The Bank’s Method ProducesThe Bank’s Method ProducesThe Bank’s Method ProducesThe Bank’s Method ProducesThe Bank’s Method Produces

Meaningless EstimatesMeaningless EstimatesMeaningless EstimatesMeaningless EstimatesMeaningless Estimates

One may think that the problem of inconsistent ratios can be avoided by using PPPs from a single base year alone, so that only one spatial comparison of purchasing power is ever involved. In fact, this is what the Bank does (within any one application of its method). This hides the problem, but cannot solve it. For the fact remains that all the results the Bank reaches – about the levels of the national poverty lines as well as about national, regional, and global poverty rates and poverty trends – are greatly influenced by the Bank’s arbitrary choice of PPP base year, which is completely irrelevant to anyone’s standard of living.

In the following diagrams, each curve represents one country’s CPI (denominated in this country’s currency), that is, nominal amounts in successive years that are deemed equivalent to one another in purchasing power. Curves rising toward the right reflect currency inflation: Ever more currency units must be earned in successive years for an income to maintain its purchasing power over commodities whose prices are rising.

With one curve, representing changes in the purchasing power of one country’s currency only, its vertical position does not matter an identifying the percentage changes in income that are necessary to maintain a fixed level of purchasing power. To represent purchasing power equivalence across different countries, however, the curves corresponding to different countries’ currencies must be placed at the correct vertical distance from one another. The Bank uses PPPs to fix such vertical distances. Comparison of a pair of hypothetical diagrams (Figures 2a and 2b) illustrate how the vertical distance between curves can be affected by the choice of base year (whose PPPs are used to fix this distance). In the illustration, using Y rather than V as the base year has the effect of lowering B’s national poverty line relative to A’s national poverty line for all years and must therefore either lower B’s poverty headcount or raise A’s poverty headcount or both. It follows that estimates of how much poverty there is in country A as compared to country B will vary substantially depending on the base year in which the Bank’s IPL is defined. Moreover, estimates of poverty for the same year and country vary substantially depending on the base year in which the Bank’s IPL is defined. It is easy to see this by thinking of country A in the diagrams as the base country in the currency of which the IPL is defined.

The poverty estimates that the Bank’s method generates when applied with one PPP base year are inconsistent with those it generates when applied with another. This suffices to discredit the method even when it is always applied with the same base year. Even then, it is still true that the estimates generated by the method are improperly influenced by the base year chosen.

As it happens, the Bank has, in the year 2000, revised its IPL from $1/day PPP 1985 to $1.08 PPP 1993. Marketing its new IPL under the same old “$1/day” label, the Bank downplays the significance of the revision and pretends that it has merely “updated” rather than redefined its IPL.3 By examining the Bank’s recalculated national poverty lines and poverty estimates, we can learn how sensitive the Bank’s poverty estimates are to its choice of PPP base year.

Economic and Political Weekly June 3, 2006

Figure 2a: Poverty Line Using “V” as Base YearFigure 2a: Poverty Line Using “V” as Base YearFigure 2a: Poverty Line Using “V” as Base YearFigure 2a: Poverty Line Using “V” as Base YearFigure 2a: Poverty Line Using “V” as Base Year

Figure 2b: Poverty Line Using “Y” as Base YearFigure 2b: Poverty Line Using “Y” as Base YearFigure 2b: Poverty Line Using “Y” as Base YearFigure 2b: Poverty Line Using “Y” as Base YearFigure 2b: Poverty Line Using “Y” as Base Year

