What we know about differences in COVID-related deaths

The debate on the main political factors influencing COVID-related infections and deaths has been examined along five main dimensions: regime type, state capacity, trust, leadership, and culture.

The relationship between regime type and public health outcomes remains a subject of debate. On the one hand, several studies have demonstrated that democracies outperform dictatorships in addressing standard public health measures, such as infant mortality rates (Annaka and Higashijima Reference Annaka and Higashijima2021; Besley and Kudamatsu Reference Besley and Kudamatsu2006). McMann and Tisch (Reference McMann and Tisch2021) examined several pandemics that occurred over the past century and found that democracies were more effective in containing their effects than authoritarian regimes.

On the other hand, early papers published in response to the onset of the COVID-19 pandemic presented a contrasting view, suggesting that authoritarian regimes, with China as a prime example, were handling the crisis more efficiently. For instance, Cepaluni, Dorsch, and Branyiczki (Reference Cepaluni, Dorsch and Branyiczki2020) found that democracies fared worse than autocracies in terms of COVID-related mortality rates during the pandemic’s first wave (see also Bosancianu et al. Reference Bosancianu, Hilbig, Humphreys, KC, Lieber and Scacco2021). Similarly, Annaka (Reference Annaka2021) showed that authoritarian regimes performed better during the first year of COVID-19 in terms of mortality rates, although he attributed this success to more extensive testing programs rather than praising anti-liberal measures. Trein (Reference Trein2020) found that countries with a strong legacy of past authoritarian institutions were more likely to implement strict lockdown policies.

Other authors have offered a more nuanced perspective on this relationship. Cheibub, Hong, and Przeworski (Reference Cheibub, Hong and Przeworski2020) found that democracies were more hesitant to adopt stringent policies during the first wave (and therefore, suffered a greater impact from the pandemic shock), but they eventually implemented such measures once death tolls began to rise, which may have compensated for the initial authoritarian advantage (see also Karabulut et al. Reference Karabulut, Zimmermann, Bilgin and Cansin Doker2021; Rodríguez-Pose and Burlina Reference Rodríguez-Pose and Burlina2021). Bayerlein et al. (Reference Bayerlein, Boese, Gates, Kamin and Mansoob Murshed2021) focused on the impact of populist leaders and found, using first-year monthly data for a sample of 42 countries, that they generated worse outcomes than less polarizing democratic authorities.

Given these mixed results, it is unsurprising that Cassan and Van Steenvoort (Reference Cassan and Van Steenvoort2021) found no relationship between regime types and mortality rates. Perhaps these variations stem from the authoritarian practice that Neumayer and Plümper (Reference Neumayer and Plümper2022) suggest may involve manipulating mortality data. According to these authors, the authoritarian advantage disappears once excess deaths are used instead of a country’s self-reported numbers.

I remain agnostic about the direction of the relationship between regime type and COVID-related mortality rates. Nevertheless, I explore whether the potential positive impact of socialist revolutionary experiences on COVID-19 containment can be attributed to the fact that countries with that legacy tend to have authoritarian institutions.

Secondly, state capacity is another commonly mentioned factor in the literature on pandemic management (Majeed and Gillani Reference Majeed and Gillani2017). It prominently featured in early analyses of the pandemic, such as Fukuyama’s (Reference Fukuyama2020). However, more rigorous statistical research methodologies have struggled to find a clear link. Bosancianu et al. (Reference Bosancianu, Hilbig, Humphreys, KC, Lieber and Scacco2021) found that state capacity measures do not predict mortality rates well. Similarly, for Kavanagh and Singh (Reference Kavanagh and Singh2020), the countries that were supposed to be better prepared to contain a pandemic – as measured by the Global Health Security index – did not outperform those with worse preparedness in terms of mortality rates during the first wave of the pandemic. Countering this result, Serikbayeva, Abdulla, and Oskenbayev (Reference Serikbayeva, Abdulla and Oskenbayev2021) claim that, for a partial period of the first wave, more capable and less democratic states dealt more effectively with the pandemic. In a similar vein, Cooray, Vadlamannati, and De Soysa (Reference Cooray, Vadlamannati and De Soysa2020) suggested that authoritarian countries with more public health capacity yielded lower mortality rates during the first wave, but only when their healthcare systems were more equitable. These findings may indicate that the explanatory power of state capacity needs to be mediated by other factors, such as regime type.

