- | Mercatus Original Podcasts Mercatus Original Podcasts
- | Ideas of India Ideas of India
- |
Lant Pritchett on Where Did Development Economics Go Wrong?
Shruti Rajagopalan and Lant Pritchett discuss economic predictions, the problems with randomized controlled trials, Indian education and more
In this episode, Shruti speaks with Lant Pritchett about economic convergence, academic skepticism about growth, flawed methodologies in development economics, the shortcomings of India’s educational system and much more. Pritchett is a development economist from Idaho. He is currently affiliated with Oxford’s Blavatnik School of Government as the research director of the RISE Programme, is the Research Director at LaMP (Labor Mobility Partnerships) and is a fellow at the London School of Economics. He previously worked with the World Bank from 1988 to 2007, living in Indonesia 1998-2000 and India 2004-2007. His publications span a wide range of development topics including economic growth, state capability, education, labor mobility and development assistance.
SHRUTI RAJAGOPALAN: Hi, Lant, welcome to the show. It’s great to see you.
LANT PRITCHETT: Great to see you.
Convergence or Divergence, Big Time?
RAJAGOPALAN: You wrote “Divergence, Big Time” in your 1997 JEP paper, just as divergence had peaked. Since then, we have seen convergence, big time. What do you think the world looks like in the next 50 years?
PRITCHETT: I would hesitate to refer to what’s happened since 1990s as big time, in the sense that in that era I was talking about was 1870 to 2000, or 1990s, when we had the data. Divergence had gone up in absolute terms by an order of magnitude. Whereas in the 1870 period, the entire range of GDP per capita across countries was 1,000 to 4,000, in current purchasing power terms. Then currently, it had gone to roughly 10 times that. Now it has moderated somewhat—more particularly, by the way, in terms of population-weighted than population-unweighted data because India and China grew very fast.
I think part of what “Divergence, Big Time” was written to do was to emphasize that in spite of the fact that Gerschenkron and others had talked about the advantages to backwardness, there had been maybe insufficient attention applied to the disadvantages of backwardness and the possibility that things could get stuck in a low-level trap. What I was trying to point out was that if and when countries found a way of breaking out of the low levels that were prevalent, they often, therefore, grew very fast.
I didn’t take “Divergence, Big Time” as a pessimistic take so much as an only conditionally optimistic take, which is when countries got their act together from very low levels, given that they were very far behind, the rates at which they could grow were very fast. I’m super pleased that in the interim since I wrote the paper, many countries have gotten into sustained periods of very rapid growth. That said, if you look at a picture of the world in terms of GDP per capita in unlogged terms, we have the hockey stick of the rich countries taking off into sustained, exponential 2% (roughly) per capita growth. Then what’s happened since is the world has come apart.
Then, of the countries that were lagging, more and more of them have gotten into the hockey stick growth themselves. The pace at which their growth happens, conditional on getting into the hockey stick, is actually much faster—much, much faster than anything that happened in history. Anyway, I think, overall, it’s been a period in which several very big and very important countries like China and India, but also Vietnam and others, have gotten into and stayed in a sustained GDP per capita growth. It’s still the case that India is probably at a tenth of the leading countries in the world in terms of levels.
Predicting the Future
RAJAGOPALAN: Where do you see this going, say, in the next 50 years?
PRITCHETT: I have to admit, my horizon doesn’t usually extend to 50 years for most things. In part because, as you know, if I had to name my biggest contribution to the economics of growth, it’s the emphasis that for whatever reasons—and we can get back to that—the growth process in the OECD countries, conditional on the takeoff that has since solidified by about 1870, has been just amazingly stable over time. The rich world is now rich not because they ever actually had super rapid growth like India and China we’ve seen, but because they had steady growth of about 2% per capita for 100 and more years.
I sometimes start with the exercise of, if you predict Denmark’s GDP in the year 2000 based on only data through 1910, the prediction is off by like 3%. That’s unbelievable. We have a 90- or a 100-year-ahead forecast. People go, “Oh, economists, we can’t forecast into the future.” For the rich—now OECD—in fact, their growth rates were just amazingly stable for a very long period.
That said, precisely my contribution to the growth literature is that growth in the non-OECD countries has been extraordinarily episodic, in the sense that it’s chugged along, again, at slightly the same, if slightly lower than the OECD on average across countries, but the dynamics have been very different. There’s been many more episodes of very rapid growth, followed by a whole variety of different outcomes in terms of just moderately slowing growth as one converges on the leaders, as we saw with Japan. I remember, one of the disadvantages or advantages of being 62 years old is that you remember lots of what, ex post, look like really stupid predictions.
I remember a very famous journalist, James Fallows, who wrote a book in the 1990s basically saying that Japan was the future of the world. That the Japanese economic system had gotten right what Anglo-style capitalism had gotten wrong, and therefore, the future of the world was Japan. In exactly a year, Japan fell into a long economic growth funk from which it has never recovered. All of that is wandering around to say I’m not super optimistic about China.
I think if we forecast China 50 years ahead, China could either have completely converged to roughly at least the level of Japan or South Korea for a 40,000 GDP per capita. But if you rolled me ahead 50 years and I was Rip Van Winkle and I woke up and China was at exactly the GDP in 50 years as it was today, I don’t think I would be that stunned. I think China has obviously had one of the most rapid and most sustained growth episodes of any country ever seen in the history of humanity.
I’ve written a paper called “Asiaphoria Meets Regression to the Mean” that argues, even if there isn’t something called a middle-class trap—which we’ll get back to; maybe there is—but even if there weren’t a middle-class trap, just regression to the mean would suggest that if you’ve seen a country growing very fast for a very long time, what we should expect is a very substantial deceleration of growth rather than a continuation. I just don’t know. I am optimistic that many countries will continue to chug along. I am optimistic that it’s possible to sustain growth and converge on the leaders to roughly full convergence.
I think COVID is epiphenomenal. I don’t think anything of significance really changed with COVID. I would be very stunned if we looked back, again, 20 years from now and COVID was a decisive turning point in anything. That said, I do expect growth to continue to be episodic in the non-OECD countries, and hence, I am much less confident. A, we definitely should not extrapolate current growth rates into the far future with any degree of confidence. B, that characteristic of this episodic growth means we should have, in essence, incredibly larger forecast errors around the forecast of growth than we tend to do.
I’ve done these literal calculations. IMF growth calculations tend to be right about on average across aggregates of countries, but just completely wrong about the variance of it. They just always forecast strong regression to the mean, which is right on average, but they don’t tend to forecast—and they do it asymmetrically. They always forecast that countries growing slower are going to accelerate, and countries that are accelerating, they’ll stay accelerating. That’s what all that is. I expect the foreseeable future to look like what I might call muddled convergence.
I think many countries will continue to grow rapidly, more rapidly than the OECD average for sure—which I suspect will slow—but I don’t forecast with any confidence anything. B, I often say, the surest way to slow growth is complacency during high growth. Everybody tends to get onto a, “Gee, we’re growing at 6%. We’ve been growing at 6% for 20 years now. Therefore, 6% is the new normal.” That can happen just before, wham, you get a serious growth deceleration. You’re unprepared to really cope with the policy terms, and you can end up in 20 years of stagnation.
Future Growth of India
RAJAGOPALAN: I’m similarly not so optimistic about India. If India grows at the rate that it grew in the last 10 years—which was already a bit of a deceleration than the previous decade—I believe it will take 43 years to catch up to pre-pandemic 2019 levels of U.S. GDP per capita. That’s conditional upon growing at 5.5, 6% for the next 43 years, which seems a little unlikely if there isn’t a more sustained push to do all the other things that you have written about—improved state capability, you bring in rule-of-law institutions as opposed to this kind of cronyistic economy and so on.
The lack of push in that direction and the more general complacency towards betting on institutions and rule of law, as opposed to this programmatic welfare growth, makes me a little bit pessimistic that India will actually catch up in 50 years. I could be wrong. I hope I am.
PRITCHETT: I have two responses to that. One is, I still think there’s a big difference between India and China in the sense that, again, if we get into this business of relatively long-range prognostication, you can run the future of India forward for a very long time without there being any truly big shock to the, broadly speaking, institutions that it has.
Because it has, in some form, all of the institutions that the now wealthy countries have. It has a free press—or it was a freer press and is less a freer press—but it has a free press. It has judicial review. It has a nominal commitment to rule of law. It has a functioning multiparty government. You don’t have to imagine there’s been any major shift. Whereas China has an authoritarian, autocratic, single party who’s running the show, which is not a form of any now-prosperous country other than Singapore that has actually reached prosperity with. It’s very hard to run your movie of the history forward 50 years and imagine China looks anywhere like it currently looks in terms of its configuration.
Then it’s precisely those shifts across configurations that become unpredictable and introduce just radical uncertainty. The first thing is India does, in fact, have a weird advantage in that it doesn’t have any necessary “revolutionary changes” that it would need to make in order to sustain growth. That said, I think you’re exactly right, that in my formulation of the ways in which countries achieve what we could call a political settlement or an order, there’s really strong state institutions of rule of law, and there’s deals economies.
