The last word in the Community Driven Development wonkwar? Scott Guggenheim responds to Howard White and Radhika Menon

July 12, 2018

     By Duncan Green     

The discussion on Community Driven Development (CDD) has been passionate, at times angry, and has surfaced some important common ground as well as differences. Here Scott Guggenheim responds to yesterday’s post, in what he hopes is the final exchange (people can always continue in the comments section). To recap for those who are arriving new to this, we’ve had 4 posts, including this one, which have generated over 20 comments, many extensive and very well informed. Taken together, I think they make up a good resource on the role, strengths and weaknesses of CDD:

My summary of a ‘bombshell’ of a critical ‘meta-evaluation’ of CDD by 3iE

A response from CDD guru Scott Guggenheim

A response to Scott from Howard White and Radhika Menon

Over to Scott:

‘So the bombshell has turned into a bombsqueak. Howard and co. now confirm that CDD does, in fact, build large amounts of useful, village-level productive infrastructure like clean water, market roads, and irrigation canals.

I wish I could be more conciliatory here – in fact I’d promised Duncan that I would be nice and constructive in a closing comment if given the soapbox – but while I’m glad to see the climbdown, I still don’t buy it. Either so many brilliant poverty specialists who read the 3ie paper mistakenly thought that 3ie had devastated CDD, or else they all somehow believe that potable water, irrigation, and market roads are no longer needed in poor and isolated communities. I prefer the read of a climbdown to thinking that all of those readers need new eyeglass prescriptions. Even in 3ie’s latest comments about a lack of comparator evidence to show cost effectiveness, they again leave out the studies from the very large programs in Afghanistan, Indonesia, Nepal, and the Philippines that show precisely that.

Let’s turn now to the more interesting question of what gives poor communities more voice in development. First, except for one $2.5m project in Sierra Leone that I’ve spent nearly a decade trying to convince my friend Rachel Glennester was actually a not very well thought through outlier, to the best of my knowledge, no CDD project has had the objective of building new social capital.  As Casey (2017) points out, mean levels of trust in control communities were already high:  95% in Sierra Leone; 93% in DRC and 85% in Afghanistan.

CDD’s claim is that we can use this existing social capital to facilitate and benefit from collective action within a community if (and only if) we can change the approach that state agencies use to partner with them. It sounds like a modest goal, but when this CDD work started in the 1990s, using local knowledge and delegating decision-making and resources to communities was about as popular in large development agencies as was farting in church. And, in fact, for all of the rhetoric today, they still don’t do very much of it – the general model remains, at best, one of setting up various flavors of ventriloquized “user groups” without any serious attention to the rules or processes for ensuring accountability, representativeness, or inclusion.

But while 3ie is “staggered” to find no impact on something that CDD has neither claimed nor disputed, our knowledge of what forms of collective action can give serious voice in the state-community encounter is still quite limited. That’s all the more reason to not discard the few tools that we have to expand that knowledge. I’m glad to see that 3ie now thinks that significantly increasing women’s attendance in village and subdistrict council meetings is indeed a worthy governance achievement, but their dissemination note for this study still says “CDD has no impact on social cohesion or governance.” And, in general, I would have thought that more local control and accountability over spending is a lot – not all, but a lot – of what any metric of changes to local governance would be measuring.

Let’s see if we can close with some sort of consensus. I agree with Howard and Radhika that at this point, evaluations of what works where and why are what is most important, rather than sweeping ‘CDD is good or bad’ claims. In my first comment I tried to explain why issues such as external validity, measurement intervals, and knowing your context variables are so important for this purpose, but for these broader issues of voice and power, I think we face a growing trade-off between increasingly precise measurements of increasingly less interesting questions about individual projects when we should be teasing out the links and interactions across different types of social action.

The most interesting work in CDD world these days is looking at complementarities between CDD and household cash transfers (Indonesia, Philippines); evolving the CDD platform into a whole of government frontline service delivery system (Afghanistan, India, Indonesia, Morocco), and coupling community development with better market access and returns for poor people (Nepal, Sri Lanka, India).  Dewi Susanti and her team in Indonesia are currently conducting a large scale RCT that is testing whether more community control over teacher’s benefits can make a dent in teacher’s high absentee rates, the kind of evaluation that, if true, would be immensely beneficial to multiple countries, just as India’s fascinating examples of CDD projects investing CDD funds to train likely to migrate rural youth in urban skills is just begging for analysis and transfer.

My own favorite – mostly because I recently visited them –  and still unmeasured example, was seeing how the activists from Indonesia’s female headed household empowerment NGO (PEKKA) used the transparency requirements of the CDD program to sponsor public debates between political candidates vying to become district heads over who would be matching those funds to help widows and other poor women.

To conclude, let’s be very careful about throwing out the baby with the bathwater. Any approach that reaches poor, neglected communities with large amounts of decent quality, pro-poor, cost-effective, self-targeting, economically productive assets should already receive a tick mark in the “plus” column. That all project models show variance in their quality is a truism, not an indictment, one which mostly puts the burden onto project owners to show that they can learn and improve. The way forward is to keep pushing boundaries, show a sufficient number of well-monitored and explained failures to prove that nobody’s getting complacent, and to keep using multiple types of analyses to cross-check and validate hypotheses.’

Last word? Let’s see…..

July 12, 2018
Duncan Green