This blog is co-authored with Renata Lemos, James Gresham, Rony Rodríguez-Ramírez, and Harry A. Patrinos
Wars devastate infrastructure and institutions, but their most profound costs are often borne by human capital. Education systems are among the first casualties of conflict: schools are destroyed, and families are displaced, cutting off access to schooling and remaining services. At the same time, governments divert education funding to military or emergency needs, causing investment in education to decline precisely when children need it most.
During Russia’s invasion of Ukraine, government spending on education fell from 17 percent of total expenditure in 2021 to just 7 percent in 2023, while between 2,900 and 3,500 schools were damaged or destroyed. By late 2022, only 30 percent of secondary schools were able to operate fully in person, while 34 percent moved entirely online and 36 percent adopted blended approaches. Despite these adaptations, Ukrainian students faced large learning losses. In 2022, Ukrainian students scored 428 in reading and 441 in mathematics on the PISA assessment, well below the OECD averages.
Underinvestment in education and other social services during conflict is often treated as unavoidable. The risk of infrastructure destruction, service disruption, and population displacement makes the delivery of public social services incredibly challenging and the returns on these investments uncertain. As a result, there is limited experimental evidence on interventions implemented during conflicts that can mitigate the adverse effects of war on human capital.
The case of an online tutoring program
In our recent working paper, we study an online tutoring program as an example of an educational investment that can be implemented during conflict. While tutoring programs have proven effective in non-war settings (Nickow et al., 2023; Gortazar et al., 2024; Carlana and LaFerrara, 2025), it is unclear whether the key components determining their impact hold up in wartime conditions. First, logistical challenges such as power outages and frequent displacement may hinder participation. Yet, students’ intrinsic motivation to recover lost learning may lead to high engagement despite adverse conditions. Second, the mechanisms through which tutoring operates may function differently during wartime. For example, structured peer interactions may both provide emotional support and facilitate learning, but they may also transmit stress or introduce distractions. Third, the broader context can moderate program impacts: the psychological toll of conflict may impair students’ ability to focus and reduce parents’ capacity to support their education.
Teach for Ukraine ran the online tutoring program as three consecutive experiments between early 2023 and mid-2024, each lasting six weeks. The program offered small-group tutoring in math and Ukrainian language for three hours a week and was adapted to the evolving context each time: the first, launched amid winter power outages, focused on core academic catch-up; the second introduced diagnostic tools to group students by ability and better support tutors; and the third incorporated psychosocial support through trauma-informed care exercises in response to students’ growing mental health needs. We evaluated each experiment through randomized controlled trials, reaching nearly 10,000 students across all regions of Ukraine by the end of the evaluation.
Feasibility: Delivering social services in the midst of war
The first key lesson is that delivering structured social services during war is feasible. Logistical barriers such as power outages, internet disruptions, and displacement were real, but demand was overwhelming. Enrollment exceeded program capacity in each experiment, and students’ take-up and attendance were remarkably high. Between 68 and 72 percent of students assigned to treatment attended at least one math session, and up to 71 percent attended Ukrainian language sessions, with average participation above six sessions per subject. Engagement levels were equally strong, with more than 95 percent of students attentive and participating during sessions according to tutor reports. Importantly, most absences were not due to dissatisfaction but to external shocks such as blackouts, illness, or school responsibilities. This pattern suggests that when services are offered, students and families are willing and able to engage even amid extreme adversity.
Effectiveness: Improving student’s learning and mental health
The second lesson is that wartime service delivery can be effective. Across the three experiments, we document sizable improvements in students’ learning (Figure 1) alongside reductions in stress (Figure 2). In terms of learning outcomes, the first experiment saw students score 0.49 standard deviations (SD) higher in math and 0.40 SD higher in Ukrainian language compared to the control group. The second experiment showed more modest gains—0.22 SD in math while Ukrainian language showed no improvement—but these were still meaningful, particularly given that this wave coincided with the end of the school year and the cancellation of final exams, which may have lowered student motivation. The third experiment again produced large effects, with math scores increasing by 0.22 SD and Ukrainian language by 0.32 SD. Regarding mental health, the first experiment saw a reduction of 0.10 SD, and the third experiment saw stress reduced by 0.12 SD. However, we do not find meaningful impact on anxiety, likely because the psychosocial content focused on stress-coping techniques such as breathing and grounding.
Figure 1. Impacts of the online tutoring program on student’s learning by experiment
Figure 2. Impacts of the online tutoring program on mental health outcomes by experiment
Pooling all three experiments, the program improved math scores by 0.32 SD, Ukrainian language scores by 0.28 SD, and reduced stress levels by 0.12 SD. These effects are large by international standards but not surprising, particularly given the severity of disruptions in Ukraine, where students had already lost 31 weeks of schooling due to COVID-19 closures, resulting in substantial learning loss before Russia’s invasion.
We also identify several potential mechanisms driving the program’s impact. Students offered a place in the program were more likely to engage with tutors and peers through the platform, showed improved attitudes toward learning, persistence, and self-efficacy (though not aspirations), and were more likely to seek extra tutoring support and spend more time using online academic resources.
Scalability: A cost-effective model with potential for national reach
The third lesson is that such interventions are scalable. Cost-effectiveness analysis shows benefit-to-cost ratios ranging from 31 to 56, among the highest reported for education interventions. In other words, for every dollar spent on the online tutoring program, we see benefits worth between 31 and 56 dollars. Even under conservative assumptions where only 20 percent of gains persist after five years, the ratios remain between 6.5 and 11.6. The program also successfully doubled its reach in the third experiment without increasing costs, keeping the average per-student cost at around $90 across all three experiments.
This scalability can also be made possible because the online tutoring program builds on existing education infrastructure by aligning sessions with the national curriculum and by hiring trained Ukrainian teachers, many of whom had spare capacity due to school closures. The intervention reached students across all oblasts, including high-conflict areas, and the sample closely mirrored the national student population on key characteristics, reinforcing the representativeness and generalizability of the results. Put simply, the program is not only cost-effective but also operationally feasible, institutionally compatible, and resilient in times of crisis.
Education continuity as a wartime necessity
The broader policy message is clear: in times of war, “keeping the lights on” isn’t just about electricity or infrastructure; it’s about people. Sustaining investments in children and families must be recognized as a vital part of essential services. Without opportunities to learn and grow, and without support for wellbeing, recovery will stall before it even begins. Governments and aid agencies can’t afford to overlook this. Investing in people is the foundation of true resilience.
The Ukrainian case shows that with the right partnerships, technology-enabled delivery models, and a focus on cost-effectiveness, social services can be delivered at scale even under conditions of ongoing conflict. This lesson extends far beyond Ukraine, offering guidance for fragile and conflict-affected settings around the world where protecting the next generation is essential for both humanitarian response and long-term development planning.
Join the Conversation