key: cord-0225926-pkbwgyrs
authors: Amodio, Emanuele; Battisti, Michele; Gravina, Antonio Francesco; Lavezzi, Andrea Mario; Maggio, Giuseppe
title: School-age Vaccination, School Openings and Covid-19 diffusion
date: 2022-03-23
journal: nan
DOI: nan
sha: 5abaedda03df509df6ac154591bb5cf6c9812f09
doc_id: 225926
cord_uid: pkbwgyrs

Do school openings trigger Covid-19 diffusion when school-age vaccination is available? We investigate this question using a unique geo-referenced high frequency database on school openings, vaccinations, and Covid-19 cases from the Italian region of Sicily. The analysis focuses on the change of Covid-19 diffusion after school opening in a homogeneous geographical territory. The identification of causal effects derives from a comparison of the change in cases before and after school opening in 2020/21, when vaccination was not available, and in 2021/22, when the vaccination campaign targeted individuals of age 12-19 and above 19. The results indicate that, while school opening determined an increase in the growth rate of Covid-19 cases in 2020/2021, this effect has been substantially reduced by school-age vaccination in 2021/2022. In particular, we find that an increase of approximately 10% in the vaccination rate of school-age population reduces the growth rate of Covid-19 cases after school opening by approximately 1.4%. In addition, a counterfactual simulation suggests that a permanent no vaccination scenario would have implied an increase of 19% in ICU beds occupancy.

The Covid-19 pandemic determined the flourishing of a substantial amount of studies addressing the health and socio-economic determinants of its diffusion, aiming also at identifying measures tied to contain its direct and indirect costs for the society. Schools, and the students' population, have been a crucial aspect of the discussion for at least two reasons. On the one hand, among the restrictive policies implemented during the first and second wave of the pandemic, when anti-Covid-19 vaccines were not available, school closure has been a measure widely adopted, together with more general measures of lockdown of economic activities. Closing the schools, and implementing distance learning, was based on the assumption that the interactions implied by attending schools might have been an important driver of the spread of Covid-19 in the population. The recent literature addressing the issue, often exploiting school openings for identification purposes, found mixed results (see Svaleryd and Vlachos, 2022 , for a detailed survey). On the other hand, closing the schools raised concerns on the costs in terms of lost opportunities of accumulating human capital, as well as on the psychological costs of the students and distress of the students' families (see Stantcheva, 2022, and references therein).

In December 2020, following the approval by the Food and Drug Administration in the US and the European Medicines Agency, anti-Covid 19 vaccines became available and recommended for individuals older than 16 years, and were subsequently approved in May 2021 for adolescents in the age bracket 12-15. In this new context, some crucial questions naturally arise: when vaccination is available to the students' population, do school openings still represent a potential triggering factor of Covid-19 diffusion? Is school closure, therefore, still to be recommended as an effective mitigation policy?

In this article we try to answer these questions by analyzing granular data from the Italian region of Sicily, comparing the effect of school openings during the school year 2020/21, when vaccines were not available, to the effect in 2021/22, when vaccines were available for the students' population and for the population at large. Considering granular data from a homogeneous territory and the same period of the year may help to account for a wide range of social and institutional confounding factors and for the effects of seasonality.

Our analysis relies on a dataset obtained by merging geo-localized data on Covid-19 cases, information on age vaccination exposure, schools' geographical location and school opening dates. In particular, we build an indicator of local (i.e. at census micro-area level) vaccine exposure from detailed data on daily vaccinations by age at municipal level, and on the demographic structure of the census area population.

Our results suggest that school-age vaccination played a major role in reducing cases diffusion after school openings in Sicily. Specifically, we show that school openings in 2021 is associated to a differential impact when compared to 2020, with almost no effects on Covid-19 diffusion at the local level. The positive effect of school openings on Covid-19 diffusion, identified by Amodio et al. (2022) , appears to be fully mitigated by vaccination, with higher effects of vaccination in areas where the share of vaccinated school age population is lower and in areas with lower population density (that in the literature is often associated with higher and denser social interactions as we discuss below). In addition, we show by a counterfactual analysis that the diffusion of vaccination is associated to a reduction of hospitalizations in ICU of approximately 19%.

The work is organized as follows. Section 2 provides a review of the relevant literature; Section 3 describes Covid-19 transmission and vaccinations in Sicily, while Section 4 introduces the dataset. Section 5 specifies the econometric models we utilize. Section 6 presents the main results, a set of robustness checks and a heterogeneity analysis; Section 7 concludes.