There’s economies in which what I call a deal is—what I mean in a geeky way, and since this podcast is a geeky show—what I call the deal is when your expectations are indexed on you. Maybe if I go to get a driver’s license in a routinely operating European or American situation, I don’t expect the outcome of my getting a driver’s license to be indexed on anything I do. I’m going to follow the rules. The quickest way to not get a routine government to work is to say, “Do you know who I am?” That line will get you treated worse than any other thing you can say. Whereas in India, in the past and maybe more so today, it’s a deals economy.
What happens to your investment has to be indexed on you either in terms of identity or indexed on you in terms of the accommodations you are taking to adapt yourself to the rules. There isn’t, in fact, a completely neutral, predictable rule of law. Deals on investments are not completely and totally secured by rule of law. Again, this is what has made episodic growth so pronounced, in my view—in the developing world versus the OECD—is that in a deals economy, shifts across deals become a major source of uncertainty in the growth process.
You can easily get stuck because the further you dig into a growth process secured by deals around a particular regime, the harder it is to reconstitute those deals when the regime shifts—even in a democracy.
Skepticism About Growth
RAJAGOPALAN: Well, also, especially in a democracy which is governed by a particular kind of identity inbuilt into its democratic structure, the way India can often be. The reason I started today with growth is I am one of the beneficiaries—the second postcolonial generation, the first generation to really see huge gains from liberalization. In fact, we’re running a project at Mercatus right now called The 1991 Project, which is looking at reforms that would gear India toward more economic growth. Of course, I was also in a position to gain from that in quite an extraordinary way.
In your recent paper with Addison Lewis, you’ve talked about how growth is not only necessary but also sufficient. And there is no country with low levels of GDP per capita that has high levels of basics. What is a good way to think about growth prescription globally, not just for rich countries or OECD countries, not just for poor countries, not just for conflict countries?
PRITCHETT: Yes, but I think there’s two things, though. I think there’s generally this intellectual pushback against the need for economic growth. It’s always hard to tell what comes into fashion and why, and I don’t want to be too simplistic about that. One thing I’m very much worried about is that the backlash against economists perhaps being overly dogmatic about there being a single, narrow recipe for growth has been—the backlash to the recipe is not “We need a different recipe,” but “We don’t need growth.”
It’s like if you say, “In order for you to have a healthy heart, you need to eat the following diet.” It could well be that people are just wrong about how narrow the necessity of that diet is, in which case you could have some skepticism. If the response to that was, “I don’t like that diet; therefore, I’m going to weigh 300 pounds and never exercise.” That’s really bad for your health.
RAJAGOPALAN: Skepticism about the lack of the poor diet is not the same thing as skepticism about the need for a healthy heart, right?
PRITCHETT: Exactly. I feel there’s way too much skepticism about the need for growth, and I think that’s just a deep confusion. It’s just confused. I want to be very careful to disentangle that I have two views. One of my views is, very strongly, up to about 25,000 GDP per capita countries just need growth in order to be able for all their citizens to have access to what anybody would reasonably consider the material basics of an adequate lifestyle. I think the focus on dollar-a-day poverty has just been morally obscene. To act as if our aspirations for humankind is for every person to get to what is now in inflation terms $1.90, that’s just morally obscene. And we could go on about how terrible poverty has been for the development crowd.
Second, I think the potential for redistribution to improve livelihoods is just radically overestimated for poor countries. One of the things about being poor is your economy tends to not be in the position to have the levers to generate large amounts of revenue. One way in which developed countries generate 40% of GDP in revenue is they have a large, formal, highly productive economy that, therefore, is easy to observe, easy to tax in relatively low-cost economic ways. That’s precisely what poor countries don’t have, so we should expect total revenue yield to be low as a fraction of GDP.
Third, by not having that, you have limited modes of redistribution in an effective, fair, transparent way. Then fourth, I just feel the political difficulties of high levels of redistribution have been gotten completely, totally wrong by the development crowd. I think all of the available variation is associated with changes in GDP per capita.
Changes in GDP per capita give you more private things that people spend on stuff they like—like sanitation, food, housing, more adequate transport. Just how revolutionary having a scooter is to people just cannot be overestimated. Imagine if your life is really circumscribed to how far you can walk. That’s how nearly all of humanity lived for nearly all of history, and all of a sudden you have the scooter—not even a car, just a scooter. Anyway, A, I think you have more private income. B, all of the data suggests at least unit elastic, if not more, buoyancy of revenue with expected GDP. So the government has more.
I think people get into this, “Oh, neoliberal economists suggest cutting taxes to accelerate growth, and therefore, they’re going to cut government revenue. Therefore, they’re going to spend less on these things like core infrastructure and investment in human development and transfers.” It’s like, “No, no, that’s completely, totally wrong.” No, there’s no sustained episodes that I know of in which revenues aren’t buoyant. And since revenues are buoyant, more growth means more government spending. Anyway, my super strong view is that growth is a necessary and sufficient condition for achieving high levels of human well-being.
I think all of the evidence is completely consistent with that. No, you can’t redistribute your way to prosperity at below-average levels of GDP per capita. Because in the end, the average is what would happen if everybody had exactly the same. Well, if your average is, as most countries are, well below the 10th percentile of a rich country, well, you’re not going to redistribute your way. Now, that said, my second strong view is we have been perhaps too dogmatic, not about the in-state, but about the transition path for my take . . .
In the sense that I am pretty convinced that something like strong institutions—and it’s a radio podcast, so I have to emphasize that my fingers make scare quotes every time I say the word “institutions.” When I used to teach, I would give, at the beginning of the semester, a student a piece of chalk and say, “If I ever use the word institutions, throw this piece of chalk at my head, because I have run out of clear and precise things to say.” That said, it’s this useful scare-quoting thing. We have institutions and rule of law, and rule of law is a little bit, to me, more specific than institutions.
It means really that individuals can form their expectations and carry out their lives on the basis that the law will pattern their interactions between them and their government, and between them and the other people, to a very significant extent. That, I’m very big on. That said, part of my “Deals and Development” work suggests countries take incredibly different trajectories through this overall space, and we shouldn’t be too . . .
Part of the work I’ve been doing, together with other people at the Kennedy School when I was there and continued, is the growth diagnostics work, which says, really to sustain growth and to initiate growth, you need to focus on identifying at least a set of binding constraints, which are what the obstacles truly are to your economy, and the context that it’s in, getting more productive.
The real trick is that part of what the obstacle to getting more productive might be, might actually involve some government active intervention to support productivity. I am as skeptical as I may sometimes be about the ability of governments at their giving capacity to act efficaciously, but doesn’t take away that sometimes they do need to act efficaciously in order to sustain growth. That was a very long answer, but when you start from a budget constraint of 90 minutes, I’m tempted to go on too long.
RAJAGOPALAN: No, this is actually helpful in terms of setting the stage. I think one of the reasons all economists—and I’m sure I’ve made this error, especially in the classroom—is we go straight towards institutions, because for all rich countries, it looks similar, right?
PRITCHETT: Yes.
RAJAGOPALAN: It’s a little bit like the “Anna Karenina,” Leo Tolstoy kind of development story. All happy families are alike and every unhappy family is unhappy in its own way. There’s a little bit of that going on here, which is, all rich countries—
PRITCHETT: You know that I have that as an epigram for my first growth episodes paper.
RAJAGOPALAN: Oh, you do? Oh, great. I would be very surprised if that is not where I got this from in my head.
PRITCHETT: [laughs] It could be, but “Anna Karenina” is much more famous than anything I’ve ever written, so you may have gotten it from the source.
Where Did Development Economics Go Wrong?
RAJAGOPALAN: Your point about the relevant context of those institutions is very well taken. But I want to go back to this paper because in this paper, you’re doing two, three different things. One is just reemphasizing the importance of growth, talking about how we’ve gone awry with the focus on redistribution. But mostly, it is a big conversation you are kicking off with your colleagues who are development economists. A lot of development economics used to just be regular economics—growth economics, trade economics, macroeconomic, stability, and so on and so forth. You’ve talked a lot about the Lucas critique and the influence of that.
There’s this massive shift from trade to aid. One is, why did this shift happen? Is it that development economists take growth as a given condition, and then start focusing on the details? Is that what is going on? Is it that they genuinely think growth is irrelevant, which seems a very odd claim to me? This is really coming from the best people. Or is it that those arguments have been made, and economics only rewards new arguments? No one saying the old stuff is really getting published in the top journals. What is going on in development economics that the shift has happened? Because you’re directly responding to that as a critique through this paper.
PRITCHETT: I actually think the shift—there were two roughly orthogonal, roughly unconnected things going on that then might have been unconnected and then got twisted together to make a very powerful rope out of some very weak threads. First, I think there was a massive geopolitical shift after the Cold War. I think, before the Cold War, countries that were doing development were at least in part doing development because they were concerned about the nature and behavior of national development.
