The World Bank’s New Poverty Estimates – Digging Deeper into a Hole Sanjay G. Reddy

The World Bank has today[1] released what it refers to as ‘updated’ global poverty estimates. These new numbers are based on a new worldwide price survey and a new benchmark international poverty line of $1.25 2005 PPP which replaces earlier benchmark poverty lines (of $1.08 1993 PPP and $1.00 1985 PPP, both widely referred to as “$1 per day”) corresponding to earlier base years. The revised figures purport to estimate world poverty figures for a range of years since 1981, and thus crucially affect our understanding of the world over the last quarter century of globalization.

Many aspects of the global order, such as the movement toward freer trade, as well as national institutions and policies, are defended by referring to their effect on the poor.  The Bank’s poverty estimates are thus central to their assessment.  Moreover, the Millennium Development Goals are defined in terms of these estimates, making this revision of great importance for determining whether the world is on track to reduce poverty by the amount required by the Goals.

Can the Bank’s new estimates be trusted? Can they be trusted more than its own earlier greatly lower poverty estimates, which they are intended to replace? Unfortunately, the Bank’s new estimates are based on the same methods it used earlier and are undermined by the same problems as the earlier estimates.

Two problems are foremost, as noted in a widely cited critique by Sanjay Reddy and Thomas Pogge.[2]  The first is that the Bank’s chosen international poverty line is far too low to cover the cost of purchasing basic necessities.  A human being could not live in the US on $1.25 a day in 2005 (or $1.40 in 2008), nor therefore on an equivalent amount elsewhere, contrary to the Bank’s claims.  Indeed, it appears to be far too low in many countries to account for the cost of purchasing basic necessities.  That this is so is self-evident in the case of the United States—the base country in whose units the international poverty line is defined. One’s daily income can be a great deal higher than $1.25 and still leave one unable to fulfill basic nutritional requirements, let alone the other requirements of a minimally decent life.  Since the international poverty line is defined in purchasing power adjusted units, meant to capture a constant level of purchasing power across countries, this incoherence is not easy to overcome.

The Bank’s claim that its poverty line is sufficient in other countries, despite being insufficient in the United States, implicitly acknowledges the second problem: that the Bank uses inappropriate purchasing power parities (PPPs) to convert its poverty line across currencies. Consider the question of how many rupiahs are needed in Jakarta to possess the purchasing power of a dollar in Washington, D.C.: the question cannot be answered without first establishing the purpose to which the money is to be put.  If the purpose is to purchase the goods needed to escape severe poverty (such as staple foodstuffs, which are internationally tradable and the prices of which tend more closely to reflect market exchange rates) the rate of equivalence may be different than if the purpose is to buy domestic services (which are typically relatively cheaper in poor countries as labor is not similarly internationally mobile).  We argue that the rates of equivalence used by the Bank typically understate substantially the relative cost of purchasing in poor countries the goods which are needed to escape poverty.  In addition, the PPPs calculated for each country inappropriately reflect irrelevant information about the pattern of consumption in third countries other than the country in which the price level is being assessed and the base country with which prices are compared (the United States).  This is because the worldwide pattern of consumption determines the weights placed on different commodities when assessing the price level in each country. The resulting problem of country irrelevance compounds the problem of commodity irrelevance just noted. The Bank’s new $1.25 poverty line is itself based on an average of poverty lines allegedly used in poor countries. However, as before many of these poverty lines have been defined by the Bank itself and they are translated into common units using the very PPPs the application of which is in question.  The claim that the international poverty line (whether the new or the old one) is representative of standards prevailing in poor countries is untenable.

Within the Bank’s current approach, these problems can be mitigated but not overcome.  The underlying source of the problems is the lack of a clear criterion for identifying the poor, and that basic deficiency remains unaddressed.  The features that supposedly constitute improvements in the Bank’s estimates lead only to increased confusion.  We have no basis to conclude that the new set of PPPs employed by the Bank generate poverty estimates which are closer to the ‘truth’, despite the Bank’s rhetoric to this effect.  Rather, we can only conclude that they are differently distorted than the earlier ones. The direction and extent of the new distortion likely varies from country to country, making it all but impossible to determine which set of estimates is more accurate.  The Bank’s declarations mask the fact that it has been building castles with sand.

