Presenting Data: Tabular and graphic display of social indicators
 
Gary Klass
Illinois State University
© 2002

Note: The website will be discontinued shortly, to be replaced by the Just Plain Data Analysis site

Home Good Tables Good Charts Divide Analyzing Budgets Election Data Education Poverty References Course page Chart of the week your comments

Measuring Poverty and Inequality

Do free markets and globalization foster greater or lesser poverty and inequality in the developing world?  Does international development assistance to third world nations alleviate or exacerbate the conditions of poverty?  Do welfare state social policies significantly reduce poverty in developed nations? What impact did the US 1995 welfare reform act have on the nation's poor? Are the relatively high rates of poverty in the United States due to the dysfunctional behavior of the poor, to racism and discrimination, or to stingy social policies?  How do these factors account for the (alleged) feminization of poverty in the United States? Is it true what they say, "The rich are getting richer and the poor, poorer"?

The arguments, policy analyses, political debates and academic research concerning these questions rely on measurements of the incidence of poverty.  Poverty rates generally measure the percent of the population living in households whose annual income (or, as we shall see, annual expenditure) falls below a predetermined poverty threshold.  There are two different approaches to determining the standard of living that constitutes poverty: absolute poverty thresholds define a level below which households lack basic necessary goods and services, while relative thresholds measure the percentage of the population living at a standard well below the average of their fellow citizens.  Neither strategy for defining poverty is ideal.   Poverty is inherently a relative concept and absolute standards of measuring it are often arbitrary, but the relative thresholds measure inequality more than they measure a consistent level of deprivation.  For these and other reasons, debates about poverty often end up being debates about the measurement of poverty.    Liberal scholars often argue that poverty rate statistics underestimate the true dimensions of poverty and that higher poverty thresholds should be used (see Chossudovsky, 1999 and Mangum, 2003), while conservatives often argue that the statistics exaggerate poverty, either because the thresholds are set too low or because the measure of income and consumption do not include all the resources available to the families who are classified as poor (Rector and Hederman, 2004).

Measuring Poverty in Developing Nations.

Since the early 1980s, the standard measures of poverty in low and middle income developing nations have been indicators designed by the World Bank.  The Bank’s poverty rate measure is defined as the percentage of a country's population living in households consuming less than $1 or $2 per day per capita.  The $1 consumption threshold is referred to as "extreme poverty" and a $2 threshold is used as the more general poverty indicator.[1]  Both threshholds are expressed in constant dollars and are adjusted for price differences between countries, using measures of national currencies’ purchasing power parity (PPP).  In a country where the average family size of 4 persons, this would be a family poverty threshold of $1424 per year.  In some countries the poverty measures are based on the number of persons with incomes below a poverty threshold, but in the poorest nations (where subsistence agriculture and non-cash economies predominate) the indicators are typically based on estimates of consumption -- the value of the goods and services consumed by persons, derived from survey responses to questions about food consumption.

In 2001, the Bank estimated that 21% of the populations of the world’s developing nations lived in extreme poverty, a total of 1.1 billion people, a reduction from 1.2 billion in 1990 and 1.5 billion in 1981.   

Regional trends in the international poverty rate (Figure 1) indicate that poverty has declined dramatically in most of Asia (particularly in China and India) over the past two decades, has increased in Africa and (since 1989) in the former Soviet Union countries, and has held relatively constant throughout the Middle East and Latin America.   In China alone, 400 million fewer people lived in poverty in 2001 than did so in 1980, with most of the decline occurring in the early 1980s.  Although the poverty rate for sub-Saharan Africa has increased only marginally, the number of poor persons in that region nearly doubled, from 164 to 316 million (Chen and Ravallion 2000). 

Figure 1: Regional trends in extreme poverty

 

The World Bank’s annual “World Development Report” is a widely respected source of statistical information on trends in poverty and related social conditions.  There, and on its website, the Bank reports detailed information national accounts (e.g., GDP, income and trade data), business conditions, governmental policies and developmental assistance in addition to the poverty data.  The United Nations provides similar data and a similar report: the annual and regional “Human Development Report”.  Although much of the data is the same, the tone of the reports do reflect the ideologies of the countries in control of the organizations.  The World Bank tends to see poverty and as something to be addressed by economic development; the United Nations stresses that addressing poverty and inequality will have the side benefit of promoting economic development.

