People working in development don’t need to be told that it  complicated, in the sense that there are lots of problems to try to solve. But there is growing interest in the idea that economic systems are complex, in a specific sense borrowed from physics and biology. Books by Eric Beinhocker and Tim Harford have popularised the idea that these processes may be at work in economics, and a new book of essays looks at how complexity thinking might affect economic policy-making.

Earlier this year, my Kapuściński Lecture considered the implications of complexity thinking for development economics and development policy. I’ve now published an updated version as a narrated online presentation which lasts about 45 minutes. You can watch and listen online; listen to the presentation – for example in the gym – by downloading it from Development Drums or via iTunes; or you can download the transcript and slides.

The presentation considers the difficulties that mainstream economics has had explaining why some nations have experienced rapid industrialisation and other nations have not. Then it explores how ideas about complexity developed in physics and biology might shed new light on these questions. This leads to a key conclusion: development is an emergent property of a complex adaptive system. The last section offers seven recommendations for development policy, for people wanting to accelerate development in their own country or wanting to support other countries.

Not everyone has time to watch a 45 minute presentation. But if you watch (or listen to) the first eight minutes, including the story of a design student who tried to build a toaster from scratch, then you will get the main gist of the presentation.

The presentation is accompanied by a series of blog posts highlighting some of the ideas and implications. The first looks at the question ‘What is Development?‘ in the light of this thinking about complexity.  The next will consider what complexity means for the British Government’s ‘Golden Thread’ development narrative.

I am grateful to the EU and UNDP for their support for the Kapuściński Lectures.

If you enjoyed this post, please consider leaving a comment below, and perhaps sharing this with other people using the buttons on the left. You can also sign up to have blog posts sent to you by email.

11 Responses to Complexity and development [presentation and podcast]

  • I thought the presentation was excellent.  Three comments: First, there was no mention of multiple equilibria, which has implications for when countries go wrong by following set of policies based on discredited economics, it is possible they could end up in another stable equilbrium that is not beencicial but very difficult to get out off.  Second, from network theory the implication of connectedness of network implies that if the system(this does depend on the system characteristics) is a freescale system then some nodes are highly interconnected, this gives resilience to smallscale pertubations, but could allow the system to collapse if the highly connected nodes are targetted or affected by global or far-reaching phenomena (eg.global warming). 
    Also it would be interesting to see your Commitment to Development Index applied internally to Europe and how higher-up institutions/poltical entities are benefitting or not benefitting other countries or regions.  Admission: I am a Scot in favour of Independence, and believe that your argument about elites capturing the system applies as easily to the metropolitan elite of the southeast & London as it applies overseas.  Interested in your views.

  • @Mark: thanks.   I agree on multiple equilibria, and the absence of any discussion about stability and resilience. These are important, and I hope that someone will do some work on the implications of all this for complexity. But it was a level of detail too far for this (already long) introduction.

    I like the idea of applying CDI internally. I’m not sure how you would do it, though! 

  • Amazing. An excellently thought-through and delivered presentation. This could be a game-changer for the way that my research is done here at UCL CASA. We’re pretty excited here that someone else is looking at modelling development in this way.

  • Excellent account – I’ve been banging on about complexity with colleagues for ages, trying to get them to read Beinhocker etc. (a prophet is never honoured in his own institute) now I’ve just got to persuade them to watch this.

    How long before this filters down to donors? To the point that these ghastly linear logframe prescriptions are abandoned and we can mention failure rather than having to blindly comply with pre-determined impossible indicators of achievement that we have to promise to get the funds and then ’game’ to achieve donor satisfaction?

  • Excellent presentation. I sat through it all in one go. I am in a non-metro part of India where net connections are unpredictable at times. Is there a way I can download this presentation and lecture?

    Owen replies: thanks Jaya – that’s kind of you to say so. There is no way to download the flash, but there is a video of it on YouTube and there are utilities available on the internet to download YouTube. Today or tomorrow there will be an audio-only version of the file, and I’ll be posting a PDF of the slides too.

  • Owen – very interesting ideas and thanks for starting the debate. I agree wholeheartedly with your views about aid agencies looking to experiment with what works rather than setting a blueprint about how change should happen (the soap nozzle approach).
     
    But surely all societies – including those in developing countries – exhibit some kind of self-organising complexity. In seeing this kind of complexity as the enabling force behind development, I think you run the risk of holding up another blueprint for success like the failed economic theories you criticise. Effectively your argument implies that there is such a thing as a ‘good system’ or a ‘bad system’ in the way that society is organised, and that by making changes to the social system we can gradually engineer better development outcomes for people.
     
    In my view, the systems by which societies are organised are produced by economic forces at least as much as they have an influence on those forces. The social and political systems in many African countries are a product of the colonial legacy and the current international economic order; it is not easy to build new systems, as you so rightly point out in your trenchant critique of technical assistance on governance.
     
