Political Epidemiology of Covid-19 Mortality: A Multilevel Analysis of Macro-Systemic Political Variables

doi:10.21203/rs.3.rs-3235104/v1

Download PDF

Article

Political Epidemiology of Covid-19 Mortality: A Multilevel Analysis of Macro-Systemic Political Variables

https://doi.org/10.21203/rs.3.rs-3235104/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

This study delves into the realm of "political epidemiology" which has gained increasing attention among scholars of political science and public health. It investigates how Covid mortality rate are influenced by macro-systemic political variables such as the level of democracy, systems of internal governance, government effectiveness, and trust in government. Employing multilevel modeling, I analyze countries' trajectories of Covid-19 mortality rates spanning from March 2020 to January 2022. I found that, on average, countries with a federal system tend to exhibit higher Covid-19 death rates compared to those with a unitary system. Moreover, democracy is found to be inversely associated with Covid-19 mortality across countries, although the specific effects of democracy on the trajectory of Covid-19 rates depend on the subgroup of countries considered. Furthermore, government effectiveness emerges as a crucial factor, with a negative association with Covid-19 mortality rates. Countries with more effective governance tend to experience lower death tolls from Covid-19. Additionally, in countries where the population demonstrates higher levels of trust in the government, the curves of Covid-19 death tolls tend to flatten.

Social science/Politics and international relations

Social science/Social policy

1. Politics as Determinants of Public Health Outcome

“Medicine is a social science, and politics nothing but medicine at a larger scale” - the founding father of social medicine and modern cellular pathology, Rudolf Virchow, once famously proclaimed. Inspired by this approach, a field of studies has emerged, considering politics as crucial determinants of health (Mackenbach, 2009, p. 181). As global diseases continue to pose significant challenges, scholars advocates for a more explicit embrace of “health political science”, recognizing that understanding policy processes is essential for influencing policy change (de Leeuw et al., 2014). This interdisciplinary quest, known as “political epidemiology”, has gained urgency among scholars in political science and epidemiology. Unlike conventional research often focuses on social factors such as living conditions, education, healthcare access, and community as key determinants of health, proponents of political epidemiology view political determinants as the underlying causes of the social determinant of health - the determinants of the determinants. These political are far more influential and insidious drivers with significant implications for public health outcomes (Bambra et al., 2005; Dawes, 2020; Kickbusch, 2015; Vu, 2011).

Viewing politics as “the process of making and executing collective decisions”, Mackenbach (2014) identifies three main political variables that impact public health: structures (states, legislative and executive agencies), processes (elections, lobbying, law-making), outputs and outcomes (the laws, taxes, social security benefits, public services). Dawes (2020) further develops a model that integrates three elements of politics that affect health in America: the systematic processes of structuring relationships, distributing resources, and administering power.

The present paper expands upon the above framework to investigate the extent to which macro-systemic political variables relate to health outcomes, specifically focusing on Covid-19 mortality data. It aims to answer the question: To what extent do macro-systemic political factors influence health outcomes, particularly in terms of mortality rates? The macro-systemic political factors under examination are as follows: 1) political regime (level of democracy), 2) institutions (systems of internal governance), 3) state capacity (government effectiveness), and 4) trust in government. These factors are essential indicators of countries’ preparedness in the face of a global health crisis like Covid-19 pandemic.

This research project contributes to the literature on politics and public health outcomes in two distinctive ways. Firstly, while the scholarship primarily focuses on the association between macro-systemic factors and long-term public health outcomes in a non-crisis condition, this study explores the relationships between these factors and health outcome amid a global crisis.

Secondly, whereas recent research has largely addressed the Covid-19 mortality as a dependent variable and focused in specific social policies, such as travel bans, lock-down,s social distancing, masking, surveillance and schools closures, (Aleta & Moreno, 2020; Haug et al., 2020; Liang et al., 2020; Migone, 2020a; Pung et al., 2021), my work delves deeper into the macro-systemic factors that may influence the formulation and implementation of these social policies.

This study seeks to investigate the degree to which countries’ political features contribute to the variability in Covid-19 mortality rates during the initial phase and throughout the period from March 2020 and January 2022. Does the presence of certain political characteristics in countries correspond to different levels of Covid-19 deaths? Are there specific macro-systemic political features that countries have developed over time, equipping them with better abilities to handle this health crisis? By examining the political and governance resources in place before the pandemic, this research aims to identify the types of countries that demonstrated better outcomes in reducing Covid-19 deaths compared to others. More specifically, I ask: Are democracies better at saving lives and whether the type of systems of internal governance influences the outcomes in a global health crisis with millions of lives at stake? Can we expect those with higher score of government effectiveness, an indicator often positively associated with economic development, to be also better at limiting mortality? Do governments with higher level of citizen trust reduce more death rates due to Covid?

I found that democracy has negative main effect on Covid mortality overall, but its effect on the trajectory of Covid rate varied depending on the subgroup of countries considered. The type of internal governance played a significant role, as counties with a federal system tend to experience higher death rates compared to those with a unitary system. On the other hand, government effectiveness emerged as a critical factor that negatively associated with Covid deaths. Additionally, in countries where people reported to have higher trust in government, the curve of death rates tends to be flatter.

This research contributes to the interdisciplinary literature, providing timely insights into the intricate connections between characteristics of political system and public health outcomes. As highlighted in the most recent evaluation of countries’ preparedness for the future global health crisis, it is evident that nations worldwide remain “dangerously unprepared for meeting future epidemic and pandemic threats (The 2021 Global Health Security GHS report).” In light of this, the analysis presented in this study holds direct implications, suggesting that that macro-systemic variables must be accounted for when formulating crisis management strategies. Looking ahead, both affluent and less affluent nations need to reconsider their approaches to safeguarding public health. Policy makers and public health experts should be cognizant of the role of political factors in shaping health outcomes during crises, as demonstrated by this research. By integrating macro-systemic political variables into future preparedness planning and pandemic response strategies, countries can better equip themselves to handle and mitigate the impact of global health threats effectively.