They were concerned about the government. They were concerned about the overall economy of these countries because they were concerned about these countries staying out of the realm of the Soviet empire. There was a huge Cold War motivation. The Cold War motivation was squarely about national development, and not necessarily so benign, therefore, for other things, but it was at least about that. Hence, they acknowledged that one of the things that was going to keep countries committed to a certain thing was sustained economic growth that benefited everybody because that could have political and other ramifications.
With the end of the Cold War, basically, the rich countries—and I have a paper like this in a journal called Horizons, and it’s called “Can Rich Countries Be Reliable Partners for National Development?” My basic argument is, once that motivation got taken out of what the mainstream bilateral, multilateral development organizations were doing, it got more and more committed to other single-issue causes, and less and less sustained focus commitment to what I regard as the interests of the South.
What that’s meant is that, A, there’s just a lot less support in the development organizations for things that focus on national development. What that allowed was a rise of concern with an attention to what can we do to mitigate the consequences of national development, as opposed to what are we doing that promotes national development?
I actually think the randomistas could have never been successful without the geopolitical shift. Without the geopolitical shift that left the North increasingly less concerned with national development as a core issue of its development agenda, I think we would have seen less interest in that because it’s just obviously unconnected to these big questions.
The second thing is the failure of structural adjustment. Again, I use that a bit, but the failure of structural adjustment to which I feel the economics profession was on the right track, which was saying, “Okay, if the prescriptions we had made were too narrowly focused and sometimes went wrong because we hadn’t done adequate diagnostics and we weren’t being adequately contextual, weren’t being adequately attuned to what was actually achievable, then we should make our growth recommendations more diagnostic-derived and more contextual.” Rather than “A pox on all your houses. We can’t say anything interesting or reliable about growth,” which is, again, another overall weak thread in this strand.
It’s weak as a thread because, in the end, I think there are a number of quite reasonably reliable things we can say. I do think once we do a diagnostic and contextual, we as economists and as development economists can give countries quite reasonable advice relative to what other people in the world are saying. Then the third thread was a more internal discipline, methodological thing about some purely methodological debate about the reliability of causal estimates, in which this weird stance of skepticism became more and more popular to where, again, you could only make a causal claim on the basis of the strongest possible identification.
That then led to achieve the interesting cutting edge of development economics developing RCTs to eke out causal impact estimates of specific interventions and programs that could be interesting if the development world was shifting away from national development towards a more redistributed, a more targeted-oriented, mitigating the consequences of national development than actually embracing the challenges of promoting national development.
Those three intertwined into this development economics. I have a paper about this whose title is very stark, “Randomizing Development: Method or Madness?” in which I come down squarely on the side of madness. But if you look at the position we’re now in, it’s just madness.
Seeing Like a Development Economist
RAJAGOPALAN: Yes. Maybe I’m reading too much into what you’re saying, or into also what you’ve written, but to me, you know how James C. Scott had this great book called “Seeing Like a State”?
PRITCHETT: Yes.
RAJAGOPALAN: Seeing like a state is to make societies more legible. Based on what you’re telling me, seeing like a Cambridge economist is about attribution. When these two groups meet, it seems like what we get is our current state of development econ policies. Am I being particularly uncharitable?
PRITCHETT: You’re not being more uncharitable than I have been at times. Certainly, you’re not wrong to read those two strands into my work. I remember reading “Seeing Like a State” and thinking, “Wow, has he been in all the World Bank meetings I’ve been in my whole life? He’s just describing life at the World Bank so precisely.” I do think this intertwined—large development organizations, by being largely government-to-government endeavors, are forced to see like a state. They want to and need to see things in terms of clear, bureaucratically delimitable, technocratically separable, easily monitorable interventions.
They love an intervention that is a cause-and-effect technocratic theory of change. Hence, those two views just love each other. It’s just amazing. One of the things that strikes me, again, is that it’s unbelievable what an incredible backward step RCTs were in the development field. Because by about the late 1990s, early 2000s, nearly everybody who had been in development as a field since the 1980s, and had really worked in countries and in the trenches, were convinced that a, if not the, major problem with the effectiveness of development as an endeavor and development assistance as an endeavor, was that it was inadequately political—inadequately embedded in a realistic, positive, political economy of how positive change actually happened, and therefore, inadequately embedded in history, society, politics.
The way the field was headed was that even to do a successful World Bank project and even if the World Bank was forbidden from doing politics, you had to be embedded in a more realistic politics. That you couldn’t fix the problems that were made to be fixed in a purely technocratic intervention mode that “Seeing Like a State” wanted you to, and that we had to grapple with that.
Then along comes the RCT crowd, who had zero experience at the time, obviously, in actually doing development, and they say to us, “No, no, no. Well, exactly the standard World Bank project mentality can work if we just do rigorous impact evaluation.” It was an unbelievably conservative, “No, no, we can keep everything we needed about the status quo and the way we do projects and the way we approve projects and the way we implement projects and the way we think about governments. We can keep all that the same; all we have to do is add an RCT to it and it’ll be fine.”
I have to say, people in the field were stunned by this level of naivete. How could you imagine that the problem with our failure to achieve takeoff and growth or to achieve adequate development outcomes was merely the result of this one tiny little technocratic fix? It was just stunning. It was just so ludicrous on the face of it that I think one of my main regrets is not having devoted more time to pointing out how ludicrous on the face of it their fundamental claims were.
That everybody that was actually working in development knew this wasn’t true and was actually willing to take on much more radical ways in the ways in which development assistance was going to work or be effective than, “Oh, everything about the World Bank projects and this modus operandi stays the same. We just add an RCT to each project.” Oh my gosh, you got to be kidding me—but they weren’t kidding me.
It’s amazing how influential that’s been, in part because it was so radically attuned to what these organizations wanted to hear, which was, “Oh, you don’t have to worry about politics. You don’t have to worry about society. You don’t have to worry about history. Just do what you’re doing in this technocratic ‘Seeing like a State’ mode, plus an impact evaluation.”
Spending Money Effectively Isn’t Always a Knowledge Problem
RAJAGOPALAN: You’ve said so many things there. Let me pull back. Again, I’ve learned a lot of this from your work. I’m reading some of the other criticism by Angus Deaton and Bill Easterly, and so on. It seems to me the RCT trajectory has four or five different stages. Initially, it just starts with, is there a more efficient way to divert money, which seems like a very reasonable question. We’re spending lots of money in all these different places. Could we be doing this a little bit better? I think that’s a narrow question. It’s not very ambitious in the larger development plan, but fair enough.
PRITCHETT: I don’t mean to interrupt, but not fair enough. I don’t agree with that. Even that stuff would require some evidence because—again, what drives me a little bit nuts, and I think most people who listen to me about RCTs agree that I’m a little bit nuts—what drives me this little bit nuts is if we look at, say we could spend our money more effectively. Well, you might want to ask, what do I think the binding constraint is on spending my money more effectively?
Do I think the binding constraint is that the government just lacks interest in doing it more effectively? Do I think that it’s not spending more effectively because there’s some deep rent-seeking going on; that is, the status quo is actually perfectly happy with the ways in which money is spent? For instance, go to India. In India, Yamini Aiyar and I did this paper called “Value Subtraction in Public Sector Production,” in which government teachers at the elementary level in India are paid a factor of five to 10 times, to maybe even higher, what equivalent-quality people will do the job for.
RAJAGOPALAN: In private institutions?
PRITCHETT: Yes, in private institutions. Even when state governments in the past, as they did in the ’90s when they were under fiscal constraints, introduced contract teachers, it was also shown that the state could attract teachers for much, much lower wages who could do the job just as well. Work about contract teacher shows that the contract teachers, I think, have been at least as effective as the government teachers. It’s not just private versus public; it’s even the public that are taxed. In that environment, you come along and you say, “Oh, well, we could show you technocratically how to be more effective at spending money.”
It’s like, “Yes, but that wasn’t that determined at that equilibrium outcome in the first place.” Nobody imagined that this was technocratically efficient and was misguided or deluded or didn’t understand the cost-effectiveness or relationships or the production function going on. Everybody could see that this was rents. First of all, you have to start with a more sophisticated positive model than the one you’ve articulated, even to start. Because you have to assert that as a positive matter of the description of the world, the binding constraint on spending more effectively is knowledge about causal relationships. I’ve never seen anyone document that that was, in fact, the case. Let me go on a little here, if I can.
RAJAGOPALAN: Please.
PRITCHETT: Again, this is part of the frustration. In 1999, well before the RCT movement took off, Deon Filmer and I had written a paper called “What Educational Production Functions Really Show.” We pointed out that there was this very long and established literature of estimating the marginal effectiveness at improving learning scores of various expenditures or actions within education. Here was the marginal effectiveness of reducing class size, here was the marginal effectiveness of teacher salary, here was the marginal effectiveness of available classroom input—long literature.
