A Cross-country Study of COVID-19 Mortality Rate: How Do Institutions Make a Change?

Review of the Literature: Institutions and Pandemics

Mounier-Jack and Coker (2006) discussed Europe’s capacity to deal with a potential influenza pandemic. The study cited some responsive and committed governmental actions in most European countries. However, such actions varied among the European countries, which resulted in some political tensions. This is in line with what has recently happened during the COVID-19 pandemic in which the European countries have closed their borders and rifts have risen among the European partners. The findings suggest that the EU needs to improve its knowledge-sharing tools on pandemic responses among the EU countries to effectively support the services and coordinate responses. In this regard, we can clearly observe such findings mainly due to the observed different strategies followed by different EU members. More specifically, unlike most other European countries, Sweden chose to go the route of herd immunity by not closing off schools or cancelling major gatherings.

Itzwerth et al. (2006) showed that the spread of influenza adds pressure on hospitals which lack sufficient staff in the case of no pandemics and hence hospitals are likely to suffer a dearth in medical personnel during pandemics. Zanakis et al. (2007) noted the importance of improving the health system, among other measures, as a means of lessening the adverse effects of HIV. Their study showed that nations could control HIV spread even if their GNP is relatively low, given that they have a good health system, low population density and effective media system. More recently, Gilles (2011) studied the effect of public trust in public institutions on the success rate of preventive measures such as vaccination campaigns. The authors argued that a lack of trust in institutions could lead to a future spread of the virus. Keogh-Brown and Smith (2008) showed that the effects of the SARS outbreak were not as significant as predicted by the media. However, the pandemic at that time impacted certain sectors more than others.

In addition to countries’ preparedness in dealing with pandemics, decision-making has been considered a fundamental tool in fighting outbreaks. Eichenbaum et al. (2020) examined the correlation between economic decision-making and epidemics by scrutinising the supply and demand effects of an epidemic. Their findings suggest that the effects of epidemics may result in a persistent recession. The results also suggest that the trade-off is between the health consequences and the severity of the short-term recession. The authors argued that these types of epidemics significantly affect long-run economic growth and substantially shape many other indicators, such as unemployment, jeopardising the supply chain and bankruptcy levels. Such findings could be worth exploring in the context of COVID-19 to forecast the economic aftermath of the pandemic.

In examining the effect of COVID-19 on the economy, Guerrieri et al. (2020) utilised the theory of Keynesian supply shocks. They concluded that resulting changes in aggregate demand are larger than the shock itself. This is mainly due to the shutdowns, layoffs and firms going out of business during this pandemic. The findings corroborate that the resulting firms’ closure has amplified the economic effect and prolonged upcoming recessions in most countries. It is worth noting that the closure policy and its costs are quite controversial due to varying levels of institutional advancement in different countries. Developed and rich countries might have more leverage in imposing these closures, while poorer countries might not be able to afford that. More specifically, Briscese et al. (2020) suggested that full enforcement of self-lockdowns in Italy was costly and hard to implement. Furthermore, the authors stated that achieving this full enforcement is always quite controversial in democratic states. Atkeson (2020), in examining the economic impact of COVID-19 on the US economy, argued that an infection rate of 1%–10% would result in significant challenges to the health system and may potentially result in severe labour shortages in many sectors, especially health care and financial and economic infrastructure. The study estimated the impact up to 18 months over several stages. They suggested that the severity of the transition between stages depends on the recovery rate, number of cases, social distancing measures and the ability to control the virus.

In scrutinising the social and economic activities determinants of COVID-19, Stojkoski et al. (2020) list a variety of determinants, such as population size, life expectancy, population density, elderly population, social connectedness, government spending and mortality from non-natural causes. These factors and others are contributing in one way or another to the spread of the virus or to the ability to control the speed of its spread. The authors show that although the factors influencing the results of COVID-19 are not many, their magnitudes are likely to vary from one country to another due to the disparities in the socio-economic qualities. Marrouch and Sayour (2020) showed that economic development is the most relevant factor for the number of COVID-19-infected people. They also found that population age, especially those ageing between 0 and 14, plays a critical role in transmitting the disease. Noting that government spending was found to be a significant indicator of how well governmental response could be to self-isolation and lockdowns. This is because such an increase in spending makes it easier for workers with symptoms to self-isolate without having to worry about their livelihood. In the absence of this employment insurance or its equivalent during COVID-19, workers will risk carrying the disease and spreading it, due to their financial need. In fact, the Canada Emergency Response Benefit has been reported as a significant tool in fending off the spread of COVID-19 thanks to the generous spending policy that has been put by the federal government in Ottawa to support wages and those who lost their income due to COVID-19 (Webster, 2020).

