C C | E E | D D | L L | A A | S S
Centro de Estudios
Distributivos, Laborales y Sociales
Maestría en Economía
Universidad Nacional de La Plata
Poverty among the Elderly in Latin America and the
Leonardo Gasparini, Javier Alejo, Francisco Haimovich,
Sergio Olivieri y Leopoldo Tornarolli
Documento de Trabajo Nro. 55
Background paper for the World Economic and Social Survey 2007
The World Ageing Situation
Poverty among the Elderly
in Latin America and the Caribbean *
Leonardo Gasparini **
Universidad Nacional de La Plata
This paper provides evidence on the incidence of poverty among the elderly
in Latin America and the Caribbean, based on household survey microdata
from 20 countries. The situation of older people is characterized in terms of
income, employment, education, health and access to services vis-à-vis the
rest of the population. The paper identifies the role played by the current
pension systems in Latin America, and assesses the efforts needed to achieve
substantial improvements toward the reduction of old-age poverty.
Keywords: elderly, ageing, poverty, Latin America, Caribbean
* We are very grateful to Ana Cortez, Robert Vos, Oliver Paddison, Marva Corley, Codrina Rada, and
Diane Horton for valuable comments and suggestions. We also thank Luis Lima, Georgina Pizzolitto,
Pablo Gluzmann, Ana Pacheco, Rocío Carbajal, Gimena Ferreyra, Luis Casanova, Carolina García
Domench, Ezequiel Molina, Adriana Conconi and Martín Guzmán for excellent research assistance. The
usual disclaimer applies.
** Corresponding author: firstname.lastname@example.org
*** CEDLAS is the Center for Distributional, Labor and Social Studies at Universidad Nacional de La
Plata (Argentina). Web page: www.depeco.econo.unlp.edu.ar/cedlas/
Poverty has a relevant age dimension. Both needs and income potential change over the
life cycle, modifying the probability of falling into poverty. This paper is focused on the
situation of the elderly relative to the rest of the population. In developed countries the
combination of strong social security systems, well-developed capital markets, and
small households contribute to higher living standards for the elderly, relative to the rest
of the population. These conditions are not replicated in many developing countries,
where pensions systems are weak and mostly favor the non-poor, the long-term formal
credit market is almost inexistent, and the elderly usually live in large extended
households sharing the budget with a large number of children.
Identifying the extent to which older persons are affected by poverty vis-à-vis the rest of
the population is essential to include the age dimension into social policy discussions.
Unfortunately, the task of measuring relative poverty across age groups is plagued by
methodological problems and data limitations. Moreover, these limitations do not bias
the results in only one direction: old age poverty may be higher or lower than what the
This paper is aimed at assessing the situation of the elderly in terms of income poverty
and other dimensions of well-being in Latin America and the Caribbean (LAC). The
evidence is drawn from a large database of household surveys from 20 LAC countries.
To our knowledge this is the first large-scale study that focuses on the poverty situation
of the elderly in Latin America based on a large comparable set of household surveys.
The rest of the paper is organized as follows. We start in section 2 by characterizing the
age structure of the population, and the household arrangements where older people
live. In addition we discuss the ageing process experienced by the region, and the
forecasts for the demographic structure of the LAC population. In section 3 we first
discuss poverty measurement issues, and then assess the incidence of poverty among
older persons in Latin America and the Caribbean under alternative proxies for
individual living standards. We compare our results to those found in other developing
regions of the world. While in section 3 we deal with income poverty, in section 4 we
enrich the analysis by including other dimensions of individual well-being: education,
health, access to the labor market and to basic infrastructure (water, sanitation, housing,
electricity). The role of the social security system is crucial in understanding old age
poverty. In section 5 we examine pension systems in Latin America and assess the
observed and potential effectiveness of pensions to reduce poverty. In section 6 we
carry out a set of microsimulation exercises in order to analyze the possible patterns
toward meeting the target of halving poverty for the elderly. In particular, we compute
isopoverty curves that show combinations of neutral growth and redistributive policies
toward the elderly capable of attaining the goal of halving old age poverty by year 2015.
