Three I's of Poverty Curves PDF

Three 'I's of Poverty Curves, with an Analysis of UK Poverty Trends
Author(s): Stephen P. Jenkins and Peter J. Lambert

Source: Oxford Economic Papers, New Series, Vol. 49, No. 3 (Jul., 1997), pp. 317-327
Published by: Oxford University Press
Stable URL: http://www.jstor.org/stable/2663596
Accessed: 26-11-2016 00:06 UTC
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
Oxford University Press is collaborating with JSTOR to digitize, preserve and extend access to Oxford
Economic Papers
This content downloaded from 175.111.91.19 on Sat, 26 Nov 2016 00:06:06 UTC
All use subject to http://about.jstor.org/terms
Oxford Economic Papers 49 (1997), 317-327
THREE 'I'S OF POVERTY CURVES, WITH AN

ANALYSIS OF UK POVERTY TRENDS
By STEPHEN P. JENKINS* and PETER J. LAMBERTt
ESRC Research Centre on Micro-Social Change, University of Essex,
Colchester C04 3SQ, internet: stephenjlessex.ac.uk
t Department of Economics, University of York
Three 'I's of Poverty (TIP) curves, based on distributions of poverty gaps, provide evocative
graphical summaries of the incidence, intensity, and inequality dimensions of poverty, and a
means for checking for unanimous poverty orderings according to a wide class of poverty
indices. The orderings may be robust to variations in the deflator used to make nominal
incomes comparable. An analysis of poverty in the UK during the 1980s demonstrates
these properties.
1. Introduction
RECENT income inequality research has demonstrated the value of graphical

devices such as frequency density, distribution functions, and (generalized)
Lorenz curves for both describing income distributions and implementing
tests of whether one distribution dominates another. In this paper we provide
a device, the 'Three "I"s of Poverty' (TIP) curve, which plays the same dual
role in poverty analysis.
TIP curves are based on distributions of income shortfalls from the poverty
line-poverty gaps-and are so named because of their ability to succinctly
and simultaneously portray what we label the incidence, intensity, and inequal-
ity dimensions of aggregate poverty. Moreover, orderings of distributions by
non-intersecting TIP curves correspond to unanimous poverty orderings
according to a wide class of poverty indices. These 'generalized poverty gap'
indices have the properties that not only does an increase in a poor person's
poverty gap increase aggregate poverty, but so too does a pure transfer of
income from a poorer poor person to a richer poor person. In fact the indices
respect Sen's (1976, 1979) influential arguments that aggregate poverty indices
should satisfy focus, monotonicity, and transfer axioms.
Atkinson (1987) and Foster and Shorrocks (1988a, b), have also proposed
criteria for unanimous poverty orderings. Our own criterion, which is based on
comparisons of TIP curves of poverty gap distributions rather than com-
parisons of generalized Lorenz curves of income distributions or comparisons
of poverty deficit curves, provides a more revealing picture of poverty and its
distribution. Moreover our construction permits the strength of the poverty
comparison to be tested, revealing that the original poverty ordering may
be preserved throughout a range of values for the deflator used to make
nominal incomes comparable. Thus the TIP curve has useful applications in
poverty comparisons across time, across regions and countries, and between
? Oxford University Press 1997
318 THREE 'I'S OF POVERTY CURVES: UK POVERTY TRENDS
population sub-groups of different household composition. We illustrate this

with an analysis of poverty in the UK during the 1980s.
2. Poverty in pictures
Let x = (x1, x2,..., x") denote a distribution of income among n income-

receiving units (e.g. persons or households), in which the incomes have been
arranged in ascending order, 0 < xi < x2 < .. < x", and let z be the poverty
line. We assume that incomes have already been adjusted to account for differ-
ences in needs and household composition using a suitable equivalence scale.
Let gx be the vector of poverty gaps associated with incomes x
gxi=max{z-xi,0} (1)
and let b, be the associated vector

incomes (Foster and Shorrocks, 1
bxi = min {xi, z} = z-gxi (2)
Given the value of the poverty line z, many familiar poverty indices may be
defined as functions of the vector gx or, equivalently, of the vector bx. See
Table 1, in which a class of poverty indices P is defined. In fact many of the
indices in P depend on the poverty line only through the ratio of each poverty
gap to the poverty line, i.e. they may be expressed solely in terms of normalized
poverty gaps
rxi= gxi/x = max {(z - xi)/z, 0} = max {1 - xi/z, 0} (3)