Now, the only thing from economists’ simple-minded microeconomic thing we should expect from estimating a production function is that if the person were, in fact, maximizing production with respect to the budget constraint, the marginal product per dollar across all inputs should be equal. That’s the implication of our normative model of spend. This is what you learn in EC 101, when you learn it—which is in graduate school—but it’s still EC 101. Then what we pointed out was, that prediction of normative economics was violated by orders of magnitude.
It wasn’t like 50% off. Many times, of course, it was infinitely off, because things on which a large portion of the education budget was spent had zero learning impact per dollar, and things that have high impact per dollar were chronically underspent on. Our whole point was, in 1999, this is a political issue. What we’re learning from education production functions is that it’s political, is that no one is attempting to maximize learning per dollar from their expending. Otherwise, they couldn’t possibly on a sustained basis be off by orders of magnitude.
One doesn’t make orders-of-magnitude mistakes from ignorance. One makes orders-of-magnitude mistakes relative to optimization because of politics. We had already come to this conclusion in 1999, with really good, solid reasoning and economics and evidence. Then along comes the RCT crowd and says, “Oh, we need to estimate the marginal product of these various interventions and education.” It’s like, “What? Are you just going to ignore everything? Was the world born when you started graduate school and everything that went before is just irrelevant to you?”
We had really well-articulated reasons about what the binding constraints of progress were, that were based on empirical evidence. Then along come these guys that make—with zero empirical evidence at all—the assertion, essentially, that the binding constraint to doing things more effectively was knowledge about what was effective. See, I’m not even willing to grant that the first step was a reasonable step to make.
RAJAGOPALAN: I can’t believe I have been placed in a position to defend the RCTs. Playing devil’s advocate, so, two things here. One, there are governments involved and private philanthropic interests involved, and not all of them are Lant Pritchetts, and not all of them have read and studied and analyzed all the development literature.
Sometimes they may just need some very, very stark evidence in their face that, look, the dollar that you spend on chickens should be spent on cash. That’s where I would leave that. That, to me, doesn’t sound crazy unreasonable. I agree with you that it may still not be the best use of the best minds working in development economics. Let’s grant that there are some interests willing to pay for this, and it’s certainly not the worst thing I can imagine.
PRITCHETT: Sorry. You just slipped something in there that’s super, super important. Let’s just be clear for all the listeners what got slipped in there. What you just slipped in there was philanthropists. I fully agree—to my mind, the only people who are conceivably interested in RCTs are philanthropists. What a philanthropist says is, “I want to spend a very limited amount of money in the grand scheme of things. I want to have attribution of the outcome to me, and I don’t want to mess with politics because I’m a small player. I know I have to work with the cooperation or tolerance of the host government, so I want to be apolitical. I want to have attribution, and I want to spend a small amount of money.”
They’re the only people to whom the RCT can give a reasonable answer about what you should do. In my mind, we never ever should lump together philanthropists, the governments or even development agencies because development agencies historically didn’t regard themselves as philanthropists.
They regarded themselves as development, by which they meant national development, and by which they meant they were partners to government about what government should be doing, and what the society as a whole, and what the rules of the game should be, and how much private-sector actors should be involved, and what the policy stance was with respect to private-sector actors and all of the big-picture questions. I agree that if we slice off philanthropists, they have a case. I want to spend a small amount of money; want to do it with the tolerance of, but not necessarily the cooperation of, the existing government; and I want to have attribution. Okay, then you’re doing TOMS Shoes giving away a free pair of shoes, and wanting to realize the impact of that is silly.
RAJAGOPALAN: No, the reason I slipped that in is that’s how this whole thing started, in my mind, right?
PRITCHETT: Absolutely.
RAJAGOPALAN: It was all about impact evaluation. It wasn’t supposed to be about development, right?
PRITCHETT: Exactly.
RAJAGOPALAN: I would categorize that as stage 1.
PRITCHETT: Right, okay.
Limitations of Randomized Controlled Trials
RAJAGOPALAN: Then you know the rest of the stages quite well. Now, just to play devil’s advocate on the question of, let’s say, RCT and education. I completely agree with you: In 1999, I was still in school, and I didn’t need a development economist to tell me that the Indian education system was hopelessly broken.
I went to one of the best schools there is, so that was crystal clear to everyone. You are right. It was a question of politics. It was a question of local fiscal federalism, as opposed to a highly centralized executive planning, like a national education system, which couldn’t be done. There were lots of different things going on. This was quite clear by the late ’90s. Now, having said that, I think the penny maybe didn’t drop for everyone in Indian education and development at that time, that there’s something systemically broken and flawed.
There’s some kind of hubris that comes with people in political power. I’ve learned quite a lot from, let’s say, Karthik Muralidharan’s RCTs. I’ll tell you the single most important thing I’ve learned. He’s done dozens of this, not just studying one question in the education system, but lots of different questions.
PRITCHETT: Yes.
RAJAGOPALAN: Is this a good way to show that the entire system is broken, that not a single thing actually works? Because very, very narrow, limited RCTs have actually shown that none of this works anywhere in any particularly sensible way. Maybe RCTs are allies in that. Again, I don’t think it’s the best use of resources, and I certainly don’t think the best minds in economics should be doing this, but that’s a slightly different question. Can we learn a lot from lots of different RCTs in a specific area to tell us there’s a systemic problem?
PRITCHETT: Yes and no. As you might know, I’ve been a big fan of Karthik’s all along. One of my current roles is the director of RISE Research, and we fund, through RISE, a country research team that’s led by Karthik, who’s done amazing stuff. Props to Karthik—love Karthik. That said, I think Karthik is of the more effective because he sees it in this bigger picture.
For instance, let’s take a really excellent paper, a recent paper of Karthik’s, which was, Madhya Pradesh is going to borrow from the U.K. this school improvement plan thing in which every school’s going to do a school-based diagnostic, and they’re going to form a school-specific improvement plan. This is what the U.K. had tried, and with some success, I understand, in the U.K. Their large scale, at scale, at government implementation shows zero impact because nobody’s behavior changes.
They literally go through all of the requisite motions to implement it because the Indian government can implement to the extent of creating the isomorphism of looking like you’ve done it. I use, all the time, Karthik’s impact evaluation to show, look, you can take things that were effective in contexts that were effective, and which were seeking effectiveness, in which the system wasn’t totally broken, and put them in a broken system, and they won’t work. 100%, right?
But the danger, the reason there’s some “no” is that—the danger is that in this “Let’s find out what works mode with RCTs,” is that if you do an impact evaluation on something that failed, the impetus is to move on to something different, not recognize that the failure had deep systemic roots. People don’t always draw the right lessons. Again, if you’re working for a philanthropist who is just saying to you, “No, no, just tell me how to spend effectively $100 million,” he doesn’t want to hear the system story.
He wants you to search around the design space or the implementation space until you find something that can be done that’s robust. And with respect to system failures, that’s a pill or an injection, or something that’s a logistically implementable, cocoonable thing. Yes, you can derive that story from it. I was just thinking like two days ago, I or someone should write this story of RCTs in India. What RCTs in India have taught us is the state doesn’t work. They did an RCT and that’s just not the conclusion they’re drawing.
Abhijit, and I forget which other collection of others, did a study together with an Indian police service officer, who’s a wonderful person and wonderful woman—Singh—where they did a randomization of police behavior. I don’t know if you’ve seen this paper, but it’s an illustrative paper. They said the public fears the police, and not unreasonably, to be honest. Part of the reason that people fear the police is the police treat the random citizen with hostility. Part of the reason the police treats the random citizen with hostility is the job is an unpleasant job. Let’s make their job better, and maybe that will let them treat the citizen better. The citizen will have more appreciation and cooperation with the police.
They set up treatment arms, and they set about implementing these treatment arms to see which ones improved the perceived attractiveness or the policeman’s perception of their well-being. What they found was that five of the seven treatment arms—I forget the exact numbers—by the time they were doing the mid-line and in-line surveys, the treatment wasn’t being implemented.
RAJAGOPALAN: Yes, and that tells you something.
PRITCHETT: Yes, exactly. The treatment was something that nominally should have been under the control of the bureaucracy. The treatment was something about scheduling a police officer: How many days in a row should they be on duty before they had a day off? It was, in principle, a completely bureaucratic, controllable thing that the IPS officer should have had control over in a hierarchical sense. And yet, when told to do it, they just didn’t do it. Now, the conclusion of this to me is that the state is broken. You’re imagining that you have this instrument under your control and the challenge is, what do I tell it to do?
Because it is under my or under some hierarchical process controlled in the way that we may imagine a bureaucratic organization is. What should the organization do if it has these fungible capabilities, and it’s just waiting to be told what to do? Whereas what we learned is there were a whole bunch of things you could tell it to do, and it wouldn’t even do them on an experimental basis, much less at scale, much less repeatable, much less with fidelity. The conclusion they drew in the paper was—because two of the things got implemented, and at least one of those things that got implemented worked—their conclusion was, “Let’s do only the things that the IPS officer can actually make a difference of because those . . .”