    The story of the student building a toaster from scratch was fascinating but your account seemed to downplay the role of power and capital in the story. After all, it’s not as if every part of the toaster is built in the UK – I’d be willing to bet that many or even most of the raw materials (even the semi-processed materials) came from developing countries. What therefore becomes important is not the complexity of the system but who has the capital to get access to all the materials, the workforce etc to set up a toaster factory. In a country where there are lots of people with that kind of capital, a competitive market can develop. Where only the government has those kind of resources, there is a significant risk that the factory will not be productive.
     
    It seems to me that countries that have escaped their position on the lower rungs of the international economic ladder have largely done so through economic strategies of diversification and attempts to escape dependency on exports of a few primary commodities. Their efforts to reach these goals may well have involved a lot of trial and error, but I do not think they started out thinking about how to make their societies more complex.

    • @Justin – Thanks for this. I agree with pretty much all of it. (In fact, I am a bit alarmed that you think this disagrees with what I am suggesting).

      Yes, all societies, including those in developing countries, exhibit some kind of self-organising complexity. But it is not the kind of self-organising complexity which gives rise to the massive improvements in human well-being which are the outcome of development, and which we are trying to bring about. So my argument does indeed imply that there are better systems (ie those which produce high levels of well-being) and worse systems (ie those which do not).

      I am not comfortable with the the word ‘engineer’, but I agree that we can try to bring about better development outcomes for people by trying to help the system be one of the better systems rather than one of the worse systems. We don’t ‘engineer’ these better systems, but we do try to make it more likely that the system will evolve in that direction. (Just as we don’t ‘engineer’ a new plant hybrid, but we do actively intervene to try to bring it about.)

      I agree with you that systems are produced by economic forces as much as they have an influence on those forces. Indeed, that is at the heart of my argument. We need to try to find ways to help those societies and economies co-evolve.

      I don’t think I down-played the role of power and capital in the toaster story. Of course power and capital are a key part of what happens. I agree that an interesting next question is about what determines whether or not a society can supply (or import) the various things needed to make toaster, and that power and capital would be part of that story, but I simply didn’t get into that.

      And finally, I completely agree that the trial and error that leads to the evolution of systems does not start out with thinking about how to make societies more complex. The trial and error that led to the evolution of the human eye, or which leads to the behaviour of an ant colony, does not start with anyone thinking about the high level goal. But the fact that many evolutionary systems do not start out with an intent does not mean that we have to be purely laisssez faire about the systems of which we are part. We don’t have to have a blueprint for society to know that if we can find a way to give more power to the poor, their interests are more likely to be part of the political and economic settlement in the future.

  • Great posts and presentation. I agree wholeheartedly with your approach, though it is not new to me. A few others (such as Frauke de Weijer) have focused on complexity and development, and I have long argued for a focus on the systems of governance (such as here: http://www.fragilestates.org/2012/06/14/how-foreign-aid-succeeds-and-fails-at-the-same-time/).
    The big difference between fragile and robust states are the natures of their societal systems. It is possible to differentiate between the two types of countries–and thus determine which countries are capable of promoting development in their current forms–by examining these informal processes. The most successful countries (with relatively good systems–none are perfectly good or bad) are those where their historical systems are intact and can be used to promote development. The least successful countries are those where colonization forced multiple societal systems (some of which might have been capable of promoting development–see Botswana and Somaliland) together in a way that undermined them (and thus destroyed what where adaptive systems with ample social capital) without putting into place any replacement “good” system. Add on the way that ethnic/religious/clan diversity combines with geography (see the DRC), foreign aid, natural resources, and the competition for the spoils of state control and you have the societal patterns (which are in effect bad systems) that predominate in many places today.
    You ignore the role of local institutions and knowledge in your discussion. Adaptive systems must be deeply rooted in local societies to work, as they depend on the knowledge and values of local peoples. Given the need to balance the new with the old, the best result is usually a hybrid, which combines both in a unique fashion suitable for local circumstances (http://www.fragilestates.org/2012/06/17/rule-of-law-developing-countries/).
    You also ignore the role of social cohesion in constructing successful governance systems. See http://www.fragilestates.org/2012/07/15/increasing-accountability-when-democracy-cannot/ and http://www.fragilestates.org/2012/03/12/horizontal-versus-vertical-social-cohesion-why-the-differences-matter/). Social cohesion is a prerequisite to “effective governance” (which produces development results not necessarily good governance indicators) at both the national and local levels, yet it is rarely talked about in development circles. See China, Vietnam, Korea, and so on. These nation states have their own societal adaptive systems that work informally and are important bases for their countries’ successes.
    Although there are some cases where outsiders can enhance how countries work (such as investing in coalitions), I am not an optimist. Better to focus on building up the “governance ecosystem”–including all the institutions – things like universities, think tanks, rule of law watchdogs, citizen pressure groups, teacher training colleges and civil service academies – that empower citizens to manage their own affairs. And better to invest in documenting local institutions such that they can form part of the solution (instead of being ignored–all too common today).
    The greater the capacity local people have to run a ministry, to build a company, to provide quality education, to organize NGOs and to hold officials accountable, the more likely it is that those in charge will do better – both because they will be more qualified and because those out of power will have a greater ability to monitor their performance.   Such a strategy works better because it allows local people, who are much more effective in navigating the local terrain, to play leading roles instead of foreigners who often know little about the country.

