IFPRI Discussion Paper 02075 December 2021 COVID-19 School Closures and Mental Health of Adolescent Students Evidence from Rural Mozambique Feliciano Chimbutane Catalina Herrera-Almanza Naureen Karachiwalla Carlos Lauchande Jessica Leight Poverty, Health, and Nutrition Division INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE The International Food Policy Research Institute (IFPRI), a CGIAR Research Center established in 1975, provides research-based policy solutions to sustainably reduce poverty and end hunger and malnutrition. IFPRI’s strategic research aims to foster a climate-resilient and sustainable food supply; promote healthy diets and nutrition for all; build inclusive and efficient markets, trade systems, and food industries; transform agricultural and rural economies; and strengthen institutions and governance. Gender is integrated in all the Institute’s work. Partnerships, communications, capacity strengthening, and data and knowledge management are essential components to translate IFPRI’s research from action to impact. The Institute’s regional and country programs play a critical role in responding to demand for food policy research and in delivering holistic support for country-led development. IFPRI collaborates with partners around the world. AUTHORS Feliciano Chimbutane (felicianosal@yahoo.com.au) is an Associate Professor of Linguistics (Languages and Education) at the Universidade Eduardo Mondlane, Maputo, Mozambique. Catalina Herrera-Almanza (cataher@illinois.edu) is an Assistant Professor in the Department of Agricultural and Consumer Economics at the University of Illinois, Urbana-Champaign, Champaign, IL. Naureen Karachiwalla (n.karachiwalla@cgiar.org) is a Research Fellow in the Poverty, Health, and Nutrition Division at the International Food Policy Research Institute (IFPRI), Washington, DC. Carlos Lauchande (lauchand59@gmail.com) is an Assistant Professor of Statistics at the Universidade Pedagógica de Maputo, Maputo, Mozambique. Jessica Leight (j.leight@cgiar.org) is a Research Fellow in IFPRI’s Poverty, Health, and Nutrition Division, Washington, DC. Notices 1IFPRI Discussion Papers contain preliminary material and research results and are circulated in order to stimulate discussion and critical comment. They have not been subject to a formal external review via IFPRI’s Publications Review Committee. Any opinions stated herein are those of the author(s) and are not necessarily representative of or endorsed by IFPRI. 2 The boundaries and names shown and the designations used on the map(s) herein do not imply official endorsement or acceptance by the International Food Policy Research Institute (IFPRI) or its partners and contributors. 3Copyright remains with the authors. The authors are free to proceed, without further IFPRI permission, to publish this paper, or any revised version of it, in outlets such as journals, books, and other publications mailto:felicianosal@yahoo.com.au mailto:cataher@illinois.edu mailto:n.karachiwalla@cgiar.org mailto:lauchand59@gmail.com mailto:j.leight@cgiar.org iii Abstract The onset of the COVID-19 pandemic, entailing widespread school closures as well as acute disruptions to household livelihoods, has presumably had substantial consequences for adolescent well-being in developing country contexts that remain largely unexplored. We present novel evidence about the prevalence of mental health challenges among adolescent students as well as educators in rural Mozambique using data from an in-person survey conducted in 175 schools. In our sample, 31% of students report low levels of well-being (though only 10% suffer from high anxiety): students enrolled in schools that used a wider variety of distance learning measures report lower anxiety, while students reporting familial shocks linked to the pandemic report higher anxiety and lower well-being. Educators experience comparatively lower levels of anxiety and higher well-being, and household-level shocks are most predictive of variation in mental health. However, well-being is negatively affected by the range of hygiene-related measures implemented in schools upon reopening. Keywords: mental health, COVID-19, school closures iv Acknowledgments We thank Daniel Maggio and Odiche Nwabuikwu for exceptional research assistance. We also thank the field team at ELIM — particularly Rosa Matine, Prince Dziva, Tatiana Mata, Ercília Mata Ubisse, and Daniel Nanjelo — for their dedication, as well as the respondents who kindly volunteered their time for the study. We gratefully acknowledge funding from the Policies, Institutions, and Markets (PIM) of the CGIAR, from World Vision (grant no. FFE-656-2019/018-00-IFPRI), and from the COVID-19 Africa Rapid Grant Fund of the National Research Foundation (grant no. COV19200616532519). 1 Introduction Since the onset of the COVID-19 pandemic in March 2020, schools around the world have experienced prolonged periods of closures that have disrupted the educational progression of hundreds of millions of students. According to UNESCO, school closures peaked around June 2020, at which point 50% of all students worldwide were out of school (UNESCO, 2021). Particularly in developing countries where schools are a crucial path out of poverty (and the technology for learning at a distance is almost entirely absent), the disruption in schooling over a prolonged period of time has meaningful implications for children’s learning, health, and emotional well-being (Engzell et al., 2021). Moreover, these closures coincided with a period in which children were potentially exposed to widespread household-level shocks linked to illness, disrupted livelihoods, and challenges in access to food and other subsistence goods (Egger et al., 2021). Adolescent students are a population that is uniquely vulnerable to the adverse effects of the pandemic for a number of reasons. Adolescents may be particularly likely to substitute away from school into work, marriage, or parenthood (especially for girls) during school clo- sures. They are also proximate to the transitions to secondary or tertiary school, rendering any disruptions to their education potentially extremely costly. In addition, evidence from previous health crises suggests that out-of-school adolescent girls may be particularly vul- nerable to the early initiation of sexual activity, or physical and sexual violence (Bandiera et al., 2019). Adolescence is also broadly a period of increased risk for mental health disorders (Allen and Sheeber, 2008). The Organization (2014) estimates that half of adults with mental health challenges had symptoms that began in adolescence, and these challenges during adolescence are correlated with a range of adverse outcomes, including lower academic achievement and non-cognitive development (Cook et al., 2009). Accordingly, understanding the effects of COVID-19 in relation to mental health is essential. Given that schools play the second largest role in children’s lives after their families, school policies can be highly influential in shaping students’ well-being. In this paper, we present novel evidence about the prevalence of mental health challenges among adolescent students as well as educators in rural Mozambique, and explore whether variation in pandemic-related schooling policies and exposure to pandemic-linked shocks is predictive of variation in these challenges. We use data from a large-scale in-person survey conducted in 175 schools between June and August 2021. Schools in Mozambique closed after a national emergency linked to the COVID-19 pandemic was declared on March 30, 2020, 1 and re-opened on a phased basis between February and March 2021. Our survey includes data from 1,383 grade seven students reporting on their households’ experiences during the school closures, their access to social support, and their mental health as measured by the WHO-5 survey of index of subjective well-being (Topp et al., 2015) and the GAD-7 index of generalized anxiety disorder (Spitzer et al., 2006). We also collected complementary data from 350 educators (the deputy school director, and one teacher per school) in the same set of 175 schools. We find that 31% of students report low levels of subjective well-being according to the WHO-5 measure, suggestive of potential vulnerability to depression, though only 10% show evidence of high levels of anxiety. We interpret this difference in prevalence rates as at least partially reflecting the fact that the WHO-5 is a broader measure of psychological well-being, though it is correlated with depression and has been validated as a measure of vulnerability to depression in several countries in sub-Saharan Africa (Garland et al., 2018; Nolan et al., 2018; Chongwo et al., 2018), including among adolescents.1 Moreover, students whose schools undertook measures to provide some remote instruction during the pandemic (such as contacting parents or engaging with students in small groups) exhibited lower levels of anxiety, suggesting protective effects of ongoing social support and opportunities, even if limited, for students to continue their educational progression. For example, in schools that reported engaging with students at home or in small groups during school closures, the probability of anxiety is 8.7 percentage points lower, relative to a mean probability of 10%.2 Self-reported experiences of negative shocks linked to the pandemic are also highly predictive of mental health outcomes. In particular, students whose families went hungry and report that a female member of their family married earlier than expected because of pandemic-related causes are much more likely to be anxious and vulnerable to depression. Finally, there is also some evidence of higher anxiety and lower well-being among students characterized by higher socioeconomic status, consistent with evidence from other contexts in sub-Saharan Africa that it is primarily non-agricultural households that have experienced a greater loss of income (Mahmud and Riley, 2021; Aggarwal et al., 2020), and 1Data collected during school closures in 2020 uncovered similar rates of mental ill-health. One study found that the prevalence of at least mild anxiety among adolescents was 15%, 10%, 40%, and 10% in Ethiopia, India, Peru, and Vietnam respectively, and was 15%, 9%, 32%, and 11% respectively for at least mild depression (Favara et al., 2021). Additionally, other studies reporting both the WHO-5 and the GAD-7 in the region have found similar patterns: a recent analysis of mental health during the pandemic among Ghanaian adults also found a higher prevalence of low well-being vis-a-vis significant anxiety (Boateng et al., 2021). 