The Politics of Data – the bit the geeks forget?

October 23, 2015 12 By Duncan Green

Had a really thought-provoking conversation with Dustin Homer of Development Gateway last week. Development Development Gateway FoundationGateway was originally set up inside the World Bank, then spun off as an independent tech organisation, and focuses on helping governments and international organizations make better use of data in their decision-making.

So far, so technocratic, but Dustin got in touch because he read my piece on the politics of SDG implementation, and it struck a chord. After years of experience, Development Gateway seems to be going through the same process of realization (and adaptation) that Twaweza in East Africa have gone through, when you see that access to data and information alone doesn’t automatically lead to any changes in policy or behaviour. DG is now doing a lot of research about how to promote data uptake, and Dustin had some intriguing initial ideas about what is really going on:

  • A lot of data collection and use is largely symbolic. High officials put pressure on low officials to collect it, then put it all together and present it to as an offering donors. But ‘the enabling environment for data-based decisions is often non-existent’.
  • Too-much-data-300x150That reporting burden may be symbolic, but it is huge, eating up staff time on data collection without leaving time/incentives/skills for useful analysis. Dustin was struck by a conversation with an agricultural extension worker who had two parallel data collections systems – one she had devised herself, which was actually useful in doing her work, and an entirely separate one she had to fill in for the bosses back in the ministry
  • Instead, Dustin has had some conversations along the lines of ‘look, if some big cheese phones up and tells me to do something, I do it, regardless of what the data tells us.’
  • Changing that requires champions at senior level – as Matt Andrews has found with PDIA, that can create a permissive environment that allows mid level technocrats to get on with some data-based policy
  • There are often pockets of data-based management, many of them at local level, where officials seem to have more latitude – and we can probably do a lot more to support them

What emerged from this is a picture of an enormous, multi-billion dollar data machine that has so far been largelyEliot on information supply driven by data providers and lobbyists. There isn’t enough user-centred design (going to ask decision makers what data they actually need to do their job, then helping them devise efficient ways to collect it). Some are calling for direction change toward demand-driven, locally-relevant data solutions, but too often, international reporting mandates win the day. International players often pre-suppose what is needed and offer solutions or systems that meet certain reporting requirements – but may not get used for much else.

Some are trying to push beyond this – for example, the government and local donors in Ghana tried to run the tables and did a gap analysis of its data needs, but they are still trying to rally support (and funding) to do something about it.

All the momentum behind Big Data, data scraping, etc is if anything flying ever higher above the ground level where decisions are actually being made, and if we’re not careful that could come at the expense of local decision-makers. It would be horribly ironic if the move to Big/Open Data saw citizens become mere data generators – the object of data, rather than the subject.

Big DataSo where’s the demand-led, bottom-up data movement? Either at the top – something like the International Growth Centre, which would offer what amounts to a free, top level consultancy to developing country governments, or more radically, some ‘barefoot data’ movement that tries to build a new bottom-up approach to collecting data (but it would have to be data that can be aggregated to the point where it helps government decision makers). Who is the Robert Chambers of data? Would love to see some examples if you have them.

Two more general points:

There is bound to be a trade off between local relevance and comparability, both between places, and over time. How much effort should go into generating beautiful data that is only relevant to one village, and can’t be compared with any others or even with the same village a couple of years later?

Thinking back to the Twaweza example, it seems to me that we have to look much more closely at the arrows in our theory of change diagrams. You know, things like Access to Data → Better decisions. The problems always arise from the assumptions implicit in the arrows – maybe every arrow should have an explanatory footnote in our project documents.