De la Cruz, Tezanos, and Madrueño (Reference de la Cruz, Tezanos and Madrueño2022) categorize countries into four groups based on their capabilities and performance against the pandemic and its associated economic crisis. Interestingly, their first two clusters predominantly include socialist countries, characterized by a combination of poor state capacity and lower mortality rates, though with variations in economic downturns. In the next section, I will explore why socialist revolutionary regimes might have been more effective in dealing with shocks like pandemics compared to countries without such legacies.

The third factor is trust, a concept that was frequently mentioned during the initial stages of the pandemic. It was posited that societies where citizens trusted each other and their political authorities would be more successful in collectively organizing against a public health challenge (Harring, Jagers, and Löfgren Reference Harring, Jagers and Löfgren2021). Preliminary evidence supported this assertion. For example, the early success of mitigation policies in certain countries was attributed to substantial social capital reserves (Liu and Wen Reference Liu and Wen2021; though Elgar, Stefaniak, and Wohl Reference Elgar, Stefaniak and Wohl2020 provide mixed evidence for this correlation). Institutional trust also played a crucial role in explaining early variations in COVID-related mortality during the first wave (Charron, Lapuente, and Rodriguez-Pose Reference Charron, Lapuente and Rodriguez-Pose2023; Oksanen et al. Reference Oksanen, Kaakinen, Latikka, Savolainen, Savela and Koivula2020). Bosancianu et al. (Reference Bosancianu, Hilbig, Humphreys, KC, Lieber and Scacco2021) conducted a comprehensive study and concluded that “the two features consistently related to mortality, conditional on healthcare system quality and population health, are trust in the government and trust in other people.” (p. 23).

However, these empirical studies are limited in scope, as measures of trust are generally available only for a subset of developed countries. This significantly restricts comparative analysis beyond the developed world. Additionally, it is challenging to isolate the singular causal impact of trust, as its determinants might also be connected to outcomes seemingly driven by trust.

Furthermore, the influence of culture on COVID-19 containment suggests that the marked differences observed in pandemic responses are driven not by institutions or specific behaviors, such as trust, but by deeper societal characteristics. Chen, Frey, and Presidente (Reference Chen, Frey and Presidente2021) observed that countries with more collectivist cultures implemented restrictions on geographic mobility more readily than those with predominant individualistic values. Canatay, Emegwa, and Talukder (Reference Canatay, Emegwa and Hossain Talukder2021) analyzed several socioeconomic factors, identifying power distance (the extent of power inequality acceptable to individuals), individualism, gender, and age as the most influential, though they did not consider institutional variables in their analysis.

The quality of leadership is the final explanatory factor. Anecdotal evidence from the first wave of the pandemic easily highlighted that some of the most successful countries were being ruled by female politicians -- such as Jacinda Ardern of New Zealand, Angela Merkel of Germany, and Tsai Ing-wen of Taiwan. In a more systematic fashion, Garikipati and Kambhampati (Reference Garikipati and Kambhampati2021) found that female political leaders did a better job at containing deaths than their male peers. However, this result has raised concerns about reverse causality (Piscopo Reference Piscopo2020) and sample limitations (Bosancianu et al. Reference Bosancianu, Hilbig, Humphreys, KC, Lieber and Scacco2021). Countries promoting women to the highest office are countries with specific characteristics that may well be better attuned to dealing with health crises. Bruce et al. (Reference Bruce, Cavgias, Meloni and Remígio2022) use a regression discontinuity design in close mayoral races in Brazil and observe that female mayors did a better job during the pandemic than male mayors. Although this evidence provides strong evidence for a causal relationship, it is not clear yet how far it can go with respect to its generalizability.

In the empirical analysis, I control for these factors – regime type, state capacity, trust, culture, and leadership – and discuss ways in which some of them may have mediated the influence of socialist revolutionary legacies over pandemic management.