Currency Units

2-WVX Y Years

Although the Bank’s two IPLs are strictly incommensurable, one can still ask how its redefinition has affected the figures it reports. Examining all 92 countries for which data are publicly available, we find that the redefinition has lowered national poverty lines in 77 countries, containing 82 per cent of the total population of the 92 countries, and raised national poverty lines in only 15 countries [Reddy and Pogge 2007, Table 5]. While no redefinition of the IPL (in terms of 1993 PPP dollars) could have yielded, for all or even for most countries, national poverty lines that match, even roughly, the national poverty lines yielded by the old ($1/day PPP 1985) IPL, the choice of a different value, higher than $1.08/day, for the new IPL could have achieved a more even balance between countries in which the new IPL is higher and those in which it is lower than the old one [see Reddy and Pogge 2007, Tables 2 and 3]. As the Bank’s researchers admit: “When we compare the most recent common year (1993) we get approximately the same poverty rate as we found in Ravallion and Chen (1997) using $1/day at 1985 PPP; the old poverty rate for 1993 was 29.4 per cent versus 28.2 per cent using the new poverty line for 1993” [Chen and Ravallion 2001:290].7 So the redefinition has reduced the 1993 global poverty count by 4.25 per cent or 58 million. This is rather a substantial achievement compared to the reported actual reduction in the number of very poor people: Over the entire 1987-2001 period, the number of persons living on less than $1.08/day PPP 1993 reportedly declined by only 7 per cent or 82.2 million: from 1,171.2 to 1,089.0 million [Chen and Ravallion 2004:153].

This lowering of the global poverty count through redefinition of the IPL might not matter if everyone understood that the two lines are strictly incommensurable. However, the media and general public cannot be expected to understand this so long as the Bank uses the same “$1/day” label for both lines. Even the Bank’s president is prone to mix lines: His statement8 that the number of poor declined by 200 million in the 1980-2001 period is based on comparing the 1980 global headcount, relative to an old IPL, with the 2001 global headcount relative to $1.08 PPP 1993. In making this statement, Wolfensohn relied on World Bank (2002:8), where the global income poverty count is charted all the way back to 1820!

Since periodic adjustment of the base year (by reference to which an IPL is defined and purchasing power parity conversions are made) is a standard feature of the Bank’s procedure, it is important to note that such adjustments are likely to introduce a systematic bias toward painting too rosy a picture of poverty trends. Adjustment of the base year reduces reported poverty headcounts insofar as international consumption is shifting toward commodities (such as services) that are very much cheaper in poor than in rich countries, and away from commodities (such as food) that are not much cheaper in poor than in rich countries. There is evidence that in recent years the share of food consumption has decreased in a wide variety of rich and poor countries (in keeping with the relationship widely known as Engel’s Law) and the share of services in consumption has increased. As the international consumption pattern shifts in this way, commodities that are very much cheaper in poor than in rich countries are given more and more weight in the calculation of generalconsumption PPPs. Using such PPPs to assess the incomes or consumption expenditures of very poor people thus becomes increasingly distorting. The distortion arises from the fact that, no matter how much the share of basic necessities in international consumption may diminish, a poor household must still focus virtually all its expenditure on such necessities. The apparent rise in the reported purchasing power of poor households due to an international consumption shift toward services can mask the fact that such households have not gained greater access to the goods they most require.

estimate of the trend (just as it influences estimates of the extent of global poverty in a given year). Given the biases just described, an IPL defined in terms of a later PPP base year will tend to be associated with lower PPPs (poor-country currency units per US dollar) and hence with lower poor-country poverty lines and lower poor-country poverty headcounts for all years.9 The impact of this effect on global poverty trend estimates depends on how this trend varies with the level of national poverty lines. In recent years, it appears that the trend in the global poverty count has been more favourable at lower levels of the IPL as discussed later. If this is any indication, then the effect of “updating” the IPL may well have been to create a more favourable picture of the global poverty trend than would have resulted from maintaining the old line, defined in terms of an earlier base year. Evidence for this conjecture is provided by the fact that the trend reported by the Bank in the last report in which it used the old ($1/day PPP 1985) IPL was unfavourable, in sharp contrast to the trend reported for a closely similar period using its new IPL. Over the 1987-99 period, the number of persons living on less than $1.08/day PPP 1993 reportedly declined by 75.5 million: from 1,171.2 to 1,095.7 million [Chen and Ravallion 2004:153]. In contrast, estimates produced by the Bank using the old ($1/day PPP 1985) IPL led to the conclusion that, “the absolute number of those living on $1 per day or less continues to increase. The worldwide total rose from

1.2 billion in 1987 to 1.5 billion today” [World Bank 1999:25].10

Price ratios between rich and poor countries vary widely across commodities. For goods easily traded across borders, prices compared at market exchange rates are about the same in rich and poor countries. For goods and services not easily traded across borders, prices compared at market exchange rates can be 50 times higher in rich countries than in poor ones. “Broadgauge” general-consumption PPPs of the kind used by the Bank average out these price ratios in a way that, roughly speaking, weights each commodity in proportion to its share in international consumption expenditure.

