Impact investing in education: Delivering financial and social returns for investors

Jishnu Das and Asim I. Khwaja joined practitioners and impact investors to discuss their research in this VoxDev webinar: Helping schools to thrive: Impact investment in global education.

The fact that low-cost private schools are a major provider of primary education around the world is well known. In Pakistan, where we work, they now account for 45% of all primary enrolments.

But two things are less well known. 

First, private schools close frequently, disrupting both the education of children and the livelihoods of owners and teachers. Our best estimates (also from Pakistan) suggest that if all private schools in a village close, the loss in welfare is equivalent to 2% to 7% of GDP per capita (Carneiro et al. 2024).

Second, we know little about how to address these costly closures and the broader constraints that prevent schools from thriving. Despite considerable progress, we still do not have answers to key policy questions such as the extent and implications of credit constraints for private schools. In fact, the IFC thought in 2017 that, as an investment opportunity, the private education sector was a no-go. As they pointed out, "the overall IFC equity return on its investments in K–12 education over a 10-year period ending June 2021 was negative," a result that the IFC believed was because “the small size and relative business immaturity of many K–12 private schools, particularly low-fee private schools, inhibit scalability."(World Bank 2022)

Our new research provides the first causal estimates of impact investments in private sector schooling (Andrabi, Das, Khwaja and Ozyurt 2025). We start with the twin observations that (a) investments need to be commercially viable even when school closures are frequent and (b) investments that affect school survival could generate sizeable social returns. We then conduct a large RCT designed to answer three questions:

1. Can interventions produce financial returns for impact investors without public sector financing even in a context with substantial school closures?
2. Do these interventions affect school survival?
3. Do different interventions have differing impacts on the types of schools that survive? 

The short answers, elaborated below, are as follows:

1. Two interventions—one cash and one ‘in-kind’— were commercially viable without any support from the public sector.
2. Both interventions helped schools survive.
3. Interventions that provide cash versus in-kind products had different impacts on which schools survive. This mirrors a literature on the importance of fungibility in interventions with SMEs.

Studying impact investment in the education sector

In 2015, we approached 3,784 schools in Pakistan and asked if they were interested in receiving a loan. Eventually, 815 schools said they were, and we experimentally allocated them to three arms: The “Finance Arm,” the “Education Products and Services” or EPS arm, and a control arm.

In the finance arm, schools were invited to apply for a loan with our partner microfinance institution, the Tameer Bank. For the EPS arm, we convened seven providers and brought them together in a mela or trade fair. Schools were then invited to attend the mela and purchase a product, if they wished. The most popular product was low-cost (but high quality) textbooks and the accompanying training, both developed and delivered by Oxford University Press. We chose finance and EPS as our two main products because our previous work demonstrated that schools lacked access to finance and pedagogic products, a characteristic of the low-cost private school sector that we reconfirmed in our baseline surveys (Andrabi and others, 2020).

What are the impacts of impact investment in education?

Four years after we started our experiment, we can first confirm that our products for private schools, financed and delivered entirely by the private sector, are commercially viable. Take-up in the finance and EPS arms was 36% and 27%. Three years after the intervention, the investments yielded an ARR of 12.8%. EPS products were purchased on site and schools paid the price that was set to be commercially viable by the provider. In the case of loans, the only criteria for receiving a loan were (a) the school’s stated interest in a loan product and (b) the usual processes of the MFI. The EPS products had a similarly simple selection process. Despite this light-touch due diligence, our returns are comparable to some of the best projects financed by development finance corporations such as the IFC.

Second, both products reduced school closures (and therefore increased school survival). Over the four years of our experiment, 32% of all schools in the control group closed down. In intent-to-treat regressions, the finance and EPS arms are estimated to reduce closures by 8 to 9 percentage points; accounting for imperfect take-up using an instrumental variables specification increases these estimates to 20 (finance) and 33 (EPS) percentage points – essentially access to either of these two products all but eliminated school closures. 

Third, the types of schools that were protected by these products are very different. We examined heterogeneity in the size of the impact by the baseline size of the school as well as the baseline performance in terms of test scores in English, math, and Urdu. The choice of these baseline attributes was guided by our previous work showing that smaller schools and those with lower test scores exit the market sooner (Andrabi et al. 2025). Figure 1 shows these differential impacts using non-parametric bin scatters and the analogous regression estimates show that there was no heterogeneity of the impact by size or baseline test scores for the EPS treatment arm. In sharp contrast, the finance treatment reduced exit rates by 1.8 percentage points for schools at the 20th percentile of the size distribution, but by 41 percentage points for schools at the 80th percentile of school size. Similarly, it reduces exits by 30 percentage points for schools at the 20th percentile of baseline test scores compared to 9 percentage points for schools at the 80th percentile of the test score distribution.

