Everything that rises must converge: How the policy and public responses to Covid-19 in the education sphere impact on virus transmission

Eleftheriou, Konstantinos; Iokimidis, Marilou; Johnes, Geraint Article Everything that rises must converge: How the policy and public responses to Covid-19 in the education sphere impact on virus transmission Review of Economic Analysis (REA) Provided in Cooperation with: International Centre for Economic Analysis (ICEA), Waterloo, Ontario Suggested Citation: Eleftheriou, Konstantinos; Iokimidis, Marilou; Johnes, Geraint (2024) : Everything that rises must converge: How the policy and public responses to Covid-19 in the education sphere impact on virus transmission, Review of Economic Analysis (REA), ISSN 1973-3909, International Centre for Economic Analysis (ICEA), Waterloo (Ontario), Vol. 16, Iss. 3, pp. 393-408, https://doi.org/10.15353/rea.v16i3.5397 This Version is available at: https://hdl.handle.net/10419/328171 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc/4.0/ Review of Economic Analysis 16 (2024) 393-408 1973-3909/2024393 393 www.RofEA.org Everything that Rises Must Converge: How the Policy and Public Responses to Covid-19 in the Education Sphere Impact on Virus Transmission Empty 15 KONSTANTINOS ELEFTHERIOU University of Piraeus Empty 15 MARILOU IOAKIMIDIS National and Kapodistrian University of Athens GERAINT JOHNES Lancaster University Empty 15 This paper studies the variation across countries in mortality rates due to COVID-19. A two-stage approach is used to model international data on the spread of COVID-19. The first stage applies a convergence club framework to identify clusters of countries within which the development of the epidemic is similar. The second stage models the membership of these clusters as a function of policy responses and public responses to the pandemic, with a particular focus on education as a determinant of the public response. Data include COVID-19 mortality per million, tertiary education completed, age structure, and voice and accountability and government effectiveness measures for each country in the dataset. We find evidence of two convergence clubs, one with a higher and one with a lower COVID-19 death rate. We also find evidence to support the hypothesis that relatively high levels of educational attainment in a population predicts a lower COVID19 death rate among individuals 65 or more years of age. We speculate that this is attributable to more informed prosocial behaviors directed toward elderly people in the form of adherence to guidelines intended to reduce spread of the virus. Keywords: COVID-19; convergence; education; world data JEL Classifications: C23; I18; I23  Eleftheriou: Corresponding author. Department of Economics, University of Piraeus, k[email protected]; Ioakimidis: Department of Business Administration, National and Kapodistrian University of Athens, [email protected]a.gr; Johnes: Lancaster University Management School, [email protected] Acknowledgement: We are thankful to the Editor and two anonymous referees of this journal for their helpful comments and suggestions. Competing interests: None declared © 2024 Konstantinos Eleftheriou, Marilou Ioakimidis, Geraint Johnes. Licensed under the Creative Commons Attribution - Noncommercial 4.0 Licence (http://creativecommons.org/licenses/by-nc/4.0/. Available at http://rofea.org. Review of Economic Analysis 16 (2024) 393-408 394 www.RofEA.org 1 Introduction The COVID-19 pandemic has had devastating effects on the lives and livelihoods of people across the world. Its spread has been accelerated by the extent of modern day connectedness, and so while pandemics have happened before, the degree of preparedness for such an occurrence has been limited. Despite (or perhaps because of) this, the policy response and the public response have varied considerably across countries in terms of both their characteristics and speed of implementation. In this paper, we focus on the public response, and in particular consider the following research question: to what extent has education shaped the public response within each country and so served to mitigate the impact of the pandemic? In doing so, we employ a modelling framework grounded in the theory of convergence (Barro and Sala-i-Martin, 1991, 1992). We hypothesise that, while countries’ experiences of COVID-19 are not synchronised, the properties of the virus will lead this experience to converge, albeit possibly on more than one equilibrium. Hence some countries converge on an equilibrium that is more benign than that approached by others. Our primary interest is in the role played by education in determining which countries cluster together in these convergence clubs. The paper is structured as follows. This introduction is the first section. Section 2 provides a brief literature review, covering both relevant parts of the emergent COVID-19 literature and methodological contributions. The methodology is discussed in greater detail in Section 3. Section 4 presents the data used in the study. The main analytical section, alongside a discussion of the results, is Section 5, and Section 6 concludes the paper and draws policy implications from the analysis. 