Quantifying Institutions Part 2 : Religion AND Politics

In the first post of this series I described recent work in empirical institutional economics and why I thought the work pursued a virtuous end but was compromised by the use of poor institutional measures. Today I will introduce a specific paper of this type that had drawn my attention and talk briefly about the other types of institutional variables that have been tried in country level regressions.

I had come into the current empirical literature whilst researching the resource curse for my honours dissertation. A Norwegian trio, Mehlum, Moene and Torvik, had created a model that predicted that a resource curse would happen only if institutions were below a certain threshold of quality. The resource curse occurred, they said, because natural wealth provided an easy source of unearned wealth; economic rents. It made becoming or usurping a thief, or a warlord or an oligarch much more attractive (and easier) compared to productive activity such as working or becoming an entrepreneur. This doesn’t help the country grow economically. The attractions towards kleptocracy and rentseeking can be mitigated or averted by good institutions, so that Australia, Norway or even Botswana could avoid the curse and even prosper from their resources, but in places like The Congo or Venezuela the gains from seizing something and sitting on it were just too great relative to alternatives.

This was very interesting, and had strong intuitive appeal. It promised insight into one of the most vexing of development issues and gave a distinct hypothesis. They even crunched data in support of this hypothesis.

Unfortunately of course, they used those damn indices.

This was little enough reason to discard an intriguing hypothesis. So I went hunting for a better measurement of institutions.

There had been other methods, but they were also abysmal. Most followed in Weber’s wake, using religion as the explanatory factor. Of course they’d have to change the variable from Protestantism to Christianity (as opposed to Islam) as some Catholic countries had become disrespectfully successful. They could also use Confucianism depending of the flavour of the time and the author’s predilections.

The fact that justifications were in flux was a big turn off. Confucianism was once blamed for China’s seemingly intractable poverty (both by Westerners and Mao) but quickly was described as the secret of Singapore’s success under Lee (albeit coded as Asian Values). Japanese values of seniority, hierarchy and group orientation went from being a unique source of economic superiority to Japan’s tragic flaw almost overnight in the early 90s. If the same belief system was used to explain both failure and success, could it explain either?

What is Protestantism/Confucianism/ Christianity anyway? The successful spread of these beliefs related to their ability to adapt to local institutions and cultures across wide expanses of the globe. Their mere presence doesn’t tell you much about the institutions they had adapted to and are now enmeshed in. Not surprisingly, the empirical work that used religious variables either gave no significant results, or sometimes found the wrong religion performing better.

Similar problems occurred with attempts to do empirical work on the legal origins theory. Whilst it was more concrete in what it was describing, the mere background of a legal system (either Common Law or Civil Law) was not sufficient to describe how it worked in practice and was only a small part of the institutions that I thought were relevant. The results were not encouraging, despite the enthusiasm of an Austrian adviser (eclectically paired with a Post-Keynesian).

Most of the empirics in these areas weren’t helped by the fact that the relevant “good” factor was generally coded as a binary variable. There’s only so much juice you can squeeze out of such an orange.

The most promising avenue appeared to be literature describing colonialism’s effect on institutions and economic development. Apart from the occasional (OK, many more than occasional) Anglophile’s preconceptions, they provided a plausible source, or at least indicator, of bad economic institutions. Whilst this would only account for a limited amount of the relevant institutions, the concept would be less circular and would still provide “degrees” of “badness”.
Because the purpose of colonialism was mercantile, extractive and usually based on zero-sum logic, it was argued that colonialism would either leave extractive institutions, or be most enmeshed where indigenous institutions would be least able to withstand such behaviour. Either way, if you could measure colonial penetration (Freudian imagery was quite explicit in the Post Colonialism literature) you should be able to get a measure of bad institutions. This measure would be exogenous to the post colonial era where the modern growth question had arisen and furthermore, this badness was the same kind of grabbing institutions that the Norwegians had described.

There were some interesting, but weakish results using the length of the colonial period and the size of trade relative to GDP (since the coloniser was presumably the sole trading partner). The promise was certainly there for the use of a real-world data source to be used as a proxy for institutions. The context was limited, but it did provide a interesting test case.

In the absense of better data, proxies seemed the best way forward and I hypothesised that I would be able to use the most basic and fascinating of social institutions: Language.

