Artificial Intelligence reshaping the role of nation-states in international arena

Top Artificial Intelligence (AI) researchers agreed on a set of 23 principles, in a recent conference. This is to contain the potential damages of AI. The principles include - value alignment, avoiding lethal AI arms race etc.

This is a significant moment in human history.

Till now, nation-states were the major players in the international arena. Governments of nation-states represented people in international forums and signed agreements on behalf of people. Individuals weren't important. This is more importantly so in case of activities that pose a threat to humanity.

For instance, major lethal weapon as of today, the nuclear weapon is in hands of only nation-states. Non-proliferation agreements were signed by the governments of nation-states.

Artificial Intelligence (AI) has changed this phenomenon. AI is now in the hands of individuals and not governments. AI also poses significant threats to humanity. In other words, AI weapons are decentralised nuclear weapons with no role of nation states. So, probably for the first time, the individuals are signing agreements to adhere by certain principles, bypassing the governments.

This is going to have interesting implications. 

One, nation-states were few, they could easily (relatively) come to an agreement. There are far too many individuals involved in case of AI. As AI becomes ubiquitous it's to be seen how the procedure of agreements change.

Two, any violation of nuclear agreements between countries could be punished by imposing sanctions against that particular country. It may not work in case of AI where individuals are involved.

Three, countries may also invest and acquire AI weapons. Countries may also sign agreements, parallel to individuals. Is this going to cause conflicts between networked-citizens and networked-countries?

Overall, the emergence of AI has potential to significantly reshape individual-individual relations bypassing nation-states, nation state - nation state relations and individual - nationstate relations. Exciting times ahead!

Land redistribution to accelerate poverty reduction

Growth reduces poverty but the amount of poverty reduction per unit growth also matters. The percentage reduction in poverty with 1% increase in growth rate (mean income) is called growth elasticity of poverty.

I blogged earlier (here and here) on the differences in growth elasticity of poverty between India and China, and lessons from Chinese experience.

In short, poverty reduction per unit growth is higher in China than in India. The evidence as of now suggests that initial inequalities determine the poverty reduction potential of growth. China bridged initial inequalities by strictly pursuing land redistribution policy that accelerated its poverty reduction.

So, the lesson from Chinese experience is that we need to bridge initial inequalities to increase the poverty reduction potential of growth.

In Indian context, the high growth states like Maharastra have low-growth elasticity of poverty reduction. On the contrary, states like West Bengal have highest growth elasticity of poverty reduction but unfortunately it has low growth rates.

If the mechanism of initial land inequalities hurting poverty reduction potential is true, we should observe that the states that actively bridged these gaps should have high growth elasticity of poverty.

IndiaSpend has a good story on land redistribution in India that collated good data. It notes that land redistribution in West Bengal till now constitutes 52.4% of total land redistributed in India. Not surprisingly, West Bengal also has the highest growth elasticity of poverty reduction among all other states. It thus supports the hypothesis that reducing initial equalities through land redistribution enhances poverty reduction, as is observed in China.

Overall, the land redistribution scenario, as per land redistribution act is disappointing. IndiaSpend notes that 101.4 million (10 crore) rural households don't own any land even today, a significant section of them being SCs and STs. Remember that total households in India would be around 25 crore and total rural households around 20 crore.

The IndiaSpend report notes judicial delays as one of the important reasons. Some states have set up land tribunals to fasten the process. Without having to mention, it needs political will. Considering that land redistribution is also electorally attractive, it's surprising that no party has pursued this except communists.


***

Follow on Facebook: @iterativeadapt
Follow on Twitter: @iterate_adapt
Email subscription or RSS Feed: Enter id in the "Subscribe" text box, on the top right of the blog.

De-privatising schools - Demonetisation equivalent reform in school education

Noises of demonetisation proposals since past few years were ignored. It was thought that such proposals don't merit any attention. Finally, it ended up happening. After it happened, people are now finding post-facto justifications. 

May be if we had paid enough attention to these early noises and snubbed it off in the initial stages itself, we wouldn't have had reached this stage.

In this context, may be it's worthwhile to think of other similar ideas floating around, and discuss them in detail now itself, so that we prevent such situations in future.

In the context of school education, there's one reform idea floating around since long. The proposal is to ban all private schools (de-privatise?) and force everyone to attend public schools. This proposal is also gaining ground slowly. Recently, Allahabad High Court ordered IAS officers and public representatives to send their children to public schools.

Like demonetisation, de-privatisation of schools has potential to cause huge disruption and there are arguments on its benefits. This post lays out the arguments from both sides, just to structure the debate.

Note that school education is supposed to be not-for-profit in India. So, the world private has a different meaning here - it means non-government ownership of schools.

De-privatising school education


The reasoning for de-privatising school education is that lack of accountability is the reason for dysfunctional public schools. Part of the reason is because the elite who have the 'voice' have shifted to private schools, weakening accountability structures. 

The fact that government teachers don't send their kids to the very school in which they work, is counter intuitive. It indicates their trust on the their own schools and performance. Imagine a chef noting eating the food cooked by himself/herself because s/he knows that its quality is poor.

It follows from this reasoning that if we were to bring all these people back to public schools, a major problem would be solved. If a child of an MLA or an IAS officer studies in a public school, imagine its functioning! It's similar to the chef example. If the chefs are forced to eat only the food cooked by themselves, it would force them to cook better food.

Famous economist Albert Hirschman gave an early theoretical framework to such proposal using what's called the "Exit-Voice-Loyalty" framework.  In simple words, people react differently to dysfunctional organisations, of which they are part of. Some people simply "exit" in search of better avenues. Some people stay within the organisation but "voice" their concerns. In some cases people turn "loyal", hoping that their loyalty will be rewarded.

This simple framework has good explanatory power to understand aspects like emigration, functioning of business organisations etc. In context of public school education, Hirschman argued that massive "exit" reduces accountability in important areas like school education. In such contexts, monopoly of public institutions is preferable because it preserves the accountability structures.

Will de-privatisation necessarily work?


There are three obvious arguments against the de-privatisation proposal.

One, there is a normative argument that de-privatising education impinges upon the liberty of citizens to educate their children in the way they prefer.

Two, the timing of implementation of this policy. It's one thing to pursue the 'only public schools - no private schools' path, if we are building a school system from scratch on a clean slate. It's a different ball game when there's already a significant section of private schools existing in the system, and we are re-setting the system to go back to 'only public schools- no private schools' stage. There are huge disruption costs involved in the latter situation.

Three, de-privatisation reduces the diversity in schooling systems. There's no 'one good method of education'. Hence, the diversity is preferable. De-privatisation hurts innovation by enforcing an artificial homogenous system.

