Andy Sumner unpicks the bias and uncertainties within future global poverty forecasts.
A Brookings paper out this week (here) does something a set of papers have sought to do recently – that is make projections about the future of global poverty.
These kind of papers have significant policy implications because it is only by understanding both the future scale and anticipated locations of poverty that properly informed debates can be had on the scale and objectives of future international aid. And of course the whole post-2015 debate is mushrooming (see the latest here).
In a new paper for the Center for Global Development, we, meaning Peter Edward and myself, add to the debate by looking across a wide range of scenarios and methods to see the level of uncertainty and bias built in to these kind of forecasts.
We have three conclusions across a range of scenarios and approaches that somewhat tally with the Brookings paper but with some quite important differences.
First, ending extreme poverty is plausible but the level of uncertainty is enormous.
Across a wide range of scenarios, using different assumptions and methods, we find that it is plausible that $1.25 and $2 global poverty will reduce substantially by 2030. However this is by no means certain.
Different methods of calculating and forecasting poverty numbers give very different results as do an approach which takes account of changes in inequality.
Uncertainties over future, and even current, poverty levels are especially high for India and China (where half of the world’s poor people currently live).
While it is likely that poverty in those countries will reduce dramatically by 2030 it is difficult to have much certainty over just how large those reductions will be. Because of these uncertainties it is possible to conceive, under different growth scenarios and different assumptions about future inequality, that $2 poverty could be eradicated in India and China by 2030 or that it could be at or above current levels.
Second, don’t get too hung up on fragile states (at least not as the OECD defines them) because global poverty may well not be concentrated in such countries in the future – at least it is not a given.
Depending on what happens to inequality much of world poverty could still be in Middle Income Countries (MICs) but better still maybe it’s time to dump such aggregate classifications by income and fragility (both conflate so many different types of countries it raises questions about their usefulness).
If we take the groupings for a moment, currently most poverty is in middle income countries (MICs) and even when China and India are removed from the picture poverty is still more or less evenly divided between MICs and low income countries (LICs). Even with those two countries excluded the forecast poverty reductions in the remaining MICs are not so large, nor so certain, as to justify in themselves the view that poverty in the future will be a matter for LICs primarily.
In fact, once recategorisations are taken into account it seems that poverty outside India and China will remain roughly evenly distributed across MICs and LICs in 2030.
And looking to other possible classifications it may be that the World Bank’s shorter list of ‘fragile situations’ that emphasizes conflict/post-conflict countries is more useful but even then the UN’s widely used Least Developed Countries (LDCs) categorization might be just as useful or more so.
Further, we find certain kinds of method produce a bias towards moving global poverty into fragile states simply by the approach taken.
Third, it is startling just how much difference changes in national inequality could make – suggesting support for greater attention to this is the key implication of all these future projections.
Forecasts of global poverty are very sensitive to assumptions about inequality. In one scenario, we found that the difference between poverty estimated on current inequality trends versus a hypothetical return to ‘best ever’ inequality for every country could be an extra billion $2 poor people in 2030.
Taking a different scenario - one of optimistic economic growth, $2 poverty could fall from around 2 billion today to 600m by 2030 – if every country returned to ‘best ever’ inequality.
However, if recent trends in inequality continue it could rise there could be an additional 400m $2 poor in 2030 compared to today.
In sum, despite all these uncertainties there is benefit in using the available data to attempt to estimate global poverty in the future.
What actually matters is recognizing the uncertainty and bias and taking the time to look across a wide range of scenarios, approaches and assumptions for commonalities in future projections and then to be informed by that rather than the next set of poverty projections.
This post first appeared on Global Dashboard.