability of refusing the COVID-19 vaccination by 4 percentage points, but this estimated effect is lower, both in terms of magnitude and significance.

When looking at the other vaccines, whether flu or pneumonia, we find that East Germans are more likely to take it. This result is at odds with the broader cross-country comparison, and likely related to the compulsory vaccination campaigns in the past. We comment further on it in Section 5.3.

Next, we dig further into the “East Germany effect” and investigate how the latter might vary with years of exposure to Communism, in Appendix D. Specifically, we create a variable that sums the years a respondent has lived in East Germany between 1949 and 1990. Table D1 suggests that the impact on vaccination increases with years of exposure to Communism, but this effect is actually driven by being born in East Germany. Since by definition differences in years of exposure across individuals can only be due to year of birth and migration in or out of East Germany, which were an extremely rare event over the whole period, years of exposure and the “East” dummy share a 92% correlation.

Last but not least, following Lippmann and Senik (2018) and Lippmann et al. (2020), we perform a permutation test by simulating the other possible divisions of the country by randomly assigning regions across East and West Germany. Our results are confirmed and the best model fit corresponds to when we assign regions correctly (see Table E1 in Appendix E).

Finally, when we regroup respondents who say that they “want to get vaccinated” (1 percent) with individuals who already got vaccinated or have scheduled an appointment, the gap in the vaccination rate remains, see Table F1 in Appendix F.

5. Exploration of channels and alternative interpretations

5.1. Is lower COVID-19 vaccination in East Germany driven by predetermined health or other factors?

Decades of Communism and the transition toward capitalism could have impacted individuals’ health, in any direction, as Communism came with a more equal but also impoverished society, and the transition was an important mortality shock (see Brainerd, 2001). A healthier population in East Germany might feel less at risk of severe illness due to COVID-19, and offer more resistance to vaccination; a less healthy population could be more scared of the vaccine potentially unknown side effects. We find (see Table G.1 in Appendix G) that older East and West Germans do not differ significantly in terms of measures of health, whether mental (EURO-D depression score), physical (number of chronic diseases), or cognitive (poor word recall). There is a difference in subjective health, with those born in East Germany less likely to say their health is excellent, and more likely to say it is “fair”. In any case, our findings hold controlling for all pre-pandemic health variables, even self-assessed health status.

5.2. Are results affected by the differential impact of the COVID-19 pandemic?

A potential concern is that the low vaccination rate in East Germany can be related to a less severe impact of the pandemic there compared to West Germany.17 We address this concern in two different ways: first, we use external information from the Robert Koch Institute (RKI) about COVID-19 cases in each Land in June 2020, after the first wave of the pandemic, and in January 2021, by the beginning of the vaccination campaign (RKI, 2020 for June 1, 2020; RKI, 2021 for January 2, 2021). In line with Bluhm and Pinkovskiy (2021), we document that initially the COVID-19 pandemic had hit West Germany more severely than East Germany. Six months later though, the cumulative number of cases (per one million inhabitants) was higher in the East than in the West (see Appendix H, Table H1). Moreover, Bluhm and Pinkovskiy (2021) further argue that state policies during the COVID-19 pandemic and access to medical services were similar across the country. To rule out that East Germans take up less COVID-19 vaccines because of fewer cases of COVID-19 during the first wave of the pandemic regardless of the later evolution of the pandemic, we repeat our main analysis adding controls for the number of COVID-19 cases in the first wave at the region-of-residence level. Our main findings remain unchanged (see Table H.2, Appendix H).

Next, we perform a validation exercise, by using self-reported information in the SHARE Corona survey: the probability of knowing someone who had COVID-19 symptoms or tested positive after the first wave of the pandemic and later by Summer (2021) (see Table H.3, Appendix H). We find evidence that East Germans were less likely to have been exposed to the virus by Summer 2020, but more likely so by Summer 2021, consistent with official statistics about COVID-19 cases in Germany. Again, our findings hold when we add these variables as controls.