2Virtually all - 98.5% – of schools reported that they provided instructional materials for students to take home, rendering it impossible to analyze variation along this dimension. 2 more adverse outcomes for girls, consistent with broader evidence that girls and women have been particularly vulnerable during the pandemic (de Paz et al., 2020). The data collected from educators shows lower levels of mental health challenges, and in the adult sample, we find little evidence that school- or education-related variables are predictive of variation in mental health. There is some evidence that COVID-19 mitigation measures decreased subjective well-being, possibly because of the added work load; however, the primary predictive variables are reported household-level shocks related to the pandemic. These findings suggest that schools that are targeting learning catch-up of returning students in the developing world may also face considerable challenges in addressing the psychosocial and emotional challenges their students are experiencing. Additionally, teach- ers’ well-being affects the well-being of both students and other teachers (Becker et al., 2014). It is important to note that we did not survey students who failed to return to school, who may be experiencing a very different set of challenges, and for the same reason we cannot assess the magnitude of pandemic-related dropout. Clearly, however, the level of disruption in schooling experienced by this adolescent population has had very serious implications for their well-being. Our paper joins a limited literature that has analyzed the effects of the pandemic on mental health in sub-Saharan Africa, particularly for adolescents. In the literature analyzing adults, Langsi et al. (2021) uses a web-based survey from seven sub-Saharan African countries to analyze the associations of demographic characteristics with COVID-related attitudes, and Chen et al. (2021) reports a meta-analysis of the prevalence of depression among adults in north and sub-Saharan Africa. However, none of the papers reported in this meta-analysis are in Southern Africa, where the only available evidence is from South Africa. Visser and van Wyk (2021) use web-based data from undergraduate students in South Africa to analyze the predictors of emotional well-being; Kim et al. (2020) uses a small (N = 250) in-person sample to analyze multivariate models predicting mental health among adults. For adolescents, Wang et al. (2021) reports cross-country data on adolescents’ experience of the pandemic in Nigeria, Ethiopia, and Burkina Faso using phone surveys, but does not analyze mental health. A recent paper drawing on a phone survey conducted in Senegal with a sample of girls and women aged 14 to 35 finds heterogeneous shifts in well-being vis-a-vis the pre-pandemic period (Dione et al., 2021). Saurabh and Ranjan (2021) analyzes the effect of quarantines on the mental health of adolescents in India, again using a relatively small sample.3 Three analyses of child and adolescent mental health were conducted in China, a 3Ochnik et al. (2021) also analyzes mental health prevalence and predictors among a cross-sectional 3 site of severe and persistent lockdowns (Duan et al., 2020; Zhou et al., 2020; Qi et al., 2020), but other sources cited in an editorial highlighting the importance of focusing on child mental health were all drawn from rich countries (Bhatia, 2020). In one very recent contribution, Favara et al. (2021) conducted a phone survey in Ethiopia, India, Vietnam, and Peru with a cohort of youth aged 19–20 and measured rates of depression and anxiety during the period of school closures. Our paper complements this work by providing evidence on well-being once schools reopened as well as evidence on educators. There is growing data on the effects of school closures on other dimensions of adolescent well-being such as learning loss, though some estimates suggest learning loss of up to one year (Angrist et al., 2021; Azevedo et al., 2021), with adverse implications for equity (Hevia et al., 2021; Sabates et al., 2021). These learning losses are also projected to have substantial economic consequences (Psacharopoulos et al., 2020; UNESCO et al., December, 2021). Our paper contributes to this literature by examining the effects of specific measures that schools implemented during school closures on the mental health of adolescent students and educators. While limited data has been collected on adolescent outcomes during school closures, this is the first study to explore the relationship between household- and school-level variation in pandemic experiences and adolescent well-being. Our study is also one of the first large-scale, in-person surveys of mental health among adolescent students (and educators) in sub-Saharan Africa. Much of the existing literature uses data collected digitally or by phone, while we use in-person data that may be more reliable especially for the measurement of sensitive attributes such as mental health. The remainder of the paper proceeds as follows. Section 2 describes the setting and the empirical strategy. Section 3 presents the results, and Section 4 concludes. 2 Empirical Analysis 2.1 Setting Mozambique has one of the lowest educational attainment rates in the world; the primary school completion rate is below 40% (Mambo et al., 2019), and less than one third of students progress to secondary school (UNESCO Institute of Statistics, 2021). In Nampula province, the site of this study, the dropout rate from primary school was 11% prior to the pandemic (UNICEF, forthcoming). Our target cohort for this analysis is students in grade seven sample of university students, but this is primarily drawn from rich countries. 4 (average age of 15), the final year of primary school; students who drop out prior to or during grade seven are not eligible to continue their education due to their failure to complete the primary school leaving exam. In Mozambique, schools closed in March 2020 and were closed for a year, during which period Nampula province emerged as one of the national centers of the pandemic (allAfrica, 2020). During the period of school closures, the Ministry of Education and Human Develop- ment (MINEDH) promoted distance learning, emphasizing the development and provision of learning materials to students and the continuation of instruction using digital platforms including radio and television (MINEDH, 2020a,b). Almost all schools in our sample (98%) provided students with take home materials, and many schools also reported contacting par- ents by phone and visiting students or conducting small group sessions. School directors, school councils, and community leaders were engaged in mobilizing parents to support their children at home; however, given the suddenness of the pandemic, stakeholders were largely unprepared and reports suggest that the challenges were significant, particularly for educa- tors (e.g. Ferrão et al., 2020; Reimers and Shleicher, 2020). Teachers were not trained in distance education or online platforms, and in rural areas, they generally lack access to the requisite equipment. Data around mental health in Mozambique is sparse, particularly for adolescents, but the available data suggests a meaningful prevalence of mental health disorders even prior to the pandemic. According to recent data, 4.1% of the population suffered from depression in 2021 (World Population Review, 2021), compared to a global average of 4.4% and 3.4% in 2017 across countries (World Health Organization, 2017). According to Cumbe et al. (2020), the prevalence of severe mental illness (including depression) in Mozambique is 5%, while the prevalence of anxiety disorders is 3% (World Health Organization, 2017). For adolescents specifically, Amu et al. (2020) measures “psychological distress” (includ- ing depression and anxiety) and finds that 21% of adolescents experienced distress, with lower rates among boys and among respondents reporting more social support. The same analysis also highlights that these challenges may pose an obstacle for Mozambique to achieve the Sustainable Development Goal of promoting mental health and well-being of all by the year 2030. Given the available data suggesting a non-trivial prevalence of mental health challenges in this population and the possibility that these challenges have been further ex- acerbated by pandemic-related shocks, examining the relationship between these shocks and adolescent student well-being is potentially very informative. As of December 17, 2021, Mozambique has reported around 150,000 confirmed cases of 5 COVID-19 infection and 1,700 deaths (World Health Organization, 2021); however, data on local COVID-19 incidence is only available at the province level. Accordingly, we are not able to identify local prevalence of cases in the study communities. 