Socialist revolutions, legacies, and collective threats

From the triumph of the Russian Revolution in 1917 to the most recent socialist revolutionary experiences in Venezuela (1999) and Nepal (2006), one of the biggest challenges to regime stability has been the occurrence of popular uprisings fueled by socialist ideals. These revolutions brought about significant changes in both the distribution of political power and economic resources (Goodwin Reference Goodwin1994).

In contrast to socialist regimes imposed by foreign occupation, such as Soviet-backed regimes in Central Europe after World War II, socialist revolutionary regimes carry an extraordinary amount of popular legitimacy. In theory, these regimes combine state capabilities, collective goals, and a popular mandate.

To begin with, when rebel-winning parties come into power, they often draw on revolutionary momentum to create more resilient states. In such cases, military, political, and economic leaders, as well as middle-to-top bureaucratic cadres, are either executed or forced to flee. The absence of hostile ruling classes allows for the restructuring of political structures without the usual constraints faced by democratic leaders (Huntington Reference Huntington, Huntington and Moore1970; Levitsky and Way Reference Levitsky and Way2022). In the absence of opposition, key new actors like the Party and the Military penetrate all social structures and gain control over their citizens. In Hassan, Mattingly and Nugent’s (Reference Hassan, Mattingly and Nugent2022) rendition of the concept of political control, socialist regimes usually excel at infiltrating and indoctrinating their populations.

What is intrinsic to socialist projects is the ambition to launch a complete reorganization of political and economic resources. If the new institutions can avoid corruption and maintain some sources of legitimacy, such as a recruitment and rewards system that promotes those with fewer resources as a means of social mobility, they can endure even after the initial revolutionary fervor subsides. This bond between the revolutionary principles and society can become even stronger when the state faces external threats, such as a foreign power seeking to overthrow the new regime (Clarke Reference Clarke2023). Foreign-imposed economic sanctions that trigger a rally-behind-the-flag effect allow revolutionary leaders to play the nationalist card and garner more support for the nascent institutions (Escriba-Folch and Wright Reference Escriba-Folch and Wright2010; Grossman, Manekin, and Margalit Reference Grossman, Manekin and Margalit2018).

Strong institutions are a prerequisite for achieving long-sought collective outcomes. A hallmark of the revolutionary program is to promote public goals such as land reform, full literacy, and some level of nationalization of commodity-extractive industries. Even if not entirely successful, these programs drastically alter the political and economic foundations of the country. Politically, they empower sectors traditionally excluded from decision-making by creating paths to mobility through official agencies. Economically, they transfer significant resources to the state, which becomes the employer or goal setter for a substantial portion of the workforce.

Most socialist regimes pursue these three collective projects, but revolutionary ones possess a surplus of legitimacy to enforce them on less receptive constituencies (Thaler Reference Thaler2018). These programs, while often not highly effective in boosting economic growth (Kalyvas Reference Kalyvas1999), enjoy broad appeal because they benefit large sections of the population. This popularity often endures even after the regime has fallen because some of these outcomes, such as literacy and land reform, are not easily undone. At times, this popularity, coupled with regime promotion, may reshape people’s beliefs about fundamental political values, such as whether the pursuit of collective goals should outweigh individual freedoms. These shared societal values may have a lasting impact on how individuals respond to public crises like pandemics.

As Stasavage (Reference Stasavage2020) has suggested, there is a trade-off between regime types when addressing collective challenges like pandemics: authoritarian regimes may have strong tools (conditional on state capacity) to contain pandemics, but their control of information can create incentives to downplay the true extent of the threat. In contrast, democracies cannot suppress the flow of information, though they face more constraints in implementing stringent anti-shock policies. Perhaps revolutionary regimes are better equipped to address collective problems due to their robust institutions, citizen collectivism, and strong popular foundations.

The intriguing question is why these tools should still be available to countries that have left their socialist revolutions behind and transitioned into other regime types. The literature on communist legacies is extensive (LaPorte and Lussier Reference LaPorte and Lussier2011; Pop-Eleches and Tucker Reference Pop-Eleches and Tucker2017) and typically focuses on institutional patterns and individual attitudes. First, revolutionary regimes rarely transform into liberal democracies (Levitsky and Way Reference Levitsky and Way2022). Consequently, these countries continue to benefit from the capacity to enforce compliance that authoritarian regimes typically exhibit. For example, security agencies established during the revolutionary period often remain untouched and survive the fall of the one-party regime. With deep community roots, these agencies may be more effective in maintaining order than comparable bodies in countries without such a history (Yashar Reference Yashar2018, chapter 7).