2 Literature review The COVID-19 pandemic has presented economies around the world with a shock that is unprecedented in recent history. The authorities were, in effect, faced by a choice between closing down transmission of the virus (or at least slowing it down enough to allow vaccines to be developed) by drastically reducing social contact or allowing the virus to infect entire populations. Either option was hugely costly. The first involved minimising close contact between people, where much of this contact occurs in places of work as a necessary input into economically productive activity. Hence lockdown sacrifices economic output, creating job loss and economic hardship for many individuals, especially young adults, minority ethnic groups, and those with unstable employment (Crossley et al., 2021). Lockdown may also adversely affect mental health by creating undue isolation (Brodeur et al., 2021). The second option likewise involves sacrificing the output of those who are infected while they are infected and in recovery. It also entails accepting a high rate of fatality, which is costly because lives have value. In any calculation of the relative merits of each option, weight needs to be assigned ELEFTHERIOU, IOAKIMIDIS Covid-19 Policy in Education and Virus Transmission 395 www.RofEA.org to the deaths that occur, in effect assigning an economic value to life. Most countries opted for some form of lockdown, hoping to minimise the number of fatalities by, in effect, closing down their economies for a period of (at least) several weeks, and then relaxing the lockdown restrictions gradually as the evolution of the epidemic permits. Both the policy response – the severity of the lockdown – and the public response – the degree to which the public have adhered to government guidelines and the extent to which they have behaved in a manner that either anticipated or enhanced these guidelines (for example keeping their children away from school) – are likely to be relevant in determining the effectiveness of the approach. The workhorse model of the development of a pandemic is the Susceptible-InfectedRemoved (SIR) model (Kermack and McKendrick, 1927). In this model, an epidemic begins when some members of a large population of susceptible people become infected by a small number of virus carriers. Eventually these people are removed (either by recovery or death), so that, over time, the numbers of susceptible people decline and the numbers removed increase. The numbers infected start small, increase as infected people pass the virus on to (other) susceptible people, and then eventually decline as the numbers of susceptible people (who are not already infected or removed) falls. The SIR model implies that the steady state of the infection rate is zero, but that the infection rate is above this during the epidemic period. Where the infection spreads rapidly in the initial stages, the infection rate will peak at a relatively high level before falling. The rate of spread of the infection has been a key consideration during the COVID-19 pandemic. The reproduction number, R0, is defined as the expected number of further cases that are directly generated by a typical case that has the infection. Early estimates of R0 for COVID-19 were between 2 and 3 (Li et al., 2020; Zhang et al., 2020). These caused concern because any number above unity implies that the virus will spread to the entire susceptible population and cannot be contained. But, since the virus, in order to spread, needs to come into contact with uninfected people, putting social distancing measures in place can serve to reduce R0. This has been the aim of lockdown measures. Indeed, Gans (2020) has suggested that, once these behavioural considerations are accommodated in the modelling, the steady state value of R0 must be one. The malleability of R0 challenges the SIR model which, in its basic form at least, assumes a constant infection rate. Consequently, infections may not decline to a steady state of zero; indeed, there may be no steady state; and in the case considered by Gans, the steady state may be path dependent. In this last context, it is appropriate to consider the processes that determine outcomes in different countries. We see here a parallel to the extensive literature on the cross-country convergence of output (Barro and Sala-i-Martin, 1991, 1992; Quah, 1996). Early papers introduced the notions of β-convergence (based on a regression coefficient) and σ-convergence (based on the spread of growth rates across countries). Simple models of convergence in output receive mixed support from the empirical literature; on the one hand convergence does seem to Review of Economic Analysis 16 (2024) 393-408 396 www.RofEA.org happen, but on the other it is excruciatingly slow. The latter observation has led to the idea that more rapid convergence may be observed within convergence clubs – aggregations of countries within which economies are converging, but between which economies may either be not converging at all or converging at a slower pace (Ben David, 1998; Canova, 2004). Econometric methods developed by Phillips and Sul (2007, 2009) are designed to accommodate the opportunity that panel data provide to accommodate heterogeneity across countries, while still identifying clusters of countries that demonstrate panel convergence (analogous to σconvergence). 