2.2 Data The survey analyzed in this paper was conducted between June and August 2021 in 175 schools in Nacaroa, Muecete and Murrupula districts in Nampula province, Mozambique. Within this sample, 142 schools included data collection with grade seven students. The remaining 33 schools served students only through grade five; in these schools, only the educator surveys were conducted.4 The surveys were administered by a local survey firm (ELIM Servicios) in Portuguese; enumerators received a week-long training on survey ethics, survey content, and appropriate COVID-19 protocols prior to survey launch. Interviews were conducted with 10 randomly selected grade seven students, the deputy school director, and one teacher in the school (the grade four teacher).5 Consequently, 175 teachers and deputy directors were interviewed, alongside 1,383 students. The surveys of grade seven students collected data on demographic characteristics (age, gender, parental literacy, household size, assets), experiences with COVID-19 (household hunger, income loss, and early marriage), and mental health (subjective well-being and anxiety).6 The surveys with teachers also included questions about their experiences with COVID-19 as well as demographic characteristics and mental health. Deputy school directors also reported school-specific information about facilities and COVID-19 measures that the school implemented during and after the school closures. Note that deputy directors were not asked about their personal experiences with COVID-19. 2.3 Estimating Equation We estimate the determinants of an individual’s mental health outcomes using the following specification: Yisd = α + β1Ssd + β2Cisd + β1Xisd + δd + εisd (1) 4A separate survey targeting grade four students was also conducted in all schoosl. 5The grade four teacher was interviewed because these data were collected as part of a larger study focused on grade four students. 6While we also collected some preliminary data on whether students reported cases of illness in their households, we found that it was challenging for them to differentiate between illness caused by different factors including but not limited to COVID-19; accordingly, we have not analyzed the data. 6 where Yisd is the mental health outcome for individual i (student or educator) in school s located in district d. We use two mental health measures: the GAD-7, which measures anxiety, and the WHO-5, which measures subjective well-being and is used as a predictor for depression. For both measures, we report the continuous raw score and indicators for having moderate to severe anxiety and low well-being, respectively.7 The vector Ssd includes the measures that a school implemented to address the pandemic- related school closures. We include indicator variables for whether a school i) encouraged students to listen to radio education programming, ii) telephoned parents to encourage them to help their children to learn at home, and iii) employed other measures, such as encourag- ing students to watch TV education programming, visiting students at home, and gathering small groups of students for lessons.8 We also employ a standardized COVID-19 mitiga- tion index capturing hygiene and health-related measures implemented at schools after re- opening, such as using different schedules across grades, seating students further apart, and encouraging students, and staff to wash their hands frequently, among others.9 Ssd also includes school-level covariates that capture school quality in two ways. The first is the av- erage grade four reading test score,10 and the second is a standardized school quality index that captures school-level infrastructure, such as the availability of latrines, electricity, and reading materials for students.11 7See the Appendix for the questions and variable definitions. For the WHO-5 index, Cronbach’s α is 0.79 for students and 0.69 for educators; for the GAD-7 index, Cronbach’s α is 0.69 for students, above or equal to the recommended cutoff of 0.7 indicative of internal reliability. Note that at the beginning of the survey period, we administered the PHQ-9 depression questionnaire to students and educators. However, these questions were viewed as disruptive in the school environment, and accordingly the WHO-5 instrument was subsequently employed; data for the WHO-5 is thus missing for part of the sample (315 students, 38 teachers and 47 school directors) who were surveyed in the first schools visited. 8The distribution of take-home materials was near-universal as previously noted, and thus we do not analyze it. 9This mitigation index was constructed using principal component analysis on the following dummy variables indicating whether a school: i) only allowed grade seven students when schools first reopened, ii) only allowed grade 3 students at first, iii) spaced students further apart, iv) ensured students and staff washed their hands frequently, v) required staff to wear masks or other personal protective equipment (PPE), vi) held classes outside when possible, vii) continued to keep libraries and other common spaces closed, viii) provided information to students and parents, ix) opened some classes before others, and x) had different students attend on different days, which was then standardized to have mean zero and standard deviation one. 10We construct this variable as the school-level average of students’ number of correct answers in the oral reading fluency section of the Early Grade Reading Assessment (EGRA) test that was administered as part of the larger study. 11We create this quality index using principal component analysis on the following dummy variables, whether a school has i) latrines, ii) latrines that are gender separated, iii) electricity, iv) Portuguese reading materials for grade four students, v) local language reading material for grade four students, vi) soap/ detergent for staff, vii) soap/detergent for , viii) hand-washing station or tippy tap, and ix) 7 Cisd is a vector of variables reporting household-level experiences of COVID-19 shocks. In the student sample, this vector includes a set of dummy variables for whether students report that at least one member of their household i) went hungry, ii) lost employment, iii) migrated, and iv) married earlier than expected (for female household members) due to the pandemic. We also include an indicator equal to one if the student reported having any social support during the school closures: if s/he shared his/her fears, frustrations, or challenges with peers from school, family members, neighbours, or others. For teachers, Ci includes a set of variables for whether the teacher reports the following consequences due to the pandemic: i) they lost income, ii) a family member lost employment, and iii) students were more difficult to manage since schools reopened. Note that we did not ask deputy school directors about the effects of the pandemic on their households. Xisd is a vector of demographic and household controls. For students, we include gender, age, a standardized household-level asset index used as a proxy for wealth, and household size.12 We also include a set of dummy variables for whether a household reports i) farm income, ii) a non-farm business, iii) completion of primary school by the father, and iv) completion of primary school by the mother. For educators, Xi includes gender, age, the number of years working at the current school, an indicator for whether the educator has completed post-secondary education and an indicator for whether s/he speaks Portuguese at home (reported for teachers only). All models control for district fixed effects (δd), and standard errors are clustered at the school level. 3 Results In this section, we report our findings including sample summary statistics, the overall prevalence of mental health challenges, and evidence around the correlates of mental health. 3.1 Characteristics of the Sample Demographics We begin by briefly describing our sample (see Table A1). Our cohort of grade seven students is 40% male and 15 years old, on average. Almost all households are engaged in farming, and 82% of adolescents worked on the family farm in the past week. Households are relatively large with approximately six household members on average, and whether the school has an appropriate water source, and then standardize the index. 