COVID-19 Cases and Vaccination Inequality: A Comparative Analysis of Political Regimes

Chinglen Laishram ( is a Ph.D. Scholar at Centre for Studies in Society and Development, School of Social Sciences, Central University of Gujarat, Gujarat, India and Pawan Kumar ( teaches at School of Law, BS Abdur Rahman Crescent Institute of Science and Technology, Chennai, India.
31 October 2022

Different regimes have different capacities to respond to pandemics. Historically, democracies outperformed autocracies in health outcomes. However, the COVID-19 pandemic exposed the shortcomings, with a sharper tone, of full democracies (having higher COVID-19 cases than authoritarian regimes) and led to the formation of two competing hypotheses among the cross-national comparative political researchers: (i) biasing autocracy: that authoritarian regime manipulated and underreported COVID-19 cases, and (ii) efficient autocracy: that authoritarian regimes can control the spread of the disease effectively than democracies. We examined these two hypotheses, employing Benford’s test and generalised linear models, using the latest data set from the World Health Organization, EIU, United Nations, and other relevant sources. Findings include having no empirical support for the biasing hypothesis. However, the efficient autocracy hypothesis acquired partial empirical support. We further examined the data on COVID-19 vaccination for reliability (using Benford’s test), and the results indicated a potential case of data manipulation.

The World Health Organization (WHO) declared COVID-19 a global pandemic in the first quarter of 2020. The reported number of COVID-19-related cases and deaths has exceeded 600 million and 6 million, respectively (WHO 2022). The numbers differed across various countries as different countries have different population sizes (Laishram and Kumar 2021). Apart from a country’s population being a factor driving the variations in COVID-19 cases and deaths, various other factors have also been identified to date. Among these, a unique factor that has acquired particular salience among scholars of cross-national comparative political research is “political regimes.” This is because, though differences across the political regimes may be contingent to other factors, authoritarian regimes had lower cases and deaths than democratic regimes (Cepaluni et al 2022; Karabulut et al 2021; Laishram and Kumar 2021). Determining the variations in COVID-19 cases and deaths using political regimes is intuitively sensible as different forms of government have different capacities in terms of implementation of COVID-19-related policies and also in terms of the development of public health infrastructures.

The observation that authoritarian regimes have lower cases and deaths than democratic regimes led to the formation of, among others, two main hypotheses: (i) biasing autocracy and (ii) efficient autocracy. First, the biasing autocracy posits that authoritarian regimes manipulated and underreported COVID-19 cases (Cassan and Steenvoort 2021). This hypothesis emerged particularly from those studies that checked conformity to the first digit law (or Benford’s law) across different countries in the world, and China’s COVID-19 data having indications to violate this law (Peng and Nagata 2020). Benford’s law describes the probability distribution for the likelihood of the first leading digits (1–9) in a set of naturally occurring collections of numerical observations such that the most frequent first digit in a set of numbers is “1” for about 30% of the observations, followed by observations of “2” (about 17.5%), “3” (about 12.5%), and so on in a successively decreasing manner down to “9” (about 4.5%) (Frunza 2015; Glen 2022; Sarkar 2018). Numeric data samples that do not conform to this law are considered “fictitious or artificially manipulated” (Frunza 2015: 235). This law has broad applications in various fields of study, including geosciences (Sottili et al. 2012), auditing and taxation (Watrin et al. 2008), astronomy (Alexopoulos and Leontsinis 2014), and financial management (Cella and Zanolla 2018), among various others. Scholars have also highlighted the application of this law to evaluate public health and epidemiological data (Idrovo et al. 2011; Koch and Okamura 2020). The study by Miranda (2020), for instance, tested whether the COVID-19 data (both cases and deaths) of the Philippines conform to Benford’s law. In the context of political regime, Kilani’s (2021) study used a data set extracted from the European Centre for Disease Prevention and Control (ECDC) to assess whether there is any evidence for data manipulation.

