Karthik Muralidharan Examines the State of the Indian State

Economists Shruti Rajagopalan and Karthik Muralidharan talk about improving India’s state capacity, streamlining public service delivery, staffing, and public expenditure in India.

SHRUTI RAJAGOPALAN: Welcome to Ideas of India, where we examine the academic ideas that can propel India forward. My name is Shruti Rajagopalan, and I am a senior research fellow at the Mercatus Center at George Mason University. 

Today my guest is Karthik Muralidharan. He is the Tata Chancellor's Professor of Economics at the University of California, San Diego. He is the author of the recent book Accelerating India’s Development: A State-Led Roadmap for Effective Governance.” We talked about the lacking state capacity in India, about improving the quality of public expenditure, fiscal federalism, methods to improve the hiring process for government, better ways of staffing and using the Indian bureaucracy, randomized control trials and development and much more.   

For a full transcript of this conversation, including helpful links of all the references mentioned, click the link in the show notes or visit mercatus.org/podcasts

I feel like I studied for a field exam on—let’s call the field the state of the Indian state or something like that, and it’s incredible what you’ve done. Let me back up. For those who haven’t read the book, the book is called “Accelerating Indias Development: A State-Led Roadmap for Effective Governance.” That’s just the title of the book. The core message of the book is that contrary to popular opinion, the Indian state is not as large as it tends to loom.

It’s actually a very chronically under-resourced, understaffed enterprise. What you look at is mainly problems of service delivery, but some things in addition to service delivery, like data management, personnel and so on. And you try and find all the research that can inform us about how this can be streamlined, and because you’re an economist, how it can be incentive-aligned. Your goal is actually really straightforward but also really ambitious. The goal is that the state should be able to deliver essential services to every single Indian, efficiently, optimizing the tax rupees without any leakages, and so on.

The streamlining that you look for in the book—this is the part that I found quite extraordinary actually—it’s in the areas of health, nutrition, education, staffing, personnel management, policing, judiciary. I’m sure I’m missing a few out because it’s got 18 chapters, and then you come up with recommendations and solutions. What I felt was it’s really like a capsule. If 200 years down the line someone comes and finds this book in a time capsule, it’s the capsule of the state of the Indian state today, and the best research we’ve had, say, in the last 30, 40 years on what the state is.

First up, I recommend everyone reads the book, and that’s also why I sound so excited about it. It’s a gift to researchers, to be honest, because it’s so much research just condensed into what I call a capsule, except the capsule is very large, and it is a bitter pill to swallow for many. That’s how I felt about the book. You’re smiling.

KARTHIK MURALIDHARAN: No, no, I’m just listening. I’m just absorbing it. Listen, I think the core reason for writing this book is that the motivation for my research and all the work I’ve done in 25, 30 years has been to see how research can inform better policymaking, better governance. Because you see that at the end of the day, the state is the entity with the largest ability to shape the lives of the poor, the common good, both by doing too much and by doing too little.

I think if I was to abstract away and say what is the single most important message in the book, I think you get most of it, but there’s I think a one-sentence way to say this, is that we spend most of our public discourse and debate focusing on what should the government do and remarkably little on how should it do it, or whether it even can do it. That’s why this is really two books in one.

The one thing which I’ve struggled with and I still have mixed feelings about is whether this should have been two volumes, where the first half of the book is really building an effective state and the second half is accelerating India’s development. The logic is really, I do believe, and this is consistent with what multiple other commentators in the Indian economy have said, is that the binding constraint to India’s development in almost every way is the effectiveness of the state itself. The ’91 reforms, and I cite The ’91 Project, the website. 

RAJAGOPALAN: I was thrilled to see that.

MURALIDHARAN: The ’91 reforms was incredibly important but it was really about getting the state out of the way, out of the way of things that it should not be doing. The second part of the agenda, which is making a state capable of doing what it should be doing is where we have to now do the heavy lifting. This is not an easy task. People have famously said building effective states is like the slow boring of hard boards, and it’s taken high-income countries 100 years to do this.

My goal in this book is that we can compress India’s cycle of building an effective state from 100 years to 25 years, then I would consider my job to have been done. I think that’s why the book is ambitious, but if I may just say something to your readers, which hopefully you can confirm, is it’s a very, very easy read.

RAJAGOPALAN: Oh, it’s a very accessible read. Actually, that’s the reason it’s a gift, that you’ve done the heavy lifting of reading all these dense research papers and RCTs. Actually, you’ve read everything from really the tough papers to randomized control trials that you’ve conducted, to blog posts to little bits of classifieds in newspaper columns, to our online project which is The ’91 Project. It really covers the gamut. I encourage everyone to read it, irrespective of whether they have any background in political science or economics, if they just want to understand their surroundings better on why certain things are broken and what works and why it doesn’t work

MURALIDHARAN: Yes, and I think the core theme—going back to just wrapping up the thing—it’s about that we talk about what to do and not enough about how to do it. I think that’s one key point of departure, which is why the building the effective state is almost the first half of the book. I think the other key unifying theme simply in one sentence is that it really is about the quality of expenditure. It means all of our public discourse focuses on what I call the top-line budget allocation and not the bottom line of how that budget translates into outcomes.

In a way, the intellectual thread that connects the whole book is, when I look at what’s the pattern emerging out of my own studies and other studies in multiple sectors—I started my work in education, then started working in health, then did a bunch of work in welfare programs—and you see that the common thread across everything is the weak governance itself and that the returns to investing in governance and state capacity are often 10 times more than the business-as-usual spending on the top line. That’s a very difficult transition to make because the top line is what you observe and that’s where the conversation is.

Whereas what the research has been about over the past 20 years has been in sector after sector, tracking the fund flows, looking at how does the spending translate into impact, and that’s why the book, though it’s been an insane amount of work, it’s also been so much fun. Because there is so much you learn over the years through all of this that never makes its way into an academic paper. The time to write a book in a scholar’s career is when you feel that you have something to say that is not being said in those papers and connecting the dots. It has been an overwhelmingly all-consuming project but hopefully, the output is worth it.

Quality of Public Expenditure

RAJAGOPALAN: I want to start with what you just talked about, which is the quality of public expenditure. This is in part one of the book. There’s a big discussion on public finance. You talk about expenditure, revenue federalism, and you parse out all the different problems. In the book, you argue that to improve the quality of public expenditure, we need to focus on things like calculating ROI, which is the return on investment, considering the second-order effects of public spending. Like if we spend too much on fertilizer or free electricity, what that does to groundwater.

How do we boost both equity and efficiency? Because India is in that weird quadrant in your two-by-two where both can actually be improved without it being an explicit tradeoff. You talk about more data-driven budgeting. You talk about more monitoring. You also address that expenditure is too centralized and basically, all your fixes, the way I read them, were about streamlining the current system as it is.

When I take one step back, when I finished reading those chapters, I think it’s chapters six, seven, eight of the book, and then again, a discussion toward the end where you reimagine the Indian state, to me, the question is, isn’t the core problem of public finance in India that politicians and bureaucrats aren’t directly accountable to taxpayers? That link is too weak because it’s too centralized and we gather revenue in the most bizarre fashion, relying heavily on consumption taxes, virtually zero on local taxes or property taxes, and the spending almost entirely is controlled top-down. In fact, it’s delegated, even when the state is doing the spending. 

My sense was, one way to fix the problem is the way you look at it, which is, these are the six different areas, let’s find a fix to streamline each of them. Alternatively, I would imagine, wouldn’t dramatic fiscal federalism just take care of this problem? Because that’s a different kind of imagination. Then you don’t have to fix each of these things. Maybe first, you can walk us through how you see the problem and the solutions, and I know, because I know you, I know you have on deep thoughts on public finance and fiscal federalism because you’ve taught some of this stuff, so maybe you can make this a two-parter.

MURALIDHARAN: Yes, thank you. Let’s start exactly with the first part because I think there’s a lot of ideas there in chapter six on the quality of expenditure, which itself I think are worth expanding. Even before that, let me preface that a little bit more for your listeners, which is something which I think is really important. When we think of waste of public money, the first thing that most listeners think about is corruption because we hear about scandals. There are scandals everywhere and you think that’s where there’s inefficiency.

I think what some really high-quality research on government waste shows is that maybe only about 17% to 20% of the waste is what is active waste or corruption. The vast majority is just inefficiency or passive waste. I think this reflects the fact that regardless of what level of government you’re at—because it’s always easy to waste other people’s money—there is just so much low-hanging fruit in terms of improving efficiencies. That’s why I think that we can do a lot. Let me also take two further steps back and come back.

See, going back to this initial point about focusing on quality of expenditure, one thing I want to highlight is that the big debates in development tend to be the growth versus development debates. Which in India has been the Bhagwati versus Sen debates where the growth people will say, “Let’s get the infrastructure, let’s get the capex, let’s reduce the logistics costs and let’s get industry growing.” The development people, led prominently by Sen, would say that the point of development is not just GDP per capita, the point is really human development, and so can we get more in health, more in education?

At one level they’re both right, because more growth gives you more human development and more human development will give you more growth. That fight effectively becomes a bit of an ideological fight where the center-right wants the capex and the center-left wants the social sector. I think the reason I’m giving you this backdrop is another core goal of the book, frankly, is not just the technocratic aspect, but it’s really to build a broad national consensus of saying, listen, we argue so much about zero-sum things, but there are things that we can do that are positive-sum that can allow us to do more of everything.

The point is that the quality of expenditure is so bad, regardless of whether you’re spending on infrastructure or you’re spending on social sector, that if you manage to focus on that, you can do more of everything. Which is why my hope is that the state capacity agenda, the effectiveness of public expenditure agenda is something that can unite both the left and the right. Because doing it better will allow you to better do the basic human development and services and improve the functioning of the state which the left cares about, but it’ll also improve value for money and build the foundations of long-term growth, which is what the right wants.

I want to give that preface by saying that there is a deeper nontechnocratic, and I would argue, almost meta-political reason for focusing. Given that everybody knows that we’re argumentative Indians and we love debating, it’s really about how do you find a way forward through all of these vexing debates. I think that’s been an undercurrent both of my thinking, as well as the writing. Hopefully, that comes across in every part that you read. Now, having said that, now let’s go into the details of public finance and expenditure.

For your listeners, expanding on what you said, that it’s not about India being at a sweet spot, it’s about the variation in the kind of expenditure. Like I said, normally I make fun of consultants and I don’t like simple two-by-twos, but this is a case where I think the two-by-two really works. It’s a simple conceptual framework which, in fact, multiple senior people in government have written to me and saying how much they appreciate that simple picture. That just having that lens in thinking about budgeting can be so powerful because that’s not the way in which people think about it.

The two-by-two is very simple. That you’ve got one axis, which is equity, and one axis, which is efficiency. You have spending proposals or any reform proposals, some of which improve efficiency, but at the cost of equity. One good example of that would be rationalizing the GST to move to a uniform GST. That would significantly improve the efficiency of tax collection, but it may come at the cost of being regressive and therefore, may hurt equity. On the other hand, you’ve got, say, programs like the PDS which presumably improves equity and food security but is inefficient for a variety of design reasons.

Now most of our public discourse happens in that quadrant II versus quadrant IV. Should we do efficiency at the cost of equity or the other way around? A big part of the book is to say that, listen, we’ve got so many large expenditure items that are in region III, which are bad for equity and bad for efficiency. We have so much low-hanging fruit of things that we could do that would improve equity and efficiency that we’re not doing enough. Again, while reasonable people may disagree on the second and fourth quadrants, can we at least put some sunlight on these huge region III expenses and build a broad consensus to say, can we move from here to R1?

I think that’s, again, the framework in which this is sitting. And I think since we’re economists, and maybe this is the podcast I get to be the nerdiest on—again, as a libertarian Mercatus economist, it’s almost axiomatic that we think that there’s a tradeoff between efficiency and equity. Which means that one of Tom Sargent’s 12 principles of economics that he talks about in Berkeley is that there is a tradeoff. The default in public finance is we think there’s always a tradeoff between equity and efficiency and that’s because to do equity, to do redistributive policy, you need to raise taxes which distorts work incentives.