In September, 2000, the United Nations adopted the “Millennium Declaration,” setting eight broad development goals for developing nations related to various aspects of poverty, education, gender inequality, health, and environment (see table 1).  Forty-eight development indicators, including the $1 a day poverty rate, are used to measure progress toward the goals.  For poverty, the goal is to reduce the $1 poverty rate by half, from nearly 28% in 1990 to 14% in 2015. 

Table 1:

 

It is generally the position of the World Bank and the International Monetary Fund that the solutions to poverty in the developing world are to be found in policies involving free market reforms and addressing the political corruption typical of many of the world’s poorest nations.  African countries generally rank among the world’s poorest nations and their governments typically rank as the most corrupt, as measured by Transparency International’s surveys of international business leaders.

Figure 2: Corruption and poverty
sources:
Transparency International; World Bank

 

Poorer nations tend to be more corrupt than wealthy nations, but whether political corruption is the cause or the effect of poverty, however, is a matter of much dispute.  Bob Geldof, activist and organizer of the Live Aid and Live 8 concerts argues that "Africa is not mired in corruption, it is mired in poverty. Corruption is a by-product of poverty, just like dying of famine or Aids." (BBC News, 2005).  The free market policies advocated by the World Bank and the International Monetary Fund (IMF) are also subject of much controversy.  Critics of the World Bank and the IMF’s argue that the organizations’ advocacy of market reforms –such as privatization , reduced government subsidies, free trade and reduced business regulation, result in reductions in necessary social services and increased poverty and inequality. 

As we see in figure 3, low poverty rates and economic development, as measured by per capita GDP, are related conditions, but not necessarily the same thing.  The dramatic reductions in East and South Asian poverty (shown in figure 1)  since  the 1980s can be credited to free market reforms in China and India.  On the other hand, the rising levels of extreme poverty in the former Soviet bloc (Eastern Europe and Central Asia) also followed the substantial market reforms following the fall of Communism.

Figure 3: National Wealth and Poverty

In Latin America where poverty rates have remained steady for the past two decades only two countries have managed to achieve substantial reductions in poverty: Chile and Guyana (see figure 3b).

Figure 3b: excel file

Both of these countries have pursued what have been the most aggressive free market economic policies of any
countries in the world, in the case of the Guyana under the impetus of IMF\World Bank structural adjustment policies.

Issues of data reliability.

While both the United Nations and World Bank strive to provide accurate and reliable measures of poverty and other social and economic indicators, much of the data is of uncertain quality.  Many of the poorest nations – such as Zaire or Zimbabwe – lack the bureaucratic infrastructure to conduct accurate demographic counts and those poor nations with large governmental bureaucracies – such as China or Cuba – often produce data designed merely to make regional or local bureaucracies look good. 

Constructing the poverty measure involves two steps: determining the distribution of income (or consumption) across a country’s households (determine how many households live at each income level), and defining what the $1 and $2 a day standard means in local currency.  In the case of the World Bank’s poverty measure, for example, the base data are derived from national surveys of household consumer expenditure or income conducted at widely varying intervals in each country (approximately every six years in India, annually or biennially in China, rarely or never in many of the poorest nations).   Typically, the household consumption surveys ask a member of the household to report the household’s food consumption and other purchases over the previous week or month.  The data are then extrapolated into yearly figures.  Similar surveys of household income are generally used in more developed countries (income surveys are less appropriate in countries with a large subsistence or non-cash economy).   The base data provide estimates of the percentage of households living at each level of consumption or income and are used to estimate the poverty rate using national standards and to generate measures of economic growth and income distribution (such as the GINI index or Lorenz curve).