Beyond these obvious arguments, there are deeper reasons that lead to skepticism about the effectiveness of such proposal.

One, the de-privatisation proposal presumes that public school system is a tightly-knit network and that accountability measures in one school transmit equally to other schools. It may not be necessarily true. This has implications for the effectiveness of the 'only public schools' policy because it is likely that the elites are going to be concentrated in few schools and may not be present in all schools. It's for two reasons.

a) Most elite (IAS, public representatives) are concentrated in urban areas. It thus excludes a vast majority of schools where there won't be any students of elites, especially those in rural areas.

b) Even within urban areas, there are also geographical constraints to education market - one would not send children to very far away schools. Given that housing in urban areas is segregated as per socio-economic status, it may still end up happening that schools in high-income localities are likely to have only children from those backgrounds and schools in low-income communities may have students of only those profiles.

If the 'de-privatisation' policy is to be effective, accountability in high income localities has to transmit to those in low-income localities. This may not happen.

Unlike other sectors, outcomes at local level are as important as aggregate outcomes in school education. It means that outcomes at each school level are as important as the aggregate outcomes. The higher quality in one school can't be a compensation for the poor quality in another school.

Two, even within schools that elite's children study, schools may respond differently as per profiles of children. It's well known that the socio-economic profile of students play a role in their outcomes. It thus calls for a different approach to engage these children. It often involves higher effort.

In responding to elites, schools may just focus on low-effort way of ensuring outcomes of elites, while ignoring the high-effort process required to engage children of non-elite parents.

Three, a mere increase in accountability to public may not yield results. Increasing school's accountability to public as 'the' solution presumes that teachers already know what's to be done and they are deliberately not doing it. Hence, increasing accountability will make them do those things. But, it needn't necessarily be true. Teachers may not know the right thing to do, they require training and our training systems are dysfunctional. In such contexts, merely increasing accountability may not help.

However, one can still counter argue that ensuring accountability is the first major step and the rest steps follow automatically. For instance, elite pressure will force school systems to restructure their teacher training system to address parents' needs and so on. Such optimistic view needn't turn out to be so in reality.

For instance, in this famous study, parents were given information on low-learning levels of children. Also, village education councils were made to meet regularly to ensure accountability mechanisms. The hypothesis was that such improvement in accountability would yield outcomes. But the study found that it didn't necessarily result in positive outcomes. It highlights the limitations of accountability through public participation.


Conclusion


Overall, there is a strong normative and good theoretical basis to de-privatise schools, forcing everyone to study in public schools. At the same time, there are also strong normative arguments and those of costs on the other side of the argument. 

Further, even if we ignore these normative arguments and costs involved in transition, there are reasons to be skeptical of effectiveness of 'all-public school' systems - elites' children may end up occupying only a handful of schools and the accountability enforced in these schools may not transmit itself automatically to other schools, schools may respond only to students of elites within school ignoring other children and there may be limitations to merely increasing accountability to public.

Even considering the limitations of the de-privatisation approach, some may still argue that 'some thing is better than nothing' - it may not bring change in all schools but it can bring change in at least some schools. Given that it's crucial to middle class and elite, this issue will always be in public discourse, unlike the current situation, and such increased attention can transmit into some action even in schools where elites' children don't study.

But will it happen and at what cost? School systems transformation takes a long time. Should children studying in otherwise functioning schools suffer in this transition? A decade is a generation of schooling.

There may be other genuine reasons for and against de-privatisation of schools and pursue 'only public schools' approach. In the context of growing popularity and appeal of this proposal, it is important to structure these debates and analyse the different possibilities now itself, rather than trying to do damage control after such decisions are taken or after they reach a popularity beyond control.


***

Follow on Facebook: @iterativeadapt
Follow on Twitter: @iterate_adapt
Email subscription or RSS Feed: Enter id in the "Subscribe" text box, on the top right of the blog.

Modest returns to village electrification in India

It's commonly believed that increasing access to energy will have multiplier effects, especially in rural areas - kids can study longer, women save time on household chores giving them bandwidth to be engaged in other productive activities, households can start small businesses etc.

A new paper by researchers of University of California, Berkeley finds that the returns to electrifying villages in India have been modest and not as much as expected, at least in medium-term. 4 lakh villages were considered in this study.

The paper further says that the results hold up in areas with high  reliability of electricity as well those with low reliability (number of hours of supply) of electricity. Results also hold up for districts where villages were electrified in the beginning and those who where it was done later, thus ruling out any bias.

It doesn't mean that villages shouldn't be electrified nor does it mean that the quality of electricity shouldn't be improved. The fact that increasing access to energy has given good returns in other developing countries but not in India suggests that we may have not have leveraged the advantage of increased energy access, by using it to increase productivity. In other words, we may have fallen short of ensuring other prerequisites that make energy access productive.

We must also note that the quality of electricity supply (reliability) is low in rural India. A survey of six major Indian states finds that villages receive only 12.5 hours of electricity per day on an average and there is a blackout of 3.6 days per month.

There are lot of challenges in improving the quality of supply - distribution issues, cost recovery etc. Smart grids is increasingly being proposed as a solution. I blogged earlier that we need to be careful about this too as the reality is different and they aren't working out in many places.

Seeing all this together, it seems that one can't escape from improving quality of grid electricity. In order to leverage this, we must also focus on other prerequisites like connectivity, skills etc. In case of energy, it calls for commitment from governments, with huge investments in infrastructure, along with innovation in bill recovery. Behavioural economics serves as a good tool in thinking through these final mile problems.



***

Follow on Facebook: @iterativeadapt
Follow on Twitter: @iterate_adapt
Email subscription or RSS Feed: Enter id in the "Subscribe" text box, on the top right of the blog.

"Authoritarian and Anti-Market traits of Russians" - A thinking fallacy

Gary Kasparov, the chess legend, is outspoken against Putin. In a recent interview on US elections, the interviewer asked Kasparov as to how would he explain huge public support to Putin? Also, the narrative that Russians have authoritarian instincts.

This is not an uncommon question considering the fact that many commentators of international relations often comment that Putin has huge public support from Russians because they have inherently authoritarian instincts and want to see Russia as a super power.

Gary Kasparov in his typical style responds - who told you that Russians support Putin and who told you that Russians have authoritarian inclinations? How can you go by Russian media narrative that is dictated by Russian? He further says that it's "bullshit" to say that some countries can't have democracy because of their character.

It reminded me of Robert Shiller's old column in Project Syndicate on this same question - Is Russia's national character authoritarian?

Robert Shiller says that some made similar argument prior to the fall of Soviet Union. The argument was that Russians don't perceive capitalism as a fair system and hence it would be to pursue capitalism mode of economic policy.

