The 2020 Budget allots around 3% of our GDP to education; independent of whether this allocation suffices, in real terms, this is a hard number to visualize: a whopping Rs. 99,300 crores (13.56 billion USD)—of which schools get 59,845 crores. Meanwhile, in 2016, we spent 80% of our education budget on teachers’ salaries. That means that most of our monies get spent on revenue expenditure—funding the schools that exist. That also means that a 1% increase in efficiency—that is, if for the same inputs, achieving 1% better outcomes—saves ~599 crores. And because money saved is money created, making sure that we get the maximal bang for the buck on spending has been a focus of public finance. When we contextualize this in education delivery, a potentially powerful tool to achieve more with less is school inspections.
Education being a concurrent subject in India, while the center regulates education, the frontline delivery of service largely rests upon states. Ultimately, with many types of schools run by different sources, this results in significant diversity in the model of inspecting schools. Generally, however, inspections come into the jurisdiction of the District Education Officer and the Assistant Educational Officers (their names vary by state); the system usually has one mandated annual inspection, likely one mandated surprise visit, and encourages as many regular surprise visits as possible. The focus of visits is typically primarily on evaluating the infrastructure and sanitation of the school although they usually also mandate some examination of students’ academic levels and teacher instruction and pedagogy.
Largely speaking, to an outsider like me, it appears like our school inspection system is broken. How do we do good better?
An important principle in political economy is to design policy by viewing the different actors as strategic players maximizing their self-interest in a game—this is an important approach as the difference between imagined policy ex ante and implemented policy ex post often comes from the incentives of those implementing policy diverging from the “ideal” incentives. This is a lesson for reformers to design policy with the sophisticated expectation that frontline actors are acting in self-interest (not to ignore that for many, self interest might rest entirely or partially upon doing good).
Further, the usual (and a very sensible) formulation of this incentive mismatch is in the form of a principal-agent problem. The principal—the citizenry—entrusts the political executive and top bureaucrats—the first level agent—who then entrust the frontline official—the ultimate agent—with service delivery. The underlying issue with this delegation is that the principal cannot observe the work of the agent, they merely observe the outcomes of the actions. That is, the work that needs to be done is un-observable. A particular confound in the context of government and, particularly, our government with its poor administrative data and lack of evaluation of public services is that not even outcomes are observable, only monetary inputs are. So, while a restaurant owner can at least observe her manager’s sales if not the manager’s customer service, the citizenry doesn’t have high quality information on the outcomes of, say, education efforts. However, the budgetary inputs are announced and observable. This results in a paradoxical state where the outcome is the input—we laud policies that spend more because that is all that is observable to us. Ultimately, neither do we have control on the bureaucrat’s effort nor do we know whether this bore fruits in terms of, say, foundational literacy and numeracy levels, dropout rates, or reading levels.
In this context, the inspection of schools induces observability into the system; it actually tells us how well government schools are using their limited resources and also how well private schools are run. But, all this, only if the inspection works…
I’ve first hand seen teachers transform into pedagogy gods when inspectors watch—new teaching aids appear, digital tools are used, and a previously empty blackboard is quickly filled up by interactive exercises. Imagine if teachers had seven days notice to the important annual inspection? That would hardly measure a “routine working day”, right? Well, don’t imagine, this is the system in, for instance, Kerala.
A worse outcome would be corruption: if you knew who your inspector was and when they will show up at your gates, as a headmaster, you are well equipped to bribe your way into a good report.
Three things can be done to structurally restrict the ability to bribe your inspector. First, for the mandated annual inspection that has the power to issue warnings, recommendations, or take more serious action, standardize panel inspections (like Delhi does) where a panel of the inspector and randomly selected local headmasters and principals inspect schools. Where it is geographically viable, select these headmasters and principals from other zones in the district that the AEO (the inspector for each zone) does not oversee. Second, introduce day-of intimation for annual inspections—this only serves the purpose of ensuring that that day is a routine working day for the inspected school. Third, for non-mandated surprise visits, randomly assign inspectors to schools. The goal-setting system (reform #3) allows inspectors to commit to visits and use a management information system to see who they’re visiting in the commited visit just before they leave their office. Further, once we introduce accuracy scores for inspectors (as proposed in reform #4) and have quantitative measures of school performance (from reform #2), we can do much more with this random assignment like assigning visits to poor-performing schools more frequently and assigning high trust score inspectors to low score schools.
Side note: there might be value in inspector discretion, but many systems like that in Delhi already have higher level bureaucrats assigning inspectors to the mandated inspection on the day of inspections.