1 Christopoulos and Eleftheriou (2020) have recently applied analytical tools drawn from this literature to the context of medicine, and we build on that work in the following sections by applying these methods to the spread of COVID-19. Any clustering of countries into convergence clubs that exhibit similar pandemic outcomes is likely to be a function of government policy responses and public responses to the challenge. These reactions may, in turn, be affected by various factors, including government effectiveness (Liang et al., 2020; Tatar et al., 2021) as well as voice and accountability (Tatar et al., 2021). Pandemic results may also be related to the degree of education among a country’s citizenry. This is suggested by findings showing education to be positively correlated with diverse improved health outcomes (Clark and Royer, 2013; Kuhlánová et al., 2014). It is possible that this correlation is partly due to more education leading to more health-friendly thinking and decision-making (Cutler and Lleras-Muney, 2006), which may result in a nation’s citizens taking increased precautions to avoid contracting the COVID-19 virus. Furthermore, evidence that increased education reduces mortality due to preventable causes (Grytten et al., 2020) suggests the possibility that more education may lead to more COVID-19-preventive behavior. It is widely known that the pandemic has affected higher education by necessitating measures to reduce student-student and student-teacher contact, with particular measures varying from country to country (Crawford et al., 2020; see also, e.g., Agasisti & Soncin, 2021 and Jung et al., 2021). But is the converse also true – that is, does higher education have an effect on the pandemic? Of particular interest in this research is whether the public response to the pandemic is related to the percentage of a population that has completed tertiary education. One effect of higher education may be to increase the number of jobs that can be done at home, thereby reducing workers’ virus exposure. Indeed, research suggests that tertiary education may be correlated with an increase in the availability of home-based work (Dingel and Neiman, 2020). Another pandemic-relevant factor positively related to higher education is prosocial 1 Phillips and Sul methodology has a number of advantages compared with other club convergence techniques, such as: independence from theoretical underpinnings based on growth theory, robustness against small-sample problems, endogenous identification of convergence clubs, appropriateness in the case of temporal transitional heterogeneity (for more details on the advantages of the Phillips and Sul approach, see Apergis et al. (2013)). ELEFTHERIOU, IOAKIMIDIS Covid-19 Policy in Education and Virus Transmission 397 www.RofEA.org orientation, which consists of attitudes of behaving in socially responsible ways toward others (Brandenberger and Bowman, 2015). Having a prosocial attitude has been found to be positively related to the adoption of health measures, such as wearing face masks and practicing physical distancing, designed to prevent the spread of the COVID-19 virus (Campos-Mercade et al., 2021). However, in contrast, a study in Switzerland found that higher-educated youth were less compliant with health measures meant to mitigate infection by the COVID-19 virus (Nivette et al., 2021), a result suggesting that an increased percentage of higher education within a population may actually worsen the public response. Furthermore, despite the correlation between education and health measures, recent studies have found little evidence that there is a causal relationship between degree of education and health outcomes (Albarrána, HidalgoHidalgo, & Iturbe-Ormaetxe, 2020; Lynch & von Hippel, 2016; Xue, Cheng, & Zhang, 2021). Given these varying findings, there is an evident need better to understand what the relationship may be between a citizenry’s education, including tertiary education, and their country’s pandemic outcomes. 3 Methodology The first step in our analysis involves applying the method of Phillips and Sul (2007, 2009) to the data on COVID-19. The data allow us to model, as dependent variable, the rate of deaths from infection, X, in each jurisdiction, i, at each point in time, t. The essence of the Phillips and Sul method is to decompose the dependent variable, Xit, into a common component, μt, and an idiosyncratic component, δit, such that Xit = δit μt (1) The idiosyncratic component is itself modelled as the sum of a time-invariant component, δi, and a term that combines scale effects, σi, random effects, ξit, and a slowly varying structure, L(t)-1t-α. Convergence, which implies δi = δ and α ≥ 0 can then be formally tested as a null hypothesis against the alternative hypothesis by estimating the following equation and obtaining a robust t statistic for the b ˆ coefficient, , ˆ log ˆ ˆ )(log2log 1 t t utbctL H H++=−         (2) where (H1/Ht) denotes the cross sectional variance ratio. If the null hypothesis is rejected, then all geographies do not converge, and Phillips and Sul recommend a procedure for identifying convergence clusters. This involves starting with all possible core clubs, and repeatedly adding states to these clubs and testing for convergence within (but not across) the clusters. Review of Economic Analysis 16 (2024) 393-408 398 www.RofEA.org A series of Stata commands in the PSECTA package (Du, 2017) allows this procedure to be conducted relatively easily. Unless otherwise noted in the text, we use default values of parameters. Given the widely different experiences of different countries affected by the COVID-19 pandemic, our prior expectation is that there is more than one convergence club. Indeed, if it were to transpire that there were only one, our approach would add little to the existing literature. Assuming therefore, for the time being, that more than one cluster is identified, our analysis proceeds to a second stage in which a logit/probit model (a multinomial logit/probit in the case of more than two clusters) is used to explain how countries are allocated across convergence clubs. 