The quality of, and methods of collecting, these base data can vary wildly from country to country and cross-country and time-series comparisons of the data are often highly suspect.  In many countries, the local population can be highly suspicious of governmental officials who come asking questions, while in others it may be suspected that forms are completed by enumerators who never leave their offices.  When household survey data is not available, but general national income measures are, the Bank uses estimates based on the assumption that the income distribution is the same as in neighboring countries.  To estimate the poverty rates for years when there were no household surveys, the Bank assumes all incomes levels have increased or decreased at the national growth rate (Reddy and Pogge, 2005, 26).  Thus, the estimates report reductions in poverty whenever a country experiences national economic growth.  This makes the data highly suspect for scholars who argue that globalization and international investment produce uneven economic growth.

Next, one has to determine what level of national consumer expenditure corresponds to the World    Bank’s $1 and $2 a day standards.  This requires an estimate of each country’s “purchasing power parity”, or PPP.  To attain this measure requires both additional household surveys to determine what commodities people in the country typically consume and business surveys of the price of those commodities.

Measuring Poverty in Wealthy Nations.

When the World Bank estimates the total world population living below the $1 a day level, it not unreasonably assumes that no one the developed world would fall below the standard.   Beginning in 1983, the Luxembourg Income Study (LIS) began compiling a collection a household income surveys conducted (usually) annually by many of the world's wealthiest countries.  The LIS database now includes income surveys from 29 countries. Under the direction of Timothy Smeeding and Lee Rainwater, the LIS center adjusts the national survey data to provide for consistent measures across nations.  From these data, LIS reports a variety of statistical measures of national income distribution.

Unlike the World Bank poverty measures, the poverty indicator most commonly reported by the LIS is a relative measure: the percent of persons living in families below 50% (or some other percentage) of the national median family income, adjusted for family size.  Unlike the US measure discussed below, the LIS measure is based disposable income (income after taxes) and includes some "near-cash" income, such as (in the US case) food stamps.

 

Figure 4: LIS relative poverty measures

 

Smeeding and Rainwater have used these data in a series of reports and a book (2003) addressing international differences in child poverty, calling attention to the high rates of child poverty (consistently the highest rates of child poverty) in the United States (see figure 4).

Measuring Poverty in the United States.

In 1963, Molly Orshansky, an economist at the Social Security Administration undertook the task of determining just how many people in the United States were poor (Fisher, 1997).  Although conservative and liberal scholars have debated the merits of the indicator ever since, the US poverty rate statistic that she developed remains today as one of the most commonly used measures of the nation's economic health.

To measure the poverty rate, it was first necessary to determine a standard for classifying people as poor.  Orshansky relied on studies of food budgets that had been conducted by the Department of Agriculture in the mid 1950s.  The USDA had developed an "Economy Food Plan", essentially a food budget designed to meet the basic nutritional needs of families.  A 1955 study of family budgets had determined that Americans spent an average of one-third of their budget on food.  For 1963, the Economy Food Plan cost a family of four (two adults and two children) an average of $1,033 per year.  Multiplying that by 3, Orshansky set $3,100 as the poverty threshold for a family of four.  Since 1963, the poverty threshold, adjusted for family size, has been adjusted for changes in the Consumer Price Index (note: not for changes in cost of the food budget). 

The U.S. Bureau of Labor and the Census Bureau report the poverty statistics annually based on the Current Population Survey (CPS).  The survey is conducted in March of each year, measuring the family incomes of the previous year for a national sample of approximately 77,000 families.

For 2004, the resulting poverty threshold for a family of four stood at $19,157.  Persons living in four person families where the total family income is less than $19,157 are thus classified as poor and those with higher incomes are classified as not poor.  In 2003, 35 million Americans were classified as poor, 12.5% of the total population (25 million families were poor, 10.8% of the families).

Figure 5: US Poverty trends


The US poverty rate fell sharply in the 1960s, reaching its lowest point in 1973; since then it has fluctuated in a narrow range (see figure 5).  Poverty trends generally reflect changes in the economy: The declines in the 1990s reflect the economic prosperity of the Clinton administration years; the increases before and after those years the economic downturns of the two Bush administrations.

Because the CPS survey includes many demographic questions about age, family structure and marital status, race and ethnicity, the data allow for very detailed analysis of the demographics of poverty. 

