In This Module We Want to Work out How to Compare Poverty Between One Country and Another

Slide #1

In this module we want to work out how to compare poverty between one country and another.

Slide #2

There are seven distinct objectives in this module. We first need to explain how the World Bank goes about measuring poverty in different countries in a way that allows one to make cross-country comparisons. One has to pick an initial year and poverty line, convert that line from U.S. dollars to local currencies, deflate the local poverty line to measure it for those years for which we have survey data, and if necessary estimate the poverty rate using data from tabular or summary statistics. This then allows us to identify where in the world poverty has fallen most quickly over the past thirty years. The third objective is to evaluate the World Bank’s approach, which may be the best we can do, but does have its critics.

Once we have measured these effects, we can then begin to analyze the links between economic growth and poverty worldwide. David Dollar and Aart Kraay conclude that what really matters for reducing poverty is economic growth; we will examine their argument, as well as Martin Ravallion’s point that inequality is bad for the poor. The fifth objective is to define “pro-poor” growth, and ask what it really means. We will then explain how to decompose changes in poverty into the components due to economic growth, and due to changes in inequality. Finally, we will try to identify the paths through which macroeconomic shocks may affect poverty, using the recent case of Thailand as an example.

Slide #3

Before we go any further we do want to ask why we want to go to the trouble of comparing poverty rates internationally. One of the most convincing reasons is to help us target our scarce resources internationally. This is also essential if we are to be able to monitor the progress that is being made towards achieving the key Millennium Development Goal, which seeks to halve the proportion of people in the world living in extreme poverty between 1990 and 2015.

Slide #4

The standard approach, and the one most widely recognized internationally, is that used and developed by Chen and Ravallion at the World Bank. Step number one is to pick a poverty line, and they have done this by summarizing the official poverty lines in about 15 of the world’s poorest countries, and arriving at a figure of $1.25 per person per day in 2005 prices. This replaces the previous so-called dollar-a-day poverty line, and it is still sometimes referred to casually as the dollar-a-day poverty line.

This poverty line is in dollars, but has to be converted into local currencies. Because of the problems involved in using official exchange rates, we have to do this using purchasing power parities. This then gives us a usable local poverty line for our base year, currently taken to be 2005.

In step three we use the local consumer price index to construct the poverty line in local currency for every year going back to 1981, and going forward to 2010. Then, for every year for which we have survey data, we can apply this poverty line and generate a poverty rate; if we do not have direct access to the original survey data, it is possible to estimate a Lorenz curve based on tabular data, and to derive a poverty rate based on this.

Household survey data are not usually collected every year, so for any given country we only have poverty rates for some years, and we need to interpolate the poverty rates for the intervening years. In the final step, we aggregate the national poverty rates to arrive at our estimate of worldwide poverty, year by year. TheWorld Bankprovides poverty rates at three-year intervals, starting in 1981.

Slide #5

The place to go for Chen and Ravallion’s calculations is the PovcalNet site of the World Bank; the URL is shown here. Their most recent revision, which was done in 2013 but refers to 2010, uses data from more than 850 surveys in 127 less developed countries. The most striking finding is that since 1981, the poverty rate in developing countries has fallen from 52% to 21% in 2010, using the$1.25-a-day standard. In China it has fallen even more dramatically, from 84% to 12%, over this period. The reduction in poverty has been slowest in Africa, but even there, there has been a clear drop in poverty over the past decade. The next three slides provide some further details.

Slide #6

In this table we have put together the data from PovcalNet for the major regions recognized by the World Bank, for less-developed countries as a whole, and for China and India. As noted earlier, more than five out of every 10 people living in developing countries were poor in 1981; this has now fallen to two out of every 10 people, using the$1.25-a-day standard. The reductions have occurred almost everywhere, although it is worth noting that the poverty rate rose in Africa until the mid-1990s before falling quite substantially in the first decade of the new century.

The picture looks slightly different if we use a $2.50-a-day poverty line. Using this standard, 75% of the population of the world’s least-developed countries were poor in 1981, and 50% were still poor in 2010. Again using this standard, there were as many poor people in 2010 as in 1981.

Slide #7

This is a relatively straightforward sample graph that shows the evolution of the $1.25-a-day poverty rate for less-developed countries as a whole – that’s the black line that snakes its way down to the right – as well as for some Asian countries. China is shown by the dashed green line that plunges from the top left to the bottom right. India has shown a more moderate reduction in poverty, as has Cambodia until recently. Poverty fell in Vietnam at a similar pace to that in China, but with a lag of several years.