12We calculate the asset index using a principal component analysis using the following variables: house- hold owns i) radio, ii) cattle, iii) mobile phone, and iv) bicycle. The index is also standardized. 8 education levels are low; only 58% of fathers and 40% of mothers have completed primary education. Turning to educators, almost all teachers are permanent government employees and report a mean of eight years of teaching experience; only 6% have any post-secondary education. A quarter are male, and they are 32 years old on average. Within the sample of deputy directors, 88% are male and they are 39 years old on average; they have approximately 14 years of teaching experience, and 23% have a post-secondary school education. In Panel D, we can observe that schools are large (on average, nearly 1000 students), and between 11% and 30% implemented additional distance learning measures during the COVID-related shutdowns. COVID-19 experiences Table 1 presents summary statistics regarding pandemic-related shocks experienced by individuals in our sample, including adolescent students (Panel A) and teachers (Panel B). For students, 53% report experiences of hunger in their household, and 27% report that a female member of the household married early. Other reported pandemic-related shocks include migration (prevalence of 13%) and loss of employment of a household member (8%). These experiences reflect severe shocks that would certainly influence a student’s mental health. Teachers also experienced significant shocks during the period of school closures: 31% report having been affected by the pandemic in some way, and of those, one-quarter report that a family member lost their job or that some of their income was lost. Additionally, 12% of teachers report that it has been more challenging for them to manage students since schools re-opened. 3.2 Prevalence of Mental Health Challenges Next, we report the prevalence of mental health challenges among students and educators in Table 2. Among adolescent students, the mean score of the WHO-5 well-being scale was 62. According to the cutoff for low levels of well-being (corresponding to a score below 50, out of a maximum of 100), 31% were classified as experiencing a low level of well-being and thus identified as vulnerable to depression. Turning to anxiety as measured by the GAD-7, the average score is five (out of a maximum of 21). According to the GAD-7 cutoff for “moderate to severe anxiety” (a score more than 10), 10% of adolescents are classified as experiencing general anxiety disorder.13 These rates are largely in line with other work conducted during 13Note that the number of observations differs across the two measures. Initially, the survey administered the PHQ-9 depression scale questionnaire. However, students found it difficult to respond to these questions so the simpler WHO-5 scale was used for subsequent respondents. The number of observations using the PHQ-9 is too small to produce meaningful estimates. 9 school closures in 2020. The prevalence of at least mild anxiety among adolescents was 15%, 10%, 40%, and 10% in Ethiopia, India, Peru, and Vietnam respectively, and was 15%, 9%, 32%, and 11% respectively for at least mild depression (Favara et al., 2021). For educators, the prevalence of mental health challenges is slightly lower: the average score on the WHO-5 scale is 72, and 12% were classified as experiencing low levels of well- being. The average score on the GAD-7 was four for educators, and 9% were classified as experiencing moderate to severe anxiety. For both students and educators, there appears to be a gap between the rates of low levels of well-being and the rates of moderate to severe anxiety. The reason for this ostensible discrepancy is that well-being encompasses more than simply depression or other mental health outcomes; it represents a person’s subjective assessment of their overall well-being in many domains. The GAD-7 measures a specific aspect of mental health, generalized anxiety disorder. While the WHO-5 does predict susceptibility to depression, it does not measure actual levels of depression (Krieger et al., 2014). 3.3 Predictors of Adolescent Students’ Mental Health We now report the results of estimating Equation (1) for the predictors of mental health for adolescent students. Table 3 analyzes anxiety in columns 1 and 2, and subjective well-being in columns 3 and 4. Columns 1 and 3 report results for binary variables (experiencing moderate to severe anxiety and having a low level of well-being) and columns 2 and 4 report results for continuous measures of the GAD-7 and WHO-5 scales, respectively. To be consistent with the anxiety measure, we have inverted the WHO-5 subjective well-being scale so that a higher number denotes lower well-being for both scales. Our findings suggest that school policies during the period of school closures are predictive of anxiety levels: students attending schools in which parents were contacted, students were visited at home, or where lessons were held with small groups of students displayed lower levels of anxiety, suggesting that this support may have mitigated some negative effects of the school closures. In schools in which either of these measures were implemented, there was a decline in the GAD-7 score of between 25% and 40% relative to the mean, as well as a decline in the probability of moderate to severe anxiety of nine percentage points for measures other than parental contacts. However, COVID-19 mitigation measures implemented following school re-opening are not associated with anxiety. Interestingly, adolescents in schools with higher test scores among grade four students experienced higher levels of anxiety. Given that student test scores are a good proxy for school quality and thus higher levels of retention 10 and progression to secondary school, it is possible that grade seven students in these schools were more anxious about disrupted learning because of its impact on their own future test scores and prospects for secondary school. The school infrastructure index is not statistically significant. Unsurprisingly, students whose families experienced large negative shocks during school closures also showed higher levels of anxiety: students who reported that a family member went hungry showed a GAD-7 score 18% higher, and students who reported that a female member of the household married early showed a GAD-7 score that is 21% higher and an eight percentage point increase in the probability of moderate to severe anxiety. By contrast, reported access to social support (a classmate, family member, neighbor, or other person to talk with) is associated with lower levels of anxiety. This result, in conjunction with the finding that schooling measures taken during closures were associated with less anxiety, are consistent with the literature showing that loneliness among children is one of the most distressing conditions of prolonged school closures (Loades et al., 2020). For demographic variables, we observe positive correlations between the probability of a non- agricultural business and a household asset index and levels of anxiety: students from these households are about four percentage points more likely to report moderate or severe anxiety. This would be consistent with the hypothesis that these households are experiencing larger shocks to their pre-pandemic income. There is also some evidence of higher GAD-7 scores among female students (a difference of 11%). A growing body of evidence shows high rates of anxiety and depression for children and youth because of COVID-19, with some studies finding that girls were disproportionately affected (UNESCO et al., December, 2021; Amu et al., 2020). Patterns in predictors of subjective well-being (also interpreted as vulnerability to depres- sion) are broadly similar as reported in Columns 3 and 4 of the same table, but with some notable differences. School policies are not significant predictors of subjective well-being. However, as with anxiety, students who report that their family went hungry during the pandemic are more likely to be classified as having a low level of well-being (by 9 percentage points) and have a continuous score that is 14% higher. Students who report early marriage of a female household member have a well-being score that is 11% lower. Once again, social support was positively associated with well-being, leading to a 16 percentage point decline in the probability of low well-being and a score that is 13% higher. For demographic variables, again we observe a substantial gender difference: male stu- dents are six percentage points less likely to be characterized by low well-being and have 11 scores around 4% higher. In line with other studies, this finding suggests that female stu- dents may have bore the brunt of the detrimental effects of the school closures, given a higher probability of domestic responsibilities at home (Mwabe et al., 2021) and girls’ greater vul- nerability to violence and risky behaviours during school closures (Bandiera et al., 2019); this would also be consistent with broader evidence that girls and women have been particu- larly vulnerable to adverse effects of the pandemic (de Paz et al., 2020). Other demographic variables do not significantly predict subjective well-being. 