5.3. Social capital is lower in East Germany

The success of any vaccination strategy relies on the share of individuals getting the vaccine. As such, vaccination -together with other strategies of contagion avoidance- create a positive externality, and can be seen as a public good.

Social capital, as the set of beliefs that promote cooperation and help to overcome the free-rider problem (Guiso, 2010), participates positively in the provision of public goods (e.g. Putnam, 1993; Herrmann et al., 2008). We therefore investigate here whether lower social capital might be a channel through which past Communist institutions lead to lower COVID-19 vaccination.

Social capital is usually proxied in data by measures that involve some prosocial behavior, such as generalized trust (or some more specific measures of trust, towards government, authorities, and others), the number of blood donations, newspaper readership (Durante et al., 2021), Google searches linked to prosocial behaviors (Guriev and Melnikov, 2016). We follow closely Conzo and Salustri (2019) and look at three measures related to social capital in the SHARE data: voluntary work, political participation, and generalized trust. All three measures are taken from the latest pre-pandemic wave, as we do not want them to be impacted by the COVID-19 context (Bargain & Aminjonov, 2020). Voluntary work -defined as whether an individual has taken part in the last 12 months into voluntary or charity work-has been commonly used in the literature as a proxy for other-regarding preferences and social capital (e.g. Putnam, 1993; Glaeser et al., 2000). Political participation is defined as whether an individual has taken part in the last 12 months to a political or community-related organization. Last, trust is measured as a linear variable from 0 to 10, recording how much “most people can be trusted or that one should be careful in dealing with people”. We extract the first principal component (the only one with an eigenvalue greater than 1) out of these three measures and use it as the social capital index. The first principal component explains 54 percent of the variation in the three variables, with the following scoring coefficients: 0.79 for voluntary work, 1.21 for political participation, and 0.38 for trust.

First, applying the same identification strategy as presented in Equation (1), we find that the past Communist regime and the transition period have eroded individuals’ social capital (Table 3 ), consistently with findings from the 2006 EBRD Life in Transition Survey. East Germans do less voluntary work, are less trusting of each other, and participate less in political organizations, than West Germans (see Table I.1 in Appendix I), but the effect is stronger (when compared to the mean of the dependent variable) and more robust for voluntary work.

Table 3.

Impact of post-communist institutions and transition on social capital in Germany.

	(1)	(2)
Dep. var.: first principal component of social capital
Mean dep. var.	−0.005	−0.005
SD dep. var.	0.820	0.820
East Germany	−0.254***	−0.228***
	(0.000)	(0.001)
	[0.001]	[0.015]
Current characteristics	No	Yes
Early-life characteristics	No	Yes
R-squared	0.022	0.137
N	1517	1517

Note: All specifications control for the constant, age, gender, and week of interview. Controls for current and early-life characteristics are seven educational ISCED-1997 categories; employment dummy; income and wealth quartiles; four levels of self-perceived health; the EURO-D depression scale; the number of chronic diseases; poor word recall; having a child; living with a partner; location (big city, town, rural area); network size; network satisfaction; having been vac-cinated at age 16; being born in a rural area; having more than 25 books and the number of home equipment at age 10. P-values of two-way cluster standard errors by the age and the region of residence are in parentheses. In square brackets, we report wild cluster bootstrap p-values. *p < 0.10, **p < 0.05, ***p < 0.01.

Going further, Table 4 shows that higher social capital in Germany is indeed associated with a higher probability to have taken the COVID-19 vaccine, and a lower probability to be against it, which is in line with Martinez-Bravo and Stegmann (2022), who find -in a different context-that a negative shock on trust led to decreased immunization rates. Looking separately at each of the social capital items, voluntary work is the one that seems more correlated with vaccination outcomes (see Table I.2 in Appendix I), which makes sense as it is also the one variable that has the strongest East-West gradient when looking at cross-country correlations, and the one that was more impacted by being born in East Germany in Table I.2, confirming a negative association between past Communism and the share of individuals doing voluntary work.

Table 4.

Impact of social capital on vaccination decisions in Germany.