In the next post I will describe why, and talk about the results I found but I’m also eager to hear if I was too dismissive of observable cultural traits like religion. I’m sure that the resident Austrians will think I was too dismissive of the legal origins theory, but I’m open to ideas on how the concept could cover actual practice as well as legal theory, or whether it could genuinely reflect broader institutions in the relevant society.

About Richard Tsukamasa Green

Richard Tsukamasa Green is an economist. Public employment means he can't post on policy much anymore. Also found at @RHTGreen on twitter.
This entry was posted in Economics and public policy, History, Uncategorised. Bookmark the permalink.

18 Responses to Quantifying Institutions Part 2 : Religion AND Politics

  1. Don Arthur says:

    Richard – This isn’t my area of expertise, so I’m just thinking out loud here. I’m guessing you’ve thought much more about this than I have.

    Unless you’re focusing only on formal institutions like law and government then isn’t part of the problem that some post-colonial nations have a variety of institutions that have developed around ethnic or kinship-based groups?

    Since national boundaries were drawn by colonisers, they often don’t correspond to any existing functional groupings. National governments might establish formal institutions, but these will not necessarily influence the behaviour of the people they are meant to apply to.

    In many cases the important institutions will be small and local. And there may be few existing norms capable of managing relationships between different groups. You’d expect this to be most pronounced in places where most of the individual groups had been self-sufficient before colonisation and trade between them had been limited.

    In these fragmented states, the most powerful norms will apply within groups and govern the behaviour of members. There will be little or no sense of national interest.

    Local norms may also be incompatible with the idea of individual (material) interest (eg demand sharing). Instead, they may encourage ambitious individuals to pursue prestige by gaining power and resources for the group.

    If this was the case, you’d expect nation states with greater linguistic diversity to have ‘lower quality’ formal institutions at the national level (unless one linguistic group was clearly dominant). Language diversity would be a proxy for institutional fragmentation.

  2. Richard Green says:

    Don – This was an issue I did spend alot of time pondering when I was looking around for alternate measures. I figured that the arbitrary nation states were inappropriate data points for the same reasons you described. I even managed to find a data set that collated data by linguistic group rather than sovereign state. Unfortunately, it wasn’t complete enough to get any strong results on the questions I was asking (particularly because I was under strong limitations. That and the data was cobbled together from country level data.

    Eventually, when I became more focused on the colonial side of things, I figured I might be able to thoughtfully disregard this. The colonial legacy would likely be most pronounced in urban areas, safely ensconced within. national boundaries. My measure also was tied to (but not actually related to) the identity of the coloniser and I thus suspected that it would largely correspond to the borders that had drawn for themselves. There were only a few countries (Somalia, Namibia, The Phillipines) where the colonising identity was a tad ambiguous.

    If I had been looking at another variety of institutional effects, I don’t think I could have continued with the national level panel regression.

    Various papers I had looked at had used “ethnolinguistic fragmentation” as a variable to see if it had an effect on instability, or assuming that. I cannot recall their findings though.

  3. Don Arthur says:

    I’m looking forward to the third installment.

  4. Richard Green says:

    Don’t get too excited. It’s only a very modest glimpse at a way forward, maybe a glimpse at a glimpse. But the empirical results got me excited and I’m almost looking for a way to temper this excitement by having someone point out an obvious mistake I missed.

  5. I’m looking forward to the third instalment too!

  6. Richard,

    I am afraid that you are in a sense on a hopeless quest in which language is not going to help you much. Let me explain: you say you are looking for the effect of institutions on desirable things like growth, peace, and more generally behaviour. A couple of points are immediate:

    – institutions will undoubtedly have effects. Laws matters in that countries that have laws where people should drive on the left-hand side indeed see the traffic moving on the left-hand side. Countries where the law is one cannot have more than one wife indeed have very view poligamous marriages, etc. Hence if the only question is whether institutions like law have an effect you can stop before you start worrying about measurement: they do and the evidence is all around you.

