A recent article in the Economist pointed to success of Ethiopia's social safety net programs. It said that social safety net programs formed 1.5% of GDP in Sub-Saharan Africa. It writes,
Ethiopia’s rural scheme is widely regarded as a success. It has reduced rural poverty and helped the poor buy food during a severe drought in 2016 that might have led to famine.
Now this is deceptive. What do we mean reduced poverty? Does it mean that people's lives have undergone a significant change? I guess all this goes back to the artificial minimalistic thresholds that we have constructed around per capita incomes to define poverty levels.
The World Bank defines Social Protection (SP) programs as consisting of social insurance (mainly public pension schemes covering old age and disability), social assistance (cash and in-kind transfers and workfare schemes, often targeted to the poor), and labor market programs (training, entrepreneurship support, unemployment benefits).
Martin Ravallion and Co explore the impact of social protection programs on the poorest, the floor level of incomes,
The bulk of the impact of SP in developing countries is due to public pensions, which lift the floor by $0.38 a day. This too is below the mean spending on such pensions, which is $0.67 per day. Social assistance on its own only raised the floor by $0.015 per day on average—merely 8% of the (already low) level of average spending on social assistance. The bulk of the impact of SP on the headcount index (5% points) is also due to contributory pensions. Social assistance on its own reduced the poverty rate by 2% points. Countries that spend more on social protection tend to have a higher floor. The correlation coefficient is 0.751. Mechanically, this relationship reflects both differing levels of SP spending and differing transfer efficiencies. Transfer efficiency in reaching the poorest varies greatly. We see that very few countries attain a value of FTE of unity or more. (Recall that this is the ratio of the gain in the floor due to SP to mean spending.) For the bulk of countries (87% of the sample), the gain to the poorest is less than mean SP spending. FTE tends to be better for social assistance on its own, for which the median value is 0.934, as compared to a median of 0.630 for all SP; 43% of countries have FTE for social assistance greater than unity. In addition to FTE, we measure the efficacy of SP in reaching the poorest 20%, giving our second measure of transfer efficiency, GTE. The two measures are correlated (r = 0.505), but certainly not perfectly; some countries are better than others at reaching the poorest people given their efficacy in reaching the poorest 20%. GTE is positively correlated with spending per capita (r = 0.656), but that is not true for FTE (r = -0.021). As countries spend more on social protection, a larger share of that spending tends to reach the poorest 20% but not the poorest.
Now the takeaways in English. We need to make the distinction between poverty alleviation (allows people to have three meals a day compared to two) and poverty eradication (allows people to have meat twice a week, or a more dignified human existence). SP programs in almost all the developing countries help address the former. It will keep people meaningfully above the biological poverty line. It does little, on their own, to help the poorest transition to any meaningfully higher income level. This is just stating the obvious - social safety nets are for poverty alleviation, and not elimination.
If this is the case, then how appropriate is to use indicators like increase in savings or increase in aggregate consumption to measure the impact of social assistance programs like cash transfers? A more relevant measure of impact would be just the change in food consumption - having enough to eat three meals against the typical one or two, or eating meat once a week, or something like those.
This is a bit like the debate about the role of women self-help group movement. It is often confused as an instrument of economic empowerment, when its more relevant utility may be as a tool of social empowerment.