Figure 6: The Demographics of Poverty


In figure 6, for example, we see that much of the disparity in black and white family poverty (but not Hispanic and white) can be accounted for by the differences family structure.    While conservatives often point to the increase in single parent families as the reason for the persistence of poverty in America, liberals see government policies as a primary cause.

Figure 7: Poverty trends


Thus Senator Daniel Moynihan (1987) often called attention the disparity in trends in poverty rates for children and the elderly.  In the 1960s, poverty rates fell for both groups, but after the early 1970s, the child poverty rate began to increase while the elderly poverty rate continued to decline -- to the point where the elderly are now less poor than the population as a whole (see figure 7).  These trends, Moynihan argued could be explained by the substantial increases in Social Security and Supplemental Security Income benefits for the elderly after 1970 and the steady decline in AFDC and other program benefits that targeted families with children.

Problems with the Poverty Statistic.

There have been many complaints that the poverty statistics either overestimate or underestimate the true poverty rate: liberals say that poverty is higher than the numbers indicate; conservatives say that the numbers exaggerate the poverty rate, although these claims are made more or less strenuously depending on which party is in power.

Liberals often complain that the poverty thresholds are too low.  Although food was one-third of a family budget in the 1950s, it is one-sixth now; taking that into account, they say, would suggest a higher threshold.  Liberals often argue that the poverty thresholds are adjusted for price inflation, rather than income.  Because, incomes generally rise faster than prices, the poverty threshold are much lower today compared to the median family income than they were in the 1960s.  Also, the poverty measures do not fully capture what we mean by poverty.  Families with incomes above the poverty line may nevertheless be deeply in debt, perhaps because of large medical bills. 

Conservatives argue that the thresholds are set too high because the Consumer Price Index that is used to adjust the poverty threshold has been shown to exaggerate the level of inflation as much as one percent per year.  They also note that the definition of income used in the measure does not include non-cash income such as Food Stamp benefits, public housing, or Medicaid benefits. Because the income measure is based on pre-tax income, it does not take into account the Earned Income Tax Credit (EITC) benefits that many families with children receive. 

Although there is some basis for both claims, one has to understand that poverty is inherently subjective concept to begin with; there is no such thing as a true poverty rate.  Most of the complaints about the poverty rate statistic concern its validity (whether or not the statistic measures what it is supposed to measures), but questions about the statistic's reliability (the consistency of the measurement) are often more to the point.

Consider the conclusion that might be drawn from the trends in child and elderly poverty rates in figure 6.    Because families with children are much more likely to receive EITC and Food Stamp benefits (which are not counted as income) than are the elderly, the disparity in poverty is not as great as the data indicate.  And because the EITC benefits were first offered in the 1980s and substantially increased in the first year of the Clinton administration, child poverty has declined at a faster rate than the data show.  On the other hand, because many of the elderly poor have savings and assets that do not enter into the calculation of the poverty rate, the poverty rate statistic may overestimate their level of poverty.

Measuring Income Inequality

Figure 8: The rich get richer


The income data used to derive the poverty measurements can also be used to address many other issues related to trends in income inequality.  Thus figure 8 shows trends in the share of aggregate family income for the richest 5% of households and the poorest 40%.   Since 1980, the richest households have seen their share of household income steadily increase.

Note, however, that conservatives sometimes object that the income distribution data reflect changes in household composition: the highest income groupings tend to comprise large households that the lower income groupings.  The top household income quintile, for example, contains 24.6% of the population, the bottom quintile, only 14.3%. (see Rector and Hederman, 2004).

One of the peculiarities in the Census surveys is that the households and families are defined at the time of the survey, while the incomes are the earnings of members of households over the previous year.  At the time of the CPS survey (e.g., March 2004) respondents are asked for the names (and relationships) of all members of the household and the earnings of each of those members over the previous year (January to December, 2003).  Thus, a family could be classified as having been poor in 2003 solely because the primary wage earner died or left the family in January of 2004.  As the rate of family disruptions have increased overtime, this may be the cause of relatively small increases in the poverty rate.