The second hypothesis, that is, efficient autocracy, posits that the authoritarian regimes, as compared to democratic ones, were able to effectively control the spread of the disease, such as by timely implementation of lockdown (even if it means curbing individual freedom), and maintenance of better social distancing (Cassan and Steenvoort 2021). Historically, democratic countries have outperformed autocracies in health outcomes (Besley and Kudamatsu 2006). However, as democratic countries had higher cases and deaths in the case of COVID-19, it is seen that autocracies outperformed democratic regimes; hence, the efficient autocracy hypothesis. Scholars such as Cassan and Steenvoort (2021) have contributed to the authoritarian or democracies debate and asserted that the efficient authoritarian hypothesis is misleading due to omitted variable bias. They consider it to have omitted variable bias because most of the studies on political regimes and COVID-19 do not consider three main factors: geographical differences (latitude and longitude), wealth (gross domestic product [GDP]), and proportion of vulnerable populations (those who are 65+ of age). Here, the case of geographical differences must be interpreted with caution because it seems to indicate that authoritarian regimes, as compared to democratic regimes, tend to locate in a better geographical condition.

Apart from these two hypotheses, there is another aspect on the linkages between political regimes and the COVID-19 pandemic that scholars have overlooked: the aspects of vaccination. This is mainly because, though the extant literature has focused on the idea of “vaccination inequality,” it is examined chiefly through the lens of economy, such as by exploring the differences in COVID-19 vaccination across different income societies (high-income  versus low-income societies) (for instance, see Ngo et al. 2022). According to Pilkington et al. (2022), the vaccination rate across different income societies is as follows: 76% in high income, 78% in upper middle income, 48% in lower middle income, and 8.3% in low income. Research has also reported the driving factors of vaccination by asserting that better governance, which generally tends to be a democratic regime, makes achieving better vaccination administration possible (Aida and Shoji 2022).

The problem, however, is that when autocracies seem to perform better in terms of having lower COVID-19 cases and deaths, scholars doubt it and check whether the data on COVID-19 cases and deaths are manipulated and underreported by the authoritarian regime. And when high-income societies (which mainly belong to democratic regimes) seem to perform better by having higher vaccination rates, the case of data manipulation is considered irrelevant. Such irrelevancy is fundamentally based on the a priori assumption that democracies should always perform better, including health administration and outcomes. Put differently, it is conceived that there is nothing to be doubted about if democracies perform well; but if authoritarians perform well, it has to be doubted. Hence, scholars need to enquire whether there are cases of data manipulation (by democracies or by autocracies) regarding the vaccination rate. Exploring vaccination rates through the lens of political regimes is worth examining because vaccination can be considered a proxy for the proactive willingness of the government (be it authoritarian or democratic) to control the pandemic. Moreover, intuitively speaking, one of the practical and visible steps a government can take to ensure stable and sound public health is by taking proactive steps such as vaccination. Here, it must be noted that asking questions regarding data manipulation of COVID-19 vaccination is by no means a critique of any forms of regimes but to affirm whether any of the regimes (authoritarian or democracies) have manipulated the rate of COVID-19 vaccination.

The Present Study

We examine, using the latest publicly available data set, whether the two hypotheses—biasing autocracy and efficient autocracy—acquire any empirical support. The biasing autocracy hypothesis—that the authoritarian regimes manipulated and underreported COVID-19 cases—made us curious to examine whether the number of COVID-19 vaccinations across the political regimes has any signs of data manipulations.

Data and Methods

a) Data

We used the four categories of classifying political regimes (full, flawed, hybrid, and authoritarian) from the Economist Intelligence Unit’s Democracy Index (EIU 2022). This democracy index ranges from 0-10 based on five different dimensions of governance: (i) political culture, (ii) political participation, (iii) functioning of government, (iv) civil liberties, and (v) electoral process and pluralism. Countries that score high in the democracy index 8-10 are considered full democracies. Countries that score between 6-8 are flawed democracies. Hybrid and authoritarian regimes scores between 4-6 and 0-4, respectively (for further information, see EIU 2022). The cumulative number of COVID-19 cases and vaccinations were extracted from WHO COVID-19 Dashboard (WHO 2022). Country-wise GDP per capita (in US dollars) was extracted from the World Bank’s national accounts data (2022). Population and population density data were extracted from the World Population Review (2022). Lastly, the average demographic age was extracted from the United Nations (2022) World Population Prospects  A total of 157 countries (20 for full democracy, 53 for flawed democracy, 31 for hybrid regime, and 53 for authoritarian regime) were included in the analysis.

