During the COVID-19 pandemic in 2020, officials in many countries used school closures to mitigate the virus spread. School closures may be an effective non-pharmaceutical measure against influenza-type outbreaks because children may amplify the virus transmission even if they do not often suffer severe symptoms themselves (Ferguson et al. 2006, Burgess and Sievertsen 2020, Oswald and Powdthavee 2020). School closures are often implemented partially, with students still attending school if they choose to (e.g. in Australia and the UK). As officials debate about using school closures to alleviate the spread of an epidemic, there are lessons to be learned from earlier epidemics.
In our recent work (Goulas and Megalokonomou 2020), we ask the following policy question: what is the effect of a relaxed attendance policy as a non-pharmaceutical measure against an epidemic on school attendance and performance? Because pandemics most harm students in the poorest neighbourhoods (United Nations 2020, Plümper and Neumayer 2020), we also ask: is there heterogeneity by socioeconomic status?
To answer these questions, we exploit exogenous variation from when Greece, fearing a swine influenza (H1N1-09 virus) outbreak, implemented a one-time policy allowing high school students to miss 30% more school periods without penalty during the spring semester of 2010. We combined novel higher-frequency attendance records and detailed transcript information from high schools in Greece and data on H1N1 cases from the Center of Disease Control and Prevention.
In particular, we use longitudinally linked transcript information from ten schools covering more than 4,000 students in grades 10, 11, and 12 between the 2005/06 and the 2010/11 school year. The transcript data contain a student identifier, school year, grade, courses taken, semester scores for each course, the grade point average (GPA) for each semester, and the number of absences (in periods) per semester. Fall semester scores reflect midterm assessment scores before the beginning of the spring semester. Spring semester scores reflect cumulative assessment scores at the end of the spring semester.
Much research has shown that having more school absences has a significant negative impact on school performance (Arulampalam et al. 2012). Our paper moves beyond previous studies in two important ways. First, we combine panel data with variation stemming from a natural experiment to mitigate two sources of endogeneity: time-invariant, individual-specific, unobserved heterogeneity such as parental supervision or personality traits; and grade-varying, common shocks such as teacher or student-age effects. Second, we use variation in higher-frequency school attendance that allows us to account for year-specific shocks.
What happened during the H1N1 pandemic?
In late spring 2009, the first sporadic cases of swine flu surfaced in Europe. The global 2009 flu pandemic involved a new strain of influenza A, subtype H1N1, and infected at least 125,550 people in Europe. There were 458 confirmed deaths in Turkey, 438 in Russia, and 299 in the UK.
In Greece, there were 20 cases by 14 June 2009; by 16 September (when schools start), the total number of cases had reached 2,149. The Hellenic Center of Disease Control and Prevention reported very few cases among people of high school age. However, the Ministry of Education, fearing an H1N1 recurrence, announced an increase in the upper limit of absences before retention, from 114 absences to 148 for that academic year only. Students did not need to provide a doctor’s note or their parents’ approval to use the extra absences.
Figure 1 shows the number of verified H1N1 cases and H1N1-related deaths for high-school-age individuals during the school year 2009/10. The Hellenic Center of Disease Control and Prevention reported that among 209,958 students attending high school that school year, 301 contracted the H1N1 strain and two of them died. The very few H1N1 cases among the high school population ease worries about a potential direct effect of the H1N1 virus on scholastic performance.
Figure 1 H1N1-infected high school students in Greece
Figure 2 plots the distributions of school absences during the semester under the standard attendance policy before or after spring 2010 and the relaxed attendance policy during spring 2010 (Pandemic Policy=1), showing a shift of the distribution to the right in spring 2010.
Figure 2 School absences distribution with and without a pandemic policy
School attendance and performance during a pandemic
To measure the effect of the relaxed attendance rule, we compare semester-specific school attendance and academic performance before/after the 2010 spring semester with during the semester when relaxed school attendance policy was in effect (Pandemic Policy=1). Academic performance measures include semester-level GPA and semester-level midterm scores in Greek, English, and mathematics. We use fall-semester GPA of grade 10 (‘early high school GPA’), which is the first instance academic performance is measured in high school, to proxy prior performance
A naive regression shows a negative association between absences and performance. We find a positive overall effect of the relaxed attendance policy on absences, which masks significant heterogeneity.
