The gendered effects of the COVID-19 pandemic on adolescent literacy and schooling outcomes in India

India hosts about 20% of the world’s adolescents (253 million aged 10–19). Before COVID-19, foundational literacy and numeracy were not universal: in 2018, only 73% of eighth graders could read a grade 2 text and 44% could solve a basic division problem, with gender gaps favoring boys. The pandemic led to prolonged school closures in India—over 18 months—following a nationwide lockdown starting 24 March 2020. Remote learning was introduced but access to devices, connectivity, and teacher support varied widely by state, school type, and household resources. Given pre-existing socioeconomic and gender disparities, policymakers need evidence on which adolescents were most affected. The study aims to estimate the effect of the pandemic on literacy and schooling outcomes among 15–17-year-old girls and boys in India, addressing limitations of prior studies by adjusting for background socioeconomic characteristics using quasi-experimental methods on nationally representative NFHS-5 data.

Multiple sources document pandemic-related learning disruptions in India. ASER 2020 reported unequal access to digital tools (e.g., 38% of 6–14-year-olds lacked a smartphone; lower provision of online content and teacher training in government vs private schools). ASER 2022 found near-universal enrollment but declines in reading proficiency across grades. A Tamil Nadu study estimated learning losses of 0.7 SD in math and 0.34 SD in language among 5–7-year-olds and, accounting for background characteristics, found no gender or SES gaps—findings not necessarily generalizable nationally. Prior literature shows strong intergenerational links and SES gradients in education in India, and boys are more likely to attend private schools, potentially widening gender gaps. Global and national estimates project substantial economic losses from learning deficits (e.g., 3.19% of India’s GDP by 2030). However, national estimates that adjust for socioeconomic differences and examine gendered effects among older adolescents were lacking; this study fills that gap.

Data: NFHS-5 (2019–2021), nationally representative across all Indian states/UTs, with two phases: June 2019–Jan 2020 and Jan 2020–Apr 2021 (paused Apr–Nov 2020 due to lockdown). The survey includes detailed household/individual socioeconomic indicators and biomarkers. Study population: adolescents aged 15–17 years; girls (~100,000) and boys (~13,700). Exposure definition: pre-COVID group surveyed before 25 March 2020; post-COVID group surveyed on/after 25 March 2020. Outcomes: (1) Literacy—binary indicator for ability to read a complete sentence in the respondent’s language (collected for all in NFHS-5); (2) Schooling—current enrollment status; (3) Years of schooling completed; (4) For out-of-school adolescents: main reason not enrolled. For girls: marriage, employment, cost; for boys: employment, cost. Methods: Employed propensity score matching (PSM) and inverse propensity score weighted (IPW) regression to adjust for observed differences. First-stage logit for post-COVID status included: age and age squared; years of schooling; height-for-age z-score; relationship to household head (self, spouse, child, in-law, grandchild); household head’s age, sex, and schooling; household size; rural residence; caste (SC, ST, OBC); religion (Muslim, Christian, Sikh); and wealth quintiles (2–5; Q1 as reference). Matching: one-to-one nearest neighbor with replacement, restricted to common support; also three-nearest neighbors as sensitivity. ATT estimated as the average difference between matched post- and pre-COVID groups. IPW: inverse-propensity weights applied in weighted linear probability models including the same covariates and a post-COVID indicator. Subgroup analyses for literacy by residence (rural/urban), caste, religion, and wealth quintile; schooling outcomes not stratified due to power. Matching quality: standardized percentage bias checked pre- and post-matching; all methods substantially reduced bias, with IPW most effective. Statistical significance at 5%.

- Sample: Girls—66,994 pre-COVID, 32,936 post-COVID (27,018 out of school). Boys—9,217 pre-COVID, 4,497 post-COVID (2,204 and 1,001 out of school respectively). - Literacy (girls): Post-COVID girls had 1.5–1.6 percentage points lower ability to read a sentence vs matched pre-COVID girls (ATT ≈ −0.016 to −0.015). Losses were larger among: • Rural girls: ≈ −1.7 to −1.8 pp (about twice urban). • Caste: OBC ≈ −1.5 to −1.7 pp; SC/ST ≈ −1.2 to −1.5 pp; upper caste not significant. • Wealth: Poorest quintile Q1 ≈ −3.1 to −3.8 pp; Q2 ≈ 0 to −1.6 pp; Q3/Q4 ≈ −0.9 to −2.1 pp; Q5 no significant loss. - Literacy (boys): No significant overall change; some subgroup declines in matching models (upper caste: −3.3 pp nearest-neighbor; wealth Q4: −3.7 to −2.8 pp) but not robust in IPW. - Years of schooling completed: • Girls: −0.08 to −0.10 years post-COVID vs matched pre-COVID across models. • Boys: −0.10 years in IPW (significant); matching estimates negative but not significant. - Out-of-school rates: • Girls: No significant difference pre vs post. • Boys: 2.2–2.4 pp lower post-COVID vs matched pre-COVID. - Reasons for being out of school: • Girls: Lower probabilities of marriage (−3.5 to −3.9 pp) and employment (−3.2 to −3.6 pp) as main reasons; higher reporting of “cost of schooling” (+5.6 to +5.9 pp). • Boys: Higher reporting of “cost of schooling” (+5.2 to +6.8 pp); employment reason unchanged. - Inequities: Pandemic-related learning loss among girls was amplified by rural residence, disadvantaged caste status, and poverty; such amplification was not observed for boys.

The study addresses whether COVID-19 exacerbated gendered disparities in adolescent learning and schooling in India. After adjusting for socioeconomic differences using PSM and IPW on NFHS-5, girls experienced a measurable decline in literacy and schooling attainment, concentrated among poorer, rural, and disadvantaged caste groups. Boys, in contrast, showed no literacy loss and slightly lower out-of-school rates post-pandemic. The findings suggest school closures likely drove declines in girls’ reading ability even in mid-adolescence, indicating that foundational literacy can erode without engagement. Structural gender inequalities—differential school quality (government vs private), access to digital tools, and time-use burdens—likely contributed to the disproportionate impact on girls. Results are broadly consistent with ASER trends, though differences in period and methods explain some discrepancies. The study underscores the need for gender-responsive recovery policies to prevent long-term human capital losses.

Using nationally representative NFHS-5 data and quasi-experimental methods, the study demonstrates modest but significant declines in literacy and schooling attainment among adolescent girls during the first pandemic year in India, with the largest losses among the poorest and other marginalized groups. Boys did not experience comparable literacy losses and were less likely to be out of school post-pandemic. Findings highlight the urgency of targeted, equitable remedial education and resource allocation to mitigate gendered learning losses. Future research should employ longer-term follow-ups to track recovery, examine pathways (digital access, time use), and assess effectiveness and cost-effectiveness of remediation programs across diverse settings.

- Unobserved confounding: Despite extensive covariate adjustment, unmeasured differences between pre- and post-COVID groups may bias estimates if correlated with outcomes. - Measurement error: During school closures, self-reported enrollment and attainment may reflect pre-closure status or assumed progression; NFHS-5 lacks detail to correct for this. - Missing pathways: NFHS-5 does not include time-use or digital access measures that could clarify mechanisms behind observed effects. - Statistical power: The boys’ sample is much smaller than the girls’, limiting detection of small effects and precluding detailed subgroup analyses. - Survey timing: Post-COVID data were collected entirely during school closures, which may influence reporting and comparability.