3.4 Predictors of Educator Mental Health Turning to educators, Table 4 displays parallel results for anxiety. The outcome variable in columns 1 and 2 is an indicator for moderate or severe anxiety, and the outcome in columns 3 and 4 is the continuous GAD-7 score. Columns 1 and 3 only report results for teachers as data on pandemic-related shocks is reported only for teachers, and columns 2 and 4 report results for the pooled sample of teachers and deputy school directors. Table 4 shows that only one of the variables included in the model is statistically signif- icant in Columns 1 and 2: teachers and deputy school directors who have a post-secondary degree or higher are less likely to be classified as having high levels of anxiety, possibly in- dicating a significant protective effect of education on emotional resilience or the ability to leverage other sources of income. More variables predict the degree of anxiety, as measured by the continuous GAD-7 score. Educators who speak Portuguese at home have GAD-7 scores that are almost 50% higher; we conjecture that speaking Portuguese at home is a proxy for higher wealth levels and consequently, these households have more to lose from a disruption of income. A family member losing employment due to COVID-19 is also strongly associated with the anxiety score, leading to an increase of 2.9 points, or 37%. Teaching experience is also positively associated with anxiety. More experienced educators may have more responsibilities at the school and consequently may have been more affected by the closures and subsequent responsibilities once schools reopened. No school-related measures predicted educators’ anxiety levels. Turning to well-being and vulnerability to depression, Table 5 displays results for a binary variable indicating low well-being in the first two columns and the continuous measure in the next two columns. Again, we include only teachers in columns 1 and 3 and include both teachers and deputy school directors in columns 2 and 4. Note again that a higher level of the inverted WHO-5 variable indicates worse well-being. We again observe that loss of income in the teacher’s household is strongly associated with 12 low well-being and vulnerability to depression (teachers are twice as likely to be classified as having low well-being) as well as lower well-being scores (38% lower). In addition, a wider range of implementing measures to reopen schools safely is associated with lower levels of well-being for both teachers and deputy school directors (a decline of around 6 points, or less than 10%). These activities may have increased the burden of work at school for educators and possibly even affected the ability to earn income from other sources. Overall, these results suggest that personal, rather than school-related shocks, are more predictive of the mental health of educators compared to that of students. 4 Conclusion This paper has provided new evidence about the associations between student and educator mental health and experiences during school closures in rural Mozambique. Our findings sug- gest a meaningful prevalence of low well-being among both students and educators, though a relatively lower prevalence of anxiety, and significant exposure to pandemic-related shocks; these shocks are then substantially predictive of adverse mental health outcomes. Among adolescents, there is evidence that the implementation of a wider range of distance learn- ing methods by schools had a protective effect, reducing the probability of high anxiety. These methods included contacting parents and meeting with students in small groups. In contrast, educators in schools that implemented COVID-19 safety measures once schools reopened experienced lower levels of well-being, possibly due to the additional workload. Relative to the existing literature, our findings build on a growing body of evidence suggesting that the COVID-19 pandemic and related shocks have had particularly acute consequences for adolescent well-being in developing countries. Though our data around mental health is novel in the region, Wang et al. (2021) find that adolescents in Nigeria, Burkina Faso, and Ethiopia report declines in food consumption as well as a total absence of access to education, in conjunction with a self-assessed decline in the ability to learn. These patterns are consistent with the evidence presented here that pandemic-related shocks are highly associated with anxiety and low well-being, while the implementation of distance learning instruction reduces student anxiety, presumably by increasing student learning or student confidence in learning. Relatedly, data from a sample of university students in South Africa — who are of course somewhat older — suggests that social isolation contributes sub- stantially to emotional challenges, and these challenges are more common among girls (Visser and van Wyk, 2021). We also observe higher rates of anxiety and lower well-being among 13 female students in this sample, while social support (reducing isolation) has a protective effect. It should be noted, however, that there is also evidence of substantial resilience in the adolescent population. A recent systematic review of estimates of mental health challenges among diverse populations in Africa during the pandemic estimated a pooled prevalence of 31% for depression and 30% for anxiety (Chen et al., 2021). While these rates may be high given the inclusion of studies focusing on populations that are particularly affected by the pandemic (e.g., health care workers), it does suggest that adolescents in our Mozambican sample may be experiencing a lower rate of mental health challenges compared to adults more broadly in the region. Again, our findings are consistent with prevalence rates of adolescents in other countries, including in Africa (Favara et al., 2021) Stepping back, it is clear that meaningful challenges lie ahead for schools in the develop- ing world in managing a range of difficulties faced by themselves and their students following school reopening. Accordingly, it is important that policymakers attempt to direct some re- sources towards supporting teachers and school management in addressing challenges linked to depression and anxiety appropriately. Moreover, teachers’ emotions and stress have been found to influence those of students and other teachers (Becker et al., 2014). This gap has been noted by international bodies (UNESCO et al., December, 2021), and we reinforce this view in suggesting that mental health awareness and simple resources could enhance the well-being, and catch-up of students and educators and is worth considering. 14 References Aggarwal, Shilpa, Dahyeon Jeong, Naresh Kumar, David S. Park, Jonathan Robinson, and Alan Spearot, “Did COVID-19 Market Disruptions Disrupt Food Se- curity? Evidence from Households in Rural Liberia and Malawi,” 2020. Center for Effective Global Action Working Paper 142. allAfrica, “Mozambique: Sharp Increase in Coronavirus Cases in Nampula,” 2020. https://allafrica.com/stories/202006030604.html, Retrieved December 1, 2021. Allen, Nicholas B and Lisa B Sheeber, “The Importance of Affective Development for the Emergence of Depressive Disorders during Adolescence,” in Nicholas B. Allen and Lisa B. Sheeber, eds., Adolescent Emotional Development and the Emergence of Depressive Disorders, 2008, pp. 1–10. Amu, Hubert, Abdul-Aziz Seidu, Wonder Agbemavi, Bernard Owusu, Bright Ahinkorah, Edward Ameyaw, and Kissah-Korsah Kwaku, “Psychosocial Distress Among In-school Adolescents in Mozambique: A Cross-sectional Study Using the Global School-Based Health Survey Aata,” Child and Adolescent Psychiatry and Mental Health, 09 2020, 14, 1–9. Angrist, Noam, Andreas de Barros, Radhika Bhula, Shiraz Chakera, Chris Cum- miskey, Joseph DeStefano, John Floretta, Michelle Kaffenberger, Benjamin Piper, and Jonathan Stern, “Building Back Better to Avert a Learning Catastrophe: Estimating Learning Loss from COVID-19 School Shutdowns in Africa and Facilitating Short-term and Long-term Learning Recovery,” International Journal of Educational De- velopment, 2021, 84, 102–397. Azevedo, João Pedro, Amer Hasan, Diana Goldemberg, Koen Geven, and Syedah Aroob Iqbal, “Simulating the Potential Impacts of COVID-19 School Closures on Schooling and Learning Outcomes: A Set of Global Estimates,” The World Bank Research Observer, 03 2021, 36 (1), 1–40. Bandiera, Oriana, Niklas Buehren, Markus Goldstein, Imran Rasul, and Andrea Smurra, “The Economic Lives of Young Women in the Time of Ebola: Lessons from an Empowerment Program,” 2019. World Bank Policy Research Working Paper No. 8760. 