Working Paper No. 342 The Dynamics Of Urban Poverty In China by Niny Khor John Pencavel May 2007 Abstract Income information on the same urban Chinese households over six years in the 1990s provides an opportunity to describe movements in and out of poverty and to compare the dynamics of poverty in urban China with poverty dynamics in other countries. Poverty in urban China resembles poverty in other countries insofar as poverty is a transient state for many of the households who experience it in any given year. At the same time, an important fraction of all poverty episodes in urban China are accounted for by a relatively few households who experience long poverty spells. Keywords: poverty, China, United States, poverty dynamics JEL Codes: D31, D63, O15 Stanford University John A. and Cynthia Fry Gunn Building 366 Galvez Street | Stanford, CA | 94305-6015
34
Embed
The Dynamics Of Urban Poverty In - Stanford University1 See, for instance, Duclos, Araar, and Giles (2006), Jalan and Ravallion (1998) and Whalley and Ximing Yue (2006) all of whom
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Working Paper No. 342
The Dynamics Of Urban Poverty In
China
by
Niny Khor
John Pencavel
May 2007
AbstractIncome information on the same urban Chinese households over six years in the 1990s provides an opportunity to describe movements in and out of poverty and to compare the dynamics of poverty in urban China with poverty dynamics in other countries. Poverty in urban China resembles poverty in other countries insofar as poverty is a transient state for many of the households who experience it in any given year. At the same time, an important fraction of all poverty episodes in urban China are accounted for by a relatively few households who experience long poverty spells. Keywords: poverty, China, United States, poverty dynamicsJEL Codes: D31, D63, O15
Stanford University John A. and Cynthia Fry Gunn Building
366 Galvez Street | Stanford, CA | 94305-6015
1 See, for instance, Duclos, Araar, and Giles (2006), Jalan and Ravallion (1998) and Whalley andXiming Yue (2006) all of whom deal with rural poverty. By contrast, this paper is directed to urbanpoverty in China.
2 The Chinese Household Income Project is sponsored by the Institute of Economics, the ChineseAcademy of Social Sciences, the Asian Development Bank, and the Ford Foundation. Additionalsupport was provided by the East Asian Institute, Columbia University. An analysis of some of itsfindings is contained in Khan and Riskin (2001). The 1996 Chinese Household Income Projectresembles an earlier survey in 1989 that collected information for 1988 (Griffin and Renwei (1993)).In this paper, we shall be exploiting information from both surveys.
THE DYNAMICS OF URBAN POVERTY IN CHINA
Niny Khor and John Pencavel
I. Introduction
The substantial and sophisticated literature on poverty and on recent changes in poverty in
China has emphasized the importance of understanding the dynamics of poverty.1 As is well known,
inferences about poverty based on information about income or consumption in a single year provide
a misleading characterization of enduring poverty insofar as society is characterized by year-to-year
income or consumption mobility. The availability of retrospective information on incomes over six
years among urban households in China affords the opportunity to examine the degree to which the
incidence and severity of poverty in a single year is an accurate indicator of poverty measured over
a longer period of time. The purpose of this paper is to report on the dynamics of poverty in urban
China in the first half of the 1990s and to compare China’s experiences with those of some other
economies at the same time.
II. Household Panel Data for China
Data Sources
The information on Chinese household incomes in this paper are drawn from the Chinese
Household Income Project, CHIP, (Riskin, Renwei, and Li (2000)), a survey conducted in 1996 of
almost 7,000 urban households and about 8,000 rural households.2 This paper examines only urban
2
3 Rural households were not asked for their incomes in all years. Urban dwellers without a formalresidence certificate (hukou) were not surveyed.
4Income consists of the sum of labor income, property income, transfer income, and “income fromhousehold sideline production”. By far the largest fraction of total income is labor income.
5 For instance, Duncan and Hill (1989) compared the answers to retrospective questions in the PSIDwith highly accurate records from the employer. The quality of the respondents’ answers variedwith the type of information sought, but the average measurement error in respondents’ reports ofannual earnings was small: the average respondent’s reports of annual earnings were within onepercent of the earnings records of the employer and the ratio of the error in variance in reported
households.3 A description of how our samples were constructed is provided in the Appendix.