b) Scheme of analysis

Benford test was used to evaluate whether the COVID-19 cases and vaccination (a proxy of governments’ proactive willingness to control the pandemic) data conform to Benford’s law. Evaluation of data normality after log transforming the required variable (COVID-19 cases) was conducted using the Kolmogorov–Smirnov test. Multivariate analysis using generalised linear models (GLM) was performed to ascertain the relationships between COVID-19 cases and political regimes by adjusting for potential covariates that were informed by the extant literature, such as GDP, population and population density, and average age (Cassan and Steenvoort 2021; Dinia et al. 2022).


The preliminary analysis was conducted in two separate ways—one was inclusive of China and one without—but the results were identical. Hence, the result that was inclusive of China is reported in this paper. The significance value of chi-square (χ2) (see the goodness-of-fit test in Table 1) was larger than 0.05, indicating that the cumulative COVID-19 cases across all the political regimes conform to Benford’s law. This result indicates no empirical support for the biasing autocracy hypothesis (see Figure 1 for visual reference).

Table 1: Goodness-of-fit in Benford’s Test across the Political Regime



Sig. (χ2)

Bayes Factor

COVID-19 cases




















COVID-19 vaccination




















Source: Authors’ estimation.            

Table 2: Descriptive Statistics and K-S Test






COVID-19 cases (Cumulative)*















Population density*





Average age





Average age*





K-S Statistic of COVID-19 cases 

(p value)









*= log10 transformed

Source: Authors’ estimation.

Figure 2: Observed frequency of leading digit of COVID-19 vaccination vs Benford distribution

Source: Authors’ illustration.

Benford’s test on COVID-19 vaccination gave an unexpected result. The significance value (see the goodness-of-fit test in Table 1) was lower than 0.05 in the case of full democracies [Sig. (χ2) = 0.037], indicating that the data on vaccinations for full democracies do not conform to Benford’s law. This finding makes it possible to question the legitimacy of vaccination data provided by full democracies. Moreover, the Bayes factor was the largest for the authoritarian regime, indicating strong evidence that the data of the authoritarian regime was not manipulated.

Figure 3: P-P Plot of COVID-19 cases

                                                                                                                                                                                      Source: Authors’ illustration

The Kolmogorov–Smirnov (K–S) test affirmed that the log-transformed data of COVID-19 cases followed a gaussian distribution (see Table 2) as the K–S Statistic was insignificant for all the categories of political regimes (p = 0.200). Visual inspection was carried using probability-probability (p-p) plot and the dispersion of data points were satisfactory (Figure 3). The descriptive characteristics of our sample (see Table 2) informed that COVID-19 cases, GDP, and the demographic average age were the highest in the full democracies.


Figure 4: Marginal means of COVID-19 cases by political regimes

Source: Authors’ illustration.

We constructed three separate GLMs to assess how COVID-19 cases have linkages with our study’s multiple predictors/covariates (political regime, GDP, population, population density, and average age) (see Table 3). The political regime variable was the only predictor in model 1 and was a significant predictor. This model indicated that the three political regimes (full, flawed, and hybrid) registered significantly higher COVID-19 cases than the reference category (authoritarian regime). The most considerable difference was observed between the authoritarian and full democracy regimes (B= 1.177, p≤ 0.01). Model 2 was constructed to examine how GDP, population, population density, and average age affect COVID-19 cases, and the results showed a positive and significant impact, except for the population density. Model 3 included all the potential covariates along with the political regime variable, and the findings were interesting as the significant differences between full democracies and authoritarian regimes (in model 1) became insignificant in model 3. This change in the significance values of the political regime variable reveals that the differences in COVID-19 cases between full democracies and authoritarian regimes are indeed attributable, as few studies have indicated, to the differences in GDP (B= 0.657, p≤ 0.01), population size (B= 0.788, p≤ 0.01), and the average age of the population (B= 0.037, p≤ 0.01). The marginal means of COVID-19 cases across the political regimes were extracted from model 1 (without covariates) and model 3 (with covariates) to graphically illustrate the differences in COVID-19 cases across the political regime. After adjusting for the covariates, an inverted U-shaped pattern of COVID-19 cases across the political regime was observed (see Figure 4).