Table 2:


Much of the research on poverty relies on data obtained from the Panel Study of Income Dynamics conducted by the Institute for Social Research at the University of Michigan.  Unlike the CPS and other Census surveys, the panel study re-interviews the same families each year (or biannually), permitting analyses of the duration of poverty status and welfare participation.  Using these data, Katherine Bradbury and Jane Katz report evidence that income mobility has declined in the the US in the 1990s: Families are more likely to remain in the same "income quintile" than they were in the past (see table 2).  More than half of the families in the poorest income quintile were still poor ten years later; more than half of those in the richest quintile remained rich. (Bradbury and Katz 2002)

The annual US poverty data are based on a CPS sample of approximately 75,000 families.  Although this is a very large sample size (and with a relatively high response rate) in comparison to many other surveys and polls, the sample size is often not large enough to provide reliable data for small demographic groups or regions.  For the US population as a whole, the sampling error (based on a 90% confidence interval) is approximately +\-.2 percent.  Thus, if the reported poverty rate is 11.8%, we can be 90% sure that the true rate is between 11.6 and 12.0%.   For subpopulations in the survey (e.g.,  children, or two parent households) the sampling errors are greater, depending on the sample size.

The annual poverty reports do include state-level poverty data, but to compensate for the small sample size (especially for the smaller states), three-year averages are used.   The Decennial Census uses a much larger (census long-form) sample, providing data for smaller subpopulations and cities and counties.  Often in educational research, the percent of students who are eligible for the federal school lunch program is used as an indirect measure of the level of poverty in a school or school district.

Table 3:


It is commonly reported the women earn less than men; in the 1970s, much was made of the fact that women earned 59 cents for every dollar that a man was paid.  Near the top of table 3, we see that when comparing male and female full-time workers, female workers earn seventy cents for every dollar male workers earn.  When we compare female and male workers at similar ages or education attainment, however, the disparities are more complex.  The disparity between men and women declines sharply with age.  Differences in education, however, do not seem to consistently account for the disparity.  Note that the earnings reported in table 3 are for year-round full-time employees, in effect controlling for differences in part time work between men and women.

Additional controls, such as seniority and occupation, might account for more of the disparity permitting us to assess the degree to which equally situated and qualified men and women earn different salaries.  Notice, though, that controlling for some factors may not be enough: Even though the earnings for women in the youngest age group are most similar to male earnings, younger women tend to have higher levels of education than men of the same age.

Spinning Poverty Statistics.

It’s not really true what they say: “You can prove anything you want with statistics.”  It is true that you can find and\or present statistics in support of just about any claim, but it takes more than just the statistics to reach any conclusion.  To “prove” something with statistics usually takes at least two things: the statistics and some reasoning.   When people make bogus claims that something is true, it is usually the reasoning, not the statistics, that is at fault.  Consider the following exchange between talk show host Bill O'Reilly and one of his radio show callers:

CALLER: Hi, Bill.

O'REILLY: Larry.

CALLER: Let's see, poverty is up since Bush took office.

O'REILLY: That's not true.

CALLER: It is true.

O'REILLY: I have the stats right here, Larry.

CALLER: I just looked at the figures. Gun crime is up since George Bush took office.

O'REILLY: All right, Larry, hold it, hold it, hold it. Let's deal with one at a time. The only fair comparison is halfway through Clinton's term, halfway through Bush's term, OK? That's the only fair comparison. You gotta go real time.

CALLER: Bill, I --

O'REILLY: Poverty is down, Larry, one full percent in real time from 1996, halfway through Clinton, 2004, halfway through Bush. That is the truth, Larry, and if you're not willing to acknowledge that's the truth, this conversation is over.

O’Reilly’s claim, repeated at least two times on his “no-spin zone” Fox News broadcasts, is based on just two entirely accurate statistics combined with a lot of very faulty reasoning.  O'Reilly’s statistics are accurate: Bush's fourth year poverty rate is lower than Clinton's fourth year poverty rate (see figure 9).   And O’Reilly’s interpretation of the two numbers is based on an entirely reasonable premise: it would be unfair to compare what President Bush has achieved in four years with what it took President Clinton four years to achieve. Where he goes wrong is in using the rate at the end of the first term rather than the change in the rate: the poverty rate went down 1.2% in Clinton’s first term; up 1.4% in Bush’s first term.  Viewed another way: O’Reilly gives Bush credit for the decline in the poverty rate since the midpoint of Clinton’s term in office, but all of that decline -- and more took -- place while Clinton was in office.