    – if you think about positive effects of this institution versus that one, then you must have counter-factuals in mind, i.e. you must engage in questions like ‘if country A had been forced to accept different institutions (legal or otherwise) would it have fared better than it did’? Then you face the problem that there is no such thing as a country without institutions, that the precis shape of any institution is tailored (by colonisers or by internal forces) to that particular society, and that institutions dynamically interact with everything else in a society: contract laws and tax laws for instance shape the type of other organisations you will see which in turn will affect the demand for different laws; The efficiency of the tax office shapes the way people try to avoid taxes which in turn shapes their capital profile, their import profile and, most likely, even where they physically live. The list of interactions is endless, but the point emerging from them is simple: if you are looking for causal stories spaced over long periods of time of such large and complex things like institutions then you will essentially have to figure out how these societies work, i.e. you need the holy grail in social science: a reasonable model of how the whole system fits together. With it you can meaningfully talk about how a different institution might affect a country, without it you are basically stuck to muddling around with correlations like the rest of us. the article you cite does precisely this by the way: it presents a stylised model for the whole system.

    – if you think about pathways, institutions will indeed work partially via its interaction with culture. For a proper understanding you then need to measure culture. Language is at best only a very vague proxy for culture and the way in which the same language by the same coloniser is twisted and turned into different uses by the colonised is not pure coincidence: it will depend on trade interactions, education structures, the level of adversity between the coloniser and the colonised, etc.. Hence I’m extremely sceptical at the outset that particulars of language will really illuminate us about the inner working of institutions on populations, let alone be a part of forming counter-factuals. At the very least you are going to have to map language into more direct measures of culture.

    The critique above does not only apply to you though, it is basic to the whole damned development field: unless one is prepared to borrow or develop a view of how the whole system works, it is nigh impossible to meaningfully talk about the effects of institutions and other complex phenomena.

    Looking forward to installment 3.

  7. Don Arthur says:

    … unless one is prepared to borrow or develop a view of how the whole system works, it is nigh impossible to meaningfully talk about the effects of institutions and other complex phenomena.

    And yet many economists persist in advocating institutional reforms on the basis of abstract economic models.

    I would have thought that reform advocates have an obligation to consider the system as a whole or go back to playing with toy societies and talking to each other.

  8. Don,

    a view of how the whole system works might well be an abstract model. you presumably mean ‘too simple’ abstract models. The usual reply to that is ‘show me a better one’. I am afraid that until such a mythical times comes as when a whole view becomes commonly accepted, that using simpler models of how the whole system works IS the best we can do. Muddling through in the causal world of models, if you like.

  9. Consider the question “how do you win a battle?” Or even “How did Hannibal win the battle of Cannae?”

    One could go and do regression analyses of all the battles in the world, but it wouldn’t help much. It would tell us some things we already know. For instance if you outnumber your opponent it helps. If you have better technology, if you have better communications etc.

    You might also turn up some surprises – for instance that better communications mattered more than technology (or vice versa) more often in one’s sample. So it could be a worthwhile exercise.

    But of course, as Paul was saying above in another context, everything is connected to everything else. So if you wanted to know the answer to the question “is encirclement a good way to win a battle?” the answer you’d get is “it depends”. On pretty much everything. Technology, numbers, the particular stage of the battle, terrain etc etc.

    This is the terrain we navigate in everyday life and we manage to get by – somehow. So there are a whole bunch of things that are accessible to our ‘commonsense’. Now one can argue that one’s commonsense is conditioned by regularities – which is true – but we also take lots of shortcuts which would be hard to justify in a quantitative environment. Our brains do this for us, and they’re far from infallible. They make lots of mistakes.

    The explicit, quantitative model building approach will only enlighten for a small subset of life’s problems – including even policy problems. Most of the decisions we make – in both life and in policy – are based on some kind of ‘learned commonsense’. The recent scepticism about the efficacy of development aid comes from experience encapsulated in a bunch of stories about how aid gets diverted and wasted. There are also regressions which suggest that aid doesn’t help growth. But that’s a very primitive model – in which regressions just go hunting for stylised facts to back up stories or hypotheses.

    So we need to appreciate that the kind of quantification that Richard is talking about is an inherently ‘special purpose vehicle’ (think of the moon buggy – only useful for some kind of terrain). If part three gives us a glipse of a glimpse of a way to inch forward I think that will be great. But keep in mind that any number of theoretical improvements of the last few decades – I’m thinking particularly of New Trade Theory, the New Geography and New Development Theory – for which Paul Krugman won the Nobel Prize (I’d have rather given it to him for his journalism, and his work on crises which actually provides insights into improving policy). In all these Krugman (and others) built new theoretical models only to bemoan how little insight they yielded.

