Schools have been fully or partially closed for an average of 39 weeks across the world due to Covid-19. The extent of the interruption has varied by region, with many poorer countries seeing a more severe impact on their education systems.
In the early weeks of the pandemic, it was estimated that 12 weeks of Covid-19 induced school closures could lead to learning losses of around 6% of a standard deviation in average test scores (Burgess and Sievertsen, 2020). While this prediction turned out to be surprisingly close to what data later showed in terms of the effect on grades, school interruptions ended up lasting for considerably longer (Engzell et al, 2021).
In early June 2021, around 65 weeks after the WHO declared Covid-19 a global pandemic, schools in the UK have been fully or partially closed for 27 weeks (UNESCO, 2019). While this has had a detrimental impact on British children’s learning and wellbeing, this is considerably below the global weighted average of 39 weeks of full or partial school closures.
Have all countries been affected equally?
As the map below shows, several countries have experienced more than 50 weeks of school closures. In the Americas and South Asia schools have been closed for a large number of weeks. Nepal, for example, has had 61 weeks of closures and in the United States schools were partially or fully closed for 56 weeks from March 2020 to May 2021. The map also reveals substantial within-continent variation. For example, in Uganda schools have been partially or fully closed for 59 weeks, compared with ‘only’ 13 weeks of closures in Burkina Faso.
Figure 1: Map of Covid-induced partial and full school closures, March 2020 to May 2021
Source: UNESCO. Note: Grey = no data
When did schools close and for how long?
In Figure 2 below, the number of partial or full school closures across four regions is displayed. As the left-most panel shows, there is substantial geographic variation in how many weeks of school children missed due to the pandemic. In Africa and the Arab countries, the enrollment weighted average is 29 weeks missed. In Latin America and the Caribbean, the average is 51 weeks missed. Again, there is substantial within-region variation. There are African and Arab countries that had more weeks of school closure than the average for Latin America and the Caribbean, and vice versa.
Splitting the pandemic into an early phase and a later phase, it is evident that different regions were affected differently across the two phases, as the two right-most panels in Figure 3 show. In the early phase, the average duration of school closures only varied by four weeks across the four regions, from 16 weeks in Africa and Arab States and in Europe and North America, to 20 weeks in Latin America and the Caribbean. The big difference across the regions emerged in the later phase: while African and Arab countries experienced on average 13 weeks of school closures in the later phase of the pandemic, countries in Latin America and the Caribbean had 31 weeks of school closures. In other words, in the first six months of the pandemic, children in Latin America and the Caribbean were 25% more affected than the children in Africa and Arab States in terms of school closures. In the following nine months, the difference was 138%.
Figure 2: Early vs late Covid-induced school closures by period
Source: UNESCO, World Bank. Note: Averages are weighted by primary school enrollment
To what extent did schools close?
Partial closure refers to schools being closed in some regions, that some year groups (or grades) have been sent home, or that there is reduced in-person schooling. While ‘full closure’ is likely an underestimate of how much children have been negatively affected in terms of school interruptions, it does provide a cleaner comparison across regions because the definition is narrower – i.e. all schools were closed at the nation-wide level due to Covid-19. Figure 3 below shows the distribution of school closures across regions by type of school closure.
The global enrollment average of 39 weeks of partial or full school closures are almost perfectly equally split across the two types of closures, with 19 weeks of partial closures and 20 weeks of full school closures. But this split between full and partial differs considerably across regions, as Figure below shows. In Africa and the Arab States, the full closures constitute 66% of all affected weeks, and in Latin America and the Caribbean the share of affected weeks where all schools were closed is 23%. In Europe and Northern America this share is only 19%.
The variation in full closures is striking and suggests that the totals conceal substantial variation in how affected children are across regions in terms of school interruptions. In Europe and Northern America, the enrollment weighted average of full school closures is ‘only’ seven weeks, which is less than 50% of second-least affected regions of Africa and Asia (at 19 and 21 weeks, respectively). The level in Latin America is almost four times higher than the level in Europe and Northern America, and no country in the latter region is above the average in Latin America.
Figure 3: Covid-induced partial and full school closures by type
Source: UNESCO, World Bank. Note: Averages are weighted by primary school enrollment
What factors can help predict school closures?
Given the variation in school closures, the next natural question to ask is whether it is possible to predict these differences based on country characteristics. The goal of this analysis is not to uncover any causal relationships. Instead, the goal is to uncover whether the burden in terms of interrupted schooling is related to a country’s resources. Given that studies have revealed substantial within-country differences in learning losses, it feasible that cross-country learning loss is even bigger. If the poorest regions – where children already complete lower levels of education – are worse affected by school interruptions, then school closures might increase cross-country inequality in the long-run.
