This is the third entry in the series on deep roots, and it concerns institutions. One of the earlier (and most widely-read) series on this blog was the “Skeptics Guide to Institutions” which had four entries: one, two, three, and four.
Those posts were particularly concerned with a critique of the cross-country empirical work on institutions that started around the mid-1990s as part of the cross-country growth regression literature, and was carried forward by the Acemoglu, Johnson, and Robinson study that used settler mortality as a source of exogenous variation in institutions. I don’t want to reinvent the wheel here, and those original posts are still a decent introduction to that set of papers. The TL;DR version of them is that I don’t find the empirical work convincing in establishing the proposition that institutions are the fundamental determinant of comparative development. The main criticisms were as follows:
The scales of institutional quality (e.g. “protection against expropriation” or “constraints on executives”) are used improperly in regressions. Treating these scales, which might run from 1 (worst) to 7 (best), as a continuous variable implies that the differences in institutions which exist can in fact be measured. That is, it implies that Liberia’s institutions (a 2) are better than Cuba’s (a 1) in exactly the same way that Australia’s (a 7) are better than South Korea’s (a 6). There’s no way that is true. Better to use sets of dummies for each group.
In many cases institutional scales are just functioning as regional dummies. If you code a measure of institutional quality for European countries in 1750, and get that the Netherlands and England have a “3”, while Spain and Portugal have a “1”, and use this in a regression, then you may as well be using an indicator for “gray winters” versus “sunny winters”.
Instrumenting for the quality of institutions in colonies using the mortality rates of European settlers doesn’t satisfy the condition that the instrument (settler mortality) have an effect on living standards only through its affect on institutions. First, the measures of institutions are picking up far more than the specifics of constraints on the executive and the like, and second, settler mortality is like a summary statistic for biological/geographic conditions, which almost certainly have other routes to influence living standards.
Hence I don’t spend a lot of time in class going through the body of literature on these cross-country institutional studies. Nevertheless, I do spend time talking a lot about institutions.
Fair warning that I didn’t proofread this carefully, so apologies for the typos that probably exist.
A key point in studying institutions, from my perspective, is being clear on what question we’re actually trying to answer. If the question is “Do institutions matter?”, then I think the answer is “Yes”, but that this answer is trivial. Of course they matter. I don’t know that you can call yourself a social scientist and not believe that institutions, writ large, matter for economic outcomes. Markets (distorted or not) are institutions, and so even if you have some hardcore belief that you only want to study the economics of something, you still implicitly think that institutions matter. So the literature isn’t about convincing us that institutions are worth studying. They are.
The second kind of question is “Are institutions responsible for sustained growth and/or comparative development?” That’s a much meatier question, and there isn’t an obvious answer here, meaning it is open for study. The original cross-country literature had this as it’s main question, with one important modification. The question was usually a form of “Are institutions solely responsible for sustained growth and/or comparative development?”, and the implied answer had to be Yes/No. That put too high of stakes on everything, as if there was some kind of Highlander-like cosmic battle going on amongst competing explanations and there could be only one. One, there is no plausible way to pick one of those answers (although you would bet on “No”) using data, which would always involve uncertainty. Two, if you wanted to make this argument theoretically, then you can make the answer “Yes” just by redefining institutions until you have a nebulous enough concept to make it true by definition.
The less severe “Are institutions responsible for sustained growth and/or comparative development?” is plausible as a question, but again runs into the issue that we probably won’t be able to find enough data across a wide enough range of economies to really make any definitive conclusions. Which means that this question is something that will probably always be best addressed in a more speculative fashion by someone who is synthesizing lots of existing findings, rather than being pinned down by any specific regression.
But most of the papers in this literature do have specific regressions in them. What are they asking? I’d say it is something along the lines of “What were the observable effects of institution X?”. This is a much more modest question, of course. Once you answer it, you will not know whether institutions are the root cause of comparative development or not. But you will have a clear idea of the effects of institution X, and that may be an input into the synthesizing that you or someone else does later on.