To address the questions raised in the Introduction about the dynamics of poverty, information
on incomes of the same households over time is needed. The 1996 CHIP survey supplies such data
in the form of responses of individuals to questions about “total income” not only for 1995 but also
for each of the previous five years.4
The Quality of the Chinese Data
The literature on recall data identifies several possible reasons to question the accuracy of the
information on incomes. One common problem arises from the respondent confusing the relevant
ranges of the reference period (telescoping) and another problem occurs when only prominent events
are recalled (the irregularity of memory). These two problems are less likely to introduce error in
annual income recalled than in the recall of other information. Errors in annual income are more
likely to occur from individuals reporting not their true incomes but the same incomes (or the same
proportion of their incomes) in successive years. If so, this will suggest less movement in incomes
and more persistence of poverty than is really the case.
While some research raises concerns about the reliability of retrospective data, other research
is more encouraging. A sweeping judgment regarding the accuracy of retrospective income data is
inappropriate.5 In our case, the respondent is asked to consult his or her income records for previous
3
annual earnings to the variance in true earnings was 0.154 in one case and 0.301 in another. See alsoKornfeld and Bloom (1999). Most job tenure data rely on retrospective information.
years and it has been suggested that, in China at this time, there was considerable social pressure on
individuals to take the survey questions seriously. A confident statement about the accuracy of the
retrospective income information in urban China in the early 1990s is impossible but previous research
on such income information suggests that it clearly warrants serious scrutiny. There is, after all, no
alternative.
A different type of problem arises from non-response: of the 6,931 households providing
information about income in 1995, 6,568 households also supplied income information for the years
from 1990 to 1994. In other words, about five percent of urban Chinese households with income data
in 1995 do not provide information on income in all the years from 1990 to 1994. This prompts the
question of whether non-response is random. If it is not, the sub-sample of households providing
income information in all years is not representative of all urban households. Table 1 compares the
values of some descriptive statistics for 1995 for the full sample of 6,931 households with the 6,568
“analysis” sample of households. The differences in median household income and in income
inequality for the two samples are relatively small. The Gini coefficient of total household income
is 0.278 for the full sample and 0.280 for the analysis sample. Therefore, we use the 6,568 urban
Chinese households providing information on their incomes in all years from 1990 to 1995. Income
in each calendar year is expressed in 1995 yuan by using the consumer price index as a deflator.
According to the data presented in Table 2, the size and composition of Chinese households
varies by income. Higher income households in urban China tend to have fewer children: on average,
those households in the top income decile have four-fifths of the number of children as the lowest
income decile. On the other hand, higher income households tend to have more adults and more
4
6 For all the results reported in this paper for China, an alternative adult equivalent scale was alsoapplied to household income, namely, one in which θ = κ = 1 and τ = 0.55. The implications of thisother scale were close to those reported in the tables below for θ = 0.5, κ = 0.7, and τ = 1.
7 If consumption information were also available we would analyze both consumption data andincome data notwithstanding the familiar problems with consumption data (the problem of imputingthe value of services from durable goods and the problem of dealing with commodities infrequentlypurchased) Retrospective information on consumption was not collected so we are not able tocompare inferences from using income data with those from using consumption data. Knight andLi (2006) describe differences between the use of income and the use of consumption expendituresin defining poverty in urban China in 1999.
workers than lower income households. Given these income-related differences in household size and
composition, it is important to assess whether our inferences about poverty are sensitive to adjustments
for demographic composition. Suppose y i denotes the total income of household i and let N Ai be the
number of adults and N Ci the number of children in household i. Then per adult equivalent household
income is
y i /[1 + θ.(N Ai - 1 )+ κ.N C
i ] τ
where κ is the weight attached to children (aged less than 14 years), θ is the weight attached to more
than one adult (aged 14 or more years), and τ is the scale economies parameter. To facilitate
comparisons with other countries later in this paper, we adopt the adult equivalent scale in which θ
equals 0.5, κ is 0.3, and τ is unity.6
The analysis of poverty requires first the specification of a threshold income corresponding
to the poverty line and second some gauge of resource inadequacy of the poor. Poverty and resource
inadequacy are measured in this paper by household income.7 With respect to the definition of the
poverty line, there are two broad approaches: one is to specify an absolute poverty line income which
identifies the income required to purchase the necessities of life; and the other is to recognize that
life’s “necessities” are dependent on time and place and to specify the poverty line income in relative
terms such as that corresponding to one-half of the society’s median income. In this paper, we present
5
8 A useful analysis of them is in Reddy and Minoiu (2006).
poverty statistics for China that use both absolute poverty and relative poverty lines.
