The Aid trilemma: are complexity, scale and measurability mutually incompatible?

I’ve been reflecting on Owen Barder’s recent post on the tensions for aid agencies between wanting to go to scale, and acknowledging that lasting development solutions have to emerge from discussions among local actors, based on local context.

Seems to me we have something of an aid trilemma here – I would add in attribution to the mix as a third element. You can have two out of three of the following:

Able to go to scale (reaching millions of people, rather than a couple of hundred)
Compatible with complex systems (inherently unpredictable, discontinuous, shaped by local context)
Measurable and attributable (being able to say change happened and that it was due to a given intervention or action)

Two out of three gives you three different permutations:

Interventions in complex systems that go to scale, but whose impact is largely unattributable:

Working on the ‘enabling environment’ (rule of law, strengthening civil society, access to finance)
Improving quality of essential services (see Lant Pritchett)
Problem-driven iterative adaptation (Matt Andrews)
Convening and brokering relationships/acting as a catalyst/other kinds of leverage rather than trying to do stuff on your own

 

Interventions that go to scale and are clearly attributable, but are not suited to complex systems

Bednets, food aid and other straightforward(ish) service delivery
Essential Services – quantity (getting kids into school) rather than quality (teaching them anything – Lant Pritchett again)
Simple replicable institutions (football/soccer, Owen thinks postal services)

 

Interventions in complex systems that are clearly attributable, but cannot go to scale

Locally grounded work on governance, accountability etc
Using process tracing and other ways to measure the impact of ‘n=1’ interventions

 

The question then arises, what (if anything) lies in the sweet spot, namely interventions that manage simultaneously to go to scale, are clearly attributable, and respect the nature of complex systems? Owen reckons a number of CGD pet schemes, such as Cash on Delivery, tick all three boxes, but I’m most dubious on the attribution corner of the trilemma. A government reduces maternal mortality, you reward them with $x thousand dollars per life saved, but how do you know that the offer of that money had anything to do with the outcome? Social franchising might work – a core ‘project in a box’ that allows for plenty of variation according to local circumstance, like Savings for Change. Other suggestions?

What do you think – is the trilemma idea worth pursuing? I’m planning to think this through further with UNDP, Owen and others – keep you posted.