    One wouldn’t have wanted to have stopped them trying with these new models, but a little more attention to the reasons similar attempts in the past had not got very far (eg. increasing returns between the wars) and a bit more serious discussion about how useful this might be along the way (as you find for instance in an earlier style of formalism – think of any major theoretical advance from Jevons to Hicks) wouldn’t have gone astray.

  10. Don Arthur says:

    Paul – Yes I do mean oversimplified abstract models.

    We’ve got good reason to think that formal models necessarily exclude many of the most important causal variables. Because there are no good data sets for these variables they get ignored.

    It is possible to take these variables into consideration without quantifying them or using formal models. In fact most decisions are made this way.

    The rigor of many formal models comes at the expense of relevance to decision making. As Nic says the “explicit, quantitative model building approach will only enlighten for a small subset of lifes problems.”

    I’m happy for economists to play formal games with each other. And I’m starting to understand the norms that govern those games. But policy advising is something different.

  11. Richard Green says:

    Paul – I’m afraid I may have appeared more ambitious than I am. I think Nick’s moon buggy analogy is an appropriate one. The terrain I’m driving over isn’t that of a general theory of society, or even economic institutions, but the narrower terrain being tread by the literature I described in the 1st post. The immediate task at hand was less to create a ground up way of encapsulating institutions (the road trip through Asia might be a better option for that) than it was to create an alternative to the indices that does not rely on a astonishingly shallow and limited concept of institutions.

    Still, as a result of your comments I have spent a great deal of time trying to properly consider exactly what I am doing. I’ll probably spend a great deal more time doing this.

    I do think this kind of approach can be very important in the long run. Despite how apparent and well articulated the problems with the current approaches are, they have real power. The 5% significance threshold can do that. Shoddy regressions on indices have already led to fatuous policy recommendations and I’m afraid that their success apparent success will find a way backwards into the formalism and thus the future empiricism.
    There can be a terrible path dependence here. The 1st generation of work will come up with a mess of concepts. The second generation will pragmatically simplify this for preliminary work, but with caveats. Subsequent generations disregard both the 1st generation and the 2nd generation’s caveats. My arch typical example would be the path macroeconomics went on following IS-LM. Instead of being taken for what it was, a tentative attempt to formalise Keynes, it became a basis to build on, and the discipline suffered as a result. Despite the received notion of scientific theory having data sought and reality observed to test the theory, our models end up being reliant on what (we think) is observable.

    It’s a bit like the joke about the economists looking for their keys under the lamppost. It’s understandable given their limitations, but then the economists that follow them go off and formalise a model of key finding based on the notion that keys are under the lamppost.
    If we can produce rough empirical work with at least the “success” on the indices, but with a measure that encapsulates a broader, more complete and less circular notion of institutions we could make sure this understanding doesn’t get lost. So in this maybe I’m just trying to light up a bit more of the pavement.

    I don’t want to wait for a Kuhnian revolution.

    I’m not sure if that makes sense though. In short, I think that in the short term a view on institutions will influence a view of “how the whole system works” more than the converse, and in the long term we’ll have a view of institutions based on the whole system view which was seeded by the limitations of our initial approach.

    I hope your skepticism about this specific language approach will be assuaged when I show it. There’s no Whorfism involved, and when I use English as data, for example, it does not relate to Englishness.

  12. Peter T says:

    My thinking on this is that language – the main tool we use to construct societies/cultures – is demonstrably too complex to be encompassed mathematically (a kind of higher-order three body problem), or (acording to some computational experts and neuropsychologists) to be encompassed by our conscious minds. It’s composed of at least 5 inter-acting parallel systems all being simultaneouly produced/interpreted.

    Analyses that focus on one system are useful – applying them to the larger whole is not. Bit like feeling your way through a jungle at night with a pencil torch – it’s what you don’t see that bites you.

