Between 2022-23 and 2025-26 alone, the share of students in government schools fell from 54% to 48%, while private unaided schools rose from 33% to 40%. Yet this shift in enrollment does not, by itself, tell us where children learn better — or why. The evidence complicates the public-versus-private debate: differences in learning outcomes cannot be explained by ownership alone.
In 2024, 35.5% of rural Class III children in private schools could read a Class II text, against 23.4% in government schools. However, since 2022, government schools have gained seven percentage points, while private schools have gained just over two in the same assessments.
A voucher lottery in Andhra Pradesh found that winning a private-school place made no difference in Telugu or maths scores after four years. However, overall scores did improve, largely driven by gains in other subjects that private schools devoted more instructional time to. Private schools achieved comparable outcomes in Telugu and Maths while spending less than half their instructional time on these subjects (government schools spent 70%), and at a fraction of the per-pupil cost. In fact, 70% of private-school students in the country pay less than ₹1,000 a month in fees.
Rather than asking whether public or private schools are better, there is merit in exploring the accountability structures in each sector: who a school answers to, and for what. Here the sectors may have more in common than one might think. Government schools are held to account mainly upward, by the state, though parents also have a voice through School Management Committees (SMCs) and other public channels. Private schools are held to account mainly by parents who can leave, but they also answer to the state, which regulates their recognition, infrastructure, staffing and fees.
Both operate in systems that collect information, conduct inspections and demand planning. Yet these processes do not necessarily make schools accountable for learning. The more important question is: what would it take for accountability to change what happens in the classroom?
Why accountability misses learning
Accountability works when three things align around the same outcome.
In the case of learning, a school needs information that shows whether children are learning, someone with the authority and capacity to act on it, and a reason to act. Without authority, information is just reporting. Without information, authority is guesswork. And without a reason to prioritise learning, neither is likely to change behaviour.
The Research on Improving Systems of Education (RISE), a seven-year, seven-country research programme, offers a useful lens into why education systems often fail to achieve learning outcomes. Lant Pritchett, its research director, argued that education systems fail when they are incoherent: when people within the system are held accountable for different goals. A district measured on enrollment, a principal on compliance, a school owner on fee collection, and a politician on new school buildings can all succeed while children are not learning.
This goes beyond public versus private, because both systems share blind spots in how they are held accountable for learning. Two important forces shape accountability in both sectors: the state and parents. Yet neither consistently has the information or incentives to make learning the central measure of school performance.
The first is the state. As their employer, it asks government schools for fund utilisation reports, improvement plans, and inspection records. It asks private schools, as their regulator, for recognition, infrastructure, staffing norms and inspection records. As one report on the private school sector puts it, the state "over-regulates inputs and entry and under-regulates the actual outcomes".
The second is parents. They may judge private schools by visible markers such as English-medium instruction, uniforms, buildings and the number of board-exam toppers; in government schools, they may focus on whether entitlements arrive on time — mid-day meals, free textbooks, uniforms and scholarships — and whether teachers turn up. These are not unimportant concerns. But they do not necessarily tell parents whether children are learning.
Evidence suggests that, irrespective of the sector, communication between schools and families tends to follow the state's lead and remain compliance-driven. A 2022 study of government, low-cost private, and NGO-managed schools in Rajasthan and Bihar found that communication channels with parents largely focused on enrollment, attendance, and monitoring, while a substantive focus on learning was "missing everywhere".
When measurement becomes paperwork
Let’s look at an example of what happens when information arrives without the authority or incentive to act. Karthik Muralidharan and Abhijeet Singh evaluated a school-quality programme across 5,435 schools in Madhya Pradesh, of which 1,774 received the intervention, between 2015 and 2016. By the end of the programme, 93% of schools had completed high-quality assessments, participated in school-rating inspections and written improvement plans. However, the intervention did not lead to changes in supervision or teacher pedagogy, nor did it improve student learning.
The programme had generated information and completed its processes, but officials were not required or sufficiently enabled to use that information to change what happened in schools. The measure of success remained paperwork submitted on time.
Block education officers studied by Yamini Aiyar and Shrayana Bhattacharya described themselves as “disempowered cogs”, passing orders down and sending information up, often viewing programme delivery as one more report to file. In the school-quality programme, therefore, assessments and school ratings provided useful information, but no one close to the school or classroom had the authority, support or reason to act on it. Learning did not improve, and school functioning remained unchanged.
Yet the state viewed the programme as a success and scaled it nationally to six lakh schools. It is a telling illustration of how a system can mistake the successful measurement of schools for the improvement of schools.