Covid-19 Diffusion: a Review of the Literature Our contribution speaks to three related strands of literature. The first one is about the effects of school openings on Covid-19 diffusion. The second refers to the efficacy of school closures as a mitigation policy. The third relates to the overall effect of vaccination in containing the epidemic.

The effects of school opening as a trigger of Covid-19 diffusion is addressed by several studies.

Among these, works such as Amodio et al. (2022 ), Chernozhukov et al. (2021 , Vlachos et al.

(2021) find a positive effect of school openings on the growth rate of Covid-19 diffusion in a range of 2%-5%. Differently, other studies find nonsignificant or mixed effects. In particular, Isphording et al. (2021a) analyze the German case taking into account the school openings after the summer break of 2020, and do not find a positive impact on Covid-19 diffusion. Yoon et al.

(2020) focus on the first wave of Covid-19 in Korea and investigate the step-wise transition from school closure to opening, both online and offline, and find that offline schooling did not lead to any substantial outbreak among the youths. Other works, such as Keeling et al. (2021) study eight different school reopening strategies implemented for primary and secondary schools in England. The authors find a high degree of heterogeneity, with half-sized classes or younger cohorts not associated to increased levels of infections, and the reductions in community social distancing exacerbating the impacts of school openings. Finally, Ertem et al. (2021) find regional differences of school openings in the US, with no significant effects in most of US states and positive effects in Southern states, in a setting where different teaching methods (in-person, hybrid, remote) is controlled for. Our paper, however, is the first one that compares the effect of school openings on Covid-19 diffusion in a period with no vaccines available to a period in which vaccines were introduced and gradually extended to different age cohorts of the population, focusing on the effect of vaccination. 1

The effect of school closures as a mitigating policy, especially during the first wave of Covid-19 diffusion is analyzed, for example, by Flaxman et al. (2020) who find a general, positive effect of lockdowns on Covid-19 transmission, but cannot identify a specific effect of school closures.

This depends on the fact that they rely on a cross-country study which does not allow to 1 A partial exception is Isphording et al. (2021b) , who also analyze a period in which vaccination was available in Germany. However, they do not focus on the effects of vaccinations but on the effect of mandatory testing in German schools, concluding that this can be an effective policy in containing the infection while schools are open. This is all the more important given that, as mentioned in Section 1, school closures imply a host of negative effects, such as losses in human capital accumulation and earnings over the lifetime (see, e.g., Agostinelli et al., 2022; Psacharopoulos et al., 2021; Fuchs-Schundeln et al., 2020, and Stantcheva, 2022, for diffusion in these two periods, we will propose an estimation by a counterfactual analysis of the reduction in ICU hospitalizations due to school-age vaccinations. This may help to give a more variegated policy answer based on cost-benefit analysis than opening or closing schools (and implementation of other kind of restrictions), insofar as there exists the additional element of health and non-health interventions.

Sicily and the South of Italy have been only marginally affected by the Covid-19 pandemic during the first wave, but the level of Covid-19 cases substantially increased from the second wave onwards. Figure 1 displays the population-weighted cases (by quintiles) at the end of the summers of 2020 and 2021. While the number of cases are substantially higher in the second year (right panel), their spatial distribution follow similar paths, with more cases in more touristic areas, such as the North-West and East coasts, and less cases in the inner areas, which are less populated and often mountainous. age 12-19, 3 and the population of age greater than 19. In particular, we see in Figure 2 (left panel) that in the school-age population the average is lower, which is expected because of the different timing of the vaccinations, but the variance is much higher than in the older group. Given these patterns of Covid-19 cases and vaccinations across space, we will consider the case in which Covid-19 contagion spreads locally and then has further spillovers. Specifically, we will follow the strategy of Amodio et al. (2022) and assume that contagion may be triggered by school openings in areas geographically close to the schools. In particular, we will assume that school openings affected the geographic micro-areas (i.e. the Italian census cells) within a ray of one km from the schools (in case of more than one school we will use a weighted average based on the number of students). In the next section we provide the details on the dataset.

The analysis relies on a set of data on 33,604 census cells observed four weeks before and To capture the effect of school openings, we build a dummy activating when at least one school attended by students in the age bracket 12-19 opens within a ray of 1km from the centroid of a census area. This variable is equal to zero before the week in which school opened and takes on a value of one afterwards, in both school years 2020/21 and 2021/22. We focus only on the 4,223 public schools in Sicily, as these contain more than 95% of students in the region, for grades where school is compulsory (in the periods considered Sicily counts approximately 790,000 students, of which 240,000 in high schools).