In section 7 we take the ageing process as giving and carry out some simple
microsimulations to estimate its impact on national and old age poverty. Section 8
closes the paper with an assessment of the results and their policy implications toward
the aim of mitigating old age poverty.
2. The elderly in Latin America and the Caribbean
The population ageing process all over the world is a well-acknowledged fact. Latin
America and the Caribbean have not been the exception from this widespread
phenomenon. According to the United Nations World Population Prospect, the life
expectancy in the region will grow 55% between 1950 and 2050: a person who will be
born in 2050 will live 28 years more than a similar person who was born in 1950 (see
table 2.1). In fact, life expectancy has been growing in LAC at rates above the world
In LAC, as in the rest of the world, the gender gap in terms of life expectancy has
widened in favor of women in the last 50 years (from 3.4 years in 1950 to 6.7 years in
2000). That gap is expected to slightly shrink in the coming decades, due to a more
intense fall in the male mortality rate.
The fact that the world, and Latin America in particular, are ageing is clear from the last
panel of table 2.1. The median age of the world population has increased from 23.9 to
26.8 since 1950, and it is expected to grow to 37.8 by 2050. The speed of the ageing
process has been faster in Latin America compared to the rest of the world, and it is
expected to continue being faster in the following decades. In fact, while in 1950 the
average Latin-American was almost 4 years younger than the average person in the
world; in 2015 a typical inhabitant of Latin America will be 2 years older than the world
Another way to illustrate the ageing process is by dividing the population in age
brackets. We consider four groups: <15, 15-24, 25-59, and +60, and label the latter
group as the elderly. This definition, although entirely arbitrary, is useful for the
analysis, as any reasonable alternative definition not based only in age is almost
impossible to implement with the usual data at hand. We follow the general practice in
LAC to define the elderly as those aged 60 or more. In some sections of this document
we assess the robustness of the results to changes in that threshold.
The ageing process discussed above has implied a substantial increase in the share of
older people in the population (see figure 2.1). This pattern holds in every continent, but
it is particularly significant in Europe. In LAC the share of the elderly in the population
increased from around 6% in 1950 to more than 8% in 2000, while it is expected to
reach 24% at the end of the century. This ageing process implies an estimate of around
200 million people older than 60 in LAC by 2050.
Figure 2.2 shows that during the last 50 years the annual rate of population growth of
the LAC elderly has been higher than the corresponding rate for the younger age
brackets. The gap between them has widened since 1980. It is expected that this gap
will continue to enlarge during the first two decades of the new millennium (reaching a
value 4 times bigger than in 1950), and then probably will start shrinking (Figure 2.2).
The intensity of the population ageing process has been heterogeneous across LAC
countries. Figure 2.3 illustrates this heterogeneity by showing the annual growth rate of
the population ratio +60/<60 in each LAC country. That ratio has substantially
increased in Argentina, Venezuela and Cuba, while it almost has not changed in
Mexico, Costa Rica and Nicaragua. Only two LAC countries experienced a substantial
fall in the ratio +60/<60: Paraguay and Haiti.
The current (2005) and the estimated future (2015 and 2050) population share of the
elderly in each LAC country is displayed in figure 2.4. In all countries the share of the
elderly is expected to substantially grow in the coming decades. All LAC societies will
have to face the challenges related to an ageing society in the near future. However, as
this and the previous figures show, the intensity of these challenges will vary across
Socio-demographic characterization of the elderly
In order to get deeper into the analysis of the socio-economic situation of the elderly in
LAC we need to go beyond the basic demographic information included in Census, and
use microdata from household surveys. In the rest of the paper we present a socio-
economic characterization of older people in LAC based on a large database of
household surveys from 21 countries: the Socio-Economic Database for Latin America
and the Caribbean (SEDLAC), assembled by CEDLAS (Universidad Nacional de La
Plata) and the World Bank’s LAC poverty group (LCSPP). SEDLAC includes more
than 150 household surveys in 20 countries: Argentina, Bolivia, Brazil, Colombia,
Costa Rica, Chile, Dominican Republic, Ecuador, El Salvador, Guatemala, Haiti,
Honduras, Jamaica, Mexico, Nicaragua, Panama, Paraguay, Peru, Uruguay and
Venezuela. For this paper we select a sample of surveys corresponding to the latest
observation in each country (see table 2.2).
