Episode 28: Tim Harford on Adapt (why success always begins with failure)
Tim Harford is a journalist at the Financial Times and the author of The Undercover Economist and, most recently, of Adapt: Why Success Always Begins with Failure.
In this interview, Tim talks about the implications for development of his idea that successful complex systems emerge from a process of trial and error.
Development Drums is hosted by Owen Barder and produced by Anna Scott at the Center for Global Development.
Running time 56 minutes; size 40.2Mb.
Download transcript (pdf)
Podcast: Play in new window
| Download
Posted in Uncategorized.
4 comments
By Owen
– February 23, 2012
4 Responses
Stay in touch with the conversation, subscribe to the RSS feed for comments on this post.
For a good example of not planning for and expecting failure consider the Independent Committee for Aid Impact’s recent publication on how it will assess aid effectiveness. In section 4 it outlines the planned use of a “traffic light” type of system for summarising overall achievements (or the lack thereof). In none of the four levels that will be used is there any outright recognition that a development program has actually failed, instead we find at the lowest level a fudge “It is performing poorly. Immediate and major changes need to be made”
You can find links to the ICAI paper, and my comments on it, here: mande.co.uk/2011/uncategorized/first-reports-published-by-uk-independent-commsssion-for-aid-impact/
Re your question in the podcast “How will we know when we have decentralised enough?” In Indonesia I saw a UN agency implement a health program across a dozen districts, over a period of years. Although district government have considerable autonomy in Indonesia, at no stage did I see the UN agency do any analysis of differences in implementation and/or outcomes across these districts. My experience is that the problem is often not one of lack of diversity but lack of attention to the diversity that already exists. This is not helped by donor/head office performance reporting requirements that typically require “totalising” accounts, and neglect to ask about range and variation.
While I see the merits of RCTs, when I heard Tim’s version of the Kenya story (trials of textbooks, then posters then de-worming) I did wonder whether if in practice the RCTs were a quite inefficient way of discovering what works, and in addition a way which is not very Darwinian – in the sense that Tim has talked about. For example, he recounts how distributing the text books was found to not work. This may be true as a generalisation across all cases, but I suspect there were a few schools where it did work. These cases of “positive deviance” (as has been described elsewhere) are surely great learning opportunities that need to be investigated and exploited. Likewise with the poster distribution, was there nowhere where these did have a significant positive effect? The search for valid generalisations needs to be balanced by the search for useful variations that contradict the average tendency. These might be accidental “mutations” of policy or more intentional variations arising from decentralised authority. This is where case studies aimed at digging into causal specifics can be very useful.
Thanks, very interesting. Except the proper soundtrack is this: www.youtube.com/watch?v=-Iiz4nGw8YQ