A number of different absolute poverty lines are available for China.8 We adopt those
computed by Khan and Riskin (2001) who use them to analyze data from the same survey as we are
drawing on in this paper. Khan and Riskin (2001) estimate the cost of consuming a diet providing
adequate nourishment (a given caloric requirement) and of consuming some non-food items. This
poverty line is lower in urban areas than in rural areas due to the assumed greater caloric requirements
of work in rural areas. Khan and Riskin (2001) were aware of a certain arbitrariness in setting poverty
thresholds and, therefore, shied away from presenting just one threshold. We shall use both their
“low” poverty threshold of 1,604 yuan and their “high” poverty threshold of 2,291 yuan for urban per
capita household income in 1995. In earlier years, these thresholds are adjusted using the official
consumer price index.
As measures of relative poverty, we use 50 percent of median household income, 50 percent
of median per capita household income, and 50 percent of median “equivalised” household income
as poverty lines. The values of these poverty lines are given in Table 3. In 1995, a poverty line of
50 percent of per capita household income represents an income between Khan and Riskin’s “high”
and “low” thresholds. However, in 1990, half of median per capita household income of these 6,568
urban households corresponds to an income that is higher than Khan and Riskin’s “high” poverty
threshold. Hence, not merely are the poverty thresholds different from one another in 1995, they do
not bear a constant relationship to one another. What do they imply about urban poverty?
III. Household Poverty in 1995
Table 4 provides information on four poverty indicators for 1995. The headcount ratio
measures the percentage of households below the poverty line. The average poverty gap measures the
6
9 Sen’s index takes the value of zero if all households have incomes above the poverty level and itassumes the value of one hundred if all households have zero income, a maximum amount ofpoverty.
10 For the full sample of 6,931 households, our poverty indicators are similar to those reported forurban areas by Khan and Riskin (2001) for 1995 (2001). Khan and Riskin (2001) report a headcountratio of 8 % using their high poverty threshold and 2.7 % using their low poverty threshold, apoverty gap ratio of 2.0 using their high poverty threshold and 0.6 for their low poverty threshold,and a weighted poverty gap of 0.8 using their high poverty threshold and 0.3 using their low povertythreshold.
amount by which income falls short of the poverty line (the income deficit) averaged over those
households in poverty. The weighted poverty gap squares each household’s proportional income
deficit and then expresses the sum of these values over all poor households as a fraction of the total
population of households (Foster, Greer, and Thorbecke (1984)) . Sen’s (1976) index combines in one
indicator the headcount ratio, the average poverty gap, and the Gini coefficient of income among poor
households.9
Naturally, the picture of urban poverty depends critically on the poverty threshold. The
relative poverty lines suggest poverty levels that lie between the two absolute poverty levels. The
headcount ratio indicates between 3 and 11 percent of urban Chinese households below the poverty
line in 1995 with, naturally, a greater incidence of poverty implied by higher poverty thresholds.10 The
values of the poverty gap and the weighted poverty gap corresponding to the absolute poverty
thresholds are very close to those reported by Kahn and Riskin (2001) for 1995.
