For this podcast, I sat down a few months ago to discuss data and development with Claire Melamed, who runs the Global Partnership for Sustainable Development Data. Apologies for delay, Claire – got caught up in internal traffic. Also apologies for length of this transcript – turns out 30m talking = 2 blog length pieces.
Duncan: Like any good Englishman, I’m going to start with an apology, which is that when you got the job, I sort of went ‘meh, data, who cares?’ And you were a bit upset because you had just got this shiny new job. And I clearly was wrong (I’m used to that). So perhaps you could tell us why we should listen to you and not me, and why we should care about data?
Claire: Well, that is very charming of you, as always. And I have to say you were not the only one, it has been part of my job, over the last seven years, to convince everybody that data is in fact, interesting and important for lots of reasons.
I think for me, the central motivator here has always been about power. And the way that data can confer power. We know, from our own lives, when we’re making decisions, the two things that we need to be able to have control. One is resources – having the money and the assets that allow us to have the freedom to make those choices. And the other thing is information to know what are the consequences of choices that you might make? What are the options available to you?
Now in the development sector, and those that are concerned with global power and inequality, we talk a lot about resources. It’s hugely important. We talk about aid, debt relief, financing instruments of all kinds. But we haven’t really talked very much about information. And for me, the exciting thing of the moment is this new interest in information as a source of power. How are we going to govern it, who has access to it, who produces it, and how we can use it to make the world a better place?
Duncan: You were saying before we started information is also a source of wealth?
Claire: Absolutely. I think we’re all quite used now to the idea that information is critical for governments to make good policy. And we understand that we need to have information on everybody in society, we need to understand the lives of people who are marginalised to be able to make good decisions on their behalf.
We realised that really acutely doing COVID, how absolutely critical information was to make good decisions and in some cases, the consequence of not having it and some of the very bad decisions that were made.
But increasingly, and this is also driving a lot of the geopolitics of data, it’s not just about the smallish world of data for government decision-making. It’s increasingly a much bigger world of data as an economic asset. People are interested in data because it’s the raw material of the future. It’s what powers AI systems, it’s what powers digital transformations in government. It’s a huge economic asset. And that’s why suddenly data is on the agenda of the G7 and the G20. Everyone now cares about it, not because they’ve suddenly developed a deep love for statistics, but because it’s driving economics.
Duncan: That’s really interesting. It’s like it’s become a commodity, but one that is very evenly distributed compared to copper or coal. Is that right?
Claire: Well, not really. The frustrating thing about data is that it could be evenly distributed – as a product of human interactions, of surveys, or censuses, or a product of human behaviour, or the data that we all produce everyday with our mobile phones. That couldn’t be more evenly distributed.
But, as is so often the way, we have managed to produce huge inequalities in a very short time, in terms of who controls that data.
Duncan: So let’s get on to that. You said right at the beginning, that you’re interested in the issue of power around data, which is obviously perfect for this podcast. So could you dig into the links between data and power?
Claire: Okay, that’s a big question. Let me try to separate that into sort of economic power and political power. We were just talking about data and economic power, and the way that the economic power that comes with data, the power to control what is this huge resource is already very unequally distributed.
One of the interesting things and this comes into debates that you and I were both involved in the early 2000s, is about sovereignty in the area of international trade and the extent to which protectionism is a good strategy. That debate is reproducing itself in the economics of data and countries are increasingly interested in making laws around data sovereignty, keeping the data which is produced in their country within their borders, insisting that even when multinational companies are collecting data, through social media apps, and so on in their countries, they shouldn’t be allowed to repatriate the data, they have to host it in data processing centres, domestically, and so on.
Duncan: Wow. So are there splits between advocates of data sovereignty v data security, like there were on food sovereignty and food security?
