Finance & Development, June 2002 - Is = World Poverty Falling?

About= =20 F&D
3D""=20

S= ubscribe

Back=20 Issues
3D""=20

Write Us

C= opyright=20 Information 3D""=20

Free Email Notification

Receive emails when we post new items of interest to = you.

Subscribe or = Modify= =20 your profile

A quarterly magazine of the IMF

June= =20 2002, Volume 39, Number = 2

Is World = Poverty=20 Falling?
Angus=20 Deaton

Real progress has been made in reducing poverty in recent = years,=20 particularly in India and China. However, there is still much = uncertainty=20 about the numbers. Using the same data, two reports released less = than two=20 years apart by the World Bank reached apparently different = conclusions on=20 whether world poverty was going up or down. How can we know = whether the=20 world poverty counts are accurate?

The first table of the World Bank's World Development Report = 2000/2001: Attacking Poverty shows that the number of people = living on=20 less than $1 a day grew from 1.18 billion in 1987 to 1.20 billion = in=20 1998=97an increase of 20 million. Less than two years later, a = headline=20 chart in another major World Bank publication, Globalization, = Growth,=20 and Poverty: Building an Inclusive World Economy, showed that = the=20 number of people living in poverty fell by 200 million from 1980 = to 1998=20 and showed no trace of an increase between 1987 and 1998. The = poverty=20 decrease was reaffirmed in the press release accompanying "The = Role and=20 Effectiveness of Development Assistance," a World Bank research = paper=20 issued before the March 2002 UN Financing for Development = Conference in=20 Monterrey, Mexico: "Over the past 20 years, the number of people = living on=20 less than $1 a day has fallen by 200 million, even as the world's=20 population grew by 1.6 billion."=20

Can these statements be reconciled? Has there been a marked = reduction=20 in poverty in the last two years? Or has the Bank revised its=20 interpretation of history?=20

Getting an accurate poverty count is important. The Bank sails = under=20 the banner "Our dream is a world free of poverty," which not only = invites=20 the use of the poverty count as a measure of the extent to which = the dream=20 is being fulfilled but also raises the issue of whether the = organization's=20 success can be convincingly measured by its own numbers. We also = need an=20 accurate poverty count to assess whether the international = community is=20 achieving one of the Millennium Development Goals endorsed by 189=20 countries at the September 2000 UN Millennium Summit=97to "halve, = between=20 1990 and 2015, the proportion of people whose income is less than = one=20 dollar a day." A lot depends on whether the scorecard is being = credibly=20 tallied, and the apparent discrepancies in the Bank's numbers = deserve=20 serious scrutiny. That scrutiny produces some genuinely good news = about=20 poverty declining but also raises some serious concerns about the = numbers.=20 With respect to the production of data, still greater transparency = on the=20 Bank's part is called for.=20

Poverty in India=20

Take the case of India. Its poverty counts are important not = only=20 because they have a large direct effect on world poverty = counts=97more than=20 one-fourth of the world's poor live in India=97but also because = the world=20 debate on globalization, poverty, and inequality has been echoed = in an=20 intense domestic debate.=20

India's economic liberalization in the early 1990s was followed = by=20 historically high rates of growth. But did this growth help or = hurt the=20 poor? Were their numbers reduced or did economic growth benefit = only an=20 increasingly wealthy urban elite?=20

The political debate has been fueled by questions about the = accuracy of=20 poverty measurements in light of the discrepancies between = estimates of=20 consumption growth based on national accounts statistics (NAS) and = those=20 based on household surveys carried out by the National Sample = Survey=20 (NSS). According to the NAS, real per capita consumption has been = growing=20 at about 3.2 percent a year since the reforms, while, at least = until=20 recently, the NSS data have shown little or no growth throughout = the=20 1990s. Reform opponents resolutely quote the NSS data, while = reform=20 advocates back the NAS growth estimates, questioning the accuracy = and the=20 integrity of the NSS data and arguing that, because the poor's = share of=20 the national pie is more or less fixed, growth must reduce = poverty.=20

The controversy only deepened last year with the release of the = 1999-2000 consumption survey, the first major survey since = 1993-94, when=20 reforms had barely begun to take effect. In the intervening years, = there=20 had been a series of smaller ("thin") household surveys showing = little or=20 no growth in per capita consumption and, if anything, a rise in = poverty.=20 Accurate or not, they provided the only numbers in town and were = widely=20 used outside India=97for example, in the World Bank's Attacking = Poverty,=20 albeit with due acknowledgment of uncertainty. Hence the belief = that=20 poverty in India had been increasing.=20

