Education Technology

Access to Computers

The earliest set of interventions in this category focused primarily on hardware provision. The theory-of-change behind these interventions was that, if students had access to suitable ICT devices (typically laptops or tablets, frequently equipped with educational software), they would focus on educational content and improve their achievement levels.

However, causal evidence on such interventions has not been promising. An early influential paper by Malamud and Pop-Eleches (2011) used a regression discontinuity design to study a policy in Romania that subsidised the purchase of personal computers for use at home for students from low-income families. They find that, while computer use increased, students did not dedicate additional time to homework or other academic activities. On core academic outcomes such as GPA in math, English and Romanian language, they find significant negative effects. Time spent playing computer games increased substantially. Thus, even early evidence suggested that merely providing hardware was unlikely to relax binding constraints for learning.

The best experimental body of evidence comes from a series of successive evaluations of the “One Laptop per Child” (OLPC) programme. Beuermann et al. (2015) present short-term effects of randomly providing laptops to primary school students in Peru. After about 5 months, they find essentially no evidence of positive effects on academic achievement, while teachers report reduced student academic effort. The study featured a two-stage design — first randomising schools into treatment and then randomly selecting students through a lottery — which allows the authors to test for spillovers (for which they fail to find any evidence).

Cristia et al. (2017) report results from a complementary randomised evaluation of the OLPC programme in Peru. This differs from Beuermann et al. (2015) in three crucial respects:(i) the sample is much larger (318 schools versus 28); (ii) all students in treated schools were provided laptops, potentially enabling teachers to incorporate computers into instruction for the whole class, and (iii) results were tracked over a longer time-frame (15 months). Despite differences in the design, Cristia et al. (2017) also find no effects on academic or cognitive achievement. Although students made substantial use of the computers in schools and at home, test scores were similar for control and treated schools after 15 months.

In a final study, Cueto et al. (2024) study the long-run outcomes of a randomly assigned expansion of the OLPC programme in 2009, using a sample of 531 schools and administrative data up to 2019. As with the previous trials, they find no improvement on academic achievement: point estimates are negative (although not statistically significant), and they find significant negative effects on grade progression in primary and secondary schools. If anything, the evidence suggests that the OLPC intervention worsened academic outcomes.[1]1. We focus on the OLPC expansion in Peru since it has been most extensively studied with well-powered experimental designs. However, similarly disappointing results have been also reported in hardware-focused expansions in other settings. See, especially, De Melo et al. (2017), who document null effects of Uruguay's nation-wide rollout involving the distribution of ∼ 1 million laptops; Meza-Cordero (2017), who document null effects in Costa Rica (with negative but statistically insignificant point estimates on test scores); and Bando et al. (2017) who report null effects in Honduras from replacing textbooks with laptops. Likewise, Barrera-Osorio and Linden (2009) also find statistically insignificant effects on achievement from a hardware-focused intervention in Colombia that provided computers and encouraged their use by teachers and students. Mo et al. (2013) report no effects of a small OLPC trial in China on language skills; although they report marginally significant effects on math skills, these are imprecise and only seen in some specifications.

The null (and possibly negative) effects of laptop distribution are among the most important findings in the literature on EdTech in developing countries. This is for two related reasons. First, and most importantly, hardware-focused interventions remain very popular with policymakers. These include not only laptop distribution programmes but also many variants of ‘smart’ classrooms incorporating smart boards, projectors, and other ICT hardware. The OLPC initiative alone distributed millions of laptops, and governments in many countries have implemented similar programmes with their own funds outside of the OLPC programme. The evidence summarised above strongly suggests that these efforts are unlikely to improve learning outcomes. Second, these interventions are often very expensive. This is particularly a concern given constrained education budgets and robust evidence that a range of alternative interventions can lead to large learning gains at much lower costs (Angrist et al. 2024).

Access to the Internet

In contrast to the substantial evidence on unsupervised home computer use and laptop distribution, there is relatively nascent evidence on the impacts of internet access.

The first study of note is Lakdawala et al. (2023), which uses an event-study design to evaluate the provision of internet access to schools in Peru. Using multi-year administrative data on student achievement, and relying on the timing of internet installation on the primary schools attended by these children, they report that internet access to schools had only very small short-run effects on achievement (∼0.02–0.03σ in the first year) but substantially stronger gains (∼0.11σ) about five years after adoption. They interpret the pattern of these dynamic effects as illustrating a period of adaptation to the new technology by teachers and schools. These gains are likely a result from both students' direct use of the internet (which they show rises after introduction of internet access) and from optimisation by teachers, who report using the internet for teaching preparation.

The second study that evaluates internet access in schools in LMICs, albeit in a very specific setting, is Derksen et al. (2022). They evaluate the effects of providing randomly assigned students in four residential schools in Malawi with access to a digital library equipped with internet-enabled Android devices. Access was carefully restricted: the devices only allowed access to Wikipedia, students’ ability to access the library was limited to designated hours, and the digital library was also supervised by an adult. Thus, the intervention is best interpreted as providing supervised access to an (expansive) encyclopaedia on the internet, rather than unrestricted internet access per se. The authors report improvements in English and Biology test scores, both of which had been primed in the intervention, and provide detailed evidence on students’ (anonymised) browsing activity on Wikipedia.

The positive results on providing internet access in schools do not, however, indicate that internet access is always positive. Jain and Stemper (2024) study the effects of the expansion of 3G internet across 82 countries using data from the OECD’s Programme for International Student Assessment (PISA).[2]2. Although most PISA countries are high-income, the sample also includes several middle-income countries in Latin America and Asia.  Their principal result is that PISA test scores decline meaningfully (by about 0.07σ in math and ∼0.05σ in reading and science). These estimates are a meaningful cause of concern, as they should be interpreted as intent-to-treat (ITT) estimates: the independent variable is access to 3G coverage in an area, not actual usage by an individual household or student. These ITT effects could hide substantial heterogeneity, with negative effects being much more pronounced for students who use the internet more intensively.

The studies above suggest that the crucial aspect of evaluating access to the internet is not access itself, but rather how the technology is used. Supervised student use may be beneficial, but it requires substantial work in developing pedagogical models that effectively integrate technology into instruction. Teachers’ use of the internet for specific purposes (e.g. accessing teacher guides for structured pedagogy, as in Gray-Lobe et al. 2022) may also be effective. In contrast, largely unsupervised internet use, especially when not restricted to a single educational website or platform as in Derksen et al. (2022), seems more likely to have negative effects. Given the rapid global expansion of mobile phone networks and smartphones, this is an important area for further research that links changes in learning outcomes to not just internet access but also to detailed patterns and information on student use.

For full reference list see the end of the conclusion chapter.