    For a sensitive discussion of how societies and institions interact in one social ecology have a look at Michael Mann on Latin America:
    http://www.sscnet.ucla.edu/soc/faculty/mann/colombia.pdf

  13. conrad says:

    I think people have looked at some of these questions a lot in social psychology, except the measures and constructs they develop differ, and no-one is proposing a measure or model of everything, which seems to appear to be your goal. There’s a huge literature on collectivism vs. individuality and how it differs across countries, for example (although social psychologists don’t tend to care about economic effects as much as you would probably like). I also think any really general question about the use of language is going to be rather unfruitful, because the scope of the problem is truly immense.
    .
    Perhaps more fruitfully, if you want to look at some effect language might have, you might be able to do small scale studies on the effect of the reduction in linguistic diversity (or increase in bilingualism) and the effect this has on trade, cultural exchange etc., as there are many places where a lingua-franca has been introduced (perhaps this is part of your colonialism). However, even then, I can’t imagine you are going to get any great predictive effects, although perhaps there might be interesting interactions that could be identified (e.g., whether learning the language of your neighbor is helpful and under what conditions), since some places it seems helpful (e.g. Southern China) and some places it doesn’t seem to do much (Africa), although those are things that need quantification. In this case, I think I’m less pessimistic than Paul above about using counterfactuals, as some of the change have been pretty quick and therefore might be identifiable and isolatable. The data might be easier to obtain also, as some is pretty recent (e.g., the effect of introducing Mandarin as on official third language in Hong Kong).

  14. Richard,

    If I trawl back to these two posts, I dont see the construction of a moon buggie with questions like ‘how would I measure trade institutions’? The question put to us is far broader. What we are asked to consider is how to measure culture, but not from the point of view of a psychologist (who could give us 100 dimensions in a heartbeat to start measuring) but rather from the point of view of an economist. The first post is thus big on wanting ‘explanatory power’ which means the desired measure of culture and institutions should fit cross-country variatoin in measured economic development (GDP or PPP). Yet, this task is complicated by Richard’s explicit wish that we should not merely look at the institutions of rich countries and then codify the institutions in the rest of the world as to whether they look like the ones in successful countries. No, what is wanted is somehow an index that captures the cultural essence of what is successful and more specifically, the thing that is in principle transportable between countries (if it cant be transported it doesnt really explain much). The lead paper mentioned above is of that ilk, i.e. if taken at face value the proposed view of the resource curse would lead ‘benevolent countries’ to withhold mineral wealth from under-developed countries because quick wealth would corrupt them. Put more strongly, if you take the story by these 3 Scandinavians seriously and take the development message coming from them, we should perhaps discuss how to take more oil and gas wealth away from East Timor rather than give them more. And if one is not willing to take the action that accompanies the causal story then the obvious quesiton becomes whether we are truly interested in the development of other countries at all!

    What Richard thus wants is already a highly theoretical construct in the sense that what is asked is what I would call ‘the exogenous cultural thing about institutions that predicts growth’. Such a thing does not exist outside of an implicit whole system view of development. The mention of religion, ‘good institutions’, etc. all fit into the same mould: these things are highly complex and we can realy only meaningfully talk about the usefulness of religion and good institutions within a particular view of the whole system.

    Don, Nick,

    of course we in reality take far more things into account in our decisions than is solveable in any formal model. Yet, for me, this merely means we are basing ourselves on an implicit view, with perhaps several formal modals in mind and rough rules of thumb as to what kinf od formal model to think of in what kind of situation. In order to have something that we can explain to others, develop and take further, whatever that implicit view is you both seem to carry in your head when making decisions should be brought into the open and codified. How else are social scientists meant to progress? Surely not by saying ‘we dare not talk about system views because we dont want to admit how vague it is. You will have to figure it out yourself as you go along, perhaps after making many enormous mistakes along the way’? Surely you would prefer your pet implicit views to be explicitly tought and subject to enquiry?