Unfortunately, in a decision whose timing could hardly have = been worse,=20 the NSS made major changes to the questionnaire design for the = 1999-2000=20 survey. Although the new survey design is sensible in itself, it = is not=20 comparable with earlier designs and almost certainly leads to more = consumption being reported, especially among the poor. As a = result,=20 measured poverty was lower than it would have been with the = previous=20 design. So when the Planning Commission issued its poverty = estimates in=20 February 2001, showing a dramatic fall in poverty=97from 36 = percent of the=20 population to 26 percent=97many cried foul. Pleas to have the = survey redone=20 to make it consistent with earlier surveys were ignored, and the = ruling=20 Bharatiya Janata Party was in no hurry to challenge estimates that = showed=20 rapid progress in dealing with India's most intractable and = longstanding=20 social and economic problem.=20

The chart shows the official estimates (the planning-commission = head-count ratios) going back to 1973-74; each circle comes from a = large=20 consumption survey. The points labeled "thin" rounds show the=20 (unofficially calculated) poverty estimates from recent smaller = surveys.=20 If the final point is blanked out, we can imagine the situation=20 immediately prior to February 2001. From that perspective, the = 1999-2000=20 estimate is nothing short of astonishing: the Indian poverty rate = fell by=20 12.8 percent over 18 months, removing 60 million persons from = poverty.=20 Progress indeed!

Over the past year or so, I and others have been working with = the data=20 to try to sort out what happened. Although the questionnaire = design was=20 changed for most of the goods in the survey, data on an important = group of=20 expenditures were collected in the same way in all of the surveys. = Fortunately, nearly all households purchase these goods, and the = total=20 amount they spend on them is a good predictor of whether a = household is=20 poor or not. The latest data show a marked increase in real = expenditures=20 on these goods, which indicates a substantial reduction in poverty = overall. The extent of the increase also allows us to estimate how = much=20 poverty has fallen. Much to my surprise, most of the officially = claimed=20 reduction in poverty appears to be real. I estimate that poverty = fell from=20 36 percent in 1993-94, not to 26 percent as in the official = numbers, but=20 to 28 percent. These calculations, although necessarily = speculative, are=20 supported by analysis carried out by K. Sundaram and Suresh = Tendulkar at=20 the Delhi School of Economics, who obtain very similar results = using=20 entirely different methods.=20

Another complication is the questionable quality of the Indian = price=20 indexes used to update the poverty lines. With some correction to = the=20 price indexes, as well as an allowance for the noncomparable = survey=20 design, the head-count ratio shows a fairly steady decline from = 1987-88=20 through 1999-2000. Indeed, since the 1970s India has made more or = less=20 steady progress in reducing poverty. (The "fully adjusted" = estimates in=20 the figure also correct for overestimation of urban poverty in the = official counts, and thus start from a lower base in 1987-88.) = Even so,=20 the estimates based on the "thin rounds" raise unanswered = questions.=20 Although the last of these, which is the most egregious, is = relatively=20 easy to challenge=97if only because the survey ran for only six = months=97there=20 is at present no obvious reason to dismiss the three earlier = observations.=20

Can we conclude that the reforms helped reduce poverty? Neither = consumption growth nor poverty reduction shows much sign of having = been=20 more rapid after the reforms. But neither is there any sign of = general=20 impoverishment as a result of the reforms=97indeed, quite the = reverse. What=20 about inequality? Again, the change in survey design precludes any = simple,=20 uncontroversial answer. But the adjustment procedures applied to = the=20 poverty counts can also be used to estimate inequality, which has = been=20 increasing in recent years, particularly between states=97the = states in the=20 south and west that were originally better off have grown most = rapidly=97as=20 well as within urban areas, which have been the greatest = beneficiaries of=20 growth.=20

Because of this growing inequality, consumption by the poor did = not=20 rise as fast as average consumption, and poverty reduction was = only about=20 two-thirds of what it would have been had the distribution of = consumption=20 remained unchanged.=20

World poverty at the Bank=20

India's story was not available to the authors of Attacking=20 Poverty, whose estimates of the decline in world poverty would = have=20 been less pessimistic if, instead of a 50 million increase in the = number=20 of poor from 1993-94 to 1998, they had used the current estimate = of a 50=20 million decline up to 1999-2000. Because China has also made = continuing=20 progress in reducing poverty, it is possible to paint a rosier = picture of=20 the decline of world poverty today than it was one year ago. = Indeed,=20 according to recent calculations by Shaohua Chen and Martin = Ravallion=20 (2000) of the World Bank using all of the household survey data = since=20 around 1980, and with due recognition of the data's many = inadequacies, the=20 best current estimate is that there are indeed around 200 million = fewer=20 people living in poverty now than 20 years ago. So it seems that = both=20 Attacking Poverty and Globalization were right: = poverty rose=20 a little from 1987 to 1998 but fell substantially from 1980 to = 1998.=20

Yet there remains much in this story to make one uneasy. Most = obvious=20 is the discomfort of having to rely on data whose margin of = uncertainty is=20 enough to allow, over a period of two years, such discordant views = about=20 what is happening to poverty in the world.=20