In Table 1, we denote by Q the class in which the poverty indices are defined
over normalized poverty gaps. Clearly Q C P.
TABLE 1
Generalized poverty gap poverty indices: examples
Class membership
Poverty indices P Q
Chakravarty (1983) I 1
Clark et al. (1981) Type 2 / v
Foster et al. (1984), FGT(ci), a > 1 / /
Hagenaars (1987) 'Dalton' type I/ some
Johnson (1988)
Pyatt (1988)
Shorrocks (1995) modified-Sen / I/
Thon (1979)* i/ /
Watts (1968) / /
P is the class of replication invariant increasing Schur-convex functions of poverty gap vectors. Q is as P,
except defined over normalized poverty gaps. * Only for the limiting form as n -+ oo (when it equals the modified-
Sen index).
S. P. JENKINS and P. J. LAMBERT 319
Cumlative sum of
poverty gaps per capita
TIP(Tp)
MrENSITY . ............................
(height)
/(curvatur
IINCIDENCE'
(length)
0 h 1
Cumulative po
FIG. 1. The TIP curve and the 3 'I's of poverty.
Our picture of poverty is obtained by ranking people from poorest to riches

cumulating their poverty gaps (or normalized poverty gaps), and plotting
them. We call the resulting picture the TIP curve.1
The TIP curve for poverty gaps, denoted by TIP(g; p) where 0 < p < 1,
plots against p the sum of the first 100% of g-values divided by the total
number of income-receiving units. Thus TIP(g; 0) = 0 and TIP(g; k/n)=
Ek l gi/n for integer values k < n; at intermediate points, TIP(g; p) is det
mined by linear interpolation: see Fig. 1. TIP(g; p) is thus an increasing con-
cave function of p, with slope at a given percentile equal to the poverty gap for
that percentile. The curve is horizontal at all p corresponding to incomes at or
above the poverty line.
The incidence aspect of poverty is summarized by the length of the TIP
curve's non-horizontal section. The headcount ratio (the proportion of the
population poor), h, is that p at which the curve becomes horizontal. The
intensity dimension of poverty is summarized by the height of the TIP curve:
the vertical intercept at p = 1 is the aggregate poverty gap averaged across all
income-receiving units. The average poverty gap amongst the poor is equal to
the slope of the ray from (0, 0) to (h, TIP(g; h)). The inequality dimension of
poverty is summarized by the degree of concavity of the non-horizontal section
of the TIP curve. If there were equality of incomes amongst the poor-if
1 In previous versions of this paper we labelled the curve the inverse generalized Lorenz curv
following the terminology introduced by Jenkins (1991, 1994a), who used the device for wage
discrimination measurement. Yitzhaki (1991) labels the curve an 'absolute rotated Lorenz curve'
in the income inequality context. In the industrial concentration literature, Hannah and Kay (1977,
pp. 58-9) refer briefly to an 'absolute concentration curve'. Shorrocks (1993) uses a similar curve,
labelled a 'duration profile', to summarize distributions of unemployment spell lengths; his survey
(1997) refers to 'deprivation curves'.
poverty gaps were equal-this section would be a straight line with slope equal
to z minus the average income amongst the poor.2
By analogy with inequality measurement for which the cases of maximum
and minimum inequality yield bounding Lorenz curves, there are also max-
imum and minimum poverty reference situations and bounding TIP curves.
Define maximum poverty as being when each person in the population has an
income of zero and hence a poverty gap of z, in which case the TIP curve is a
straight line from the origin with slope z (and vertical intercept z at p = 1). At
the other extreme, when no person is poor, the TIP curve coincides with the
horizontal axis.
The TIP curve for normalized poverty gaps, denoted TIP(r, p), has the same
shape properties and thus also shows the 'three "I"s of poverty'. Its right-hand
vertical intercept is the Foster et al. (1984) poverty index FGT(a) with para-
meter a = 1, and Sen's (1976, 1979) income-gap ratio index is equal to the
slope of the ray from (0,0 ) to (h, TIP(r; h)). The area under TIP(r, p) equals
one half of the Shorrocks (1995) modified-Sen poverty index which equals
[(n+1) THON-FGT(1)]/2n, where THON is the Thon (1979) poverty
index. (The area approaches THON/2 as n -* oo.) In the maximum poverty
case, the curve follows the 450 line, and the area underneath it equals 0.5.
Analogous to the Gini coefficient, which equals the ratio of the area above
the Lorenz curve to the area above the maximum inequality Lorenz curve, the
modified-Sen index (and the Thon index in the limit) equal the ratio of the area
under the TIP curve for normalised poverty gaps to the area under the max-
imum poverty TIP curve.
No other graphical device used in the poverty literature pictures poverty as
completely and transparently as the TIP curve does. The forte of generalized
Lorenz curves (Shorrocks, 1983) and poverty deficit curves (Atkinson, 1987) is
that they permit the checking of poverty dominance criteria.3 But TIP curves
also have this property, as we shall show.
3. Poverty dominance
Given the poverty line z, denote a typical index in the class P by P(xlz), and a
typical index in the class Q by Q(xlz). If one TIP curve lies wholly above
another TIP curve, we have a situation of TIP dominance.4 TIP dominance
2 If a country's social assistance benefit programme provided a universal minimum income,