	COVID-19 vaccine			Other
	Got	Do not want	Undecided	Flu
	(1)	(2)	(3)	(4)
Panel I: main controls
Social Capital	0.022**	−0.012*	−0.003	0.023*
	(0.024)	(0.084)	(0.431)	(0.062)
	[0.027]	[0.111]	[0.457]	[0.100]
East Germany	−0.077**	0.037*	0.010	0.110***
	(0.021)	(0.070)	(0.147)	(0.010)
	[0.077]	[0.155]	[0.229]	[0.038]
R-squared	0.045	0.017	0.011	0.036
N	1517	1517	1517	1514
Panel II: controlling for current and early-life characteristics
Social Capital	0.015*	−0.008	−0.003	0.017
	(0.092)	(0.224)	(0.301)	(0.364)
	[0.044]	[0.148]	[0.249]	[0.206]
East Germany	−0.072**	0.034*	0.003	0.118***
	(0.023)	(0.090)	(0.679)	(0.010)
	[0.071]	[0.141]	[0.649]	[0.026]
R-squared	0.11	0.066	0.039	0.079
N	1517	1517	1517	1514

Note: All specifications control for the constant, age, gender, week of interview, and being born in East Germany. In Panel II, controls for current and early-life characteristics are seven educational ISCED-1997 categories; employment dummy; income and wealth quartiles; four levels of self-perceived health; the EURO-D depression scale; the number of chronic diseases; poor word recall; having a child; living with a partner; location (big city, town, rural area); network size; network satisfaction; having been vaccinated at age 16; being born in a rural area; having more than 25 books and the number of home equipment at age 10. P-values of two-way cluster standard errors by the age and the region of residence are in parentheses. In square brackets, we report wild cluster bootstrap p-values. *p < 0.10, **p < 0.05, ***p < 0.01.

How much of the East-West Germany vaccination gap does social capital account for? Being born in East Germany is associated with an estimated social capital index that is lower by 0.254 compared with being born in West Germany. A one unit decrease in the social capital index is associated with a 2.2 percentage point lower probability of vaccination. This suggests the social capital mechanism to account for (0.254*0.022)/0.082, i.e. roughly a 6.8 percent of the estimated effect of being born in East Germany on getting the COVID-19 vaccine.

Last, we believe our social capital index does not capture all dimensions of social capital that are relevant to explain vaccination. Our measure of trust for instance, is a measure of generalized trust, or trust in others, which does not necessarily reflect the confidence in public State institutions that are those “imposing” the vaccine in the COVID-19 context. As shown by Costa-Font et al. (2021), which uses cross-country aggregate data on attitudes towards vaccines and several dimensions of trust, exposure to Communism is associated with a drop in government trust that is twice that in interpersonal trust. Algan et al. (2021) collects panel data from 12 countries and finds that trust in scientists is the key driving force behind individual support for and compliance with compliance behaviors, and for favorable attitudes towards vaccination. Had we had access to more adapted measures of trust, we believe our estimates of the impact of social capital might have been of higher magnitude, and that they would have taken away part of that of being born in East Germany.

5.4. Do East Germans have a lower general exposure to vaccines?

Our results on East Germany getting less vaccinated against COVID than West Germany do not extend to more traditional vaccines such as the long-known vaccine against influenza, which was created in the 1940s.18 When we replace the vaccine against flu in 2021 with -vaccine against pneumonia or vaccine against flu before pandemic-we still systematically find a positive effect of being born in East Germany on vaccination (Columns 3–4 in Table K.1, Appendix K). This is likely to be related to the long-standing mandatory vaccination campaigns in East Germany (at the center of the econometric strategy in Bluhm and Pinkovskiy, 2021), which allows us to discard an overall vaccine skepticism, and lends support to the hypothesis of a lack of trust that seems more specific to the novel COVID-19 vaccine. This is also in line with our findings on social capital, as low social capital individuals are more likely to offer resistance to a new vaccine that is aggressively promoted by the government, whereas they were already used to the influenza vaccine.

