Well written, and largely persuasive on the tautology problem. But I think it undersells the legitimate role of vague, high-level concepts in research programs. Bad research often precedes good research. AJR's sweeping claim may have been the necessary precursor to Boehm & Oberfield's precision. 'Institutions matter' functions as a Schelling point: it coordinates researchers, policymakers, and donors around a productive question even before the machinery exists to answer it rigorously. Just having people talking about and debating institutions is a big step forward from Samuelson-esque optimization as a lens to view global wealth differences.
The alternative isn't obviously better. 'Property rights' is itself a bundle of nebulous moving parts that you can always decompose further. At some level, broad categories that gesture at real phenomena aren't always a failure of rigor. I think it's often how disciplines orient themselves before the data and fancy econometrics catches up. The question is whether the AJR program generated useful downstream work. I think it did.
I think it's fair to say that the best contribution of the institutions literature might have been paving the way for what came next, and maybe the work I think is better might not have happened without the institutions literature.
The prescription to study specific institutions, not "institutions," is absolutely right. But even "reduce court backlogs" names a goal rather than a mechanism. The harder question is not which institution, but which operational layer: case management, listing procedures, judicial productivity norms. That's where the capacity to deliver actually sits.
The package-deal correlation that defeats cross-country identification is partly a measurement artefact. Studying the major global indices, I saw that scores for subcomponents of governance seemed to be somewhat anchored by the measuring institution's overall view of the country. So the package is probably less entwined in fact than appearance. The within-country evidence bears this out.
India's states vary enormously in court congestion, property rights, and regulatory predictability — Tamil Nadu and Bihar aren't the same package, which shows up in Boehm and Oberfield's own data. Nigerian states show the same pattern. Sub-national is where the identification strategy the post calls for starts to become possible.
The cross-domain version is more striking. Within-country variation across sectors regularly exceeds between-country variation. Think of Nigeria's Central Bank vs its National Petroleum Company. Also, Brazil's 95 federal agencies show almost no correlation between capacity and autonomy — high-performing and dysfunctional institutions running side by side under the same government (States in the Developing World, Ch. 6). If a single country can 'score' well and badly on institutions at the same time, the national label is doing no analytical work. Which is to say: the granularity argument has at least one more level to go.
A few years ago, I edited a couple dissertations on refinements/extensions of the Varieties of Capitalism framework, which deals with similar themes, and I was struck by the gap between the high-level constructs being discussed (e.g. innovation) and the proxies used to measure them, to the point that I started to feel like the research might actually have made a slightly negative contribution to our understanding of the world.
But as with all things, there's a balance—the push toward specificity is important, but there will always be a countervailing pull toward generalization. When you call for answering specific questions about specific institutions, is the implication that we should simply resist the urge to generalize, or that we should learn to generalize better?
When you call for answering specific questions about specific institutions, is the implication that we should simply resist the urge to generalize, or that we should learn to generalize better?
@karthiktadepalli I would love to know what you think!
Some of your criticism, sure looks like the old causation criticism. At least that seems to be the most natural interpretation of your point that of course the correlation between institutions and development is strong. Since we specifically are testing for that as we noticed the developed countries tend to have specific kinds of institutions while those countries who have not developed do not. To the extent that your criticism differs from it, it appears to overstate its case. Since for example, if we were confident at some combination of rule of law constraints on the executive property rights, et cetera is good for economic growth. That’s actually very useful to know if you are, for example, drafting a constitution or thinking of making major changes to the government. It’s obviously a pity that we can’t get more specific answers like exactly how important different things are, but that’s old lack of data for you and it doesn’t mean that it would be a bad idea to aim for the package of everything. It’s frustrating when making trade-offs, obviously, and I can understand why a politician wondering about what things to spend political capital on might be frustrated, but claiming that the research is meaningless seems too strong as meaninglessness generally implies that the research is useless, even in theory. It is also a criticism that appears applicable to just too much. For example, what if I started criticising the research on court “congestion because it did not go specifically into how important cases under different topics or about different laws are. Sure seems like that lack of specificity would be frustrating for, for example, a chief justice, trying to figure out which cases to prioritise in terms of resolution time. Reality, just unfortunately suffers from resource constraints and lack of data, making it difficult to get maximally specific.