Claire: Exactly. And in the same places. So there’s two debates going on internationally at the moment about data, one is in the G20 and one is in the G7. The battle lines are not nearly as hard as they were on trade policy discussions 20 years ago, but there is a sort of school of thought that says that data sovereignty is really important that countries should be looking to hold their data internally. UNCTAD did a really good report on that. The G7 is more interested in the free flow of data across borders – a globalisation type approach to data
Duncan: So Data Liberals v Data Protectionists?
Claire: Exactly. I mean, the weird thing about data, of course, unlike a commodity like copper is that data can be in more than one place at once. You can have data that is both in India and in California. So, we’re kind of trapped, I think, in this way of thinking, which is based on our experiences of trading goods, and I’m not sure that that’s doing us a great service when it comes to imaginative solutions for trade policy.
Duncan: Okay, this is all very interesting, and quite meta. You also talked a little bit about yet another very not topic at the moment: decolonization. Data is a project – that became clear when the SDGs were being designed. The Millennium Development Goals, the Sustainable Development Goals are a multi-billion dollar project, controlled by a certain number of thinkers, countries and institutions. How’s it doing on the whole decolonization question, the idea that power needs to be distributed or is the driving intellectual and financial force behind it still quite centralised?
Claire: This comes I think, onto the second power question, which is data and political power, and the way that data is collected and used by public sector organisations. There are lots of different forces at work here. First of all, I think we have to recognise that most governments have not cared that much about data. If the government has a choice between recruiting a couple of more statisticians or building a new health clinic, we know what they’re going to choose, even if the statisticians could help them to avoid wasting a lot of money on health experts situated in places where they’re not needed.
To me, so the politics of data are really, really important. The economics and the politics are now feeding off each other. Governments are becoming more interested in data, because of the economic drivers. And this is spilling over into some governments thinking about their whole data system, including the data that they need for decision making.
The politics of data are different in every context. But there’s a couple of patterns here. One is, of course, as you would expect, people that have less power are less visible in the data. Just like everything else in policy, data reproduces patterns of inequality. So, famously, women tend to be underrepresented in data. If you look at things on which there is less data, they are often things like gender-based violence, things that tend to affect people that have less power and less money.
There is an amazing First Nations advocacy group in Canada that that we work with, who’ve done a huge piece of work on trying to make sure that they’re represented in the data; that the Canadian dataset reflects their life. They say, ‘Look, whenever the government comes to collect data from us, all they want to do is collect data about how ugly we are, and how poor we are, and how dysfunctional we are. They’re not asking about all the great things in our communities, the community spirit, the way that we work together, all of the things we’re really proud of – they never get into the dataset. So they’re trying to make sure that they’re collecting data on their own community that reflects the totality of their life and feeding that into policymaking.
So there’s a power element at the national level, the way that the resources that are allocated to data and the assumptions about what is important enough to be collected. But there’s also a global element to this, and this is where I come back to your question on decolonization, which is about the international architecture of data collection.
For those like me, whose recent career has been very much underpinned by the Sustainable Development Goals, and the belief that having global norms and global goals on things does help, at least somewhat to drive positive change, we have, I think, to think hard about the way that this has sometimes played out.
On data, first the Millennium Development Goals, and then the Sustainable Development Goals created a whole international architecture of data collection, a whole set of things that are important – the indicators, which we’re going to measure, and then create an institutional architecture at national and international level. The UN agencies are responsible for that and they are accountable for data collection. Countries that didn’t have strong institutions or resources, they had lots of agencies coming in and collecting data and a lot had very good relationships between UN agencies and governments that needed the data. But it’s really important that they set their own strategies.
Duncan: So I remember, back in the early 2010s, in the first discussions on the SDGs to replace the MDGs started taking place. What really struck me was that the driving force for the whole discussion was ‘Oh, my God, we’re not going to have a reason to collect data any more. We’re not going to have this mechanism for generating internationally comparable data. It was a bunch of UN people that were the driving force behind this conversation and right from the beginning, it seemed very striking that that the issue of comparability was really, really crucial for them. But if you’re in a country, you don’t really care about comparability. You want to know what’s relevant to your country, maybe comparability within the country, but if you’re Bangladesh, you don’t really care what the score is in Tanzania. So it seems skewed by the fact that it was born from this constituency.