Drawing from the existing literature on macro-systemic political determinants of public health and exploring the underlying mechanisms through which these factors influence health outcomes, this section presents four testable hypotheses:

Democracy

The relationship between democracy and public health has been a subject of extensive debate in the literature. Two mechanisms are argued to contribute to the potential benefits of democracy for public health. Firstly, democratic regimes empower a broader segment of the population to participate in political decision-making, leading to increased responsiveness from politicians (Franco et al., 2004; Ncayiyana, 2004). For instance, child mortality was reduced thanks to the enfranchisement of women in democracy (Miller, 2008), or in counties that underwent democratic transition (Kudamatsu, 2012; Pieters et al., 2016). Secondly, democracy may indirectly enhance public health outcomes by improving socioeconomic factors, such as income, income inequality, education, and social access to health care services (Bollyky et al., 2019; Safaei, 2006).

However, the literature also presents contradictory findings and raises questions about the non-effect of democracy on health or instances where non-democratic regimes may perform better. For example, a cross-national study by Ross (2006) spanning 168 states from 1970 to 2000 found that although democracies spend more resource on education and health than nondemocracies, these benefits often favor middle and upper income groups, leaving the low-income group little or no benefit, thus renders an overall non-effect on infant and child mortality rates. Lin’s (2015) examination of 150 countries between 1995 and 2009 reveals that state capacity, regime type, and their interactions determine the magnitude of human losses during disasters. Specifically, democracies demonstrate greater effectiveness in preventing human losses due to predicable disasters, but not when the causes are unpredictable.

Moreover, certain cases illustrate how non-democratic regimes can effectively manage public health crises. For instance, China's response to the Severe Acute Respiratory Syndrome (SARS) epidemic in 2003 was marked by non-democratic practices, such as government censorship and control of information, which impacted the Chinese population during the early phase of the epidemic (Ruger, 2005). However, the authoritarian Chinese regime effectively implemented draconian quarantine measures, closed public spaces, screened a large segment of the population, and ultimately contained and ended the SARS outbreak (Hesketh & Zhu, 2004). Also, according to these authors, China has historically been successful in eradicating some diseases by aggressive closure of brothels and opium dens by using dictatorship power.

As the world faced the devastating Covid-19 pandemic, which claimed millions of lives, the relationship between democracy and public health was once again subject to urgent scrutiny. In this context, this paper contributes to the extensive empirical literature seeking to identify the effect of democracy on health. I first hypothesize:

Hypothesis 1

Countries with higher level of democracy tend to have lower Covid mortality rates.

Systems of Internal Governance (Formal State Organization)

The system of internal governance, or formal state organization, is the way a state organizes its political power over its territory. Political scientists typically classify the system of internal governance into two categories: federal or unitary. In a federal polity, power is diffused among multiple centers, each with its authority guaranteed by the constitution, fostering a noncentralized structure (Elazar, 1987, p. 34). Approximately forty percent of the global population resides in federal and quasi-federal countries today. Conversely, unitary states concentrate power at the national level and establish subordinate bodies at lower levels.

The debate surrounding the efficiency of power concentration in public management and crisis resolution has led to both advocates and critics of the federalism model. On the one hand, federalism can facilitate good governance thanks to its mechanism of including multiple interests (Palermo & Alber, 2015). Some economists argue that encouraging grassroots initiatives can foster effective disaster recovery (Storr, 2015). The Austrian school of thoughts posits that local communities have the local knowledge that would best help them overcome challenges during crises, making a less centralized power system, polycentric approach beneficial for communities (McGinnis, 1999).

In contrast, federal systems may be perceived as less effective than unitary systems in managing significant national crises due to the structures of shared rule and self-rule potentially hindering rapid, focused, complementary, and coordinated responses – traits are more commonly associated with unitary systems (Kincaid et al., 2010). For instance, during the Covid-19 pandemic, Canada experienced mixed effects of its federalism on emergency management, enabling specific responses to local realities, but potentially misalign critical policy steps (Migone, 2020b). Similarly, Italy’s decentralized system exhibited a slow and incremental response to the fast-moving crisis, indicating low capacity in handling the pandemic (Capano, 2020).

Nevertheless, not all federal states responded to the pandemic in the same way. Hegele and Schnabel (2021) observed considerable variation among federal states along two dimensions of federal decision making: the degree of power centralization and the mechanism of decision-making, whether unilateral or coordinated. Austria and Switzerland adopted a centralized approach to introducing containment measures, while Germany employed a decentralized decision-making approach. Moreover, Austria and Germany utilized coordinated schemes, whereas Switzerland predominantly relied on unilateralism. To further examine the relationships between the systems of internal governance, or the form of state organization, and health outcome in the Covid-19 context, I hypothesize that:

Hypothesis 2

Federal states tend to exhibit lower Covid mortality rates than unitary states.

Government Effectiveness

The influence of governance effectiveness and corruption control on economic development and social equity is widely acknowledged (Arora & Chong, 2018). These factors may manifest in various ways that can have life-saving implication during a public health crisis such as the Covid-19.

Areas where effective governance can make a critical difference include the distribution of urgent healthcare resources, coordination of human resources, and formulation of policies and spending to control the spread of Covid-19. Studies have demonstrated that administrations that were well-prepared structurally, learned from preceding crises, and displayed high-quality network cooperation with other administrations and civil society performed significantly better, on average, during crises like refugee migration and the early phase of the Covid-19 pandemic (Schomaker & Bauer, 2020). Based on these observations, the following hypothesis is put forward:

Hypothesis 3

Countries with higher scores of governance effectiveness tend to have lower Covid mortality rates.