The sample covers all countries in mainland Latin America and three of the largest
countries in the Caribbean – Dominican Republic, Haiti and Jamaica. In each period the
sample of countries represents more than 92% of LAC total population. Most household
surveys included in the sample are nationally representative. The main two exceptions
are Argentina and Uruguay, where surveys cover only urban population, which
nonetheless represents more than 85% of the total population in both countries.
The population structure drawn from household survey microdata for each LAC country
in our sample is presented in table 2.3. On average, older people (60+) represent around
10% of total population. Figure 2.5 illustrates the heterogeneity within the region. While
older people in Guatemala and Nicaragua represent 6% of their total population, in
Uruguay and Argentina that share is 3 and 2 times greater, respectively.
The share of the elderly in the population is larger in rural areas than in cities (Figure
2.6). One possible reason behind this fact is that urban areas offer a wider range of labor
possibilities for younger people, which may encourage them to migrate into those areas
in order to improve their socio-economic situation.
Table 2.4 shows the population ratios between the elderly and the rest of the population
in each country. On average, the elderly are 32% of the children (<15). Figure 2.7
shows the heterogeneity within LAC. While in Uruguay the number of older people is
roughly the same as the number of children (<15), in Guatemala the proportion is 1
older people for around 8 children. As expected, the population ratio elderly/children is
greater for women than for men. In Uruguay, for instance, there are 15% more older
women (>60) than girls (<15), but there are 25% fewer older men than boys.
As expected from the differences in life expectancy shown above, the gender structure
differs by age group. In almost all countries the share of women among older people is
substantially larger than the corresponding share for the youth (figure 2.8). The average
masculinity index, defined as the ratio between the male population and the female
population, is 13% higher for the youth (0.97) than for the elderly (0.86) (table 2.5).
Older people tend to live in households of smaller size than younger people (table 2.6).
On average in LAC, the elderly live in households with 1.5 persons less than the rest of
the population. This gap varies from 1 person in Colombia and Venezuela to around 2
in Guatemala, Argentina and Bolivia. Even though the average family size in rural areas
is larger than in cities, we do not find significant differences within the older population
Table 2.7 helps us to learn on the type of households where the elderly live. On average
in Argentina a typical older people lives in a household with 1.37 older people
(counting herself), 0.77 adults, 0.26 youngsters and 0.28 children. There is not much
variation across countries in the number of people older than 60 living in households
with older people (from 1.40 in Peru to 1.25 in Nicaragua). Differences are sharp in
terms of children and adults. The average old person in Venezuela lives with 1.52
adults, while the average old Uruguayan lives with 0.64 adults. In rural areas the
average LAC old person lives with 20% more children than in cities.
Around a quarter of all LAC households are headed by an older person. Once again
there are dissimilarities within the region. For instance, older household heads in
Bolivia represent 17% of all heads, while in Argentina and Uruguay that proportion
goes up to 31% and 41%, respectively (table 2.8). In rural areas the share of older
household heads is higher than in urban areas (figure 2.10.b).
3. Old age poverty
In this section we provide evidence on the incidence of poverty among older persons in
Latin America and the Caribbean based on a large set of household surveys. Poverty is
certainly a multidimensional issue. However, in this section we restrict the concept of
poverty to that of income deprivation. In section 4 we extend the analysis to other
relevant variables as education, health, housing, water, sanitation, and labor market
An individual is considered as poor if her living standard indicator is lower than a given
threshold, known as the poverty line. The practical implementation of this definition
requires the choice of a proxy for the individual well-being and a poverty line. Most of
the economic literature suggests using household consumption adjusted for
demographics as the welfare variable, and a poverty line that combines a certain
threshold (largely arbitrary) in terms of consumption of calories, with the consumption
habits of the population, and the domestic prices of goods and services.1
Although household consumption is a better proxy for welfare than household income,
in this study we follow the literature in LAC and use income as the well-being indicator.
