Size isn’t everything: COVID-19 and the role of government

Abstract

The emergence of the global COVID-19 pandemic in 2020 quickly generated claims that the crisis had demonstrated the superiority of extensive welfare states and a failure of market economies. We conduct cross-sectional statistical analyses to test this claim with regard to first response and reported COVID-19 related deaths per 31 December 2020 (N = 164–200), using government spending as a central variable. The analyses confirm some of what is known from other studies: COVID-19 deaths associate positively and robustly with ageing and more obese populations. However, we find no statistically significant associations between various measures of government size and the number of COVID-19 deaths, alone or when controlled against demographic, political and economic factors. If anything, the general effectiveness of government services and the availability of hospital beds seem more important than the simple size of government or level of health care expenditures.

1 Introduction

Fighting pandemics is a typical example of public good provision with potential free-riding problems (Barrett, 2007), and hence where governments and their policies should be able to make a visibly positive difference for the outcomes—at least ideally speaking. Nonetheless, how government policies and social conditions may influence the spread and fatality of diseases is not a topic that has usually been widely studied by economists and political scientists, including public choice analysts (Leeson & Thompson, 2021).

However, barely had the COVID-19 pandemic become a global phenomenon in January 2020 before researchers, journalists and commentators drew very hard and specific conclusions about the possible social, economic, and political causes of variation in how affected countries were, as well as about the wider policy implications. Some suggested that the crisis was proof that larger, more intrusive governments are better equipped to handle such crises: “We need big government to save us from the pandemic” (Time, 17 April 2020), “A crisis shows the value of big government and big business” (Bloomberg, 16 April 2020), and “The coronavirus shows the era of big government is back” (New Statesman, 19 March 2020).

Others went significantly further and declared the pandemic to be no less than the gravestone of market economies, capitalism, and globalization: “Will coronavirus signal the end of capitalism?” (Al Jazeera, 3 April 2020), “Has Covid-19 killed globalization?” (The Economist, 14 May 2020), and “Welfare states will be big Covid-19 winners” (Reuters, 13 May 2020). Renowned Columbia economist Jeffrey Sachs claimed that economic inequality was one of the primary, “structural” causes of COVID-19 fatalities and that the Gini coefficient could explain much of the variation (Sachs, 2020).

A more fundamental point seems to loom in these claims: That bigger, more redistributive governments supposedly would be better equipped to weather such a pandemic, at least in its early phases. In media coverage some Nordic welfare states—notably Denmark, Finland, and Norway—were held out as illustrations of how a large public sector might alleviate the spread and fatalities better than other societal models. However, even a cursory glance suggests that the picture at least may be more complex (cf. Bejan & Nikolova, 2022). Countries such as Sweden, UK, The Netherlands, and Belgium—not known for small governments—chose somewhat different strategies from the aforementioned Nordic states and exhibited quite different trajectories, at least in the short run.

So very strong conclusions about causes and implications seemed at least in some cases to be possible instances of partisan or ideological bias influencing the interpretation of policy effects and whom to assign blame or praise (cf., e.g., Bisgaard, 2015). However, making such sweeping statements, often bereft of data and while the COVID-19 pandemic was only beginning to rage, was a bit akin to judging what sports team’s strategy is best while only ten minutes into a game. Ultimately, such conclusions ought to be based in data rather than be taken as premises.

The present study attempts to consider the potential importance of the size of government during the initial year of the global pandemic. Specifically, we seek to test a simple hypothesis in line with the drastic claims made early: That larger governments were better at handling the pandemic than smaller ones—i.e., a claim that would suggest a negative correlation between the size of government and the number of COVID-19 deaths. Doing so we are not putting forward a theoretical model but merely submitting the supposed wisdom of many pundits and politicians to a simple and straightforward test: Did countries with bigger governments have relatively fewer COVID-19 deaths during the first full year of the pandemic? Along the way, we shall consider other socio-economic and political factors that have been claimed as important factors or at least conceivably could be so.¹

In the following, we shall first describe the data to be considered (Sect. 2) and subsequently use these for both bivariate and multivariate statistical tests (Sect. 3). The analyses suggest quite unequivocally that there is no evidence to support a claim that countries with larger governments fared better during the first year of the crisis than those with smaller. Any such analyses must, of course, be taken with considerable caution, so in the end we discuss the reliability and implications of these results (Sect. 4) and draw some wider perspectives (Sect. 5).