Regarding behavioral legacies, some collective-oriented beliefs persist even after the regime’s downfall. In line with Bondar and Fuchs-Schundeln (Reference Bondar and Fuchs-Schundeln2023), Pop-Eleches and Tucker (Reference Pop-Eleches and Tucker2017) find that citizens in post-communist societies tend to favor strong state intervention more and express lower support for democracy and markets. They also have more equalitarian gender attitudes (Beloshitzkaya and Reilly Reference Beloshitzkaya and Reilly2023). Initially, these perceptions were thought to be influenced by having lived through the communist period. However, there is growing evidence of intergenerational transmission of values, indicating that “the effects of political regimes on preferences can be extremely persistent” (Bondar and Fuchs-Schundeln Reference Bondar and Fuchs-Schundeln2023, 1).

In any case, it is not particularly surprising, as the success of socialist revolutionary movements has been associated with specific socio-economic conditions (Talhelm et al. Reference Talhelm, Zhang, Oishi, Shimin, Duan, Lan and Kitayama2014) or particular kin structures (Todd Reference Todd2019) that facilitated these beliefs. Regardless of whether these beliefs existed before the revolution or were nurtured by it, they tend to persist and are not easily replaced.

In summary, I argue that countries with a socialist revolutionary past performed better in containing the COVID-19 pandemic than countries without such an experience, regardless of whether they remain socialist.

Data and results

In this section, I first describe the data used to test the link between revolutionary socialism and pandemic management. Then, I offer basic comparisons of means in support of my hypothesis and finally move on to discuss more robust tests.

I am interested in investigating why some countries performed better than others in containing the death toll associated with COVID-19. To this end, my dataset includes all countries in the world with more than 1 million inhabitants. As health policies are devolved to sub-state political units in some countries, there is a significant body of literature investigating systematic patterns at the regional level (Charron, Lapuente, and Rodriguez-Pose Reference Charron, Lapuente and Rodriguez-Pose2023). However, there is a lack of regional data beyond the developed world, and furthermore, decentralization has not been identified as a key explanatory factor for COVID-19 mismanagement (Kuhn and Morlino Reference Kuhn and Morlino2022). Nonetheless, I control for whether countries have a federal structure as a way to account for intra-state heterogeneity.

The dependent variable in my analysis is the COVID-related death rate by country. I utilize the raw total counts from John Hopkins University from the beginning of the pandemic to May 1, 2022 (downloaded from https://github.com/CSSEGISandData/COVID-19). I examine death rates per 100,000 inhabitants because it provides a more informative measure for evaluating the actual impact of the pandemic when comparing countries with vastly different population sizes. I also re-run my analysis using “excess deaths” associated with COVID-19 data compiled by the World Health Organization (https://covid19.who.int/data). The WHO’s method compares all reported deaths during the pandemic, even if seemingly unrelated to COVID-19, with death trends during pre-pandemic years and attributes the increase (“excess deaths”) to the pandemic’s effect. This method has the advantage of circumventing government efforts to downplay the true number of victims. However, it may impute negative rates (i.e., countries where nobody seemed to die from COVID-19) for approximately 14 percent of the countries in the sample. It also faces challenges in distinguishing deaths where COVID-19 may be the primary cause from those where the virus was merely an intervening factor. Nevertheless, there is a relatively high correlation between the two measures (p-corr value = 0.70; see Figure AII.1 in the online appendix for a visual representation of the two dependent variables).