Table 3: Generalised Linear Models of COVID-19 Cases and Political Regime


Model 1

Model 2

Model 3


B (SE)

B (SE)

B (SE)


5.277 (0.11) ***

-3.255 (0.41) ***

-3.547 (0.41) ***

Full (versus Authoritarian)

1.177 (0.21) ***


-0.036 (0.13)

Flawed (versus Authoritarian)

0.844 (0.16) ***


0.248 (0.08) ***

Hybrid (versus Authoritarian)

0.149 (0.18) ***


0.207 (0.08) ***



0.589 (0.08) ***

0.657 (0.09) ***



0.774 (0.05) ***

0.788 (0.04) ***

Population density


0.012 (0.05)

0.003 (0.05)

Average age


0.042 (0.01) ***

0.037 (0.01) ***





















LR Statistics




Prob (LR statistic)




Note: ***= p≤ 0.01; B= Unstandardised coefficient, SE= Standard error, AIC= Akaike’s Information Criterion, AICC= Finite Sample Corrected AIC, BIC= Bayesian Information Criterion, LR= Likelihood Ratio.


Source: Authors’ estimation.


Further, model 3 also showed that even after adjusting for the potential covariates, the difference between flawed and authoritarian, and hybrid and authoritarian, is significant. In model 3, GDP and population had the largest coefficient compared to other predictors. We illustrated this finding using contour plots for each political regime (see Figure 5). A contour plot is a graphical illustration technique to project a three-dimensional surface on a two-dimensional plane. The plot shows the values of two predictors (GDP, Population) and the corresponding fitted probabilities of COVID-19 cases. For all the categories of political regimes, the highest probabilities of COVID-19 cases are in the top-right corner (see Figure 5), indicating that the COVID-19 cases are linked with the increment in GDP and population. Further, the plot also reveals that the hybrid and authoritarian regimes tend to have low COVID-19 cases when GDP is low; however, this is not the case for full and flawed democracies.

In terms of model statistics, all three models were significant [Prob (LR statistic) < 0.05]. However, based on the AIC and BIC values, we concluded that model 3 has the best fit of all the models. Overall, the GLM allowed us to affirm that the role of political regimes cannot be negated in determining how different countries have been affected by COVID-19.


The study examined two competing hypotheses related to COVID-19. First is the biasing autocracy. And the second hypothesis is efficient autocracy. As far as biasing autocracy hypothesis is concerned, our study’s major finding is that the cumulative COVID-19 cases across all the political regimes conform to Benford’s law. This finding indicates that there is no empirical support for the biasing autocracy hypothesis (see Table 1 and Figure 1, where the WHO COVID-19 data have been utilised). In other words, the idea that authoritarian regimes manipulated and under-reported COVID-19 cases is not true. Our finding is similar to the forensic analysis of COVID-19 data from 198 countries by Farhadi and Lahooti (2022), the objective analysis of the WHO’s situation reports by Idrovo and Manrique-Hernández (2020), and also the study by Koch and Okamura (2020). These three studies employed the Benford’s test—a data analysis technique similar to our study. However, our findings contrast with the studies that employed regression analysis—a data analysis technique we did not adopt for testing biasing autocracy hypothesis. In this regard, mention can be made of Cassan and Steenvoort (2021) and Annaka (2021). More pertinently, Annaka’s (2021) study posited that data transparency is positively associated with COVID-19 fatality and that one of the reasons why democracies reported higher COVID-19 fatality is because they reported the actual numbers transparently.

Thus, as per our analysis, it can be concluded that COVID-19 cases, for all the categories of political regimes, conform to Benford’s law. The difference in the results could be due to the differences in the “analytical method.” It must also be highlighted that Benford’s test is particularly tailored to detect fraud and data manipulation, whereas regression analysis is tailored to estimate the relationships between a dependent variable and one or more independent variables.