Shiller surveyed attitudes of Russians and Americans towards markets and democracy. The survey was conducted in 1990 and again in 2015 so that one can both compare and observe the trends. 

The survey had questions like "On a holiday, when there is a great demand for flowers, their prices usually go up. Is it fair for flower sellers to raise their prices like this?"; "The press should be protected by the law from persecution by the government"; "It is better to live in a society with strict order than to allow people so much freedom that they can bring destruction to the society". People had to agree or disagree.

Summarizing the results, Shiller says
while there are differences, the results do not lend strong support to the idea that recent events have a simple explanation in terms of differences in deep attitudes toward free markets or authoritarianism. It’s wrong to write Russia off as fundamentally different from the West. In 1991, we concluded that the Russian national character was not an obstacle to creating a market economy in Russia – and were proven right. We hope we are right again, and that national character will not prevent Russia from becoming a truly democratic society someday.

So, Gary Kasparov may have a point. Russians are inherently neither anti-market nor anti-democratic.

This is a good example of one of the fallacies of thinking that attributes "observed difference" to inherent traits of people.

Some examples are: "Many famous chess players are from Russia - Russians have special traits of intelligence that makes them good at chess", "There aren't any girls in top 100 of Chess (only 1) - It's because girls aren't cut out for Chess."

Such reasoning attributes the reason behind "observed differences" only to the individual traits ignoring the role of external factors.

In Russia's case, the long time sustenance of communism in Russia is falsely attributed to character of people ignoring the other external factors. As Shiller shows, Russia had authoritarian governments not because people have authoritarian inclinations, but because of external factors (it's forced on people etc.).

Similarly in case of 'girls in chess', it's a fact that there's an observed difference in players' rankings. But, it doesn't follow from this fact that it's due to some inherent traits of girls. It could also be due to external factors like low number of players, vulnerability to stereotype threat etc.

We should hence be careful of such line of reasoning that attributes "observed differences" (a fact) to inherent traits ignoring external factors.

ASER 2016 - Stagnation of outcomes and enrolment in Indian low-cost private schools?

The much awaited ASER 2016 is now released. Many eagerly awaited ASER 2016 report because it didn't release the report in 2015.

This post discusses the broad findings of ASER 2016, followed by my comments on stagnation of outcomes and enrolment in Indian low-cost private schools. In the end, I also share some of my queries on ASER's methodology regarding the error margin in survey.

ASER 2016's findings on learning outcomes


ASER reports that there's a slight increase in reading levels
The proportion of children in Std III who are able to read at least Std I level text has gone up slightly, from 40.2% in 2014 to 42.5% in 2016.
The same is true for arithmetic too but it's a much smaller increase
In 2014, for the country, 25.4% of Std III children could do a 2-digit subtraction. This number has risen slightly to 27.7% in 2016.
This improvement has come primarily from government schools where the percentage of Std III children who could do a 2-digit subtraction increased from 17.2% in 2014 to 20.2% in 2016

Stagnation of outcomes in Indian low-cost private schools?


ASER 2016 discussed above is only for rural areas. So, we can reasonably assume that a majority of private schools mentioned in the survey are low-cost private schools.

ASER 2016 is the first year that reports increase in arithmetic levels since 2010. Between 2014 and 2016, public schools' arithmetic levels (grade 2 students who can do subtraction) increased from 17.2 to 20.2 as quoted above, while private schools' levels increased only from 43.4 to 44.

Note that the arithmetic levels have fallen from 42.4% in 2007 to 25.4% in 2014.

Similarly, the reading levels (grade 3 students who can read grade 2 text) of public schools has increased from 17.2 to 19.3, while that of private schools has increased from 37.8 to 38 (page 52, ASER 2016)

In other words, we notice that the learning levels in public schools are increasing (both reading and arithmetic), while it has stagnated in private schools.

ASER 2016 also reports that the improvement in arithmetic levels is primarily due to increase in levels of public schools. Result of increased attention to education?

This raises an important question - Have the (low-cost) learning levels in private schools hit a limit?

This is inline with the hypothesis that it is not theoretically possible for low-cost private schools to improve outcomes beyond a point. One can reason it as follows.

Learning outcomes is a function of governance and pedagogy. Public schools lack both governance and pedagogy while the private schools lack pedagogy.

Seen from this perspective, it is argued that low-cost private schools can't improve beyond a point because they already reached maximum levels of governance. Any resulting improvement should be an improvement through pedagogy.

It can be argued that such improvement in pedagogy in low-cost private schools isn't possible due to two reasons. One, teachers in low-cost private schools aren't trained. Low-cost private schools typically recruit grade 10, grade 12 graduates. So, they are less responsive to training. Two, low-cost private schools don't have money to invest in training. 

These limits to governance and pedagogy means that the learning levels (should) stagnate after a point in low-cost private schools.

On the other hand, after fixing the governance issue, government schools can't hit such limits because public school teachers are more responsive to training due to their higher qualifications and pre-job training. Also, government has funds to invest in pedagogy.

If the stagnation in private schools' levels, and the increase in public schools continues as per this year's data, does this support the above argument that low-cost private schools can't improve beyond a point and public schools have more potential to improve?

May be we should wait for 1-2 years more to get a clear picture. Or may be this is just because public schools have a low baseline levels or may be the increments are just a measurement error (more below).

The argument of competition increasing learning levels in low-cost private schools needn't hold true because there's already enough competition in urban areas. A survey in Patna reports that there are 9 and 93 private schools within a 1 km radius of every public school, with the median being greater than 50. If this competition can't result in improvement, then it gives little reason to assume that further competition improves, with other conditions as it is. In rural areas, there may not be great scope for more schools due to limited number of students.

This doesn't mean that one should shut low-cost private schools. One can assist them by training the teachers in these schools. I discussed several possible approaches in an earlier post. For instance, these teachers  can be given Skill India coupons to get training etc.

Since we are struggling a lot to improve governance in public schools, why not utilize low-cost private schools where you at least don't have first order problems like teacher absence etc. Also, the argument that teachers in low-cost private schools can't be trained is too far-fetching. May be they can be trained and may be they will do much better than public school teachers after the training, given their incentives. We have such examples of Shiksha Karmi programme in Rajasthan where 8th grade and 10th grade passed tribal girls were trained and they caught up over a period of time.

We should at least try training teachers in low-cost private schools once. One should not look at it in terms of - why  should government support private schools? One, providing such training has huge externalities. If children outcomes is the one we care about, we shouldn't desist from doing something that could possibly help us achieve that. Two, why not think of this as an investment? Government does give tax incentives to many big industries and corporates. Why not give to low-cost private schools when it may have long-term impact?

In a situation with low-learning levels, we should grab every opportunity that helps us improve the situation. Letting off anything isn't wise.

