Luxembourg Income Study
Working Paper Series
Luxembourg Income Study (LIS), asbl
Working Paper No. 557
Policies To Reduce Child Poverty:
Child Allowances VS. Tax Exemptions For Children
THE MIDDLE CLASS IN LESS DEVELOPED AMERICAN NATIONS
Department of Economics and Finance
West Long Branch, NJ 07764
The purpose of this paper is to extend my previous work on how the middle class fares
throughout the world. Pressman (2007, 2010) provided a working definition of ‘the
middle class’ as well as estimates of the size of the middle class in several nations,
mainly developed Western countries. These papers argued that differences in the fraction
of households that are middle class, across nations and over time, were mainly due to
government tax and spending policies. The more progressive the national tax system, and
the more generous and more extensive government spending programs, the larger the
proportion of middle-class households in the nation.
This work was done using the Luxembourg Income Study (LIS) Database, an
international database containing extensive income as well as socio-demographic
LIS databases center around particular years, called ‘waves’. Each wave is
around five years apart, with Wave #1 beginning in the early 1980s. The most recent
data, Wave #6, covers the mid 2000s.
Until recently, Mexico was the only less developed American country in the LIS
database. This changed during the summer of 2009, when the LIS released Wave #6
http://www.lisproject.org/techdoc.htm (multiple countries; accessed 29 August 2009
through 28 August 2010).Those interested in more information about the LIS Database
can consult a number of excellent summaries (Smeeding et al. 1985; Smeeding,
Rainwater, Buhmann & Schmays 1988) or the LIS homepage at www.lisproject.org
databases for five Latin American nations (Brazil, Columbia, Guatemala, Peru and
Uruguay). In conjunction with previously available Wave #6 data for Mexico, these new
LIS datasets provide a unique opportunity to assess the size of the middle class in less
developed American countries, and to compare the size of the middle class in these
countries with the size of the middle class in developed nations.
This issue is important for a number of reasons. Aristotle (1932: Book IV) first
noted that political communities with a large middle class would likely be well governed
and society would not be dominated by either of the income extremes. More recently,
Thurow (1985) argued that “a healthy middle class is necessary to have a healthy
democracy” because social unrest increases when incomes and people become polarized.
Barro (1999) provides some empirical support for this view – countries are more likely to
be democratic the higher the share of income going to middle-class households—
although he admits that the causal connection can run in both directions.
There are also psychological reasons why a large middle class is important.
Attaining a middle-class living standard comes with feelings of success and personal
accomplishment. As Malthus (1803: 594) first pointed out in his second Essay on
Population, “Our best grounded expectations of an increase in the happiness of the mass
of human society are founded in the prospect of an increase in the relative proportions of
the middle parts.”
Psychological optimism is likely to spill over to economic optimism, leading to
more consumption, more investment and more rapid economic growth. In addition, just
having more money go to the middle class should lead to greater consumption and more
growth, since the poor do have money to spend and the wealthy have too much money to
spend. Landes (1998: 217-21) identifies a rising middle class as being the main reason
Britain was the first country in the world to industrialize. Taking a broader perspective,
Adelman & Morris (1967) argue that a growing middle class was the driving force behind
economic development in all of Western Europe.
A robust middle class is likely to be just as important for the less developed
American nations as it was for Europe. Estache & Leipziger (2009) argue that income
distribution is more important for less developed countries because a rising middle class
leads to greater economic development, and not just economic growth, and that it also
provides a driving force for democracy and political support for economic policies that
will aid both middle-class and low-income households. Easterly (2001) provides some
empirical support for this view. Furthermore, it is fairly well-known that income
distribution in Latin America is the most unequal in the world (Galbraith 2002;
Krozeniewicz & Smith 2000), and government spending programs there tend to be rather
regressive in nature (see Goldstein & Estache 2009).
The remainder of this paper measures the size of the middle class in some less
developed American countries during the mid 2000s and seeks to identify some
determinants of the size of the middle class in these nations. It begins by defining ‘the
middle class,’ and estimating the size of the middle class in less developed American
nations. It then goes on to assess the role of socio-demographic factors, government fiscal
policies, and a few labor market factors in the development of a large middle class in
these countries. The Appendix contains information on the particular year and the
original national survey for each LIS dataset used in this paper.