2 Data

In order to study the possible role that the size of government and other political and socio-economic factors may have played in the spread of the virus and generating fatalities during the initial phases of the pandemic, we shall utilize the COVID-19 data collected by Max Roser et al.’s Our World in Data website, whose daily updated data on reported cases and reported COVID-19 related deaths all come from the European Centre for Disease Prevention and Control (ECDC). All COVID-19 fatalities are as reported per 31 December 2020, thus essentially covering a full year of observations, and with the national numbers relative to population size (million inhabitants) rather than raw numbers. Only looking at the first year takes out the later effects of vaccination programs and hence variations in these (Albrecht & Rajagopalan, 2022). Summary statistics for these, as well as all other variables considered here, are given in Table 1 (N = 164–220).²

We focus only on COVID-19 related deaths, as it is widely recognized that there were significant measurement problems regarding reported COVID-19 cases.³ While the number of cases relative to population may give some measure of the spread of the disease, there are simply too many uncertainties involved to use these as the dependent variable. Some of these uncertainties are medical-methodological (e.g., the quality of the tests used, with considerable differences in the number of false positives/negatives). Others include that differences in the number of cases may have been a function of alternative test strategies—so, e.g., that a country that tested more extensively would tend to also have more reported cases, which itself may be a function of the variables we are interested in. Including reported cases in multivariate analyses of the causes of deaths might therefore introduce serious problems of both reverse causality and post-treatment bias. Nonetheless, reported cases may give at least a rough indication of how widespread the pandemic was, so we shall in the preliminary, bivariate analyses make one lonely comparison of deaths and cases, mostly for illustrative purposes.

To test the possible relationship between the size of government and COVID-19 fatalities across countries, we have chosen the simplest of all measurements of the independent variable: Total government expenditures as percentage of GDP (2018), as reported in the International Monetary Fund’s World Economic Outlook database. However, we shall also—albeit briefly—consider another possible measure of the size of government, namely one of the major components of the Economic Freedom of the World index (cf. Gwartney et al., 2019). The index contains five sub-indices, one of them measuring the size of government as a composite of government consumption, transfers and subsidies, government investments, and top marginal tax rates. The scores of the subindex on government size have here been reversed in order to make interpretation more straightforward. Unsurprisingly, the measure is highly correlated with that of government spending (r = 0.531; p = 0.01; N = 162).

However, other factors may obviously have played a role for both the spread of the virus across nations and the number of associated fatalities, and we shall accordingly consider a large number of other demographic, geographic, socio-economic, and political factors. It is, for example, by now well-established that some of the major characteristics associated with reception of and vulnerability to COVID-19, aside from being male, has been obesity and age (cf., e.g., Jordan et al., 2020; Palaiodimos et al., 2020; World Obesity, 2021). Accordingly, we have included a measure of the percentage of the male population that is obese (2016), as well as the share of the male population that is 70 years or older (2017). It is also conceivable that the virus may have spread more easily in more populous countries or more densely populated areas (Allain-Dupré et al., 2020), so we include measures for those too (2020 or most recent year available). These demographic data are all based in Our World in Data but ultimately derive from the United Nations and the World Bank.

It has frequently been suggested that corona viruses spread more slowly in warmer climates, so we include a dummy variable signifying whether or not a country has a tropical climate (cf. Banik et al., 2020). To consider other potential regional effects in the form of proximity to places with the disease rampant, we have furthermore included dummy variables for each of the continents.

The risk of spread of a virus may affect countries with different levels of development differently, so we include GDP per capita (2011-USD, most recent year, also derived from Our World in Data). In particular, how ‘globalized’ an economy has been suggested to be correlated with how vulnerable it is to the spread of a virus (Ruiz Estrada & Khan, 2020): The more people travel and trade with each other, the more likely it is that they will come in contact with infected individuals. Accordingly, we include the KOF Globalisation Index as a variable (2017, cf. Gygli et al., 2019).