What do these poverty thresholds imply for changes in poverty over the years from 1990 to
1995? For this, we use the retrospective information on incomes asked in 1996. Figure 1 graphs the
poverty headcount ratios while Figure 2 graphs the values of Sen’s poverty index from 1990 to 1995
for our four poverty thresholds. These indicators all indicate a reduction in poverty between 1990 and
1995 with the poverty reduction being steeper for the absolute poverty thresholds than the relative
7
poverty thresholds. According to the “high” absolute poverty thresholds, the poverty headcount ratio
fell remarkably from 31.3 percent in 1990 to 11 percent in 1995 and Sen’s index in 1995 is almost
one-quarter of its value six years earlier. To the best of our knowledge, these are the first poverty
statistics for urban Chinese households in the first half of the 1990s that follow the same households
over time. They suggest a different time pattern from those implied by some other studies. For
instance, Khan and Riskin (2001) estimated that the widening distribution of income between 1988
and 1995 resulted in an increase in urban poverty, but these figures were not based on examining the
movements in income of the same households.
These indicators of poverty are all based on a “snapshot” of income in a given year. They tell
us nothing about whether the same households are in poverty from one year to the next nor whether
the poverty experiences of households are long or brief. By tracing the annual incomes of the same
households in urban China each year from 1990 to 1995, we aim to provide a different perspective on
poverty, one that emphasizes the persistence of poverty and the transitions of households into and out
of poverty.
IV. Household Poverty Dynamics in Urban China
The first question to be addressed is the degree to which annual poverty rates of households
signal longer run poverty. To effect this, consider the information contained in Table 5 which uses
the income information from 1990 to 1995 to compute the annual poverty rate, that is, the fraction of
households who are poor in each year from 1990 to 1995 and then averaged over the six years. On
average over the six years from 1990 to 1995, using the high absolute poverty threshold, twenty
percent of urban households were poor in a single year. On the low absolute poverty threshold, almost
eight percent were poor in a single year. The column headed “ever poor” reports the percent of
households who were poor in at least one year out of the six. According to the high absolute poverty
8
threshold, some 37 percent of households experienced poverty in at least one of those years. On the
low absolute poverty threshold, seventeen percent experienced poverty in at least one year.
The column headed “always poor” presents the fraction of households who were poor in all
six years. By the criterion of the high absolute poverty threshold, six percent of households were
always poor and, by the low absolute poverty threshold, just over one percent were always poor.
Finally, the column “permanently poor” in Table 5 presents the percentage of households for whom
the sum of income over all six years is below the sum of the poverty thresholds over six years.
According to the high absolute poverty threshold, almost 16 percent of households were permanently
poor and according to the low absolute poverty threshold almost five percent of households were
permanently poor.
In general, the poverty indicators using the poverty thresholds based on relative income are
not far from those based on the low absolute poverty threshold. The fraction of households who are
always poor is substantially below the fraction of households “ever poor” and this suggests that
poverty spells are not long: of those ever poor, sixteen percent are always poor on the high absolute
poverty threshold and, of those ever poor, about seven percent are always poor on the low absolute
poverty threshold. The ratio of the permanent poor to the annual poverty rate is between 62 and 80
percent implying that, although the annual poverty rate overstates long-term poverty, nevertheless the
annual poverty rate does not overstate permanent poverty by a great deal.
What observable characteristics of households are associated with their poverty status? This
question is addressed by the maximum likelihood logit estimates given in Tables 6 and 7 which reports
the relationship between the differing poverty probabilities and various household attributes. The
estimates in Table 6 use the high absolute poverty level as the poverty threshold while the estimates
in Table 7 use 50% of equivalised income as the poverty threshold. The results reported in Tables 6
9
11 In a study of poverty using annual household consumption expenditures, Meng, Gregory and Wan(2007) also found a strong positive effect of household size (holding constant the number ofhousehold members who are working) on the probability of a household ever being poor. Inaddition, they report that households with a larger fraction of members who are children are morelikely to be poor in a given year.
and 7 suggest that a household’s poverty probability is lower when the household is headed by a
woman, is lower when the household head has more schooling, is higher in households with more
children, and is lower for households located in a Coastal province. The effects are particularly large
for the probability of ever being poor over these years. According to the estimates corresponding to
the high absolute poverty threshold in Table 6, the probability of ever being poor was 28 percent lower
for a household whose head had a college education than for a household whose head had less than
elementary schooling; the probability of ever being poor was almost 5 percent lower for a household
whose head was a member of the Communist Party than for a household whose head was not a
member. The probability of ever being poor was 41 percent higher in a household with two more
children than another household that was otherwise identical.11
Comparing the values of the probability derivatives across columns of Table 6 or Table
7 for each variable, the effects of observed variables on the probability of ever being ever poor are
larger than the effects of always being poor. Thus a household head whose highest schooling
attainment is a college education has a 28 percent lower probability of ever being poor, a 14 percent
lower probability of being permanently poor, and a 4 percent lower probability of always being poor
than a household whose head has at most an elementary education.