2. WHO IS MIDDLE CLASS?
Unfortunately, there is no official economic definition of ‘middle class’, and no definition
that most social scientists seem willing to accept. This paper employs a suggestion
advanced by Birdsall et al. (2000), Pressman (2007, 2010), and Thurow (1985), where
middle-class households are those households with incomes in the middle of the national
income distribution. In particular, a household is considered middle class if and only if its
adjusted household disposable income falls between 75 and 125 percent of median
adjusted household disposable income. Such a relative definition of ‘middle class’ is
justified by numerous studies showing that after a certain level of income is achieved,
relative incomes matter most to people (Frank 2007; Layard 2005; Luttmer 2005), and by
the work of many scholars who have argued that people are social beings (Fuchs 1967;
Easterlin 2001; Sen, 1999: 71) continually comparing themselves to others. This work
implies that we need relative definitions of ‘the middle class,’ ‘poverty’ and ‘human well-
being’ rather than absolute definitions of these terms.
In order to measure economic well-being we need to adjust household income to
account for differences in household size. For example, an income of $24,000 can
support a single individual in the US reasonably well. In 2009, it would have provided
more than twice the poverty-level income for a single person. But for a family of 5, an
income of $24,000 provides each person with just $4,800 on average. This cannot support
the same lifestyle as $24,000 for a single individual; in fact, according to the US Census
Bureau, a family of five would have been considered poor with this income in 2009.
One way to deal with household size differences is to treat the income needs of all
household members the same and look at per capita household income. But this ignores
economies of scale in living arrangements. Two people can live more cheaply together
than separately and will have a higher standard of living than two single individuals with
the same combined income. What we need is some middle ground between assuming no
economies of scale in living arrangements and assuming that household size does matter
at all for household living standards.
In what follows, we adjust household incomes using the OECD (1982)
recommendations regarding equivalence scales for household size. According to this
standard, income requirements for children are 50 percent of the requirements of the
household head, and income needs for additional adults in the household are 70 percent
of the requirements of the household head. These are pretty close to the implicit
household adjustments or equivalence scale in the set of poverty lines developed by
Mollie Orshanksy (1965, 1969) when she established the official US poverty lines for
different household sizes. Since the Orshansky poverty lines came from surveys of food
consumption and expenditures for different households, this approach provides a good
empirical foundation for using the OECD standards when adjusting income to account for
household size. However, not a great deal depends on this decision. Other adjustment
formulae have been suggested and tested, and studies have found that this decision makes
little difference to the broad results that one gets when using the LIS (Smeeding,
Buhmann & Rainwater 1988); obviously, though, the actual figures will differ.
3. THE SIZE OF THE MIDDLE CLASS IN LATIN AMERICA
Table #1 provides information on the size of the middle class for our sample of Latin
American countries. Looking at our main definition (adjusted household disposable
income between 75% and 125% of median adjusted household income), 21% of
households can be regarded as middle class. Uruguay, by far, has the largest middle class,
with nearly 27% of all households falling into this category. This is followed by
Columbia at 22.6%. The rest of our countries are bundled around 19%. By means of
contrast, for Wave #6 LIS databases, the middle class in developed countries averages
around twice as much—40% of all households. And in the Scandinavian nations of
Denmark, Norway and Sweden, the middle class approaches 50% of the entire
population. For the other two American nations, Canada and the US, Pressman (2010)
estimates the size of the middle class at 35% and 29%, respectively.
The figures in Table #1 are roughly consistent with what we know from other
sources about these countries and their distribution of income. The LIS summary
statistics of income distribution for the six nations in Table #1,
ranks Uruguay as having
(by far) the most equal distribution on various distribution measures, including the Gini
coefficient and the Atkinson coefficient, and the ratio of income received by someone at
percentile relative to someone at the 50
percentile of the income distribution. At
the other extreme, it is generally recognized (and confirmed by LIS summary statistics)
that Brazil has one of the most unequal distributions of income in Latin America and the
entire world; and, as Table #1 shows, Brazil has the smallest middle class of our six
Because there is no generally accepted definition of ‘the middle class’, and
because the choice of any income range will be arbitrary to some extent, Table #1 also
provides a sensitivity analysis of our estimates. The latter columns employ three
additional income-related definitions of ‘middle class’-- (D2) household adjusted income
between 75 percent and 150 percent of median adjusted disposable income, (D3)
These are available from the LIS website at www.lisproject.org.
household adjusted income between 75 percent and 175 percent of median adjusted
disposable income, and (D4) household income between 75 and 200 percent of median
adjusted disposable income.