Trust in Government

The notion of citizens’ trust in government has been extensively discussed in political science, particularly concerning its relations to development and prosperity. Fukuyama (1995) introduces a thesis of trust and economics, which suggests that social trust can influence a country’s economic development, leading to a more prosperous economy by lowering transaction costs and enhancing market efficiency. Notably, overall trust in government that is established prior to a crisis can have a lasting effect, shaping more positive perception of the efficiency of public management during a time of crises (Keele, 2007; Khan, 2016). Although several studies have explored the connections between trust in government and resolving social and health problems, these investigations have primarily focuses on routine times rather than crisis situation (He & Ma, 2021; Van de Walle & Bouckaert, 2003).

In the context of the Covid-19 pandemic, trust in government assumes a crucial role, as citizens' belief in the government's efforts to protect public health influences their behavior and support for intervention policies aimed at combating the pandemic (Robinson et al., 2021). That is, higher levels of trust in government lead to increased legitimacy granted to the government and greater willingness among citizens to cooperate in handling the crisis (Mizrahi et al., 2021). Schraff (2020) found that the rising number of Covid-19 cases was associated with increased political trust, while two standard determinants of political trust, economic evaluations and social trust, lost explanatory power. Political trust, as defined by the author, refers to diffuse support for political institutions, encompassing perceptions of their competence and legitimacy, and was measured through a household survey conducted in the Netherlands in March 2020.

In this study, the linkage between the level of trust in government and Covid deaths across countries is examined, leading to the formulation of the fourth hypothesis:

Hypothesis 4

Countries with higher level of trust in government tend to have lower Covid mortality rates.

The data utilized in this analysis is derived from a compilation of reputable public sources, including Our World in Data, the Economist Intelligence Unit Democracy Index, The World Bank’s World Development Indicators, the OECD Trust in Government Indicator, and the Global Health Security Index. The dataset comprises repeated measurements, where the Covid mortality rates calculated at each time point represent the varying component. Conversely, all other covariates are regarded as time-invariant. Except Covid mortality data, all other variables are for year 2019. The deliberate choice of 2019 data, as opposed to earlier or later years, aligns with the study’s specific objective: to investigate how pre-existing political characteristics impact countries’ capacity to respond to the Covid-19 pandemic.

Table 1

Descriptive statistics
Statistic	N	Mean	St. Dev.
total_deaths_per_million	3,384	480.46	769.77
gdp_percap	3,384	15,251	20,893
democracy	3,384	56.04	21.75
gov_eff	3,339	0.02	0.98
health_index	3,315	44.01	14.18
trust_gov	873	49.89	16.22
unitary_federal	3,384	0.14	0.35
unitary_semi_federal	3,384	0.22	0.41
Notes: Number of observations (N) is the sum of the product of the number of the time point measurements each country has and the number of countries. For three variables gdp_percap, democracy, unitary_federal, the sample consists of 151 countries, of which 115 countries have 23 time point measurements, 20 countries have 22, 7 countries have 21, 6 have 19, the last 3 countries each has 14, 13, and 11 points, respectively, hence 11523 + 2022 + 721 + 619 + 14 + 13 + 11 = 3384 total points of measurements. The same method of calculation is applied for other variables.

Outcome Variable: Covid-19 Mortality Rates

The Covid mortality rate data was retrieved from Our World in Data in February 2022, encompassing 3384 observations of monthly Covid deaths per million population. The dataset spans from April 1, 2020, to February 1, 2022. Although the original data has measurements of January and February 2020, I do not include these two months because there are too few observations during this time. For the purpose of this analysis, the measurements for January and February 2020 were excluded due to limited observations during this period. Consequently, the data is partitioned into 23 monthly time points, with month 0 commencing on April 1, 2020, and month 22 concluding on February 1, 2022.

Time-invariant Predictors of Covid-19 Mortality

Democracy Index

The EIU Democracy Index assesses the quality of democracy in 167 countries, assigning continuous scores ranging from 1 to 100. This comprehensive index is derived from five distinct categories: electoral process and pluralism, political participation, the functioning of government, civil liberties, and political culture. For this analysis, the specific value utilized is the "Democracy index (EIU)" for the year 2019. The choice to employ the EIU measurement over alternatives, such as Polity V, is driven by the EIU's continuous scale, which provides a more nuanced evaluation of the level of democracy in each country.

Systems of Internal Governance

In this study, two distinct methods were accounted to categorize systems of internal government. The first method identified 26 federal states in the world, resulting in 22 federal states included in the dataset. The second method extended this classification to encompass an additional 11 states that have devolved governance power structure, regionalized unitary, and federacy. In these semi-federal states, central governments delegate a portion of governance power to regional and local level governments. Consequently, this second approach recognizes a total of 37 federal states worldwide, resulting in 32 federal states in my dataset. It is worth noting China poses a unique case. Despite having two semi-autonomous regions, Hong Kong and Macau, which have some economic autonomy, China was coded as a unitary state in this study. This coding decision reflects the fact that China maintains highly centralized power of pandemic control. The distinction in coding systems of internal governance resulted in varying outcomes, which will be presented in the subsequent analysis section.

Government Effectiveness

Government effectiveness is gauged through an indicator featured in The World Bank's World Development Indicators, which spans a range from − 2.5 (indicating lower effectiveness) to 2.5 (indicating higher effectiveness). This multifaceted indicator assesses the quality of public services, civil service, policy formulation, policy implementation, and the credibility of a government's commitment to its policies. The data for this analysis was retrieved from the World Bank's "data360r" data package, specifically using the indicator "Government Effectiveness, Estimate" for the year 2019, encompassing data from 190 countries.

Trust in Government

Trust in government data is collected by the OECD from 40 advanced industrial countries. However, I am not aware of an equivalent measurement for trust in government in other countries. This limitation results in a reduction in the overall sample size when estimating the impact of trust in government on Covid deaths. Despite this limitation, it is important to note that countries within this subset tend to exhibit higher scores in all other covariates, such as democratic index scores, GDP per capita, and Health Security Index. Nevertheless, it is argued that the sample size of 38 data clusters (countries) with 23 repeated measurements still provides sufficient statistical power to estimate the effect of each covariate.