Large social disparities might conceivably be a cause for differences in how health problems are treated. In order to consider the possible effects from distributive factors, we include Gini coefficients on income inequality derived from Frederick Solt’s Standardized World Income Inequality Database (SWIID) (Solt, 2020a, b).

Ultimately, the way a society weathers an emerging pandemic may, of course, be affected by its policies and governance and therefore also its previous experiences and basic health care set-up (Allain-Dupré et al., 2020). We therefore include a dummy variable for whether or not a country had any fatalities during the 2002–2003 SARS epidemic (cf. Bremmer, 2020), based in WHO data for number of cases. We also include a variable for the number of available hospital beds per 1,000 inhabitants, just as we include a country’s total health care expenditures per capita (in both cases obtained from Our World in Data). However, there may not necessarily be a connection between the amounts of resources devoted to a policy area and how well they are utilized, so we also consider the expert survey data on government effectiveness derived from the World Bank’s Worldwide Governance Indicators (2018).

Table 1 Summary statistics

^£ Original data range index reversed. ^$ Original data range index from 0 [least] to 10 [most]; here transformed to a dummy variable with democracies being countries with index value ≥ 5. ^§ Data as downloaded January 2021.

It is conceivable that countries with different types of political leadership might respond differently to a crisis. It has both been argued that authoritarian regimes will be better equipped to implement ‘tough’ but necessary policies, and that democracies will be more focused on providing public goods than autocracies. To investigate this aspect, we have utilized The Economist Intelligence Unit’s Democracy Index (2019), which has been transformed into a simple dummy variable (for countries scoring 5.0 or above on the index) for whether or not a country is democratic.

In the Spring of 2020 it was frequently claimed by journalists and commentators that countries with female heads of government had fared remarkably better during the crisis than governments led by men,⁴ supposedly because women would take a greater interest in health care. A research paper also added statistical support for this (Garikipati & Kambhampati, 2021) and concluded that deaths were lower in countries led by female politicians. To consider this, we have included a dummy variable signifying whether in June 2020 the head of government in parliamentary and semi-presidential systems was a woman, or in presential systems the head of state.

It was also suggested that left-wing governments were better at handling the crisis, whereas the seemingly erratic strategies of, e.g., the US, the UK and Brazil were seen as at least partly the reflection of the more or less conservative ideologies of Donald Trump, Boris Johnson and Jair Bolsonaro respectively (cf. Högl et al., 2020; Piscopo, 2020). We have accordingly included a dummy variable signifying the ideological color of a government—in this case whether the head of government was the representative of a political party that is self-declared left-wing, defined as social democratic, socialist, communist, or left-radical (as per June 2020).

Ideally a cross-sectional study should also consider when and how governments have responded to the pandemic. The Oxford COVID-19 Government Response Tracker (OxCGRT) of the Blavatnik School of Government of the University of Oxford (Hale et al., 2020) constitute a daily updated database, systematically collecting information on various policy responses that governments have taken, using a set of 17 indicators and with data from more than 160 countries. From this we have included (a) the date in 2020 when COVID-19 related policies first were made; the level of stringency of government responses on (b) the 26 February 2020 and (c) the 18 April respectively. This will allow us to get a glimpse of both how relatively early/late governments responded and how stringent their policies were. However, just as with the number of reported cases, we may here be faced with problems of post-treatment bias: If a country is hit hard by fatalities and therefore adopts very severe policies, any apparent correlation is likely to be flawed, so we only use these variables for the bivariate analyses to get a preliminary glimpse of any relationships.

3 Analysis

The following empirical analysis is split into two parts: First, we run a set of simple, bivariate correlations between on one side a number of potentially relevant variables and on the other frequency of COVID-19 related deaths. Second, we conduct a series of multivariate analysis between some of the former and with deaths as the dependent variables and with the inclusion of a number of control variables.

