Complementarity in economic development policies

[This is a guest post by Anna Tompsett.]
If you work in development, or think or read about it, you’ll be familiar with the idea of complementarity.  You may not have called it by that name, but you’re sure to be familiar with the idea; that a package of interventions can be much more effective than interventions on their own.  
For example, if there is no road access to a village, then people inside can’t travel to access healthcare, teachers can’t get to schools in order to teach and farmers can’t get their goods to market.  Improving the clinic in the nearby town, or paying teachers extra to show up on time, or creating a mobile phone price information system, has little impact, because of the constraints of the existing infrastructure.  On the other hand, building the road doesn’t, in itself, improve the clinic, change incentives for teachers, or resolve agricultural market failures.  

Some schoolgirls in a very isolated area of Nigeria 
(Dadiya Hills, Gombe State) looking out across the 
valley.  You can't see the road, because there's a 
two-hour walk and a thigh-deep river in between,
but that's kind of the point.

It’s a critically important idea in policy; it has informed the Millennium Development Goals, and it’s a large component of the philosophy behind theMillennium Villages (love them or hate them).  Yet we have staggeringly little evidence for how important these complementarities really are.  If they are significant, they could cause us to systematically underestimate how effective development interventions could be in conjunction when we assess them in isolation.   

However, there’s a very good reason why there is little robust evidence on their magnitudes; it’s hard enough to design an effective field test, or find a natural experiment, when you’re interested in a single intervention or policy.  And the size of your required sample increases with the square of the number of interventions whose interactions you would like to measure.  (Anecdotal evidence suggests that the amount of luck required to find natural experiments may even increase with the cube of that number.)

Every time I get disheartened by these odds, however, I read an article like this one, and I’m reminded of why I shouldn’t lose give up looking for a context in which to study this issue with the rigour it deserves. 


  1. Fantastic piece. I wish we had talked about the intricacies and practicalities of this stuff more instead of just passively acknowledging it and moving on.

    Perhaps we could map out the overlapping systems dynamics of different development interventions, even if we can't measure their isolated impact. It's complex topics like these that require mixed (qualitative and quantitative) methods to articulate them properly. Could you imagine if that were one of the outputs of our program every year as we each furthered our research? An expanding integrated systems map of human and global prosperity?

    Awesome. Thanks for this piece, Anna.