The critical independent variable is whether the country has a socialist revolutionary experience in its past. I define a “socialist revolutionary” regime as one in which new rulers emerged from popular uprisings with an economic program that nationalized key assets and granted the state central-planning powers, along with a political program that imposed plurality-limiting institutions – a revolutionary regime akin to Skocpol’s concept, but with a socialist orientation (Skocpol Reference Skocpol1979; Goldstone Reference Goldstone2001, 142). The definition excludes Scandinavian social democracies because they never assumed central control of the economy nor limited political plurality. It also does not include Salvador Allende’s short-lived period in power, as he failed to dismantle existing democratic institutions. I do code Venezuela and Nepal as positive cases, where leaders came to power through elections riding the wave of large grassroots movements. In Venezuela, Hugo Chavez, despite his fiery rhetoric, may not have initially qualified as a socialist regime. However, after the failed coup against him in 2002, he swiftly moved the country toward more political and economic control with a socialist underpinning (Azicri Reference Azicri2009). In Nepal, Maoist victories in the battlefields forced the state to negotiate a peace plan that ended the authoritarian monarchy and paved the way for their ascent to power (Adhikari Reference Adhikari2014). In both cases, the lack of an electoral pathway into power may not have been sufficient to impede their ultimate takeover.

Although my causal argument centers on socialist policies, it also takes into account that those policies were implemented with significant social support – the socialist governments were inaugurated through popular revolutions. In this sense, I make a distinction between endogenous socialist regimes and those where socialism was adopted through foreign imposition (e.g., USSR-backed regimes in Central Europe) or by military coup without grassroots support (such as in Myanmar in 1962 and Libya in 1969). In a few cases, socialist regimes incorporated some of the features associated with revolutionary socialism (such as stronger institutions and large-scale public programs), but without popular support, these policies may have faded away following the fall of these regimes. Most of the time, however, the absence of genuine domestic demand for a socialist takeover left little trace of these policies once their backers were removed from power. (Online) Appendix I includes a list of the regimes I classify as “socialist revolutionary” and their duration.

My initial empirical approach aims to demonstrate that when using straightforward mean comparisons, socialist revolutionary regimes have outperformed other types of regimes. Figure 1 presents four bar comparisons with different subsets of socialist countries (see the list of countries included at the bottom). The left panel shows that the six countries that still identify themselves as socialist have experienced a four times smaller COVID-related death rate than non-socialist countries, though their variance is large enough to push the comparison beyond standard significance levels (it is noteworthy that this variance is driven by Venezuela’s underperformance). The center-left comparison includes all countries that had a self-identifying socialist regime at some point after World War II, and the effect diminishes. Similarly, the center-right comparison adds countries that became independent after the dissolution of socialist federations (the Soviet Union and Yugoslavia), and the correlation reverses its direction, becoming significant but in the opposite direction: socialist countries appear to have fared worse than non-socialist countries during the pandemic.

Figure 1. Socialism and COVID-related death rates (as measured by JHU). Current socialist countries: China, Cuba, Laos, Nepal, Venezuela, and Vietnam [North Korea is not included because of lack of reliable data]. Current and former socialist countries: Afghanistan, Albania, Algeria, Angola, Benin, Bulgaria, Cambodia, Cape Verde, China, Congo, Cuba, Ethiopia, Hungary, Laos, Libya, Madagascar, Mongolia, Mozambique, Myanmar, Nepal, Nicaragua, Poland, Romania, Russia, Serbia, Somalia, Venezuela, Vietnam, and Yemen. Current, former and successor countries: All previous, plus: Bosnia, Croatia, Czechia, Estonia, Georgia, Germany, Kazakhstan, Kyrgyzstan, Latvia, Lithuania, North Macedonia, Moldova, Montenegro, Slovakia, Slovenia, Tajikistan, Ukraine, and Uzbekistan. Socialist revolutionary countries: Afghanistan, Albania, Algeria, Angola, Cambodia, Cape Verde, China, Congo, Cuba, Ethiopia, Laos, Madagascar, Mongolia, Mozambique, Nepal, Nicaragua, Russia, Serbia, Venezuela, Vietnam, and Yemen.

