The youth mental health crisis: Quasi-experimental evidence on the role of school closures

Abstract

During the COVID-19 pandemic, the youth mental health crisis has reached unprecedented levels. To which extent school closures, one of the most heavily debated pandemic measures, have contributed to or even caused this crisis is largely unknown. We seek to narrow this blind spot, by combining quasi-experimental variation in school closure and reopening strategies across the German federal states at the onset of the pandemic with nationwide, population-based survey data on youth mental health and high-frequency data from the largest crisis helpline. We show that prolonged school closures led to a substantial deterioration in youth health-related quality of life, precipitating early signs of mental health problems. The effects were most severe among boys, younger adolescents, and families with limited living space. We further provide evidence that family problems are a major issue that adolescents were struggling with when denied access to school. Overall, school closures largely explain the deterioration of youth mental health over the first pandemic wave.

Fig. 1.. Variation in school closure and reopening strategies.

Data are compiled on the basis of a comprehensive screening of the state-specific corona protection ordinances (22). Each blue dot represents the date when the respective federal government (partially) reopened schools for selected grade levels, possibly restricted to the grade levels of certain school tracks (A) and the resulting variation in the reopening dates by grade levels across the federal states and possibly school tracks (B). As shown in fig. S1, substantial variation remains when displaying the variation within the different school tracks separately.

Fig. 2.. Results.

(A) The short-run impact of school closure on HRQoL and several screening devices for mental health problems (including the 99, 95, and 90% confidence interval) for the effect of one additional week of school closure on standardized measures (mean, 0; SD, 1) of adolescents’ self-reported HRQoL (KIDSCREEN-10), and screening instruments for mental health problems ( HBSC-SCL, SDQ, CES-DC, and SCARED). For all scales, except KIDSCREEN-10, higher values express more health problems. For exact values, see table S7A. Each bar results from a separate two-way fixed-effects regression (controlling for a set of state and school track–specific grade level fixed effects as well as for age and sex) using ordinary least squares with sampling weights and SEs clustered at the state*grade level*school track (see Eq. 1). (B) The decline in HRQoL (higher values express less health problems) and depressive symptoms (CES-DC, higher values express more health problems) (including the 95 and 90% confidence interval) between 2017 and 2020. Each bar represents a separate estimation, with the first estimation in each panel not accounting for school closure and the second estimation including it. All estimations control for age and sex and use ordinary least squares with sampling weights and SEs that are clustered at the state*school track *grade level (as detailed in Eqs. 2 and 3). For detailed regression results, see table S7 (B and C).

Fig. 3.. Results by sociodemographic characteristics.

The short-run impact of school closure on HRQoL and screening devices for mental health (including the 95 and 90% confidence interval) for the effect of one additional week of school closure on standardized measures (mean, 0; SD, 1) of adolescents’ self-reported HRQoL (KIDSCREEN-10), and screening instruments for mental health problems (HBSC-SCL, SDQ, CES-DC, and SCARED) separately by age (A), sex (B), and living space per child (C). For all scales, except KIDSCREEN-10, higher values express more health problems. For more details see Eqs. 5 to 7 and table S10 for the detailed regression results.

Fig. 4.. Changes in crisis helpline call volume in 2020.

(A) Twenty-eight–day moving averages of call volumes. Call volume is the residuals of a regression in which the number of calls times their average duration is regressed on a set of month and weekend dummies. Residuals are standardized to have mean of 0 and SD of 1 for 2019. The shaded area indicates the 95% confidence interval. (B) The same moving average as in (A) but adolescents returning to school at or before 18 May and adolescents returning after 18 May, respectively. The vertical lines correspond to the date of nationwide school closure (16 to 18 March 2020), and the respective dates of school reopening for the early returners (from 20 April 2020 onward) and late returners (from 18 May 2020 until 15 June 2020). The gray area indicates the time when COPSY was conducted. For detailed regression results, see the table S12.