Controlling Factors

To discern the specific influence of political factors, this analysis takes into account potential confounding variables that could impact the relationships between the explanatory factors and the outcome of death tolls. Two controlling variables considered in this study are public health infrastructure, as measured by the Global Health Security Index, and the level of national wealth, represented by GDP per capita.

The Global Health Security Index serves as a comprehensive assessment and benchmarking tool for evaluating health security and related capabilities across the 195 countries that constitute the States Parties to the International Health Regulations (IHR [2005]). The index employs a scoring system ranging from 1 to 100. For this analysis, the data used is the "Global Health Security Index" for the year 2019.

To gauge the wealth level of countries, GDP per capita is employed as a proxy. By controlling for the wealth level, this study aims to isolate the impact of political determinants on Covid deaths, allowing for a clearer understanding of their effects independent of other material factors. The GDP per capita data is collected by the World Bank and was retrieved from the R package "data360r" for the year 2019.

In this analysis, a multilevel method is employed, as it proves more suitable than a traditional general linear model (GLM) for two compelling reasons. Firstly, the data with repeated measurements, where each country unit forms a data cluster, violates the fundamental assumption of GLM models regarding residual independence and identical distribution. Secondly, the research question at hand inherently involves a multilevel structure, focusing on how predictors at level-2 (countries' characteristics) influence the outcome at level-1 (measurements at different time points). A multilevel approach offers distinct advantages in handling clustered data and addressing such research inquiries.

Multilevel modeling provides flexibility in representing time, rendering it particularly advantageous for longitudinal data analysis. This modeling approach can effectively accommodate variations in the number of measurement occasions between countries, allowing for a robust examination of the research question (Raudenbush, 2002).

Level 1 Variables

At level 1, the variables include unique country IDs, monthly Covid mortality rates per million population, and time points. At level 2, the country-level political characteristics are incorporated, encompassing the level of democracy, the form of internal governance system (a binary covariate with federal states coded as 1 and unitary states as 0), the level of governance effectiveness, and the level of trust in government. To facilitate interpretation, the continuous measurements of democracy, government effectiveness, and trust in government were standardized during model fitting. The analysis also controls for two potential confounders at level 2: GDP per capita and the Health Security Index.

The modeling process involved three steps. Firstly, an unconditional growth model was applied to serve as a baseline for comparison with subsequent models. In the second step, conditional multilevel growth models were employed to investigate model results. Lastly, country characteristics were added as additional predictors of Covid mortality. The two conditional models consisted of 1) a level-1 model specifying Covid mortality rates as predicted by a quadratic function of time of measurement, and 2) a level-2 model determining the prediction of level-1 model parameters based on country characteristics. The initial level-1 model can be described as follows:

𝑀𝑜𝑟𝑡𝑎𝑙𝑖𝑡𝑦 𝑡𝑖 = 𝛽0𝑖 + 𝛽1𝑖 (𝑡𝑖𝑚𝑒) + 𝛽2𝑖 (𝑡𝑖𝑚𝑒 𝑠𝑞𝑢𝑎𝑟𝑒𝑑) + 𝑟𝑡𝑖

Where 𝑀𝑜𝑟𝑡𝑎𝑙𝑖𝑡𝑦 is the mortality rates for country 𝑖 at time 𝑡. 𝛽0𝑖 is the initial status or intercept for country 𝑖 at time 0, 𝛽1𝑖 is the linear rate of change, 𝛽2𝑖 is the curvature that represents the acceleration or deceleration in each country’s Covid deaths growth trajectory. The term 𝑟𝑖𝑗 represents the residual for each country. A quadratic model was chosen based on data inspection and statistical testing of alternative growth functions, which demonstrated a significant improvement in model fit compared to the linear model (Chisq = 267.12, p < 0.001).

Level 2 Variables

At level-2, the level-1 parameters are modeled by incorporating countries' characteristics. The dichotomous predictor "unitary_federal" is uncentered, while all other continuous variables are standardized. Consequently, the coefficient for the dichotomous predictor represents the effect for the group coded as 1, which pertains to countries with a federal form of state organization. On the other hand, the coefficients for continuous predictors (democracy, government effectiveness, trust in government) represent the effect for a country with an average score in that specific predictor. The level-2 equations for the mortality initial status and growth rate parameters are as follows:

Initial Status, 𝛽0𝑖 = 𝛾00 + ∑𝛾0𝑘(𝑃𝑟𝑒𝑑𝑖𝑐𝑡𝑜𝑟 𝑘) + 𝑢0𝑖

Linear Rate of Change, 𝛽1𝑖 = 𝛾10 + ∑𝛾1𝑘(𝑃𝑟𝑒𝑑𝑖𝑐𝑡𝑜𝑟 𝑘) + 𝑢1𝑖

Curvature, 𝛽2𝑖 = 𝛾10 + ∑𝛾2𝑘(𝑃𝑟𝑒𝑑𝑖𝑐𝑡𝑜𝑟 𝑘) + 𝑢2𝑖

In these equations, 𝛾00 represents the Covid mortality rates intercept at the first time point in the dataset, i.e., April 1, 2020, for all countries. Each 𝛾0𝑘 represents the average partial regression coefficient relating the predictor of interest to the country's initial status. The term 𝑢0𝑖 represents the residual between the fitted predictor value for each country and the observed Covid mortality rates at the first time point. For each rate of change parameter (i.e., 𝛽1𝑖 and 𝛽2𝑖), the rate of change for each country, 𝛽p𝑖, is modeled as a function of the average Covid mortality rates of change, 𝛾pk. Each 𝛾pk represents the average partial regression coefficient relating the predictor of interest to the country's rate of change. The term 𝑢p𝑖 is the residual between the fitted predictor value for each country's rate of change and the observed rate of change.