From ISS (2020) we also gather data on vaccinations, which includes information on the number of administered first and second doses by age. Differently from the Covid-19 cases, these data do not contain the indication of the census cell of residence of the vaccinated individual, so we can only utilize data on vaccinations at municipal level. We use these data to develop a set of indicators on vaccination coverage by age. First, for consistency, we create a variable on the total number of vaccinated individuals by class age and municipality. Then, we combine the vaccination data with population data from ISTAT to build two indicators on the share of the 12-19 population and on the share of the population of 19 or older, who received the first and second dose in a given municipality. Finally, we combine vaccination information at municipality level with age-structure information at cell level from the 2011 Census to develop a cell-level indicator measuring exposure to vaccination in a census cell i as follows: This logic of this variable is that the higher the share of vaccinated population in an age group in a municipality, and the higher the share of population of that age group in a census area, the more exposed the census area is to vaccination. The implicit assumption is that the within a city vaccination rates for an age group (i.e. 12-19) are quite similar across census cells. 7

Finally, we collected data on the total number of beds in intensive care units (ICU) in the municipality's hospitals, and on the distance of the municipality centroid from the closest hospital with a ICU, gathered from data on the health structures from the Ministry of Health. 8

For a municipality hosting at least one hospital with ICU, the distance indicator takes value zero. For a municipality without a hospital with ICU, the number of beds takes a value equal to zero, while the distance takes a real positive value. Table 1 reports the descriptive statistics of the variables, both at the municipality and census cell level, that we will use in the econometric analysis. 9 In the next section we describe our empirical strategy.

The empirical strategy relies on three different specifications. The first specification captures the differential effects between school opening in 2020/21 and 2021/22. We do this by using a Diff The empirical specification takes the following form:

In Table 1 by "share" we refer to the number of vaccine doses divided by the population. For example, an individual that received two doses is counted twice in the calculation of the share. For this reason, the share referred to receiving at least one dose can be greater than one. 10 With respect to Amodio et al. (2022) , the time windows under consideration are shorter. This is done to avoid further confounding factors, such as the beginning of the colder season and the spread of the Omicron variant, which was observed from November 2021 in Italy.

with i = 1, 2, . . . , n; t = 1, 2, · · · , T ; y= 2020/21, 2021/22;

where ∆lnCovid i,t,y denotes the growth rate of Covid-19 cases in census area i in week t and school year y. The equation includes the lag of the dependent variable and the term lnCovid i,t−j , denoting the natural logarithm of Covid-19 cases in the same census area i, measured at times t − 3 and t − 4, following an approach similar to Chernozhukov et al. (2021) , who show that this can be seen as an empirical specification derived from a theoretical SIR model (see also Amodio et al., 2022) . The main explanatory variable of interest is the coefficient λ of the dummy S i,t−2 on school openings, which takes value 0 in the four weeks before the opening, and 1 after the week of school opening. As in Amodio et al. (2022) , this variable enters with a 2-week lag to account for the time to detect the contagion, defined also as the serial time of infection. The coefficient λ captures the impact of school opening on Covid-19 diffusion and can be interpreted as a percentage increase in the growth rate. The term η denotes the coefficient of the interaction between the school opening dummy S i,t−2 and the year dummy Y ear 2021/22 , and it aims at capturing the differential effect of school opening in the second

year of the pandemic. 11 The model includes census area fixed effects C i , which account for short term time-invariant unobserved characteristics, such as the population profile or level of education, and week dummies T t , which account for common shock in time, such as an increase in the number of Covid-tests available from a given week onwards. Finally, the term u i,t is a robust error term clustered at census area level.