  15. conrad says:

    Reading the above post, I have second suggestion for a possibly tractable project based on these ideas (I think!).
    .
    You could pick something that you apriori know is good and you can apriori quantify (even if messily), and then look at cultural factors which help/hinder it’s development. That way rather than try and make a measure of everything and then see how it spreads to everyone else, it would work in reverse, where the important things would be identified based on the things you know are apriori good and economically analyzable. In this way you would get some idea of the cultural factors involved by looking at the problem (i.e., it would be exploratory) which could then be plugged into your apriori specified economic model. If you were then motivated after finding out the factors via exploratory methods, you could see if you could conglomerate them into some simpler construct and still see whether they have predictive power on your economic model (i.e., your cultural factor would simple be one parameter in a model that is already known).
    .
    For example, I’m not an economist, but I’m going to assume that most economists think that women in the workforce are apriori a good thing (I can’t think of any decent economy where women don’t work), and that you can model the increase in workplace productivity of them entering the workforce already. However, the extent that women have entered the workforce and have entered different professions is culturally constrained. Sometimes it has occurred because of a cultural change within a country, and sometimes it has occurred due to cultural exchange from other countries. It beats me what all the factors behind this are, but it would be interesting to know, and there’s sure to be a huge qualitative literature that you could use before trying to quantify it. It would also be a statistically interesting question because in some countries, the process has been very gradual, whereas others (Korea I think is an example) it’s been exceptionally fast, as I imagine cultural factors were previously acting as a bottle neck. No doubt there are also probably reciprocal relationships between cultural and economic factors. If you did this, you would end up with a nice model with two parameters at the top level — one is the old model that people are happy with, and the other is your model of women entering the workforce. Hopefully your extra parameter then allows you to explain something like workforce productivity based on participation rates better than the old model without it (and even better if it could predict some of the qualitative differences between countries), and you get to look at lots of interesting cultural things whilst constructing it. In addition, because the project would be initially exploratory, you would find the cultural factors that are important rather than try and apriori specify them. This would make things simpler and more tractable.

  16. Richard Green says:

    Paul – I’m confused, but only because Paul decided to starting talking to Richard in the third person after addressing him.

    The wording I used in the posts probably was misleading. I was trying to get attention after all, including from non economists who may have tuned out if I had been more specific.

    I recognise the caveat that there is an implicit whole system going on behind this and I’m working on trying to identify what assumptions I’m making, but I also feel that the strength of this implicit system is shaped by little stuff like these models and regressions. The implicit whole system will start to become explicit and less malleable and begin to exclude parts, but the work that does this will do so based on this kind of work (and unaware that the system was once implicit). I think there is, therefore, virtue in doing this work without yet fully grasping the implicit system that encompasses both my work in this territory and Acemoglu et al, the Norwegians etc..

    I will neglect to discuss the specific dynamics and the implications of competing theories of the resource curse. The use of the Norwegians’ model was a bit of an artifact of the honours course. I wanted to try out an alternate measure, but also needed to demonstrate I could test a specific model rather than just do ad hoc regressions. The Norwegians’ model lent itself to testing with my measure. A better economist would have found or made a better model but c’est la vie (c’est ne pas moi). Nonetheless, I don’t think the particular topic of the resource curse has too many implications on what I’m talking about mainly and might be a (fascinating) distraction). I may as well spoil some of the ending and say that my measure failed to replicate their main finding of interaction between resources and institutions whilst looking promising in other ways.

    I think I can address the issue of explanatory power and exogeneity to an extent if I can isolate a set of institutional factors that has a source exogenous to the sample in a particular study (in this case before the growth period) and can have this source be captured in the institutional measure used. I don’t think the “transportability” of the institution is crucial to explanatory power if there is a viable cause that is also captured in the empirics.

  17. Don Arthur says:

    Paul – So it’s better to be precisely wrong than more or less right?

    I’m all for tight definitions, precise measurements and explicitly set out relationships. But not at the expense of realism.

    I certainly don’t want to use bridges and office towers designed by people whose motto is “that looks about right to me”.

  18. Richard Green says:

    I also just realised that there might be a slight apparent inconsistency when I say that the specific resource curse paper above was merely convenient, when the post says “I had come into the current empirical literature whilst researching the resource curse “.

    The explanation here is actually that I did first find the institutional work this way, and I decided to try something there instead before coming back to the resource curse at a later date when I decided it was convenient.

    Just in case any of you were so fastidious and disturbingly time rich to check the comments I was making.

Leave a Reply

Your email address will not be published. Required fields are marked *

Notify me of followup comments via e-mail. You can also subscribe without commenting.