So... I wanted a picture for this blog but couldn't find anything to fully express the geekiness of what follows. Here's a nice picture of some cliffs instead.
There’s an iron rule of development research which says that when two or more researchers are gathered together at some point they’ll start complaining about how awful the data is. This is quite true. But of course it doesn’t stop any of us who do research from expounding our insights about poverty and inequality with great certainty. And the strange thing is that we know least about the things that we talk about most.
Usually when we talk about poverty and inequality we’re talking about incomes. And what do we know about incomes? Well, the gold standard is probably the World Bank’s Living Standards Measurement survey. It’s wonderful geek-fodder: huge amounts of information on incomes, assets, educational outcomes etc etc, all disaggregated for regions and gender and everything. But…..only five countries in Sub-Saharan Africa have LSMS surveys which include income data available on the Bank’s website. Of which only two are after 2000. This is perhaps not surprising – these surveys are expensive and costs millions of dollars to do properly.
It’s not quite as bad as it first looks. Another 30 countries have other surveys done after 2000 which measure what households earn or what they spend, all collected in one place by the quite brilliant International Household Survey Network. It’s data from all these surveys which are crunched together by the World Bank and used to compile the World Development Indicators and the PovCal Net database. These aren’t all exactly the same – some ask about income, some about how much households spend, but it’s possible to add them together. Together, they cover 82% of the population of Sub-Saharan Africa. Which is not bad but not great. It’s rather like making confident assertions about poverty and inequality in the European Union without any data on the UK, Greece, Switzerland or Sweden.
But though incomes data is patchy, we know much more about the assets owned by poor households, including land, housing and things like bicycles and radios thanks to the USAID programme of Demographic and Health Surveys (DHS) and UNICEF’s Multiple Indicator Cluster Surveys (MICS). Together information from these surveys collected after 2000 covers 99% of the population of Sub-Saharan Africa. Researchers have used these to create measures of poverty based on assets (e.g. here), but this approach hasn’t really hit the mainstream.
It seems odd that we know more about assets at a household level, but talk more about incomes when we’re talking about poverty. One reason might be that income is such a (superficially) clear, simple and persuasive indicator. Especially with the extra development of ‘purchasing power parity’ that lets you compare with apparent confidence between countries (though see Angus Deaton’s deconstruction of the PPP methodology – it’s not quite as simple as it looks, folks). By contrast, assets are tricky things to add up and compare with each other. How would you compare, for example, a household that had a tin roof and a radio, with one that had a bicycle and two chickens; or one hectare and a cow versus two hectares and a goat. It has been done, but it’s a lot harder to explain than ‘Ms X earns less than one dollar a day’.
Another reason might well be the politics. As my clever colleague Milo Vandemoortele argued very eloquently to me in an email exchange, if policy makers start to talk about asset inequalities then they might be expected to act on them. Land redistribution anyone? Probably not. People tend to steer clear of data that tells them things they really don’t want to know. And of course, an exclusive focus on incomes fits very nicely with a story about poverty reduction that begins and ends with economic growth. But once you bring in assets, things look a bit different and a lot more political.
I thought that we didn’t talk about asset inequality because the information wasn’t there. But it is, and in fact it’s better than our incomes data. Using the two together would provide much better information on poverty – think about the difference between someone in the UK earning £30,000 who owns their own house, and someone on the same income without any property.
Researchers are toiling away for hours to collect this stuff. Let’s try to use it as well as we can, no?