Poverty Spells
The comparisons between the “ever poor” and the “always poor” indicate movements in and
out of poverty and the length of poverty spells over six years. Information on the extent of poverty
transitions among urban Chinese households in the first half of the 1990s is provided by the length of
10
spells of poverty over six years and by movements in and out of poverty. Table 8 presents the
distribution of years in poverty among those ever poor in any year from 1990 to 1995. According to
all the different poverty thresholds, among those ever poor, most households are poor for one year and
then the next most frequent occurrence are those poor for two years. Depending on the poverty line,
between 44 and 55 percent of households poor in any year are poor either in one or two of the six
years from 1990 to 1995. Among those households in poverty in any year of the six years from 1990
to 1995, the average number of years in poverty was at most a little over three years and, by other
poverty criteria, between two and three years. For these urban households in the 1990s, poverty seems
to be predominantly a transient state.
Another indicator of the mobility of the poor is to calculate the total number of household-
years in poverty (that is, the sum over all households of all years in poverty) and determine what
fraction of these household-years in poverty consist of households who experience x years in poverty
where x takes on values from one to six. This is answered by the data reported in Table 9. By the
standard of poverty given by the high absolute poverty line, 53.1 percent of household-years in
poverty are accounted for by households in poverty for five to six years. Indeed, according to each
poverty line, five or six year spells of poverty account for a minimum of 40 percent of household-years
in poverty. A typical year spent in poverty is experienced by households who are in poverty for a long
spell so although there is much movement in and out of poverty, much of that movement appears to
be associated with households who experience repeated negative income shocks.
Movements of households in and out of poverty are given in Table 10. Typically
between one and three percent of non-poverty households in one year fell into poverty the next year
while between one-fifth and one-half of households in poverty one year left poverty the following
year. During the first half of the 1990s, poverty was far from being a fixed state for many urban
11
12 For the United States, the data below relate to the years 1987-89. The information below forOECD countries is drawn from OECD (2001) and Valletta (2006).
Chinese households. Among those households who were ever poor, most were poor in one or two
years and yet, a large fraction of poverty episodes across all households consisted of households in
poverty for five or six years. Hence there is the apparent contradiction that, while the typical poverty
interval was short, long poverty episodes accounted for a large fraction of all years in poverty
experienced by all households.
V. Comparisons of Individual Poverty in China with Other Countries
How does this description of poverty in urban China compare with that in other
countries at the same time? To contrast our findings on the dynamics of poverty in China with
existing findings for some OECD countries, we adopt the convention from the previous research of
making the individual - not the household - the unit of analysis. This means that, if a household’s
income is below the poverty threshold, then all individuals within that household are deemed to be
poor.
Existing research on poverty in OECD countries in the 1990s takes a three year window
- in most cases, 1993-95 - to analyze the dynamics of poverty12 and, to maintain comparability with
this previous research, we shall do the same with these data for China. From our data on China from
1990 to 1995, two windows of three years are specified: the first is the three years from 1990 to 1992
and the second is the three years from 1993 to 1995. We shall use both of these three year windows
for China.
Also to be compatible with the previous OECD research, we use the definition of
“equivalised” household income from above. Finally, in line with the OECD research, we employ a
12
13 Whereas the data for China are restricted to urban individuals, those for other countries describeindividuals in both urban and rural areas.
relative concept of a poverty line, namely, one that is 50 percent of a country’s median value of
eqivalised household income in a given year. We have seen above that, for China, such a poverty
threshold generates values regarding poverty that tend to lie between the low and the high absolute
poverty thresholds. Clearly, when engaging in international comparisons of income, the use of a
relative poverty threshold avoids the need to use purchasing power parity exchange rates (which are
especially difficult to compute for China) to convert the incomes of different countries into the same
units and this makes the analysis of international comparisons of poverty more tractable. This
comparative analysis is designed not to compare poverty levels across these countries, but to compare
the character of poverty and, in particular, to compare the dynamics of poverty across countries. That
is, regardless of the precise poverty threshold, are individuals in poverty permanently and what is the
probability that someone in poverty in one year will leave poverty by next year? How do the answers
to these questions differ across countries?