Slide #8

Here is another way of looking at the situation; the URL for an animated version of this graph is given at the end of this module. This is a bubble chart; the horizontal axis shows GDP/capita, the vertical axis shows the poverty rate, each bubble refers to a major region, and the size of each bubble reflects the number of people who are poor. In 1981, East Asia, shown by the red bubble, was the poorest region, followed by South Asia (which is given by the blue bubble); Africa, represented by the bubble with elephants under a tree, was somewhat richer, and had fewer poor people, both relatively and absolutely.

The decade of the 1980s saw reductions in poverty in China and India that brought them down to a par with Africa. Fast forward a decade, and we see that Africa has not moved much, and East Asia – which is dominated by China, of course – now has a lower poverty rate, and fewer poor people than South Asia. By 2010, East Asia has continued to reduce its poverty, but South Asia and Africa are moving down to the right too, although the larger African disk shows that the reduction in the poverty rate there was not fast enough to reduce the number of people in poverty.

Slide #9

There are of course plenty of methodological issues involved when we are trying to make international comparisons of poverty. The first concerns the choice of poverty line. It is essentially arbitrary, but a case can be made that it does not really reflect the cost of basic needs, as it would be impossible to live on $1.25 a day in a rich country.

The second issue is the purchasing power parity exchange rate conversion. The idea here is that we take $1.25 and we find how may rupees, for instance, would buy the same bundle of goods in India. Unfortunately, the PPP conversion is not terribly accurate even at the best of times, and moreover it is based on the cost of an average basket of goods, and not necessarily the basket of goods that a poor person would consume.

Third, within a given country, we have to adjust the poverty line so that we have a poverty line series over time, but ideally the price index should reflect the cost of goods consumed by the poor, which is almost never done.

If we do not use the Chen and Ravallion method, what might we do instead? Pogge and Reddy suggest perhaps it would be more satisfactory to measure the cost of basic needs in each country – for example, based on the cost of calories, plus a provision for non-food items – and then to compare that across countries and over time. This would require a great deal of computation, and it would be very difficult to standardize the protocols for making these calculations on an internationally-comparable basis.

Slide #10

Once we begin to think about international comparisons of poverty, we can begin to do some analyses that we might not otherwise have considered to be feasible. One of the most important such studies is the one done by David Dollar and Aart Kraay about a decade ago, in which they argue that growth is good for the poor. They gathered data from 137 countries, both developed and poor, between 1950 and 1999, and identified 418 episodes, which are intervals of at least five years for which there was a household survey at the beginning and at the end. This allowed them to measure both average income growth, and the growth of the income of the poorest quintile.The question they ask is very simple: If the economy grows, what happens to the income of the poorest 20%?

Their main conclusion is clear and striking: over the long term, the income of the poorest fifth rises in line with average income. For instance, a regression of the log of the per capita income of the poor on the log of average income per capita fits well – the R2 is 0.88, as shown here – and the coefficient on income per capita is close to one. This mean that, say, a 10% increase in average income is ultimately associated with a 10% increase in the incomes of the poor. This is a robust result, and the fit is not markedly improved by the addition of any other variables. It means that in the long run, one helps the poor by ensuring the economic growth is sustained at a good rate. This does not mean that growth alone matters, because with greater affluence, most countries find ways to help the poor directly, especially by putting social protection policies in place.

In the short-run, the picture is somewhat less clear; while a change in average incomes is associated with a comparable (proportionate) change in the incomes of the poor, the fit is not nearly as close. So in the short-run one might well have a country where average income is growing, but the incomes of the poor are stagnant. This may leave a greater role for public policy, to cushion short-term shocks to the poor.

Slide #11

Here are two of the most fundamental graphs from the Dollar and Kraay study. The first graphs the log of per capita income of the poor on the vertical axis against the log of average per capita income; the close fit is clearly evident, and we take this to reflect the long term relationship. The right-hand panel graphs the change in the log of per capita income of the poor against the change in the log of average per capita income. The relationship is still positive, but the fit is much looser. This may be interpreted as showing the short-run relationship.

Slide #12

In a subsequent study done in 2004, Aart Kraay more specifically looked for ways in which policies might help the poor over the medium term, and had difficulty finding any. In effect he was looking for possible policies that might make economic growth more “pro-poor”. This is an elusive term, about which there is little agreement. Let us consider two definitions, with the help of the diagram, which plots the growth rate for the poor on the vertical axis and the overall growth rate on the horizontal axis.