In a second specification we study whether the differential increase in cases during the four weeks following school opening in 2021/22 and in 2020/21 can be attributed to school-age vaccinated population, and how this effect compares to the one associated to the vaccinated population of age higher than 19. 12 To do so, we average our dependent variable across the four weeks after school opening in every census cell i, and take the first-difference of this new variable across the two school years. This is done to avoid zero-inflated regressions given the high level of granularity of our dataset, and the fact that the post-summer period brings to few cases in many census areas. Defining k as the week of school opening we have, therefore,

The new model, having Y i as dependent variable, takes the following form:

where Exp V accines age=12−19,i and Exp V accines age>19,i are two indicators of the exposure to vaccines for individuals of age between 12-19 and above 19 at census area level, built following the procedure introduced in Section 4. Given that vaccination was implemented only from December 27th 2020, these shares are equal to zero for the school year 2021/20, making ∆Exp V accines age=12−19,i = Exp V accines age=12−19,i and ∆Exp V accines age>19,i = Exp V accines age>19,i , respectively. The term Y i,k−4 denotes the lag of the dependent variable measured the four weeks before the school opening. Additionally, the control variables also include a set of municipality-level dummies to capture residual local unobserved characteristics (we exclude them in the initial specification because the municipality dummies are dropped by differencing). Equation (3) can be considered as a first-difference version of Equation (2) where the dependent variable is averaged across four weeks, to avoid zero inflated regression, and where the census area fixed effects and the week dummies are canceled out. In this framework, we still account for the initial conditions of the Covid-19 process at school openings, to control for possible differences in the phase of the epidemic between 2021/22 and 2020/21. Finally, ω i is the error term clustered at the census area level.

In this section we present the results of the estimation of Equations (2) and (3). In particular, Section 6.1 illustrates our benchmark findings, while Section 6.2 contains the results on the identification of heterogeneous effects and of robustness tests. Section 6.1 also contains the results of a counterfactual analysis in which we estimate the effect of vaccinations on the number of of Covid-19 cases and on ICU beds occupancy by Covid-19 patients. Table 2 reports the results of the estimation of Equation (2). In particular, Columns (1) and (2) show the coefficients from Equation (2) The estimated coefficient for school opening in 2021 is not significantly different from zero, indicating that school opening is not associated to an increase in the growth rate of Covid-19

cases. 14 This finding suggests that school opening after vaccination was made available did not affect the diffusion of Covid-19, pointing to the fact that a restriction such as school closure may be not justified under these conditions. This result is confirmed when we estimate the specification on the full sample. Resulst in Column (3) show that the coefficient for school opening remains positive and significant, while the coefficient of the interaction between school opening and the year dummy for 2021 is negative, generating a null overall effect. Once again, this provides an indication that school opening in 2021 had no significant effects on Covid-19 cases diffusion, as also confirmed by the t-test on the equality of the absolute value of two coefficients reported at the bottom of Column (3).

Finally, we slightly modify the baseline specification by adding an indicator on the number of vaccinated individuals (with second dose) of school age calculated at the time of school 13 The consideration of other time windows is discussed in Section 6.2. 14 While in principle school opening may be considered as endogenous, in practice in 2020 it corresponded to a staggered design due to a national referendum for which some schools were used as polling stations (see Amodio et al., 2022 Table 3 reports the coefficients obtained from the estimation of Equation (3), where the effect of school-age vaccination is correlated to the average growth rate of Covid-19 cases in a window of four weeks after the school opening. Column (1) shows that school-age vaccination, measured at cell level through the exposure variable (second dose), is associated to a strongly significant negative effect. This effect is robust to the inclusion of: i) the initial conditions,

i.e. the value of the average number of cases in the four weeks preceding the school opening (Column (2)); ii) the inclusion of Municipality FE (Column (3)); iii) vaccination exposure in the population older than 19 years (Column (4)). Indeed, the coefficient related to vaccination of school-age population increases in magnitude with these controls. In terms of magnitude, the coefficient on the exposure reported in Column (4) suggests that a 1% increase on the average level of exposure to vaccination of individuals aged 12-19 is associated to a decrease of about 0.14% in the growth rate of Covid-19 cases. 15 Notes: ***, **, and * denote significance at 1%, 5%, and 10%, respectively. Standard errors are clustered at census area level.

As an implication of the results in Table 3 in what follows we propose an estimation of the reduction in Covid-19 cases and of hospitalizations in ICU for Covid-19 patients, implied by the vaccination of the school-age population. Specifically, we compute the cumulated value of the cases we would have observed had vaccinations not be administered, by using the estimated coefficients from Column (4) of Table 3 , and setting to zero the values of school-age vaccination shares.