3.1 Bivariate correlations

Table 2 contains results of simple, bivariate correlations between the potential explanatory variables described in the previous section and reported COVID-19 deaths. The coefficients (Pearson’s R) for deaths per 31 December 2020 are here ordered from largest to smallest in the second column.

Many associations that are statistically significant at conventional levels are as expected from other studies: Most obviously and least surprisingly, there was in the first year of the pandemic a strong correlation between reported cases and deaths. Also, higher age and more obesity, as expected, associate with more deaths. However, several less straightforward associations are also visible, with more deaths seeming to associate with countries being more highly developed: Higher prosperity, more democracy, a more globalized society, and a location in Europe, thereby suggesting that the first onslaught perhaps hit countries with more travel and perhaps slightly more indoors interaction more immediately. In contrast, a location in Africa and Asia and having a tropical climate associate with fewer deaths.

Perhaps surprisingly, there are no visible and statistically significant, positive associations between, on one hand, how early governments reacted to the virus or how stringent or early their anti-virus and lockdown policies were and, on the other hand, the rate of fatalities.

Probably equally surprising, there are positive correlations between COVID-19 deaths and such variables as government spending, health care expenditures, availability of hospital beds, and government effectiveness—all statistically significant—thereby suggesting that countries with more extensive and more effective governments have seen more and not fewer COVID-19 deaths. Income inequality also exhibits a statistically significant association, but with the opposite sign of what should be expected if social distribution was a driver of the pandemic, as suggested by Sachs (2020).

The ideological color or gender of the head of government, as well as population size, population density, and prior SARS-experience exhibit no statistically significant associations. The same is the case for the alternative measure of government size.

Table 2 Bivariate correlations, COVID-19 deaths per million. Pearson’s R

Against the preceding analysis it might be objected that the date of 31 December 2020 is arbitrary. A case can be made that it makes sense to have observations from before the vaccinations set in, but in order to prevent criticism for cherry picking we have included a second column with similar correlations but with the date of observation being 11 May 2023—chosen as it was the date the last COVID-19 restrictions were lifted in the US. The important point to note here is that despite there being two and a half years between the dates of the two observations, the results are essentially the same even if the size of the coefficients might have changed in size: All the correlation coefficients that were statistically significant for the 2020 data, were so too for the 2023 data. Only three that were not statistically significant for the 2020 data had become so by 2023: The regional dummies of Asia, South America, and Oceania. None of the statistically significant correlations had changed signs. This also entails that even for the 2023 data there is nothing to suggest that countries with higher levels of government spending or any of the other measures of government activities had experienced fewer deaths.

3.2 Multivariate analyses

The foregoing, simple analysis cannot prove the causal importance of any factor. Furthermore, while many of the factors positively associated with COVID-19 deaths suggest that the pandemic hit economically developed, mostly Western societies particularly hard (at least in the first year), it should be equally obvious that many of these variables probably are highly correlated with each other. This calls for multivariate analyses, which due to the nature of the research question will need to be cross-sectional in nature and will be done using Ordinary Least Square (OLS) regression analysis.

We approach this by first creating a baseline model, including only variables that clearly are ex ante to the pandemic itself and therefore unaffected by it, and which simultaneously makes sense as political and social factors that potentially could influence the dependent variable: Government spending levels, health care expenditures, government effectiveness, age, obesity, and hospital beds. We also include democracy and income inequality. We would have wanted to include both globalization and prosperity (GDP per cap.), but the two are strongly and significantly correlated with each other (r = 0.596; p = 0.000; N = 180). This also hints at multicollinearity, with the globalization index’s Variance Inflation Factor (V.I.F.) = 10.55 if added to the baseline model. To handle this, we omit globalization and retain prosperity.