15 Becker, Eva Susann, Thomas Goetz, Vinzenz Morger, and John Ranellucci, “The importance of teachers’ emotions and instructional behavior for their students’ emotions — An experience sampling analysis,” Teaching and Teacher Education, 2014, 43, 15–26. Bhatia, Richa, “Effects of the COVID-19 Pandemic on Child and Adolescent Mental health,” Current Opinion in Psychiatry, 08 2020, Publish Ahead of Print. Boateng, Godfred O., David Teye Doku, Nancy Innocentia Ebu Enyan, Samuel Asiedu Owusu, Irene Korkoi Aboh, Ruby Victoria Kodom, Benard Ekumah, Reginald Quansah, Sheila A. Boamah, Dorcas Obiri-Yeboah, Epa- phrodite Nsabimana, Stefan Jansen, and Frederick Ato Armah, “Prevalence and Changes in Boredom, Anxiety and Well-being Among Ghanaians During the COVID-19 Pandemic: A Population-based Study,” BMC Public Health, 2021, 21. Chen, Jiyao, Nusrat Farah, Rebecca Kechen Dong, Richard Z. Chen, Wen Xu, Allen Yin, Bryan Z. Chen, Andrew Delios, Saylor Miller, Xue Wan, and Stephen X. Zhang, “The Mental Health Under the COVID-19 Crisis in Africa: A Systematic Review and Meta-Analysis,” medRxiv, 2021. Chongwo, Esther, Derrick Ssewanyana, Carophine Nasambu, Patrick N. Mwan- gala, Paul M. Mwangi, Moses K. Nyongesa, Charles R. Newton, and Amina Abubakar, “Validation of a Swahili Version of the World Health Organization 5-item Well-being Index among Adults Living with HIV and Epilepsy in Rural Coastal Kenya,” Global Health Research and Policy, 2018, 3. Cook, Mary N, John Peterson, and Christopher Sheldon, “Adolescent depression: an update and guide to clinical decision making,” Psychiatry (Edgmont), 2009, 6 (9), 17. Cumbe, Vasco, Alberto Muanido, Nelia Manaca, Hélder Fumo, Pedro Chiruca, Leecreesha Hicks, Jair Mari, and Bradley Wagenaar, “Validity and Item Response Theory Properties of the Patient Health Questionnaire-9 for Primary Care Depression Screening in Mozambique (PHQ-9-MZ),” BMC Psychiatry, 07 2020, 20. de Paz, Carmen, Miriam Muller, Ana Maria Munoz Boudet, and Isis Gaddis, “Gender Dimensions of the COVID-19 Pandemic,” 2020. Dione, Malick, Codé Lo, Moustapha Seye, Abdou Salam Fall, Melissa Hidrobo, Agnès Le Port, Jessica Heckert, and Amber Peterman, “Women and Adoles- 16 cent Girls’ Experience with COVID-19 in Rural Senegal,” 2021. IFPRI Project Note https://doi.org/10.2499/p15738coll2.134274. Duan, Li, Xiaojun Shao, Yuan Wang, Yinglin Huang, Junxiao Miao, Xueping Yang, and Gang Zhu, “An Investigation of Mental Health Status of Children and Adolescents in China During the Outbreak of COVID-19,” Journal of Affective Disorders, 2020, 275, 112–118. Egger, Dennis, Edward Miguel, Shana S. Warren, Ashish Shenoy, Elliott Collins, Dean Karlan, Doug Parkerson, A. Mushfiq Mobarak, Günther Fink, Christo- pher Udry, Michael Walker, Johannes Haushofer, Magdalena Larreboure, Su- san Athey, Paula Lopez-Pena, Salim Benhachmi, Macartan Humphreys, Layna Lowe, Niccoló F. Meriggi, Andrew Wabwire, C. Austin Davis, Utz Johann Pape, Tilman Graff, Maarten Voors, Carolyn Nekesa, and Corey Vernot, “Falling Living Standards During the COVID-19 Crisis: Quantitative Evidence from Nine Developing Countries,” Science Advances, 2021, 7 (6), eabe0997. Engzell, Per, Arun Frey, and Mark D. Verhagen, “Learning Loss due to School Clo- sures During the COVID-19 Pandemic,” Proceedings of the National Academy of Sciences, 2021, 118 (17). Favara, Marta, Richard Freund, Catherine Porter, Alan Sánchez, and Douglas Scott, “Young lives, interrupted: short-term effects of the COVID-19 pandemic on ado- lescents in low-and middle-income countries,” Covid Economics: Vetted and Real-Time Papers, 2021, 67, 172–98. Ferrão, Jorge, Ivan Collinson, and Carlos Lauchande, “Educação, Capital Humano e Economia, em Tempos de Covid-19, em Moçambique,” Sémata: Ciencias Sociais e Humanidades, 07 2020, pp. 02–07. Garland, Ann, Negussie Deyessa, Menelik Desta, Atalay Alem, Tigist Zerihun, Kristopher Hall, Nicole Goren, and Irving Fish, “Use of the WHO’s Perceived Well-Being Index (WHO-5) as an Efficient and Potentially Valid Screen for Depression in a Low Income Country,” Families, Systems, & Health, 2018, 36 (2), 148–158. Hevia, Felipe J., Samana Vergara-Lope Tristan, Anabel Velásquez-Durán, and David Calderón Mart́ın del Campo, “Estimation of the Fundamental Learning Loss 17 and Learning Poverty Related to COVID-19 Pandemic in Mexico,” International Journal of Educational Development, 2021, pp. 102–515. Kim, Andrew Wooyoung, Tawanda Nyengerai, and Emily Mendenhall, “Evaluat- ing the Mental Health Impacts of the COVID-19 Pandemic: Perceived Risk of COVID-19 Infection and Childhood Trauma Predict Adult Depressive Symptoms in Urban South Africa,” Psychological Medicine, 2020, p. 1–13. Krieger, Tobias, Johannes Zimmermann, Silke Huffziger, Bettina Ubl, Carsten Diener, Christine Kuehner, and Martin Grosse Holtforth, “Measuring Depression with a Well-being Index: Further Evidence for the Validity of the WHO Well-Being Index (WHO-5) as a Measure of the Severity of Depression,” Journal of Affective Disorders, 2014, 156, 240–244. Langsi, Raymond, Uchechukwu Levi Osuagwu, Christopher Piwuna, Em- manuel Abu, Khathutshelo Mashige, Bernadine Ekpenyong, Godwin Ovenseri- Ogbomo, Chikasirimobi Timothy, Chundung Asabe, Tanko Ishaya, Richard Oloruntoba, Obinna Nwaeze, Deborah Charwe, and Kingsley Agho, “Prevalence and Factors Associated with Mental and Emotional Health Outcomes among Africans dur- ing the COVID-19 Lockdown Period - A Web-Based Cross-Sectional Study,” 01 2021. Loades, Maria Elizabeth, Eleanor Chatburn, Nina Higson-Sweeney, Shirley Reynolds, Roz Shafran, Amberly Brigden, Catherine Linney, Megan Niamh McManus, Catherine Borwick, and Esther Crawley, “Rapid Systematic Review: The Impact of Social Isolation and Loneliness on the Mental Health of Children and Adolescents in the Context of COVID-19,” Journal of the American Academy of Child Adolescent Psychiatry, 2020, 59 (11), 1218–1239.e3. Mahmud, Mahreen and Emma Riley, “Household Response to an Extreme Shock: Evidence on the Immediate Impact of the Covid-19 Lockdown on Economic Outcomes and Well-being in Rural Uganda,” World Development, 2021, 140, 105318. Mambo, F., M. Basso, V. Salvucci, and R. Santos, “An Analysis of School Dropout in Mozambique, 2014-15,” WIDER Working Paper, 2019, No. 2019/49. MINEDH, “Medidas Adicionais Para a Prevenção do COVID-19 nas Instituições Públicas de Educação Geral, Formação de Professores, Centros Internatos e Lares e Salas de Es- tudo.,” 2020. 18 , “Prevenção do COVID-19 - Medidas Adicionais de Reforço a Serem Aplicadas nas Insti- tuições Públicas e Privadas de Educação Geral e Formação de Professores.,” 2020. Mwabe, Julie, Karen Austrian, and Sheila Macharia, “Promises to Keep: Impact of COVID-19 on Adolescents in Kenya,” 2021. Nolan, C. P., P. J. M. O’Donnell, B. M. Desderius, M. Mzombwe, M. L. Mc- Nairy, R. N. Peck, and J. R. Kingery, “Depression Screening in HIV-positive Tanza- nian Adults: Comparing the PHQ-2, PHQ-9 and WHO-5 Questionnaires,” Global Mental Health, 2018, 5, e38. Ochnik, Dominika, Aleksandra Rogowska, Cezary Kuśnierz, Monika Jakubiak, Astrid Schuetz, Marco Held, Ana Arzenšek, Joy Benatov, Rony Berger, Elena Korchagina, Iuliia Pavlova, Ivana Blažková, Imran Aslan, Orhan Çınar, and Yonni Angel Cuero Acosta, “Mental Health Prevalence and Predictors Among Uni- versity Students in Nine Countries During the COVID-19 Pandemic: A Cross-national Study,” Scientific Reports, 09 2021, 11, 18644. Organization, World Health, “Health for the World’s Adolescents: A Second Chance in the Second Decade,” Report, World Health Organization, Geneva, 2014. Psacharopoulos, George, Victoria Collis, Harry Patrinos, and Emil- iana Vega, “Lost Wages: The COVID-19 Cost of School Closures,” 2020. http://dx.doi.org/10.2139/ssrn.3682160. Qi, Meng, Shuang-Jiang Zhou, Zhao-Chang Guo, Li-Gang Zhang, Hong-Jie Min, Xiao-Min Li, and Jing-Xu Chen, “The Effect of Social Support on Mental Health in Chinese Adolescents During the Outbreak of COVID-19,” Journal of Adolescent Health, 2020, 67 (4), 514–518. Reimers, F.M. and A. Shleicher, “A Framework to Guide an Education Response to the COVID-19 Pandemic of 2020,” 2020. Sabates, Ricardo, Emma Carter, and Jonathan M.B. Stern, “Using Educational Transitions to Estimate Learning Loss due to COVID-19 school closures: The Case of Complementary Basic Education in Ghana,” International Journal of Educational Devel- opment, 2021, 82, 102–377. 19 Saurabh, Kumar and Shilpi Ranjan, “Compliance and Psychological Impact of Quaran- tine in Children and Adolescents due to Covid-19 Pandemic,” Indian Journal of Pediatrics, 2021, 87 (7), 532–536. Spitzer, Robert L., Kurt Kroenke, Janet B. W. Williams, and Bernd Löwe, “A Brief Measure for Assessing Generalized Anxiety Disorder: The GAD-7,” Archives of Internal Medicine, 05 2006, 166 (10), 1092–1097. Topp, Christian, Søren Østergaard, Susan Søndergaard, and Per Bech, “The WHO-5 Well-Being Index: A Systematic Review of the Literature,” Psychotherapy and psychosomatics, 04 2015, 84, 167–76. UNESCO, “Education: From disruption to recovery,” 2021. https://en.unesco.org/covid19/educationresponse, retrieved December 6, 2021. UNESCO Institute of Statistics, 2021. http://data.uis.unesco.org, retrieved December 1, 2021. UNESCO, UNICEF, and World Bank, “The State of the Global Education Crisis: A Path to Recovery,” December, 2021. UNICEF, “Longitudinal Assessment of Dropout in Mozambique,” forthcoming. Visser, Maretha and Eloise Law van Wyk, “University Students’ Mental Health and Emotional Wellbeing During the COVID-19 Pandemic and Ensuing Lockdown,” South African Journal of Psychology, 2021, 51 (2), 229–243. Wang, Dongqing, Angela Chukwu, Ourohiré Millogo, Nega Assefa, Christabel James, Tara Young, Bruno Lankoande, Firehiwot Workneh, Elena C. Hem- ler, Michelle L. Korte, Josiemer Mattei, Abdramane Bassiahi Soura, Ali Sie, Ayoade Oduola, Yemane Berhane, and Wafaie W. Fawzi, “The COVID-19 Pan- demic and Adolescents’ Experience in Sub-Saharan Africa: A Cross-Country Study Using a Telephone Survey,” The American Journal of Tropical Medicine and Hygiene, 2021, 105 (2), 331 – 341. World Health Organization, “Depression and Other Common Mental Disorders: Global Health Estimates,” 01 2017. , 2021. https://covid19.who.int/region/afro/country/mz, retrieved December 20, 2021. 