To detect multicollinearity, a potential problem in which a predictor is already explained by other predictors, variance inflation factor (VIF) was calculated. All VIF values were found to be smaller than 10, a commonly recommended limit indicating acceptable levels of multicollinearity.

Average Observed Growth of Covid Mortality by Country

First, I conduct an observational analysis (Figure. 1), which shows that most typical trajectories are curvilinear in form, exhibiting accelerating growth across time points. All countries show increasing variability in Covid mortality over time. Notably, five countries that have highest mortality rates at initial status in 2020 (Belgium, Spain, Italy, Netherlands, and France) have relatively low slopes. On the other hand, the top five countries with the highest mortality rates at the ending point (Peru, Bulgaria, Bosnia and Herzegovina, Hungary, and Montenegro) exhibit relatively higher slopes than other countries. Despite some variation, all countries demonstrate positive growth over time, with notable differences in both linear rate of change and rate of curvature. These observations will be tested using multilevel statistical models in the next section.

Partitioning Within- and Between-country Components

To further analyze the data, the variance in the outcome variable, Covid mortality per million population, is partitioned into its within- and between-country components. The sample consists of 151 countries, with a total of 3384 points of measurement. There is some degree of imbalance in group sizes, as the number of monthly data point measurements for each country ranges from 11 to 23. However, this imbalance can be addressed effectively using multilevel modeling, where the maximum likelihood estimator accounts for variations in group size.

The fixed effect for the intercept indicates that the grand mean of Covid mortality rates per million population is 472.48 for the pooled data (across all time points in all countries). The level-1 residual variability is 299012, while the level-2 random intercept variance is 291718. The intra-class correlation (ICC) is calculated as 291718 / (291718 + 299012) = 0.494, indicating that approximately 49.4% of the variance in Covid mortality is associated with between-country differences. In other words, within a given country, Covid mortality rates have a meaningful correlation of 0.494. These results suggest that there is significant dependence in the Covid mortality per million population rates that requires further explanation based on countries' characteristics.

Linear and Quadratic Growth Curve Model

The estimated random effects reveal that the variance among the intercepts is 40216. When accounting for the time factor, this represents a substantial reduction (70%) compared to the unconditional model discussed earlier, where the variance associated with countries was 291718. Similarly, the variance among the slopes is 3297, and the level-1 residual variance is 20943. These results indicate variability among the country intercepts and slopes. Additionally, the random effects of the intercepts and slopes show a negative correlation of -0.53, suggesting that countries with high death tolls at the early phase tend to experience slower increases over time, while countries with lower death tolls tend to experience faster increases. Even after accounting for time, there remains significant variability at level 1. The fixed effects analysis reveals a significant mean linear slope (𝛾 = 52.02, sd = 4.68, p < 0.001), indicating an average increase of 52 deaths per month. However, the intercept with a negative value is not meaningful in this context and can be disregarded.

Moreover, the results from the quadratic growth curve model provide strong support for the inclusion of a quadratic component in the analysis. The fixed quadratic effect of 0.78 is statistically significant (sd = 0.06, p < 0.001). Furthermore, the formal test shows a significant improvement in model fit with the inclusion of the quadratic component (Chisq = 148.4, p < 0.001). Therefore, the quadratic term will be included in the succeeding models to better capture the data's curvilinear nature.

Systems of Internal Governance

Starting from this step, I expand the model to include level-2 predictors. The first predictor added is the system of internal governance of a country, i.e., a federal or unitary system, while controlling for GDP per capita (log) and health security index (standardized). Two methods of categorizing federal states, as mentioned in section 3.2.2, result in two distinct estimate outputs. Using the first method that identifies 22 states in the dataset as federals, no significant effect of the system of internal governance on Covid mortality is observed. However, when using the second method that recognizes 32 states in the dataset as federal, the system of internal governance becomes a significant factor associated with higher Covid deaths, even when controlling for GDP per capita (log) and health security index.

Regarding the fixed effect results, the main effect of internal governance system (federal coded 1) is significant. The model implies that federal states have an intercept higher than unitary states by 92.8 Covid deaths per million population (t = 2.42, SE = 38.2, p = 0.016). However, when including the cross-level interaction between time and federal, the regression of trajectory slopes on federal is not significant. This indicates that the model-implied slopes do not differ in magnitude between federal and unitary states. The figure below visually illustrates that the intercepts of the trajectory of Covid mortality are significantly higher for federal states compared to unitary states, while the slopes of the trajectories do not significantly differ from each other.

Democracy

In this section I added democracy as an additional predictor at level 2 to examine its effect on Covid mortality, while controlling for GDP per capita and Health Security Index. The original scale of democracy index is a continuous, with a 1-point incremental increase. To have a more meaningful interpretation, I standardized democracy.

M𝑜𝑟𝑡𝑎𝑙𝑖𝑡𝑦 𝑡𝑖 = (𝛾00 + 𝛾01𝑡𝑖𝑚𝑒 + 𝛾0Democracy) +

(𝛾01∗𝑡𝑖𝑚𝑒 + 𝛾0 ∗ Democracy) + (𝑢0𝑖 + 𝑢1𝑖 Democracy + 𝑟𝑡𝑖)

The fixed effects indicate significance in both main effect and interaction effect of democracy. Countries with higher democracy score have lower intercepts and steeper regression slopes. One standard deviation in the democracy score is associated with 63.89 less Covid deaths per million population (p = .004). The significant interaction between time and democracy (𝛾^ = 21.78) shows that the slope of the trajectory of Covid mortality increases by approximately 21.78 units for each one standard deviation increase in the predictor democracy (p < 0.001).

The plot above provides further insights into the relationships between democracy and Covid mortality. It displays three lines representing the democracy values at the mean, one standard deviation above the mean and one standard deviation below the mean. Notably, one standard deviation of democracy index in the dataset is equivalent to 21.75 points.

The plot reveals that countries with higher democracy scores exhibit lower number of Covid deaths per million population at initial time point. However, as time progresses, their mortality rates increase at a quicker pace. On the other hand, countries with lower democracy index scores started out with higher death rates, but their mortality rates increase at a slower rate overtime. In other words, higher democratic countries had lower death rates early in the pandemic, but as time went on, their mortality rates increased faster compared to those of less democratic countries. This trend becomes apparent starting at around time point 3, which corresponds to July 2020. From that point onwards, higher democratic countries experienced a greater impact from Covid deaths in comparison to their less democratic counterparts.

Government effectiveness

Subsequently, government effectiveness was added to the model as an additional predictor, and it was also standardized for ease of interpretation. The results of the model highlight the significant impact of government effectiveness on Covid mortality. Specifically, one standard deviation increase in government effectiveness was associated with 68.77 fewer Covid deaths (p = 0.043).

Additionally, a model with a cross-level interaction term between the slope (time) and government effectiveness was examined, but it yielded no significant effect of this predictor. This suggests that while government effectiveness plays a crucial role in influencing the overall Covid mortality rates, it does not significantly influence the trajectory of mortality rates over time.

Trust in Government

Next, trust in government was included as an additional predictor, and it was found to have a negative effect on the slope of Covid mortality. For each standard deviation increase in trust in government, the slope decreased by 1.78 units (p = 0.026). It is important to note that the model now consists of 873 observations from 38 countries, as opposed to the previous 147 countries, due to the limited availability of the indicator trust in government. Consequently, caution should be exercised when comparing variance estimates in the random effects with the previous models. Both the between-country variance (31278) and within-country variance (1874) were reduced in this smaller sample size. Additionally, the correlation between time in months and the intercept, -0.46, is smaller than in the previous models.

Interestingly, in this smaller sample size, democracy shows a significant negative effect on the slope of the Covid trajectory, which is in the opposite direction compared to the previous model. Specifically, one standard deviation increase in democracy is associated with a 2.78 lower unit in the rate of change of Covid mortality. This difference in the effect of democracy on Covid mortality may be attributed to the distinctiveness in the scores in the two groups of countries. In the smaller sample of 38 countries, democracy is much less varied, ranging from 31.1 (Russia) to 98.7 (Norway), with a mean of 78.18 and standard deviation of 13.61. This is in contrast to the score in the larger sample size of 148 countries, which ranges from 13.2 to 98.7, with a mean of 56.04 and standard deviation of 21.75. Essentially, the subgroup of countries classified as democracies in this model comprises those with much higher levels of democracy than those in the larger sample.

In this joint model using the smaller sample size, the government effectiveness factor maintains its negative main effect on Covid deaths. That is, one standard deviation increase in government effectiveness is associated with 2.67 fewer deaths (p = 0.025). The table below summarizes the results of hypotheses testing.

Table 2

Analysis result
Variables	Model 1	Model 2	Model 3	Model 4
time_month	34.644***(5.01)	34.800***(4.69)	34.912***(4.714)	109.285***(15.06)
I(time_month^2)	0.793***(0.065)	0.794***(0.065)	0.797***(0.065)	0.014(0.146)
GDP per capita (log)	2.683(13.345)	5.907(14.172)	31.251(18.718)	228.876**(80.442)
Health Security Index	2.682(1.494)	3.106(1.611)	4.680**(1.753)	5.753(3.922)
Federal system (Yes = 1)	92.857*(38.272)	93.612*(38.349)	87.149*(38.956)	159.318**(61.088)
scale(democracy)		-63.891**(22.2)	-50.582*(23.645)	153.128*(77.777)
time_month:scale(democracy)		21.777***(4.46)	21.581***(4.482)	-27.869*(12.216)
scale(gov_eff)			-68.772*(33.759)	-266.969*(113.229)
scale(trust_gov)				-9.196 (37.384)
time_month:scale(trust_gov)				-17.894*(7.640)
Num.Obs.	3315	3315	3293	873
RMSE	136.50	136.50	136.94	161.43
* p < 0.05, p < 0.01, * p < 0.001 Notes: Standard errors shown in parentheses.

Table 3

The study's hypotheses and summary of findings
H1	Countries with higher level of democracy tend to have lower Covid mortality rates.	Partially supported by the data. In the sample of 148 countries, countries with higher level of democracy have lower death rates at the beginning, but the death tolls tend to increase faster than those with lower level of democracy. In the group of 38 OECD countries, those in higher democracy experience slower rate of Covid deaths.
H2	Federal states tend to have lower Covid mortality rates than unitary states.	Data shows an opposite effect. Federal states tend to have higher rate of Covid deaths compared to those in unitary system.
H3	Countries with higher score of governance effectiveness tend to have lower Covid mortality rates.	Supported by the data. Government effectiveness significantly associated with lower rate of Covid deaths.
H4	Countries with higher level of trust in government tend to have lower Covid mortality rates.	Supported by the data. In the groups of 38 OECD countries, those with higher level of trust in government experience significantly slower rate of Covid deaths.

The results of the analysis presented in this study both support and challenge some findings in the existing literature on the political determinants of public health. Contrary to the argument that a more decentralized power structure would enable local entrepreneurship and community revival after a disaster, I found an opposite effect of power decentralization in relation to Covid deaths. That is, states under a federal system tend to have higher Covid deaths.

Democracy has a negative main effect on Covid mortality overall, indicating that higher democracies experienced fewer Covid deaths on average. However, the effects of democracy on the trajectory of Covid rate increase over time differ depending on the subgroup of countries considered. In the group of 148 countries, democracy shows a positive relationship with the trend of Covid deaths, meaning that higher democracies also tend to have a higher increase in death rates. In the smaller group of 38 OECD countries, democracy shows a negative relationship with the slope of Covid deaths, indicating that countries with advanced democracy witnessed slower rates of Covid deaths over time.