I believe there is some latent factor that you can productively call “good institutions” that truly is maybe the biggest cause of development disparities. In that sense I think the causal criticism is offbase. But I also think our particular constructions of that latent factor are nothing more than plausible-sounding stories.
“Drafting a new constitution” is maybe the only situation where “revamp all institutions to be more like rich countries” is a plausible policy prescription. (Even there, the strong insinuation of the institutions literature is that institutions are historical legacies rather than something you can decide to change with a new legal framework or something, but set that aside.) In every other situation, the prescription boils down to “get a better history”. I claim that this is useless even in theory.
This doesn’t apply to more specific claims like “court congestion reduces manufacturing productivity” because reducing court congestion actually is a meaningful policy option. It isn’t an easy problem to solve, and it could be helpful to do even more specific research on the types of cases that are most important to prioritize, but at least the research can point to court congestion as a load-bearing factor in manufacturing productivity (which can correspondingly increase the priority that policymakers give it). There is no analogous possibility with institutional prescriptions.
I think I actually agree with most of that, although I think I disagree that research should necessarily be generally focused on answering question about what you should do directly. I think having a general theoretical background of how the world works is helpful, even in situations where you can’t directly extrapolate an answer in terms of policy recommendations because such an understanding is helpful when developing policy proposals in the long run, even if it’s not helpful in the short term. It’s kind of like how a lot of theoretical mathematics or cosmology looks pretty useless until suddenly becomes relevant to solving real-world challenges. So I think having a general understanding of the World is one of those things that’s helpful.
Although I do find it very believable that perhaps the economics community should invest less effort into this specific line of research, compare to what it’s doing now, but that’s an empirical argument that is quite difficult to assess because the optimal level of effort that should go into particular lines of research is one of those things that is very difficult to calculate and highly dependent on a lot of factors that are next to impossible to measure.
I’m also under the impression that for example, the research on whether democracy causes development or not is downstream of this research into institutions in general and it does seem particular enough to satisfy many of your objections. Although obviously there’s a problem where it’s not likely to influence things on the ground much due to having little impact on the actual actions of relevant actors, for example, there are not many dictators who are going to step down in hopes of increasing GDP growth by 1% or people on the ground who will start protesting if the research findings change. Though keep in mind that this problem where research findings don’t affect the behaviour of many actors on the ground because they have bad incentives or care about the wrong things is a problem for a huge amount of obviously valuable economic research.
I also do think that the research is helpful for situations like a legislature considering a constitutional amendment or a court, interpreting the Constitution and other such big legal changes, where the impact will be obviously much smaller than designing a constitution from scratch, but it will still have some impact and you can still nudge things in the right direction. Although I agree that these things can only have a limited impact, given the importance of things like history and existing cultural equilibrium.
The Boehm and Oberfield India finding is the most useful thing in this piece. Court congestion forcing firms to produce in-house instead of building specialised supply chains directly explains why Indian manufacturing is stuck at 11 percent of GDP. That is not a vague institutions problem. That is a specific, fixable one. Reducing court backlogs never appears in election manifestos. Which is probably why it has not been fixed.
I think the biggest problem is their definition of “institution.” It is far broader than what most people think of as institutions. A tighter definition would be far more useful.
The argument posited by the Acemoglu and Robinson isn't (1) "good institutions cause development". The argument is (2) "inclusive institutions are conducive to development". (1) and (2) are not identical analytic a priori, ergo the claim is not tautological. Whether inclusive institutions are good has to do with the extension of the term "good" and not its definition.
As someone with years of experience in retail, reviewing performance by area, by store, by category, I can see why this is applicable.Telling a developing country to "have better institutions" is like telling a struggling store to "have better operations." Okay, which ones? In what order? Which lever actually moves the needle? The only research that helped was when economists got specific. Court backlogs affecting supply chains is the equivalent of fixing a planogram problem in a specific aisle. That's when you get a real answer and a real fix. Really enjoyed this one, Karthik.
Thank you for a thoughtful piece! I enjoyed reading it. The Gru meme is great as well.
> How exactly do you put institutions in a linear regression?
>In other words, we can make productive statements about institutions – but to do so, we have to stop talking about “institutions” and start talking about a particular institution.
Is your suggestion that we then put a "particular institution" into a linear regression? Could you be more specific about what you mean here?