In the previous section, we argued that it is advisable to exclude all countries where socialism was imposed from the top through foreign occupation and/or a military coup. Most successor states would fall into this category, except for Russia and Serbia (Hoepken Reference Hoepken1999), which were the origins of their respective socialist revolutions. The right-hand comparison in Figure 1 takes this distinction into account and shows that revolutionary socialist (past and present) countries experienced two-and-a-half times fewer COVID-related per capita deaths than countries without that legacy. This effect is statistically significant and does not appear to be driven by the inclusion of specific cases that may be contentious, such as Afghanistan (Halliday Reference Halliday1978) or Yemen (Ishiyama Reference Ishiyama2005), where socialist-fed rebellions were supported by Soviet intervention (Porter Reference Porter1984).

From this point forward, my focus is on examining the impact of countries with a socialist revolutionary history on pandemic management. To assess this experience, I employ two indicators. First, I use a straightforward dummy variable that takes on the value of 1 for countries that have undergone a successful revolutionary experience in the past. However, this approach does not account for the duration of the socialist regime and the time elapsed since its collapse (Pop-Eleches Reference Pop-Eleches2007). Therefore, I introduce a second indicator that takes into consideration both of these aspects, which I define as follows:

$$socialist\;duration = {\rm{ln}}\left( {1 + \left( {{{SR\;years} \over {\left( {1 + years\;since\;collape} \right)}}} \right)} \right)$$

By design, countries with lengthy and continuing experiences of a socialist revolution, such as Cuba, Vietnam, and China, score high on this indicator. Countries with brief and bygone experiences, such as Mozambique, Afghanistan, and Yemen, receive low scores. Those without any socialist revolutionary past score zero.

To account for competing arguments, I include a set of independent variables that were discussed in the review section for their explanatory power. First, I incorporate the regime-type debate into the equation by measuring the extent to which each country was “liberal-democratic” in 2020 (all variables are measured the year before the start of the pandemic unless otherwise indicated). I utilize the Varieties of Democracy’s “liberal-democracy” index (Varieties of Democracy Project 2022).

Secondly, I control for a country’s wealth using the World Bank’s GDP per capita at purchasing power parity, which is a traditional measure of state capacity (https://data.worldbank.org/indicator/NY.GDP.PCAP.PP.CD). I also include the squared GDP per capita to explore potential non-linearities in the relationship between income and COVID-19, as some developing countries like Brazil, Mexico, and South Africa notably underperformed during the pandemic.

Thirdly, social trust has been identified as one of the best predictors of pandemic outcomes. However, obtaining reliable indicators for a globally representative sample of countries can be challenging. I employ Klassen’s (Reference Klassen2018) compilation of the social trust index, which ranges from 0 to 100 (higher values denote higher social trust). To maximize the number of observations, I include all countries with any data points during the 2000s, resulting in a sample of approximately 130 countries. As this inclusion results in the loss of some countries with socialist revolutionary experiences, I conduct all models with and without this social trust index.

Fourthly, it has been suggested that more open countries are more susceptible to the spread of viruses. I capture this effect using the “Social Globalization Index” measured in 2018, as developed by Gygli et al. (Reference Gygli, Haelg, Potrafke and Sturm2019). Fifthly, I account for the potentially greater intra-country heterogeneity in federal regimes by measuring whether the regime is confederal/federal or not (data sourced from Wig, Hegre, and Regan Reference Wig, Hegre and Regan2015). Sixthly, I code those countries where the main executive office was held by a woman before June 2020 (when the pandemic had affected nearly every nation in the world).

Seventhly, I control for health factors that may influence the varying lethality of COVID-19 (Basellini and Camarda Reference Basellini and Camarda2022; Cegan et al. Reference Cegan, Trump, Cibulsky, Collier, Cummings, Greer, Jarman, Klasa, Kleinman, Surette, Wells and Linkov2021). I focus on the percentage of inhabitants aged 65 years and older, the prevalence of diabetes in the population, and population density (as a proxy for crowdedness), all collected from the World Bank’s database. Finally, one of the most puzzling results of the pandemic has been its lower impact in Africa (Nguimkeu and Tadadjeu Reference Nguimkeu and Tadadjeu2021). To control for regional and cultural differences, I include simple regional dummies, with Africa serving as the baseline category.

Table 1 presents the results, and Figure 2 illustrates the effects of the main independent variables. I use two different measures of COVID-related death rates, one based on the John Hopkins University database, and the other drawn from the World Health Organization’s excess deaths reports.