Stagnation of enrolment in low-cost private schools?


ASER 2016 reports
the proportion of children (age 6-14) enrolled in private schools is almost unchanged at 30.5% in 2016, as compared to 30.8% in 2014
It is commonly believed that there is an increasing shift towards private schools. So, this comes as a surprise.

We may have to wait for at least next year or two more years to say something about this. May be it's just an aberration in this year's data or may be it's the trend. But if it's true, it's a good insight. We need to understand this better. Some possible reasons that I can think of as of now are

1. RTE made it difficult to open new private schools or shut down existing ones, forcing students to join public schools.

2. Quality in public schools has improved, as per parents' perception. So, there is a reverse shift to public schools or it at least stopped the outflow from public schools.

3. We have hit the limit - may be this is the maximum that we can achieve in terms of private school enrolments. The remaining people can't afford to send their kids. 

If this is true, it's also an indication of percolation of benefits of growth. It may be the case that growth hasn't led to increase in incomes of the remaining people (those currently in public schools) and hence they aren't able to afford to shift their kids to private schools. If that's so, then the shift could re-start once people's incomes increase.

Precision of ASER's estimates


There's an error margin in measurement in all surveys and calculations. In crude terms, if the error margin is x%, the results could have been within +/- x% range of the reported figures.

Error margins can turn crucial if one is dealing with numbers less than error margin. For instance, election surveys report error margins of 2% to 3% but the problem is that the election outcomes can drastically change with just 1% difference in vote share.

ASER doesn't report standard errors (error margin) but this year's report quotes an external study 
a study done on the precision of ASER enrolment and learning estimates shows that margins of error are well within 5% at the state level
This just got me thinking - if the error margin is 5% (even if it's 2 or 3), how optimistic should we be of learning level increments that are in the range of 2 to 3%? 

For example, learning outcomes increased 2.3% between 2014 and 2016.  Is the increase just a measurement error? For our understanding, the trend of reading outcomes from 2011 to 2014 is as follows: 40.4, 38.8, 40.2, 40.3. It has now increased to 42.5.

Similarly, can the public-private comparison of increments also be just attributed to measurement error?

Finally, the usual reminder applies. We are talking of basic minimum competencies here and also at a low level. Minor improvements shouldn't hence make us complacent as there's a lot to be achieved.

Jhumpa Lahiri - School uniforms - RTE - Social Inclusion - Social Capital

In Conversation with Tyler,  responding to one of the questions, Jhumpa Lahiri talks about her experience of trauma in school where she had to struggle daily to choose a dress to wear (they didn't have uniforms) and associated psychological complexes.
If I had to choose, I would choose the safety of the uniform because, of course, the whole piece, the whole little essay begins with the memory of being a child and being traumatized by having to dress myself. Because it just churned up so many problems and was a source of true anguish for me, as a child, to have to choose clothes and put them onand this has economic ramifications, this has cultural ramifications, this has all sorts of ramifications, because clothes are things we buy in stores, etc., etc. 
I had this crazy envy, admiration, obsession with my cousins’ school uniforms in Kolkata because they were all the same. They just put on what they had to wear to school every day and it was the same thing. And I dreamed about that. I dreamed of being able to wake up in the United States and just put on my blue skirt and my white shirt and my black shoes, and going to school, and nobody commenting on what I was wearing. I was always so terrified because people were always commenting on what I was wearing. And they were teasing me or whatever.
I think there’s this: Where do you stand between wanting to express yourself and be free and being afraid of that freedom, being actually vulnerable to that freedom? I think America represents Freedom with a big capital F. And it always has, and we hope it always will for the good. But there’s also the danger of that, even as a young girl in the ’70s, as a kid, a child of immigrants, I knew what it meant to shop in one store versus another store. 
I saw what the girls in my class were wearing: the kinds of shoes, the kinds of purses. I knew that my parents weren’t taking me to those stores, that they thought that was a waste of money and that we’re not going to pay $40 for Nike sneakers, or whatever it is, because it’s a waste and you’re going to grow out of them in six months. Whereas my schoolmates had these things, and suddenly there was the gap between me and them, reinforced by these things. For a child, at least for me, these things were traumatizing. And I imagine for others as well.
Often people ignore the role of uniforms in schools but as Jhumpa Lahiri succinctly puts, there are wide cultural, economical, sociological and psychological ramifications behind it.

It highlights the of role of schools in the society. While there is issue of learning outcomes on one hand,  schools and education also have a much larger role in shaping the citizens. Schools are a venue to help build traits of empathy etc, bridging the psychological divide that otherwise exists in the society outside.

However, over the time, schools in India have become segmented, as per socio-economic profiles of parents. Rich parents send their kids to a set of schools, while poor parents send their children to a different set of schools.

To put in words of Robert Putnam and Prof. Ashutosh Varshney, such segmentation reduces the 'social capital', which is not good for society in long-term. 

To give a short brief, social capital refers to the collective value of all "social networks" [who people know'] and the inclinations that arise from these networks to do things for each other ["norms of reciprocity"].

Robert Putnam in his book "Making Democracy Work" argues that social capital is key to high institutional performance and maintenance of democracy. Reduction in social capital means reduction in cooperation among people, which Putnam attributes as a reason for difference between South and North Italy.

In Indian context, Prof. Ashutosh Varshney has argued that associational forms of engagement (sports clubs, film clubs etc.) explains why the inter-ethnic conflict was contained in certain areas of India, while their surrounding areas experienced such conflict. It illustrates the role of social capital.

Returning to our context, segmentation of schools as per socio-economic profiles reduces the interaction between different sections of society. It thus reduces the social capital, hurting the society in long term. Over time, such segmentation has completely absolved schools of its role in building social capital.

It's time to correct this deviation. Schools have to i) prevent the external socio-economic differences from reflecting within class rooms and schools; ii) refrain themselves from taking steps that make the socio-economic differences reflect within the classrooms and schools; and iii) promote inclusion within classrooms and schools

On preventing front, as Jhumpa Lahiri narrates, the school uniform precisely does that role of providing uniformity.

A recent incident of TN private schools is an example of schools' policies that can allow the socio-economic differences to be reflected in the classroom. In TN, a particular school had started a new system of differentiated fee for children attending the same school, that allegedly gives differential treatment to them. The government objected to this and didn't allow it.

In the discourse that followed, some argued that government shouldn't interfere with the autonomy of schools, and should be allowed to design business models as they wish. It was questioned - if trains and aeroplanes can have differential pricing for the facilities, why not schools?

Such arguments under-appreciate the role of schools in the society. As discussed earlier, schools have a larger role in society that the trains don't have. Allowing such discriminatory practices within the school, will bring the outside socio-economic differences into the school and classroom. 

Prevention and self-restraint conflict with principles of liberty (freedom) but as Jhumpa Lahiri nicely puts, freedom shouldn't be extended to a point that makes some vulnerable. And, it shouldn't lead to long-term damage which in fact makes people unable to exercise their very freedoms.

A mere prevention and self-restraint isn't enough. Schools also have to promote inclusion by actively bridging the often invisible gaps.

Reserving 25% seats in private schools to students from low-income communities is a 'revolutionary step' in this regard. Its importance is under recognized but there's a silent transformation going in the classrooms, which will have sociological implications in long term.

The initial hiccups in building inclusion within classrooms highlights the weaknesses in our system. Our schools had to never face the problem of promoting inclusion till now, as they were dealing with a homogenous set of students. The new heterogenous mix is a sudden shock to them. There are also instances of discrimination reported. It takes time to adapt to the new paradigm. At least, as per some evidence on this, this step has been successful in building pro-social preferences among children, without affecting the academic scores.

Government should also take appropriate steps to resolve schools' issues like delays in reimbursements, lack of clarity in admission norms etc., and make schools willing and active participants in this larger step.

In summary, schools also have an important function of building social capital in a society. Over time, they got absolved of such responsibility due to segmentation of students as per socio-economic profiles. It's time to correct this deviation by following policies that both prevent socio-economic discrimination from being reflected within classrooms and also actively promote inclusion. Making schools an active partner in building social capital will result in long-term benefits for the society as a whole.

Data and Dogma: The great Indian poverty debate

Anyone who wants to understand why Angus Deaton got the Nobel Memorial Price in Economics for 2015, should read his 2004 paper on Indian poverty. Deaton is appreciated for his work of careful estimates of poverty. Deaton's book on household surveys is the go-to book for any development economist conducting surveys.

The usual figures reported in media like 'consumption', 'income', 'assets' etc. are not as easy to measure as they seem. 

For instance, how do you measure consumption of a household in surveys? If you ask them, what exactly should the question be? How should it be phrased? Should we ask, what the respondent ate yesterday or should we ask what they ate a week ago? 

Similarly, income is not a straight number that households give you when you survey them. Employed people can directly quote the monthly income but what about those who have irregular income from diverse sources? These things have to be measured carefully. In one of the surveys that I ran, our income module ran into 4-5 pages! Usually, people think income is just about only one question in the survey.

More important is the comparison across surveys. How do you account for change in the text of the question, recall period and so on? Similarly, how to correct for missing data or non-response. It's a complex job.

Owing to this uncertainty, there's a lot of scope for diverse interpretations making it a fertile ground for ideologically motivated people to dress figures to suit their narrative.

Deaton's 2004 paper "Data and Dogma: The great Indian poverty dabate" captures the essence of many arguments around growth-poverty-inequality data. In this paper, Deaton carefully deconstructs difference in data across surveys and how certain sections of economists have (deliberately) misinterpreted it. It also reflects the larger issue of politics around the data.

Some comments in this paper on Surjit Bhalla's work are insightful to read to get a sense of the politics around data and statistics.
Bhalla does not address the detailed arguments by Minhas against this sort of adjustment, nor Kulsheshtra and Kar’s demonstration of the inferiority of the numbers that Bhalla treats as correct compared with those that he rejects. Indeed, Bhalla does not reference their papers.
Further, Bhalla compares data across surveys that are incomparable, to prove his point.
Bhalla’s argument that there has been no increase in inequality is based on measures that appear to be taken directly from the unadjusted 1999–2000 survey, and are compared with similar measures from earlier surveys. But as we have seen, the unadjusted data from the 55th Round understate measured inequality because of the change in response periods for the low frequency items such as durables and clothing. 
Deaton makes these final comments on Bhalla's work
Bhalla’s work is important, not so much for its calculations and conclusions, which are not credible, but because it represents an important and widespread strand in recent Indian thinking, that the reforms have not only been associated with rapid growth of national income, but with the virtual elimination of poverty in India.
Anyways, I highly recommend reading Deaton's paper to get a rich understanding of the nuances behind poverty and inequality data and how even a slight change (for example, change in recall period in survey question) can lead to completely different set of results. 

At a time, when working on these seemingly minor pedantic things wasn't a fashion as compared to other research themes in economics, Angus Deaton persistently worked on them bringing rigour and clarity. Deaton rightly deserved the Nobel Memorial Prize for this.

On constraints in recruiting teachers and larger issue of low capability trap

IndiaSpend has a good article on teacher shortage in Delhi public schools. It lists two primary constraints.

On recruitment procedure
The Delhi Subordinate Services Selection Board (DSSSB), which recruits teachers in Delhi, also hires patwaris (village registrars), nurses and malaria food inspectors for the government. This means delays, with further time lost during the exchange of candidate dossiers between the board and the education department.
On shortage of qualified teachers
There are candidates who hold a Bachelors of Education (B.Ed) certificate and have cleared the teacher eligibility exam at the state and central level but do not pass the DSSB exam. 
Three years ago, the board conducted an exam to fill 5,013 graduate teacher vacancies. Of these, 2,133 were hired, according to the board official. In the case of 239 candidates, caste-reservation documents needed further scrutiny. The rest could not make the 40% pass score. 
The board also  had 1,596 unfilled vacancies for special educators to teach disabled students because candidates failed the exam.
This is a symptomatic of a larger systemic problem of low capacity trap where the short term measures are in conflict with the long term measures.

There are two issues here - the relevance of degrees and the shortage of teachers.

1. One needs trained teachers for classrooms. Degree and exams are usually used as a metric. But, in a context, where degrees don't mean anything, how much should one rely on these?

Many private schools complain that it's difficult to get teachers who are both good and have a B.Ed degree. For instance, a teacher might be good but without a degree and those with degrees might not be good. But, government regulations mandate to recruit only those with B. Ed degrees.

2. In case of severe shortages, excessive reliance on (meaningless) degrees hurts in short term because it artificially creates teacher shortage.

If government schools aren't able to recruit teachers with these requirements, imagine the condition of private schools who have to recruit teachers with similar stipulated government regulations.

One can argue that B.Ed as a requirement should be done away with, considering its irrelevance and that one should focus more on 'on-job training'. This might be useful in short-term but is it good in log term?

The other argument is that if degrees are the problem, teacher education should be revamped instead of removing it as a criteria. It's easier said than done since revamp takes a lot of time. Considering the fact that many governments aren't attempting to do anything on this front, the short-term damage is probably going to continue longer. In that context, isn't it wise to relax regulations till the teacher education is revamped? 

But then again, if such regulations are removed, will the governments have the incentive to improve teacher education? It's also important to note that even on-job training is poor in government schools. So, there's no hope of training them later too.