The first definition states that growth is pro-poor if the incomes of the poor are rising, as is the case of all the countries graphed here, with the exception of Zambia, although we might also note that in four of the countries the income of the poor is growing more slowly than the national average. The second definition is that growth is pro-poor if the incomes of the poor grow faster than those of the country as a whole; by this definition, Ghana and Zambia have pro-poor growth, even though the incomes of the poor are actually falling in Zambia.

Slide #13

It may be useful to decompose the reduction in poverty into the part that is due to economic growth, and the part that is attributable to changes in the distribution of income. Datt and Ravallion have proposed just such a decomposition, as set out in the equation here.

The growth component is obtained by taking the initial poverty rate, raising everyone’s income by the average growth rate, and computing the new poverty rate. This is equivalent to asking what would happen to poverty if there were no change in the distribution of income. The redistribution component is computed by taking the poverty rate in the end period, reducing all incomes proportionally so that the average is the same as in the initial period, and comparing the resultant poverty rate with the one actually observed initially. There is also a residual term, which is usually quite small.

This table provides some information that illustrates the situation. Take the case of Peru, for instance; between 1996 and 2002, the poverty rate actually rose from 28% to 32%; economic growth should have reduced the rate by 5.7 percentage points, but rapidly rising inequality more than negated this effect, as the poor got left behind. For a very different experience, consider the case of urban China. Between 1996 and 2001 the poverty rate fell from 9.7 to 6.5 percent, fuelled by solid economic growth that outweighed the rising inequality. As the period under consideration lengthens, the stronger is the growth effect relative to the distributional effect, because distributional changes are usually much slower.

Slide #14

To repeat: even if growth dominates in the medium and long term, distribution really does matter in the short term. Martin Ravallion does make another important point, which is that inequality is bad for the poor. Economic growth will reduce poverty less quickly in a society that is unequal, as we illustrate in the next slide. However, Ravallion finds no clear link between economic growth on the one hand, and inequality on the other, but this is an area of continuing controversy.

Growth matters; and inequality matters; but we are still left with a central challenge, which is whether we can find policies that can help the poor directly, and if so, what they might be. We return to this fundamental issue in module 8.

Slide #15

This slide illustrates graphically the effect of growth on poverty. The top left panel shows an initial distribution of the log of income, which then moves right by 50% in each step, or about what one would expect every decade in a rapidly-growing country. The poverty rate is given by the proportion of the curve that is to the left of the poverty line. Starting with an initial poverty rate of 50%, the poverty rate will fall to 31%, and then to 16%, and eventually to 2%, as shown in the bottom-left table.

Now consider a society that starts with the same mean income and 50% poverty rate, but has a more unequal distribution of income, as shown in the top right panel. With economic growth, the poverty rate will decline, but more slowly, eventually dropping to 9% rather than 2%. This underlines Ravallion’s point that while growth pulls people out of poverty, inequality makes it harder and slower.

Slide #16

One of the things that interest us is the way in which shocks to an economy – such as a rise in the price of food imports, or a reduction in exports – affect poverty in a country. The links between external shocks and poverty are weak, they are unclear, and they are specific to the country and the time period. We can illustrate these themes using the case of Thailand; the “great recession” of 2008-2009 was associated with a 19% fall in Thai exports, a 14% reduction in the number of foreign tourists, and a 2.3% drop in real GDP. The graph illustrates this, using quarterly data from 1994 through 2011: the solid line shows the annualized growth of exports, and from it we see how remarkable the fall was in 2008-2009. The bars show the annualized growth rate of GDP, which fell sharply during the Asian Financial Crisis of 1997-1998, and less dramatically in 2008-2009. So now we are ready to ask what happened to poverty in Thailand during this recent recession.

Slide #17

The answer to the question of what happened to poverty in Thailand during the recent recession there may essentially be seen in these graphs. Both graphs show monthly data from 2007 through the middle of 2010, from the excellent Socio-Economic Surveys. The data have been deseasonalized in order to reveal any underlying changes. The vertical axis gives the log of per capita expenditure. The blue shaded area highlights the period of recession, which ran from October 2008 through September 2009. In the upper graph, the top line refers to Bangkok, and the other four lines to the major regions of Thailand. The most striking finding is that expenditure per capita was maintained, during the recession, in all of the regions of the country, despite the significant fall in GDP. The bottom graph shows the evolution of spending for the top decile, given by the dashed line, the fifth decile, and the bottom decile. Here too there is no evidence of a fall in spending during the recession. There are at least three good explanations for the modest effect of the recession on poverty: first, the worldwide recession kept some prices in check, including many commodities; second, households were able to smooth their consumption, confident that the recession would not persist; and third, the government responded actively, sending checks to the poor and elderly, and increasing subsidies to some utilities and to education.