The headcount ratios of individual poverty in the different countries in a three year
window in the 1990s are presented in Table 11.13 The “annual poverty rate” in this table presents the
percent of individuals below the poverty threshold as calculated for each of the three years and then
averaged over the years. Given the concept of relative poverty being applied to define the poverty
threshold, these values of the incidence of poverty reflect the degree of annual income inequality at
the lower part of the income distribution in these countries. Hence, among these countries, using this
measure of relative poverty, the annual poverty rate is highest in the United States and is least in
Denmark where income dispersion is lower.
Relevant to our principal concern with the dynamics of poverty, the column of Table 11 headed
13
14 The entry and exit rates are measured in the same way as those reported in Table 10 except thatthose in Table 10 describe households and those in Table 12 describe individuals.
15 “Big drop” is defined as, of those entering poverty in year t, the percent whose income in year t-1was at least 60% of median income. “Big rise” is defined as, of those leaving poverty in year t , thepercent whose income in year t + 1 was at least 60 percent of median income.
“ever poor” lists the percent of individuals poor in at least one year out of the three. The column
headed “always poor” presents the percent of individuals poor in all three years. The “permanent
poor” is the percent of individuals for whom the sum of three year (equivalent) income is less than the
sum of the poverty thresholds over three years. The final column of Table 11 reports the ratio of the
percent “always poor” to the percent “ever poor” and, in all countries, less than one-half of those “ever
poor” are “always poor”. In 1993-95, the incidence of “always poor” to “ever poor” in urban China
is the same as that in France. In all these countries, the average annual poverty rate is above the
percent permanently poor which suggests that measures of poverty based on annual income overstate
the incidence of permanent poverty.
Further information on the dynamics of poverty is provided by Table 12 which compares
annual rates of entry into poverty and exits from poverty in urban China with other countries.14 Among
those ever poor, the mean years in poverty during the period 1993-95 were 1.8 in urban China and 1.7
in Europe. In China in 1993-95, as indicated by the entries for “big” drop and “big” rise, 35 percent
of those leaving poverty and 35 percent of those entering poverty experience sizeable changes in their
income.15 This suggests a volatility of income towards the bottom of the income distribution in China
in these years
These figures on movements in and out of poverty in urban China suggest a considerable
fluidity of income and of poverty status. This is also the inference from studies of OECD economies
so, in this sense, poverty in urban China, resembles that in these other countries. For many
14
individuals, poverty is a transient state. At the same time, of all poverty episodes over several years,
a sizeable fraction is experienced by a relatively small number of individuals who experience a
succession of negative shocks.
In general, these indicators of poverty dynamics would not identify urban China as a poverty
outlier in any meaningful sense.
VI. Conclusions
In urban China in the 1990s, because the percent of households who were always poor
was considerably less than the percent of households who were ever poor in any year, one may infer
that poverty sprells were not long. The ratio of the fraction of households permanently poor to the
fraction poor based on income in a single year is about two-thirds or three-quarters. Hence the annual
poverty rate overstates long-term poverty somewhat. Of those households poor in any year, most are
poor for one year. For the six years from 1990 and 1995, of the households poor in any year, about
one-half are poor in one year or in two years. In urban China between one and three percent of non-
poverty households in one year fall into poverty the following year while between one-fifth and one-
half of poverty households in one year leave poverty the following year. At the same time, long
poverty episodes account for a large fraction of the sum of all poverty spells experienced by all
households: between forty percent and fifty-three percent of all years in poverty experienced by all
households in urban China from 1990 to 1995 are accounted for by households in poverty for either
five or six years. The implication is that some households experience a succession of negative
income shocks.