It's a tough balancing act between short-term and the long-term. If governments don't act immediately, the choice gets more difficult in future.

On a similar note, this issue also highlights the quality of recruitment exams. The above story talks of discrepancy in performance measured by Teacher Eligibility Test of State government, Central government and DSSB exam. We don't know the good exam amongst the three, that can be used for comparison. It might also turn out that all the three exams are poorly designed.

A more fundamental question is - is it actually possible to measure the skills required for teachers through exams with multiple-choice questions?

Shortage of human resources and mismatch between job demands and degree requirements isn't an issue in education alone. For instance, former Union Health Secretary K. Sujatha Rao argues that we need to relax qualification requirements for nurses in operation theatres. In a context with large nurse shortage, is it wise to require a 4-5 year degree to perform the job of just passing on surgical equipment? Won't it be better achieved by a shorter training through diplomas?

Relation between No Detention Policy and Decline in Learning Levels

Learning levels (percentage of grade 3 children who can read grade 1 text) in public schools of India decreased by 10.7% between 2010 and 2014, as per ASER. I had earlier blogged on the declining learning levels in India, trying to bring some structure to the debate, using the available data.

There can be many reasons behind the decline in learning levels in India. Of all, No Detention Policy (NDP), that allows students to progress grades without failing, is attributed to the decline in learning levels. It is argued that learning levels declined after introduction of NDP through RTE in 2010.

There is a widespread belief on this hypothesis in policy discourse. For instance, 18 states agreed to revoke NDP. Recently, (Jan 13th, 2016), there's a news report where NITI Aayog seems to attribute decline in learning levels post 2010, to No Detention Policy.

It may not be possible to establish or disprove a causal relationship between NDP and RTE. However, certain questions are to be answered before we strongly attribute NDP to decline in learning levels. 

Some observations from ASER report below. Learning levels here means the percentage of grade 3 students who can read grade 1 text.

1. Decline in learning levels between 2006 and 2010: During 2006 and 2010, when there was no NDP, there was a decline of 3.3% in public schools. More specifically, 15 states out of 27 showed a decline during 2006-2010.

In other words, there was already a declining trend of learning levels by the time NDP was instituted.

If NDP is the major constraint as it is being made out to be, why was there a decline pre-2010, when there was no NDP? 

2. Reversal in declining trend in some states post-2010. There are 5 states that reversed the declining trend in public schools post 2010, i.e. their learning levels declined between 2006 and 2010 but increased between 2010 and 2014. How do we explain that?

3. Reversal in declining trend in private schools: Learning levels in private schools declined by 0.8% between 2006-2010 and increased by 1.4% between 2010 and 2014. Also, the number of states with declining trend has reduced from 10 in 2006-2010 to 7 in 2010-2014.

It means that the situation has actually improved in private schools post RTE. They arrested the pre-2010 declining trend and have instead improved their levels between 2010 and 2014.

No-detention policy applies to private schools too. How do we explain this reversal of trend? May be, the argument should then be that NDP affects public schools more than private schools. If it is so, then isn't the problem really with schools and not necessarily with the policy?

4. Low-level of learning outcomes during no NDP era: The learning levels were low even before 2010, the no NDP era. Why were the learning levels low in the no NDP era?

5.  The number of students who had to repeat grades even when there was no NDP was low: The percentage of overage children in primary school, those above 11 years of age, fluctuated between 5.2 and 6.2 during 2005 and 2009, when there was no NDP.

Over-age enrolment can be due to two factors - late admission in school or repetition of grades. Even after assuming that all the over-age children in primary school are due to repetition of grades, they  still constitute only 5%-6%.  It means that only a small section of students were being made to repeat grades, when NDP was absent.

Assuming a similar trend post-2010, how far could these kids have led to decline in learning levels? It is possible that some kids from the other 95% have started putting less efforts due to NDP, but then such argument takes us back to the four questions above.

In summary, the learning levels were low and there was already a declining trend before 2010, the number of students who had to repeat grades before 2010 was low, there are states that reversed the decline post-2010, learning levels increased in private schools between post -2010. How do we explain all this if NDP was the major reason behind the low learning levels and the decline post-2010?

In usual discourse, there is a strong tendency to report the superficially visible issue as the binding constraint, attributing all failure to it. For instance, in a school with 60 students per class, a teacher would attribute the low learning levels to the huge class size. It seems a reasonable explanation but it needn't be true because learning levels may not improve even if the size is reduced to 30 because the actual issue is the lack of efforts by the teacher. S/he may be just reporting what s/he perceives as pressing problem and using it as an alibi for non-performance. Now, it doesn't mean that the class sizes shouldn't be reduced. It just means that it's not the main reason why students don't learn.

Is something similar happening with NDP too? If NDP was the main constraint that's leading to the decline, how do we answer the above five questions?

Does all of this mean that the actual problem lies elsewhere and NDP is just being used as an alibi for non-performance?

We may not know answers to these questions as of now. Understandably, there might be other reasons for revoking NDP, which may be fair and reasonable. But, the above 5 questions are to be asked and answered before we attribute declining levels to NDP. If not, we might be misdiagnosing the problem and by revoking NDP, we might get complacent thinking that we have resolved a major constraint. It's better to step back, ask tough questions, and get closer to the truth, before we spend our enormous energies on something that might turnout not to be the actual constraint.

Also read: My earlier post that explores the theme of declining learning outcomes in more detail -"What's the reason behind declining learning levels?"

Update: As one of the readers commented, there can be other reasons behind the dynamics of changes in learning levels. Hence, NDP should not be seen in isolation. I agree with it and I would respond it in two ways.

One, people who argue that NDP is behind the decline in learning levels implicitly mean that NDP is the main driving factor, if not the only one. Hence, I applied same standards above.

Two, in an earlier post, I discussed the contributions of other possible factors in detail. This blogpost is in some sense an excerpt of the earlier post. I did so because most of the debate is around NDP and hence thought of putting it as a separate post.

Antibiotic over prescription - The alarming state of health care in rural India

Consider these stats. Nearly 60% of healthcare spending in India is out of pocket. 47% of hospital admissions in rural areas and 31% in urban areas are financed by sale of assets. 30% of those needing care in rural India and 20% in urban areas go untreated because of inability to pay. Health care  costs push 40 million people into poverty every year.  These are alarming figures by any standards.

While this highlights the financial burden on the poor due to health care costs,  there's another danger looming in India, unnoticed. It's the irresponsible and excessive prescription of drugs. Antibiotics have a unique property. Bacteria builds resistant to these drugs. So, each time, you need to increase the dosage. It means that if you end up taking high dose for something that could have been dealt with low dosage, next time you need to take a higher dosage (higher than previous high dose). The ladder escalates. The worst part is that once the bacteria develops resistance, even a person who has never taken antibiotics has to take high dosage if infected with drug resistant bacteria.