4 Data The data we use in this study come from a variety of sources. The data for new COVID-19 deaths per million inhabitants (ndpm) were extracted from https://ourworldindata.org/coronavirus and cover the period from May 4 through August 31, 2020. These data allow us to conduct the first stage of the analysis. For the second stage, however, the latest available data on a wider range of variables are needed. Data on educational attainment in each country are drawn from the updated Barro and Lee (2015) data set from which we obtained projections for 2020 of tertiary education completed in each country. Data on population age structure were retrieved from https://ourworldindata.org/coronavirus. Data on government effectiveness and voice and accountability come from the World Bank’s governance data set (available at www.govindicators.org). From these data we used estimates ranging from approximately -2.5 (weak) to 2.5 (strong). The World Government Indicators (WGI) methodology paper (Kaufmann et al., 2010) provides details on data sources, aggregation method, and indicator interpretation. Descriptive statistics for all variables are reported in Table 1. For most variables we only have one observation per country – and the value of this observation does not vary across the short period under consideration. For the COVID-19 deaths variable, however, we have dayspecific observations within each country. These data cover 54 countries over the period from May 04, 2020 through August 31, 2020, and so there are 6,480 daily observations forming a balanced panel. The countries in the sample are varied in character, including some from all continents and all levels of development. Since some countries suffered their first experience of COVID-19 earlier than others, at the stage at which these data were collected the countries represented here are at various points in the cycle of the epidemic. ELEFTHERIOU, IOAKIMIDIS Covid-19 Policy in Education and Virus Transmission 399 www.RofEA.org Table 1: Descriptive statistics N Mean S.D. Min Max ndpm 6,480 0.824 2.523 0 119.34 ndpm_w 6,480 3.724 2.523 2.900 122.24 ndpm_m 6,480 1.120 2.523 0.296 119.64 aged 65 older 54 12.04 6.686 1.144 27.05 voic_acc 54 0.330 1.007 -1.450 1.730 gov_eff 54 0.622 0.946 -1.460 2.230 tertiary 54 14.55 9.585 0.400 43.60 days_100 54 20.70 36.646 0 116 Notes: ndpm = new COVID-19 deaths per million inhabitants; ndpm_w = weekly-corrected new COVID-19 deaths per million inhabitants; ndpm_m = monthly-corrected new COVID19 deaths per million inhabitants; gov_eff = government effectiveness index; voic_acc = voice and accountability index; tertiary = percentage of population aged 25-64 who have completed tertiary education; aged 65 older = percentage of population aged 65 and above; days_100 = number of days since 100 total COVID-19 deaths. To account for this effect, we time-correct the ndpm variable as follows: First, we regress ndpm on weekly (or monthly) dummies and save the estimated coefficient for each dummy. Second, we subtract the above estimated coefficients from ndpm to create the time-corrected variable. We chose to analyze data from 04 May to 31 August 2020 because this was the interval during which early pandemic-inspired lockdowns had passed for most, if not all, of the countries in our sample. Note that, owing to lack of data availability and the need to obtain a balanced panel, some countries have been excluded. We wanted to learn how policy and public factors affected COVID-19 outcomes during the period when restrictions were somewhat loosened. During this interval, citizens had greater freedom in their behavioural responses to the pandemic. We especially wanted to learn how individuals with a tertiary education responded to the pandemic when they had greater latitude in choosing to exercise or not exercise preventive measures such as wearing masks, social distancing, and not attending large gatherings. Beginning in September 2020, the emergence of the so-called second wave of the pandemic resulted in governments again enforcing stricter guidelines and in some cases instituting a lockdown. Citizens no longer had the degree of latitude they had enjoyed during the previous four months in deciding the extent to which they would engage in behaviours to prevent the spread of the virus. The undertaking of preventive actions became more a function of governmental decrees than it had been previously. Thus, the period from May through August offers a unique window through which to view whether having a higher education was related to pandemic outcomes that might be attributed to relatively self-determined preventive behaviours. 