While working at CPAN, I have been introduced to Growth Incidence Curves (GICs), a graphical technique which shows how different percentiles of a population benefit from growth during a certain period. They can be used to identify pro-poorest growth (PP2G) episodes in developing countries.

What is pro-poorest growth?

An easy way to define it is it to say that PP2G is the relative version of pro-poor growth. It is not simply about favouring the poor households, but rather targeting the poorest of the poor - the ones who have been stuck under the poverty line for generations - and enabling them to catch up with the non-poor. For a matter of simplification, the chronically or severely poor are also sometimes referred to as the bottom 10% or 20% of the income distribution, depending on the poverty level in the country. CPAN has been focusing on this category of the population by studying the causes of chronic poverty and how it can be reduced.

It is rather common to state that economic growth is a powerful tool to reduce poverty. However, poverty reduction is not an automatic consequence of growth. When thinking about poverty reduction the question is not only to know how much income has been generated but how it has been distributed among the population. If growth is not distributed evenly across the whole population and favours mostly the richest, inequalities will increase and poverty will persist.  At CPAN, we advocate that, for poverty to be reduced efficiently and sustainably, growth, at least for a period, has to favour the poorest more than the non-poor. The bottom of the income distribution should be benefiting from a higher growth in consumption expenditures than the average, or than the non-poor. This is why pro-poorest growth is important.

How can we identify PP2G episodes?

In the framework of a CPAN project on economic growth, I have been trying to identify developing countries which have experienced pro-poorest growth episodes between 2000 and 2010. The aim of this project is to understand the context in which these episodes occur and to advocate for the incorporation of the notion of pro-poorest growth in the international debate around the Sustainable Development Goals. In April 2016, CPAN organised a conference in Manila on this topic. Andrew Shepherd and Chiara Mariotti, two CPAN researchers have recently published a Challenge paper discussing the policy implications of PP2G. Several working papers and the next Chronic Poverty Report will also be focusing on this and will published in the upcoming months.

One way to identify PP2G episodes is to use GICs. As mentioned above, these graphs show how different percentiles of the population have benefited from growth (in consumption) over a certain period. If the curve presents a negative slope, it suggests that the growth episode is benefiting the poorest more than the richest. Here are a few examples:

In the case of Tanzania, between 2000 and 2007, growth in consumption expenditures was positive for all percentiles. However, the poorest 10% experienced the lowest rate, while the richest 10% saw their consumption boost in the space of seven years. During this period the level of inequalities, measured by the Gini coefficient, increased from 37.3 to 40.3 (World Bank Data). As a result, from the GIC, we can see that the case of Tanzania is clearly anti-poor. On the contrary, in the second graph, as the slope is negative, the Philippines appear to present a PP2G episode.

But how accurate is the use of GICs in identifying pro-poorest growth episodes?

GICs are a very attractive method and easy to understand, but their interpretation requires critical thinking and careful consideration of the broader context. After performing some sensitivity checks, a few results have shown that misinterpretations can easily occur. Let’s take the case of Nigeria for example. Over the 2004-2011 period, the graph shows a nice and clear PP2G episode: the bottom 10% benefit from a 5% growth in consumption while the bottom 50% experienced a 3% growth rate. However, if the period is shortened by a year, the shape of the GIC drastically reverses: the growth of per capita expenditure is higher for the richer percentiles than for the poorest percentiles, making the period clearly anti-poor.

It seems very unlikely that the growth trend suddenly and drastically reversed in 2010. This shows that we cannot conclude anything from the Nigeria GIC and the PP2G trend needs to be verified by investigating the literature and performing other robustness checks. Indeed, the literature tends to confirm the anti-poor trend: the decade in question was characterised by an increase in relative and absolute poverty (Kale, 2012).

In theory we would like to check the robustness of a GIC with specific econometric techniques using time series (yearly) data. However, the data comes from household expenditure surveys, which are not performed on a yearly basis. The use of GICs restricts us to the analysis of periods between two surveys, leaving what happened between those years unknown, which can potentially bias the result.

Besides the period-sensitivity of this technique, GICs can also hide geographical disparities. In general the incidence of poverty is higher in rural areas than in cities. This can seriously affect the national trend, which will then end up not reflecting the reality (Grant, 2005). It is important to therefore look at rural-urban GICs and compare then when interpreting the national level curve. Once again, the issue is the availability of this kind of data. There are very few datasets which allow us to do this comparison. This is not a limitation to the GIC method itself, but to the type of data used to plot them.

There are a few papers which state other limitations of the GIC method and offer some alternative methods, based notably on ‘non-anonymous’ GICs. You can read Bourguignon (2010) and Grimm (2005) to go further.

So what do we do about GICs?

Despite the limitations of the GICs in identifying actual PP2G episodes, this method remains very useful.  It is a simple and comprehensive approach, which can be easily understood by a non-specialist audience. Interpretation of GICs, however, needs to be done with precautions and accompanied by verification processes. After plotting a GIC, the next step is therefore to investigate the past literature, looking for poverty trends, wage levels, the existence of pro-poor policies and programmes, infrastructure accessibility and so on. The trends of the GICs will give a good indication about where to look in the past literature . The GIC method can also give us reasons and opportunities to adopt a new perspective for certain countries. In the Philippines for example, the literature suggests that poverty reduction has not been responsive to economic growth over the 2003-2009 period. However, the GIC shows a rather pro-poor episode. This result might be wrong, but it is worth looking into it more and start thinking out of the box.

This is what I have learnt about GICs. Use them but stay critical

This blog post is written by Sophie Bridonneau, CPAN Research and Communication Assistant.