A comparison of the poverty dynamics of individuals in urban China in the first half
of the 1990s with that in other countries suggests a broad similarity: the annual poverty rate is always
15
above the percent who are permanently poor. In all these countries, for those toward the bottom of the
income distribution, a substantial volatility in incomes is suggested. While poverty is a transitory
state for many individuals, of all years spent in poverty, the typical year in poverty is experienced by
people undergoing a poverty episode lasting several years.
16
REFERENCES
Duclos, Jean-Yves, Abdelfrim Araar, and John Giles, “Chronic and Transient Poverty: Measurement
and estimation, with Evidence from China”, unpublished paper April 2006.
Duncan, Greg J., and Daniel H. Hill, “Assessing the Quality of Household Panel Data: The Case of
the Panel Study of Income Dynamics”, Journal of Business and Economic Statistics, 7(4), October
1989, 441-52..
Foster, James, Joel Greer, and Erik Thorbecke, “A Class of Decomposable Poverty Measures”,
Econometrica, 52(3), May 1984, 761-66.
Griffin, Keith, and Zhao Renwei (1993), “Chinese Household Income Project, 1988" [computer file],
Hunter College Acadedmic Computing Services [producer], ew York, NY 1992, Inter-university
Consortium for Political and Social Research [distributor], Ann Arbor Michigan, 1993.
Jalan, J., and Ravallion, M. “Transient Poverty in PostReform Rural China”, Journal of Comparative
Economics, 26 (2), 338-57.
Khan, Azizur Rahman, and Carl Riskin, Inequality and Poverty in China in the Age of Globalization,
Oxford University Press, New York, 2001.
Knight, John, and Li Shi, “Three Poverties in Urban China”, Review of Development Economics, 10
(3), 2006, 367-87.
17
Kornfeld, Robert, and Howard Bloom, “Measuring Program Impacts on Earnings and Employment:
Do Unemployment Insurance Wage Reports from Employers Agree with Surveys of Individuals?”
Journal of Labor Economics 17(1), January 1999, 168-197.
Meng, Xin, Robert Gregory, and Guanhua Wan, “Urban Poverty in China and Its Contributing
Factors, 1986-2000", Review of Income and Wealth, 53 (1), March 2007, 167-189.
OECD, “When Money is Tight: Poverty Dynamics in OECD Countries”, Employment Outlook, Paris,
June 2001, pp. 37-87.
Reddy, Sanjay G. and Camelia Minoiu, “Chinese Poverty: Assessing the Impact of Alternative
Assumptions”,, Columbia University unpublished paper, version 4.31, 2006.
Riskin, Carl, Zhao Renwei, and Li Shi, “Chinese Household Income Project, 1995" [computer file],
ICPSR version, Amherst, MA, University of Massachusetts. Political Economy Research Institute
2000, Inter-university Consortium for Political and Social Research [distributor], An Arbor Michigan,
November 2000.
Sen, Amartya, “Poverty: Am Ordinal Approach to Measurement”, Econometrica, 44(2), March 1976,
219-31.
Whalley, John and Ximng Yue, “Rural Income Volatility and Inequality in China”, National Bureau
of Economic Research, Working Paper 12779, December 2006.
18
Robert G. Valletta, “The Ins and Outs of Poverty in Advanced Economies: Government Policy and
Policy Dynamics in Canada, Germany, Great Britain, and the United States, Review of Income and
Wealth, 52 (2), June 2006, 261-84.
19
Figure 1
Percent Poverty Headcount Ratios, 1990-1995, Urban China
“absolute high” and “absolute low” denote, respectively, Khan and Riskin’s “high” and “low” poverty
thresholds. “50% MPCI” stands for fifty percent of median per capita income. “50% MEHI” stands
for fifty percent of median equivalised household income.
20
Figure 2
Sen’s Index of Poverty, 1990-1995: Urban China
“absolute high” and “absolute low” denote, respectively, Khan and Riskin’s “high” and “low” poverty
21
thresholds. “50% MPCI” stands for fifty percent of median per capita income. “50% MEHI” stands
for fifty percent of median equivalised household income.
22
Table 1
Measures of Urban Household Income and Income Inequality in 1995 for Two Groups: Urban China
6,568 households
providing information on
income in all six years
6,931 households
providing information on
income in year 1995
total household income
median in 1995 yuan 12,455.50 12,524.00
Gini coefficient 0.280 0.278
ratio of 90th to 10th percentile 3.365 3.331
per equivalent adult household income
median in 1995 yuan 6,432.78 6,477.5
Gini coefficient 0.271 0.269
ratio of 90th to 10th percentile 3.252 3.222
Notes to Table 1: The adult equivalent scale deflates household income by adjusted household size.
Household size is adjusted by applying a weight of unity to the first adult, a weight of 0.5 to each
additional adult (aged 14 or more years), and a weight of 0.3 to each child (aged less than 14 years).
23
Table 2
Household Size and Composition by Household Income Decile: Urban China, 1995
income decile
1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th
N C 0.711 0.728 0.784 0.749 0.744 0.699 0.706 0.597 0.546 0.566
N A 2.094 2.234 2.275 2.286 2.407 2.394 2.466 2.645 2.755 2.966
N A + C 2.805 2.962 3.059 3.035 3.151 3.093 3.172 3.242 3.301 3.532
N W 1.236 1.472 1.656 1.656 1.799 1.779 1.826 1.906 2.079 2.282
N C denotes the average number of children, N A the average number of adults, N A + C the average
number of household members, and N W the average number of workers. Children are defined as those
aged less than 18 years and adults as 18 or more years. There are 6,568 urban Chinese households.
The 1st decile is the lowest income decile and the 10th is the highest.
24
Table 3
Alternative Poverty Lines for Household Income in Urban China in 1995 Yuan
absolute poverty lines defined
by
household per capita income
relative poverty lines
50% of median ....
“high”
threshold
“low”
threshold
total income per capita
income
“equivalised”
income
1990 1,249.91 875.10 4,490.68 1,466.34 2,430.62
1991 1,292.28 904.77 4,875.29 1,602.64 2,619.89
1992 1,374.75 962.51 5,332.75 1,777.58 2,852.93
1993 1,576.80 1,103.94 5,630.28 1,870.71 2,986.67
1994 1,956.60 1,369.86 5,503.33 1,814.93 2,902.40
1995 2,291 1,604 6,227.33 2,037.58 3,216.39
“Eqivalised” income deflates household income by an adult equivalent scale that applies a weight of
unity to the first adult, a weight of 0.5 to each additional adult, and a weight of 0.3 to each child.
No. of Adults 0.097 (0.009) 0.016 (0.002) 0.056 (0.005)
No. of Children 0.205 (0.013) 0.041 (0.003) 0.116 (0.007)
Coastal Province -0.176 (0.012) -0.041 (0.004) -0.097 (0.008)
This equation is fitted to 6,568 observations. Estimated standard errors are in parentheses. Forcontinuous variables, marginal effects are partial derivatives while, for discrete variables, the effectsshow the change in the value of the dummy variable from zero to unity. These effects are evaluatedat the mean values of the right-hand side variables. “Age” measures the years of age of the householdhead. “No. of Adults” is the number of people aged 18 or more in the household and “No. ofChildren” is the number of people aged 18 years or less in the household. All the other variables aredichotomous. Age, Schooling, Minority, Communist Party membership, and Woman are allcharacteristics of the household head. “Schooling 1" takes the value of unity for someone with acollege education. “Schooling 2" takes the value of unity for someone with a professional schooleducation. “Schooling 3" takes the value of unity for someone with a middle level professional,technical, or vocational school education. “Schooling 4" takes the value of unity for someone withan upper middle school education and “Schooling 5" takes the value of unity for someone with a lowermiddle school education. “Schooling 6" takes the value of unity for someone with an elementaryschool education. The omitted schooling category is “Schooling 7" which relates to someone with less
28
than elementary schooling. “Coastal province” takes the value of unity for households residing in1995 in Liaoning, Jilin, Jiangsu, Zhejang, Shandong, Guangdong, Fujian, and Beijing.
Table 7
Marginal Effects from Logit Estimates of the Probability of Being Poor where Poverty is Defined
according to 50% of Equivalised Income: Urban China