Measurement has three issues right now: first, it focuses on infrastructure over academics (I speculate, a part of the hangover of stage 1 of independent India’s education effort—ensuring access before being able to think of quality); second, it is entirely qualitative and subjective to the frontline inspector enabling corruption; and third, the report once submitted and shared with the school is allowed to collect dust without use (for many states, the only information available pre-visit seems to be last date of inspection and the corresponding inspector).
It follows then that the measurement system must do three things: first, increase focus on examination of students’ levels and teachers’ instruction; second, create clear quantitative scores for different categories like student levels, instruction, teacher and student attendance, dropout rates, infrastructure, sanitation, etc. and compute an index summarizing school scores (Kendriya Vidyalayas do something similar); and, third, explicitly measure improvement in these categories vis a vis past years.
This set of reforms ensure that what is being measured is indeed what matters for students and ensures that inspectors enter inspections with rich records of past inspections. These inform what they should spend time examining and allows them to see and reward improvement. To further enhance this, inspectors could be given readymade tests that include, say, scientifically approved reading level tests to use and be given a MIS with searchable, high quality data from past reports.
The state delegates its duty of care over minor citizens to parents precisely because (in the majority of instances) parents are actors who care about their children and are (at least relatively) well informed about their children’s lives. Soliciting parental feedback and complaints is an easy way to build state capacity without doing much. Allow the end-user to supplement reports with their high quality information on their experiences with the real routine working day.
Kremer et al. (2005) find that 25% of Indian public school teachers are absent on any given day and less than half of them actually teach even if they attend class. Remember, 80% of our budget goes to paying teachers. Presumably then, surprise visits are an important check on teacher absenteeism. Unfortunately, not in the real world. There are many anecdotal accounts on record of teachers and administrators that testify that surprise visits are a joke—they are reported at the whim of the inspector and very rarely does a visit that occurred on paper actually occur.
If we are to exploit surprise visits, we must first make each officer set goals for the number of visits they will complete in a year ex ante. This way, we can include their visits in their performance appraisals (re: reform #5). But, we also need to ensure that the visit actually occurs—perhaps here we should exploit technology by creating tablets or apps that lodge surprise visits and ensure same-day report submission. This tech tool could require photographs of the visit and could record the geo-coordinates of the inspector from the time of the visit till the time that it ends. As discussed, a more powerful solution would allow inspectors to commit to a visit at a certain date-time and then only tell them who is being visited just before. An even more powerful one would randomly assign inspectors to schools because knowing your inspector enables corruption and assign high score inspectors to poorer-performing schools.
While inspectors can help mitigate the principal-agent problem in education delivery, ironically, inspectors are also agents who induce the same principal-agent problem. So, appraising their performance, tracking their efforts as much as possible, and rewarding better inspectors, say, via pay and promotions, is important. This furthers rewarding good, honest work and reduces the relative seductiveness of corruption. Appraisal could also track the zone that the inspector operates in and reward them for actual outcome improvements. Further, backchecks (where a smaller subsample of the total sample are re-tested to determine the accuracy of the first test) could be used to generate scores of accuracy of inspectors.
Quantitative measurement does not mean that teachers’ creativity and freedom should be shackled. Moreover, an explicit purpose of visits is to give teachers feedback on their pedagogy. So, job aids should be created for inspectors to be able to give better feedback for teachers. Multi-day visits that focus on visiting each teacher’s class and giving them detailed feedback should be encouraged and rewarded in the performance appraisal system. Many states already require somewhat regular meetings between education officers and local headmasters and principals: these should also be tracked and enforced as in reform #4, and inspectors should be given pedagogy materials and external support as viable to do more with these meetings.
To defeat the principal-agent problem in education delivery, inspections could be exploited. However, to defeat the principal-agent problem in education inspections, we need to introduce new extrinsic incentives like pay and promotion based on actual performance, restrict the possibilities of corruption and collision by reducing notice time, randomizing inspectors, and tracking the trust levels of inspectors, and improve measurement by talking to parents, changing what and how we measure, and using measurement to do more. Some of these require robust technological tools.
Ultimately, as we improve our data on schools and understand which schools work, we can then figure out why. That way, we evaluate policies, compare them with others, and constantly strive to do more with the same resources. The data is also an end by itself—some computed indexes on school scores could be released publicly to inform parents for school choice. The stakes are large here.
*This post is inspired by the thinking of Dr. Santhosh Mathew, Country Lead of Public Policy and Finance at Gates Foundation India, whose state capacity approach I have been working on under him.
Git: 
Twitter: