When floods wash away learning: Why India’s disaster planning must protect children

Floods are usually measured in terms of lives lost, homes damaged, crops destroyed and roads washed away. Far less visible is another consequence that can persist long after the water recedes: the loss of children’s learning. A school may reopen within days or weeks after a flood, but that does not mean education has returned to normal. Interrupted schooling, damaged learning materials, displacement, illness, disrupted transport and household economic distress can leave children permanently behind.

A new study, “Floods, community infrastructure, and children’s heterogeneous learning losses in rural India,” by Nazar Khalid, Jere R. Behrman, Emily Hannum and Amrit Thapa, published in the Economics of Education Review (2025), makes an important contribution to understanding this largely overlooked dimension of disaster vulnerability. The paper uses longitudinal data from the India Human Development Survey (IHDS) to examine not merely whether floods affect learning, but whether community infrastructure can cushion those losses and whether it does so equally for children from different social and economic backgrounds.

The researchers also looked at whether different types of infrastructure affect children equally. Their analysis considers differences related to caste, religion, gender, household income and parental education. The findings show that the relationship between infrastructure, flooding and learning is complex. Most of the physical infrastructure examined in the research was not significantly associated with lower flood-related learning losses. Some forms of social infrastructure, however, were associated with better outcomes.

The hidden educational cost of a flood

The researchers used the India Human Development Survey, a longitudinal survey covering more than 41,000 households across 971 urban blocks and 1,503 villages. For this analysis, they examined 7,250 children from households included in both survey rounds. The children were between eight and 11 years old. Their learning was assessed through reading, mathematics and writing tests conducted at home.

This approach is significant because it measures learning directly rather than relying only on school enrolment or attendance. The analysis found that children living in villages exposed to floods had learning skills that were, on average, 0.084 standard deviations lower than children living in villages that were not exposed to floods. Greater exposure to floods at the state level was also associated with lower learning outcomes.

Flood exposure was common among the children in the sample. About one-third lived in villages that experienced floods between 2004 and '05 and 2011 and '12. Exposure was also uneven across social groups. Hindu, Other Backwards Class, Hindu Scheduled Caste and Muslim children had relatively high exposure to flood-affected villages. Children from poorer households and those whose parents had no schooling were also more likely to live in villages exposed to floods.

The effects of a flood extend beyond the school building. When roads become impassable, children may not be able to reach school. When families lose income, children may be required to help at home or with livelihoods. Floodwaters can damage books, uniforms and other learning materials. Illness, displacement and disruption to household routines can make returning to school more difficult.

Physical infrastructure does not tell the whole story

The research examined both physical and social infrastructure. Physical infrastructure included electricity, roads, schools, healthcare facilities, piped water and toilets. Social infrastructure included social organisations, community safety, collective problem solving, teacher attendance, perceptions of teacher fairness and parent-teacher engagement. Some physical infrastructure was associated with better learning outcomes even outside flood conditions. Children living in households with electricity had higher learning outcomes. Greater toilet ownership was also positively associated with learning.

Distance from towns was another factor. Each additional kilometre from the nearest town was associated with a 0.004 standard deviation reduction in learning. For the average child living about 14 kilometres from a town, this amounted to approximately 0.056 standard deviations. However, these findings do not mean that every physical infrastructure investment automatically reduces learning losses during floods.

The research found that most physical infrastructure components were not significantly associated with lower flood-related learning losses. Proximity to towns was an important exception. For every additional kilometre from the nearest town, children in flood-affected villages experienced an additional reduction of about 0.005 standard deviations in learning.

The findings related to social infrastructure were more pronounced. Safety from crime was associated with substantially smaller learning losses in flood-affected villages. Higher baseline educational quality, perceptions that teachers were fair and unbiased, and regular parent-teacher engagement were also associated with smaller learning losses.

The estimated association with community safety was particularly large. Children in flood-affected villages perceived as safer had learning losses that were 1.032 standard deviations lower than children in otherwise flood-affected villages perceived as unsafe.

Social organisations, including community groups and self-help groups, can also provide channels for communication. They may help identify children who have stopped learning, mobilise support and connect vulnerable families with government services. The research also cautions that social infrastructure is not automatically beneficial. Social structures can reproduce existing exclusion and privilege if interventions do not take local inequalities into account.

The same flood does not affect every child equally

One of the important findings of the research is that flood-related learning losses vary between social groups. Disaster planning often refers to children, households or communities as broad categories. The evidence from this research shows that children within these categories can experience very different levels of vulnerability.

Caste and religion

The study did not find statistically significant differences in flood-related learning losses associated with physical infrastructure between marginalised Hindu and Muslim children and privileged Hindu upper-caste children.