The use of such PPPs is quite inappropriate for poverty assessment and severely distorts the resulting poverty estimates. To illustrate, consider once more our simple two-country world. Suppose, not unrealistically, that $1 buys about three times as much food in the US as Rs 10 buy in India and also that Rs 10 buy about three times as much services in India as $1 buys in the US. If food and services have equal shares in international consumption expenditure, conventional methods of calculating PPPs will then equate the purchasing power of Rs 10 with that of $1. This may be fair enough guidance for assessing average incomes in the two countries. But not for assessing very low incomes: Poor Indians’ disadvantage of lesser command over basic foodstuffs is not compensated by their advantage of greater command over services. Even if richer persons spend much of their income on maids, drivers, haircuts, etc, the poor do not, and cannot, do so. They must concentrate what little income they have on basic necessities. The calculated PPP (Rs 10 per $1) thus grossly overstates the true value of their incomes.

Very low incomes should therefore be assessed by relating them not to the prices of all goods and services, but only to the prices of those commodities they must consume to meet their basic needs. This would ensure that the IPL is meaningfully related to the circumstances under which people actually live. Under the Bank’s procedure, by contrast, people living at the IPL may differ greatly in their command over basic necessities.

We do not currently possess all the data needed to estimate poverty worldwide in this more sensible way, although it should be possible in the future to collect it. The data we do have suggest how global poverty estimates would be different if a more credible procedure were used. Existing data about the prices of foodstuffs and, more specifically, of staple bread and cereals, show that these items (a large part of the consumption requirements of the poor) cost far more in poor countries than general-consumption PPPs suggest. The same is true for many basic necessities other than food.11 It is very likely that the Bank, were it to use PPPs more closely related to the needs of the poor, would translate its $1/ day standard into substantially higher national poverty lines for most poor countries.

How much higher would these national poverty lines be? If prices of foods, or more specifically of breads and cereals, rather than prices of all commodities had been used to convert the Bank’s existing IPLs, then poor-country poverty lines would likely be some 30-40 per cent higher on average,12 which would raise the estimated global incidence of severe income poverty substantially. For a small number of countries for which we were able to make estimates, we found that increases in national poverty lines by 30-40 per cent entail increases in poverty headcounts of a similar magnitude [Reddy and Pogge 2007, Table 10].

The question of whether the Bank’s IPLs have an interpretation in terms of the resources needed by human beings to achieve elementary capabilities, to meet their basic needs, may also be approached in another way. If the PPP conversion factors employed by the Bank correctly captured the currency amounts necessary in different countries to possess equivalent purchasing power, then the IPL would have this interpretation in poor countries only if it had such an interpretation in the US. Does it?

The Bank Sets Its IPLs at an Arbitrarily Low LevelThe Bank Sets Its IPLs at an Arbitrarily Low LevelThe Bank Sets Its IPLs at an Arbitrarily Low LevelThe Bank Sets Its IPLs at an Arbitrarily Low LevelThe Bank Sets Its IPLs at an Arbitrarily Low Level

One way to investigate this question is to ask what is the cost of achieving a particular elementary requirement of human life (for example, the ability to be adequately nourished) in the US. The Thrifty Food Plan, produced by the US department of agriculture as a guide for low-income households and government agencies, offers one answer to this question. Adopting a thorough and careful analytical methodology, the plan estimates that the least cost of meeting a minimal calorie constraint (varying between 1,600 and 2,800 calories depending on age and gender) and a set of other minimal nutrient constraints (while minimising the deviation from the existing pattern of consumption of low-income Americans) is $ 98.40 (1999) per week for a reference family consisting of a male and a female age 20 to 50, and two children age 6 to 8 and 9 to 11[USDA 1999, ES-1].13 According to the Bank’s IPLs, by contrast, this reference family needed in 1999 only $43.35 ($1/day PPP 1985) or $34.72 ($1.08/day PPP 1993) to avoid poverty, that is, to meet all its basic requirements – not only for food, but also for clothing, shelter, and minimal healthcare. Even if we allow that the USDA has taken a generous view of what it takes to be adequately nourished, it is clear that the Bank’s IPLs are too low to reflect even the basic nutritional requirements of human beings – let alone all their basic requirements.