A simple reason justifies this practice: few countries in the region routinely conduct
national household surveys with consumption/expenditures-based questionnaires, while
all of them include questions on individual and household income.
The elements needed to construct a poverty line are idiosyncratic to each community, a
fact that leads to wide differences in the national lines across countries, and introduces
serious comparability problems. For this reason cross-country comparisons are usually
made in terms of some simple international line. The most popular one is the USD1-a-
day line proposed in Ravallion et al. (1991). It is a value measured in 1985 international
prices and adjusted to local currency using purchasing power parities (PPP) to take into
account local prices. The USD 1 standard was chosen as being representative of the
national poverty lines found among low-income countries. The line has been
recalculated in 1993 PPP terms at USD 1.0763 a day (Chen and Ravallion, 2001). The
USD-2-a-day line is also extensively used in comparisons across middle-income
countries, like most in LAC. Although the USD-1 or 2-a-day lines have been criticized,
their simplicity and the lack of reasonable and easy-to-implement alternatives have
made them the standard for international poverty comparisons.2 For instance, the United
Nations’ Millennium Development Goal 1 – eradicate extreme poverty and hunger – is
stated in terms of USD-1-a-day poverty – halving between 1990 and 2015 the
proportion of people whose income is less than USD 1 a day.
The measurement of poverty among the elderly poses some additional relevant
problems. The first one is related to the lack of consumption data. Some older people
may be living on the assets they accumulated during their lifetimes. The sale of an asset
is not usually included as current income, and then not considered in a poverty analysis.
While this could be the proper practice for, say, a young adult that sells his car to later
buy a new one, it might be incorrect for an older person who periodically sells assets to
keep his/her living standard.
1 See for instance Deaton and Zaidi (2003).
2 See Srinivasan (2004), Kakwani (2004) and Ravallion (2004) for a discussion on the merits and
demerits of the USD-1-a-day line.
An additional problem is posed by the fact that resources may be unevenly allocated
within households. The typical information included in an income-based household
survey does not allow identifying the specific allocation scheme adopted by each
household. For these reason the usual practice is to assume complete within-household
equality in living standards.
Another relevant problem arises from the fact that older people usually live in
households with a significantly different demographic structure than the rest of the
population, as documented in the previous section. That difference makes the poverty
comparisons between the elderly and the non-elderly population highly dependent on
the assumptions about the impact of the household structure on individual well-beings.
In particular, older people tend to live in households of smaller size, which impedes
them taking advantage of the household consumption economies of scale.
In summary, although we recognize that poverty is a multidimensional complex
problem, data limitations restrict this paper (and most of the literature) to simply
consider the poor as those individuals living in households whose per capita income is
lower than a certain international poverty line in terms of PPP dollars. Most researchers
and practitioners seem to agree that this is a reasonable approximation to a complex
problem. In this paper we use that widespread definition and assess the robustness of the
results to some methodological changes (economies of scale, adult equivalents and
We provide evidence on poverty by age groups for a sample of 20 LAC countries.
Evidence is drawn from microdata of the SEDLAC database described in section 2.
Even after agreeing on the income variable and the poverty line, a large number of
methodological problems should be solved to compute poverty in each country. Specific
details on methodological issues could be found in the SEDLAC web page.3
Poverty rates significantly differ across LAC countries. Table 3.1 shows the headcount
ratios for the USD2-a-day poverty line. While the share of persons with household per
capita income below that line is 5.1% in Chile, the share climbs to 78% in Haiti.
Poverty is substantially higher in rural areas.4
The correlation between national poverty and poverty in any age group is very high. For
instance, the linear correlation coefficient for the case of the elderly (older than 60) is
0.95. Figure 3.1 illustrates this close relationship. It is interesting to notice that most
points lie close but below the 45° line, implying lower poverty rates for the elderly
when compared to the rest of the population. That is the case in both urban and rural
areas. This piece of evidence does not imply that poverty is always decreasing in age. In
4 See Cicowiez et al. (2006) for evidence on the urban-rural differences.
fact when compared to the adult population in most countries poverty is higher for the
elderly (figure 3.2). Defining the elderly as those older than 60 or those older than 65
does not make a significant difference.