You need to provide “freebies” which may reduce incentives to work and subtly, when I’m targeting benefits to the poor, I need to phase them out. That creates a very high marginal tax rate for the poor as you’re earning more. Put together, it’s almost seen as axiomatic that I cannot do welfare without hurting efficiency, but intellectually what makes development economics so much fun is the fact that both at the individual level and the aggregate level, when you are close to a poverty trap, when you’re close to subsistence, you’ve got a whole class of models that shows how you can have well-designed, well-implemented interventions that can improve both equity and efficiency.

RAJAGOPALAN: We’re not at the frontier yet, basically.

MURALIDHARAN: We’re so far from that. Even when I teach graduate development economics, the first question in the first lecture I ask—which most students still struggle with—I say, when I teach Ph.D. development economics is, what makes development economics different from just doing applied micro in developing countries? Or what makes it different from doing growth economics? The key difference is, in both cases, it is the idea of a poverty trap, meaning, at the level of both individuals and countries.

The point therefore is, because we are still a low-middle-income country, and because we still have large amounts of poverty and large amounts of other market failures and frictions, that there is a space for well-designed interventions that can improve both equity and efficiency. That’s what gives us this free lunch. A big point of this book is there is a free lunch. There is a free lunch if we get more analytical about our expenditure and start allocating. Coming back to things that are in quadrant III or region III, so much of our expenditure is bad for both equity and efficiency.

Which means, free electricity for farmers, probably the single worst policy in the country in terms of just being—it’s the most inequitable thing you could do. It’s the most ridiculously regressive thing, where the roughly top 5% of landowners get 50% of the subsidies, and it’s bad for equity, it’s bad for efficiency and it’s bad everywhere.

RAJAGOPALAN: And the environment, right? Because the second- and third-order effects are horrific.

MURALIDHARAN: Exactly. Or reverting to the old pension schemes. Again, you’re benefiting existing government employees. We have tons of research showing that unconditional pay increases have no impact on outcomes. Part of the problem in personnel is you pay incumbents too much and you don’t hire enough, and so that goes back to your original point about the understaffed state. Anyway, I think this is just laying the framework for saying that chapter six is just laying out this framework for how should we think about quality of expenditure.

In the case of capex, you would think about what is the return on investment, and think about that like any capital budgeting exercise. In the case of welfare, it’s a little trickier because there isn’t necessarily a direct return on investment. There, you can think about the quality of welfare expenditure along the axis of, say, targeting, meaning the axis of delivery, and then distortions. Then you see, and what it’s showing is a taxonomy of how we can do better on each of these things. Then it lays out this framework for how you can improve quality of expenditure.

Decentralization vs. Federalism

Now having said that, let me jump to your second part, which is, why is the answer not just more decentralization where there is a lot more linkage at the local level just between services provided and accountability? I would say there’s at least three reasons. Now, clearly, I believe in more decentralization, and that’s there in chapter eight on federalism and decentralization.

RAJAGOPALAN: Actually, it’s there in all the chapters, it’s pretty much everywhere you’re talking about how we need to not do everything at this tiny little elite group at the top that doesn’t have eyes on the ground.

MURALIDHARAN: Right. It cuts through, but there are three reasons why I’m not putting all my eggs in the decentralization basket. I think one way to think about the book is that, see, it is a menu of options, and so recognizing the political reality. Which means that if I say this is the one silver bullet, and then that doesn’t happen, then effectively it’s like, okay, this is this one-trick pony, and then this is politically not going to happen, and therefore, we will not make any progress. I find that deeply unsatisfying. I think part of the book is really—I say this in the preface.

In the preface, I say it reflects a lifelong journey at the intersection of ethics, economics and politics. Where you’ve got an ethical side of underpinning of what kind of society you want to build. That is constrained by the economics of saying what are the resource envelopes within which I need to achieve this? Then the last part of the politics is saying, let’s think about the distributional consequences of every idea seriously, so that you can build broad coalitions to make things happen. It’s a book that is a combination of idealism and pragmatism. It’s the pragmatism piece that then informs my approach.

Which is to say, see, we are not going to get from a 3% share of GDP spent at local on this thing to 50% overnight. In fact, it’s interesting, even the finance commissions over the years, when they did a gradualism, when they did 1% each, that stuck, but when the 14th tried to jump from 32 to 42, it was too big, which is why then you get a bunch of offsetting behavior, which is then suboptimal in other ways. You got the government of India increasing state shares, you got an increase in cesses and surcharges that were outside. Because that 10% was too big a shift for the system to deal with.

I think, at the end of the day, we’re 1.4 billion people. We’ve got more people than the entire Western hemisphere: North America, South America, Central America put together. Which means that this is a slow-moving aircraft carrier. It doesn’t take violent turns. You’ve got to pivot this gently in the right direction. While we are doing the decentralization piece, I think there is still—anyway, I’ll give you three reasons. The first is just the pragmatic reason. That even the decentralization agenda needs an element of gradualism for it to be sustainable.

I’ll come back to thoughts on how we might do this. That’s the pragmatic reason. I think there is also a deep substantive reason, and the deep substantive reason is, and I say this in chapter eight, it is not obvious that we need more decentralization in everything. There are some areas where actually you need more centralization, and the subtle point—and that’s why I go into the conceptual framework of how does country size affect the quality of governance. There are areas where size is good because of economies of scale, of coordination externalities and a bunch of other things.

There are areas where being small is good because you need to cater to diverse preferences, and you need better local information and the ability to act on that. Broadly speaking, functions of service delivery that require accountability over the front-line service providers are better done at the local level, but functions like infrastructure, or even welfare, which gets very subtle, because as you have many more internal migrants, the way to then think about it—and we can talk more about the Seventh Schedule and things in that. In fact, since given your own interest in constitutional law and stuff, I’d love to talk about that.

There’s three answers; the first is the pragmatic political reason, that the decentralization agenda is going to be a gradual one and not a shock therapy one. There is no state of the world where I see that we’re suddenly going to go from one equilibrium to the other. Putting all the eggs in that basket, I think would not be pragmatically wise. The second reason is that just practically by first principles of fiscal federalism, certain functions do need to sit at high levels, will need coordination, including things like welfare. Then I think there’s a third reason which contributes to that, which is that, I think the history of more decentralized societies comes from more decentralized revenue.

If you want to have authority, then you are contributing with your money directly. The U.S. has had a much stronger history of decentralization because the property tax regimes are much stronger. China is decentralized, but China, what people don’t appreciate enough is how much of the basic services include user charges. People are actually paying for it. When you’re paying locally, that gives you more locus standi to say, “I am paying; I want the control.”

Now, in any polity where the revenue is coming from a completely different place and the expenditure is in a different place, how is that money going to get there? That then gets mediated by a political process at a higher level of aggregation. 

RAJAGOPALAN: That’s exactly what my problem is that no matter how much you streamline, create data management systems, create monitoring, that’s eventually where you have to reckon with that.

MURALIDHARAN: Exactly. Which is why I think my pragmatic view of how we will get more decentralization then connects to chapter seven, which is, we will not get as much decentralization by clamoring for more devolution because that is effectively, it is a local body saying, “I want money,” but it is somebody else’s money. The pragmatic way to get more decentralization is going to be to strengthen, say, property taxes. That’s where I’m most optimistic about the decentralization agenda, because if you go to chapter seven on revenue, we clearly need more revenue, but the focus in that chapter is not just on quantity of revenue, but the quality of revenue.

How do you raise revenue in the least distorting ways? I think economists, we all agree that property taxes is something that we should have more of. One way of doing this is that on the margin, the Indian state needs more revenue, and if that marginal revenue is being generated at the local government level, then you are automatically shifting the share of local government control over expenditure.

To me, that is a more promising way of pushing the decentralization agenda, rather than saying, “I’m going to have some rules that devolve more money.” They’re not mutually exclusive, but I think on the margin, you’re going to get more traction by focusing on revenue raising at the local level, which will then automatically create the local spending authority.

RAJAGOPALAN: I think actually now that I’ve heard you, we’re not that far apart because my interest is actually not decentralization, it’s fiscal federalism. I’ll tell you what the difference is to me. Maybe I’m being nerdy and nitpicking, but to me, decentralization is basically the central authority has the power, and it’s willing to part with that power or devolve that power, but whenever it wants, it can get it back. Whereas genuine federalism means that there is more than one locus of power, which is the Seventh Schedule stuff that you were talking about.

That there genuinely is more than one power center, and in fact, there should be three power centers. This works out much better in places like the United States. In Switzerland, it works remarkably well. In India, you talk about this in the book, because of colonial legacies and central planning, the locus is very much at the union level, and not even at the state level to a large extent. To that extent, we are very much aligned, that I think what we need is not so much decentralization, we need more fiscal federalism, and the federalism has to be fiscal in nature.

You have to raise your own revenue, both through property taxes and through user fees. Actually, I think the poor in India—and you know this better than anyone—actually end up paying a lot in user fees. It’s just to private sector people because they have exited the system, as opposed to having a process within the state system that they can mediate their grievances and actually demand the services and so on. There’s also room for other things, which is not exactly devolution through the finance commission, 1% up and down, but something like when you’re designing the GST, for instance.

At the design level, you can say that these are consumption taxes. The consumption taxes are obviously based on people in a particular location, and that location is good or bad to the extent that it can attract people and have an economic engine of growth which leads to more consumption. Why not at the design level imagine a GST which directly splits everything at maybe 15% to 20% at the local level, another 30% at the state level and then the remaining to the union government, or something like that.

Right now we have a split between union and state. But I feel like even there, as opposed to the kind of devolution we’ve experienced in the past, there is much more to be done at the design level itself. Which again, it’s not that I was disappointed when I was reading the book, but I was like, this is also part of the low-hanging fruit, Karthik. This is very much part of that two-by-two matrix that you’re talking about. You’re not going to lose that much in efficiency if you have a single-rate GST which at the design level says 20% or 15% goes straight to the local government.

MURALIDHARAN: Again, I don’t disagree with any of this. Here’s the irony, the irony is, we’ve got the opposite problem hopefully of what I say in the bureaucracy chapter. The joke I talk about, there are two people at the restaurant, where the first one says, “The food here is terrible,” and the second one says, “The portions are too small.” Which is the truth or story of our bureaucracy.

RAJAGOPALAN: This is the Annie Hall, right? 

MURALIDHARAN: Exactly. It’s more that the state is too small, but the design is horrendous. I’m facing the opposite problem. Maybe this is a humble brag, but the opposite problem is people are saying, “The content is good, you should write more,” and then it’s like, “The book is so big.” 

RAJAGOPALAN: No. I’ll tell you my charitable answer to that, and my slightly cheeky answer to that. Of course, you’re right, the book is already enormous, and there’s only so much you can fit in it, and so on, so forth. My cheeky answer is, I felt at various points, you either held back because you thought this is politically too difficult to do, or because there isn’t enough of a robust, empirical literature on it, because you are really keen on the empirical literature.

Sometimes that’s where my disappointment kicked in, because I was like, “Karthik is a fabulous theorist too. He knows this stuff. He knows this really well. I know that he knows this really well. Why isn’t it there in the book?” I was like, that’s probably because we haven’t done big empirical studies. We haven’t done the different state comparisons on which one’s devolved more versus less, or which states have better working finance commissions and so on. That’s why the focus is so much on streamlining, because there you have evidence for that marginal rupee expenditure. Am I on the right track?

MURALIDHARAN: Yes. No, absolutely. I think that’s exactly right. Even with the decentralization chapters, you will probably pick it up. It’s not just that I’m summarizing research, you can almost see the joy I’m feeling when I find a good study I need, that then allows me to be more confident about the claim that I’m making, because I think—and we can talk later about the macro and the micro—I think the macro is an incredibly powerful way of having the frameworks. I think one of the things I’ve tried hard to do in the book is discipline everything by saying, is there good evidence on this? 

RAJAGOPALAN: I want you to go a little off script, actually. I feel like you’re almost tying your hands too much as an economist. It’s almost like if I can’t back this perfectly with a gold standard study which was published well and done by credible people, I’m not sure I’m going to venture into it very much.

MURALIDHARAN: Yes, also, listen, I think there is a course, and maybe once that this book is done—I know Amit has said I should be writing blogs and newsletters and stuff like that. There may be a course for a continued conversation that says, “Listen, we don’t have the perfect evidence but based on my considered judgment this is what I would do.”

RAJAGOPALAN: Yes, we have lots of theory on it, which also economics is a fantastic provider of both theory and rigorous evidence.

MURALIDHARAN: Absolutely right. I think in a way the book is already way more than what most economists would do. So even what I’m doing in chapter two or three of these grand narratives of the Indian state and stuff, that is taking well-founded micro evidence and then placing this in macro framework. I think I’m already venturing beyond the comfort zone of just a well-identified microeconomist, but this is the level of venturing which I felt was appropriate for what is—I think one of my discussants somewhere called it a magnum opus because it is my grand synthesis, and part of my scholarly brand, if I may say so, is that—

MURALIDHARAN: —is that rigor. Anyway, I’m not ruling this out, maybe that’s why we do the podcast. We do the podcast, we do all these things.

RAJAGOPALAN: Also, maybe the next book, because there’s a lot of work to be done through, say, comparative studies, case studies, which are not exactly in the same realm. 

Welfare Expenditure

Let me move to another part which you talked about, which is on welfare expenditure. If we go back, again, you’ve taught public finance, so you know this stuff. I love the thumb rule that Vincent Ostrom gave us, which is you want the level of government to be only as large as the extent of the externality that it’s trying to solve. The larger the externality, the higher the level of government that should be equipped to solve it.

That’s an excellent thumb rule for externalities, public goods and so on. Why not also extend that thumb rule to something like welfare entitlements? For instance, one of the things that you talk about in the book—and this is in a fair amount of detail on the expenditure and also in other chapters—is that in poorer regions, most of the expenditure gets captured in some way, even the stuff that’s not leaking out, which is like the fertilizer subsidy, the free electricity, water, so on. Most taxpayers, at the local government level, wouldn’t want their local government to spend on fertilizer subsidies and free electricity.

They would demand health and education and public sanitation and public goods and so on. When the system is centralized, that gets broken. Here we have a wealth of literature globally on this. I know we don’t have very, very rigorous studies on India, but we do have a wealth of literature globally both in public finance, in political science, plus we have some very, very solid theory from folks like Vincent Ostrom or Richard Wagner and so on. I guess that’s what I’m asking, even for welfare expenditure, why is that not an equally useful lens to look at? I understand it’s not in the book, but what would be your way to think about that?

MURALIDHARAN: That’s a very good question. The answer is very simple. The answer is, the fiscal federalism—and this applies even to that earlier point you were making. See, I think the fiscal federalism framework is a very powerful framework, but it is a deeply conservative framework in the sense that its basic, its fundamental existential flaw is that it wants to sequester revenue in narrow areas, which essentially completely bypasses the question of equity and distribution.

Coming back to the federalism chapter, the U.S. is this deeply conservative constitution because it’s a bottom-up federalism where the DNA of the entire society was local communities solving their problems and then reluctantly giving some power to the federal government for defense because recognizing that there’s economies of scale over there. Which is why the U.S. Constitution is wonderful in every way except it doesn’t really care about justice or at least distributive justice. 

When we talk about one of the reasons the Indian constitution is then so over-centralized, reflects the fact that local elites were just not trusted, and even today. The moment the U.S. tries to implement the Voting Rights Act, you need federal enforcement, which means it doesn’t happen in local level. I think that is the original reason for why everything in the Indian structure is so much more centralized, because the moment you care about equity and justice, then there’s both fiscal and social reasons for why that gets aggregated. Which is why what I’m doing here is being aware of that history.

In saying that therefore the practical way to get the level of decentralization we want is that we are never going to get away from this problem of needing taxation at a higher level of government and redistribution for our horizontal equity needs. Like you said, why are we doing the subsidies? Why are we doing this? Why are we not doing education? I think within that larger fiscal and constitutional framework that we have, that this is a constitution origin story that tried to build in equity and justice into its original moment, at least in intention, if not in practice.

There is no way, I mean to say, even today, when you think about horizontal devolution in the finance commission, Bihar would not be able to fund its education system without its central transfers. Now, it’s true that it’s massively inefficient and therefore there are ways to then make those transfers more efficient. Part of the thing I’m talking about in that chapter, for example, is to say the problem with centrally sponsored schemes—and this is getting increasingly difficult as the divergence across the country is growing—is that the logic of a centrally sponsored scheme is to make sure that every state has a certain baseline of spending.

The problem is that the norms are often benchmarked for the more laggard parts of the country. That doesn’t work for a Kerala, Tamil Nadu. One way to do that is to have sectorally tied but line-item untied grants that say, as long as it’s spent on education, because I do care that a certain amount of education is happening. That’s why even everything I’m saying about decentralization is what I call a nuanced decentralization that recognizes the use unique additional objectives that the Indian state is trying to do, which, frankly, was just not part of the lexicon of when we think of Switzerland.

These are deeply decentralized societies to begin with, that grudgingly gave up some power for defense, which is why the origin stories are different than the Indian one. I think my approach takes that path dependence as given, and then saying, given the path dependence, how do we get toward a better place?

RAJAGOPALAN: Here there are maybe a couple of more follow-up questions. I agree with you on the origin story. I would add one thing though. I think a very big part of the Indian origin story is not just this focus on justice or equity, which we say was not there in many of the other origin stories, but I think there’s a very big emphasis on uniformity. I don’t think that’s gotten us anywhere, because the attitude is we need uniformity between Kerala and Bihar, and we don’t care. We’re willing to trade off uniformity for agency. We’re willing to trade off uniformity for some kind of efficiency, especially when it comes to how we spend the marginal tax rupee and so on.

I think our obsession with uniformity is what has landed us in this mess. Because you’re absolutely right; there are many other ways of supplementing a particular region or a particular state which is lagging behind. The United States, you have Pell Grants and federal subsidies and those sorts of things. In India, you can very easily imagine a direct transfer for education. You can imagine a voucher system. You can imagine just a direct central scheme which will in some way supplement schools that have better learning outcomes or PISA scores. You’re the expert in this. There are like 50 different ways to figure out how to target that, but that’s a second-stage targeting issue.

You’re right; in the first stage design issue, our original design is now so complicated that you are stuck with that. I understand that you’re trying to maneuver and find some efficiencies within that and somehow I feel so weird that between you and me, I’m the radical one because I just don’t think of myself as that radical person, but clearly in this conversation, that’s true. You’ve taken that there is an existing Indian state. I don’t want to burn everything to the ground and start from scratch, so given what we have, let’s make some changes here and there and maneuver. There’s a lot of merit in that.

MURALIDHARAN: It’s not just merit. I think the one place I want to slightly push back on is the sense that, okay, therefore, this is only an incremental agenda that can only get us a little bit. Right? I think part of the point—

RAJAGOPALAN: No, it can get us a lot.

MURALIDHARAN: It can get us a lot.

RAJAGOPALAN: Depending on which area we’re looking at.

MURALIDHARAN: See the other thing that we should not overlook, right, is, see, part of this is, again, the chicken and egg of low state capacity. Right? I think one of the biggest chicken and eggs is like, see, we are a low-income country, right? The state capacity itself requires a certain amount of fixed costs and investment. In the original period, you don’t have the capacity, even today. Let’s take today, 70 years after independence, 75 years after independence. 

I remember K.P. Krishnan saying this, when he was doing land, he was additional secretary land. He said something like, technically land, the central government, there is no constitutional place for it. It is a state-level issue. He said every state-level counterpart was delighted when there was a central scheme. That’s because the state-level bureaucracy doesn’t have the capacity to then think through the issues. That is this delicate balance again. That is even worse from state to center.

Now it is chicken and egg. It’s chicken and egg in the sense that you don’t decentralize because you say there’s no capacity and you don’t develop the capacity because why bother with capacity when you don’t decentralize. Which is why, again, it has to happen slowly, right? And we see this with NREGA; Panchayats have now become more powerful because there is real money that is flowing down that pipe and therefore that gives a certain amount of discretion. I think that’s why it’s interesting. I think my goals are often quite liberal, but I’m very conservative in a Burkean conservative sense. In the sense of saying that this is a large, complex ship. When you try radical things, it doesn’t usually end well.

We have to go slow. Which is why even my ideas of bureaucratic reform, every set of reforms will often say, “Listen, let’s grandfather what’s already in, but let’s think about the margin. On the margin, you can start moving this 800-megaton aircraft carrier in a more productive direction.”

Personnel for the Indian State

RAJAGOPALAN: Here, one of the things that keeps coming up is the bureaucracy. Actually, this is there in a couple of chapters in detail, but it’s there throughout the book again: education, health, everywhere. It feels like half of India’s development problem is an HR problem. We need to make you the human resources czar of the Indian state or something. Or at least whoever is that czar needs to imbibe the ideas in this book.

I’m very keen to talk about personnel within the Indian state because, okay, you and I might disagree on what is the role of the state. Let’s say that maybe we both agree that fertilizer subsidy is a bad idea, but there might be some other area where you think a particular welfare entitlement is a good idea, and I may think it’s a bad idea or vice versa. Whatever the role of the state, we know that we need to be able to staff it efficiently. We need the best quality people at each level.

We need the least amount of mismatch or misallocation between the talent and the actual role. Somehow, we need to accomplish all of this at the least cost to the taxpayer, given our ambition. Not just make it a self-serving enterprise where the bureaucrats are like the elite English babus who are going to be a level up above everyone else. You’ve made some fantastic suggestions in the book on how to create, say, competence-based career recruitment and career progression, how we can have practical training programs, how we can have apprenticeship programs, how we can have short-term contracts. How can we create an empaneled pool of UPSC and other state government exams where you’ve already managed to find the right people?

Now, why not just put them in a larger fence or a pool, and then people can hire out of it? Just like so many suggestions, I don’t even think I covered half of it. First question I have is, just for those who haven’t yet read the book, how is it that we actually recruit and staff government officials in India, right from the most elite civil services to the ward and branch officer?

If you can just give us the CliffsNotes version for the listener, and then you can talk about some of the fixes and ideas. Because I think those are some of the places where I feel like we’re going to get the maximum bang for the buck from your research. Right? That’s not a two times, three times marginal change. It’s like 100 times change because it can dramatically transform how award functions or how primary care health center functions and so on. Again, this is a two-parter. Sorry, my questions are so long, Karthik. The book is so long. Yes. Again a two-parter. Now I’ll leave you to it.

MURALIDHARAN: Thank you. In fact, I really think in fact, that chapter 5, the personnel chapter, is one of my favorite chapters in the sense that I think it’s not just because it’s got a lot of research. I think that’s perhaps the chapter with the most counterintuitive set of insights, like for a lot of people who don’t get into the inner innards of the Indian state. There is just so much basic misconception, including the sense that, “Oh, we pay too little, the state is bloated.”

It’s exactly the opposite. Anyway, so I think the factual question is easy one. The factual question is that for most government jobs, we select by exams, and this reflects the sense that the process needs to be objective and fair, and that exams were how mandarins were selected, over the years. We’ve just maintained that system. That’s why there’s so much to discuss in personnel, which is why I broke it into two chapters.

There’s the chapter on bureaucracy, which just describes certain structural issues of the bureaucracy overall and ends on saying, what are the political actions that need to be taken for state capacity? Then the personnel chapter gets into the weeds of ideas that can be implemented by bureaucratic leadership themselves. There is so much low-hanging fruit. Now before I get there, the other thing I want to highlight is that the book is also, in some ways, it’s an ode to research because we have learned so much in the past two decades.

The quality of empirical analysis of personnel policies, we didn’t know much of this stuff before. The Bloom and Van Reenen agenda on measuring management practices of getting into the black box of what is the A that you put in the production function. What Solow famously called the Solow residual was a measure of our ignorance. Okay? You can see that part of the management research agenda is to say, “Can we get into that black box of our ignorance and start putting some structure so that there are things we can actually work on?”

One of the very robust results from that management literature is that by far the most important—they measure management in very systematic ways. They do this for operations management, personnel management. I think there’s other dimensions, but it turns out that personnel management is by far the most important driver of organizational effectiveness. That’s fact number one. Fact number two is that personnel management in the public sector is like standard deviations worse than in the private sector.

Then once you put that together, that says, okay, this is hugely important for us to work on. The answer to your first question is just that we recruit with an exam. Let me go a little bit more than that and just explain why the status quo is dysfunctional. The dysfunction of the status quo comes from the fact that because government jobs are so lucrative, both at the very, very high end, for the very top IS officers, you are being underpaid relative to the scale of your responsibility and what that talent would command in the private sector. But once you go one level below the top, essentially the public sector—

RAJAGOPALAN: It’s 7.2 times, 6 times for teachers, it’s bananas.

MURALIDHARAN: The public sector multiplier is just insane. Which is why you got to see that figure to believe it.  The reason this is so important because it matters not just for the state, it matters for the overall economy. The broader distortion of talent that creates and that then feeds into the education and the skills and the jobs and all of those chapters. The reason I didn’t break this book into two volumes is because everything Those cross-connections is what makes the whole thing an intellectual enterprise that I’ve enjoyed doing. Coming back to the core issues of personnel. The basic problem of the Indian state is that you pay too much and you hire too little. Now, so people might think that, “Wait, if I’m paying well, do I maybe at least get good talent?” That’s where I think the research becomes important because it turns out that actually paying too much is incredibly perverse and can hurt public welfare.

There’s many reasons. Okay. The first is just that when I pay too much, I can’t hire enough. That’s just the basic reason. We’ve got studies now showing that unconditional doubling of pay of incumbent government teachers—that was Indonesia, but I’m sure we find the same in India—had zero impact. Whereas high, even modest amounts of pay linked to performance, or even just hiring more staff, with that same money, gets you significant improvement.

First challenge is just that your core problem of state capacities, you don’t have the money to hire enough because your incumbents have taken all the money in these pay commission increases. It’s more perverse than that. It’s more perverse than that because the government is by far the most lucrative employer in town. You end up with this kind of spectacle of hundreds if not multiple hundreds of thousands of applicants for every government job. That then creates a whole bunch of downstream problems.

Everybody looks at this and says, “Oh, this says that the private sector is not creating enough jobs.” No, it’s not about the private sector jobs. The private sector is creating market-clearing conditions. It’s the government that’s distorting the market. That is, I think, one of the simple aha moments where the connection between the perverse structure of the public sector labor markets and the inefficiency in the overall—these general equilibrium effects is not something that people have actually thought about or analyzed enough. But it creates so many problems because what you have is you have candidates taking government job exams because it’s like a lottery. Your chance of winning this lottery ticket is one in 300, one in 400. It’s rational for the individuals to take as many.

I think there was a print article that Shekhar Gupta tweeted just yesterday, saying this guy who’s been taking exams til he’s mid-40s. It’s a tragedy both for the candidates, but it’s also a tragedy for the government because what happens is then the people who are getting into a job often have no interest in that job itself. You’re only there because it’s a government job. People are attempting everything from being a forest guard, to a teacher, to a railway clerk and it just doesn’t matter. It is, do I get a government job?

The cost of that is you’ve then got a huge mismatch between what people are intrinsically interested in and the job you’re hiring them for a lifetime. From a citizen welfare perspective, we have designed the worst possible system because people put in all the work before you join and the day you join, you have lifetime employment, at which point you’re set for life. It’s perverse from a selection perspective, it’s perverse from an accountability perspective, it’s perverse from a corruption perspective because when the stakes are so high, it is not surprising that you have a whole bunch of these recruiting scams because it’s too lucrative.

Then you combine with lifetime employment, then you’ve got this mismatch between skills and jobs, and you’ve got this general equilibrium problem of the entire labor market. Because remember, so why do we have an education system that produces rote learning and produces exam-taking skills? Why do we have 80% of our graduates who are essentially unemployable in the private sector? That’s because the most lucrative employer in the economy cares only about exams and not skills. That is the downstream demand that shapes the upstream education system itself. 

RAJAGOPALAN: The entire labor market, especially for the tertiary sector.

MURALIDHARAN: Everything. Today, I’ve talked to people who run skilling, and people drop out of skilling courses because it’s too hard. It’s not that the private sector is not creating good jobs. The private sector is creating market-condition jobs, and the government has distorted the labor market by creating these completely lucrative jobs. I think there’s a stunning number from the study in Tamil Nadu by Kunal Mangal. They estimated that over 80%—

RAJAGOPALAN: Yes, he’s been on the podcast.

MURALIDHARAN: —80% of the unemployment among educated youth in Tamil Nadu is attempting government exams. I saw a recent—


MURALIDHARAN: —LinkedIn profile of somebody who literally on his LinkedIn profile said, for 10 years, 10 years, the status was UPSC aspirant.

We’ve set up this incredibly perverse system that’s not serving the state, it’s not serving the candidates and it is not serving the economy. Why are we doing this? This is why I’m saying that in a counterintuitive way, reducing the stakes of the government job lottery will actually be better on multiple dimensions. This is why it’s so counterintuitive.

It’s a little naive to say, “Oh, people should not be aspiring for UPSC, they should be doing other things.” You’re starting to see some of the public discourse emerge around the UPSC structure, but I don’t think anybody’s actually put it together of how all of these pieces come together. Like you said, what I’ll also tell the readers is each chapter is like a mini book. You just have to read 30 pages at a time, and you don’t have to read all 600 pages, but just the chapter 3 will then give you the sense of, here are the problems and now how do you fix it? 

RAJAGOPALAN: Before we get to the fix, I think there are also second- and third-order effects, some of which you hint at, but we didn’t talk about here. They also clog up the education system because you need certain minimum qualifications. Also, this was my experience when I was at Delhi University at the faculty of law, getting a law degree, a lot of the people will enroll in LLB because even though they have no interest in becoming a lawyer, because you know it will count toward one paper in the UPSC exam, and if they can’t clear UPSC, they can join the judicial service or something like that.

If none of those get cleared, at least at the end of it, you have something to show for it. You can go back to Bihar or UP or Rajasthan with a law degree so that your parents don’t think this was a complete waste. We also have this bizarre second-order effect where Delhi University should be producing lawyers, or at least people who wish to be lawyers, not just handing out LLB degrees to anyone who needs to do this as a byproduct of writing exams or something else. A lot of the MA/MPhil courses in India are exactly clogged up for this reason. A lot of the teacher training requires additional degrees or a master’s degree or an education degree.

Just to get that hike in pay or to qualify for that promotion, people are just clogging up the state education system in this bizarre way. I don’t even know if this is a second-order effect or a third-order effect. Like you said, when it causes distortions, it’s so big, that distortion.

MURALIDHARAN: Frankly, I’m not bothered about the fact that they’re in an education program while taking the exam. Okay, that’s fine. What bothers me is that there is no real skill being accumulated in that period other than just taking exams. Let me maybe just talk about one of these solution ideas or two of these solution ideas. It is a Gordian knot. It is a wicked problem.

I guess the problem is you can’t say, “Oh, we are paying too much. We are going to reduce pay.” That’s a political nonstarter. You can’t say that I’ve got a long left tail of people with lifetime employment who are not upgrading the skills and are not working. That’s a political nonstarter. Which is why you’ve got to combine this with the art of the possible and saying on the margin, every new cohort is where you can actually change this.

Now, for the incumbents, you can do some things like I’m talking about. I think even though my own work has shown that performance-based pay can be highly effective, I’m also realistic that that’s a proof of concept study that shows how much slack there is in your day that you can get so much. But in practice, implementing that in government is next to impossible at scale because of the measurement challenge. You can at least do competency-based promotions because that’s fully in your control.

This is what, to be fair and give it credit, that’s what Mission Karmayogi I think is trying to do, and eventually heading down that path of saying that let us create these competency passbooks for every government employee so that there’s a set of skills, that there’s continuous education, there’s upgrading, and let’s at least do that.

I think coming back to two or three of my important reform ideas, one, is the practicum. I think the practicum, I would say it’s perhaps the most important idea in the book in terms of how it cuts across so many sectors. It channels a body of research that says, “You can have locally hired contract teachers who are paid one-tenth of the government teachers and who are as effective, if not more effective.” This is what multiple studies show.

Here’s the problem. The problem when you try to “just scale that up” which some states tried, it gets into three difficulties. The first is the professional difficulty, which is the APEX certification institution, say in education, like a NIEPA or NCERT, who are the APEX, they absolutely viscerally detest the idea that you would deprofessionalize education and get in a bunch of people with no teacher training, which means it’s dead on arrival. Because whatever the economists may say, those are the bodies that have a seat at the table every time these policies are made.

Remember, the secretaries come and go every two years, but these guys are there forever. You cannot win without coopting and working with those institutions. That’s problem number one. Problem number two is what you would say is the legal problem. Because you tried to do this. The mistake was say with shiksha karmis in Madhya Pradesh is that then governments tried to do this in the cheap. They said, “Okay, we will not hire regular teachers and we’ll keep hiring these informal teachers.” Then it’s unsustainable because you go to courts and saying, “It’s equal pay for equal work. We are side by side doing the same job. Why should I not get regularized?” One fine day, the court will pass an order, and then you’ve got a problem.

The third is the political problem, which is you’ve got the pressure cooker building up of these contractual workers, and no politician can resist then at the time of election to say we are going to regularize you. Then you get the worst of both worlds because you don’t even have the qualifications or the exams and you’ve gotten regularized. Which is why the approach I’m proposing is to say, listen, as a professional educator, a professor, I believe in training. It’s not like I think credentials don’t matter. It’s just that the nature of our credentialing, why does all the research show that having a teacher training qualification has no impact on your effectiveness in the classroom? One simple reason is that many of these degrees are fake. If they’re fake, that obviously doesn’t matter. 

RAJAGOPALAN: They’ve never seen a classroom during the training.

MURALIDHARAN: The second one is even when it is not fake, even when it’s genuine, a lot of the training focuses on history, theory, philosophy, sociology of education, as opposed to the practical aspects of teaching and managing a classroom. Then there’s obviously aspects of incentives, accountability, et cetera. Long story short is that training is good. Training does matter, but what the global evidence suggests is that for service delivery functions, the most effective training comes on the job. It comes from doing it and ideally doing it under the eye of somebody with more experience who you can just constantly turn to.

What I’m proposing is a very, very, very simple modification of saying, “Let us not think about this as changing public sector recruitment. Let us think about this fundamentally as changing skilling in human capital,” because when you think about the future of jobs in a world of automation and AI, where are the nonautomatable jobs? They are in service-intensive areas that are physical that require the human judgment, the situational awareness in the classroom, with the patient, in a law-and-order situation, in a conflict situation.

Those are the places where it’s not just about jobs. Think about where does human intermediation add social value. It’s going to come in those service sectors. If you’re skilling, if you think about this agenda, as it’s an agenda of skilling for the service sectors, that is going to help both the government and the private sector because these are all jobs where the private sector employs four times as many as the government.

If you say that tomorrow, we are going to launch this practicum-based credentialing programs where after 12 standard, like in many cases, you just say, “We’ll take the top candidates in every Panchayat, bring you to the district headquarters for three months of modular training. Then there is nine months of practical training that happens in your home village where you are just posted in your school as an apprentice.”

You do this for a maximum of four years. Then at the end of the four years, you have gotten a degree. It is a degree, it is a credential with much better actual human capital because you’ve learned on the job. What you’re doing is, you’re paying a stipend which is much lower than the pay that is tenable in every way because even legally, this is now ring fenced within your training and capped it four years. It’s not like you can be exploited for 10 years on that arrangement, it’s part of that training.

Then the key point is to say, at the time of regular recruitment for government jobs, you don’t change that, but you provide extra points for every year that you’ve spent in a practicum. The practicum doesn’t guarantee you that job. There is still that exam-based selection, but there is now weightage on actually having done this job. 

RAJAGOPALAN: As opposed to just mindlessly taking exams.

MURALIDHARAN: Instead of mindlessly taking 25 exams. The person who is committed to a sector and has shown that commitment by doing the practicum-based training has a significant advantage in the recruitment process. Then finally, even with this, there’s not going to be enough government jobs for everybody. The key point is that these are all sectors that also have large private sector schools. Then what you’re doing is, you’re working with the private sector to get this credential recognized so that it is not seen as a hiring program. It is fundamentally a skilling program, but where the apprenticeship and practical training is happening in the context of public facilities.

That allows you this win-win-win of saying, how do I augment state capacity and staffing at a fiscally affordable way, in a way that is also boosting training? In a way, if this practicum is local, I think this is a particularly big deal for young women, when we think about female labor force participation.

RAJAGOPALAN: Absolutely.

MURALIDHARAN: Study after study shows that you’ve now got in almost every state in India villages with high school past women, who would like to work, but for a variety of sociological, patriarchal and safety reasons, don’t want to leave the village, correct? How do you create those jobs? I’ll give you the stunning statistic. Tamil Nadu—and I talk about this in the education chapter—post-COVID, Tamil Nadu did this probably one of the world’s most ambitious COVID remediation programs where they hired 200,000 young women at a stipend, an honorarium of 1,000 rupees a month to do 60 to 90 minutes of after-school training.

We show that the program is highly effective, cost-effective, improved equity, but what is really stunning is that even in a stipend of 1,000 rupees a month, you had four applicants per job. This is Tamil Nadu, this is not even UP or Bihar[.

RAJAGOPALAN: Absolutely.

MURALIDHARAN: Then the qualitative interviews, when you go talk to these women, what they’re valuing the most is the ability to leave the house and the ability to have that income.

RAJAGOPALAN: Get some skills.

MURALIDHARAN: Have that income, have that respect, have that standing in the community that comes from then being that kind of local service provider. That is such a win-win-win in terms of the untapped human potential that we are sitting on. It can mean that just a little bit of creative unlocking. 

RAJAGOPALAN: I knew these women naturally have excellent core skills because women in India are raised to have excellent core skills. They can solve problems. If the electricity goes away, if there’s a crisis in the school, these women are basically fundamentally solid at doing that sort of thing. They’re not going to be deer in headlights. I feel like that’s another part of it, but my favorite part of your practicum program is actually slightly different.

This is a little bit off-tangent. I was reading John List book, “The Voltage Effect” and how he talks about when you get excellent results in a pilot program or a randomized control trial, and then you try to scale it, you may not be able to scale because you’ll have what he calls a voltage drop. This is a classic teacher training example. When you do a small little pilot in one particular district or one particular suburb of Chicago, you can hire the best school teachers, but if you need to do this across the state of Illinois—ding, ding, ding—we are not training enough high-quality teachers at a scale and at a quality that now you can actually deploy them across all the kindergartens in Illinois.

We are going to have the exact same problem in India. What I like the most about this program is how scalable it is, not just from a fiscal point of view, but because here you don’t have those standard, inelastic supply problems, which you have virtually everywhere else. You may also have inelastic supply problems in other countries where the median age is 10 years or more than India, like in the case of China or somewhere else. In India, we’re not going to have that problem.

You are not going to get the voltage drop because it’s literally happening at the lowest level where there’s this massive elastic supply of labor with core skills, may not be sophisticated skills, but core skills. That’s actually my favorite part of what you suggested but you haven’t talked about that.

MURALIDHARAN: In a way, what you’re saying is you’re bringing back the Lewis model here. You’re basically saying that there is an infinite supply, there’s elastic supply because you’ve got this demographic. 

RAJAGOPALAN: Absolutely.

MURALIDHARAN: I’m glad you caught onto this because this is something I do write about in some of my papers on scaling. I think the reason I feel particularly confident, and this is not just about the practicum, the other no-brainer, low-hanging fruit, which I would recommend that the new government does the first thing literally when they come up is to literally hire a second Anganwadi worker everywhere, and that’s because—and again, all of this is coming from high-quality evidence. We’ve got this large-scale RCT, in fact just came out in the JPE, in Journal of Political Economy, this month, and this is on 320 Anganwadi centers, four districts in Tamil Nadu, 160 randomly selected to get this extra worker.

You get this stunning improvement in learning outcomes, but also a reduction in child malnutrition because you freed up the time of the existing worker to spend more time on health and nutrition. We estimate that the ROI on this is between 12 to 20 times the cost. That’s a classic example of a quadrant one expenditure. That improves equity because the children who are benefiting are the most marginalized and has a 20 times ROI. If you’re a private investor and given that kind of opportunity, you’ll say, where do I put more money?

We are not doing it because all of the money is going to unconditional salary increases of the incumbents, because as you are a public choice person, you will be delighted to know that I agree with you that the political economy is like all the powers with the incumbents who then grab the lion’s share of the money. Sorry, but coming back to this point about voltage effect, the reason I cite that study is because it exactly did what the scale up would look like, which is on the extensive margin. Hire one extra person in every village, and that labor supply existed.

What the Tamil Nadu Illam Thedi Kalvi shows is, you could do that for two lac women, and they existed. There was four times as many applicants. So here’s the basic point: Of a public sector pay so out of whack that at the current rate, you get like 200 applicants per job, you could reduce that wage by a fifth and still get 10 applicants per job. And I think the other subtle point, which is underappreciated in our distorted personnel policies is, yes, if I offer a wage that’s 10,000 or 8,000 as opposed to 40,000, yes, I’m going to get less qualified people.

The problem in our system is we have obsessed on paper qualifications and the idea of getting “the most qualified person” even though they are not connected to the community. The person who’s passing that test and even if they’re local, once you’re paying 40,000, 50,000, their aspiration is to be urban. Where does the absenteeism problem comes from? Half of it comes from the fact that everybody is doing “up down.” Here’s the thing, when you talk about absence, people say, “Poor things are teachers live so far.” I’m like, “Boss, the endogenous question is, where you live is a choice.”

RAJAGOPALAN: Why do they live so far?

MURALIDHARAN: Where you live is a choice variable. That’s not something that is fixed. Again, we have evidence on all of this stuff. In health, one of the most effective interventions of the past 30, 40 years has been the ASHA program. The code of the ASHA program, which is what Dr. Abhay Bang and Rani Bang developed in Gadchiroli, in Western Maharashtra, backward area, tribal area.

The whole point was that instead of saying that we want highly skilled doctors to come and relocate here, how do you upskill people in the local community who are invested in the community and having that faith in the community because they have skin in the game, okay?

RAJAGOPALAN: They saved us during COVID when you couldn’t have migration and when you had all these other problems. The ASHA workers are the absolute heroes of India’s COVID outcome, especially in the better districts.

MURALIDHARAN: Everywhere. I think the point I want to make is not, yes, of course, when you have the capacity in times of crisis, the capacity saves you. The more systemic point is that there are two or three keys to that model, which is local talent. Invest in upskilling the local talent, and a compensation system that is actually a piece-rate because you’re paid on institutional deliveries, paid on vaccinations.

Better Approach to Skilling

That gets to topics in chapter 12 about the dysfunction of the—anyway, listen, there’s a lot going on here, but I think this is one simple idea that then has applications in education, applications in health, applications in policing, which in fact may be one of the highest-return areas, with the same approach. Think about, sometimes I joke in saying the politically viable slogan, and maybe there is some Hindi-speaking, chief minister who’s listening to this, is literally what I want is a Teen Devian program. Of course, it was this famous Dev Anand movie.


 MURALIDHARAN:  The Teen Devian  program would basically just say that every village or every habitation would then have these practicum-based apprentices in health, in education and in police. You could even over three years or four years, imagine giving them uniforms with one stripe, two stripes, three stripes, that tells you which year in your program you are. Because, again, we have such a status-conscious society that wants these little markers. These are ways of providing jobs, of creating huge social value and of creating dignity.

RAJAGOPALAN: And skilling. Our human capital building is not happening in schools, so it has to happen somewhere.

MURALIDHARAN: Then you pay it forward because effectively, the contract that the government is then signing with the 18-year-olds is to say, “We are investing in you for you to invest in the children.” Then what you’re getting is this credential that is then going to be valuable both in the public and private labor markets. Frankly, for health, you could imagine that for the top candidates, they might qualify then for the credentialing that would be internationally valid.

Then where is the big global opportunity today? It’s in the caregiving jobs in aging societies. There is a global part of what India needs to negotiate with the aging West, which is to say, “Let’s have high-quality guest worker programs.” You don’t want the migration because obviously there’s political costs of migration, but there’s economic needs for workers in the caregiving economy.

There are hundreds of thousands of jobs in the caregiving economy, where if you can get the certification and the standards right and have a clear work visa program, where you come back in three years. It’s a three-year, six-year program. Imagine in this practicum, you’re starting all the way at the bottom, but there are prizes. In fact, there might be a prize bigger than a government job, because the bigger prize in the government job is to be so good that you get the foreign posting.

That requires being good enough that that entity that is going to credential you for providing those quality of, say, care services to the geriatric population in Japan or Germany or the U.S., that’s going to be at such a high standard. What you’ve done is you built an aspirational career ladder whereby people are being rewarded for real skills as opposed to rewarded for exam taking, which has zero returns in the labor market. There’s somebody very, very senior in the government who said, “If we managed to do this one thing in the next five years, this book would have served its purpose.”

RAJAGOPALAN: No, I really love this idea so much, and it has so much practical value. You can think about extending this, right? You’ve talked about Anganwadi workers, you’ve talked about teachers’ apprentices, you’ve talked about police apprenticeship. We can also think about this in the context of something like the short service contract programs that the government tried to start when it came to the armed services, right?

Now, again, that may not work well for someone who’s at the frontier or someone who’s going to be an air force pilot, but there are so many other jobs at cantonments to be done, right? You need a health worker at the cantonment. You need gardeners at the cantonment, right? You need so many other support services, which right now is just simply out of question because fiscally, it’s an impossibility to even suggest something like that.

MURALIDHARAN: I’m so glad you said that. Because I think, see, my own view on Agniveer is like, it’s about 80% right, but I would tweak it. 

RAJAGOPALAN: Yes, I don’t want this to be Agniveer, by the way. I want it to be what you are suggesting as an extension into the armed forces.

MURALIDHARAN: I don’t think the model works for fighting soldiers. It doesn’t work for fighting soldiers for all three reasons. Because I don’t even think it’s cost-effective, because the fixed cost of training somebody to bear arms is way, way, way higher.

RAJAGOPALAN: The skill levels, the recruitment, the training, all of it.

MURALIDHARAN: I don’t think it makes sense even economically, frankly, in terms of the cost of the training. That’s one. The second is, at the end of the day, being asked to lay down your life is the ultimate sacrifice which people do not do very often for some abstract notion of country, but for a concrete notion of my unit and my buddies. That sense of solidarity in the fighting soldier, I think, is very precious. The last, and I think what I worry about the most, frankly, is the military is about applying lethal force, and there is no civilian use case.

RAJAGOPALAN: We don’t want to train people in that. Absolutely.

MURALIDHARAN: There’s no civilian use case for the ability to apply lethal force. We have this long tooth-to-tail ratio. For all of the support staff, the cooks, the drivers, the gardeners, there you could do this for four years, and then it’s a huge boost because then the discipline of having been in the army and having a skill that is needed in the real economy then gives you a boost back out there. 

RAJAGOPALAN: I didn’t say Agniveer because I don’t want it to be that. The army employs so many people who are loading and unloading in difficult cantonment areas, people who are drivers of really heavy-duty vehicles, people who are actually doing security for the cantonment, people who are managing the supply chain of the cantonment, the pantry. These are really detailed tasks. I have been to army cantonments, and it’s like running a small city. You also have a school, they have a temple, they have so many things.

MURALIDHARAN: That’s exactly why I mentioned it, because that was the context in the military, right? I think we’re completely aligned on that.

RAJAGOPALAN: Yes. I was actually nervous when they put that program out for all the reasons that you mentioned, especially the lethal force issue. I have one of my colleagues here at George Mason, Chris Coyne, has talked so much about how we’ve imported the wars that the United States has conducted abroad, we’ve imported that back home because a lot of the weapons come back and then they get deployed at these local police chiefs and so on.

A lot of the veterans come back and then they work in the policing units and veterans get an advantage. Then you have this really big problem of police brutality and lethal force being used in the United States and all the downstream effects. Forget everything else and the fiscal and all. Just on the lethal force issue alone, I’m totally with you.

MURALIDHARAN: We’re fully aligned. Coming back to just bureaucracy, personnel, et cetera, I think the other thing which is different, which is this distinct idea. See, this approach of practicum, I think, works for frontline services. In fact, the one place which I thought you’d be excited is, in fact, it can also work with courts, where I have mentioned, where here it’s not a full-time thing, but just saying, let’s give every district—

RAJAGOPALAN: The apprenticeship model.

MURALIDHARAN: —give two legal clerks that you hire through a state-level exam. I think there was a Madras high court or somebody said that 85% of law school degrees in India are worthless. Then that again has the downstream effect of saying, because the government is the largest employer in town, you can shape the entire ecosystem of the skilling economy by recruiting for the right things. Saying, I’m going to have an exam to recruit for legal clerks.

Then these clerks, it’s a two-year clerkship that you spend—because it’s not easy to augment judge capacity, even if you have the fiscal allocation, because where do you find the judges? Correct? That way, this becomes a more practical way of augmenting judicial capacity for the research, the writing the first drafts and all of that, the way clerks work. Then that’s a two-year thing. That also has, I think, many of these benefits.

RAJAGOPALAN: I was less excited about that because some of our courts already have an informal clerkship program. The supreme court has a slightly more formal clerkship.

MURALIDHARAN: I’m talking about—

RAJAGOPALAN: Not in the district courts. You’re right. That I’m excited about.

MURALIDHARAN: 86% of cases are there. 

RAJAGOPALAN: That I agree. We need to train people to be the next district court judge. We need case management. We have so many tasks.

MURALIDHARAN: My focus is on district courts. Because again, see, elites focus on what’s happening at the very, very top. The bulk of those cases are sitting in the district courts, right?


MURALIDHARAN: Then anyway, I think you also mentioned this empanelment. Let me just talk about that quickly.

RAJAGOPALAN: One sec. Aside from this, the empanelment. I am most excited about empanelment. I’ll let you describe it first, but I’ll tell you which element of empanelment I’m excited about. Because when I was reading that part, I was like, “This is the best thing I have read so far.” Just briefly, the idea of empanelment is all these people are taking these UPSC and state service exams. We only manage to recruit the top 0.1% which is the cream. Then everyone else who took the exam, it’s a total waste. Now they’re going to go back and spend some time studying, but there isn’t that much difference in quality between the 0.1% and the top 1% or the top 2%. You can still use that. The fact that these people have cleared that exam is now useless. Your all-India rank is useless, nobody cares. Private sector doesn’t care. It doesn’t turn into a credential. It doesn’t turn into anything that can actually be leveraged outside of that exam. What you suggest is we create a pool, we empanel them such that governments across the board can recruit contractors from that pool. Then again, they have the same point system, and so on. I’ll let you describe the rest of it, but then I’ll tell you what I’m excited about.

MURALIDHARAN: Again, it’s just coming back to taking a step back about the book itself. The book, I think, what is unique about it is, and the reason again, this is just apologizing for the length, is that every chapter is what I call half science and half engineering. There’s the first half, which is saying, “Here are the facts. Let’s just understand the facts and the reasons. Now, let’s build a better mousetrap.” Here the core fact is that the state is hollowed out. We just don’t have enough people in the state. One measure of this is you go to any government office in state or center, they are strewn with consultants, all these consultants.

RAJAGOPALAN: Oh, my God, this is a nightmare in economic policy right now.

MURALIDHARAN: Everywhere. The reason is that it is both a cause and a consequence of the atrophy of state capacity, correct? If you are a secretary today and you need a document, you need something, you need a PowerPoint made, there’s nobody in your office who can do it. So effectively because you are under pressure—the consultants have many advantages. They are accountable, they are flexible and they have better skills. All of that is good. There’s at least three problems. The first is that they get used very tactically.

The second is the consultant is only as good as the RFP. You can do what you’re asked to do, but you can’t think for the government, you can’t think in the public interest. When you start doing that, then we get into other murky territory. Then the last problem is just the lack of institutional memory. As people rotate out, rotate out, there’s a project and then that is not the state. Then you’re actually further hollowing out the state because instead of investing in the capacity of your people because that’s a long-term project—it’s both a cause and a consequence, correct?

RAJAGOPALAN: I would add a fourth, actually, which is that a lot of the times, the consultants get their next gig for agreeing with the state, which is typically not the job of the people at the highest levels, especially the thinking jobs. There should be people who give some feedback mechanism, but to get the project, you have to say, “Oh.” Even if they don’t have to agree with the state, there’s this imagination that if I tell them what they want to hear—


RAJAGOPALAN: —if I please the client, then it’s good, which is exactly what’s happening.

MURALIDHARAN: They don’t have the locus standii to push back, frankly, right?

RAJAGOPALAN: Yes, absolutely.

MURALIDHARAN: Because at the end of the day, the government treats them as vendors for hire. You’ve been asked to make this report, you do the report. I want a nice flashy report that looks good. Multiple officers have told me, they’ve said, “Listen, if I want a good flashy report that I can inaugurate, these consultants will just crush it.” 

RAJAGOPALAN: If they want something good, then they have to ask Karthik to write the report. That’s basically the gap right now.

MURALIDHARAN: The problem is you can’t scale me. That’s why I’m building CEGIS, but that’s a different story. I can maybe talk about that some other time. Just coming back to this core problem. The core problem is that there just isn’t enough capacity within the government. There isn’t enough head count and there isn’t enough skills.

One thing we didn’t talk about the problem of the status quo is the problem of lifetime employment. That lifetime employment basically means that it is not just that you’re wasting money on people with obsolete skills and no incentive to upgrade the skills. Often, even if you’re willing to pay the salaries, their presence makes it harder to modernize. The best example I use of this is, say, the bank modernization. Banks couldn’t computerize because their existing staff were just too reluctant to do it. The only way the banks could do it was BRSing and getting the vintage cadre out and getting in a new cohort of people. It’s not just the salary. It is what you’re doing in terms of slowing down the technology adoption within the largest player in the economy because of these—

intra odd issues. The bottom line is consultants have huge—there is value. There’s flexibility, there’s accountability and there’s current skills. This is not taking away from that.

RAJAGOPALAN: Fiscally, they are cheaper than the lifetime employment of whoever you hire, to a large extent.

MURALIDHARAN: Of course. It looks like government is hiring consultants all the time. Once you integrate over that, it’s not obvious because of the margins you see.

RAJAGOPALAN: Presumably, it could be fiscally cheaper.

MURALIDHARAN: There’s at least flexibility, right?

RAJAGOPALAN: Yes, there’s flexibility.

MURALIDHARAN: I think the problem here is the different kinds of lateral entry have been talked about, but that doesn’t work, again, because there is a process, and departments often lack the fixed cost of paying for that. Then even if they do it, then you worry about irregularities. Are you hiring your friend, your nephew? We are always nervous about lack of transparency in due process in public sector hiring.

My idea is therefore very, very, very simple. Government desperately needs people. And we have no shortage of people who want to serve in government, but our current public sector recruitment, with an acceptance rate of 0.1% or 0.2% and if the average student studies for two years, they often study for more, the opportunity cost of every one person hired in government is over 1,000 man-years, or person-years—

MURALIDHARAN: —of time that people are spending studying for this exam. My simple idea is to say, instead of the top 0.1%, you take the top 1% or 2% and you just empanel them. Then you’re saying, “Once you are in the top 1%, you’ve already shown that in raw smarts, you are in the pool that is at the very top.” Now, let’s give you a foundation course. We’ll give you a three-month foundation course that could be virtual, with some field visit, some interaction, whatever. At the end of the foundation course, which is certified by UPSC and some combination of central training agencies, you’ve got a pool of empaneled talent, who then any government department anywhere in the country can just appoint on a three- to five-year contract. And that department is capable and fully empowered of having its own additional requirements.

Now, let’s say I want to hire in the tax department. Today, UPSC doesn’t care what your major is, but in the tax department, I would love to have somebody who’s got some accounting background, some finance background, some law background. In forests, I would love somebody who studied botany and zoology. Today, there is no matching at all between the course of study and your job. Now, what you have is the exam—see, the main value of the exam is because of the shoddy and variable quality of our degrees. The exam serves as a GRE/SAT type of common clearing ground to say, “This is a high-integrity exam. If you pass this, you are of a certain basic standard.” That purpose is still useful, but what it doesn’t do is not anything in terms of getting the fit between your own training, your own aptitudes and allocation of talent. 

RAJAGOPALAN: Allocation of talent.

MURALIDHARAN: Let us use the exam for what it’s worth, but tweak where it is not working. Now, what you do—

RAJAGOPALAN: It’s selection, not allocation. Then allocation can be done through other mechanisms.

MURALIDHARAN: Exactly. It is objective, you’re getting the objectivity and you’re getting the transparency in a way, and you’re paying the fixed cost because it’s a high fixed-cost enterprise. In fact, one thing we did not talk about is one perverse thing people wonder is, why are so many vacancies unfilled, even though they’re sanctioned positions?

RAJAGOPALAN: Oh, yes. That’s my favorite part.

MURALIDHARAN: It’s because nobody wants to run the recruitment process because the fixed cost of running the exam for an officer, there’s only downsides because there’s corruption.

RAJAGOPALAN: It’s crazy.

MURALIDHARAN: It’s crazy, so why would you run the process? Which is why you choose to do it once in four years, once in five years, sporadically. The politicians like doing that before the elections. Therefore, the nature of public sector jobs, instead of being about serving the public, it becomes about the benefit I’m giving to the people who got the jobs, which is again, completely the wrong way to think about it.

Coming back to the empanelment, what you’re doing is you’re getting all of the benefits of the UPSC, but then you’re saying, “Being empaneled makes you eligible for the job. It doesn’t guarantee you a job.” You still have to show up, you still have to be effective, you still have to be productive, you still have to invest in your skills and it allows departments to have additional requirements for their specific job. Because of empanelling 10 times, they will be the people with the zoology, with the botany, with those specific interests who want to then go work in the forests. 

Then what happens is because these are three- to five-year contracts and the renewal is not guaranteed, you have the motivation to stay current and keep your skills current. You’re eligible for all of the training programs and everything. Then you could imagine that maybe you have a three-year contract, and then a five-year contract and then a seven-year contract. Then after 15 years, you say, “Listen, you’ve done this so well. Now we absorb you.” Because you still need certain permanent positions.

RAJAGOPALAN: You get the same points that you were suggesting in the apprenticeship program.


RAJAGOPALAN: You can have a point system even here.

MURALIDHARAN: The other difference is because here, this becomes a way of encouraging people to rotate out of government. You do this for three years and five years and say, “Fine. I’m going to go to the private sector. I’m going to go to civil society. I’m going to learn different sets of skills. Having done my empanelment and having done this, I’m always in the pool of people who can then come back into government.” Then you build a system of lateral entry that is quality-controlled by the UPSC, right?