I wish the listeners could see your reaction to this. This is horrific. This is exactly the wrong conclusion to draw, that since the IPS officer who nominally has control (and in a functional organization would have control over behavior), since he or she doesn’t in fact control behavior, let’s just focus on the narrow little domains of this morass and mess of a dysfunctional, corrupt police service that this officer can control. That’s exactly the wrong response and exactly the wrong lessons learned from the empirical findings of the RCT.
In principle, you’re right. In principle, you could do these RCTs, and in principle, these RCTs could demonstrate this is really a massive system. RISE, as part of our education work, we’re doing case studies on, when do we really get huge effects out of interventions? In most of them, we can show that most of the interventions that have huge effects fix a number of interlocked features of the system in order to unlock this kind of complementarity of system and design. Again, I still think, yes, maybe, but you have to see it as that.
Learning the Wrong Lessons from RCTs
RAJAGOPALAN: I have a great RCT story for you, and you’ll get a chuckle out of this. There was a series of papers on time-stamping schoolteachers and how that reduces absenteeism. Different governments started implementing schoolteachers need to be time-stamped, and they had different ways of doing that. I think it was the Andhra Pradesh government, the state government, which adopted it first, and then it spread to other governments.
Now, I know that this is not what the RCT crowd is suggesting. They fully understand the external validity problems of their research, but I don’t think government bureaucrats do. My mother is an Indian classical musician. She used to be with the All India Radio symphony orchestra.
As the Indian state would have it, once upon a time, the orchestra used to be fantastic. It had Ravi Shankar as its composer and conductor and so on. There was a peak, many decades ago, and then slowly, with other influences and opening up of telecom and TV and all those things, and the budget to the symphony orchestra started shrinking. I think at some point, I don’t remember exactly which year, they gave the option to a number of the musicians to retire, like the voluntary retirement to reduce the size of government.
What used to be a 50-, 60-member orchestra suddenly shrank to like 12 or 14 members. They used to get people on contract the few times a year they would perform. They don’t have the same amount of work to do as they used to when it was a 70-person orchestra, when they used to perform more often. Now there’s a lot of absenteeism in this motley crew of musicians.
My mother actually also was involved in production, so she would actually have to go in every single day, but a lot of the other musicians simply had nothing to do. To fix this absenteeism, time-stamping was brought to All India Radio.
PRITCHETT: [laughs] Right.
RAJAGOPALAN: That’s stage 1. He was carrying 12 timestamp cards for the other 11 musicians to punch in and punch out, or some scam like that. That was going on in India. You know the Indian systems, right?
PRITCHETT: Of course, right.
RAJAGOPALAN: Then they figured out that this is what’s happening, and they said, “We need biometric entry. It obviously can’t be faked.” Now these musicians enter All India Radio, and there’s a back door that you leave wedged open. You enter, you sign the register, you leave through the back door. You give your private music classes. You come back at 6 p.m. and you check out through the biometric exit. Now, in none of this did anyone ask, “Is the government supposed to be owning and running a symphony orchestra?”
Narrowly focused on the question of there is absenteeism with the symphony orchestra members, when, in fact, there was really nothing for them to do on a daily basis—this is, I think, one of the worries I have with the RCT literature penetrating the larger governance system in India. It sounds really fantastic that, if only you put some kind of fingerprinting, biometric entry system, then you could solve all problems of government bureaucracy, inefficiency and absenteeism.
Then that’s where you have no question of what the state should be doing. Should it be doing it in the following way? Should it be doing it at the union level, state level, local level? You know all the arguments.
PRITCHETT: All the questions for which economic analysis is actually really helpful have been how does the emergent order operate? Which, by the way, is exactly where we’ve gone. This is a perfect anecdotal illustration of exactly where we’ve gone. The phrase I’m now using with education people in India is, you don’t make Pinocchio into a real boy by adding more strings. If you have lost purpose and drive and seeking of performance inside the organization, you don’t revivify a dysfunctional symphony by making them all show up and timestamp their work.
The idea that you’re going to fix the state level, the government-owned schools in India, by biometrics is just an insult to anyone’s intelligence, right?
RAJAGOPALAN: Yes.
PRITCHETT: You need a system that’s recommitted to purposes that are shared by the people inside the organization. I am sure if your mother’s symphony came back to its former glory, inside or outside the government, and had actual music to be played, the musicians would love to show up and play, right?
RAJAGOPALAN: Yes.
PRITCHETT: Yours is a perfect illustration of attempting to restore the efficacy of something without actually a true commitment to its purpose, without any understanding of the purpose. This fits the whole “Seeing Like a State” technocratic mindset. We don’t actually need to understand music. We don’t actually need to understand education. We don’t actually need to understand economics, which is about how humans behave and how they respond. We can adopt these technocratic, IT-enabled schemes, but you don’t make the story of Pinocchio the puppet.
You make Pinocchio into a real boy with magic, and not with IT and putting a locator chip injected into the arm of every teacher and track every minute of it. Even if you could do that, it’s not the way in which any effective organization—or effective education system, or health system, or police force, or anything sophisticated and complicated which the government does have a necessity to do some things that are actually sophisticated to carry out organizationally—needs to happen.
Again, the thing is that the RCTs have already taught us this, though. The Band-Aid on a corpse was a very early RCT, and it basically showed—because the history of this is the cameras-in-classrooms thing that showed that the cameras could make teachers show up and lead to higher education outcomes. That was in private NGO schools. It was never in government schools. Then when they took the analogy of that and put it into auxiliary nurse midwives in subcenters, it didn’t work at all. It broke down completely. Within months, it broke down, in which people were just essentially buying exemption-from-duty certificates.
Then there’s been another biometric study, and this is one of my favorites. It has the title “Deal with the Devil,” by which they meant the devil that was in the details, but I took it to be accepting to do this inside the government. I think they clarified the title later. Anyway, the point was that they did biometrics—I think it was in Karnataka, in PHCs, in their primary health centers. And what happened is, the staff got so much more hostile because of the pain in the ass that biometrics were that even though they attended more, the patients saw quality deteriorate, the patient perception.
In which case, more of the births moved into higher-level facilities, so birth outcomes got better. The conclusion is, through biometrics in PHCs, we made birth outcomes better by inducing people to not have children in PHCs—which is the whole point of PHCs, is to not have them give birth in the super high-cost facilities. And so, a completely perverse outcome. Again, RCTs are showing this radical lack of external validity, meaning you’re getting opposite outcomes of the intended outcomes, just not assuming the outcome you want.
Yet, the fear at work is that if this is seen as a search for what works, and we’re only going to take what works and scale it, all of the RCTs that fail are going to get ignored. “Because we didn’t really learn anything from that because we didn’t learn what philanthropists want to know, which is, how do I spend my $100 million?” Whereas from a development point of view, we are learning a ton from RCTs. It’s just not at all the way they’re being portrayed. We’re not learning what the RCT crowd is portraying these as learning.
I just come back to—this is really fundamental economics. I teach some master’s students. I’ll teach this session on education economics, where I’ll show them J-PAL’s chart of cost-effectiveness of various interventions. They have a whole bunch that are zero, and they have some that are infinitely cost-effective. Not surprisingly, because they’re infinitely cost—again, the podcast listeners should see you grasping your head. Any economist looks at an estimate that’s infinitely cost-effective and immediately asks the question, “If, in fact, it’s infinitely cost-effective, why isn’t it already being done?”
I show them this. I say, “Hey, this would be a really good way to guide your investment decisions because you should do the things that are really cost-effective and not do the things that aren’t,” and they all agree. Then I teach them a little bit of actual normative economics of the producer, that you should equalize marginal product per dollar.
Then I say, “Look, if your positive is your normative, if your theory about the way the world works is that people are optimizing, how the hell did we get to infinite cost-effectiveness? If we’re in a world in which there’s infinite cost-effectiveness, what’s our theory about how this world got there? If we don’t have a theory about how this world got there, how in the hell are you making recommendations? How can you be making recommendations about what should happen in a world you clearly don’t understand?”
Yet that’s the essence of the RCT world, is that we can ignore understanding the world and having an adequate positive model that actually explains why we have the results of behaviors that we have, and yet we can make recommendations about what they should be without understanding what reality is.
I don’t know. I just don’t know. Like I said, this is EC 101. Because they’re basically starting with policy recommendations based on a figure that shows you, by orders of magnitude, marginal product per dollar in generating learning is not being equalized across these variety of experiments. Now, you could go back to 1999 and say, “A, we do that already, at least as confidently as we know it from your random grab bag of experiments. And B, the correct conclusion from that is, what is your positive model why that was happening, and what would a recommendation mean in such a positive model?”
That’s not at all where they go. They actually just proceed blithely as if—again, if your only person to whom you’re making recommendations is a billionaire philanthropist who wants to do something small, attributable and tolerable, but not necessarily scalable by government or even scalable by a market, then sure, fine, but that’s not development economics. That’s charity economics.
Indian Education: Schooling vs. Learning
RAJAGOPALAN: Coming back to Indian schooling, which I think is the most important problem facing Indians today—
PRITCHETT: Yes, very much so.