Claire: I think so. And I think that does speak to who has power internationally. We all take it for granted – we’re so used to the idea that international comparability is important. It helps us to make a case for what we want. And it is really critical. But as you say, for national level policymaking it’s less useful. We don’t want to go too far and say it’s pointless, because countries are connected to each other. We are in a global world where there are global policies on aid, trade, debt relief and so on. So we don’t want to go too far here and say, international comparability is not important. Of course, it’s important.
But I think we want to retain a balance between the two. I mean, we’ll never know for sure, (I would say this, wouldn’t I!), but I think on balance, the kind of global focus on data has been positive, because I’ve seen how that has galvanised the national level conversation on data. There’s a sort of global solidarity of the sort of people who care about data that has been created by the fact that there’s a global conversation on data driven by the SDGs, which has also led to national progress.
Duncan: Earlier you mentioned of gender and data. What examples have you seen where data is gender blind?
Claire: One of the data stories that I think is really interesting and important comes from BRAC, which is originally Bangladeshi but now it’s global. They’ve done so much innovative work that has driven so much of this sort of thinking across the whole sector.
They worked on a project in some of the urban slums in, in Bangladesh, where maternal mortality rates were really high, there were no services. And often this was at least in part due to invisibility – informal settlements hadn’t been mapped; there wasn’t a census.
They wanted to invest in services for maternal health, and had to start out with a whole mapping exercise – actually go round, count the houses, count the people, how many pregnant women were there? And then collect data on what was it that people wanted? And where should they put these new health posts etc? One exercise was social mapping, where they talked to women, and found out that these informal settlements often have very poor infrastructure. Women who were pregnant, felt uncomfortable crossing some of these very fragile bamboo bridges that were often created to get over ditches.
So once they knew that, they built the maternity centres so that pregnant women had the minimum number of bridges to cross to get to them. And that was one of the things that helped to increase uptake of maternity services. It’s about understanding that everything, even something as basic as drawing a line on a map, has a social context. And you need to understand that if you’re going to get the right data and use it in the right way.
Duncan: A lot of the development sector at the moment is increasingly focused on places where states are either absent or predatory – fragile and conflict-affected states. Now, those are the places where not only do they not have the capacity, but they probably don’t even have the interest in collecting data. So do you just wait for those to stop being fragile and conflict-affected and then go and say, ‘Hey, let’s collect data’ or can you do anything while they are still really messy and difficult?
Claire: We have agonised over this a lot from our earliest days. We do have to recognise the limits of data – there’s no single solution to everything and that includes data. Some of the UN agencies that work in some of these humanitarian contexts have actually been at the forefront of innovations in data, so there’s a lot of ways in which data is relevant in conflict situations, but that’s not always true of their governments.
But some of the governments care. For example, the government of Somalia is putting a huge focus on data. I remember meeting the new Minister of the Environment at the last Climate COP. She was saying, ‘Okay, I’m the first Minister of Environment there’s been in Somalia, and I need to do two things: I need to physically build myself a building, because we don’t have one. And I need to build myself a data system so that I understand what I should be doing. That was a good day!
Duncan: I’ve been involved in research in Myanmar, where it became clear that people were so suspicious that they didn’t want to give data, because they were worried about what the government was going to do with it.
Claire: Yes, and in some cases, that’s absolutely justified. The fact that there are some governments doing great things with data doesn’t mean that other governments are not using it to oppress people. Like any powerful thing, we have to be careful how we use it.
Duncan: I get the impression that you’re in this shrinking world of what Karl Rove called the ‘reality-based community’. You’re trying to get data into the hands of policymakers, on the that policymakers will use that data to make better decisions, to help their constituents, whether citizens or institutions. But then you think of Duterte, Bolsonaro, Trump. So do you feel like you’re in a sort of shrinking world of data-relevant policy? And outside is all barbarians?