Attacking Poverty, whose preparation was much influenced = by=20 nongovernmental organizations and groups opposed to globalization, = frequently takes a skeptical view of the role of economic growth = in=20 promoting poverty reduction. Globalization takes a much = more=20 sanguine view. A chart showing the number of people living on less = than $1=20 a day from 1820 to 1998 shows a decline from about 1980 that = accelerates=20 in the latest years. The historical data in this chart were = assembled by=20 Fran=E7ois Bourguignon and Christian Morrisson in a paper that is = to appear=20 in the American Economic Review. They derive their = estimates by=20 applying (sometimes sketchy and outdated) distributional = information to=20 the consumption figures from national accounts data, a technique = that is=20 almost certainly the only methodology that would allow the = construction of=20 data for a century and a half. But as we have seen from India's=20 statistics, such methods can sometimes beg the question of whether = growth=20 reduces poverty. After 1993, when the Bourguignon and Morrisson = data end,=20 Globalization uses the poverty estimates that were = assembled by=20 Shaohua Chen and Martin Ravallion prior to Attacking = Poverty. But=20 Chen and Ravallion's data from 1987 to 1993, which is when poverty = increased, are dropped from the chart. In consequence, and without = any new=20 information, we go from an assessment that the number of poor = people in=20 the world was showing little or no decline from 1987 to 1998 in=20 Attacking Poverty to an assessment, in = Globalization, of a=20 continuous and accelerating decline from 1980 to 1998.=20

One of the Bank's greatest achievements over the past 20 years = has been=20 its extensive investment in primary data, particularly the = household=20 surveys on which the poverty counts are based. Before the late = 1980s, the=20 Bank calculated and forecast poverty from data on aggregate = consumption,=20 together with whatever distributional information might be = available,=20 sometimes from an earlier period or even from a different, but = "similar,"=20 country. Since they were first calculated in 1987, the $1 a day = poverty=20 counts have been based on direct observations of what poor people = earn or=20 consume, using the several hundred income and consumption surveys = that are=20 now available to the Bank. This statistical advance is what makes = it=20 possible for us to have a serious debate about what is happening = to=20 poverty in the world. And it is clear that the Bank's best = economists do=20 the calculations to the highest standards, and there has never = been any=20 suggestion that the counts are anything other than what the data = show. But=20 as the case of Attacking versus Globalization shows, = it is=20 possible for these underlying data to be used in ways that seem to = support=20 very different factual statements about trends in world poverty.=20

Although external access has improved over time, most of the = data used=20 in the count are not available to independent researchers or to = agencies=20 outside the Bank. Nor, given the arrangements that the Bank makes = with=20 member countries, is it possible for the Bank to make all of the = data=20 available. So the Bank is effectively the only producer of world = poverty=20 counts. National governments around the world carefully separate = and=20 insulate the production of data from the policymakers who rely on = them and=20 are often judged by them. A similar arrangement would surely be = wise for=20 the world poverty counts, if only to protect those who create the=20 statistics. Perhaps there should be an independent outside agency = or=20 auditor who either produces the numbers under a carefully = specified=20 contract or verifies their construction. Second best would be for = the Bank=20 to continue to calculate the numbers but to visibly insulate them = from the=20 internal policy debate and to institute a process of collection = and=20 production that is fully transparent to outside observers. The = Bank's very=20 success in producing these numbers, and in making them a central = part of=20 the policy debate, argues for an upgrading in the way they are = produced=20 and maintained. Otherwise, if its reports continue to contradict = each=20 other, the Bank risks losing the authority to monitor its own = success, and=20 no one will know whether or not the Millennium Development Goals = are being=20 achieved.

References:

Fran=E7ois Bourguignon and Christian = Morrisson,=20 2002, "Inequality among World Citizens: 1820-1992," DELTA Working = Paper=20 (Paris), American Economic Review, forthcoming; =

Shaohua=20 Chen and Martin Ravallion, 2000, "How did the world's poorest fare = in the=20 1990s?" World Bank Research Paper (Washington); http://econ.worldbank.or= g/docs/1164.pdf=20

Angus Deaton 2001, "Computing prices and poverty rates in = India,=20 1999-2000," Research Program in Development Studies Working Paper, = Princeton University; http://lnweb18.worldbank.org/sar/sa.nsf/At= tachments/adp/$File/prices+poverty+55thRound+all.pdf=20

World Bank, 2001a, Globalization, Growth, and = Poverty:=20 Building an Inclusive World Economy (New York: Oxford University = Press for=20 the World Bank).

=97=97=97, 2001b, World Development = Report 2000/2001:=20 Attacking Poverty (New York: Oxford University Press for the World = Bank).=20

=97=97=97, 2002, "The Role and Effectiveness of Development = Assistance,"=20 World Bank Research Paper (Washington); http://econ.worldbank.org/ =

Angus = Deaton=20 is Dwight D. Eisenhower Professor of International Affairs = and=20 Professor of Economics at Princeton University=20