< < z, its nature would be revealed by the TIP curve. The curve would have a linear segment
with slope z - q from the origin up to percentile pz,, the proportion of the population receiving the
minimum income.
3 The curves are all related of course. For instance, the TIP curve shows, at each p, the vertical
distance between two generalized Lorenz curves: one for the censored income distribution bx, the
other for the distribution in which each person is attributed with an income equal to z (a ray from
the origin with slope z). Atkinson's (1987) poverty deficit curve, zFGT(l), shows how the aggregate
poverty gap per capita varies as the poverty line varies. It is a dual representation of the generalized
Lorenz curve (with cumulation by income rather than income receiver).
4g TIP dominatesf if TIP(g; p) > TIP(f; p) for all p E [0, 1]. Strict TIP dominance demands that
this inequality be strict for some p.
of un-normalised poverty gaps is equivalent to a unanimous poverty ordering

by all indices in P for all poverty lines set at z or lower, whilst TIP dominance
for normalized gaps is equivalent to a unanimous poverty ordering by all indices
in Q, for all poverty lines set at z or lower. These results are proved in Jenkins
and Lambert (1996). Although the headcount ratio, the mean poverty gap
amongst the poor, and the income-gap ratio are not members of P or Q,
orderings by these indices can also be seen from TIP curve configurations.
If TIP curves intersect once, and if FGT(2) is higher in the initially dominant
one, then poverty is greater in that distribution for a sub-class of P or Q for
which aversion to income inequality amongst the poor is sufficiently large: see
Jenkins and Lambert (1994) for details.
Poverty comparisons are typically made conditional on some prior choice of
scaling factors used to make the income distributions comparable in real
income terms. There are three main examples of such scaling: (i) intertemporal
comparisons, in which distributions for the same population at different points
in time are made comparable using a price index which accounts for inflation;
(ii) inter-regional or cross-national comparisons, in which distributions from
two different regions of a country (e.g. rural and urban), or for two different
countries, are made comparable using an exchange rate or purchasing power
parity index; and (iii) inter-group comparisons, in which household type and
composition is the same within two population sub-groups but differs between
them, and the component distributions are made comparable using an equiva-
lence scale.
If TIP dominance of gy over gx obtains, and is found to be strong, then not
only is poverty less in x than in y according to the class P, but there may be
room to re-scale upwards the incomes in y, the poorer distribution, whilst
preserving the poverty ranking between x and y for the sub-class Q, or equiva-
lently to lower the poverty line for distribution y alone, all the while preserving
the poverty ordering for the sub-class Q. The extent to which this may be done
is revealed by the poverty gap data themselves, and the test to be conducted is
fully described in Jenkins and Lambert (1996). The intuition is that when the
TIP curve for y lies everywhere above that for x, there will be scope for low-
ering TIP(ry, p) while maintaining non-intersecting TIP curves of normalized
poverty gaps and the ordering Q(ylz) ? Q(xlz). Such downward shifts are
brought about by re-scaling the incomes in y upwards (equivalently, lowering
the poverty line in y, the poorer distribution). The potential for doing this will
be greater the larger the vertical distance between TIP(gy, p) and TIP(g, p) at
each p < hy, and the nearer TIP(gy, p) is to the line p.z of maximum poverty.
Our analysis of UK poverty which follows demonstrates the usefulness of the TIP
dominance property and also shows what can be said when TIP curves intersect.