Table 1. Regression analysis of COVID-related death rates

t statistics in parentheses.

* p < 0.1.

** p < 0.05.

*** p < 0.01.

Figure 2. Predicted effects of key factors over COVID-19 death rates.

Remarkably, countries with a socialist revolutionary experience have encountered lower numbers of deaths associated with COVID-19, regardless of the measure used. Countries with a socialist revolutionary legacy have, on average, 48 fewer COVID-related deaths per 100,000 inhabitants compared to countries without that legacy (using the JHU indicator). For an average country with a population of 40 million people, this translates into 29,365 fewer COVID-related victims in a country with a socialist revolutionary legacy, which is approximately sixty percent fewer than expected in countries without such a legacy (49,475 victims).

The other covariates mostly align with our expectations. Democracies do not appear to fare significantly worse than autocracies. Developing countries are more impacted by COVID lethality rates than both poorer and more affluent ones. Notably, countries with female prime ministers overwhelmingly outperformed those without, experiencing a lower rate of 35 deaths per 100,000 inhabitants compared to 122 deaths in countries led by male prime ministers. This intriguing result may be attributed to institutional features that facilitate the election of women to office, but it could also be influenced by leadership effects that are challenging for standard datasets to capture.

In contrast, the social globalization index and the health indicators seem to have a minimal impact on the count of COVID-related deaths. Finally, Europe and the Americas performed notably worse than African nations in managing the pandemic.

To assess the robustness of these findings, I conducted several additional checks, all of which are reported in (online) Appendix II. Firstly, I replicated the main results with additional independent variables, such as major pre-pandemic death causes and public spending on health as a share of GDP. I collected information about the prevalence in 2019 of the three most typical death causes associated to COVID-19 (ischemic heart disease, hypertension, and diabetes) (Elezkurtaj et al., Reference Elezkurtaj2021). The data come from the “Global Burden of Disease Study 2021 -- Cause-Specific Mortality 1990–2021” (Global Burden of Disease Collaborative Network, 2024). The results do not vary (see Table AII.1).

I also collected information about how much countries were spending on public health as a share of GDP in 2019, just before the start of the pandemic. The data come from the World Health Organization’s Global Health Expenditure Database (https://apps.who.int/nha/database/Select/Indicators/en). The indicator goes in the expected direction (higher spending produced lower deaths), and although not always statistically significant, it seems to marginally make a dent on the impact of revolutionary socialism (see Table AII.2). I explore this more in depth in the next section.

Secondly, I replicated the main results using different dependent variables, including earlier cut points for fatality data and log transformations for two different time points: 02/28/2021 and 05/01/2022. In general, the results remain highly consistent, although, as expected, the impact of having a socialist revolutionary experience appears to be relatively stronger for earlier time periods (Tables AII.3 and AII.4).

Next, I introduced controls for stringency measures, as compiled by Hale et al. (Reference Hale, Noam Angrist, Goldszmidt, Kira, Petherick, Phillips, Webster, Cameron-Blake, Hallas, Majumdar and Tatlow2021). The results indicate that, while lockdowns may slightly diminish the impact of socialist revolutionary legacies, they do not entirely cancel them (Table AII.5).

Lastly, I conducted sensitivity analysis by excluding two of the most successful socialist countries in managing the pandemic: China and Vietnam. The results remained essentially the same (Tables AII.6 and AII.7).

Mechanisms

Countries with a legacy of socialist revolutions, regardless of their current socialist status, consistently exhibit lower COVID-related death rates compared to countries without such a legacy. This finding remains robust even when using data from the World Health Organization on “excess deaths” and controlling for additional causal factors. I now shift my focus to investigate the mediating factors through which this legacy exerts an influence on contemporary outcomes. I have identified three key mechanisms: authoritarianism, state capacity, and mass mobilization.