5.5. East Germans are less willing to comply with anti-COVID-19 measures

COVID-19 vaccination being the most important measure advocated by public authorities in order to curb contagion, it is likely that individuals who were reluctant to maintain social distancing or to comply with other preventive measures would also be more willing to take a vaccine.

We explore this hypothesis by looking at preventive behaviors in the first and the second waves of the SHARE Corona Survey. For the first wave, we use the five preventive measure variables that do not involve the company of others, i.e. social distancing, cough covering, hands washing, sanitizer use and mask wearing. We follow Bertoni et al. (2021) and run a principal component analysis instead of looking separately at each type of preventive behavior, avoiding potential multiple hypothesis testing issues. We obtain two components with an eigenvalue above 1 that explain 71 percent of the total variance. For the second wave, only two of the abovementioned five variables are available: social distancing and cough covering. As expected, East Germans comply less with preventive measures, but the only significant result has to do with social distancing in 2021 (see Table J.1, Appendix J). Nevertheless, our models do not consistently show a positive relationship between social capital and preventive measures (see Table J.2, Appendix J). Because other studies have shown such a link (e.g. see Durante et al., 2021; Bargain & Aminjonov, 2020; and evidence on the role of ethnic diversity Egorov et al., 2021), we believe our measure of social capital does not cover enough of the trust spectrum, to yield significant results here. Still, we find coefficients that are either extremely close to 0, or exhibit a positive sign, as we shall expect.

5.6. Further discussion on other potential channels and alternative interpretations

Could it be that our estimate of the impact of being born in a former Communist regime captures something else than exposure to Communism?

First, as mentioned in the introduction, our estimate is a mixture of the effect of exposure to Communism and to the transition out of Communism, which has been shown to be a particularly hard period in terms of financial hardship, life satisfaction, particularly for the elderly (e.g., Guriev and Zhuravskaya, 2009; Easterlin, 2009), in terms of health, and more generally speaking human losses (Brainerd, 2001). While this has sometimes been ignored in studies that use the East-West German divide to identify the role of past Communist institutions on a certain outcome, we believe the dual interpretation of the estimate should be acknowledged.

Second, we have uncovered evidence that individuals are more skeptical towards the COVID- 19 vaccine in East Germany compared to West Germany. If this is the case on the receiving end of the vaccine, there is no reason why it would not be the case on the dispensing end. This applies to all kind of medical workers, who are supposed to incentivize their patients to get vaccinated: if physicians, nurses, and other medical care workers instead share their concerns regarding the COVID-19 vaccine, and do so systematically more in East Germany than in West Germany, patients who are still hesitant about the vaccine might well decide to refuse it. Even those who had decided to get the vaccine could change their mind if people who know more about health than they do warn them against the vaccine. Likewise, if the public discourse, for instance from local politicians, or local media (local newspapers, TV, or even social media, as one's social networks online is mostly made of geographically close people), tends to relay more anti COVID-vaccine opinions in East than in West Germany, then there is a kind of feedback effect, as individuals who are more inclined to refuse the vaccine are more exposed to similar anti-COVID-19-vaccine ideas. The media have been shown to influence compliance with social distancing during the COVID crisis (Simonov et al. 2020 about Fox News in the US), or the spread of “anti-vax” ideas (see Chiou and Tucker 2018, about fake news on social media and the anti-vaccination movement) before the COVID-19 crisis. These effects are amplified by “filter bubbles”, or “ideological frames”, resulting from website algorithms that offer to the user information they would like to see based on their online history. As a result, users become locked down within their information bubble, separated from other points of view.

Whether through medical care workers, local politicians, newspapers, or social networks, this would lead to the propagation of more anti-vaccination stories, and eventually produce an amplification of the negative effect of exposure to the past Communist regime on individuals’ decision to get vaccinated.