Well written, and largely persuasive on the tautology problem. But I think it undersells the legitimate role of vague, high-level concepts in research programs. Bad research often precedes good research. AJR's sweeping claim may have been the necessary precursor to Boehm & Oberfield's precision. 'Institutions matter' functions as a Schelling point: it coordinates researchers, policymakers, and donors around a productive question even before the machinery exists to answer it rigorously. Just having people talking about and debating institutions is a big step forward from Samuelson-esque optimization as a lens to view global wealth differences.
The alternative isn't obviously better. 'Property rights' is itself a bundle of nebulous moving parts that you can always decompose further. At some level, broad categories that gesture at real phenomena aren't always a failure of rigor. I think it's often how disciplines orient themselves before the data and fancy econometrics catches up. The question is whether the AJR program generated useful downstream work. I think it did.
I think it's fair to say that the best contribution of the institutions literature might have been paving the way for what came next, and maybe the work I think is better might not have happened without the institutions literature.
The prescription to study specific institutions, not "institutions," is absolutely right. But even "reduce court backlogs" names a goal rather than a mechanism. The harder question is not which institution, but which operational layer: case management, listing procedures, judicial productivity norms. That's where the capacity to deliver actually sits.
The package-deal correlation that defeats cross-country identification is partly a measurement artefact. Studying the major global indices, I saw that scores for subcomponents of governance seemed to be somewhat anchored by the measuring institution's overall view of the country. So the package is probably less entwined in fact than appearance. The within-country evidence bears this out.
India's states vary enormously in court congestion, property rights, and regulatory predictability — Tamil Nadu and Bihar aren't the same package, which shows up in Boehm and Oberfield's own data. Nigerian states show the same pattern. Sub-national is where the identification strategy the post calls for starts to become possible.
The cross-domain version is more striking. Within-country variation across sectors regularly exceeds between-country variation. Think of Nigeria's Central Bank vs its National Petroleum Company. Also, Brazil's 95 federal agencies show almost no correlation between capacity and autonomy — high-performing and dysfunctional institutions running side by side under the same government (States in the Developing World, Ch. 6). If a single country can 'score' well and badly on institutions at the same time, the national label is doing no analytical work. Which is to say: the granularity argument has at least one more level to go.
I will be covering many aspects of this in upcoming posts on https://hiddenrules.substack.com
A few years ago, I edited a couple dissertations on refinements/extensions of the Varieties of Capitalism framework, which deals with similar themes, and I was struck by the gap between the high-level constructs being discussed (e.g. innovation) and the proxies used to measure them, to the point that I started to feel like the research might actually have made a slightly negative contribution to our understanding of the world.
But as with all things, there's a balance—the push toward specificity is important, but there will always be a countervailing pull toward generalization. When you call for answering specific questions about specific institutions, is the implication that we should simply resist the urge to generalize, or that we should learn to generalize better?
@sandstoneviolin
I really like this question:
When you call for answering specific questions about specific institutions, is the implication that we should simply resist the urge to generalize, or that we should learn to generalize better?
@karthiktadepalli I would love to know what you think!
Some of your criticism, sure looks like the old causation criticism. At least that seems to be the most natural interpretation of your point that of course the correlation between institutions and development is strong. Since we specifically are testing for that as we noticed the developed countries tend to have specific kinds of institutions while those countries who have not developed do not. To the extent that your criticism differs from it, it appears to overstate its case. Since for example, if we were confident at some combination of rule of law constraints on the executive property rights, et cetera is good for economic growth. That’s actually very useful to know if you are, for example, drafting a constitution or thinking of making major changes to the government. It’s obviously a pity that we can’t get more specific answers like exactly how important different things are, but that’s old lack of data for you and it doesn’t mean that it would be a bad idea to aim for the package of everything. It’s frustrating when making trade-offs, obviously, and I can understand why a politician wondering about what things to spend political capital on might be frustrated, but claiming that the research is meaningless seems too strong as meaninglessness generally implies that the research is useless, even in theory. It is also a criticism that appears applicable to just too much. For example, what if I started criticising the research on court “congestion because it did not go specifically into how important cases under different topics or about different laws are. Sure seems like that lack of specificity would be frustrating for, for example, a chief justice, trying to figure out which cases to prioritise in terms of resolution time. Reality, just unfortunately suffers from resource constraints and lack of data, making it difficult to get maximally specific.