Moving ahead, to test the second hypothesis (the efficient autocracy hypothesis), we adopted multivariate GLM. We found that authoritarian regimes have significantly lower COVID-19 cases compared to the three other political regimes (full, flawed, and hybrid). However, after adjusting for key potential covariates (GDP, population, population density, and average age), the significant difference between full democracies and authoritarian regimes became insignificant. It indicated that the observed difference in the initial analysis is partially attributable to the variations in GDP and population. That said, the difference between flawed and authoritarian, and hybrid and authoritarian, remains significant. This finding is suggestive of the need to prioritise safe and efficient public health measures by flawed and hybrid political regimes.

After adjusting for the covariates, we observe that the marginal means of COVID-19 cases across the political regimes follow an inverted U-shaped pattern (see Figure 4). Based on this observation, an appropriate question that needs to be asked is: Why is the marginal means between full democracies and authoritarian regimes different from that of the flawed and hybrid regime? Is there a more substantial similarity between the two extremes of the political regime scale (full democracy and authoritarian regime) than those political regimes in the middle (flawed and hybrid regime)? These questions arise because our findings suggest that full democracy and authoritarian regimes are similarly effective (covariate-adjusted) at handling COVID-19 compared to the flawed and hybrid regimes. It may be that though full democracy and authoritarian regimes are effective, the means (or the process) of being effective is different in that while the former takes the decentralised civil procedures, the latter takes the state-centric centralised procedures. Whatever the case, the multivariate GLM allowed us to affirm that the role of political regimes cannot be negated in determining how different countries have been affected by COVID-19.

Another focus of our study was to examine whether the data on COVID-19 vaccination conform to Benford’s law. An exploration on “vaccination inequality” literature reveals examining the phenomena through the lens of economics (see Aida & Shoji 2022; Ngo et al. 2022; Pilkington et al. 2022). However, we consider it worthy to examine COVID-19 vaccination through the lens of political regimes. The necessity of examining the political regime aspect arises because vaccination can be considered a proxy for the proactive willingness of the government (be it authoritarian or democratic) to control the pandemic. Unexpectedly, the goodness-of-fit test (in Benford’s test) was significant for full democracies, indicating that the data on COVID-19 vaccination by the full democracies may be doubtful. This finding, however, must be interpreted cautiously as future studies with newer and different samples might inform against our findings.

If our findings can be verified and reaffirmed (or replicated) in other studies, scholars need to come up with probable explanations. In other words, why should the full democracies manipulate their data on COVID-19 vaccination? One potential reason is that the COVID-19 pandemic exposed the shortcomings in the public health infrastructure across all the regimes but with a sharper tone for full democracies (higher cases and deaths than other regimes). Further, such an exposure led various scholars to doubt the efficacy of democracies in ensuring sound public health measures (see Cepaluni et al. 2022; Narita & Sudo 2021). Perhaps, to mend/cover up such an exposure, the government of full democracies might have given extra effort to appear they are delivering higher vaccination rates compared to other regimes (flawed, hybrid, authoritarian). Whatever the case, vaccinating higher numbers of people to prevent from the deadly diseases is desirable. And if the full democracies have indeed been able to do so for large proportions of their citizens, it is something that the other three regimes can follow.

Though the present study had a few key findings, it is not devoid of limitations. For instance, the GLM we employed had a limited number of predictors (only five). Future studies may address this issue by considering other potential predictors of COVID-19 cases and deaths. Moreover, our study did not indicate causal relationships between the predictors included in the analysis and the outcome variable. Future studies may also inquire whether the relationships between political regimes and COVID-19 cases should be less about theoretical questions than exploring the empirical facts so that practical decisions can be made to ensure stable public health.

We thank Dr. Homen Thangjam, faculty at the Indira Gandhi National Tribal University (IGNTU-RCM), for his valuable comments on the manuscript.

Chinglen Laishram ( is a Ph.D. Scholar at Centre for Studies in Society and Development, School of Social Sciences, Central University of Gujarat, Gujarat, India and Pawan Kumar ( teaches at School of Law, BS Abdur Rahman Crescent Institute of Science and Technology, Chennai, India.
31 October 2022