India is being called the superbug (bacteria resistant to all known antibiotics) capital of the world. "While antibiotic usage worldwide in the first decade of the 21st century rose by 36%, in India, the count went up by 62%." Consequently the drug resistant strains also increased and the number of people dying from such infections. For instance, consider this: The resistance to Klebsiella pneumonia to carbapenems, the antibiotic of last resort is 57% in India, as compared to below 5% across Europe.

The over prescription of drugs is rampant in rural India, where there is a proliferation of quacks (untrained medical doctors). Manoj Mohanan of Duke University finds that only 20% of children suffering from diarrhoea are given appropriate drugs in his study in Bihar. The drugs prescribed to the rest 80% are "scandalous" and "criminal". The rest are given powerful antibiotics even for those diarrhoea cases that don't require antibiotics. In a significant number of cases, they are even given steroids and even chemotherapy drugs. It wrecks the health of children in long term.

One may then infer that the trained medical doctors are better at this. Unfortunately it's not so. They are no better than the untrained quacks as found in several studies.

One might then infer that giving short term training to the quacks on these aspects will constrain their quackery. Abhijeet Banerjee, Jishnu Das et al. conducted an experiment in West Bengal that provided training to quacks. While it improved their adherence to medical protocols on other aspects, it didn't reduce the over prescription of drugs.

'Competition', the need to attract patients seems to be the main reason behind this. Powerful antibiotics and steroids give instant relief while causing long term damage. Patients can only feel short term pain relief but not long term damage. Hence, they prefer to go to those doctors who give them short term relief. Doctors act accordingly by over prescribing drugs. It's a supply induced demand! It's also the reason why some doctors answer appropriately in tests of knowledge but act differently while treating.

It's thus a difficult problem that calls for doctor's self regulation, which is difficult to achieve on a large scale. Standardization and use of technology like tele medicine could be a hope to address self regulation issues. Bihar tried a social franchise model trying to standardize treatment protocols and use tele medicine. The idea is that a medical practitioner can get a franchise - s/he gets list of protocols to be followed. The doctor can observe their mentor's clinical practice remotely and also seek help from them whenever required. It was hoped that this could help to that extent.

Mahanan's large scale RCT finds that this approach also didn't help. To start with, the take up was low. It's because no one has an incentive to invest in getting a franchise, adhering to protocols and invest in gaining extra knowledge because they already have a thriving practice.

Though the attempts to standardize protocols and use of technology are limited, the existing one's don't show positive picture, in their current state. Self regulation seems to happen only in some top tier hospitals. For instance, in some top tier hospitals, prescription of antibiotics beyond certain strength has to be approved by a senior doctor.

The larger picture is that arresting over prescription of drugs is a complex problem. It's not yielding to any traditional instruments  - training informal doctors, using technology etc. The commonly used instrument of getting rid of discrepancies with competition also doesn't help in this case because competition is the primary culprit here.

In an ideal world, this should call for high levels of government attention and some urgent steps with innovative thinking. You know what our governments did instead? They banned transport of blood samples outside the country arguing that western researchers are spoiling India's name by highlighting the antibiotic resistance problem, hurting its medical tourism! 

Not stopping this problem is a criminal negligence to say the least.

What's the reason behind declining learning levels in India?

It is a well known fact in policy circles that learning levels in India are poor.  The other lesser known fact is that learning levels are declining. As per ASER, percentage of grade 3 students who can read grade 1 text reduced from 42.5 in 2010 to 31.8 in 2014. Young Lives survey that tracks children longitudinally over the years also found a similar pattern.

The decline is puzzling, especially considering the increased attention to education in recent years. There could be many reasons for this decline. In particular, RTE was introduced in 2010. Some of the reasons are listed below.

Hypotheses for declining learning levels


1. No-detention policy came into effect from 2010, as part of RTE. It means that a child can progress to next grade regardless of the  child's learning levels. It is argued that it has taken out fear factor from students and parents leading to lower efforts from them, that resulted in this decline.

2. Continuous and Comprehensive Evaluation (CCE) also came into effect from 2010, as part of RTE. As per this, a teacher has to conduct a number of low stakes tests across the year. It is argued that the CCE framework involves lot of record keeping and it's taking a toll on teachers, taking time out of their teaching and planning activities. Since inspections place emphasis on record keeping, teachers spend more time on that as compared to other activities.

3. The expansion in schooling without adequate expansion of monitoring mechanisms might have resulted in the decline.

4. Private school enrolment has increased over the years. It could have impacted the learning levels in three ways. One, the better motivated students are taken out of public system. It thus reduces the peer-learning effects (children learning from each other).  Two, the shift from public schools might have lowered the parent accountability. Three, private schools aren't able to adapt to the profile of students who have joined newly.

5. Learning levels were already in decline by the time RTE came into effect.

6. May be the declining trend nationally is only due to a couple of big states and situation is better in other states.

7. May be there's an issue with ASER's methodology

Of all these, the first reason is the most cited one. There's a strong uproar against the no-detention policy (NDP) from many states. They want it to be repealed. The decline in learning levels is attributed to NDP.

In reality, no one knows the real reason. We also have limited data to trace the causal mechanisms. In this context, this post tries to provide some structure to the debate using the limited data available. The aim of the post is not to give precise answers but it's to raise more questions! Since RTE is the primary anchor of the debate, we focus mainly on trends before 2010 and post 2010.

I have compiled ASER data from 2006 to 2014. In particular, I have considered only reading data. It's primarily because one has to manually enter data point of each year for each state from pdf to excel. Please feel free to mail me at iterativeadaptation @ gmail if you want the compiled data. I am happy to share.

Percentage of grade 3 students who can read grade 1 text is used as the metric in following observations. Learning levels refer to the percentage of grade 3 students who can read grade 1 text. All the data is sourced from ASER's report on learning level trends.

Observations from the ASER trends for the public schools:


1. Learning levels were already in decline by the time RTE came into effect. Between 2006 and 2010, learning levels declined by 3.3%. However, the decline seems to have accelerated post 2010. It declined by 10.7% between 2010 and 2014.

2. In terms of number of states, 15 out of 27 states already had a declining trend of learning levels between 2006 and 2010. The number of states with declining trend of learning levels increased to 21 post 2010, by 2014.

If RTE is the reason for decline, how do we explain the declining trend in 15 states pre-2010? May be the argument then could be RTE accelerated an already existing declining trend. It's a difficult hypothesis to verify.

These are trends at national level. The trends at state level have a mixed pattern. Some trends follow the national pattern while some states follow the opposite pattern. The picture is thus mixed.