In the absence of experimental data, I turn to mediation analysis to examine these channels (Imai, Keele, Tingley, and Yamamoto Reference Imai, Keele, Tingley and Yamamoto2011; for an application, see Carreras, Irepoglu, and Bowler Reference Carreras, Irepoglu and Bowler2019). Mediation analysis relies on the assumption of sequential ignorability, which posits that (i) given pretreatment confounders, the treatment (in this case, the socialist revolutionary experience) is independent of all potential values of the outcome and mediating factors, and (ii) the observed mediator is independent of the outcome, given the actual treatment status and pretreatment confounders (Imai, Keele, Tingley, and Yamamoto Reference Imai, Keele, Tingley and Yamamoto2011). As this assumption cannot be directly tested with observational data, I control for all pretreatment confounders that were included in the previous models.

Imai, Keele, Tingley, and Yamamoto’s (Reference Imai, Keele, Tingley and Yamamoto2011) method also provides a sensitivity analysis to assess “how severe the violation of the sequential ignorability assumption would have to be for the estimation of the average causal mediation effects (ACME) to be biased” (Carreras, Irepoglu, and Bowler Reference Carreras, Irepoglu and Bowler2019). I will briefly describe the mechanisms and then discuss the results.

Results

The models estimate the predicted values of each of the three mediators (autocracy, state capacity, and social mobilization) based on the exposure variables (socialist revolutionary dummy and socialist duration). These predicted values are then used to estimate the indirect effect (the ACME), the direct effect, and the total effect on COVID-19 death rates. The results are presented in Tables 2 and 3. For the sake of clarity, they exclude the confidence intervals, but the latter can be found in Tables AII.8 and AII.9. The Appendix also includes similar mediation models with the WHO’s dependent variable, reporting very similar results (Tables AII.10 and AII.11).

Table 2. Mediation analysis, with socialist revolutionary regimes as dummy

“Autocracy” models are estimated with logit equations; the three other models were estimated with least squares. The results are computed via the “Medeff” package in Stata (Hicks and Tingley, Reference Hicks and Tingley2011). ACME = Average Causal Mediation Effect.

Table 3. Mediation analysis, with socialist duration

“Autocracy” models are estimated with logit equations; the three other models were estimated with least squares. The results are computed via the “Medeff” package in Stata (Hicks and Tingley, Reference Hicks and Tingley2011). ACME = Average Causal Mediation Effect.

In general, the indicators for the “social mobilization” mechanism show the strongest mediating impact. For instance, 47 percent of the effect of the duration of a socialist revolutionary legacy is mediated by the success of the vaccination rollout process (Table 3, model/column 6, row 4). The non-profit Vaccine Alliance (Gavi) ranked the ten low-income countries with the strongest efforts to distribute vaccines to their populations. Interestingly, two out of the top four countries have socialist revolutionary legacies (Nepal and Nicaragua) (see https://www.gavi.org/vaccineswork/which-lower-income-countries-saw-biggest-covid-19-vaccination-coverage-gains-2022).

The indicators for the “authoritarian” mechanism are next in terms of mediating impact. Socialist revolutionary regimes face fewer constraints on policy making while simultaneously allowing more leeway for experimentation. This could have fed strong incentives for local officials to contain the spread of the pandemic (Aofei et al. Reference Aofei, Luo and Duckett2022). State capacity indicators display, on average, a smaller effect.

A note of caution, however, comes from the rho coefficients reported in the last row of the two tables. This can be interpreted as the degree of correlation required between the errors in the two estimated equations (the mediating and the outcome equations) for the ACME coefficient reported in the table to not be different from 0. In simpler terms, a higher correlation indicates a more robust point estimate. The success at administering the vaccine seems again to offer the more consistent estimate. The models also indicate a non-negligible direct effect of the legacy on pandemic outcomes, suggesting that other mechanisms untapped in this analysis may still exist.

Given that the vaccination rollout only started after one year of pandemic, it is worth asking whether other mechanisms had a larger impact during its first year. I reran all the mediation analyses with February 2021 data on death rates (see Tables AII.12 and AII.13 in the Appendix). The share of vaccinated people is no longer the stronger causal path, as expected. Instead, public health spending and the existence of protections for minorities and women in the country stand out as more relevant factors to account for the impact of socialist revolutions on the early development of the pandemic. Against popular perception, authoritarianism did not seem to be that important during the early stages of the pandemic – at least for socialist revolutionary regimes.