4. Empirical application: poverty in the UK
We analyze UK poverty in 1979, 1981, and 1988/89 with the Households Below
Average Income (HBAI) data set already used by the Department of Social
Security (DSS) to provide official statistics on low incomes. The data are
particularly suitable for our purposes as they are the best quality UK
income micro-data which are publicly available, but HBAI reports (see for
example DSS, 1992) summarize the amount of poverty using only headcount
ratios. The HBAI results are therefore ripe for re-analysis using TIP curves and
distributionally-sensitive poverty indices. Moreover the time periods covered
include both macroeconomic recession and boom.
We use the same income distributions as the DSS did: we have their micro-
data (for 1988/89 this is pooled data for 1988 and 1989). For each calendar
year, the sample comprises at least 7,000 households, covering more than
18,000 individuals. 'Income' is household money income from all sources,
including cash social assistance and social insurance benefits, and after the
deduction of direct taxes (but prior to the deduction of housing costs). Incomes
are expressed in April 1992 pounds per week using the HBAI 'RPI excluding
local taxes' price index; for example, nominal incomes for 1979 are scaled
up by a factor of about two to make them comparable with incomes for
1988/89.5 To account for differences in household composition, money incomes
are deflated by the official McClements equivalence scale. Scale rates are
normalised at 1.0 for childless married couple households, and 0.61 for
single adults (rates for children are less, and depend on age). Equivalized net
household incomes are assumed equally shared within households, and the
income-receiving unit is the individual. The data are also adjusted and
weighted to make them nationally representative.6 In what follows, we
shall simply refer to 'income' rather than 'equivalized real net household
income'.
We investigate the case in which the poverty line is taken to be half 1979
average income, i.e. ?92.58 pounds per week. This real income cut-off is
also about half 1981 average income, but less than 40% of 1988/89 average
income: see Table 2. Although there is no official poverty line in the UK-
indeed the HBAI report always refers to 'low income', and never to 'poverty'-
half the average income is used as the poverty line by many commentators. Our
analysis is in terms of half the 1979 average, but it also provides conclusions for
poverty lines equal to half the contemporary mean (see below).
We begin with a comparison of 1979 and 1988/89. Figure 2 shows the TIP
curves of un-normalized poverty gaps, and it is clear that the 1988/89 distri-
5 More precisely, index values include January 1979 = 52.9, January 1981 = 70.6, January
1987=100.0, January 1988=103.1, January 1989=110.8, and April 1992=138.7. For each
year, incomes from different interview months were converted to January prices using the c
sponding within-year deflators.
6 The only difference between our income distributions and the DSS's is that, for compatibility
with the theory, we have re-set zero income values in a given year at the smallest positive value
minimum for that year (negative income values have already been censored at zero in the HBAI
data). The numbers of persons affected, and the minimum positive values, were 34 and ?5.08 for
1979, 38 and ?1.21 for 1981, and 144 and ?0.81 for 1988/89. We have checked that neither (i)
excluding these cases, nor (ii) leaving the data unadjusted, affects the ordering results reported
below.
2 - maOmumnthi seho I hi
I I I L . l | .~~~~~~~~~~~~~~~I
A 8 ; X 1.73
0.
1.5 -- , 4
>/
0.
E 30~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3 06~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
0~ ~ ~ 19794 .6.09.01 .
.5.
3 CL~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
0 .02 .04 .06 .069 .081 .1