5 Results Review of Economic Analysis 16 (2024) 393-408 400 www.RofEA.org Applying the Phillips and Sul (2007, 2009) methods to the time-corrected new COVID-19 deaths per million inhabitants (ndpm) data allows identification of two distinct convergence clubs in each case. The membership of these clubs is reported in Table 2. There is a sharp distinction between the two groups in that the mean number of weekly-corrected (monthlycorrected) new deaths per million is 5.6515 (2.2582) in convergence club 1, but is 3.4369 (0.7594) in convergence club 2. That said, the composition of the two clubs warrants some comment. Some countries with high incidence of COVID-19 death, such as Belgium and the United Kingdom, appear in convergence club 2 alongside other countries, such as New Zealand, that have been more successful in restricting the death count. Meanwhile, convergence club 1 likewise comprises both high incidence countries such as the United States and Brazil, and low incidence countries such as Australia. In interpreting our results, it is important therefore to note that membership of one club rather than the other concerns the process of convergence to an equilibrium, not the mean value of ndpm itself. All in all, while the countries of each convergence club may exhibit high variability in terms of COVID-19 mortality, they will eventually converge in the long-run. Figure 1 illustrates the transition paths in each case. These confirm that the divergence between the two clubs rises with the passage of time. Figure 1. Transition Paths Notes: ndpm_w = weekly-corrected new COVID-19 deaths per million inhabitants; ndpm_m = monthly-corrected new COVID-19 deaths per million inhabitants. 01234 01may2020 01jun2020 01jul2020 01aug2020 01sep2020 date Club 1 Club 2 Transition paths for ndpm_w 01234 01may2020 01jun2020 01jul2020 01aug2020 01sep2020 date Club 1 Club 2 Transition paths for ndpm_m ELEFTHERIOU, IOAKIMIDIS Covid-19 Policy in Education and Virus Transmission 407 www.RofEA.org Discacciati, A., Orsini, N., Greenland, S. (2015). Approximate Bayesian logistic regression via penalized likelihood by data augmentation. The STATA Journal 15(3), 712-736. Du, K. (2017). Econometric convergence test and club clustering using Stata. The STATA Journal 17(4), 882-900. Gans, J. (2020). The economic consequences of R=1: towards a workable behavioural epidemiological model of pandemics, https://osf.io/preprints/socarxiv/yxdc5. Grytten, J., Skaub, I., Sørensen, R. (2020). Who dies early? Education, mortality and causes of death in Norway. Social Science & Medicine 245, 112601. Jung, J., Horta, H., Postiglione, G. A. (2021). Living in uncertainty: the COVID-19 pandemic and higher education in Hong Kong. Studies in Higher Education 46(1), 107-120. Kaufmann, A., Kraay, A., Mastruzzi, M. (2010). The worldwide governance indicators: a summary of methodology, data and analytical Issues. World Bank Policy Research Working Paper No. 5430. http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1682130 Kermack, W., McKendrick, A. (1927). A contribution to the mathematical theory of epidemics. Proceedings of the Royal Society A 115(772), 700-721. Kulhánová, I., Hoffmann, R., Judge, K., Looman, C., Eikemo, T.A., Bopp, M., Deboosere, P., Leinsalu, M., Martikainen, P., Rychtaříková, J., Wojtyniak, B., Menvielle, G., Mackenbach, J.P. (2014). Assessing the potential impact of increased participation in higher education on mortality: Evidence from 21 European populations. Social Science & Medicine 117, 142149. Li, Q., Xuhua, G., Wu, P., Wang, X., Zhou, L., Tong, Y., Ren, R., Leung, K., Lau, E., Wong, J., Xing, X., Xiang, N., et al. (2020). Early transmission dynamics in Wuhan, China, of novel coronavirus infected pneumonia. New England Journal of Medicine 382, 1199-1207. 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, 12567. Lynch, J.L., von Hippel, P.T. (2016). An education gradient in health, a health gradient in education, or a confounded gradient in both? Social Science & Medicine 154, 18-27. Nivette, A., Ribeaud, D., Murray, A., Steinhoff, A., Bechtiger, L., Hepp, U., Shanahan, L., Eisner, M. (2021). Non-compliance with COVID-19-related public health measures among young adults in Switzerland: Insights from a longitudinal cohort study. Social Science & Medicine 268, 113370. Phillips, P., Sul, D. (2009). Economic transition and growth. Journal of Applied Econometrics 24(7), 1153-1185. Phillips, P., Sul, D. (2007). Transition modelling and econometric convergence tests. Econometrica 75(6), 1771-1855. Quah, D. (1996). Empirics for economic growth and convergence. European Economic Review 40(6), 1353-1375. Review of Economic Analysis 16 (2024) 393-408 408 www.RofEA.org Tatar, M., Faraji, M.R., Montazeri Shoorekchali, J., Pagán, J.A., Wilson, F.A. (2021). The role of good governance in the race for global vaccination during the COVID-19 pandemic. Scientific Reports 11, 22440. Xue, X., Cheng, M., Zhang, W. (2021). Does education really improve health? A meta-analysis. Journal of Economic Surveys 35(1), 71-105. Zhang, S., Diao, M.Y., Yu, W., Pei, L., Lin, Z., Chen, D. (2020). Estimation of the reproductive number of novel coronavirus (COVID-19) and the probable outbreak size on the Diamond Princess cruise ship: A data-driven analysis. International Journal of Infectious Diseases 93, 201-204.