  • Dear Owen,

    First, Development Drums is great!

    I’m writing because I just finished listening to the one entitled ”Complexity and Development” on my walk home from Bole to Kazanchis. I’m a student of complexity, and machine learning – a math guy, rather than biologist or economist – and I had a couple thoughts to share on the development of the logic in the podcast.
    I want to suggest some mathematical perspectives that might reinforce some of the ideas, as well as raise a couple questions. Literally less than 48 hours ago, I saw a video lecture by a Stanford professor describing how he handles situations like the development community is facing, where the models don’t actually work. Specifically, depending on a test I describe below, a test I will argue has already been done in development economics, one can make an informed decision about how to spend time solving a complex problem.

    The Theory

    I assume you’re familiar with the concept of cross-validation of statistical models – you find the ”best fit” of your model (could be simple, like linear regression or complex and non-linear, like a neural network) to a subset of the data, and then see if the model also explains the data you left out, the cross-validation set. If it is much worse at explaining the cross-validation set than it is at explaining the data it was fit to, then it’s ”overfit”. Overfitting, according to this Stanford professor, can be solved by collecting more data and then refitting the model. However, if the model performs poorly on both the fitting data and the cross-validation set, then collecting more data points won’t help. Instead, a good use of time would be either considering a different type of model or collecting additional ”features” – i.e. adding new details or alternative details about the data.

    The Application

    As you explain in the podcast, economists came up with models of development that seemed to fit what they knew. However, as time passed, the models were tested on new events – kind of like a cross-validation set. They were wrong and many countries didn’t develop. In other words, the situation of development economics was one of overfitting. This implies that we should either (a) consider different kinds of models for explaining economies or (b) collect new and different details. You suggest a bit of both in the podcast, I think.

    I’d caution against putting too much stock in pursuing (a), though. A major reason for the overfitting is likely (perhaps ominously and ironically) publishing bias in development economics, be it manifest in academic journals or on Twitter. Although individual economists may themselves be disciplined, the industry as a whole does data dredging all day long – lots of people looking for successful models and the successful ones rise to the top and, importantly, meanwhile, no one is keeping track of the number of models that are tried. It’s statistically acceptable to use cross-validation to measure the success of 20 different models and pick the best, but you’re only allowed to do that if you re-cross-validate, because really you’re just fitting one giant model that has a parameter with 20 different options, 1 for each sub-model, and all of a sudden you’re not actually doing any cross-validation.

    Of course, (b) has its perils, too. Your example of Haile Selassie’s court not wanting to tell him about starvation in the countryside is just one example of many that I’m sure you became all too familiar with in your time in Ethiopia, of information control for the sake of maintaining power (okay, maybe that’s me encountering a lot of that, but I’d guess you saw more than a little, too). You can imagine that a powerful elite might control information about the things that would let outsiders come and ”develop” their country because that would topple them from power. Being on the other side of the curtain, where they can see the content of that information, it’s plausible that in many cases they could control this information much better than development economists could do at figuring out how to get at it.
    I’m not sure how to deal with these issues (although for (a) I think something like JPAL’s site where you can register your hypotheses in advance could be neat, at least).

    One Last Thought

    You certainly demonstrate in the podcast that
    randomized selective (evolutionary) processes are awesome
    some such processes exist in economies
    development economics is complex
    but I don’t think it is at all established that any of the following are necessarily true
    randomized selective processes are the best tool to try next for development
    randomized selective processes (co-evolution) are the most significant cause of complexity to consider in development economies

    I think the strongest that the argument can get at this point is that (to correspond with the numbers immediately above)
    thinking about randomized selective processes might provide valuable new data about an economy
    co-evolution is one potential cause of so-far debilitating complexity in development economics

    I would suggest that with economy-wide development in mind, lists should be developed of both (1) interesting new types of data to be considered for economies and (2) other sources of complexity. Some extremely subjective comparisons could be made to justify whether thinking like a biologist is the best new idea to consider, but I would suggest that a justification warranted, because it seems that, aside from just applying many many more interesting models, finding new features of economies and their parts to keep track of is the only way forward on answering what may be the most important question in the world.

    Conclusion

    Basically, a lot of thoughts about complexity and evolution in there: evolution causes complexity, complexity is dealt with nicely by evolutionary processes, development is complex, development has co-evolution in it. The ideas do not all have to come as one package.

    Hope this was somewhat decipherable. Do send questions, or not, as suits your interest.

    Thanks for Development Drums.

    -Michael Johnston

Follow Owen




Email subscription

I want to receive new posts by email

Owen on Twitter