RAJAGOPALAN: —just especially post-pandemic. Everyone has slid back in terms of learning. You’ve talked about this a lot, how we need to focus on learning and not schooling, right?
PRITCHETT: Yes.
RAJAGOPALAN: We need to focus on outcomes, not inputs, and that’s very valuable. I still think there’s a very long-run problem with Indian education. Sometimes I hate to blame everything on colonial legacies, but some of this goes back to Macaulay’s legacy, which is, they were trying to create a system of selection. They wanted to select the best civil servants. Macaulay has written about this.
Now you come to post-independent India, and Nehru wants to select the best engineers because the focus of the Indian socialist state is heavy industry. You need engineers and statisticians and economists who are actually going to make your planning system work. It almost seems like this 150-year-long legacy of schooling, which is focused entirely on selection and not at all on learning. It assumes that people have learned, and now we just need a mechanism to select the best and the brightest.
It’s almost like they missed the point. Now, given this situation, what do we do with the Indian schooling system? Do we just burn the whole thing down and start from scratch? Because it seems so dire to me that those in the fifth grade can’t write their own name. There’s virtually no numeracy. It’s pretty dire. The longer they stay in schooling without learning, the worse they get. That’s the other problem. Is there a way to think about the Indian education system which is fixable, given its current state? That’s an easy question.
PRITCHETT: Right. I remain cautiously optimistic that there is. I say this because I first started working on Indian education issues in 1995, so I’ve been at this a fairly long time—A. B, India is the only country that, when I start talking about fixing the public education system, where I get serious, consistent pushback that it’s unfixable and that the Indian public sector has sufficiently collapsed, that it’s beyond redemption. We should just acknowledge that and move to some public-private engagement as the core. I take that seriously.
I think these aren’t silly, ideological, private-minded people. They’re people that are realistic. I remember having a conversation with Gurcharan Das, who said to me, “I didn’t really take seriously how deep the problems with Indian public education were until I was out in front of my house one day, and the street sweeper came up to me and wanted to ask my opinion about private schools.” He said, “I took for granted that my kids would go to private schools, and I took for granted that my internal household people would go to private schools, but assumed that somebody was using the public schools. If the street sweeper is taking their kid out of the public schools, what’s the future of it?”
Now, that said, I think it has to be what I would call a back-from-the-brink strategy. The only way it can work—and this is going to be very difficult, so this is why I’m cautious about optimism—is if the seriousness and direness is acknowledged. And you really say, “We’ve got to bring this organization back from the brink, and we have to reinvigorate this system with purpose.” I do think that you could potentially reinvigorate it if you acknowledge that this has been a selection system, and we are going to become an education system. That we’re going to radically remake Indian elementary education, as a first go, from a selection system to an education system.
Then, in other countries, municipalities in Brazil, large-scale things in Africa have done this, where they’ve just said, “We have completely lost our way. We need to reconfigure the curriculum, reconfigure the attitudes, reconfigure what the goals are.”
I have some presentations that aren’t yet papers where I talk about—the problem is, most organizations, they have these shell of service functions: IT, procurement, HR, around a core. The core is purpose and practice. Here’s a purpose to which the organization is committed. Here are the practices that we feel reach those purposes. I’m describing generically a university, a religion, an NGO, a unit of the armed forces, the police. The problem is, if you don’t have purpose and practices aligned and at the center, then—I call this the donut—you’re trying to fix the organization, but it becomes a donut. Because all there is a shell of HR is hiring teachers and procurement is buying desks.
These service functions are continuing to service a hollow core. Reform of the Indian education system has to start with revivifying the core of, what can we agree is the purpose of it? What can we say are the practices that lead to that purpose? I don’t think there’s a substantial understanding of India of what those could be. I think it’s growing over time, but it’s radical. It’s radical. No incrementalism, and again, I think Karthik’s RCTs have been instrumental in showing that incrementalism is not an option at this stage. We really need to revivify the whole thing around purpose and practice and rebuild it from the core out.
Like I said, other places that were equally weak can do it. Then the only way I disagree with your characterization of the Indian schooling is that you, in your description, said people were assuming there was learning, and then there was selection. I think it was worse than that. I think teachers went into it assuming that only certain kinds of kids could learn.
RAJAGOPALAN: Oh, yes.
PRITCHETT: That what the education system was meant to do was filter that out very fast. Rukmini Banerji, who’s a good friend of mine and just won the Yidan Prize for her wonderful work on education, she was telling that they were working in a state where they did this summer camp where, in six weeks of the summer camp with public-sector teachers, they could achieve a year’s worth of progress in literacy. Then those same teachers were encouraged to use the same methods in the classroom setting. When you moved the teachers with the same, in some sense, techniques and knowledge into the classroom setting, it stopped working.
When you asked the teachers why it wasn’t working, they would say, “These kids can’t learn, these kids can’t learn.” Rukmini would say she would just, “Wait a second, you just taught them over the summer, these same kids. And you saw progress with your own eyes. You saw progress. How can you now . . .” Once they were reintroduced to the schooling mentality, the selection just overtook them again. It’s like, “Look, I’m not really about teaching every kid to get to some minimum threshold of adequate functional literacy and numeracy.” What they really meant is this kid is never going to get there to the super-high threshold anyway.
RAJAGOPALAN: So what’s the point?
PRITCHETT: What’s the point, right?
RAJAGOPALAN: Yes.
PRITCHETT: Anyway, I am cautiously optimistic. [crosstalk]
RAJAGOPALAN: And this interacts with the caste system, right?
PRITCHETT: Oh absolutely, absolutely.
RAJAGOPALAN: In the sense that their perception of who can learn and who cannot learn is so deeply entrenched with who their parents are, and their grandparents are.
PRITCHETT: No, that’s right.
RAJAGOPALAN: It’s really heartbreaking to see it in practice.
PRITCHETT: Oh, in India, yes. As a non-Indian, I’ve worked in India a long time, but I’m not an Indian. Because one time I went to visit, I was writing a paper with Chandra Bhan Prasad.
RAJAGOPALAN: He’s one of our affiliate scholars, yes.
PRITCHETT: Oh, yes. We wrote a paper together about the change in status and life well-being of Dalits over the reform period, which, if I had to name my most under-cited paper, I think it’s this paper. Because it’s very important, but most people don’t want to hear it. Anyway, but we went to visit his village. We’re walking around the village with a Dalit. We’re sitting around with a group of maybe 16 men, and we’re talking to somebody who is a Dalit that had been a public schoolteacher—a very respected guy, obviously, in his community and the broader community.
We’re sitting around, and he tells this joke. The joke he tells is, “You’re walking through the forest, and you come across a snake and a teacher. What do you do?” He says, “Oh, you pick up a stick and you beat the teacher.” Everyone just rolling on the ground laughing, and I’m like, “What in the—?” I’ve clearly got this mystified look on my face, and he says, “Oh.” He looks at me, he says, “Because the snake, it’s just a brute; it doesn’t know what it does. But the teacher should know better.” Again, I’m not saying that’s a huge sample, but that is, in some sense, the social reality.
Even though the massive expansion in public government has allowed the Dalits into school, they still see the teacher as being not an instrument of their liberation, but as an instrument of social selection. That they’re not really being treated in these schools in a way that is, “We expect you to come up to our standard of learning.” Yes, I think, which again, leads to this environment of frustration.
Premature Load-Bearing
RAJAGOPALAN: One very interesting strand which I found—it’s the bridge between your work on schooling and your work on state capacity—is the idea of premature load-bearing, right?
PRITCHETT: Yes.
RAJAGOPALAN: The Indian state does too much given its state capacity, and that’s not only undesirable, but it’s also harmful because it destroys state capacity actively, right?
PRITCHETT: Absolutely, yes.
RAJAGOPALAN: Alex [Tabarrok] and I have written a little bit on this.
PRITCHETT: I looked at that paper, I thought, “Oh, I’m anxious to read it now.” I just saw it. I thought, “Yes, I think this is an interesting idea that needs to be explored more and more.”
RAJAGOPALAN: You have the same idea for education, which is that actually teaching less may lead to learning more, right?
PRITCHETT: Yes, absolutely.
RAJAGOPALAN: Now, both of these are counterintuitive ideas.
PRITCHETT: Very much so, yes.
RAJAGOPALAN: I’m going to ask you to defend this claim because I’m completely on your side. I have also been arguing that the Indian state should do less, and whatever it does, it should do well. That’ll help build the capacity to do more in the future. You’re thinking about this as a long-run pathway to development. Now, I’ll tell you why these ideas need defending, and it may have something to do also with our discussion on caste just previously. The idea that by teaching less in education, students could learn more—is this another kind of discrimination we’re introducing in the state education system?
For those kids, it’s enough for them to just learn how to write their name and how to do some basic numeracy. For the rest of us, of course, we need to be able to write papers that Lant Pritchett can read. Do you worry about this at all when you extend the premature load-bearing idea to schooling—in some sense, that we’re against stratifying students based on a birth accident in terms of which school they end up at?