Cumulative population share, p
FIG. 2. 1988/89 poverty gaps TIP dominate 1979 poverty gaps (poverty line =
income).
bution TIP dominates the 1979 one.7 Hence poverty in 1988/89 is higher than
in 1979 according to all poverty indices belonging to P and for all common
poverty lines z < ?92.58 per week. Both TIP curves lie well below the line of
maximum poverty (represented by a ray from the origin to a vertical intercept
of ?92.58 per week at p = 1), except in the very poorest percentile where the
greater incidence in 1988/89 of very small incomes is clearly apparent.
Table 2 confirms the ordering result, showing estimates for several members
of P. Interestingly the poverty ordering by indices in P is the opposite of that
provided by the headcount ratio (8.1% in 1979, 6.9% in 1988/89). Thus,
although there was a small fall in the proportion who were poor, poverty
gaps for those people who were poor became larger: for instance the average
per capita poverty gap increased from ?1.48 to ?1.73. The mean poverty gap
amongst the poor increased from ?18.22 to ?25.21, an increase of some 38%
compared to the increase in mean income of 28%. In sum, changes in the
intensity and inequality dimensions of aggregate poverty more than offset
changes in poverty Incidence between 1979 and 1988/89.
In fact the dominance is sufficiently strong that this poverty ordering would
be preserved for a range of intertemporal price deflators bounded by the base
case deflator and by one some 6.8% higher. In other words poverty is higher in
7 There is an issue to be resolved here concerning the degree of detail to which the comparisons
should be done. (This is related to issues of statistical inference. We defer study of these topics to
later work.) Our checks reveal that there is TIP dominance at each p. Figure 2 focuses on the lowest
percentiles because that is where is the 'action' is.
TABLE 2
Poverty in the UK, 1979, 1981, and 1988/89: a numerical summary
(poverty line half 1979 mean income)
1979 1981 1988/89
Mean income 185.16 183.79 236.11

Poverty line 92.58 92.58 92.58
(as percentage of mean income) (50.0) (50.4) (39.2)
Poverty indices
Headcount ratio [FGT(0)] 0.081 0.090 0.069
Mean poverty gap per capita 1.48 1.78 1.73
Mean poverty gap per poor person 18.2 19.7 25.2
FGT(1) x 1,000 15.9 19.2 18.7
FGT(2) x 1,000 7.0 9.6 10.8
FGT(3) x 1,000 4.6 6.9 8.3
Modified-Sen x 1,000 1.9 2.6 1.9
Watts x 1,000 25.2 35.7 42.3
Number in sample (persons) 18,123 20,380 36,587
Incomes expressed in April 1992 pounds per week. Source:

files (see text). Poverty indices described in Table 1 and tex
1988/89 even if real incomes in 1988/89 relative to those in 1979 were taken to
be almost 7% higher than previously assessed.
Our empirical results can also be seen from another perspective by interpret-
ing the robustness of the ordering in terms of variations in the poverty line for
the poorer distribution (rather than variations in the intertemporal price
deflator). It is common in the UK and Europe to use poverty lines defined
as fractions of contemporary means or medians, on the grounds that income
adequacy should be assessed relative to contemporary living standards. If we
were to identify poverty using half the contemporary mean as the low income
cut-off-a fixed relative rather than a fixed absolute real income standard as
before-the poverty lines for 1979 and 1988/89 would be ?92.58 and ?118.54
respectively.8 Now, conditional on the original intertemporal price deflator, we
find TIP dominance of normalized poverty gaps, showing that there was
greater poverty in 1988/89 than in 1979, in fact for all 1988/89 poverty lines
equal to or greater than 37% of the 1988/89 mean [0.37 z (1 - 0.068) (92.58/
236.11)], a range which includes half the 1988/89 mean as well as half the 1979
mean. Hence our conclusions about the increase in UK poverty between 1979
and 1988/89 are robust to whether or not an absolute or relative poverty line is
used, or indeed something in between.
We also find that the poverty in 1981 was higher than in 1979 according to
all poverty indices belonging to P and for all common poverty lines z < ?92.58
per week (TIP curve not shown). In 1981 the headcount ratio was 9.0%, the
8 The DSS's HBAI reports use both of these types of relative and absolute cut-offs. For recent
cross-national poverty comparisons also using both, see Blackburn (1994) and Hagenaars et al.
(1997).
I , I I I I I
2 Maxwnum oha,,. hissi
;,'1988/89 ' - _1.78

1.5 =; 1.73
0 .0
~0
.5
CL 0~~~~~~hlZ
1979aerg1 ncm) CL~~
0 .02 .04 .06 .069 .09 .1

Cumulative population share, p
FIG. 3. The 1988/89 TIP curve intersects the 1981 TIP curve once from above (poverty line =
half 1979 average income).
average per capita poverty gap, ?1.78, and the mean poverty gap amongst the
poor, ?19.74; i.e. the orderings by the headcount ratio and average poverty
gaps coincide in this case.
Figure 3 illustrates the comparison of the 1981 and 1988/89 distributions,
and shows that the 1988/89 TIP curve intersects the 1981 TIP curve once from
above. The average poverty gap (and headcount ratio) were larger in 1981, but
poverty gaps were less unequally distributed. Although we do not have a
poverty ordering according to all members of the classes P or Q, the results
of Jenkins and Lambert (1994) enable us to identify the sub-class of indices for
which there is one. Since FGT(2) was higher in 1988/89 than in 1981, we can
say for example that aggregate poverty was greater in 1988/89 than in 1981 for
all members of the Foster et al. (1984) class FGT(a) with a > 1, and for all
Clark et al. (1981) Type 2 poverty indices for which the parameter summarizing
aversion to income inequality amongst the poor is greater than 0.86. This
includes the Watts index (for which the aversion parameter equals one). See
Table 2 for calculations confirming these ordering results. Interestingly the
modified-Sen index, which is not a member of the sub-classes of P or Q
relevant in this case, provides the opposite ordering.
Further conclusions about UK poverty trends can be drawn using results
linking the Atkinson and Foster/Shorrocks poverty ordering methodology
with ours. For example, we find that the 1988/89 generalized Lorenz curve
lies below the 1979 curve up to p = 13.7% (corresponding to an income of
?103 in the 1979 distribution, ?108 in the 1988/89 distribution), and that the
1988/89 curve is everywhere above the 1979 one for all p > 13.7% (see Jen-
kins, 1994b, for details). Equivalently, we have generalized Lorenz dominance
of 1979 incomes over 1988/89 incomes for censored distributions in which all
incomes of ?103 and above are replaced by ?103. Hence, by Jenkins and
Lambert's (1996) Theorem 2, we can say that, according to all poverty indices
belonging to P, poverty is higher in 1988/89 than 1979 for all common poverty
lines up to ?103, i.e. 56% of average income in 1979. This result extends the
range of poverty lines for which we have TIP dominance beyond the popular
50% fraction which we used earlier.
5. Summary and conclusions
We have introduced a new construct, the 'Three "I"s of Poverty' (TIP) curve,
which provides a succinct and evocative graphical summary of the incidence,
intensity, and inequality dimensions of aggregate poverty. Comparisons of
TIP curves also provide a means for checking for unanimous poverty orderings
by a wide class of poverty indices, and for checking robustness of these order-
ings to variations in deflators used to render nominal income distributions
comparable (equivalently, to a change in the poverty line in one distribution).
This result has useful applications in poverty comparisons across time, across
regions and countries, and between population sub-groups of different house-
hold composition.
Our empirical analysis applies the TIP dominance criterion and reveals new
perspectives on UK poverty trends during the 1980s which are not apparent
from the official low income statistics. We have shown that for a range of
poverty lines encompassing ones commonly adopted in public debates, poverty
unambiguously increased between 1979 and 1988/89 according to all poverty
indices in the classes P and Q. Moreover in the case in which the poverty line is
fixed at half 1979 mean income, we have shown that the poverty ordering
would be preserved even if real incomes in 1988/89 relative to those in 1979
were anything up to 6.7% higher than previously assessed. The TIP method-
ology also highlights the changing features of poverty during recession and
boom years during the 1980s. The incidence of poverty was higher in 1981 than
1988/89, but there were also relatively fewer large poverty gaps. Thus whether
aggregate poverty was higher in 1981 than in 1988/89 depends in part on how
averse one is to income inequality amongst the poor.
ACKNOWLEDGEMENTS
We dedicate this paper to the memory of the late Aldi Hagenaars. Jenkins acknowledges financial
support from the Joseph Rowntree Foundation. We thank seminar participants and anonymous
referees, Frank Cowell, David T. Johnson, Martin Ravallion, and Tony Shorrocks for their com-
ments, and the DSS for the HBAI micro-data. Revision of the paper was completed with financial
support from EU Human Capital and Mobility Network 'Distribution and Redistribution of
Income' contract ERBCHRXCT940647.
REFERENCES
ATKINSON, A. B. (1987). 'On the Measurement of Poverty' Econometrica, 55, 749-64.