I believe there is some latent factor that you can productively call “good institutions” that truly is maybe the biggest cause of development disparities. In that sense I think the causal criticism is offbase. But I also think our particular constructions of that latent factor are nothing more than plausible-sounding stories.
“Drafting a new constitution” is maybe the only situation where “revamp all institutions to be more like rich countries” is a plausible policy prescription. (Even there, the strong insinuation of the institutions literature is that institutions are historical legacies rather than something you can decide to change with a new legal framework or something, but set that aside.) In every other situation, the prescription boils down to “get a better history”. I claim that this is useless even in theory.
This doesn’t apply to more specific claims like “court congestion reduces manufacturing productivity” because reducing court congestion actually is a meaningful policy option. It isn’t an easy problem to solve, and it could be helpful to do even more specific research on the types of cases that are most important to prioritize, but at least the research can point to court congestion as a load-bearing factor in manufacturing productivity (which can correspondingly increase the priority that policymakers give it). There is no analogous possibility with institutional prescriptions.
I think I actually agree with most of that, although I think I disagree that research should necessarily be generally focused on answering question about what you should do directly. I think having a general theoretical background of how the world works is helpful, even in situations where you can’t directly extrapolate an answer in terms of policy recommendations because such an understanding is helpful when developing policy proposals in the long run, even if it’s not helpful in the short term. It’s kind of like how a lot of theoretical mathematics or cosmology looks pretty useless until suddenly becomes relevant to solving real-world challenges. So I think having a general understanding of the World is one of those things that’s helpful.
Although I do find it very believable that perhaps the economics community should invest less effort into this specific line of research, compare to what it’s doing now, but that’s an empirical argument that is quite difficult to assess because the optimal level of effort that should go into particular lines of research is one of those things that is very difficult to calculate and highly dependent on a lot of factors that are next to impossible to measure.
I’m also under the impression that for example, the research on whether democracy causes development or not is downstream of this research into institutions in general and it does seem particular enough to satisfy many of your objections. Although obviously there’s a problem where it’s not likely to influence things on the ground much due to having little impact on the actual actions of relevant actors, for example, there are not many dictators who are going to step down in hopes of increasing GDP growth by 1% or people on the ground who will start protesting if the research findings change. Though keep in mind that this problem where research findings don’t affect the behaviour of many actors on the ground because they have bad incentives or care about the wrong things is a problem for a huge amount of obviously valuable economic research.
I also do think that the research is helpful for situations like a legislature considering a constitutional amendment or a court, interpreting the Constitution and other such big legal changes, where the impact will be obviously much smaller than designing a constitution from scratch, but it will still have some impact and you can still nudge things in the right direction. Although I agree that these things can only have a limited impact, given the importance of things like history and existing cultural equilibrium.
The Boehm and Oberfield India finding is the most useful thing in this piece. Court congestion forcing firms to produce in-house instead of building specialised supply chains directly explains why Indian manufacturing is stuck at 11 percent of GDP. That is not a vague institutions problem. That is a specific, fixable one. Reducing court backlogs never appears in election manifestos. Which is probably why it has not been fixed.
I think the biggest problem is their definition of “institution.” It is far broader than what most people think of as institutions. A tighter definition would be far more useful.
The argument posited by the Acemoglu and Robinson isn't (1) "good institutions cause development". The argument is (2) "inclusive institutions are conducive to development". (1) and (2) are not identical analytic a priori, ergo the claim is not tautological. Whether inclusive institutions are good has to do with the extension of the term "good" and not its definition.
As someone with years of experience in retail, reviewing performance by area, by store, by category, I can see why this is applicable.Telling a developing country to "have better institutions" is like telling a struggling store to "have better operations." Okay, which ones? In what order? Which lever actually moves the needle? The only research that helped was when economists got specific. Court backlogs affecting supply chains is the equivalent of fixing a planogram problem in a specific aisle. That's when you get a real answer and a real fix. Really enjoyed this one, Karthik.
Thank you for a thoughtful piece! I enjoyed reading it. The Gru meme is great as well.
> How exactly do you put institutions in a linear regression?
>In other words, we can make productive statements about institutions – but to do so, we have to stop talking about “institutions” and start talking about a particular institution.
Is your suggestion that we then put a "particular institution" into a linear regression? Could you be more specific about what you mean here?