20 World Population Review, “Depression Rates by Country,” 2021. Zhou, Shuang-Jiang, Li-Gang Zhang, Lei-Lei Wang, Zhao-Chang Guo, Jing-Qi Wang, Jin-Cheng Chen, Mei Liu, Xi Chen, and Jing-Xu Chen, “Prevalence and Socio-demographic Correlates of Psychological Health Problems in Chinese Adolescents During the Outbreak of COVID-19,” European Child & Adolescent Psychiatry, 2020, 29, 749–758. 21 Tables Table 1: Experiences with COVID-19 Mean SD Observations Panel A: Adolescent Students Household member went hungry 0.530 0.499 1383 Household member lost employment 0.083 0.276 1383 Household member migrated 0.131 0.337 1383 Female household member married early 0.275 0.447 1383 Panel B: Teachers Personally affected by COVID-19 0.309 0.463 175 Household member lost employment 0.080 0.272 175 Teacher lost income 0.080 0.272 175 Difficult to manage students 0.120 0.326 175 Notes: This table shows the prevalence of COVID-19 shocks experienced by adolescent students (Panel A) and teachers (Panel B). Column 1 reports the mean, column 2 the standard deviation, and column 3 the number of observations. Note that these measures were not recorded for deputy school directors. 22 Table 2: Adolescent Student and Educator Mental Health Mean SD Observations Panel A: Adolescent Students GAD-7: Moderate or severe anxiety 0.102 0.303 1383 GAD-7: Score 4.925 3.973 1383 WHO-5: Low well-being 0.308 0.462 1068 WHO-5: Score 61.940 26.221 1068 Panel B: Teachers GAD-7: Moderate or severe anxiety 0.086 0.281 175 GAD-7: Score 3.943 4.048 175 WHO-5: Low well-being 0.124 0.331 137 WHO-5: Score 72.321 18.775 137 Panel C: Deputy School Directors GAD-7: Moderate or severe anxiety 0.069 0.253 175 GAD-7: Score 3.251 3.403 175 WHO-5: Low well-being 0.125 0.332 128 WHO-5: Score 75.031 19.119 128 Notes: This table displays means, standard deviations, and the number of observations of mental health outcomes for adolescent students (Panel 1), teachers (Panel 2), and deputy school directors. WHO-5 is a subjective well-being score ranging from 0-100 and low well-being is represented by a score below 50. Higher values represent higher levels of well-being. GAD-7 is a measure of generalized anxiety disorder and ranges from 0 to 21. Moderate to severe anxiety is present when the score is below 10. Higher values represent more anxiety. Note that 315 students, 38 teachers, and 47 school directors did not complete the WHO-5 module, and instead completed the PHQ-9 module during the first survey phase; this module was phased out due to respondent distress, and is not analyzed. 23 Table 3: Predictors of Adolescent Mental Health GAD-7 - Anxiety WHO-5 - Well-being Binary Continuous Binary Continuous inverted (1) (2) (3) (4) School closures: school promoted radio education 0.0383 1.014 -0.0958 -2.933 (0.0424) (0.680) (0.0792) (3.825) School closures: school contacted parents -0.0176 -1.205∗∗∗ -0.0130 -2.298 (0.0219) (0.449) (0.0547) (3.556) School closures: school took other measures -0.0874∗∗∗ -1.901∗∗∗ -0.0487 -6.154 (0.0265) (0.582) (0.0747) (4.847) School reopening COVID-19 mitigation index -0.00893 0.0435 0.0133 1.397 (0.0102) (0.182) (0.0375) (2.034) Average grade 4 test score 0.00128∗ 0.0263∗∗ -0.00131 -0.0399 (0.000650) (0.0122) (0.00168) (0.120) School quality index 0.00286 -0.0258 -0.0262 -1.951 (0.0145) (0.237) (0.0371) (2.426) COVID-19: Household member went hungry 0.0226 0.906∗∗∗ 0.0992∗∗∗ 8.423∗∗∗ (0.0185) (0.298) (0.0371) (2.519) COVID-19: Household member lost employment -0.0261 -0.471 0.00547 -2.167 (0.0254) (0.406) (0.0593) (3.246) COVID-19: Household member migrated -0.0447∗ 0.391 0.00253 1.663 (0.0252) (0.352) (0.0420) (2.241) COVID-19: Female household member got married 0.0846∗∗∗ 1.030∗∗∗ 0.0752 7.041∗∗∗ (0.0225) (0.295) (0.0477) (2.628) Any social support -0.00407 -0.721∗∗ -0.158∗∗∗ -8.155∗∗∗ (0.0209) (0.288) (0.0450) (2.899) Male -0.0170 -0.545∗∗∗ -0.0626∗∗ -3.857∗∗ (0.0128) (0.197) (0.0297) (1.617) Age 13-14 0.0281 0.436 0.0562 1.406 (0.0252) (0.371) (0.0517) (3.178) Age over 14 0.0337 0.272 0.0353 -0.420 (0.0310) (0.393) (0.0627) (3.669) Household asset index 0.000999 0.182∗∗ -0.00259 -0.429 (0.00683) (0.0899) (0.0145) (0.800) Household reports farm income -0.0279 -0.698 0.0864 3.812 (0.0430) (0.558) (0.0746) (3.371) Household reports non-farm business 0.0365∗∗ 1.219∗∗∗ 0.0198 1.629 (0.0173) (0.234) (0.0314) (1.798) Household size -0.00395 -0.0292 -0.0112∗ -0.491 (0.00368) (0.0489) (0.00568) (0.348) Father completed primary school 0.00223 -0.0114 -0.00332 1.303 (0.0183) (0.256) (0.0364) (2.072) Mother completed primary school 0.00331 0.0176 0.000184 -0.602 (0.0234) (0.274) (0.0366) (2.388) Mean dependent variable 0.10 4.92 0.31 61.94 Observations 1383 1383 1068 1068 Adjusted R2 0.031 0.135 0.050 0.094 Notes: This table shows predictors of adolescent students’ anxiety and well-being. The outcomes in columns 1 and 2 are an indicator equal to one if a student is classified as having moderate or severe anxiety according to the generalized anxiety disorder (GAD-7) scale or low well-being according to the WHO-5 subjective well-being scale, respectively. The outcomes in columns 2 and 4 are continuous measures of the GAD-7 measure of generalized anxiety disorder, which runs between 0 and 21 and the WHO-5 scale, which runs between 0 and 100. Here, the WHO-5 scale has been reverse coded so that a higher value is a worse outcome, consistent with the anxiety score. See the Appendix for variable definitions. All regressions include district indicator variables and standard errors, reported in parentheses, are clustered at the school level. * p < 0.1 ** p < 0.05 *** p < 0.01. 24 Table 4: Predictors of Educator Mental Health: Anxiety GAD-7 Binary GAD-7 Continuous Teachers Teachers/Directors Teachers Teachers/Directors (1) (2) (3) (4) School closures: school promoted radio education 0.0455 0.0618 -0.990 0.351 (0.0860) (0.0552) (1.074) (0.781) School closures: school contacted parents -0.0415 -0.0325 -0.112 -0.116 (0.0446) (0.0307) (0.697) (0.538) School closures: school took other measures 0.0127 0.00769 1.077 0.757 (0.0768) (0.0619) (0.990) (0.849) School reopening COVID-19 mitigation index 0.0200 0.0103 -0.0226 0.209 (0.0255) (0.0158) (0.309) (0.194) Average grade four test score -0.000369 -0.000563 -0.0195 -0.00621 (0.00172) (0.000904) (0.0265) (0.0137) School quality index 0.0465 0.00926 0.313 0.137 (0.0299) (0.0165) (0.363) (0.255) COVID-19: Teacher lost income 0.0185 1.090 (0.0714) (0.978) COVID-19: Household member lost income 0.112 2.846*** (0.119) (1.067) COVID-19: Hard to manage children 0.121 1.881* (0.0967) (1.081) Male 0.0401 -0.0292 0.252 -0.624 (0.0564) (0.0310) (0.771) (0.390) Age -0.00216 0.000309 -0.0388 -0.0175 (0.00270) (0.00175) (0.0481) (0.0265) Speaks Portuguese at home 0.0918 2.103*** (0.0565) (0.734) Years working at current school 0.00829 0.000517 0.247** 0.147* (0.00753) (0.00530) (0.108) (0.0849) Post-secondary education -0.0783* -0.00562 0.242 0.333 (0.0466) (0.0412) (0.748) (0.539) Mean dependent variable .09 .08 3.94 3.60 Observations 175 350 175 350 Adjusted R2 0.062 -0.004 0.101 0.002 Notes: This table shows predictors of educator (teacher and deputy school director) anxiety outcomes. Columns 1 and 3 report results for teachers only as COVID-19 experiences were not measured for deputy school directors. Columns 2 and 4 report results for both teachers and deputy school directors. The outcome in columns 1 and 2 is an indicator equal to one if a teacher or a teacher/deputy school director is classified as having moderate or severe anxiety. The outcome in columns 2 and 4 is a continuous measure of the GAD-7 measure of generalized anxiety disorder, which runs between 0 and 21. See the Appendix for variable definitions. All regressions include district indicator variables and standard errors, reported in parentheses, are clustered at the school level. * p < 0.1 ** p < 0.05 *** p < 0.01. 