The New York Review of Books
November 21, 2002



By Howard Nye, Sanjay Reddy, Thomas W. Pogge, Reply by Benjamin M. Friedman

In response to "Globalization: Stiglitz's Case" (August 15, 2002)

To the Editors:

In Globalization and Its Discontents, Joseph Stiglitz attacks the economic order and policies the "Washington Consensus" has imposed upon the world. Citing World Bank statistics, he claims that severe poverty has persisted at high levels throughout the 1990s.

Benjamin Friedman's review ["Globalization: Stiglitz's Case," NYR, August 15] discusses this deeply contested and highly important empirical issue. He correctly reports that there is controversy about how many live below the international poverty lines of $1/day and $2/day: Stiglitz gives the World Bank headcounts of 1.2 billion and 2.8 billion, respectively, while Xavier Sala-i-Martin gives 286 million and 980 million instead.

However, Friedman is wrong about the source of this huge discrepancy. It is not true that the World Bank uses market exchange rates to convert foreign currency amounts into US dollars. Rather, the Bank has always been using purchasing power parities (PPPs) just as Sala-i-Martin does.

What then is the source of the discrepancy? The Bank seeks to measure consumption expenditures as assessed through household surveys. Sala-i-Martin relies on data about each country's GDP and income distribution, and thus attributes to each household not only its consumption expenditure, but also a proportionate share of national savings and of all government expenditures. In projecting this much broader measure of a household's standard of living onto the same $1/day and $2/day thresholds, Sala-i-Martin reduces the count by counting something else.

Neither headcount is meaningful, because both use general PPPs to convert foreign currency amounts. In the context of poverty assessment, this is a fatal mistake. General PPPs are related to average price levels for all commodities, weighted by their share in international expenditure. A low-income household must, however, concentrate its spending on a quite narrow subset of all commodities: basic foodstuffs and other necessities. These are cheaper in poor countries, but typically not as much cheaper as general PPPs suggest (see for extensive evidence). A household's income per person per day thus may have as much general purchasing power as $1 has in the US and yet much less purchasing power with respect to basic necessities. Such a household falls below $1/day in the sense that matters.

Whether a household is poor in the very acute sense here at issue should thus be judged by relating its consumption or income not to the local price level of commodities in general, but to the local cost of a basket of basic necessities, generically described to allow for variations in how the poor in different locations meet their nutritional and other basic needs. Specifying a plausible such basket would also help fix the international poverty line at a more meaningful level than the arbitrary $1/day and $2/day standards.

It is likely that both the World Bank and Sala-i-Martin would reach quite different conclusions about the extent and trend of severe poverty if they adopted such a sounder method. So far, no one has made the attempt. Yet the task is urgent. Without it, we simply do not know the extent, trend, and geographic distribution of severe poverty. We therefore cannot accurately monitor and effectively attack such poverty—the first of the International Development Targets ( /declaration/ares552e.htm) endorsed by nearly all governments and international organizations.

Thomas W. Pogge, Howard Nye, Sanjay Reddy
Columbia University
New York City

Benjamin Friedman replies:

I welcome the helpful clarification by Messrs. Pogge, Nye, and Reddy of why the World Bank's estimates of how many people live on less than the equivalent of $1 or $2 per day are so much larger than estimates like those by their Columbia colleague Xavier Sala-i-Martin. I also agree that thinking of poverty in this context as the inability to purchase some basic basket of consumer necessities—presumably including food, shelter, clothing, and so on— is a superior strategy, albeit a difficult one to implement.

On the more specific issue of whether to include in people's "income" for this purpose their proportionate share of what their government spends, the answer depends on what the government is buying. To the extent that the government is paying for the public's medical care, or for subsidized housing, that spending plausibly counts as part of a poor family's purchase of necessities. Spending on corruption or military adventures, which alas often bulks too large in many poor countries, should not. Professor Sala-i-Martin, in a follow-up paper to the one I cited, applied his methodology to data for each country's consumption spending only—that is, omitting all of government spending, as well as national saving. Not surprisingly, the results indicated more people living on less than $1 or $2 per day than in his earlier study. But importantly, these estimates too indicated that the number of people in such dire straights has clearly been falling in recent decades.

Readers may also be interested to know that the United States is one of the few countries that now attempts to define its own "poverty line" for domestic purposes in something like the way Pogge et al. suggest—and moreover, that this practice has fallen under widespread criticism. The US Census Bureau recently reported that the number of Americans living in poverty rose from 31.6 million in 2000 to 32.9 million in 2001, thereby reversing the downward trend that had begun in 1994. These reports hinge on a set of specific poverty lines, updated each year to allow for changing consumer prices. (For a family of two adults and two children, the 2001 poverty line was an annual income of $17,960.) The Census Bureau's original method for deriving these poverty lines was to calculate what it cost the "standard" family to buy the largest single component in its basket of necessities, namely food; then multiply by three (because the research on which the Census Bureau relied showed that poor families spent about one third of their total budget on food); and then figure in adjustments, either up or down, for families of different sizes. This procedure is presumably less accurate than directly calculating the cost of the full basket of consumer necessities, as Pogge et al. suggest, but the underlying idea is the same.

By contrast, most countries around the world either define their poverty line as some set percentage of the country's median income—in the UK, for example, the Blair government has set a benchmark of 60 percent of the median—or, alternatively, do not even attempt to define ''poverty'' but simply report the numbers of people living on incomes below various percentages of the national average. The distinction is important conceptually, and it can be important politically as well. The approach used in most countries abroad treats poverty as a relative concept; if the average person's income moves up, then to stay out of poverty someone at the bottom of the scale needs more income too. The American approach treats poverty as an absolute concept; if the average person's income rises, but the prices of consumer necessities (as represented by food) remain unchanged, then it takes no more income to stay out of poverty than before. In 1995 a National Academy of Sciences panel recommended that the United States replace its longstanding absolute poverty line by a poverty concept that would change over time in relation to the national average, but the government has never accepted this proposal.

Is it inconsistent to favor an absolute notion of poverty for purposes of international comparisons (as Pogge et al. do) while also conceiving of poverty in relative terms for purposes of the domestic discussion within any one country (as most countries other than the US do, and as the National Academy panel recommended for the US as well)? Not necessarily. But the tension between these two ways of thinking about this all-important problem suggests that the matter is hardly settled.