BLACKBURN, M. L. (1994). 'International Comparisons of Poverty', American Economic Review
(Papers and Proceedings), 84, 371-4.
CHAKRAVARTY, S. R. (1983). 'Ethically Flexible Measures of Poverty', Canadian Journal of
Economics, 16, 74-85.
CLARK, S., HEMMING, R., and ULPH, D. (1981). 'On Indices for the Measurement of Poverty',
Economic Journal, 91, 515-26.
DEPARTMENT OF SOCIAL SECURITY (DSS) (1992). Households Below Average Income. a Statistical
Analysis, 1979-1988/89, HMSO, London.
FOSTER, J. E., GREER, J., and THoRBEcKE, E. (1984). 'A Class of Decomposable Poverty Indices',
Econometrica, 52, 761-6.
FOSTER, J. E. and SHORROCKS, A. F. (1988a). 'Poverty Orderings and Welfare Dominance', Social
Choice and Welfare, 5, 179-98.
FOSTER, J. E. and SHORROCKS, A. F. (1988b). 'Poverty Orderings', Econometrica, 56, 173-8.
HAGENAARS, A. J. M. (1987). 'A Class of Poverty Indices', International Economic Review, 28, 58
607.
HAGENAARS, A. J. M., DE Vos, K., and ZAIDI, M. A. (1997). 'Patterns of Poverty in Europe,' in
S. P. Jenkins, A. Kapteyn, and B. van Praag (eds), The Distribution of Welfare and Household
Production: International Perspectives, Cambridge University Press, Cambridge, forthcoming.
HANNAH, L. and KAY, J. (1977). Concentration in Modern Industry, Macmillan, London.
JENKINS, S. P. (1991). 'Aggregation Issues in Earnings Discrimination Measurement', Economics
Discussion Paper No. 01/91, University of Bath.
JENKINS, S. P. (1994a). 'Earnings Discrimination Measurement: a Distributional Approach',
Journal of Econometrics, 61, 81-102.
JENKINS, S. P. (1994b). 'Winners and Losers: the UK Income Distribution in the 1980s', Economics
Discussion Paper No. 94-07 (Report to the Joseph Rowntree Foundation), University of
Wales, Swansea.
JENKINS, S. P. and LAMBERT, P. J. (1994). 'Ranking Poverty Gap Distributions', Economics
Discussion Paper No. 94-09, University of Wales, Swansea.
JENKINS, S. P. and LAMBERT, P. J. (1996). 'Three "I"s of Poverty Curves: TIPs for Poverty
Analysis', Economics Discussion Paper No. 96/50, University of York.
JOHNSON, D. T. (1988). 'The Measurement of Poverty in Australia: 1981-82 and 1985-86',
Australian Economic Review, 3rd quarter, 13-24.
PYATT, G. (1987). 'Measuring Welfare, Poverty and Inequality', Economic Journal, 97, 459-67.
SEN, A. K. (1976). 'Poverty: an Ordinal Approach to Measurement', Econometrica, 44, 219-31.
SEN, A. K. (1979). 'Issues in the Measurement of Poverty', Scandinavian Journal of Economics, 81,
285-307.
SHORROCKS, A. F. (1983). 'Ranking Income Distributions', Economica, 50, 3-17.
SHORROCKS, A. F. (1993). 'On the Measurement of Unemployment', Economics Discussion Paper
No. 418, University of Essex.
SHORROCKS, A. F. (1997). 'Deprivation Profiles and Deprivation Indices', in S. P. Jenkins, A.
Kapteyn, and B. van Praag (eds), The Distribution of Welfare and Household Production:
International Perspectives, Cambridge University Press, Cambridge, forthcoming.
SHORROCKS, A. F. (1995). 'Revisiting the Sen Poverty Index', Econometrica, 63, 1225-30.
THON, D. (1979). 'On Measuring Poverty', Review of Income and Wealth, 25, 429-39.
WATTS, H. (1968). 'An Economic Definition of Poverty', in D.P. Moynihan (ed.), On Under-
standing Poverty: Perspectives from the Social Sciences, Basic Books, New York, NY.
YITZHAKI, S. (1991). 'Necessary and Sufficient Conditions for Dominance Using Absolute Lorenz
Curves', unpublished paper, Hebrew University, Jerusalem.