We first build the counterfactual difference between Covid-19 cases's growth rates in 2020

and 2021 with school age shares of vaccinated individuals equal to zero. Then we replace this difference to the Covid growth rate of cases in 2020 to build the counterfactual growth rate of 2021. Finally, starting from the level of the Covid cases 2021 before school openings, we derive the cumulated cases of 2021 4 weeks after school opening. This exercise shows that, without accounting for the vaccinations, the new weekly cases at municipality level would have increased on average by 11% (407 vs 362). The average municipality number of ICU beds in Sicily is 1.02 but this number is driven by small municipalities without ICU beds, otherwise the average number would be 13.8. 16 Then, in a very conservative scenario for this smaller set of municipalities with ICU beds, this means reaching the critical ICU parameter for the declaration a status of "yellow zone" and "red zone" respectively with the second and fifth occupied ICU bed by a Covid-19 patient.

To build a counterfactual scenario in terms of ICU beds occupancy, consider that the ratio between active cases and ICU patients was 236.5 in Sicily in the four weeks after the 2021 school openings of September 16th. This implies that the additional cases implied by our counterfactual scenario with no school-age vaccinations would have implied a 19.03% increase of ICU bed occupancy by Covid-19 patients, a remarkably high number.

We carried out a set of exercises to test whether the relationship between Covid-19 cases, vaccination and school opening may be heterogeneous across a set of socio-demographic characteristics that may be linked to Covid-19 diffusion. Specifically, we test for heterogeneity across population density, classroom size and vaccination share of the whole population, by estimating Equation (3) on two subsamples deriving from splitting the sample by the median value of these dimensions. 17 Figure 4 reports the whisker plot of the coefficients obtained from this exercise, while Table A1 in the Appendix reports the full sets of coefficients.

As Figure 4 shows, the coefficients of the exposure variables are all significant and negative, and the confindence intervals for most of them largely overlap. Interestingly, the only dimension for which the two coefficients do not largely overlap is the density of population, suggesting that the effect of vaccination exposure is larger in absolute terms for areas with lower population density.

This result is similar to Amodio et al. (2022) , who find a larger effect of the 2020 school opening in low population density zones. This finding is likely due to the higher degree of social interactions in smaller sized communities as pointed out in Sato and Zenou (2015) . Finally, we test the robustness of our main results through a set of additional exercises, that we compare to those of our preferred specification Column (3) of Table 2 , which are reported in Column (1) of Table A2 , which contains the results of the robustness tests. First, we test for eventual bias stemming from the inclusion of the lagged dependent variable in the dynamic process of Equation (2). This is done following the work of Chernozhukov et al. (2021) , which uses the Analytical Bias Correction estimator of Chen et al. (2019) to take into account the potential Nickell bias in this process. Column Column (2) of Table A2 shows that results are substantially unchanged. Then, we check whether the definition of the time-window may affect the results. We therefore extend the time window up to 8 weeks before and after the school opening and run the baseline specification, without not finding any appreciable difference (Column (3) of Table A2 . We then take into account potential spillover effects due by the school opening, by using the process of Conley (1999) , implemented by Colella et al. (2020) .

Column (4) of Table A2 ) shows that the coefficients remain consistent. Finally, we model school opening in 2020 through propensity scores, to take into account the small flexibility implied by the referendum, following the same procedure of Amodio et al. (2022) . Column (5) of Table A2 shows that he results remain robust when accounting for this potential source of endogeneity (see Amodio et al., 2022 , for more details on the use of propensity scores).

We also carried out a set of robustness tests on the specification of Equation (3), where we modified the definition of the dependent variable and used the share of population that received at least a single dose of vaccine. As reported in Table A3 , whose Column (1) contains our benchmark results, i.e. those in Column (4) of Table 3 , the coefficients appear slightly more variable in Columns (2) and (4), but this is likely to depend on the change in the scale of the dependent and independent variables, which are hereby computed differently from before.

Specifically, in Column (2) of Table A3 we consider an alternative specification in which we rearrange the time span of the regression as follows: (1) the dependent variable is calculated as the growth rate of Covid-19 cases in the 4 weeks after school reopening in 2020 and 2021, and (2) the control variable is given by the level of Covid-19 cases in the 4 weeks before school opening in 2020. Differently, in Column (4) of Table A3 we consider at least the first vaccination shot to build the exposure variable, whose average is 2.3 times bigger than in the benchmark case, so that if we apply this change to the coefficient we obtain -0.06, which is less than half of the benchmark value (-0.14) but it remains negative and significant.