Definition D2 is noteworthy because, in a recent analysis of the middle class in
the US, the Pew Research Center (2008) maintained this was the most appropriate range
for identifying middle-income households and used this range throughout its study of the
middle class. Definition D4 is noteworthy because it yields an income range for the
middle class close to what US households actually report when asked about the income
levels necessary to put a family of four in the middle class (Pew Research Center 2008;
Cashell 2008). There are some drawbacks to D4, however. Including those with 200% of
median income as part of the middle class means that only around 3 percent of the US
population gets classified as wealthy or upper class. Moreover, this upper limit is
substantially above what most scholars regard as providing middle-class income levels
and lifestyles. Nonetheless, D4 still reflects popular views of the middle class; and it may
make more sense to use this definition for the less developed nations of the Americas. On
D4, the upper class constitutes 10% to 20% of households in the majority of our six
nations, and for Guatemala and Peru the upper class approaches 25% of all households.
Finally, definition D3 is included as a midway point between the Pew suggestion and the
range suggested by empirical surveys in the US.
The good news is that these definitional issues do not make much substantive
difference. Pressman (2010) shows that the actual income range for defining ‘middle
class’ matters very little for developed nations, and any decision about where to draw the
line does not drive the empirical results obtained from adopting any one income range.
Of course, the estimated size of the middle class increases as we include larger and larger
income ranges in our definition. However, the size of the middle class varies across
nations in the same pattern for all income ranges. In the developed world, countries with
a relatively large middle class on our main definition also have a relatively large middle
class on all our alternative definitions. The Scandinavian nations always have the largest
middle class; the Anglo-Saxon countries always do worst; and continental European
nations (France, Germany, Italy and Luxembourg) fall in the middle.
For less developed American countries our results are similar. Under each of our
definitions, Uruguay does best by a substantial amount, and Columbia always does the
next best in terms of the size of the national middle class. At the other extreme, Brazil
always does the worst of our six countries. And in each case, Guatemala, Mexico and
Peru are always bunched together in the middle.
This shows that our main results do not seem to depend on definitional issues. For
most reasonable income ranges, countries with a relatively large middle class on one
definition also have a relatively large middle class on other definitions. Countries that do
poorly on one definition also do poorly on other definitions.
The next three sections address some of the main factors that might lead to a
relatively large middle class in some countries and a relatively small middle class in
others. First we examine some possible socio-demographic determinants of the middle
class. Then we look at government fiscal policy and how it affects the size of the middle
class. Finally, we examine labor market factors that work to promote or hinder the middle
3. SOCIO-DEMOGRAPHICS AND THE MIDDLE CLASS
Table #2 presents estimates on the size of the middle class for our less developed
American nations based on a number of socio-demographic characteristics of households.
For a number of reasons, various socio-demographic factors might explain changing
income distribution and changes in the size of the middle class.
It is well-known that income (on average) depends on age. Young, inexperienced
workers, typically get paid less than other workers. Income rises with age and tends to
peak at around age 50; shortly thereafter it tends to fall with age. Consequently, if the age
distribution of the working population changes, this should lead to changes in income
distribution and possibly to changes in the size of the middle class. If a large fraction of
households are in their prime earning years, income equality should increase and there
should be more middle-class households; on the other hand, if there are large fractions of
households headed by someone who is young or old, income inequality should increase
and there should be fewer middle-class households.
Living arrangements might also affect the size of the middle class. Since women
earn less than men, a larger fraction of female-headed households should increase income
inequality and reduce the size of the middle class. On the other hand, more married
couples should increase the overall percentage of middle-class households. However,
where child-rearing responsibilities limit employment possibilities, women may not be
able to work as much as they can in households without children. This should make it
more difficult to obtain middle-class status.