To further document the age-poverty profile in figure 3.3 we show non-parametric
(kernel) estimates of the poverty headcount ratio by age in each LAC country. The
curves are clearly downward sloped along all the age range for the set of Southern Cone
countries (Argentina, Brazil, Chile and Uruguay). In contrast, for the rest of the
countries poverty is clearly decreasing only until around the age of 40, and then
becomes either constant (e.g. Paraguay, El Salvador, Nicaragua), slightly increasing
(e.g. Bolivia, Ecuador, Venezuela) or substantially increasing (e.g. Colombia, Mexico).
For the South American countries with well-developed pensions systems poverty
reaches its minimum levels in the older age brackets (table 3.2). In Argentina and Chile
the poverty rate for those older than 60 is around a third of the poverty rate for the total
population. That proportion drops to 20% in Brazil and just 10% in Uruguay. In
contrast, in some other LAC countries old age poverty is more than 20% higher than the
national rates. That is the case of Jamaica and Mexico.5
The shape of the age-poverty profiles is surely dependent on factors like the extent of
the pension system and the age-education profile. We postpone a discussion on these
factors to first investigate another likely determinant of the poverty gaps by age: the
demographic structure of households.
The role of the demographic structure
So far, we have measured poverty using per capita income as the individual well-being
indicator. It has long been argued that needs differ across age groups and that
households can take advantage of their size by exploiting consumption economies of
scale (Deaton, 1997). These economies allow a couple to live with less than double the
budget of a person living alone.6 According to this approach individual well-being is
proxied by total household income deflated by an equivalence scale, defined as a
function of the size of the household and its demographic composition. There is a long-
standing literature on equivalence scales (see Deaton and Paxson, 1998). We follow the
approach of Buhmann et al. (1988) and Deaton and Paxson (1997) by assuming a
parametric form for the equivalence scale and examining the consequences of changing
the parameters. In particular, we assume that the living standard of an individual i living
in household h is given by
5 Notice that although the ratio in Costa Rica is high, the difference in poverty points is small, and even
probably not significant.
6 For instance two persons can save costs by living together and having one restroom to share.
where A is the number of adults, C1 the number of children under 5 years old, and C2
the number of children between 6 and 14.7 Parameters α allow for different weights for
adults and kids, while θ regulates the degree of household economies of scale. When
θ=1 there are no economies of scale, while in the other extreme when θ=0, there are full
economies of scale, meaning that all goods in the household could be shared completely
(i.e. they are all public goods, with no rivalry in consumption). In very underdeveloped
economies where people spend nearly all their income in food, there is no much scope
for economies of scale (a potato eaten by one member of the household cannot be eaten
by another member). In developed economies where a much larger share of the budget
is spent in housing, entertainment and other goods easier to share, consumption
economies of scale are more important. Following the suggestion of Deaton and Zaidi
(2002) for middle-income countries like those in LAC we take intermediate values of
the αs (α1=0.5 and α2=0.75) and θ (0.8) as the benchmark case.
To illustrate the adjustment for economies of scale, consider two households, labeled as
A and B, for simplicity comprised only by adults, with the same household per capita
income ($1000) but with different household size (2 persons in A and 5 persons in B).
Using θ=0.8 implies that, despite per capita income is the same in both households,
equivalized income is 20% higher in household B ($1380 in B, and $1149 in A).
Notice that in the same way as in the above example, countries where family
arrangements imply larger households can take advantage better of the consumption
economies of scale, even with a common parameter θ. In addition, one can assume or
estimate different parameters θ across countries based on different consumption budget
structures (see Deaton, 1997), but this is well beyond the scope of this paper.