Claire: I think, in any world, we all create bubbles for ourselves. There’s a certain group of people who I don’t engage with much in my work; I’m not out there every day on the barricades trying to persuade Trump supporters that the election wasn’t stolen or whatever, because that’s not a good use of my time in my job. But that sets the context massively and I find that quite galvanising: if that is the world we’re in, that makes this work even more important to defend things that we didn’t ever think we were going to have to defend.
Duncan: So this is defensive strategy and strengthening the good guys?
Claire: I mean, I want someone to take on the bad guys. I don’t particularly think that’s my job but I’m not sure anyone’s found a very successful way of doing that yet. I think we just have to come up with stories that are as good as theirs, we have to make the truth as attractive as the fiction through storytelling. And I’m just not sure that we’ve ever found a way to do that. And that’s the story of every campaign ever, pretty much!
Duncan: Research shows that if you show climate deniers the data on climate change, they become stronger climate deniers. So, so as you say, the issue there is what story you’re coming at it with, not ‘we have the data and you’re wrong’.
Claire: Linked to climate, I think we also have to be clear about when lack of data is the problem and when it’s not. The lack of political action on climate is not because we don’t have the data on climate change – we’ve had that for decades. The lack of action is because these are really difficult political challenges and entrenched interests in the way of progress. So we have to understand, you know, where is more data the solution? I think data can help. Once you know you want to tackle climate change, then data can help you to decide how to do it.
Duncan: That’s interesting. So there’s an overlap there. Going back to my initial scepticism, I was at the height of thinking about how change happens, and stakeholder mapping and incentives and so on. But I guess there must be a role for data in understanding the political economy that is preventing change or driving change. I mean, stakeholder mapping, for me is an incredibly vague science; people wave their hands around and say, ‘Oh, faith leaders’ or whatever. Have you seen good exercises in data collection around power around politics, and stakeholder mapping?
Claire: I, personally, am probably just about enough of a Marxist to believe that you can read a lot of the politics from the economics. And I think that’s true of data, too. You know, who’s gaining? Who’s losing? Where are the assets? What are the trade patterns? Who owns what? Who’s controlling the economy. All of that you could collect data on and that gives you a lot of clues about the stakeholder mapping.
Duncan: But the other sources of opposition things like ideas, ideologies, are harder to collect data on, but not impossible?
Claire: No, not impossible, and maybe easier now that we have social media and we have a bit more of a glimpse into what people think about because they’re telling us much and in ways that we can collect.
Duncan: I’ve always felt that when people talk about networks and complex systems, that you have the geeks who think they can crack the code, and the hand wavers, who say, it’s so complex, there’s no point and therefore we have to ‘cross the river by feeling the stones’, to try things out and see how it goes. I’m definitely in the second camp, partly because I’m so rubbish at technology. But you’re suggesting that there may be something that can be done to crack the code?
Claire: Absolutely, but I also don’t understand why you have to choose between the two. You could crack the code, and then you can use that to inform the strategy. And then if the feeling the stones bit suggests that that strategy is wrong, then you can go back to the code. Why would we not do that?
Duncan: I wanted to finish on the other topic at the moment. Where does AI fit into this? So is this turbocharging your job by making data more valuable? More worrying? More exciting? How does it feel seeing the AI wave break over your world?
Claire: All of the above! The lens through which I look at AI is unsurprisingly that it’s all about data, because the raw material of AI is data, it’s built on data, you can’t have AI without big datasets. So a lot of the problems that we worry about with AI, like the problem of algorithmic bias, are problems with the function of the datasets as much as the algorithms. They’re both absolutely critical.
Duncan: Claire Melamed, thanks very much for coming on the podcast. That’s been fascinating.
Claire: Thank you. That was a lot of fun.