It is obvious that, were the Bank to set its IPL at an appropriately higher level, it would arrive at a much higher global poverty count. It is less obvious, but true, that the Bank would then also be likely to calculate much worse poverty trends.14 Using its $1.08/day PPP 1993 IPL, the Bank calculates that the number of poor worldwide has fallen from 1,171.2 million in 1987 to 1,089.0 million in 2001. Relative to a more appropriate doubled IPL of $ 2.15/day PPP 1993, however, the number of poor worldwide has, by the Bank’s own estimate, increased from 2,477.5 million in 1987 to 2,735.4 million in 2001 [Chen and Ravallion 2004: 153]. Note that a 1999 US reference family living at this doubled IPL would still have only $ 69.44 per week – well below the $ 98.40 the USDA says such a family needs for food alone.

The general consumption PPPs currently in use are based on limited and highly questionable evidence. A large number of countries containing poor people – including China and India, which contain the largest number of the world’s poor – have not participated at all in recent “benchmark” price surveys of the International Comparison Programme. As a result, a massive element of guesswork and gap-filling underlies current poverty estimates.15 The current estimates of PPPs for China that are considered to be plausible differ by a factor of two.16 Shifts in China’s poverty line resulting from the adoption of alternative PPPs would entail massive variations in the global poverty count. India has not participated in a benchmark survey since 1985. As observers of India are well aware, there are likely to have been significant changes in its internal price structure since then. Consequently, current estimates of India’s poverty headcount are subject to substantial uncertainties.

While the Bank reports global poverty counts with five-digit, even six-digit precision, suggesting possible errors of a few thousand at most, the data used in producing these estimates (PPPs in particular) are subject to vast uncertainties. This fact sharply undermines any claim to precision, and should be more fully acknowledged.

In addition to these reasons why the Bank’s estimates of the extent of poverty are uncertain, there are additional causes to believe that poverty trends in particular are uncertain, and that they may currently be represented in an overly favourable manner: When the Bank lacks up to date data about the distribution of consumption expenditure in a country, it assumes that the distribution is unchanged. In particular, the Bank assumes that the consumption of all individuals has grown in proportion with the rate of growth of consumption in the national income and product accounts. There are several reasons to doubt the validity of this procedure. First, it is well known that in recent years income inequality has risen in many countries, again including China and India.17 It may thus be overly optimistic to assume that the consumption of the poor has grown in tandem with mean consumption in their country. Second, as argued powerfully by Deaton (2003), the measure of consumption in the national income and product accounts is a broad one and is likely to reflect growth in forms of consumption that have little significance for the poor.

Data about income poverty are of great importance for the design and evaluation of policies, agencies and social institutions. To be sure, there are other important sources of information about people’s standard of living: data about their health status and educational attainments and about mortality and morbidity, for example. Such data do and should inform overall judgments concerning the extent, distribution, and trend of poverty in the world. Income poverty data are nevertheless an essential part of the picture.

poverty assessment within countries if more meaningful global poverty estimates are to be created. It is notable that such efforts have been lacking, although they are feasible. An example of what is possible is given by the diffusion throughout the world of a common system of national income and product accounting through the involvement of the United Nations18– an achievement once thought highly improbable. Although the development of a credible system of global poverty monitoring will require time and resources, these costs are small compared to the potential costs of continued ignorance. Without an effort to develop such a system, it will be difficult to accept that the agencies charged with monitoring global income poverty are serious about the task.

[:l

Email: tp6@columbia.edu

NotesNotesNotesNotesNotes

for which data were available. If all-food PPPs are used in lieu of generalconsumption PPPs to convert $1 PPP 1985 into national currencies,national poverty lines of poor countries increase by 40 per cent on average.Using bread-and-cereals PPPs to convert $1 PPP 1985, the averageincrease is 34 per cent. Using all-food PPPs to convert $1.08 PPP 1993,the average increase is 31 per cent. Using bread-and-cereals PPPs toconvert $1.08 PPP 1993, the average increase is 40 per cent.

ReferencesReferencesReferencesReferencesReferences