 

Figure 9: How President Bush lowered the poverty rate.


The truth is that the poverty rate goes up and down regardless of who the president is and social and economic events and trends well outside the control of either the president or Congress have much more to do changes in the poverty rates than any policies the presidents pursued.   If President Clinton does deserve credit for policies that resulted in an eight year decline in poverty, why didn’t those policies continue to have the same beneficial effect during the Bush years? 

Surprisingly, O’Reilly did not resort to the more common method office holders use to misrepresent statistical trends that are going the wrong way: the average.  The average annual poverty rate for the eight years of the Clinton administration was 13.5; for the Bush administration’s first four years, 12.1 – even better one could compare Bush’s numbers to the average annual rate in just the first four years of the Clinton presidency.


References:

BBC News (2005), “G8 leaders 'real stars of show'” 2 July, 2005 http://news.bbc.co.uk/2/hi/uk_news/4643451.stm

Katharine Bradbury and Jane Katz, "Are lifetime incomes growing more unequal? New evidence on family income mobility in the
1990s suggests that the answer may be yes," Regional Review, Quarter 4 (2), pp 3-5.
http://www.bos.frb.org/economic/nerr/rr2002/q4/issues.pdf

Chen, Shaohua and Martin Ravallion (2000), “How have the world’s poorest fared since the early 1980s? 
Policy Research Working Paper 
3341, World Bank, 2000

<http://ideas.repec.org/p/wbk/wbrwps/3341.html>

Michel Chossudovsky (1999). "Global Falsehoods: How the World Bank and the UNDP Distort the Figures on Global Poverty" The Transnational Foundation For Peace and Future Research http://www.transnational.org/features/chossu_worldbank.html (viewed 12/20/04).

Deaton, A. S. and Kozel, V. (2004), “Data and Dogma: The Great Indian Poverty Debate”, Princeton University, mimeo Available at:
http://www.wws.princeton.edu/~rpds/downloads/deaton_kozel_datadogma.pdf (accessed 06.15.2006)

Gordon M. Fisher (1997). "The Development of the Orshansky Poverty Thresholds and Their Subsequent History as the Official U.S. Poverty Measure," US Census Bureau -  Poverty Measurement Working Papers http://www.census.gov/hhes/poverty/povmeas/papers/orshansky.html (viewed 6/1/05).

Garth L. Mangum, Stephen L. Mangum and Andrew M. Summ, (2003) The Persistence of Poverty in the United States (The Johns Hopkins University Press).

Lee Rainwater and Timothy M. Smeeding. (2003). Poor Kids in a Rich Country: America's Children in Comparative Perspective (Russell Sage Foundation)

Media Matters for America “O'Reilly defended false Clinton-Bush poverty comparison as ‘the only accurate measuring stick’” http://mediamatters.org/items/200509160002

Moynihan, Daniel Patrick. (1987).  Family and Nation (Harcourt Brace Janovich).

Robert Rector and Rea Hederman, Jr. (2004) "Two Americas: One Rich, One Poor? Understanding Income Inequality in the United States,"  The Heritage Foundation backgrounder #1791 August 24
http://www.heritage.org/Research/Taxes/bg1791.cfm

Reddy, Sanjay G.. and Camelia Minoiu (2005), “Chinese Poverty: Assessing the Impact Assumptions”, available on www.columbia.edu/~sr793/china.pdf

Reddy, Sanjay G.. and Camelia Minoiu (2005b), “Has World Poverty Really Fallen During The 1990s?”, available on
www.columbia.edu/~sr793/sensitivityanalysis.pdf

Reddy, Sanjay g. and Thomas W. Pogge (2005), “How Not to Count the Poor” Version 6.2 http://www.columbia.edu/~sr793/count.pdf

U.S. Census Bureau (2004), "Income, Poverty, and Health Insurance Coverage in the United States: 2003" Current Population Reports, P60-226, U.S. Government Printing Office, Washington, DC.