5.7. Current or inherited environment? That is the question

Finally we use movers from East to West to provide more insights about whether the current or the inherited environment matters more for vaccination decisions. Accordingly, we define a new variable which can take on four values: to be born in East and to live in East now, to be born in East and to live in West, to be born in West and to live in East, and to be born in West and to live in West now, which we could label as EE, EW, WE, WW (first letter is birth, second letter is where the person lives). We can identify “pure birth” effects from the comparisons EW-WW and EE-WE, and “pure where-you-live” effects from the comparisons WE-WW and EE-EW. Unfortunately, we do not have enough movers to identify the double differences (with standard errors) precisely. This makes no doubt for the case of Germany (6 individuals moving from West to East Germany, 61 from East to West Germany), and is likely in the case of the rest of Europe. Even if there was a sufficiently high number of movers, the cross-country analysis is still not an ideal setup for uncovering a causal impact of exposure to Communism on vaccination decisions, for reasons already explained before.

Nevertheless we present our findings in the case of Europe (excluding Germany), in Table 5 and Fig. C1: both “where you live” and “where you were born” seem to matter, but it is hard to pin down their relative magnitudes. Those who are less vaccinated against COVID-19 are those who were born in the East, and who were still living there when surveyed, both in the case of Europe and of Germany. Conditional on currently living in the West, those born in the East seem to be less vaccinated than those born in the West, in line with inherited environment playing a role. On the other hand, Table C1 is also trying to get at “pure birth” coefficients, and suggests there is indeed a “pure birth” effect, since the “East” coefficients are large and significant, but at the same time, the fact that the coefficients are smaller than in the main specifications suggests that “where you live” also matters.

Table 5.

Impact of post-communist institutions and transition versus living in a post-communist country on vaccination decisions: Evidence from Europe excluding Germany.

	COVID-19 vaccine			Other
	Got	Do not want	Undecided	Flu
	(1)	(2)	(3)	(4)
Mean dep. var.	0.822	0.083	0.063	0.387
SD dep. var.	0.382	0.276	0.244	0.487
Panel I: main controls
Born in West and moved to East	−0.202**	0.107	0.029	−0.412***
	(0.030)	(0.108)	(0.392)	(0.000)
	[0.033]	[0.129]	[0.445]	[0.003]
Born in East and moved to West	−0.147**	0.043	0.089***	−0.093
	(0.019)	(0.220)	(0.006)	(0.163)
	[0.054]	[0.245]	[0.049]	[0.222]
Always lived in a post-Communist country	−0.198***	0.095**	0.097**	−0.448***
	(0.010)	(0.037)	(0.006)	(0.000)
	[0.021]	[0.059]	[0.049]	[0.000]
R-squared	0.132	0.067	0.050	0.186
N	34,173	34,173	34,173	34,127
Panel II: controlling for current and early-life characteristics
Born in West and moved to East	−0.309***	0.166**	0.025	−0.948***
	(0.006)	(0.023)	(0.511)	(0.000)
	[0.005]	[0.005]	[0.551]	[0.001]
Born in East and moved to West	−0.085*	0.002	0.081***	−0.064
	(0.065)	(0.937)	(0.006)	(0.34)
	[0.114]	[0.962]	[0.072]	[0.395]
Always lived in a post-Communist country	−0.314***	0.142***	0.093**	−0.951***
	(0.000)	(0.000)	(0.019)	(0.000)
	[0.000]	[0.005]	[0.032]	[0.005]
R-squared	0.082	0.034	0.039	0.184
N	23,383	23,383	23,383	23,352

Note: All specifications control for the constant, age, gender, week of interview. Other controls are characteristics of the current country of residence: population (January 2020), GDP per capita, number of hospital beds per capita, governmental and out-of-pocket health expenditure as a share of GDP (all in 2019), life expectancy in 2019, median age in 2019, the percent of individuals with upper education among the older population, the cumulative number of COVID-19 infections by June 1, 2020 and by January 1, 2021, and vaccination policy at the moment of the interview. In Panel II, controls for current and early-life characteristics are seven educational ISCED-1997 categories; an employment dummy; income and wealth quartiles; four categories for self-perceived health; the EURO-D depression scale; the number of chronic diseases; poor word recall; having a child; living with a partner; three location categories (living in a big city, in a town or in a rural area); network size; network satisfaction; being vaccinated at age 16; being born in a rural area; a dummy for having more than 25 books and the number of equipment in the house at age 10. P-values of two-way cluster standard errors by age and region of residence are in parentheses. In square brackets, we report wild cluster bootstrap p-values. *p < 0.10, **p < 0.05, ***p < 0.01.