In practice it is convenient to work with a transformation of the above equation to make
poverty estimates comparable to those obtained with household per capita income and
the USD-2-a-day line. The need for an adjustment comes from the fact that by deflating
instead of by just the number of family members
the indicator of individual welfare x
We alleviate (although not eliminate) this nuisance by following the procedure
suggested by Deaton and Paxson (1997), and multiplying the above equation by
, where C
children under 5, children between 6 and 14 and adults in the “base” household. We
take the average number of children and adults in each country to construct the base
ih increases, and then poverty estimates go down.
10, C20 and A0 are the number of
Table 3.3 shows older people relative poverty using four alternative income variables:
(i) per capita household income, (ii) household income per adult equivalent, (iii)
household income adjusted for economies of scale, and (iv) household income per adult
equivalent adjusted for economies of scale. The consideration of these demographic
7 Van Praag has suggested the possibility of using different weights for the elderly as their nutritional
needs may be lower than those of the adult population. The argument loses strength when expanding the
needs to other goods and services (e.g. health).
factors implies an increase in the relative poverty of the elderly. As seen in section 2
older people live in smaller households, and then they are not able to take advantage of
consumption economies of scale. Also, the increase in equivalized income after the
adjustment for the lower needs of children does not particularly favor the elderly, who
on average live in households with a smaller number of children.
Figure 3.4 illustrates the change in poverty when carrying out the adjustments. Relative
old age poverty significantly increases in Bolivia and Mexico when considering adult
equivalents and economies of scale. The effect goes in the same direction in the cases of
Argentina and Brazil, although the impact is quantitatively less relevant. This is not
surprising, given the smaller household size (and number of children) in Argentina and
Brazil, compared to Bolivia and Mexico.
The impact of considering different parameters for economies of scale is analyzed with
the help of Table 3.4 and Figure 3.5. As the parameter goes from 1 to 0 consumption
economies of scale internal to the household turn more important, and relative old age
poverty increases in all countries. In many countries the sign of the poverty comparison
between the elderly and the rest does not depend on the parameter of economies of scale
(given the adult equivalent scale used). For instance, in Bolivia old age poverty is
always higher than national poverty, while the opposite is true in Brazil, regardless of
the degree of economies of scale. In some other countries the sign of the difference
depends on the parameter: that is the case of Guatemala, El Salvador, Nicaragua,
Paraguay and Venezuela. The curves for other countries like Argentina and Chile also
cross the unity line, although they do so at improbable values of the parameter of
economies of scale.
Characterizing poverty-age profiles
As shown above, old age poverty substantially differs across LAC countries. Countries
are different not only in terms of total old age poverty, but what is more relevant for this
study, also in elderly poverty relative to the rest of the population. What are the factors
behind the country differences in poverty-age profiles? This question is important since
it helps to understand why in some countries old age poverty is not a particularly urging
problem, at least when compared to poverty for other age groups. Unfortunately,
disentangling the complex process leading to old age poverty, even in a single country,
is a very difficult topic that goes beyond the scope of this paper. Rather than attempting
econometric estimations that will face all sort of data and endogeneity problems, in this
section we just show some simple correlations that motivate the possible links between
certain characteristics of the economy and old age poverty.
For many reasons, for most people the income potential diminishes after a certain age,
and then income poverty is more likely to occur. Societies all around the world have
developed pension systems to shield older people against these risks. Old age poverty is
then expected to be highly correlated with the development of the pension system. The
first panel in figure 3.6 shows a simple scatter plot between relative old age poverty
(+60/-60) and the share of old people in the population receiving pension payments. The
linear correlation is -0.85 suggesting a strong positive relationship between both
variables.8 The relationship is driven by the presence of two clearly different set of
countries: those Southern Cone countries with a relatively-well developed pension
system where more than half of the population is covered (on average, 66%), and the
rest of LAC countries where on average only 14% of the elderly is covered. Within this
group the correlation poverty-pensions is not statistically significant.
Older people might be poorer just because they are less educated than the younger
generations. As will be documented in the next section, all LAC countries have
experienced an education upgrading process which implies that younger people are
more skilled and hence better prepared for the labor market. The second panel in figure
3.6 shows that there is a positive relationship between relative old age poverty and the
gap in years of education between the elderly and the adult population. The correlation
coefficient however is small and barely significant (0.27).