When information reaches someone who can act (and has a reason to)
When information is provided to people who have a lever to act on it, it can shift behaviour. In rural Pakistan, Tahir Andrabi, Jishnu Das and Asim Khwaja gave families report cards showing their child’s test scores alongside the scores of every school in the village. Parents could now compare schools on learning rather than just on ‘visible markers’. Learning in schools that received report cards rose, on average, by 0.11 standard deviations (over 42% more than the gain in comparison villages). Weak private schools improved or closed, stronger ones cut fees (compared to control schools), and public schools in the area improved. The information was independent and comparable, and in the hands of people who had both the power and the incentive to act on it.
However, information alone is not enough. In Jaunpur, Uttar Pradesh, only informing communities about village education committees or helping them test their own children raised awareness but not their engagement with the school or the children’s learning. Parents in Pakistan could switch schools, while parents in Jaunpur had fewer practical options to act on the information.
An evaluation of the journey of the Teaching at the Right Level (TaRL) programme as it was integrated into government schools further illustrates the link between authority, information and incentive. TaRL, developed by Pratham, begins with a simple test of each child and groups children by learning level rather than grade. How well it worked depended not just on the teachers, but on what the system around them enabled and expected them to do. In Bihar’s 2008 summer camps, where implementing TaRL was their only task, for the children who attended, teachers raised language and maths scores by about 0.4 standard deviations; because only about a quarter attended, average gains were more modest between 0.07-0.09 standard deviations. During the school year, the same approach largely failed, except in schools with volunteer support. Teachers had been trained and valued the materials, yet at most one in ten classes was grouped by level, because they “also have to complete the syllabus”.
In 2012-13, Haryana changed all three conditions. On authority, the state mandated a dedicated extra hour, so teachers did not have to choose between completing the syllabus and addressing children’s learning needs. On information, a baseline assessment told each teacher where every child stood. Block-level coordinators, trained by Pratham and placed in schools, trained, mentored and monitored teachers. On incentive, the mandate “sent a signal that the intervention was government-mandated”; and because the dedicated hour was visible, implementation could be monitored. Over nine in ten schools grouped children by level, and Hindi reading scores rose by 0.15 standard deviations more than the comparison schools.
Madhya Pradesh measured schools without improving them, and in Bihar, teachers attended the training but largely went on teaching as before. Haryana used mandates, supervisors and monitoring too, but aligned them with learning: protected time gave teachers room to act, mentors provided support and information about their classrooms, and the government mandate signalled that learning was a priority. What Haryana got right was not more accountability, but accountability oriented towards learning.
What it would take
With students almost equally divided between private and government schools in the country, the more important question is not which sector performs better, but how we hold schools accountable for learning, irrespective of ownership or management.
Two possible solutions could help bridge the gap between intent and action.
First, while a comprehensive school-rating programme like Ofsted in the UK may be an eventual aspiration, a starting point would be low-stakes, independent census assessments and public reporting of results across all schools. Most large-scale assessments in India currently sample schools and are designed for policy-level decision-making, rather than school-level use.
To be credible, it would need to be run by an independent body; NEP 2020 already proposes one: a State School Standards Authority (SSSA). This would enable parents and policymakers to compare schools across sectors on the basis of learning outcomes. NEP 2020 also endorses such assessments, recommending key stage assessments (for grades 3, 5 and 8) to assess the health of the education system while remaining diagnostic and low-stakes. Keeping assessments low-key would prevent schools from being incentivised to teach narrowly to the test or exclude children who are struggling. Organisations such as CSF are already working towards such reform in Rajasthan and Gujarat (for key stage assessments) and Assam and Uttar Pradesh (for SSSA); we believe more state governments should take heed.
Second, information must lead to support and incentivised action, rather than become another target to chase. For government schools, this would mean results trigger support before teachers are “pulled up”, as in Haryana, where teachers were supported by mentors when children fell behind, alongside protected teaching time. School recognition should also take account of measured progress rather than absolute scores, recognising that schools serve students with different starting points and circumstances. Further, embedding incentives into the design is key; performance-linked bonuses have been shown to affect student outcomes in India and Tanzania. These incentives must be linked to independent third-party assessments rather than the census assessments, thereby keeping them low-stakes. For private schools, the levers exist: recognition under the RTE Act, state fee regulation laws and the SSSA. Regulators could shift emphasis from inputs towards requiring participation in independent assessment and disclosure of comparable learning outcomes, enabling parents to make more informed choices, as the report cards did in Pakistan.
The test for a learning-oriented accountability system is simple to state but hard to pass: when a child is not learning, does someone close to that child know, have the power to respond, and have a reason to act?