In Figure 4 below, scatter plots of weeks of school closure against the logarithm of the number of Covid-19 cases per 100,000, the logarithm of the number of Covid-19 related deaths per 100,000, the logarithm of GDP per capita (PPP adjusted), the average years of schooling in the country, and the Human Development Index are each displayed. The first two columns suggest that the regions that were more affected by the pandemic had larger school interruptions. But it is worth repeating that these relationships are not necessarily causal. If school closures are an effective remedy to reduce the spread of the virus, we could observe a negative relationship between school closures and the severity of the pandemic.
Figure 4: School closures versus schooling level, HDI, Covid cases, Covid deaths and GDP
Source: UNESCO, World Bank, UNDP, WHO. Note: Fitted lines are weighted by primary school enrollment
The absence of a negative relationship does not tell us that school closures are not effective for at least two reasons. One reason is that school closures are likely to be implemented together with other initiatives to reduce the spread of the virus which are not controlled for. Another reason to refrain from a casual interpretation is reverse causality – where school closures were implemented in reaction to a large increase in cases, not to avoid an increase in the future. While there is some evidence that school re-openings had limited impact on the spread of the virus, these findings are likely to be very dependent on the context and other initiatives to reduce the spread of the virus (Isphording et al, 2020). We will therefore not argue whether it was a good or bad strategy to close the schools to reduce the impact of the pandemic.
Looking at the two lower rows of Figure 4 suggest that countries with a high GDP per capita, more schooling, and a higher human development index had more weeks of partial school closures, but fewer weeks of full school closures. These associations should be taken with a pinch of salt given the weak correlations, but they do suggest that countries with fewer resources are more affected by full school interruptions. These trends also suggest that countries that are already behind in terms of schooling (measured in average years of schooling) are more affected by full school closures.
Is the data reliable?
The data on school closures comes from UNESCO’s global monitoring of school closures. As school closures might be implemented at a sub-national level and vary across primary, secondary and tertiary education, it is difficult to both get a grasp of the big picture and capture all nuances. The distinction between full and partial is an attempt to take these nuances into account. While we have no reason to doubt the quality of the UNESCO data, we compared the values with data from other sources for a few countries and found identical or very similar levels of school closures.
What have we learned?
Children around the world have been affected by Covid-19 induced school closures. 65 weeks into the pandemic children have experienced on average 39 weeks of partial or full school closures. But the global average conceals substantial variation across countries and regions. While children in Europe and North America have experienced on average 37 weeks of partial or full school closures, children in Latin America and the Caribbean have experienced 51 weeks of closure.
The regional differences in school interruptions are even more striking when focusing on full closures. Children in Europe and North America have experienced an average of 7 weeks of nationwide school closures. This number is almost three times larger when looking at the global average, at 20 weeks, and more than five times higher in Latin America and the Caribbean.
Looking at what predicts the variation in school closures, there is some suggestive evidence that regions with lower GDP per capita, and where children complete less schooling, have been more affected by full school closures. Partial school closures on the other hand seem to be correlated with higher GDP per capita and more years of completed schooling. These correlations are worrying for at least two reasons. First, mitigating the harmful effects of school closures will require additional resources, and children living in regions with fewer resources to start with (as measured by GDP per capita) experienced longer periods of full closures. Second, relatively speaking, these differences across regions are even larger as longer periods of full school closures are observed in regions where children already complete fewer years of schooling. Recovering from learning losses will already require substantial public spending, and longer periods of interruptions mean that children in poorer countries appear to have a greater mountain to climb in terms of catching up.
Where can I find more?
- A number of studies have assessed the impact of school closures on the learning loss (see Patrinos and Donnelly (2021) for a review). But all these studies focus on Europe, North America, or Australia. A few studies have attempted to quantify the economics of a learning loss in the long run (Hanushek and Woessmann, 2020, Psacharopoulos et al, 2021).
- The Center for Global Development collects blog posts, data, and evidence on the impact of COVID-19 on school closures and education system. Earlier this year, the Center for Global Development published a blog post with a detailed analysis of the global impact and suggestions on how to react (Evans et al, 2021).
Who are economic experts on this question?
- Simon Burgess, University of Bristol
- Anna Vignoles, University of Cambridge
- Lindsey Macmillan, UCL and IFS
- Stephen Machin, LSE
- Jo Blanden, University of Surrey
- Hans H. Sievertsen, University of Bristol and The Danish Center for Social Science Research