As argued above, the size of the household could be linked to the degree of income
poverty. The third panel of figure 3.6 shows the scatter plot of relative old age poverty
and the gap in household size between those older than 60 and the rest of the
population. In countries where that gap is large, that is where older people live in
households substantially smaller than younger people, relative old age poverty is lower.
However, this positive link is entirely driven by two countries, Argentina and Uruguay,
with low old age poverty and family arrangements such that a large fraction of the
elderly, many of whom receive pension payments, lives alone. The linear correlation
coefficient is 0.42; it falls to 0.34 when computing poverty with household income
adjusted for economies of scale and adult equivalents, and vanishes to zero when
deleting Argentina and Uruguay from the sample.
In a cross-country regression (with only 20 observations!) the coefficient of the size of
the pension system is always significant, even when controlling for education and
household size. In contrast, when controlling for the pension system the coefficients of
education and household size become non-significant. In summary, this preliminary
evidence suggests that there exists a strong negative relationship between relative old
age poverty and the development of the pension system. The evidence about the links
between old age poverty with education and household size is weaker.
Older people in the income distribution
Table 3.5 shows the distribution of people older than 60 across quintiles of the income
distribution. The elderly are over-represented in the top quintile of the household per
capita income distribution in all countries, except Jamaica. When considering the
distribution of equivalized household income (θ=0.8, α1=0.5, α2=0.75) the elderly
8 The correlation coefficient is -0.87 when computing relative poverty as +60/-15, and -0.85 when
computing poverty with household income adjusted for economies of scale and adult equivalents.
become under-represented in five countries. While in the first panel the share of old
people in the top quintile exceeds 25% in 15 countries, that number falls to 5 countries
in the second panel. When using equivalized income as the welfare indicator, in more
than half of the LAC countries the share of the elderly in the bottom quintile is larger
than 20%, implying over-representation of older people among the poorest.
Another way of showing the location of the elderly in the income distribution is through
concentration curves. Each curve shows the cumulative share of the elderly for the
poorest p percent of the population. Figure 3.7 shows these curves for a sample of
countries. If the curve lies above the diagonal (the perfect equality line) means that the
distribution of older people is biased toward the low-income strata. Suppose the
government implements a transfer of $1 to each old person. That policy will be pro-poor
(pro-rich) in those countries where the concentration curve lies above (below) the
Some results are worth mentioning. First, the curves do not locate too far from the
diagonal, meaning a not particularly biased distribution of the elderly in the population.
Second, there is not a homogeneous location of the concentration curves across LAC
economies. In some countries the curves lie below, in others above, and in others they
cross the diagonal.
Box 3.1: Income vs. consumption poverty. The case of Nicaragua
In this box we illustrate the differences between income and consumption poverty of the
elderly vis-à-vis the rest of the population, using the Living Standard Measurement
Survey of Nicaragua, 2001. This LSMS is one of the few Latin American surveys with
reliable information on both income and consumption. The following table shows the
ratio of poverty levels between age groups using the two alternative indicators of well-
Source: own calculations based on the EMNV 2001.
Notice that old age poverty relative to the rest of the population is lower when measured
with consumption rather than income. Figure B3.1 shows that while when measured
with income, poverty slightly increases for the elderly (with respect to adults), it
actually goes mildly down when measured with consumption. As expected old age
poverty is a less worrisome problem when measured with consumption data.
Box 3.2 Subjective poverty and the elderly. The case of Colombia
An alternative approach to determine whether a person achieves a minimum standard of
living consists in asking if they consider themselves to be poor. It is interesting to study
whether subjective poverty is higher among the elderly, independently of objective
measures of deprivation.
Colombia’s Encuesta de Condiciones de Vida asks household heads (or their spouses)
whether they would rate themselves as poor. Figure B3.2 illustrates the relationship
between self-assessment of welfare and age. As people age, the negative perception of
their economic well beings tends to slightly increase.
