I have been enjoying Tim Harford's 'Adapt (Why Success Always Starts with Failure)'. Some interesting lessons here in terms of the decentralisation debate.
Harford argues that trail and error is an effective tool for solving problems. As the world gets more complicated, it appears that a bottom-up trial and error approach may work better than the old top-down hierarchies. For those interested in pushing decentralisation in government, this will sound familiar. Indeed, the fact that decentralisation allows for experimentation is one of the big pluses highlighted by advocates of moving power away from Whitehall. Those converted to this argument will enjoy the examples, from the military and the private sector, that Harford uses in his book.
A note of caution, however. Harford argues that experimentation is only part of the solution. Organisations need to be prepared to fail, to identify failure and to respond to such failure by changing course. These are things that do not come easily to government (of course, they don't come easily to business either). In some quarters David Cameron's willingness to change his mind is already opening him to criticism (and even ridicule). Personal experience suggests that both politicians and policy makers tend to start with the working assumption that their policies are working. Contrasts the randomistas that Harford describes conducting socio-economic experiments in developing countries who justify random assignment by assuming that their intervention will have no effect. These different outlooks often make policy makers averse to proper open evaluation of their policies. Harford explains this aversion by appealing to several 'nudge type' explanations of why we struggle to learn from mistakes. Policy evaluation might not matter so much if there were other effective feedback loops in place. But the type of group incentive mechanisms that Harford describes in his book may be difficult to implement in the public sector (where outcomes may be harder to evaluate and assign to specific groups) than in the private sector. Local democracy may, unfortunately, prove a rather weak feedback mechanism (especially when strong vested interests are involved). All this suggests that the coalition government needs to give as much thought to the problem of the feedback on, and adaption to, failure as it is currently giving to the promotion of experimentation.
A few further thoughts.
For those of us that struggle with many aspects of the planning system the book highlights the key problems for those who create visionary top-down plans (rather than the kind that adapt e.g., to market signals). It also provides a great example, the 'Merton rule', which explains why some of us worry about the unintended consequences of the government's decision to go for Zero-Carbon homes (you'll have to read the book).
In the chapter on development, Harford considers the problems that occur when trail and error isn't enough. What if we need some kind of big push coordinating lots of changes at once? The problem is well recognised in the development literature and Harford goes for experimenting with Paul Romer's solution - 'charter cities'. I am a lot more wary - and would draw the parallel with Enterprise Zones. In short, you need to worry a lot about displacement (in the short to medium run) and whether any changes then spill over to changes elsewhere in the longer run. In the spirit of Harford's book - this is an experiment that may be worth trying - but surely on a small scale first.
There are some other specific issues on which I am not so convinced, but let me finish with a more general one. Specifically, I struggle with the leap from 'trial and error is important' to the conclusion that 'we should use evolutionary approaches to model economies'. Harford explains that evolutionary modelling is better at describing certain aspects of firm behaviour (e.g. the life cycle of firms). That may be true, but that doesn't necessarily advocate shifting whole-scale to evolutionary type modelling because these models often can not replicate aspects of firm behaviour that may be equally, or more important, for our understanding. (To be fair, I am not sure whether Harford is advocating this - but at a couple of points he certainly appears to come close to it). I am reminded of the debate in urban economics about zipfs law. This is the idea that the second largest city is often half the size of the largest, the third largest city a third the size of the largest etc. For a while, it was fashionable in economics to try to build models of city systems that replicated this particular pattern. But, as pointed out by Gilles Duranton in a paper in the American Economic Review, these models failed to replicate other aspects of the behaviour of cities (for example the tendency of some cities to specialise) that had far stronger empirical support. Duranton argues, and I agree, that is far better to have models that match these facts and only approximate zipfs law. I think the same critique could be applied of the evolutionary approach to clusters, which is popular right now with some economic geographers. These evolutionary models can replicate some of the complex network of interactions between firms that we might see in real life, but at a cost that they lose focus on basic insights about the costs and benefits of clusters. Again, in keeping with the spirit of Harford's book, I am happy to let a thousand flowers bloom - but I still think that many of those flowers need to be of the non-evolutionary type.
Regardless, much food for thought (and perhaps a useful addition to Ed Milliband's summer reading).