Table 3 contains an estimate for such a baseline regression including these variables (model 1). Several of the independent variables of the baseline model have the expected signs: An ageing and more obese population associate with more deaths and are statistically significant, while a more effective government associates with less fatalities. However, somewhat surprisingly higher government spending and higher health care expenditures associate with more deaths, whereas higher income inequality also associates with more; however, the coefficients of the variables are all statistically insignificant at conventional confidence levels. All in all, the baseline model has a nontrivial explanatory power, explaining ca. 44 pct. of the variation in death rates and suggesting that government spending, including on health care and hospital beds, played no role for how fatality rates developed across countries in the first year of the pandemic. Only age (positive), obesity (positive), government effectiveness (negative) and perhaps surprisingly democracy (positive) seems to be of robust importance.

Table 3 Ordinary Least Squares regression, COVID-19 deaths per millions, 31 December 2020

Against the baseline model, it might be objected that it could be problematic to include both government spending levels and health care expenditures in the same regressions, since they could be seen as capturing much of the same, and indeed the correlation between the two is high and significant (r = 0.482; p = 0.000; N = 181). However, this need not be so, given that the one measures a percentage relative to the economy overall, while the other is measured in real spending per capita. Nonetheless, in order to consider the effects of using the one or the other, we run two models with the baseline model variables except only health care spending in the one (model 2) and only government spending level in the other (model 3). As is evident from Table 3, this has no virtually significant effect on the results of model 1. The overall explanatory power is similar (ca. 44 pct.), and the coefficients’ signs and statistical significance remain unchanged, with the exception of democracy in model 2 which falls just below the 0.95 level.

However, models 1–3 show some signs of possible multicollinearity, with the measures of government effectiveness and health care expenditures having V.I.F. values ≥ 5, most likely due to high correlations with prosperity (r = 0.776 and 0.827 respectively). So, model 4 is a ‘best fit’ version of the baseline model with only those variables included that contribute to the overall explanation and without signs of multicollinearity (in this case V.I.F. < 2.5), i.e., government spending, prosperity, and income inequality are dropped. Instead, the model confirms the usual suspects: Higher age and more obesity associate with, more deaths, while more effective governments associate with less. Somewhat surprising democracy remains positively associated with higher death rates and statistically significant.

Model 5 contains the full set of all ex-ante relevant variables considered here,⁵ i.e., excluding some that are ex-post to the initial phases of the pandemic (reported cases of COVID-19, governments’ anti-virus policies, etc.), i.e., the baseline model combined with other ex-ante factors (geography, demographics, etc.) added as control variables. Altogether, model 5 has non-trivial explanatory power (ca. 50 pct.) but not much more than models 1–4 and almost all of the coefficients lack the levels of statistical significance conventionally required. In fact, only obesity (positive), health care expenditures (positive) and a location in Asia, Africa and North America (negative) are significant.

However, health care expenditures exhibit some, moderate signs of multicollinearity in model 5 (V.I.F.=4.44), no doubt partly due to high correlation between those and prosperity (r = 0.827; p = 0.000; N = 175). So, we run a final model (6) in which we again (as in model 3) exchange health expenditures for government spending and make a best-fit version. Model 6 explains ca. 52 pct. of the variation, with the only statistically significant variables being obesity (positive) and democracy (positive), as well as the regional dummies of Asia, Africa, North America and Oceania (negative).

Looking across all models, it is evident that government spending per se has no robust, explanatory value: If anything, it seems in fact that larger governments generally associate with higher numbers of COVID-19 deaths, although in all models the coefficients are statistically insignificant. Figure 1 graphically illustrates the partial correlation of government spending and COVID-19 deaths in model 3 (i.e., when controlled for hospital beds, government effectiveness, age, obesity, and democracy) and quite visibly displays the non-relationship.

Not even the size of health care expenditures exhibits any negative association with COVID-19 deaths: If anything, higher spending consistently associates with more fatalities, both in models where overall government spending is included and where only health care spending is. The only other almost consistently noteworthy variable is the availability of hospital beds, which as expected is negatively associated albeit statistically insignificant.

Fig. 1

Partial correlation between government spending (pct. of GDP) and COVID-19 deaths (per million) (OLS-regression, model 3, Table 3)

4 Discussion

The previous results raise a number of interesting points, first and foremost the apparent non-importance of political factors in general and the size of governments in particular. In the following we shall briefly elaborate on these aspects, as well as the question of the gender of the heads of government.