Whereas on average around 66% of people aged 25 to 59 consider themselves as poor,
that share increases to 70.3% for people older than 60 (table B3.1). Notice that around
90% of the elderly living in rural areas are poor under this subjective measure.
For people older than 25, we estimate a basic probit model for the probability of being
poor according to the subjective perceptions of individuals. The set of control covariates
includes two age dummies (“old people” is the omitted category), a gender dummy, a
set of educational dummies, an urban dummy, household size and dummies regarding
labor status. As table B3.2 shows, once controlling by observable characteristics the
conclusions are different. The likelihood of rating oneself as poor is not significantly
different for the elderly and adults aged 50 to 59. Moreover, individuals between 25 and
49 years old are more likely to be poor according to this approach than old people. The
higher non-conditional likelihood of being poor of the elderly seems to be due to
differences in other observable characteristics, like educational levels. The other
estimated coefficients, in general, show the expected sign.
Is income inequality higher among the elderly? The answer seems to depend again on
the relevance of the pension system in each country. Table 3.6 shows that the Gini
coefficient for the income distribution among the elderly is lower than for the rest of the
population in Argentina, Brazil, Chile and Uruguay. The results are robust to the change
in the individual well-being indicator.
The economic literature has discussed whether ageing societies tend to be less unequal.
There is a strong presumption in favor of more equal economies, at least in terms of
incomes, in ageing societies with well-developed pension systems. The serious analysis
of the interplay between the demographic structure of the population and the income
distribution is beyond the scope of this paper. As an exploratory analysis we present in
figure 3.8 a simple scatter plot across LAC countries between mean age and the Gini
coefficient for the distribution of per capita income. The seemingly negative correlation
is driven by Uruguay. As soon as we delete that observation, the negative correlation
vanishes (the linear correlation coefficient becomes non-significant). Similar results Download full-text
arise when using the share of older people in the population instead of mean age. At
least in the context of LAC where pension systems are poorly developed, the equalizing
effect of an ageing society does not show up, at least at a first glance.
Old age poverty in the developing world
The evidence on relative old age poverty in the developing world is still too scarce and
non-systematic to identify a clear pattern. Comparisons across studies are mined by all
sort of methodological problems, arising from the choice of different poverty lines,
different measures of well-being, and different definitions of later life (Barrientos et al.,
2003). But even within a specific study patterns are not easy to identify. As we have
found for the case of Latin America, other studies report that in other regions of the
developing world while in some countries old age poverty is lower, in others it is higher
than national poverty. Moreover, the results of these comparisons are affected by the
assumptions on economies of scale and adult equivalents (Lanjouw et al., 1998). In
contrast to the mixed results for the developing world, most studies find that in
advanced economies poverty is significantly lower among older people (Whitehouse,
In table 3.7 we reproduce some results of previous studies on developing countries.
Deaton and Paxson (1997) conducted a detailed analysis of old age poverty for
countries of different regions, while Grootaert and Braithwaite (1998) and Lanjouw et
al. (1998) use information from the Household Expenditure and Income Data for
Transition Economies. In none of these studies a clear pattern for old age poverty arises.
In a recent study Kakwani and Subbarao (2005) find that while poverty is higher among
households with older persons (particularly in rural areas) in Malawi, Uganda and
Zambia, this is not the case in Madagascar, Mozambique and Nigeria, where children
were assessed to be in worse situation. Based on these pieces of evidence Barrientos et
al. (2003) conclude that “poverty in later life broadly reflects aggregate poverty”. This
conclusion seems correct on average, but does not apply to many countries, as table 3.7
and the LAC evidence shown in this paper suggest.
4. The socioeconomic situation of the elderly
In this section other dimensions of well-being are explored. So far, only income
deprivation has been taken into consideration. However, well-being is a multi-
dimensional concept. Clearly, variables such as health, education, basic infrastructure
and security affect the quality of life. These variables have a positive correlation with
income, but the correlation is far from being perfect, due in part to the impossibility of
buying some attributes of well-being. Another well-known difficulty that reinforces the
necessity of examining other dimensions of individual welfare is the biases resulting
from measuring poverty with current income (as opposed to permanent income). As