Social infrastructure showed more differentiated outcomes. Safer villages were associated with reduced learning disparities among marginalised Hindu children. Lower levels of conflict were associated with smaller learning disparities among Muslim children. At the same time, the presence of parent-teacher associations was associated in some cases with larger disparities during floods. These findings show that community institutions can have different effects depending on the social context in which they operate.

Gender

The research found no significant evidence that physical infrastructure affected flood-related learning losses differently for girls and boys. Some social infrastructure characteristics appeared potentially protective for girls. These included teachers not using corporal punishment and the presence of parent-teacher associations. However, these findings did not remain statistically robust after corrections for multiple inferences.

The research also found that higher historical educational quality in a village was associated with greater flood-related learning losses for girls relative to boys. The authors indicate that this finding requires caution in interpretation.

Household income and parental education

The findings also point to differences based on household circumstances. Widespread access to piped water was associated with smaller flood-related learning losses among poorer children compared with richer children. This suggests that access to basic services can sometimes reduce differences in how children experience disasters.

Social organisations and communities with less conflict and collective problem-solving were associated with lower learning losses among children whose parents had no schooling compared with children whose parents had some schooling.

Together, these findings show why a single disaster response may not work equally well for every child. An intervention that helps one group may have little effect on another. In some circumstances, it may even widen existing differences.

What this means for disaster and education planning

The research does not suggest that physical infrastructure is unimportant. Rather, it shows that infrastructure is part of a wider social system. The relationship between infrastructure and flood-related learning losses is complex in a country where social and economic inequalities shape people's exposure to disasters and their ability to recover. This has several implications for disaster and education planning.

Identify vulnerable children before floods

Flood-prone districts could map schools and vulnerable children at village level before the monsoon. Such mapping could combine education information with disaster risk information. It could identify schools that may become inaccessible, settlements that are likely to be cut off and children who may face greater barriers to returning to school.

Plan for learning continuity

Flood preparedness plans could include specific measures to keep children learning when schools are inaccessible. These could include rapidly replacing textbooks and uniforms, providing temporary learning spaces, identifying safe transport routes, allowing flexible assessment schedules and offering remedial education after schools reopen. The recovery of a school should therefore be assessed not only by whether its building is functioning but also by whether children have been able to resume learning.

Strengthen school and community relationships

The findings point to the role of teacher attendance, teacher fairness, parent-teacher engagement, community safety, collective problem-solving and local organisations. School Management Committees and parent groups could be included in disaster preparedness alongside their regular responsibilities. Community organisations can also help identify children who are unable to return to school and connect families with available support.

Consider differences between children

Flood recovery programmes can use information on gender, caste, religion, poverty, parental education and location to understand different barriers to learning. For example, a girl from a poor household where parents have little formal education may face different challenges from a child in a household with greater resources.

Children from marginalised communities may also face different access barriers depending on where their settlements are located and their proximity to transport and public services. The research shows why these differences need to be considered when programmes are designed.

Measuring the learning cost of disasters

Disaster assessments commonly measure damage to roads, houses, crops and public infrastructure. The educational cost of disasters receives less attention. The research points to the value of measuring factors such as lost instructional time, damaged learning materials, absenteeism, dropout risks and learning deficits. Without such information, the effect of floods on education can remain largely invisible in disaster loss assessments. Tracking learning recovery could also help governments understand whether children have caught up after schools reopen or whether some groups remain behind.

Moving from emergency relief to learning resilience

The central message from the research is not that physical infrastructure should be replaced by social interventions. Instead, it highlights the need to consider both physical services and the social environment in which children live and learn. A school needs a safe building and reliable access. Children also need teachers, families, community support and functioning services around them.

The research shows that these factors can interact differently with flood exposure and with the social characteristics of children. This has relevance for how India approaches flood risk management. Educational continuity can be considered alongside other elements of disaster preparedness. Flood planning can identify schools and children at risk, prepare alternatives for periods when schools cannot function normally and include measures to support learning after reopening.

At the same time, investments in physical infrastructure remain relevant because access to electricity, water, sanitation, transport and other basic services influences children's everyday learning conditions. The evidence suggests that resilience cannot be reduced to a single intervention.

Keeping children connected to learning

Floods cannot always be prevented. Their effects on children's education, however, can be identified and addressed as part of disaster planning. The evidence from rural India shows that children can lose learning when floods disrupt schools, households, transport and basic services. It also shows that these losses are not distributed equally.

Children's ability to recover can be influenced by their household circumstances, social identity, location, access to services and the strength of relationships within their communities. For disaster management, this means that the return of floodwaters to normal levels is only one measure of recovery. A school may reopen, but a child's education may still be disrupted.

Understanding and measuring these learning losses can help education and disaster management authorities identify where additional support is needed and how recovery can be made more responsive to the children and communities affected.