4.1 Health care expenditures and the size of government

How should one explain the seemingly consistent result that in the first full year of the COVID-19 pandemic higher health expenditures associated with higher death rates (albeit not always statistically significant)?

One possible interpretation would seem to be that expenditures as such are a bad indicator for the quality of the service—because what goes into health expenditures in practice are used to cover many other things than an effective health care vis-à-vis a pandemic. Seen from a public choice perspective two interpretations may be at hand: That government run health care systems have other, politically and bureaucratically determined priorities, including the special interests of politicians and administrators (cf. Niskanen, 1971). Or that regulatory and oligopolistic interests in partly privately financed or privately managed health care systems drive up costs (cf. Stigler, 1971). That is, variations of the theme that government failure may be the cause and indeed be pervasive (cf. Murtazashvili & Zhou, 2023), either because policies are not efficient relative to other strategies or because they may lead to considerable, often negative unintended consequences (cf. Leeson & Thompson, 2021; Hebert & Curry, 2022; Chakraborti & Roberts, 2023).

However, another, simpler explanation—or at least a part of it—may be that wealthier and more globalized economies were hit harder because of more travel, etc., but these are also typically the ones who can better afford—and have—higher health care expenditures. Unsurprisingly, there is a very strong, positive correlation between GDP per cap. and health care expenditures (r = 0.827; p = 0.000; N = 175).

Could the same be the explanation for the empirical non-support for the claim that big government is necessary in order to handle a pandemic such as COVID-19, or at least better than small governments at doing so? This would be possible to test, and it seems the answer is no. While GDP per capita and government spending levels also are positively correlated, they are less so (r = 0.427; p = 0.000; N = 177). Furthermore, if we run models 1 and 3 of Table 3 but omit prosperity, we get models 7 and 8 in Table 4, and this actually marginally improves the explanatory power of the former and does not change the overall picture of either.

Table 4 Ordinary least squares regression, COVID-19 deaths per millions

However, maybe government spending is not the ideal measure for the size of government? After all, the ‘size’ of government might be more than its spending. For that purpose, we omit government spending and health care expenditures and instead include the government size variable, derived from the Economic Freedom of the World index (model 9). The result may be seen by comparing models 8 and 9, which are basically identical, even if the measures of government size are different. In all three models, only ‘the usual suspects’ of age, obesity and government effectiveness are statistically significant and certainly with stronger and clearer effects than over-all government spending, the level of health care expenditures, or the availability of hospital beds.

4.2 Female heads of government

The previous analysis does not lend support to the much popularized result that countries with female heads of governments should have weathered the crisis best. In their paper, which exists in two different versions, one with (Garikipati & Kambhampati, 2020) and one without (Garikipati & Kambhampati, 2021) multiple regressions analysis, the authors consider data largely similar to several of those used here, although with several important differences: Their data on fatalities stop half a year to nine months earlier than those considered here, because they only focus on outcomes at the time of lockdowns, and their dependent variable is total number of deaths rather than deaths compared to populations. Secondly, they test for much fewer possible covariates than done here, including not testing for obesity. Also, for much of their analysis they use a ‘nearest neighbor’ matching methodology to compare countries headed by female leaders with their closest ‘neighbors’ along a range of covariates. They show that on average less reported cases are found and less people have died in countries with female heads of government.

However, given that the number of cases with female heads of government is quite small,⁶ it is also the case that the results may be very weak. What the researchers did not make explicit (cf., e.g., the OLS regression analysis in their original paper (Garikipati & Kambhampati, 2020: Table 2, p. 7) were four non-trivial aspects. First, that due to omission of a number of relevant variables (notably obesity) the over-all explanatory power of their multivariate regression model is fairly modest: A mere ca. 14 pct. of the variation. Second, the by far largest part of the explanatory power is derived not from the gender-variable but from the other variables included (first and foremost population in urban agglomerations). Third, the coefficient of the gender variable is not even statistically significant at conventional levels (p < 0.1; N = 167). Finally, the supposed empirical size effect of gender is difficult to interpret in a meaningful way since the measure is the total number of deaths rather than relative to population.