Two other factors, government effectiveness and trust in government, are statistically significant predictors of Covid mortality. Government effectiveness persists as a significant factor that is negatively associated with Covid deaths. This result aligns with Liang et al. (2020) who also found a negative association between government effectiveness and Covid-19 mortality during the early phase of the pandemic. Trust in government is an important factor that is negatively associated with the trajectory of Covid deaths. In countries where people reported higher trust in government, the curve of death tends to be flatter.

This research contributes to the emergent studies that examine the relationships between major political and governance features and public health outcomes. The paper establishes a correlation, not causal relationship, between macro-systemic factors and a specific public health outcome, Covid-19 mortality rates. Nevertheless, by showing how countries' political characteristics relate to the consequences of Covid deaths, the paper strongly suggests that politics matters a great deal in crisis management. It also supports the notion of politics being "the determinants of the determinants" developed by Dawes (2020) as introduced at the beginning of this paper.

A major part of political science works focuses directly on the impact of political factors on the economy, which can be indirectly inferred to population health. A better economy, arguably, can lead to better public health outcomes. However, as witnessed in the current pandemic, richer countries are not necessarily better at saving more lives. As this paper points out, political and governance factors play a significant role in limiting deaths. My work is an effort to establish an association between macro-system political factors and Covid mortality. Further examinations in terms of causal inferences will be needed to deepen our understanding of politics and public health.