PRITCHETT: I worry about it, yes, a little bit, but I don’t worry about it very much—because I just have a paper that’s out as a working paper that basically looks at the recent PISA-D. They did PISA, which is the large-scale evaluation that was for developing countries. One of the countries they did, which I’ll use an example, is Zambia. What we find is that, in Zambia, PISA is normed so that the average OECD student is 500. In the sustainable development goals, they chose a minimum level of learning for a middle school student at 420.
They take reaching 420 as the minimum literacy standard. We looked in Zambia at the advantaged elite. Let’s take all the inequality off the table. Let’s take urban, male, native speakers who are in the 95th percentile of the socioeconomic status variable. These are advantaged kids by birth and who are the elite in SES. Their average score was 360. The problem of over-ambitious curriculum affects everybody.
Everybody is getting a crappy education. One of the few times I’ve gotten a spontaneous standing ovation was, I was giving a talk about Indian education to a group, mainly of IAS officers—and who were elites of the education system, who have emerged as elites from the system. I was talking about the deficits in the Indian education system. I just said, “Look, I know everybody in this room is intrinsically smarter than I am by a lot, because imagine where you would be had anyone given a shit about what you actually learned.”
They just did a spontaneous standing ovation because they realize there is someone who actually gets what’s going on. I disagree, in some sense, that the Indian education system is actually demanding in a good way. Even for the elite and advantaged, it’s still a selection system. It’s still just putting you through some brute, brute and brutal torture test to say if you can do this. And yet no one comes out of the Indian education system feeling they were appreciated, feeling their skills were acknowledged, feeling that the best of their potential had been drawn out.
They come out of it thinking, “Okay, I survived that torture test in order to become an elite. Now let me forget everything that I might have learned because I could probably—maybe didn’t really learn anything and get on with my life.” What I come back to is most of the countries with really low performance don’t in fact have higher inequality in performance than rich countries. The gradients of performance with respect to SES were no higher in these low-performing countries and high-performing countries.
It was that the curriculum is so out of touch and the teaching is so out of touch with good teaching that by not learning the fundamentals early, it’s not the case that the elite are getting a great education. Indonesia is another place I’ve worked and lived and have a lot of love for. When the OECD did adult literacy tests, fact of the matter is that adult literacy of tertiary graduates in Jakarta—this is just a sample of the most elite city, the most elite—were lower than high school dropouts in the OECD.
Again, I think there’s this illusion that there’s this very steep gradient in a low-performing system and that the elite are coming out with super good educations. I agree that the elite coming out of the Indian system are super impressive, amazing people because no way could I have survived the education system. I’m not claiming the Indian elite aren’t unbelievably world-competitive with any elite everywhere. There’s a reason why the Indian elite come out and are now CEOs of major American corporations, but it isn’t because the education system has been this wonderfully value-added process.
It’s been an unbelievably brutal selection process, which select on a bunch of overcoming features like grit and determination and drive that then might be good signals of who could be an effective CEO. If you can survive the Indian education system, of course you can run Google.
RAJAGOPALAN: It’s interesting what you’re saying, that everyone is suffering. Both the elites and the nonelites are suffering. But the disenfranchised don’t have all the additional benefits that the elites have to actually use their networks to get where they need to go even if the learning hasn’t taken place, which seems to be the fundamental problem. In fact, the gains of reducing the curriculum, the larger share of the gain, so to speak, might go to those who are in the disenfranchised category because they have even fewer networks and other access points that the elites have.
PRITCHETT: One of the points of universalizing education is that we believe that IQ is actually not that steep a gradient with respect to anything. You need to expose people to enough to see if they even have any talent. I think just the tragedy of the Indian system is that there are geniuses in all kinds of domains in the arts and music and literature and mathematics and engineering. There are just absolute geniuses that just never even get to the point where it can become obvious that they're geniuses. They never even get exposed to enough of learning that they can show that they can do it, and therefore they just get washed out.
Indian Education System 2.0
RAJAGOPALAN: In your complete reboot, what is the level at which you would set the curriculum, and how would it progressively increase? I’m not expecting a detailed educational system from you, but just to get a sense of where we’re going with this.
PRITCHETT: Yes. Where we’re going with this is that I think the World Bank, for instance, Jaime Saavedra at the World Bank is on the right track with Learning Poverty. And Learning Poverty isn’t “Read fluently by grade X.” I would say grade three, he says grade four, just to give it an extra—again, one of the shocking things about this is that to actually be useful reading, just neurologically, you have to read enough words a minute that you remember what you’re doing. Your short-term memory has a super binding constraint on how your brain operates. If you can’t read 40 words per minute, you can’t really read effectively.
When you look at early-grade reading assessments—not in India, where I haven’t really seen them as much—kids can make out the letters. But they’re literally having to make out the letters word by word by word, which means they’re never getting to fluency. First of all, if you had to focus all of the first three grades on getting every child to read fluently by grade three, it would almost be worth it, except for numeracy. Then this gets complicated, and I’m going to make all kinds of shapes with my hands, which aren’t very helpful to a radio listener or a podcast listener, but I would emphasize with numeracy depth rather than breadth.
We have these studies by Michelle Kaffenberger and others—Julius [Atuhurra], whose last name I’m forgetting—on the curriculum. What they find is that in the curriculum, there’s this tradeoff between exposure and depth of understanding. Depth of understanding goes from, “Can I carry out this arithmetic operation if I see it exactly like I saw it before? Can I then apply this arithmetic operation to some context that’s almost exactly like one I’m seeing before?” to where, “When I’m confronted with a novel situation for which this arithmetic operation be helpful, do I see that and use it?” That’s breadth.
The second thing I would do is currently, if you think of breadth as horizontal access and coverage of topics as being vertical access, currently the curriculum is this very tall, thin tower. The students are being rushed through exposure, and they never really understand some quite deep fundamentals. For instance, in Indonesia, which does substantially better than India on mass education—doesn’t necessarily lead in, but the mass education—one-third of kids who finish senior, 12 grades of school, can successfully add fractions. That’s shocking.
The reason is they never, in some sense, really understood fractions on a deep level—that a fraction is a representation and that there are multiple—it’s a complex concept, but one-third and two-sixths are exactly the same number represented in different ways. It’s a deep concept. The point is that it would be better if children understood on a deeper level what numbers are and how they work on many fewer topics, so the base of understanding was much fuller, rather than the exposure. Again, the exposure is part of a selection system. I’m going to expose you to this and move on. I’m going to expose you to that and move on. I’m going to expose you to that and move on.
To some extent, what really bright, academically capable kids are able to do is that they can memorize each thing. But then, since they never really had a base, when you present them with novel situations, even on concepts that they have demonstrated mastery over, they don’t have. My answer is two things: Read fluently by grade three, and have a deep and sophisticated understanding of arithmetic operations by grade three. If I could repurpose the Indian education system so that my first three years of schooling, nothing else happened, this would be a radical, history-transforming improvement in the system.
RAJAGOPALAN: I’ve got to admit I’m with you on this. I still think I’m quite amazed that I learned anything at all through that system. I remember learning a lot, but I don’t know how much of it I actually learned, which should tell you everything.
PRITCHETT: That’s a great phrase. I remember learning a lot. No, no. Because you remember revision, right? You remember that “I studied really hard, and they said, ‘Shruti, do this for us under pressure.’ And I did it and they said, ‘Oh Shruti, you can now go on, and you’re good at this.’”
RAJAGOPALAN: No. By grade three, we needed to learn our math tables. I remember, I think by grade three it was 16 times 16. You need to know all your tables up to times 16 and up to 16. Right? Each year they would tag on a few more. I remember, even now, learning my tables actively, I don’t remember it coming to me. The fraction example you gave, it assumes to know that one-third is the same as two-sixths, you need to know simple division, simple multiplication. You need to know more specific parts of the puzzle.
PRITCHETT: It’s a conceptual thing to understand that, right? After all, you can’t memorize all the different ways to represent one-third.
RAJAGOPALAN: Yes.
PRITCHETT: There’s an infinite number of them, right? You have to understand in such a way that you then use it. Because the study that ASER did of youth—because usually, they focus on basics. But a couple of years ago, just before COVID, they did a study of 14- to 18-year-olds. That is the most shocking because the thing is that even kids who’ve gone to lots and lots of schooling in India have not made this transition to thinking in the ways that schooling has taught them to think when it comes to apply to their life.
Even when you showed people a really straightforward, real example, like here’s a picture of a shirt with a price tag marked 300 rupees, and then it’s marked with a 10% discount. How much is the shirt going to cost? These are kids that have been to 8, 10 years of schooling. This is the simplest possible application, but almost universally, with very, very low levels of being able to do that. The one that was even more shocking was, they showed this picture of girl going to bed at 10:30 at night, girl waking up at 6:30 a.m. How many hours did she sleep? Yes, exactly.
RAJAGOPALAN: Wow.
PRITCHETT: Again, your facial reaction—you should do this as a video.
[laughter]
RAJAGOPALAN: Maybe that’s why I don’t do it as a video.