Figure 1: Investing in private Schools decreases closures—but the types of schools that are helped depends on the specific investment

Notes: This figure shows binscatter plots by treatment and school type. Each plot groups the school type variable into equal-sized bins, calculates the mean of the school type and closure within each bin, and presents these data as a scatterplot. Additionally, each plot includes a linear fit by treatment.

Implications for the private sector’s role in the education sector in developing countries

With millions of children enrolled in low-cost private schools around the world, improving outcomes in these schools remains an urgent priority. We have been very interested in understanding the types of interventions that can be carried out entirely by the private sector, as a counterpart to more standard approaches of public financing for private provision. However, these kinds of impact investments in education have been held back by a lack of fundamental research demonstrating that such investments can generate commercially viable returns while still producing social returns. Seen in this light, the biggest contribution of our paper is to show that impact investments in education can generate both financial and social impact at the same time.

If impact investing is to grow, as we think it should, an entire programme of evaluation now needs to be put in place. Our understanding of this sector is that the demand for such evaluations is starting to grow, but it will need nurturing. A consortium of investors who drive evaluations for key projects could yield knowledge products with potentially massive returns for multiple organizations operating in this space.

Lessons on evaluating the impact of investments in education

Two implications of our work are worth highlighting for how such evaluations should be structured.

First, the space of products is large—and the impacts of these products on schooling outcomes will differ both by the product and the school. In our previous work, we showed that unconditional cash grants to private schools increased test scores, but only when given to all schools in the village (Andrabi et al. 2020). Thus, investments in the entire schooling market versus investments in specific schools generated different social returns. Here, we showed that the fungibility of the investment matters. This is because loans, like cash, are fully fungible, but EPS can be monetized only through greater demand, either by enrolling more children or raising fees. Our accompanying theory shows that it is this critical difference that gives rise to the patterns of differential impact that we observe in our data. An implication is that, at least for the time being, there is not “one” winning product in this space; rather, we need to encourage experimentation across a range of products and contexts.

Second, market structure matters. The impact of interventions will vary based on the nature of underlying competition and entry and exit constraints. In related work on the impact of investments in public schooling we show that such investments not only raises quality for public schools, but also private schools that are in competition with such schools – public investments can crowd-in private quality (Andrabi et al. 2024). This, however, is not a universal result and, indeed, evidence from the US showed that large investments in public schools can also lead to the closures of private schools, rather than an increase in their quality. The key point is that understanding the underlying structure of the market can help design educational investments and interventions that maximise the aggregate (social and financial) impact. Smarter design can pick whether our investments result in a positive or negative multiplier effect. 

Our research takes an important step towards demonstrating that impact investments in education consistently improves the lives of children, while generating financial returns for investors. How to do so efficiently and consistently must now become the aim of a larger program on this topic.

References

Andrabi, T, J Das, A I Khwaja, S Ozyurt, and N Singh (2020), “Upping the ante: The equilibrium effects of unconditional grants to private schools,” American Economic Review, 110(10): 3315–3349.

Andrabi, T, N Bau, J Das, and A I Khwaja (2025). "Heterogeneity in School Value Added and the Private Premium." American Economic Review, 115(1): 147-182.

Andrabi, T, N Bau, J Das, N Karachiwalla, and A I Khwaja (2024), "Crowding in private quality: The equilibrium effects of public spending in education." The Quarterly Journal of Economics 139(4): 2525-2577.

Andrabi, T, J Das, A I Khwaja and S Ozyurt (2025), "Helping Schools Survive: Experimental evidence on the impact of Financial and Educational Support to Private Schools". Working Paper.

Carneiro, P, J Das, and H Reis (2024), “The value of private schools: Evidence from Pakistan,” Review of Economics and Statistics, 106(5): 1301–1318.

Dinerstein, M, and T D Smith (2021), "Quantifying the supply response of private schools to public policies." American Economic Review 111(10): 3376-3417.

World Bank (2022), An evaluation of International Finance Corporation investments in K-12 private schools, Independent Evaluation Group, World Bank.