A lot has been said recently about the reduction in global poverty over the past few decades. Although the positive coverage of these encouraging statistics is certainly justified, important questions still remain. Many of these questions relate to the dynamics of poverty, rather than simply snapshots of static poverty. In short, static poverty measurements (which are used when estimating global poverty at given points in time) cannot distinguish between people who have been in poverty their entire life and people who happen to have had an unfortunate circumstance in the year the poverty data was collected.
These distinctions matter, a lot. If poverty is chronic, meaning the same people (or households) are poor year after year after year, then policies and programs should focus on breaking this persistence of poverty. If poverty is transient, meaning different people (or households) are poor in different years, then policies and programs should focus on proving temporary assistance to ensure that unfortunate circumstances do not lead to persistent poverty.
Answering questions about poverty dynamics are challenging and complicated by data availability. Obviously, since the question asks about poverty status of the same person (or household) at two different points in time, this analysis requires panel data. Additionally, this panel must have a sufficient lag between survey rounds, because we don’t expect much to change in, for example, just one year. This leads to another challenge that the longer the lag between survey rounds, the harder it becomes to track down the same people (or households). This leads to survey attrition and weakened statistical analysis. There is some classic research on this question (see Lybbert et al. 2004 and Adato, Carter, and May 2006), but these studies focus on a specific region within a specific country and lack important external validity.
Two World Bank economists, Hai-Anh Dang and Andrew Dabalen, aim to address this gap in our collective knowledge. In a paper, entitled “Is Poverty in Africa Mostly Chronic or Transient? Evidence from Synthetic Panel Data“, recently published in the Journal of Development Studies, the authors address this question the best they can given the challenges discussed above. Due to the lack of panel data with meaningful coverage of African countries and with necessary lag between survey waves, the authors create a “synthetic panel”. This method essentially manufactures a second wave of data using information gathered with a nationally representative cross-sectional survey.
This method has both advantages and disadvantages. One advantage is since the second wave of data is not actually collected at a different point in time, there is no survey attrition between rounds. A key disadvantage is that the panel is not real, it is estimated. Therefore the validity of these results rest on the level of confidence one has in the estimation method. Another disadvantage of the entire approach is coverage of African countries is still lacking. The analysis includes only 21 countries in Africa. This is only about one third of all African countries. This is troubling because it is reasonable to presume that countries without data are also countries with more chronic poverty. It is worth noting that the 21 countries included in this study have relatively large populations, as the the study covers about two thirds of the entire population of Africa. Nevertheless, the authors aim to answer an important question as best they can with current data limitations.
Here are some results. The table below shows the decomposition of poverty rates in various categories: chronic poverty (poor in both time periods), downwardly mobile (non-poor in the first and poor and the second period), and upwardly mobile (poor in the first and non-poor in the second period).
The results of this analysis are interesting. The bottom row of the table above shows that one third of their sample is chronically poor, about 13% are downwardly mobile, and about 17% are upwardly mobile. The rest of the sample (if you are keeping track of the math yourself) are never poor. The authors highlight some interesting take-aways about comparing static and dynamic poverty between countries.
[C]ountries that are similar in terms of [static] poverty rates may be dissimilar in terms of poverty dynamics. For instance, Swaziland and Uganda both show a similar headcount poverty rate that hovers just above 40 per cent in the most recent period (column 2), but chronic poverty rate in the former (18%, column 5) is almost half of that in the latter (32%, column 5). Similarly, a country may have both more headcount poverty and less chronic poverty than another at the same time. For example, Tanzania has a poverty rate that is 10 percentage points higher than Senegal (that is, 48.8% versus 39%), but its chronic poverty is two percentage points lower than that of Senegal.
If you are a more visual person, here is a figure that illustrates the results presented in the table above. The countries are ranked in an increasing order based on their rate of chronic poverty.
The full paper includes a number of additional findings. For example, the authors find evidence of “pro-poor” growth in Africa with a 5% reduction in poverty and a 28% increase in the size of the middle class. At the same time, a category which the authors define as “vulnerable” (e.g. people who are at risk of falling back into poverty) also grows by 12%. The paper is also seems relatively accessible to non-economists, so I’d encourage all those interested to dig into it.
Research like this is important not because it has an air-tight identification strategy answering age old questions about what causes poverty or development (it doesn’t), but because it helps us (both researchers and practitioners) dispel misconceptions and ask the right questions. Yes, poverty is falling (on average) in Africa (for the countries in which we have data). This rate of poverty reduction is perhaps slower than in other regions of the world. However, within the diversity of contexts in the continent there is quite a bit of variation of experiences with poverty dynamics. Large shares of people are escaping poverty and a slightly smaller but also substantial share of people are falling back into poverty. At the same time, about one third of people (again, for whom we have data) remain in poverty year after year after year. Therefore, in the presence of “pro-poor” growth, some are being left behind. This suggests that greater efforts should aim to help the “poorest of the poor” out of poverty.