PRITCHETT: The shocking thing was that the percent correct of the youth 14 to 18 who were enrolled in tertiary schooling was 54%.
RAJAGOPALAN: Wow. That’s just counting, at some level.
PRITCHETT: Exactly. I wouldn’t even do arithmetic for that. I would count on my fingers. Right?
RAJAGOPALAN: Yes. I count on my fingers.
PRITCHETT: In some sense, they don’t even understand telling time in a useful way, in a way that that’s how they actually go about things. Anyway, if we could get to deep understanding of very arithmetic operations, not more of them and not memorizing the times tables—this is conceptual. You need to master the concept of what a fraction is or what division is doing. Because again, you only use these techniques if you actually engage your world back and forth in them.
The Lant Pritchett Production Process
RAJAGOPALAN: How can the economics profession get more Lant Pritchetts? What have you been through? What does that production process look like?
PRITCHETT: I’m laughing because I know very few people who would want that, to be honest. You’re a very small minority. I think you want more of me. I’m quietly regarded as a pain in the ass.
Part of this answer is I didn’t grow up in academia. I got a Ph.D. from MIT, and I went to the World Bank. When I went to the World Bank, I was in the research group. Even in the research group, we were forced to go out on operational missions and hear the questions that policymakers were asking. If you do that with an amount of respect that people are struggling with real hard things, I think it leads you to realize that the world is a hard, complicated place. And that to be marginally useful, we have to really understand the constraints in all their entirety of what people are facing and doing.
I remember, I was almost 40 years old when I went to work in operations full time in the bank. I was interacting with people in the Indonesian government. I realized that they were just incredibly cleverer than I was. I’d written many papers and published this and published that, but they just had all kinds of practical skills and understood, at an intuitive and yet sophisticated level, the real nature of the problems they face in ways that it’s taken me a long time to catch up with them. Anyway, I think some amount forced exposure to real problems . . .
One of my mentors, Larry Summers’, advice goes, “Write papers about the economy. Don’t write papers about economics.” And that’s a very hard thing to keep in mind. I do think method detracts from this. I have a friend, Scott Guggenheim, who was one of my development heroes, who’s just extraordinarily intelligent and clever guy about getting things done in the world so things actually happen and making things better.
When we were working in Indonesia together in the middle of the East Asian crisis, Ben Olken came out as a kid, just a kid. I think he was still in graduate school. I’m sure he was still in graduate school. He trumped around with us bit. He went out and he visited projects. Ben Olken has a very, quite famous paper on corruption and ways of monitoring corruption. Scott Guggenheim makes the point that Ben Olken had been in and around Indonesia and in and around the project that Scott was a task manager for the Bank on for five years before he came to Scott and said, “I have an idea, a way we could use your project to learn about the world.”
Scott said, “Since I knew that Ben had taken the time and effort to invest in understanding what I was doing and why I was doing it and what Indonesia was like, et cetera, before he came to me with a research idea, then I was happy to deploy the resources of the project to create this treatment and control and learn.” He says, “Now these students of Ben Olken’s step off the plane and say, ‘Here’s my idea that I want to test for my dissertation.’”
He’s like, “Take a hike, kid. What are you even thinking? Why would you think—you’re not with us. You’re not seeking to understand the reality that I’m living in. You don’t understand what we’re doing. We don’t exist to do a tiny little field experiment to test some pet theory of yours.” Again, I think the true greats like Michael Kremer and like Ben Olken, they’re doing the method on top of this really deep, sophisticated understanding of the reality they’re trying to understand. I think getting to that is the key to getting at good research, because good research teaches us important conceptual insights into the world that we wouldn’t have necessarily had, or been able to make without the method, but for which the method is perhaps necessary, but certainly not sufficient.
RAJAGOPALAN: How has your Utah-Idaho background, lots of exposure to Mormonism—how has that affected or shaped your perspectives on economic development?
PRITCHETT: Okay. I’m laughing in part because no one ever asked that, and I think I was thinking about this the other day, though. There’s two ways, right? One is just Utah, Idaho. Both of my grandfathers were just completely working class. They finished grammar school but not higher, and one became a cement finisher and benefited from the fact that they were building a freeway across Utah. He used to say that he knew Utah well because he crawled across it backwards, finishing cement. And one was a carpenter.
I always felt that I am in some ways closer, by being a nonelite, not that far from pretty hardscrabble, rural backgrounds. I have an empathy with the development experience that sometimes is harder today for kids coming from the generation behind me. They have to look back several generations to find equivalent conditions of how people survive and thrive. Whereas, again, my parents were born in the Depression, and my grandparents survived it and lived in Utah in a time when it was a very poor place. My grandparents cut ice out of the lake in the winter to keep their food cool. I do think that background has helped.
The second thing is—and I’ve never said this to anyone before—but Mormonism is an American religion that was founded because a 14-year-old kid named Joseph Smith was living in upstate New York during a religious revival in which there were all these tent preachers and denominations coming around, preaching the true word of God. And Joseph Smith at 14 years old said, “It sounds like a bunch of nonsense to me. I don’t believe any of it,” and went off and had his own revelation and started his own church.
He believed that organized religion conveyed dogma but not any true, sincere truth. I’m a member of a denomination who believes that the devil had introduced religion into the world. I think I came away from a religious upbringing with a really enormously high level of doubt that anything that anybody says to me is true, which I take as a characteristic of the founder of our religion, that it was a religion that was grounded. That before you had faith, you had to have doubt because if you just believe what you’re told, that’s not even real faith. I’m a huge believer in doubt, if that makes sense.
RAJAGOPALAN: That has more implications for development economics than we imagine.
Writing Process
RAJAGOPALAN: What is your writing process?
PRITCHETT: My writing process is that I get up and write, meaning, what I do first—because the world conspires against research. Research is a very tender flower and the weeds around it, the pressing will always crowd out the important. Part of my writing process is that I’m an incredible procrastinator, because what I do is I procrastinate everything I ought to be doing in order to write. And then use, in some sense, the fact that I can do things quickly. I enjoy writing. I enjoy the research part of it.
Even when I was working at the World Bank, I would come in and spend the first five hours of my day writing, and then the last three hours of the day, I would do my job. That often led to very nerve-wracking situations in which I was, having procrastinated, really scrambling around to complete it. But in the end, I got more done than had I spent more time doing my job. I think I’m one of the world’s most effective procrastinators. I’m both effective at procrastinating and effective while procrastinating, which is not necessarily a strategy for the faint of heart, but it is that.
RAJAGOPALAN: It’s worked well for you. I read your bio, just to get to know a little bit more about you, and I think it might be my favorite economist bio. This is the part I’m talking about: “He has been happily married to Diane Pritchett since 1981 and has three children and two grandchildren. And nothing else. He hasn’t climbed Everest. He hasn’t played the cello, doesn’t write poetry. Being a decent husband, father and economist was all he could do. At the end of the day, I just watched reruns of sitcoms or Netflix.”
It brings me to my final question and probably the most important. During the pandemic, what have you been binge-watching?
PRITCHETT: I’ve binge-watched this really interesting British show about the internal affairs division—which is called a different name—of a big, large Northern English town where the police force is completely rife with corruption. I think it’s called “Line of Duty.” It was really fascinating. It was great drama. Have been binge-watching “Seinfeld” now that it’s back. Now, I don’t binge-watch “Seinfeld” because it’s too corrosively cynical. If I watch more than about two, I get more cynical.
RAJAGOPALAN: Do you have a favorite “Seinfeld” episode?
PRITCHETT: I was in Boston just a few weeks ago and there was a blizzard. The news, of course, the local news goes crazy covering a blizzard because it’s so fun. I kid you not, the woman reporter had been sent to stand in front of the sea with the seawalls with waves coming over. She says into the microphone on live TV, “The sea today, it’s like an old man trying to send back a soup in a deli.”
RAJAGOPALAN: “The Marine Biologist”
PRITCHETT: “The Marine Biologist.” That—
RAJAGOPALAN: That’s my favorite episode.
PRITCHETT: It was like, life imitates art. She literally just, “It’s like an old man trying to send the soup back at deli.”
RAJAGOPALAN: That’s so great.
PRITCHETT: [crosstalk] I’m not sure whether she was actually making a direct reference, but that was just so—
RAJAGOPALAN: That’s great. It has the best ending of any “Seinfeld” episode, the best punch line ending, which is “hole-in-one.”
PRITCHETT: Yes, hole-in-one. It pulls out the Titleist ball—the marine biologist episode, which I just saw a few weeks ago.
RAJAGOPALAN: That’s great.
PRITCHETT: I had just seen it a few weeks ago. Then, like I say, I’m in Boston, and she says, completely straight-faced, “The sea today like an old man trying to send back the soup in a deli.” Wow, what influence. That episode was 30 years ago. Yet it’s part of the parlance. I often think, with respect to advertising jingles and sitcoms, if I could write one thing in life as clever as Taco Bell’s commercials with the chihuahua, I would be a happy man.
RAJAGOPALAN: This has been super fun. Thank you so much for coming on the show, Lant.