RAJAGOPALAN: Absolutely.

MURALIDHARAN: That then incentivizes the skills, incentivizes the experiences, creates diversity. Again, I just feel that this is such low-hanging fruit that you can easily hire five in a district. You go sit in a district collector’s office, they are harried. They’ve got like 50 departments to handle, no analytical capacity. You put like three of this kind of people in there who just be able to analyze data and there is so much low-hanging fruit because Devesh Kapur and Aditya Dasgupta have this very nice paper looking at staffing in block development offices and they’re all understaffed and the staffing really matters. The chicken-and-egg problem today is you’re understaffed, but you’re fiscally nervous about hiring because you know the moment they come in is the day they stop working.

You’ve got the Woody Allen problem which is why you have to rethink how you do this. Then if we have time, I’ll talk about the Telangana example that showed that this is not just a pipe dream. It is doable.

RAJAGOPALAN: It’s doable, yes.

MURALIDHARAN: I think the CEGIS part of my story of the past four or five years is the reason we built it was just to prove to ourselves that these are not just like ivory tower ideas, but these are actually doable in practice.

RAJAGOPALAN: Now my favorite part of this whole thing—so I’m obviously coming from this, from the judicial side of it. When I was looking into the vacancy problem, there are multiple problems. One is, of course, the fiscal problem and so on, but the other part is there are a lot of jobs where the budget allocation has been done. That seat has been created or that position has been created, but they haven’t even put a call out. They haven’t conducted the exam for that local engineer or the local forest officer and so on, and that vacancy will go unfilled for years on it.

One of the reasons for this, from what I understand, is entirely judicial, which is if you don’t craft the perfect advertisement for the exam, anyone can then challenge the exam saying it wasn’t fair. If you say something like, “We’re looking for forest officers and we want people who have BSc in botany and zoology,” someone will raise an issue and say, “Why not a BTech in botany or zoology? Why didn’t you add the BTech?” At the level of the advertisement, the entire thing is dead on arrival. Then you have at each stage, if you conduct an exam, then people can challenge the exam. The court will hold it up. You have all these problems.

The politician who actually created the job is very keen that someone from their party or caste or region or language actually gets that position, but they can’t do it because the moment they do it again, the courts will hold it up. My favorite thing about this pool of people who have been in some sense vetted and credentialed is that it makes this entire thing bulletproof in that it completely de-risks hiring. No one can be pulled up as a corrupt politician or a corrupt bureaucrat, or someone who put out an unfair ad, or someone who treated people who have BTechs poorly as opposed to BScs or government programs differently from private programs. That problem goes away.

Right now, we have so much risk and we put on so much accountability with virtually no reward on the people who are doing the hiring. It’s completely thankless job and no one does it. That’s actually my favorite part of this. My second favorite part of this, and now I am being a little bit tongue in cheek, is I think it’ll strongly devalue this whole over-investment in engineering and MBAs because the reason that’s happening right now is that’s another selection process for the private sector. For the public sector, the selection process is the UPSC.

For the private sector, even if they’ve learned nothing related to the job you actually want to hire them for, you’re doing private equity, but you want an engineer with an MBA or some kind in marketing or whatever because you think they’re credentialed. They clearly have some skills, they have some endurance, they have some ability to learn, we can train them for the rest of it. If people come with this kind of credentialing and some real-world experience for three to five years, the private sector would actually love to hire people like that. These are the people who will be running the district office for Hero Cycles or a tire company distribution or ITC distribution. All the things for FMCG is happening at the ground level.

I can imagine so many other misallocations we have in our economy, which goes away because we start creating a better system of credentialing based on the exams we’ve already taken, which is not as rigid as the Chinese merit exam system. This is adapted for Indian conditions. It is made in India, some Indian jugaad for Indian situations, accommodating preferences that Indians already have. That’s actually my favorite part of that. I’m a design person, and according to me, really cracked a nice design problem.

MURALIDHARAN: Thank you. I think that’s why the book, in a way, doesn’t fit any genre. 

It is because most academic books will not get into this level of practical detail and if you can have a practical thing, it’ll not be as grounded in the research. This is just my way of apologizing for the length. 

RAJAGOPALAN: I’m constantly teasing about the length. Actually, I read it really quickly. It’s very accessible. The second time I wanted to read certain chapters carefully because I wanted to go back and forth with the end notes, and the book is a little heavy and my very delicate wrists and lack of physical strength was hampering. If you have time for one last question, this is a little bit more about you and the meta question. Again, I think I’m going to sound a bit cheeky when I say this, but I felt this the most when I was reading the education and health chapters because that’s where you’ve done so many field studies, randomized control trials and so on.

The sense I got was that for the last 20 years, starting with young Karthik in the Ph.D. program till now, you are feeling out a really large creature. You’re trying to feel out that large creature from multiple angles with multiple different co-authors. Some people felt the flaps and thought that’s a bird. Some people thought it’s a mammal, but has two tails. You somehow, because you are so prolific and have published so many papers, you managed to feel all of it out and come to the realization that, “Oh, no, it’s an elephant. The flaps are not wings. The flaps are the ears of the elephant. This is not a two-tailed mammal. It’s got a trunk, and this is actually an elephant.”

Now coming to the elephant in the room, sorry for the pun, it almost seems like after all of this and feeling across every aspect of the failed state, you figured out that the development problem is at the state level. The engineering plumbing fixes that RCTs typically suggest, whether it is to fix absenteeism or service delivery or improving learning outcomes, or the million other things that you’ve done, that’s not going to solve the problem. Finally, what you need to do is fix the state and the way Lant Pritchett says that you’ve managed to figure out what is the problem at the ontological level, not just the problem at the surface or the minor level.

The Backdrop of the Field of Economics

One, I know I’m being cheeky, but is this a good way to think about how you had your own career progression and how this book has come about? If yes, how do you think about the micro-foundations, field studies, randomized control enterprise, now looking back?

MURALIDHARAN: I don’t think it’s cheeky at all. You characterize the intellectual journey broadly correctly. See, where I think I’ve been very privileged is that I also had exposure to the older ways of doing development. Which is when I was an undergrad, I had taken graduate development, which at that time that class itself was called structural transformation, historical perspective. I think the problem sometimes with the younger development economics Ph.D. is that for economic history, used to be a requirement in Ph.D.s has been taken out. History of political thought, history of economic thought essentially doesn’t exist now. Even at the undergrad level, it’s gone into social studies programs.

RAJAGOPALAN: It exists at George Mason. For the listeners who are listening, we have a field at the Ph.D. level.

MURALIDHARAN: Good for you. Again, I think it is true that those things were taken out in the sense that they were not core to the creating of the modern professional economist whose job is to go deep into specific areas and write papers. I feel uniquely privileged in terms of having had exposure, which is why you will never hear me rant against RCTs. You will never hear me ranting, but it is a very eclectic sense of saying, let’s take the best out of everything and just pull it together and just keep learning and keep learning and keep learning. It’s still amazing. It’s so easy to criticize RCTs.

I think the one thing I would say about RCTs is that I think framing the purpose of an RCT is saying, let us find out what works so that we can scale it up, I think is a mistake. I think a lot of the pushback probably comes, and rightly so, from, I would say the early framing errors. Like me saying, “Oh, this is going to be about finding out what works and that’s how we are going to advance development.” I think the way I have approached RCTs has always been to say, “What does the RCT tell me about what the binding constraint is in this particular setting?” Which is why the way I solve the external validity problem is I’m never trying to say, “This thing worked here and therefore we should scale it everywhere.”

The way I approach the RCT is what is the principle that is being illustrated by this high-quality study? Now, the value of the RCT—and I’ll give you the example from perhaps the most important paper I feel I’ve written in my 20 years, which is the paper on NREGA, where we found that improving NREGA implementation not only improved wages, but it also increased private sector employment. That goes contra to every ECON 101 we learn. We learn demand curve slopes downwards.

Now, the point is that this then goes back to Sherlock Holmes and saying, when the facts change, I change my theory. The value of the RCT is that with the well-designed experiment, you don’t question the facts. Then that sometimes make you go back and question the theory. I mean, in saying, so what is it? Then you realize that no, we are taught the new classical ECON 101 in our textbooks but under imperfect competition, labor market power, things can be different, and that’s what you illustrate.

So what I feel I hope is unique about the book, and again, there’s an intellectual timeline, there was an error in the early 2000s—it’s easy to criticize the credibility revolution, but the credibility revolution is coming after an era when people were writing papers like I just ran 2 million regressions. Empirical work had lost all credibility because you could get whatever result you wanted by whatever control you chose to put in. Which is why I think the field went through this phase of saying, “I would rather answer small questions well than big questions badly,” because what were the big questions? The big question was growth.

Growth is the big question, but you had a decade of a cottage industry of cross-country growth regressions. That was essentially garbage. That was the 2 million regressions. Everything was correlated and so you couldn’t say anything. Now, some things are more robust than others but I think that is the epistemic revolution the field went through, which then went through a longer time of better access to microdata, better computers, better computing power, and saying, “Let’s do it.” Then the experimental movement can be seen as a part of that way of saying, “Listen, let’s get credible estimates.”

What I think the top people in the field have always been able to do is to then kind of say, “Let’s look at each of these experiments, and then let’s synthesize what is the big picture that’s emerging from that,” which is never the same as mechanically counting studies and saying, “Oh, this is what we should do.” Because there’s no substitute for judgment and you do need the judgment. I think the way I combine the micro and the macro is I feel much more confident saying what I’m saying. I’ll give you a simple example.

We talk about decentralization. Theoretically, there’s clearly something as too much decentralization. Theoretically, there is an optimum amount of things that happen centrally in a decentralized way. The beauty now, and this is a paper that’s R&R of the ER by Jeff Weaver and Veda Narasimhan. It’s not an RCT, but it’s beautifully identified micro evidence of the impact of having a smaller polity using a regression discontinuity. Then that shows that a smaller polity had better outcomes in every way. Because we then see consistent patterns. 

I had a Ph.D. student and it’s a project I’m also working on. I’m just looking at this reform in Andhra Pradesh under N.T. Rama Rao, where you went from blocks to mandals, which is one of the most important governance reforms that’s not been studied enough, where a block serves 250,000 people, the mandal serves about 60,000 to 65,000. What we find is that on almost every measure of service delivery, because you’ve got this RD where villages now get closer to the district block headquarters, service delivery improves.

All of this then builds high-quality micro evidence consistent with this grander theoretical point that we are over-centralized. Therefore, I don’t see the contradiction and I think I said this in a chat with Amit once, of saying that I hope to take both a bird’s eye view and a worm’s eye view. That you’re both telescoping back to see what this big picture is and then zooming in all the way in on the details. 

I hope it’s every I guess senior scholar’s life is their own personal intellectual journey. I think to the extent that there are younger people listening and seeing what to do, there is a sense in which that the daily practice of being a researcher requires you to put your head down and just do your one particular problem really well. There’s a phase for that and then there’s a phase for the big picture and ideally, you don’t lose that and find a way to balance that. I’ve just been uniquely lucky. 

Even you’ll see one thing that in this book and I think which is not common to book by economists is I’m unapologetic about having an ethical frame. That shows up in many of the chapters and like I say, even those six sectoral chapters I’m doing, which is education and skills, health and nutrition, police and public safety, courts and justice, social protection and jobs.

RAJAGOPALAN: It tells me what you think is the role of the state, basically. That’s what I got out of it.

MURALIDHARAN: It is true but it’s also there’s something deeper than that, which is to say that there are a lot of people who’ll say, “Oh, we should do education, health because that’s what we need for growth.” But the point with these six things is I don’t have to choose between the Bhagwati and the Sen views because they are intrinsically important for human welfare. I care that you’re better nourished, I care that you’re better skilled, regardless of what it does for productivity. From a pure ethical perspective, I want to crack that problem. Because you see, if I say, “Do this because it’s going to contribute to growth,” somebody will come back and say, “Oh, maybe it’s better to invest in higher education for growth.”