Three I's of Poverty Curves PDF

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Three I's of Poverty Curves PDF

Uploaded by

Copyright:

Available Formats

Three 'I's of Poverty Curves, with an Analysis of UK Poverty Trends

Author(s): Stephen P. Jenkins and Peter J. Lambert

THREE 'I'S OF POVERTY CURVES, WITH AN

RECENT income inequality research has demonstrated the value of graphical

? Oxford University Press 1997

population sub-groups of different household composition. We illustrate this

Let x = (x1, x2,..., x") denote a distribution of income among n income-

and let b, be the associated vector

rxi= gxi/x = max {(z - xi)/z, 0} = max {1 - xi/z, 0} (3)

FIG. 1. The TIP curve and the 3 'I's of poverty.

Our picture of poverty is obtained by ranking people from poorest to riches

2 If a country's social assistance benefit programme provided a universal minimum income,

of un-normalised poverty gaps is equivalent to a unanimous poverty ordering

4. Empirical application: poverty in the UK

0 .02 .04 .06 .069 .081 .1

1979 1981 1988/89

Mean income 185.16 183.79 236.11

Incomes expressed in April 1992 pounds per week. Source:

;,'1988/89 ' - _1.78

0 .02 .04 .06 .069 .09 .1

5. Summary and conclusions

ATKINSON, A. B. (1987). 'On the Measurement of Poverty' Econometrica, 55, 749-64.

You might also like