This post originally appeared on Let's Talk Development.
The World Bank’s classification of economies as low-, lower-middle-, upper-middle-, or high-income has a long history. Over the years these groupings have provided a useful way of summarizing trends across a wide array of development indicators. Although the income classification is sometimes confused with the World Bank’s operational guidelines, which set lending terms and are determined only in part by average income, the classification is provided purely for analytical convenience and has no official status.
The classification is based on gross national income per capita, calculated annually in U.S. dollars using a three-year average exchange rate, the so-called Atlas methodology. Because the thresholds are frozen in real terms while real incomes have risen, the number of countries in the low- and lower-middle-income groupings have decreased and poverty rates have declined, even as hundreds of millions of people remain in absolute poverty. This is hardly news. When China crossed the threshold from low- to middle-income twelve years ago, a quarter of the world’s poorest people moved with it. And after India crossed the same threshold nine years later, a majority of the people living on less than $1.25 a day lived in middle-income economies. In another decade it is likely that no country will remain below the current low-income threshold, even though absolute poverty, whether measured by dollars or life expectancy or education levels, will certainly persist. As Martin’s note suggests, this calls into question the usefulness of the current classification scheme.
For this reason the Data Group at the World Bank is reviewing income classification system. To inform this review, we welcome inputs from all users and colleagues in the research and development community. We would appreciate any comments or suggestions on the following questions:
- Does a classification of economies by development status have some usefulness in analysis and research?
- Should the classification be based on average income or some other indicator or combination of indicators? Should income be measured by exchange rates or purchasing power parities (PPP). Should we go “beyond GDP,” and, if so, how should such a measure be constructed?
- Should poverty rates be used as a classification criterion? Does Martin’s analysis of “capacity for redistribution” have any application?
- How should the categorical thresholds be set? Are there “natural” dividing lines between categories measured along a GNI per capita scale or by some other indicator?
- How and how frequently should the classification be updated? Should the thresholds be updated or only the rankings of countries?
Please send your inputs to SBadiee@worldbank.org
Join the Conversation
Glad to see this discussion taking place as it's important, whether for cash transfers or any other intervention we want to study. At GiveDirectly our approach since this initial RCT has been to require researchers to randomize at the community level. We received enough qualitative feedback through our call center within months of the first experiment to convince us that within-village randomization... imposed some costs. It's just not a great experience to be poor, eligible, but lose a lottery to your neighbor. If anything the statistical results here are weaker than we would have guessed. We've also begun the process of treating controls from that original experiment. I'd also like to take the opportunity to request that the Bank give David unlimited time to work on inventing ways to increase the power of cluster-randomized designs.
Read more Read lessGreat (and important) discussion. I think expectations play a big role here, eg Berk's academic example may be slightly different because both junior professors hope (even expect) to get tenure, whereas most villagers - at least in the past - didn't expect large amounts of unconditional cash. Similarly I think Paul's statement above is too strong: if "losing" a lottery means you get $x instead of $y,... where x
Read more Read lessWhy not the following: every eligible household in a village is asked if they want to be in the study. They are told that if they say yes they will be randomized to get either $x or $y, where x 0. Most people will say yes because it is free money (hence the researchers don't have to worry much about selection), and those who "lose" the lottery still get $x and understand that the process was fair... and transparent and voluntary on their part. Of course you don't observe a 'pure' control group ($0), although you can still have control villages separately, but for x
Read more Read lessReally interesting discussion. Minor thing, the Cunha, De Georgi, and Jayachandran 2015 paper link isn't working. Second, to follow on Paul's comment about power calculations: someone should devise new power calculations that take negative spillovers into account. But until then, we have to keep in mind that there are also tradeoffs associated with cluster-randomized trials since they have lower power... and therefore cost more. Maybe there is an intermediate solution where we randomize at the individual level but tell the control group that they will receive the cash (in the case of cash transfers) later. We could then take the savings from not doing a cluster-randomized trial and give them to eligible households in treatment villages after the study is complete. A redistribution from enumerators to villagers?
Read more Read lessEthics in Development Economics I have raised my ethical concerns about this RCT in a blog post http://blogs.csae.ox.ac.uk/2015/11/ethics-in-development-economics/ There are a number of things I find objectionable about this research. This lottery seems to have been imposed on people. They did not consent. And unsurprisingly the 'untreated' households (Table 2) are less happy, have lower life satisfaction,... experience more stress and have lower indicators of psychological wellbeing. We did not need an experiment to document this standard human response. As researchers we should not treat poor communities/countries as convenient labs to address minor research questions and instrumentalize fellow human beings for our ends. Does the publication of a paper justify the treatment of other people in this manner? I would like to pose two questions in connection with this research: 1. Why do economists just discuss the identification strategy and other statistical issues rather than the fundamentals behind this research? 2. How this study ever get past the ethics committee in Princeton?
Read more Read less