We estimate the effect of the relaxed attendance policy on absences and performance for students in different quintiles of prior performance. The results show that, as we move from lower to higher early high school performance, the estimated effect of the relaxed attendance policy on absences increases. At the same, the higher a student’s early high school performance, the higher the magnitude of the negative impact of relaxed attendance policy on GPA.
We find that students at the highest quintile of prior performance took 14 additional absences as a result of the precautionary relaxed school-attendance policy and their school performance decreased by 0.06 standard deviations. Students at the lowest quintile of prior performance kept going to school despite the relaxed attendance policy and their school performance increased by 0.072 standard deviations, which is not statistically different from zero. Figure 3 summarises the effects of the H1N1 pandemic-related relaxed attendance policy on absences and GPA.
Figure 3 Who goes to school during a pandemic?
Notes: Lighter colours represent statistical significance at weaker confidence levels.
Early high school GPA and neighbourhood income
Academically weaker students are less likely to make use of a relaxed school-attendance policy during a pandemic. One may expect that academically weaker students may be more likely to be found in the poorest neighbourhoods. If that is true, low socioeconomic status may be a driver of the diﬀerential eﬀects of a relaxed attendance policy.
We combine data from Goulas et al. (2018) and Goulas et al. (2020) to build a dataset on early high school performance and income (in 2009 euros) at the school’s postcode, covering 123 schools in Greece in the same year-span as our results presented above, to test whether prior performance is associated with neighbourhood income.
Figure 4 shows that early high school performance is positively associated with income at the school’s postcode, a proxy for socioeconomic status (ρ = 0.15). This suggests that students in the poorest neighbourhoods may be less likely to distance themselves from school during a pandemic.
Students in the poorest neighbourhoods may have limited out-of-school resources to make up for lost school learning during a pandemic, while their families may have limited work ﬂexibility, potentially decreasing the students’ propensity to distance themselves from school during a pandemic.
Figure 4 The association between early high school GPA and postcode income
The day after a pandemic
Our findings speak to policymakers considering precautionary school closures as a non-pharmaceutical measure to protect student outcomes from health-related situations. Our results provide a benchmark for the minimum effect a recovery education policy must have to balance out the loss in learning from precautionary school closures.
At the same time, our findings speak to the literature on the externalities in education for students across the distribution in prior performance. Students of lower prior performance suffer significant negative externalities, reflected in their school performance, from the school attendance of students with higher prior performance.
Arulampalam, Wiji, Robin A Naylor and Jeremy Smith (2012), “Am I missing something? The effects of absence from class on student performance”, Economics of Education Review 31(4): 363–75.
Burgess, Simon, and Hans Henrik Sievertsen (2020), “Schools, skills, and learning: The impact of COVID-19 on education”, VoxEU.org, 1 April. https://voxeu.org/article/impact-covid-19-education
Ferguson, Neil M, Derek A T Cummings, Christophe Fraser, James C Cajka, Philip C Cooley and Donald S Burke (2006), “Strategies for mitigating an influenza pandemic”, Nature 442(7101): 448–52.
Goulas, Sofoklis, and Rigissa Megalokonomou (2020), “School attendance during a pandemic”, Economics Letters 193: 109275.
Goulas, Sofoklis, Silvia Griselda and Rigissa Megalokonomou (2020), “Comparative advantage and gender gap in STEM”, IZA Discussion Paper 13313.
Goulas, Sofoklis, Rigissa Megalokonomou and Yi Zhang (2018), “Does the girl next door aﬀect your academic outcomes and career choices?”, IZA Discussion Paper 11910.
Oswald, Andrew, and Nattavudh Powdthavee (2020), “Individual COVID-19 fatality risk (and the consequences for universities)”, VoxEU.org, 6 June.
Plümper, Thomas, and Eric Neumayer (2020), “Wealthier districts were hit by Covid-19 first in Germany, but their lockdowns were more effective”, VoxEU.org, 11 June.