Aleta, A., & Moreno, Y. (2020). Evaluation of the potential incidence of COVID-19 and effectiveness of containment measures in Spain: a data-driven approach. BMC medicine, 18(1), 157-157. https://doi.org/10.1186/s12916-020-01619-5
Arora, P., & Chong, A. (2018). Government effectiveness in the provision of public goods: the role of institutional quality. Journal of applied economics, 21(1), 175-196. https://doi.org/10.1080/15140326.2018.1550593
Bambra, C., Fox, D., & Scott-Samuel, A. (2005). Towards a politics of health. Health promotion international, 20(2), 187-193. https://doi.org/10.1093/heapro/dah608
Bollyky, T. J., Templin, T., Cohen, M., Schoder, D., Dieleman, J. L., & Wigley, S. (2019). The relationships between democratic experience, adult health, and cause-specific mortality in 170 countries between 1980 and 2016: an observational analysis. The Lancet (British edition), 393(10181), 1628-1640. https://doi.org/10.1016/S0140-6736(19)30235-1
Capano, G. (2020). Policy design and state capacity in the COVID-19 emergency in Italy: if you are not prepared for the (un)expected, you can be only what you already are. Policy & society, 39(3), 326-344. https://doi.org/10.1080/14494035.2020.1783790
Dawes, D. E. (2020). The political determinants of health. Baltimore, Maryland : Johns Hopkins University Press.
de Leeuw, E., Clavier, C., & Breton, E. (2014). Health policy - why research it and how: health political science. Health research policy and systems, 12(1), 55-55. https://doi.org/10.1186/1478-4505-12-55
Elazar, D. J. (1987). Exploring federalism. Tuscaloosa, AL : University of Alabama Press.
Franco, Á., Álvarez-Dardet, C., & Ruiz, M. T. (2004). Effect of democracy on health: ecological study. BMJ, 329(7480), 1421-1423. https://doi.org/10.1136/bmj.329.7480.1421
Fukuyama, F. (1995). Trust : the social virtues and the creation of prosperity. New York : Free Press.
Haug, N., Geyrhofer, L., Londei, A., Dervic, E., Desvars-Larrive, A., Loreto, V., . . . Klimek, P. (2020). Ranking the effectiveness of worldwide COVID-19 government interventions. Nature human behaviour, 4(12), 1303-1312. https://doi.org/10.1038/s41562-020-01009-0
He, A. J., & Ma, L. (2021). Citizen Participation, Perceived Public Service Performance, and Trust in Government: Evidence from Health Policy Reforms in Hong Kong. Public performance & management review, 44(3), 471-493. https://doi.org/10.1080/15309576.2020.1780138
Hegele, Y., & Schnabel, J. (2021). Federalism and the management of the COVID-19 crisis: centralisation, decentralisation and (non-)coordination. West European politics, 1-20. https://doi.org/10.1080/01402382.2021.1873529
Hesketh, T., & Zhu, W. X. (2004). Effect of restricted freedom on health in China. BMJ, 329(7480), 1427-1427. https://doi.org/10.1136/bmj.329.7480.1427
Keele, L. (2007). Social Capital and the Dynamics of Trust in Government. American journal of political science, 51(2), 241-254. https://doi.org/10.1111/j.1540-5907.2007.00248.x
Khan, H. A. (2016). The idea of good governance and the politics of the global South : an analysis of its effects. New York : Routledge, is an imprint of the Taylor & Francis Group, an Informa business.
Kickbusch, I. (2015). The political determinants of health—10 years on. BMJ (Online), 350(jan08 2), h81-h81. https://doi.org/10.1136/bmj.h81
Kincaid, J., Tarr, G. A., & Wälti, S. (2010). Federalism and the Global Financial Crisis: Impacts and Responses: Introduction. Europe en formation, n° 358(4), 3-14. https://doi.org/10.3917/eufor.358.0003
Kudamatsu, M. (2012). HAS DEMOCRATIZATION REDUCED INFANT MORTALITY IN SUB-SAHARAN AFRICA? EVIDENCE FROM MICRO DATA. Journal of the European Economic Association, 10(6), 1294-1317. https://doi.org/10.1111/j.1542-4774.2012.01092.x
Liang, L.-L., Tseng, C.-H., Ho, H. J., & Wu, C.-Y. (2020). Covid-19 mortality is negatively associated with test number and government effectiveness. Scientific reports, 10(1), 12567-12567. https://doi.org/10.1038/s41598-020-68862-x
Lin, T.-H. (2015). Governing Natural Disasters: State Capacity, Democracy, and Human Vulnerability. Social forces, 93(3), 1267-1300. https://doi.org/10.1093/sf/sou104
Mackenbach, J. P. (2009). Politics is nothing but medicine at a larger scale: reflections on public health’s biggest idea. Journal of epidemiology and community health (1979), 63(3), 181-184. https://doi.org/10.1136/jech.2008.077032
Mackenbach, J. P. (2014). Political determinants of health. European journal of public health, 24(1), 2-2. https://doi.org/10.1093/eurpub/ckt183
McGinnis, M. D. (1999). Polycentric governance and development : readings from the Workshop in Political Theory and Policy Analysis. Ann Arbor, MI : University of Michigan Press.
Migone, A. R. (2020a). The influence of national policy characteristics on COVID-19 containment policies: a comparative analysis. Policy design and practice, 3(3), 259-276. https://doi.org/10.1080/25741292.2020.1804660
Migone, A. R. (2020b). Trust, but customize: federalism's impact on the Canadian COVID-19 response. Policy & society, 39(3), 382-402. https://doi.org/10.1080/14494035.2020.1783788
Miller, G. (2008). Women's Suffrage, Political Responsiveness, and Child Survival in American History. The Quarterly journal of economics, 123(3), 1287-1327. https://doi.org/10.1162/qjec.2008.123.3.1287
Mizrahi, S., Vigoda‐Gadot, E., & Cohen, N. (2021). How Well Do They Manage a Crisis? The Government's Effectiveness During the COVID‐19 Pandemic. Public administration review, 81(6), 1120-1130. https://doi.org/10.1111/puar.13370
Ncayiyana, D. J. (2004). Is democracy good for people's health? A South African perspective. BMJ, 329(7480), 1425-1426. https://doi.org/10.1136/bmj.329.7480.1425
Palermo, F., & Alber, E. (2015). Federalism as decision-making : changes in structures, procedures and policies. Leiden, Netherlands ; Boston, Massachusetts : Brill Nijhoff.
Pieters, H. D., Curzi, D. P., Olper, A. P., & Swinnen, J. P. (2016). Effect of democratic reforms on child mortality: a synthetic control analysis. The Lancet global health, 4(9), e627-e632. https://doi.org/10.1016/S2214-109X(16)30104-8
Pung, R., Cook, A. R., Chiew, C. J., Clapham, H. E., Sun, Y., Li, Z., . . . Lee, V. J. (2021). Effectiveness of Containment Measures Against COVID-19 in Singapore: Implications for Other National Containment Efforts. Epidemiology (Cambridge, Mass.), 32(1), 79-86. https://doi.org/10.1097/EDE.0000000000001257
Raudenbush, S. W. (2002). Hierarchical linear models : applications and data analysis methods (2nd ed. ed.). Thousand Oaks : Sage Publications.
Robinson, S. E., Ripberger, J. T., Gupta, K., Ross, J. A., Fox, A. S., Jenkins‐Smith, H. C., & Silva, C. L. (2021). The Relevance and Operations of Political Trust in the COVID‐19 Pandemic. Public administration review, 81(6), 1110-1119. https://doi.org/10.1111/puar.13333
Ross, M. (2006). Is Democracy Good for the Poor. American journal of political science, 50(4), 860-874. https://doi.org/10.1111/j.1540-5907.2006.00220.x
Ruger, J. P. (2005). Democracy and health. QJM : monthly journal of the Association of Physicians, 98(4), 299-304. https://doi.org/10.1093/qjmed/hci042
Safaei, J. (2006). IS DEMOCRACY GOOD FOR HEALTH? International journal of health services, 36(4), 767-786. https://doi.org/10.2190/6V5W-0N36-AQNF-GPD1
Schomaker, R. M., & Bauer, M. W. (2020). What Drives Successful Administrative Performance during Crises? Lessons from Refugee Migration and the Covid‐19 Pandemic. Public administration review, 80(5), 845-850. https://doi.org/10.1111/puar.13280
Schraff, D. (2020). Political trust during the Covid-19 pandemic: Rally around the flag or lockdown effects? European journal of political research. https://doi.org/10.1111/1475-6765.12425
Storr, V. H. (2015). Community revival in the wake of disaster : lessons in local entrepreneurship. Houndmills, Basingstoke, Hampshire ; New York, NY : Palgrave Macmillan.
Van de Walle, S., & Bouckaert, G. (2003). Public Service Performance and Trust in Government: The Problem of Causality. International journal of public administration, 26(8-9), 891-913. https://doi.org/10.1081/PAD-120019352
Vu, T. (2011). Epidemics as Politics with Case Studies from Malaysia, Thailand, and Vietnam. Global Health Governance, IV(2).

https://www.ghsindex.org/report-model/
https://github.com/owid/covid-19-data/tree/master/public/data/excess_mortality
https://www.gapminder.org/data/documentation/democracy-index/
https://www.eiu.com/topic/democracy-index
https://cs.mcgill.ca/~rwest/wikispeedia/wpcd/wp/l/List_of_countries_by_system_of_government.htm
OECD (2022), Trust in government (indicator). doi: 10.1787/1de9675e-en (Accessed on 14 February 2022) https://data.oecd.org/gga/trust-in-government.htm
https://www.ghsindex.org/about/
There are 115 countries (76.2% of all countries) that have full 23 time point measurements, 20 countries have 22 measurements, 7 countries have 21 points, 6 countries have 19 points, one has 14, one has 13, and the last one has 11 time point measurements.

There is NO Competing Interest.

Download PDF

Version 1

posted

You are reading this latest preprint version

Political Epidemiology of Covid-19 Mortality: A Multilevel Analysis of Macro-Systemic Political Variables

Status:

Version 1

Abstract

Figures

1. Politics as Determinants of Public Health Outcome

2. The Impact Mechanisms and Hypotheses

3. Data

4. Analysis

5. Result

6. Conclusion

References

Footnotes

Additional Declarations

Status:

Version 1