The new PPPs produced by the International Comparison Program do indeed incorporate certain methodological improvements –but these are largely improvements which have little benefit for poverty assessment. For instance, the better measurement of the quality of government services, which is an important source of the discrepancies between the most recent PPPs and the last, is quite important for assessing real national incomes but largely irrelevant for determining the number of poor.

Even if the latest PPPs present a better picture of relative prices in 2005, that does not make them a better basis to judge poverty across countries in the previous years in which poverty must also be estimated to assess trends. In fact, as we have shown in our work, the relative extent of poverty in different countries and years, and the estimated trend, is

so dependent on the base year chosen for the exercise that there is no convincing basis to pick the estimates corresponding to one base year over those corresponding to another.

These fluctuations are inherent in the way that present PPPs are constructed, because PPPs reflect the relative costs for a worldwide pattern of consumption prevailing at only one moment in time, and this pattern is constantly changing.  The notion that the new PPPs constitute merely an “update” which better captures the reality over the entire period being assessed is badly mistaken.  Rather, they merely present a snapshot of relative prices across countries at a point in time which is no more authoritative for intervening years than similar snapshots of the relative prices across countries taken at points closer to those years. The use of national consumer price indices to identify the local equivalent of the international poverty line in years other than the base year further diminishes comparability across country-years as each such index refers to the price of a basket of goods with a composition entirely different from the pattern of world consumption which is used to calculate price differences across countries in the base year. The discrepancies can be substantial.  The Bank implicitly admits this by substituting the new $1.25 international poverty line for the 2005 U.S. equivalent of its earlier $1.08 1993 poverty line (which is close to $1.45 in 2005 prices).  Estimates of the level of poverty in each country and of the relative extent of poverty in different countries depend greatly on the choice of base year (and associated international poverty line).

The new estimates of the proportion of the world’s population in poverty suggest that the number of poor is almost fifty percent more than the Bank had previously proposed.  Although the rate of poverty reduction since 1990 is almost the same under the new estimates as the old, this appears to be a fluke and plausibly attributable to data errors.  If the final year of the comparison is moved backward by just three years to 2002, for instance, the rate of reduction of world poverty appears notably less favorable under the new estimates.  The only region that appears to have had a faster rate of poverty reduction under the new estimates, regardless of whether the period is taken to begin in 1980 or in 1991, is Latin America. Europe and Central Asia as well as the Middle East and North Africa have much lower rates of estimated poverty reduction according to the new estimates than they did previously.    The estimated proportion of the population that is currently poor in Latin America has barely changed under the new estimates, whereas it has risen by between twenty and thirty percent in sub-Saharan Africa and in South Asia and by multiples elsewhere (by a factor greater than four in Europe and Central Asia, greater than three in the Middle East and North Africa and almost two in East Asia).

The enormous fluctuations in the Bank’s poverty estimates, of a kind which would be unacceptable for most economic statistics, make them unfit for use. The Bank has already undertaken two revisions of base year (and associated international poverty line) and wreaked havoc to its poverty estimates each time. Does it intend to continue on the same path? The next global price survey which will necessitate such a revision is scheduled for 2011. The Bank can at that point choose between pulling the rug from underneath itself again as it has done today, continuing to use PPPs from an earlier base year despite their growing apparent irrelevance, or admitting that it is wrong.

The Bank’s failure to take criticisms seriously and to develop new poverty estimates in a transparent manner mean that the excuse that it is doing the best that can be done is increasingly flimsy.  There exist  feasible alternative methods These involve careful coordination of household surveys and poverty line construction across countries, ensuring comparability from the first and invalidating the need to use ever shifting PPPs for the purpose. Such an effort would be along the lines of the coordination of national accounts — a previous crowning achievement of the United Nations and its member countries. The subject is too important for the world to defer to unscrutinized claims of expertise.

[1] sr793@columbia.edu . I would like to thank Thomas Pogge for his invaluable comments and suggestions

[1] August 26th, 2008

[1] “How Not to Count the Poor”, forthcoming in Stiglitz, J., Anand, S. and Segal, P. ed., Debates on the Measurement of Global Poverty, Oxford: Oxford University Press.