As a poverty economist, I know how difficult it is to forecast poverty numbers for even the next year. But, since the World Bank Group (WBG) announced the goal of ending extreme poverty by 2030, I have been projecting poverty numbers of some countries and regions for nearly two decades into the future. I could not help but ask myself “What am I doing this for?”
Forecasting versus Benchmarking
I looked for answers by reading recent papers on similar topics. Most seemed to be competing to prove that their projections were the most plausible. Some seemed better than others, but they were not convincing enough to erase my doubt. Before giving up the search, I reread Ravallion (2012, 2013) , and at last got an answer. In both papers Ravallion projected future poverty under the assumption that the developing world maintains its current pace of growth and poverty reduction. Although his methodology ensures that his projection predicts future poverty accurately if this assumption holds, it is still uncertain whether this assumption will hold and, consequently, whether his projection will prevail.
But, Ravallion does provide a benchmark. His projection tells us whether the developing world should focus on maintaining the status quo or whether it needs to accelerate its pace. Furthermore, by changing assumptions, differences between new projections and the benchmark can inform us of key bottlenecks. Hence, I became convinced that poverty projections are useful if conducted for benchmarking and policy simulations and, along with my colleagues, began working on our recently published paper, “Is Extreme Poverty Going to End?”
A New Benchmark
The usefulness of benchmarking is determined by the underlying assumptions. Ravallion called his assumptions “optimistic.” Although I initially thought they looked reasonable, after studying them carefully, we realized some of his assumptions are indeed optimistic, possibly too optimistic.
Here are the three key assumptions of Ravallion’s and of our paper for comparison:
The growth rate of 4.5 percent is not so unrealistic, but it is unrealistic to assume all countries grow at the same rate. According to PovcalNet, there have been large variations in the growth rates of developing countries. In fact, poorer countries in general have grown more slowly; therefore the assumption of a uniform growth rate for all countries is optimistic and likely overestimates the progress of poverty reduction. The other two assumptions also seemed optimistic; thus we replaced these with more realistic assumptions and created a new benchmark.
Our benchmarking analysis suggests that if each country maintains its current pace of economic growth, population growth, and changes in distribution, then the global poverty rate will decline to 8.6 percent by 2030. This is significantly higher than the extreme poverty target of 3 percent. According to our decomposition analysis (figure 1), a key contributor to the high poverty projection is the large variation in growth rates of developing countries. Indeed, switching from the uniform growth assumption to the country-specific growth assumption raised the poverty projection by 5.7 percentage points. Introducing country-specific population growth rates and considering recent within-country inequality trends also raised the poverty projections, but the impacts were not as strong.
Figure 1: Decomposition Analysis
What Will It Take to End Extreme Poverty?
Two papers helped us organize our findings: first, Dollar, Kleineberg, and Kraay showed that the main driver of poverty reduction in the past has been growth, and that the contribution of reducing inequality has been minimal. Second, Basu clarified with an interesting hypothetical example that their findings explain well what happened, but do not necessarily indicate what is going to happen or what is needed to meet the extreme poverty target, and also argued for the importance of promoting shared prosperity.
Our findings are consistent with Basu and Dollar, Kleineberg, and Kraay. For example, like Dollar, Kleineberg, and Kraay, we found that within-country inequalities will have a minimum impact on future poverty if they maintain the trends of the last 10 years. But, like Basu, we also found that accelerating growth alone is not enough and that promoting the sharing of prosperity is necessary to meet the extreme poverty target. Indeed, our projection suggests that unless GDP per capita in the developing world grows at an unprecedented pace (around 10 percent) until 2030, the target cannot be achieved by accelerating growth alone. The only difference is that Basu focuses on the sharing of prosperity within a country, while our analysis suggests a slight extension of this concept—the sharing of prosperity across countries—is also important.
Our findings show interesting links among the WBG twin goals. It is clear that the extreme poverty target cannot be met without promoting the sharing of prosperity within and across countries. Also, meeting the extreme poverty target requires strengthening environmental and fiscal sustainability. Given the challenges in achieving the extreme poverty reduction goal, it is crucial to minimize countries’ risks of facing an economic crisis or a natural disaster—and if either of these fates does occur, it is critical to help countries recover as quickly as possible.
I remain skeptical of the use of projections to forecast poverty into the distant future, but I am convinced that projections can be useful if they are used for benchmarking or policy simulations, as shown by our study. Our work on this topic will continue and any comments or suggestions would be appreciated.