3. Five states, Manipur, Meghalaya, Himachal Pradesh, Karnataka and West Bengal have a declining trend before 2010 but there's an increasing trend post 2010. Manipur and Meghalaya have a significant improvement.

4. In several states like AP, Chattisgarh, Punjab and UP etc the increasing trend pre-2010 turned into a declining trend post 2010. Punjab seems to be the worst. It went from a 13.6% increase between 2006 and 2010 to 10% decrease between 2010 and 2014.

UP especially is quite worrying, given its lower levels. The percentage of grade 3 students who can read grade 1 text in 2014 is  astonishingly low at 13.4% in UP and it came to this level from 26.5% in 2010.

5. There are some states that continued on a continuous declining path - Gujarat, Haryana, Rajasthan, Kerala etc.

6. Two states, Madhya Pradesh and Bihar deserve a special note. The learning levels in Madhya Pradesh declined from 64.8% in 2006 to an astonishingly low figure of 16.2% in 2014. What a fall! Bihar fell from 51.7% to 25.3% during the same period (2006-2014), a huge fall too.

All of this is for public schools. The story is bit different for private schools.

Observations from ASER trends for private schools:


1. At national level, in private schools, the learning levels declined by 0.8% between 2006 and 2010. However, it increased by 1.4% post 2010, between 2010 and 2014.

2. ASER has enough data for only 19 states. Out of these 19 states, 10 states had declining trends pre-2010. This number reduced to 7 post 2010.

Overall, it suggests that learning levels in private schools have moved from a declining  trend pre 2010 to an increasing trend post 2010. But, as like public schools, this is only at national level. We find mixed patterns when we look at state level data.

Coming back to our original questions - What's the learning from the above story? Is RTE responsible for the decline in learning levels post 2010?

The answer still is "we don't know" but the above data helps us appreciate the problem better. We can make the following observations on the seven hypothesis listed in the beginning, from the above data.

Summing up


1. On No-Detention Policy (NDP)

Data highlights important contradictions to the hypothesis that No-Detention policy (NDP) is behind the decline. They are:

1. There are 5 states that reversed the declining trend in public schools post 2010. How do we explain that?

2. The situation has actually improved in private schools post RTE. They arrested the pre-2010 declining trend and have instead improved their levels between 2010 and 2014. Also, the number of states with declining trend has reduced from 10 to 7.

No-detention policy applies to private schools too. How do we explain this reversal of trend? May be, the argument should then be that NDP affects public schools more than private schools. If it is so, then isn't the problem really with schools and not necessarily with the policy?

3. If NDP is the major constraint as it is being made out to be, why was there a decline pre-2010, when there was no NDP? We may note that 15 out of 27 states were already on a declining trend pre-2010 during which there was no NDP.

4. More importantly, the learning levels were low even before 2010, the no NDP era. Why were the learning levels low in the no NDP era?

5. The number of overage children in primary school, those above 11 years of age, fluctuated between 5.2% and 6.2% during 2005 and 2009, when there was no NDP. It's important to note that not over-age enrolment can be due to two factors - late admission in school or repeating grades.

Even after assuming that all the over-age children in primary school are due to repetition of grades, they still constitute only 5%-6%.  It means that only a small section of students were being made to repeat grades, when NDP was absent.

How far could these kids have led to decline in learning levels?

Does all of this mean that the actual problem lies elsewhere and NDP is just being used as an alibi for non-performance?

We may not know answers to these questions as of now but these questions should definitely be asked in the debate on NDP.

2. On CCE

It's difficult to comment on CCE using this data but we there's one RCT on CCE conducted in some districts of Haryana which suggests that CCE didn't improve learning outcomes. It suggests that CCE at least didn't show any negative effects, as this hypothesis suggests. It might be difficult to extrapolate it to all contexts but it's the best evidence available.

3. Rapid expansion of schooling leading to degradation of oversight: We can't directly verify this by just using ASER data. However, if all the other hypotheses are ruled as 'major causes', then by the method of elimination, we are only left with this.

However, we can cross check this hypothesis using enrolment data. Enrolment data is the proxy for expansion of schooling. The Net Enrolment Ratio (NER) (percentage of kids of a particular age group attending the school) has increased from 84.5 in 2005 to 98.6 in 2009. It's a matter of judgment as to whether a 14% increase in enrolment leads to enough burden to decrease learning levels.

[Expansion in schooling should ideally be measured using increase in number of schools vis-a-vis increase in number of teachers. I have to yet collect data on that. I will update the post after it's done.]

4. Increase in private school enrolment:

One question that the data answers more clearly is that private schools aren't necessarily failing at adapting to new group of students. If it were so, then we should have seen a declining trend in private school learning data, with increase in shift from public to private schools. It isn't the case, nationally.

It's a possibility that may be private schools aren't able to adapt and the decline due to new students is being offset by some other factor. But, I can't think of any such positive factor that could offset the plausible negative effect.

5. Already declining learning levels by 2010

As we have seen, it is true that learning levels were already in decline before 2010. Nationally, the decline accelerated between 2010 and 2014 but there's a mixed pattern if it's analyzed at state level.

6. On contribution of major states

There might be some truth to it but we don't know the extent though. We notice that all the major states in terms of population are on a major declining trend, especially UP, MP, Bihar and Punjab. Arresting the decline in these states is going to be crucial in coming years.

7. ASER methodology

Apprehensions over ASER's methodology needn't necessarily be true for two reasons. One, ASER has been publishing yearly data for past decade. So, if there is an issue with methodology, there would have been huge discrepancies across years. It isn't the case. Two, the declining trend is also observed in other independent surveys like Young Lives.

Way forward


Overall, this suggests an urgent need to invest in research to understand these problems better. Any such analysis should be ideally at a state level or a district level because there's a mixed pattern as we have seen. There are important stories at state level that are to be captured and matched with the data. For instance, Seema Bansal of BCG attributes the arrest of declining trend in Haryana (the decline is arrested post 2012 but it's yet to reach 2010 levels) to a series of reforms that the state has taken up in partnership with BCG. We need more of such stories to understand the problem better. Only then we will get a true picture.

As a matter of caution, one shouldn't get carried by terms like "arrested decline", "improved learning levels" etc because we are talking of very low learning levels and that too based on a metric of minimum competency (ability to read). We are far behind the goal of achieving true learning and that of higher order skills. It's important to remember this because often there are heated debates on relative comparisons between states and between types of schools based on the relative performance on metrics of basic competencies. Even if something is performing relatively better, we haven't even scratched the surface on an absolute scale.

Finally, to reiterate, there's an urgent need to arrest decline in the big states like UP, MP, Bihar, Punjab, Gujarat etc, since they constitute a large number of school going children.

PS: All the data is sourced from ASER's report on learning level trends.