There is an ethical code in what I’m doing here that says, “No, regardless of what it does for productivity, we need to do this,” but the bonus here is it also helps productivity. Then when you look at the police chapter, being safe is an intrinsic right of a citizen but it’s also going to boost female labor force participation. On justice: 70% of people incarcerated are under trial without being found guilty.

And so it’s a human right to improve the functioning of our court system. But if you look at Manaswini’s beautiful paper, increasing judicial capacity unlocks block factors.

RAJAGOPALAN: That is a fantastic paper.

MURALIDHARAN: Unblocks factors of production that is also boosting economic productivity.

RAJAGOPALAN: It’s like such a marginal improvement. It’s not like she’s increasing the number of judges by a hundredfold or something. It’s such a tiny increase but the return is big.

MURALIDHARAN: It is a 25% improvement at the district court. 

RAJAGOPALAN: Exactly. It’s a huge, huge improvement.

MURALIDHARAN: Then this again goes back to the micro and the macro and the purpose of a book like this. There are thousands of scholars doing a lot of high-quality work but the purpose is to select the right tails of that body of work and make it accessible to a general reader and say, “Listen, this is why research matters and this is how you bring the research together in this big picture to then identify a path forward for the country.” That’s why I think the endorsement I like the most—there’s lots of lovely endorsements—but I think what Dr. Kelkar said was particularly gratifying. He said, “I wish people would just do it.”

RAJAGOPALAN: I don’t want to rant about RCTs with you. For that, I think you’re the wrong person. I like to rant about RCTs with Lant Pritchett, my favorite person. I have two episodes on it, which are fantastic. He’s just so insightful about it. I do want to get into one point with you, which is right now, we’re not just talking about what is good research or what is worth doing or what is good evidence and rigorous evidence.

On that, I completely agree with you. The credibility revolution happened for a reason. There was too much compute, people were publishing garbage, we needed better results, so on and so forth. I think the point where the disagreement starts kicking in is that there is an entire industrial complex of people who believe that RCTs will get us development in the old-school sense of the word. I think you know the people who don’t know that difference and so on.

MURALIDHARAN: The reason I want to push back on that is even among the high priests in the movement, I think there has been—here is my own take on this. My own take on this is that if you’re trying to move the epistemic center of gravity of how a field functions—so it is somewhere in one corner—you take a position at the opposite extreme to try to move this. Now, the victory has been won in some ways. Here is where I’m a little more optimistic because if you look at what’s being published in the journals, if you look at the top job market candidates, a simple RCT doesn’t get you a top job anymore. A simple RCT doesn’t even get you into the AER unless you’re now getting into mechanisms or embedding that in models that then gets you to think more about which aspect.

There’s a relatively narrow window in economics where you could publish well just by saying, “I have an RCT.” In a way, credible identification is now table stakes, so we move the bar. Again, the RCT movement, yes, it gets criticized for a variety of reasons but sometimes I think we underappreciate how much it’s delivered for a better scientific process. Even like pre-registration of trials. If you look at P-hacking, if you look at the studies of P-hacking, what they consistently show is that RCTs have a lot less P-hacking.

RAJAGOPALAN: Yes. Again, on this, we don’t disagree. I think they’ve made certain kinds of academic studies better. They brought in particular kinds of rigor, they templatized things, there’s pre-registration. On all of those things, I agree with you. The trouble is, and this is especially the trouble downstream, that the idea of what is development economics became quite synonymous with RCTs. Whereas if a young person came to ask you today, “Hey, what should I study? I deeply care about development,” my sense is you are going to tell them, “Look at public service delivery. Look at how we staff state capacity, look at personnel, look at fiscal federalism. Look at why don’t we have a property tax regime. Look at how we can act.”

My sense is the development questions never get picked up and answered because those questions can’t be answered. It’s like a fiscal federalism discussion, right? You don’t go into it because you don’t have credible evidence so that just gets left on the table, and it’s almost as if downstream people think, especially very young, impressionable students think that, “Oh, that must not matter for development, or that’s not development economics. That’s some kind of political science or something else.”

MURALIDHARAN: The reason I’m less concerned if you look at the program of BREAD, which was the Apex Development Economics Conference that we just hosted here at UCSD this past weekend, yes, there are a bunch of RCTs. But there is almost no paper that I saw that was being presented as saying, “Here is an RCT of what works.” The RCTs were being presented as saying—I’ll give you a simple example. There was this very, very nice paper about targeting LPG subsidies and the subtle point here was that you want to subsidize the adoption among the poor whereas, right now, the subsidy is the same regardless of the cylinder size.

The simple point is because the poor have a higher opportunity cost of time, you can target the subsidy better by subsidizing the smallest cylinder. At one level, it feels small bore, but it is actually quite central to discussions of climate adaptation, discussions of allocating budgets and even I would say it’s property taxes. 

See, what I keep telling my students, and I think many people do, most people do now, is that the question comes before the method. I don’t think it’s correct to say that development means RCTs, though a lot of people do have that mistaken impression. Partly, perhaps because that’s where the pipeline of RAs comes from. But one way to think about it is that getting that RA experience is now a way of getting field experience of just going to the field and understanding what the ground reality is. It doesn’t mean you have to do an RCT.

On the other hand, I think the way the framing has changed—there was a wonderful paper on female labor force participation. It was an RCT, but the nice thing is that there are so many important micro questions that are perfectly amenable to RCTs. If I’m looking at what happens if I give an opportunity to work out of home to women, and what does that do? That is an individually randomizable question. Now, if you come back and say, “why am I optimistic,” is that macro development is also doing well.

Macro development has always been about building these broader general equilibrium models. We are starting to see now more discipline in these macro models that is informed by well-identified micro estimates. The good news is I feel that the market in journals—the standards keep going up. There is enough of a market test here to say that, “Okay, just the plain old vanilla RCT doesn’t get you there.” I think, yes, there are other problems with RCTs.

Personally, I would say the bigger problem with RCTs right now is that it’s not about the method. It’s about the fact that because data collection for individual RCTs is expensive. If I was a public funder of research, I would fund high-quality administrative data sets, high-quality general-purpose thing because that would lower the entry barriers to hundreds of Ph.D. students to come work on important topics. 

RAJAGOPALAN: Also, lots of comparative studies.

India’s a laboratory because it’s federal, lots of different states, wards.

MURALIDHARAN: Where I think the misallocation is happening is not in terms of are we crowding out important questions because there’s enough of a market test. You see? The referees, you know how difficult it is to please referees, and referees are no longer happy just with having an RCT. That I’m not worried about. 

Again, the people who’ve seen this kind of enterprise, we are all aware of this. J-PAL itself was actually investing in this administrative data initiative whereby there would be attempts to improve the quality of admin data that could then be made available, both to lower the marginal cost of RCTs and reduce the barriers to doing work.

Lots of improvements are possible, but at least for young students or people thinking about development, at least for me, development is not RCTs. If you look at the BREAD program, you’ll see that’s this whole bunch of exciting papers being presented that are not RCTs. That being said, if the question you are working on is something that is amenable to modular interventions then the RCT is going to be more credible than arguing your way out.

The number of shoddy studies you’ll see about let’s look at what happens to women who work inside or outside the house without the 2,000 confounding variables they will be, that’s the place where you want to see if I offer you the opportunity to work out of home, what does that do? The RCT buys you a lot in terms of methodological credibility and there I think it’s not crowding out a big question. Yes, that’s how I think about it.

RAJAGOPALAN: I guess the big worry I have is not that a simple garden variety RCT gets published in a top journal. I couldn’t be bothered about that problem at all. I’ve checked out of that system. My bigger concern is the misallocation of talent that a lot of people come in wanting to work on big development questions and the way our Ph.D.s are designed is you have to specialize, they end up specializing in a method. They end up sinking a lot of time in one or two field studies that take six, seven years to get published. Then this is the thing that you do and there are very few people who are able to switch out of it or make the transitions that you’ve made.

My philosophical issue with the whole thing is, I have a fundamental problem with the idea that we should only pursue those sorts of policies for which we have gold-standard evidence, as if common sense is dead, as if experimentation is dead, as if we created the Indian constitutions through RCTs, as if we threw our geographical boundaries and our election commission through RCTs.

That’s my meta-philosophical problem with it, that I do deeply appreciate as an economist, how much you focus on rigorous studies. You don’t pick up garbage, you’re so careful in parsing out every sentence, every claim made in the book. I really appreciate that, but I also grieve a lot of the good common sense ideas that I know that you are fundamentally backing that are not in the book, that I think would make the book even better. Though it would make it longer, but I would think it would make it even better.

MURALIDHARAN: Listen, I think the good news is this. Let’s think about the difference between an academic economist and a policy economist. I think the same economist—and it’ll be interesting at some point, you should talk to Dean Karlan and saying, “Here was Dean, one of the leading lights of the RCT movement. And how did he function differently when he’s now the chief economist of USAID?” Or talk to Rachel, when she was the executive director of J-PAL, a leading light of the RCT movement, and then was chief economist at FCDO and she’s now just become president of the Center for Global Development. I think basically what happens here is that the way you function is a function of what your core job is.

As an academic, your job is to create high-quality pieces of knowledge and not necessarily worry too much about what the policy decision is. That is, when you’re the policy economist, your job is to say that, “I’m going to synthesize the best that there is out there and make the best decision,” and you’re right. Many times, it’s not like the world is waiting on RCT evidence for decisions. Maybe what we need is better training for policy economists. And maybe the book that way, hopefully, serves that purpose by focusing a lot on principles that will apply to any person in a policy role saying, “Okay, here are principles that I should constantly be thinking about in the context of the two-by-two.”

There’s gazillions of public finance expenditure items that I’m not covering, but that’s a framework that can now be used in any government finance department saying, “When I’m getting a budget proposal, can I even start thinking about these issues of economic incidents of the expenditure or spatial incidents?” None of this is RCTs that’s coming from my background in public finance. I think the book, in a way, represents this broader synthesis, but again, I’m less worried, frankly, because the reason I talk about what’s being published in top journals is because that speaks to the pipeline of young talent.

I also think that there are increasing vehicles for people to move into policy roles, variety of fellowships where what you need is not to be a producer of knowledge, but a high-quality synthesizer and absorber of knowledge with the caveats of knowing what does it buy you and what doesn’t it buy you. I agree. Every methodology has its flaws, and I’ll be the first person, and I’ve written openly about limitations of RCTs. The art of doing good economics is knowing what method works for what problem, and to always put the problem first.

RAJAGOPALAN: And ask the big questions. Not be scared of asking the big ambitious questions.

MURALIDHARAN: In a way, what I’m doing with the book is saying, “My primary identity is a micro development economist.” Because the questions I have focused on like education and health are amenable to modular implementation and variation, those are questions that are well suited to RCTs. When I’m writing a book like this, I’m clearly, in fact, I think Junaid (Kamal Ahmed)], one of the endorsers says that I draw well beyond the world of RCTs and that’s what the book is about.

RAJAGOPALAN: Yes. No, this was a pleasure to read. You have to promise to come back and talk about so many other things that we both are interested in, including the work that CEGIS is doing, how do we think about improving nuts-and-bolts state capacity at the state level and so on. This was such a pleasure. I’m so happy that you did this. Thank you.

MURALIDHARAN: Thank you. My pleasure. I’ve always been a fan of the podcast. In fact, I don’t know if you saw, in addition to citing ’91 Project— 

RAJAGOPALAN: Yes. You mentioned it in the acknowledgment. I was so thrilled. I got some serious credit because people took screenshots. Listeners took screenshots and sent us that you actually mentioned us in the acknowledgments, which we were very thrilled about. I’m so glad about that.

MURALIDHARAN: Yes. No, you guys are doing great work and I think the broader project of creating knowledge, transmitting that knowledge and then acting on that knowledge is essentially how most human progress has happened. That’s why we do what we do.

About Ideas of India

Host Shruti Rajagopalan examines the academic ideas that can propel India forward. Subscribe in your favorite podcast app