This is the current state of the literature on institutions, from my perspective. In retreating a little on the scope of the question, we’ve been able to get a wave of very nice papers on specific institutions and their influence on economic outcomes over time. That’s the literature I focus on in the class.
One of the features of this latest literature is the care it (usually) takes with identifying causal effects, motivated by the general evolution of economics in the last few decades. In the case of institutions, however, this has meant that nearly all of the work now ends up studying consequences of a broader historical event, colonization.
The reason for this involves an implicit assumption about institutions themselves. If you think that institutions are themselves a reaction to existing conditions (geographic, economic, political, etc..), and evolve to solve very specific problems, then there is essentially no way to evaluate the effect of institutions across any two units of observation. You could call this the “Ostrom problem” after Elinor Ostrom, who noted across a long career that local institutions to manage common pool resources (almost?) always had a clear rationale and were able to prevent environmental overuse.
Why does this create an empirical problem? Let’s assume that I’m trying to compare several different societies (or countries, or villages, or whatever) with some difference in observable institutions. Even if I control for other observable characteristics (e.g. environmental or whatever) the the Ostrom problem is that the differences in observable institutions are, almost by definition, related to some other unobservable characteristics. In short, the Ostrom problem is that differences in institutions can never be thought of as random.
Which brings us back to colonization. Most current work on institutions uses an aspect of colonization to find variation in institutions, because it might be plausible to argue that these institutions were imposed from outside, and not related to unobservables. In certain situations, you might even be able to argue that the institutional differences were the result of random chance.
A good example here is the paper by Lakshmi Iyer on the effects of direct versus indirect British rule in India. There, she argues that there was plausibly random variation in direct/indirect rule due to death of local rulers without heirs. For a specific time period, if there was such a lapse, the British took control directly, while in areas where an heir existed they took control and rule remained indirect. With this kind of quasi-random variation in institutions, she can do a well-defined empirical comparison, and finds that directly ruled areas have fewer public goods (e.g. schools, clinics, roads), even today, than indirectly ruled places.
This only works because there was a colonizer that had a level of control of the institutional structure, and thus the organic local institutions (which may have used adopted heirs or some other means of choosing a successor) were shut down. This same kind of logic is at work in most of the other institutional papers that have come out recently. By leveraging a colonizers power over a given colony, we can see some kind of clear exogenous variation in institutional structure that otherwise would be correlated with unobservables.
I talked about this in my original Skeptics Guides, but once we’re looking at the literature on effects of specific (often colonial) institutions, we’re talking much more about historical persistence than institutions per se. That is, a common goal in these papers is to establish that there are current differences in living standards associated with some historical differences in institutions. That’s the purpose of Dell’s paper on the mining mita, and the purpose of Iyer’s paper on British rule in India.
The institutions that they study no longer exist, of course. So the finding that the mita or British direct rule had negative effects on living standards is not a statement about the direct effects of those institutions on contemporaneous populations. Instead, these papers establish that while once those institutions had some negative effect on the affected economies - although this is usually some kind of black box - these areas remained behind even after the institution itself was removed. These papers are evidence that there is some path dependence in development, and if you push down a place far enough, it won’t be able to catch up again. Note that this runs counter to our standard stories of growth and development, where negative shocks have temporary effects and forces of accumulation, catch-up, and/or convergence bring the economy back to normal.
In the sense that institutions like the mita or British rule can push an area below the threshold beyond which they cannot recover, then institutions matter for development for sure. And the evidence seems to keep accumulating that institutions of various kinds have these kinds of effects. The broader conclusion is thresholds exist, and that in turn means history matters for development in a powerful way. This gives us another reason for looking backwards for the sources of comparative development. Moreover, while the institutions literature was crucial in establishing that thresholds exist, that doesn’t limit us to thinking only of institutional reasons for falling behind. Anything, including geography or culture, could be a reason for falling below a threshold. It’s not institutions OR geography OR culture. It’s institutions AND geography AND culture.