Despite the cited literature above perceiving institutions as having a critical role in combating pandemics, other studies have argued that institutional quality may not be important as compared to other factors. Carstensen and Gundlach (2006) investigated the relationship between institutions and economic development within a pandemic scenario, namely malaria. They presented two arguments: the first, assign more weight for institutional quality as the main factor of economic development as opposed to geographic factors while the second argument states the opposite. They found support for the second argument showing that institutions are secondary for geographic factors in attaining economic development. Similarly, Bhattacharyya (2009) showed the literature on economic development has been divided between what Bhattacharyya refers to as the ‘institutions view’ and the ‘disease view’. The literature on the former argued that institutions are the sole cause behind economic development, whereas the literature on the latter argued that the emergence of diseases is as important as institutions in affecting economic development. Bhattacharyya (2009) proposed a third strand on this matter, showing that in the initial phase of economic development, surmounting prevailing diseases is superior to improving institutional quality, which is deemed of greater importance in a later phase.

In sum, there seems to be no consensus on whether the quality of institutions in a country has a critical role in combating pandemics as some studies showed it to be effective while other studies have argued that institutional quality may not be as vital as compared to other factors. Accordingly, this article contributes to the literature by attempting to unveil the hidden nuances behind the latter relation between institutional quality and a country’s ability to fight pandemics.

Data Sources

This article uses the 2018 cross-sectional data of 124 countries around the world (see the Appendix for the list of countries) to regress with the COVID-19 mortality rate up to April 2020. It is essential to point out here that regressing with lagged values of explanatory variables may be an unusual case in a cross-country study, but it cannot be denied that in reality, the response of one variable to another variable is rarely instantaneous. Therefore, it is reasonable to regress the COVID-19 mortality rate on the lagged values of explanatory variables. Indeed, the importance of lagged variables is widely acknowledged in both the econometric and economic literature (e.g., Barro, 1991; De Long & Summers, 1991; Wooldridge, 2016). The sample countries are primarily based upon the availability of data and the significance of the mortality rate caused by the coronavirus disease. The data of this study are mainly extracted from several sources, particularly the World Development Indicators (WDI), the World Governance Indicators (WGI), the Global Competitiveness Report (GCR) and the Worldometer. In addition, we also collected relevant information such as the English legal of origin, latitude and religion from the Central Intelligence Agency (CIA) World Factbook and the published work of La Porta et al. (1999). The detailed sources of data including the unit of measurement and the descriptive statistics are summarised in Table 1.

OPEN IN VIEWER

Table 1.

Summary of Statistics.

Variables	Sources	Obs.	Min.	Mean#	Max.	Std. Deviation
CMR i	Worldometer	124	0.003	3.986	65.607	10.129
GDP i	WDI	124	210.804	18,081.06	110,746.6	21,920.43
AGE 50 i	WDI	124	7.354	24.881	46.280	11.950
HEALTH i	WDI	124	0.188	3.839	9.222	2.288
SCHOOL i	GCR	124	9.155	60.970	94.666	20.760
INSQ i	WGI	124	4.915	53.301	98.077	25.912
ENG i	CIA and La Porta et al. (1999)	124	0	0.274	1	0.448
REL i	CIA and La Porta et al. (1999)	124	0	0.581	1	0.495
LAT i	CIA and La Porta et al. (1999)	124	0.011	0.327	0.722	0.196

Note: WDI, World Development Indicators; GCR, Global Competitiveness Reports; WGI, Worldwide Governance Indicator; CIA, Central Intelligence Agency World Factbook; CMR COVID-19 mortality rate (deaths caused by COVID-19 per 100,000 population); GDP per capita real GDP ($ at 2010 price); AGE 50 population age 50 and above (% of population); HEALTH public spending on health (% of GDP); SCHOOL average years of schooling (scaled 0–100); INSQ total quality of institutions (scaled 0–100); LAT absolute value of the latitude of the country (scaled 0–1). Both ENG and REL are dummy variables, where ENG = 1 if the country adopts English common law in the legal system and REL = 1 if the majority of the population in the country follows or practises the Catholic and Islam religions, while otherwise 0. #For dummy variables, the value represents a percentage, whereas, for continuous variables, the value represents the mean.