When we here replicate the analysis using quite similar data, we get the results in model 10 of Table 5. The results are almost identical to those of Garikipati and Kambhampati (2020: 7), and again the gender variable is not statistically significant at a 95 pct. level of significance, and only the age variable (positive) is. Together the included variables explain ca. 36 pct. of the variation.

Adding the variables that we have here found to be potentially important (obesity, hospital beds) (model 11), the significance of gender becomes even less. Only age (positive), obesity (positive) and hospital beds (negative) are statistically significant, whereas the explained variation increases to 42 pct.

Removing those variables that do not contribute to the over-all explanation, we get a ‘best fit’ model (12), where gender of head of government is dropped and only three of the variables of model 11 remain (age, obesity, prosperity) but where the explanatory power is about the same (ca. 41 pct.). To accept the claim that the gender of the head of government is very important, one would have to disregard such factors as age, obesity, prosperity, and availability of hospital beds.

On this basis it is impossible to conclude that a relationship between the gender of the head of government and COVID-19 deaths is of any robust importance. Rather, as pointed out by Selck et al. (2020) the dual fact that Garikipati and Kambhampati use total deaths rather than deaths relative to population and that female heads of government predominantly were found in small states, heavily skewed the data resulting in a most likely spurious relationship, produced by a fortuitous combination of non-political factors and political-economic decisions made sometimes long before the pandemic (cf. Piscopo, 2020).

Table 5 Ordinary Least Squares regression, COVID-19 deaths per millions

5 Conclusion

The present analysis finds no support for any claim that countries with larger, more intrusive governments somehow were better at handling the pandemic in its first stages. Countries with higher general government spending did not have lower fatality rates in the 12 months of 2020, when the pandemic first raged. There is also no solid support for a claim that countries with greater income equality did better.

However, the present findings are to some extent null findings, suggesting that there is no relationship between the size of government and how successfully a country has weathered the pandemic in terms of deaths. Null findings are rarely popular as research results, but they are in no way unimportant when it comes to research connected to public policy. To falsify views held widely, by the populace, the media or among elites, is also of literally vital importance—and no less so when it is something as dangerous and costly as a global pandemic (Eichenberger et al., 2020).

All in all, it thus seems as if it is not the size of government—the amounts of resources consumed by it—that is important for its ability to handle such social ills as epidemics. Rather, it is the effectiveness with which it uses its resources and the quality of what is actually delivered, as well as the behavior of citizens (cf. Zhang et al., 2020). As such the present results are largely compatible with those of Bejan and Nikolova who analyzing data from various welfare states found that liberal societies “fared much better” during the pandemic than those with social democratic or conservative/corporatist welfare states (Bejan & Nikolova, 2022).

The bivariate analyses also suggest, at least prima facie, that there is no support for the claim that those countries that imposed more stringent anti-COVID-19 policies had lower fatality rates. At some level this corresponds with the empirical finding that there may not be any correlation between the extent of emergency powers given to governments and how well those states weather a large scale emergency—or even the counterintuitive result that there may be a negative correlation (Bjørnskov & Voigt, 2021; cf. Koyama, 2023).

The finding that there is no robust association between policy stringency and deaths is also consistent with the findings of Bjørnskov who—also using the Blavatnik index on government response and Our World in Data’s COVID-19 death data—investigated the possible relationship between severity of lockdown and deaths and found no connection (Bjørnskov, 2021).

The COVID-19 pandemic led to many calls for more ‘big government’: More regulation and more government spending (Potrafke, 2021)—often with considerable public support, even if the policies implemented often had very little solid data to support them and the costs often were not very visible (Congleton, 2021; Mulligan, 2021). The cross-sectional data from the first year of the pandemic considered here, highlight that the initially more boisterous claims that it was the countries with larger, more intrusive governments that were best able to minimize the negative effects, find no support. There is, on the basis of the data considered here, no empirical basis for claiming that only large governments can combat pandemics, or even more modestly that they are significantly better at doing so.⁷