25 Table 5: Predictors of Educator Mental Health: Well-being WHO-5 Binary WHO-5 Continuous Teachers Teachers/Directors Teachers Teachers/Directors (1) (2) (3) (4) School closures: school promoted radio education 0.0498 -0.0628 4.586 0.447 (0.126) (0.0698) (5.881) (4.299) School closures: school contacted parents -0.0584 0.0126 -2.924 0.252 (0.0549) (0.0604) (3.274) (3.677) School closures: school took other measures 0.0859 0.0927 2.010 2.333 (0.0969) (0.0799) (4.814) (4.665) School reopening COVID-19 mitigation index 0.0524 0.0631*** 5.851*** 6.073*** (0.0387) (0.0222) (1.666) (1.298) Average grade four test score 0.00172 0.00179 0.247 0.194 (0.00299) (0.00177) (0.174) (0.120) School quality index 0.0115 0.0280 2.210 2.135 (0.0321) (0.0248) (1.817) (1.558) COVID-19: Teacher lost income 0.0364 0.783 (0.0793) (3.924) COVID-19: Household member lost income 0.542*** 27.46*** (0.146) (5.557) COVID-19: Hard to manage children -0.0296 -1.361 (0.0953) (4.127) Male -0.0167 0.0177 -2.950 -2.369 (0.0627) (0.0403) (3.486) (2.015) Age -0.00228 0.000513 -0.106 0.0436 (0.00367) (0.00299) (0.217) (0.173) Speaks Portuguese at home 0.00327 -0.146 (0.0634) (3.411) Years working at current school 0.00215 0.00251 -0.00622 0.0340 (0.0104) (0.00864) (0.541) (0.444) Post-secondary education 0.0742 0.00486 5.838 -1.430 (0.0937) (0.0632) (5.904) (3.825) Mean dependent variable .12 .12 73.32 74.18 Observations 137 265 137 265 Adjusted R2 0.175 -0.006 0.214 0.058 Notes: This table shows predictors of educator (teacher and deputy school director) subjective well- being outcomes, which can be interpreted as a measure of vulnerability to depression. Columns 1 and 3 report results for teachers only as COVID-19 experiences were not measured for deputy school directors. Columns 2 and 4 report results for both teachers and deputy school directors. The outcome in columns 1 and 2 is an indicator equal to one if a teacher or a teacher/deputy school director is classified as having low well-being. The outcome in columns 2 and 4 is a continuous measure of the WHO-5 measure of subjective well-being score, which runs between 0 and 100. Here, the WHO-5 scale has been reverse coded so that a higher value is a worse outcome, consistent with the anxiety score. See the Appendix for variable definitions. All regressions include district indicator variables and standard errors, reported in parentheses, are clustered at the school level. * p < 0.1 ** p < 0.05 *** p < 0.01. 26 Appendix A: Additional Tables 27 Table A1: Characteristics of the Sample Mean SD Obs Panel A: Adolescent Students Male 0.402 0.490 1383 Age 15.076 2.047 1382 Household reports farm income 0.960 0.197 1383 Household reports non-farm business 0.495 0.500 1383 Household size 6.142 2.187 1383 Father completed primary education 0.579 0.494 1153 Mother completed primary education 0.401 0.490 1207 Number of younger siblings 2.260 1.514 1383 Panel B: Teachers Male 0.246 0.432 175 Age 32.006 7.418 175 Teaching experience 7.949 6.384 175 Has post-secondary education 0.057 0.233 175 Permanent teacher 0.983 0.130 175 Speaks Portuguese at home 0.309 0.463 175 Panel C: Deputy School Directors Male 0.880 0.326 175 Age 38.863 7.649 175 Teaching experience 14.486 6.729 175 Has post-secondary education 0.229 0.421 175 Panel D: School Characteristics School size (enrollment) 948.132 740.432 1383 School closures: radio education 0.110 0.313 1383 School closures: school contacted parents 0.295 0.456 1383 School closures: school took other measures 0.133 0.340 1383 Notes: This table shows summary statistics of demographic characteristics of adolescent students (Panel 1), teachers (Panel 2), deputy school directors (Panel 3), and schools (Panel 4), reporting means, standard deviations, and the number of observations. The average grade four test score is the score on the Early Grade Literacy Assessment (EGRA), which was conducted as part of a larger study. The school quality index and school reopening COVID-19 mitigation measures index are standardized indices at the school level. See the Appendix for details on variable definitions. 28 Table A2: Variable Definitions COVID-19 Experiences Variable Definition Adolescents Household member went hungry Indicator equal to one if the student reported that a household member went hungry over the past year. Household member lost employment Indicator equal to one if the student reported that a household member lost employment in the past year. Female household member married Indicator equal to one if the student reported that a female member of the household married earlier than planned in the past year. Teachers Personally affected by COVID-19 Indicator equal to one if the teacher was personally affected by COVID-19 at all in the past year. Household member lost employment Indicator equal to one if the teacher reported that a house- hold member lost their employment in the past year. Zero is recorded for teachers not affected by the pandemic at all. Teacher lost income Indicator equal to one if the teacher lost a source of their income. Zero is recorded for teachers not affected by the pan- demic at all. Difficult to manage students Indicator equal to one if the teacher reported that students were more difficult to manage after the reopening of schools. Zero is recorded for teachers not affected by the pandemic at all. 29 Mental Health Variable Definition WHO-5: Score Total score of responses on the five questions included in the WHO-5 subjective well-being scale. The five questions are: I have felt cheerful in good spirits; I have felt calm and relaxed; I have felt active and vigorous; I woke up feeling fresh and rested; My daily life has been filled with things that interest me. Response options are: all of the time (5 points), most of the time (4 points), more than half the time (3 points), less than half the time (2 points), some of the time (1 point), none of the time (0 points). Total score is multiplied by 4 to calculate the%age score, which runs from 0-100. A higher value indicates a higher level of subjective well-being. In Table 2, this score is reported. In regression tables, the score is inverted so that higher numbers represent worse well-being. WHO-5: Low well- being Indicator equal to one if the WHO-5 score is less than 50 points. GAD-7: Score Total score of responses to seven questions included in the GAD-7 generalized anxiety disorder scale. Respondents are asked to the frequency with which they experienced seven feelings in the past two weeks. Respondents choose between the following options for each question: Not at all (0 points), several days (1 point), more than half the days (2 points), nearly every day (3 points). The questions are: Feeling ner- vous, anxious, or on edge; Not being able to stop or control worrying; Worrying too much about different things; Trouble relaxing; Being so restless that it is hard to sit still; Becom- ing easily annoyed or irritable; Feeling afraid, as if something awful might happen. The scale ranges from 0-21. GAD-7: moderate to severe anxiety Indicator equal to one if the GAD-7 score is above 10. Demographics Household asset index First principal component of the following variables: student’s household owns a radio, bicycle, cattle, and mobile phone. Index is standardized. Household reports farm income Indicator equal to one if the student reports that their family receives income from agriculture. Household reports non-farm business Indicator equal to one if the student reports that their family owns a non-farm business. 30 Variable Definition Permanent teacher Indicator equal to one if the teacher has a permanent govern- ment contract. Speaks Portuguese at home Indicator equal to one if one of the educator’s spoken lan- guages at home is Portuguese. School size Number of students enrolled in the school across all grades. Average grade four test score Average score among grade four students on the Oral Reading Fluency (ORF) section of the Early Grade Reading Assess- ment (EGRA) School quality index Index using the first principal component of the following variables: whether a school has i) latrines, ii) latrines that are gender separated, iii) electricity, iv) Portuguese reading materials for grade four students, v) local language reading material for grade four students, vi) soap/detergent for staff, vii) soap/detergent for students, viii) hand-washing station or tippy tap, and ix) whether the school has an appropriate water source. Index is standardized. School closures: radio education Indicator equal to one if the school encouraged students to listen to educational programming on the radio during school closures. School closures: school contacted parents Indicator equal to one if the school contacted parents to en- courage them to help their child continue their education dur- ing school closures and/or encouraged them to send their chil- dren back to school after schools reopened. School closures: school took other measures Indicator equal to one if the school took any of the following measures during school closures: giving students materials to take home, encouraging students to watch television educa- tion programming, visiting students at home, and gathering small groups of students for lessons. School reopening COVID-19 mitigation index Index using the first principal component of the following vari- ables: only allowed grade seven students when schools first reopened, only allowed grade 3 students at first, spaced stu- dents further apart, ensured students and staff washed their hands frequently, required staff to wear masks or other per- sonal protective equipment (PPE), held classes outside when possible, continued to keep libraries and other common spaces closed, provided information to students and parents, opened some classes before others, had different students attend on different days. Index is standardized. 31 ALL IFPRI DISCUSSION PAPERS All discussion papers are available here They can be downloaded free of charge INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE www.ifpri.org IFPRI HEADQUARTERS 1201 Eye Street, NW Washington, DC 20005 USA Tel.: +1-202-862-5600 Fax: +1-202-862-5606 Email: ifpri@cgiar.org https://www.ifpri.org/publications?sm_content_subtype_to_terms=6&sort_by=ds_year&f%5B0%5D=sm_content_subtype_to_terms%3D1&f%5B1%5D=sm_content_subtype_to_terms%3A88 http://www.ifpri.org/ mailto:ifpri@cgiar.org