The Covid-19 pandemic hit all countries as an unexpected shock in 2020 when few pharmaceutical tools were present. This implied that the first set of governmental reactions was mainly based on restrictions on individuals' mobility to lower the frequency of interaction and the consequent spread of the virus. Subsequently, medical treatments, mostly in the forms of vaccines, allowed to increase and improve the set of tools to contrast the Covid-19 diffusion, which started to imply a trade/off in using the restrictions. The evidence presented in this paper shows that school age vaccination had a substantial role in reducing, basically neutralizing, the effect of school openings on Covid-19 diffusion. With a focus on Sicily and on a rich set of granular data, while school openings were a substantial driver of cases in 2020, this effect disappears in 2021, when school-age vaccination was available.

In particular, our results show that an additional 1% of vaccination in the age cohort 12-19 (attending middle and high school) is associated to a decrease in 0.14% of Covid growth rate in post-summer school reopening period at local level. In addition, in a counterfactual analysis we show that school-age vaccination is associated to an estimated reduction of 19% occupancy of ICU beds by Covid-19 patients, which implies a sizeable effect on the possibility of municipalities to escape the restrictions which would be otherwise implemented by the State.

Several points are left to future agenda, in which we can consider how they may make our results appear to be a lower bound or an upper one. Specifically, in this paper we assumed that attitudes of vaccinated individuals do not change, 18 while individual restrictions affecting unvaccinated individuals in Italy (even if only for people older than 17, a small share of our class age) may imply that vaccinated individuals have higher mobility, so that the mitigation effect caused by vaccination should be counterfactually lower. Then, we could take into account that the school opening period is a post-summer period so that, due to the seasonality, we would observe less cases. This means that our results could be a lower bound, because in a period of larger spread of diffusion this effect could have been larger.

In this Appendix, we present some additional results. Table A1 contains the results on heterogeneity of the effects estimated from Equation (2). Table A2 contains the results of the robustness tests for the results from the estimation, respectively, of Equations (2) and (3). Notes: ***, **, and * denote significance at 1%, 5%, and 10%, respectively. Standard errors are clustered at census area level. Table 2 . ***, **, and * denote significance at 1%, 5%, and 10%, respectively. Standard errors are clustered at census area level. Notes: ***, **, and * denote significance at 1%, 5%, and 10%, respectively. Standard errors are clustered at census area level.

When the great equalizer shuts down: Schools, peers, and parents in pandemic times

The effects of school closures on covid-19: A cross-country panel analysis. Applied Health Economics and

Schools opening and covid-19 diffusion: Evidence from geolocalized microdata

Anticipation of covid-19 vaccines reduces willingness to socially distance

Elasticities and the inverse hyperbolic sine transformation

The early phase of the covid-19 outbreak in lombardy, italy

Mastering panel metrics: causal impact of democracy on growth

The association of opening k-12 schools with the spread of covid-19 in the united states: County-level panel data analysis

Optimal levels of vaccination to reduce covid-19 infected individuals and deaths: A global analysis

Acreg: Stata module to perform arbitrary correlation regression

Gmm estimation with cross sectional dependence

Impact of us vaccination strategy on covid-19 wave dynamics

The impact of school opening model on sars-cov-2 community incidence and mortality

Impact of nonpharmaceutical interventions (npis) to reduce covid-19 mortality and healthcare demand

Estimating the effects of non-pharmaceutical interventions on covid-19 in europe

Review of covid-19 vaccine subtypes, efficacy and geographical distributions

The long-term distributional and welfare effects of covid-19 school closures

No causal effect of school closures in japan on the spread of covid-19 in spring 2020

Safety and efficacy of nvx-cov2373 covid-19 vaccine

Does re-opening schools contribute to the spread of sars-cov-2? evidence from staggered summer breaks in germany

Schools under mandatory testing can mitigate the spread of SARS-CoV-2

Dati indice riproduzione

The impact of school reopening on the spread of covid-19 in england

Vaccination strategies and transmission of covid-19: evidence across advanced countries

The covid-19 cost of school closures in earnings and income across the world

How urbanization affect employment and social interactions

Inequalities in the times of a pandemic

Covid-19 and school closures

Characterizing the spatiotemporal heterogeneity of the covid-19 vaccination landscape

Effectiveness of 3 versus 6 ft of physical distancing for controlling spread of coronavirus disease 2019 among primary and secondary students and staff: a retrospective, statewide cohort study

The effects of school closures on sars-cov-2 among parents and teachers

Stepwise school opening and an impact on the epidemiology of covid-19 in the children