According to the statistics reported in Table 1, the mean and maximum values of COVID-19 mortality rate CMR i are 3.986 and 65.607, respectively, showing that the incidence of death from the diseases is approximately four people and it can reach nearly 66 people in every hundred thousand of population. Moreover, our descriptive statistics clearly show that the data set used in this article covers a wide range of countries with different levels of economic development and quality of institutions. As such, the findings of our article are considered comprehensive and reflect the global perspective. More specifically, the per capita real GDP carry the largest standard deviation followed by the institutional quality. The per capita real GDP ranges from $210.804 in Burundi to the maximum value of $110,746.6 in Luxembourg. Additionally, we find that the institutional quality ranges from the lowest score of 4.915 in Venezuela to the highest score of 98.077 in New Zealand. In an effort to control such variation, all the variables (except for ENG i and REL i as they are dummy variables) are transformed into natural logarithm to induce homoskedasticity and may further improve the quality of inferential statistics.

Estimation Methods

This article attempts to investigate the role of institutions and other major factors in explaining the global behaviour of the COVID-19 mortality rate. We adopt the mortality model developed by Tang (2019), Subramaniam et al. (2018) and Frey and Field (2000) to examine the drivers of the mortality rate of the COVID-19 pandemic. Our model is presented in Equation (1) below:

where ln denotes the natural logarithm, CMR i is the mortality rate of COVID-19 measured by the number of COVID-19 death cases per 100,000 population, GDP i is the per capita real gross domestic product, SCHOOL i represents education attainment (years of schooling) or a proxy for human capital, HEALTH i represents health quality measured by the ratio of public spending on health to GDP, AGE 50 i represents the size of the ageing population measured by the ratio of the population aged 50 and above to the total population, INSQ i represents the total quality of institutions measured by the average value of six different ‘indicators, namely control of corruption, government effectiveness, political stability and absence of violence, regulatory quality, rule of law and the voice and accountability’ (see Kaufmann et al., 2011).

Another interest of the present article is to examine the facilitating effect of institutions on the impact of education on the mortality rate. Therefore, an interaction term of ln SCHOOL × INSQ i is accommodated into the model as shown below:

However, North (1990) highlighted that the quality of institutions may not be strictly exogenous. Therefore, there is a possibility of an endogeneity problem in our article because mortality and institutional quality can be affected by a third factor like economic development. For instance, economic development promotes better health care to control morality (Tang, 2019). Moreover, economic development also encourages the demand for freedom and rights, thus enhancing institutional quality (Wang, 2013). In the presence of endogeneity, the ordinary least squares (OLS) estimator is biased. In light of this, the present article employs the instrumental variables estimator such as the two-stage least squares (2SLS) to mitigate the endogeneity problem caused by institutions.

According to the institutional literature, the quality of institutions is correlated with the origin of the legal system, religion and geographical factors (Acemoglu et al., 2001; La Porta et al., 1999; Landes, 1998). In the aspect of the origin of the legal system, Beck et al. (2003) documented that countries with the British legal system tend to provide a higher level of property rights compared to the French legal system. In addition, Acemoglu et al. (2001) also found a marginally significant positive effect of the British legal system on institutions. Therefore, they surmised that British-colonised countries are likely to inherit better institutional quality. Besides, the institutional theorists also agree that religious affiliations of the population play a significant role in shaping the institutional quality of a country through culture. For example, government performances are likely to be inferior for the country with a large population of Catholics and Muslims because they acquired cultures of intolerance, closed-mindedness and xenophobia that detriment development (La Porta et al., 1999; Landes, 1998). In addition to the two mentioned factors, La Porta et al. (1999) and Hall and Jones (1999) also found that geographical factor such as the distance from the equator (also known as the latitude) has a significant impact on government performance. More specifically, Hall and Jones (1999) found that countries with large degrees of latitude tend to enjoy higher per capita income due to the influence of good governance and institutions from the Western region. Motivated by the earlier studies (e.g., La Porta et al., 1999; Lee & Law, 2017), the present study incorporates the English legal system ENG i , religion REL i and latitude LAT i as the potential instruments for institutions to estimate the following first-stage regression:

In the first-stage regression, the endogenous variable, lnINSQ i , is regressed on four suggested instrumental variables and other exogenous covariates Z i . To materialise the second-stage estimation, the predicted values derived from the first-stage regression, namely ln INSQ ^ i , will be used to regress the behavioural equations instead of using the original endogenous regressor, lnINSQ i . However, a battery of diagnostic tests, for instance, the endogeneity test (Durbin, 1954; Hausman, 1978; Wu, 1973), the under-identification test (Anderson, 1951), the weak instrumental test (Stock & Wright, 2000) and the over-identification test (Sargan, 1958) are performed to validate the models and estimators.