Besides, migration is an endogenous choice variable which might relate to individual unobserved preferences that themselves relate to preferences for the Communist regime and for vaccination, so that we cannot rely on this exercise to provide “clean” evidence in support of our hypothesis.

6. Conclusion

All European countries succeeded in guaranteeing universal availability of the COVID-19 vaccine to their citizens. Yet, the vaccination rate against COVID-19 varies significantly between Central and European countries, and Western countries.

We exploit novel data from the second COVID-19 SHARE wave to answer this research question. First, we show that post-Communist countries have a lower COVID-19 vaccination rate even after controlling for potential confounders. Next, to get closer to a causal estimate we switch to the quasi-natural experiment provided by the separation and later reunification of Germany. In contrast with the higher prevalence of flu vaccination, we document lower COVID-19 vaccination among older East Germans when controlling for individual socio-economic and demographic characteristics. Exploring possible mechanisms, we find that social capital - measured as voluntary work, political engagement and trust in people - is lower among East Germans than West Germans. Furthermore, social capital correlates negatively with COVID-19 vaccine uptake. However, only a small fraction of the East-West Germany Covid-19 vaccination gap can be explained by our measures of social capital.

Our findings are relevant for policymakers to understand better reasons for non-compliance with COVID-19 policies and, as a result, to tailor vaccination campaigns depending on targeted groups or targeted places.

CRediT author statement

Conceptualization All authors. Methodology All authors. Software All authors. Validation All authors. Formal analysis All authors. Investigation All authors. Resources All authors. Data Curation All authors. Writing - Original Draft All authors. Writing - Review & Editing All authors. Visualization All authors. Supervision All authors. Project administration All authors. Funding acquisition All authors.

Handling Editor: Social Epidemiology Editorial Office

Appendix A. Supplementary data

The following is the Supplementary data to this article.

Data availability

This paper uses data from SHARE Waves 1, 2, 3, 4, 5, 6, 7, 8 and 9 (DOIs: 10.6103/SHARE.w1.800, 10.6103/SHARE.w2.800, 10.6103/SHARE.w3.800, 10.6103/SHARE.w4.800, 10.6103/SHARE.w5.800, 10.6103/SHARE.w6.800, 10.6103/SHARE.w7.800, 10.6103/SHARE.w8.800, 10.6103/SHARE.w8ca.800, 10.6103/SHARE.w9ca800) see Börsch-Supan et al. (2013) for methodological details.(1) The SHARE data collection has been funded by the European Commission, DG RTD through FP5 (QLK6-CT-2001-00360), FP6 (SHARE-I3: RII-CT-2006-062193, COMPARE: CIT5-CT-2005-028857, SHARELIFE: CIT4-CT-2006-028812), FP7 (SHARE-PREP: GA N°211909, SHARE-LEAP: GA N°227822, SHARE M4: GA N°261982, DASISH: GA N°283646) and Horizon 2020 (SHARE-DEV3: GA N°676536, SHARE-COHESION: GA N°870628, SERISS: GA N°654221, SSHOC: GA N°823782, SHARE-COVID19: GA N°101015924) and by DG Employment, Social Affairs & Inclusion through VS 2015/0195, VS 2016/0135, VS 2018/0285, VS 2019/0332, and VS 2020/0313. Additional funding from the German Ministry of Education and Research, the Max Planck Society for the Advancement of Science, the U.S. National Institute on Aging (U01_AG09740-13S2, P01_AG005842, P01_AG08291, P30_AG12815, R21_AG025169, Y1-AG-4553-01, IAG_BSR06-11, OGHA_04-064, HHSN271201300071C, RAG052527A) and from various national funding sources is gratefully acknowledged (see www.share-project.org).