Finally, economic development is usually associated with a rising middle class.
Historically, a good fraction of this stems from rural to urban migration. Households
engaged in subsistence agriculture move to large cities where higher paying jobs are
available. As households start to move to urban areas, income inequality begins to
increase due to rising incomes in urban areas.
To examine the impact of socio-demographic factors on the size of the middle
class we look at several key sub-groups of the entire population. In order to make these
comparisons easy to follow, we begin by repeating the main results from Table #1—the
percentage of all households that are middle class.
Table #2 shows that, with just a few exceptions, socio-demographic factors do not
seem to affect the size of the middle class. The percentage of middle-class elderly
households (those whose head is greater than 64 years old), the percentage of middle-
class prime-age households (whose head is between 35 and 50 years old), the percentage
of middle-class female-headed households, the percentage of middle-class married
couples, the percentage of middle-class households with children and the percentage of
middle-class households living in urban areas are not very different from the aggregate
figures presented in column 1. These results are consistent with the results for developed
nations—socio-demographic factors seem to matter very little for the size of the middle
class in most cases (Pressman 2007, 2010). In the developed world, there are very few
countries where socio-demographic characteristics seem to be a factor in determining
middle-class status, and few where demographic changes can explain a large portion of
the changing size of the middle class.
However, two differences between developed Western nations and Latin
American nations are worth noting. First, in contrast to the developed Western nations,
where households with children are more likely to be middle class (see Pressman 2010),
in less developed American nations, households with children are just as likely to be
middle class as all households. One possible explanation for this, which we will explore
later, is that government policy helps families with children more in developed nations.
Second, Table #2 shows that elderly Latin American households are more likely to be
middle class in Brazil and Uruguay, but are less likely to be middle class in Mexico and
Peru. This, too, may be due to government policy—a generous public pension system in
some nations but not in others.
Another socio-demographic factor that might affect the size of the middle class
concerns marriage propensities; in particular, whether a household is middle class may
depend on wife earnings and their income level relative to the income of their husband.
The argument here is rather simple and straightforward. If high-income men become
more likely to marry high-income women, the result will be increased inequality and
fewer middle-class households since the marriage of two people who each earn a good
salary is likely to push them into the upper class. This has become a rather contentious
issue among scholars looking at the impact of wife earnings on income inequality in the
US (Burtless 1999; Cancian & Reed 1998, 2001; Esping-Andersen 2007). Several studies
have found an increasing correlation between husband and wife earnings over time in the
US (Shorrocks 1983; Lerman & Yitzhaki 1985; Karoly & Burtless 1995). This means
that high-wage men are more likely to marry high-earning women, which increases
overall inequality. On the other hand, Gottschalk & Danziger (2005) show that changes
in female hours of work and female earnings inequality over time have made income
distribution more equal in the US, while Cancian & Schoeni (1999) found that wives
earnings tend to mitigate inequality among married-couple families in most developed
countries. In perhaps the best summary of this diverse literature, Cancian & Reed (1998)
argue that whether wives’ earnings have either increased or decreased equality in the US
or had no impact on income inequality-- depending on how we do the measuring.
Table #3 takes a first stab at examining this issue for less developed nations. It
looks at how the size of the middle class is affected by wife earnings in some less
developed nations. To analyze this we start by calculating the percentage of non-elderly,
husband-wife families that are middle class. Then we subtract wives’ earnings from
household income and recalculate the percentage of middle-class, married-couple
families that are middle class.
Based on this exercise, wives’ earnings appear to have little impact on the size of
the middle class in our sample of less developed American nations. In several nations
(Columbia, Guatemala and Peru) wives’ earnings actually lower the size of the middle
class. In these cases, wives’ earnings are more likely to push incomes up and push
households into the upper income category than they bring low-income households into
the middle class. For comparative purposes, at the bottom of Table #3, we add data for
the two developed American nations—Canada and the US. The US is unique here in that
wives’ earnings seem to matter a great deal for middle class status. In fact, without these
earnings, the US would look a lot like our less developed American nations in terms of
the size of middle-class married couples. Despite having a much larger middle class than
the other countries in Table #3, Canada looks more like Latin American nations in one
respect—wives earnings do little to bring household income to middle-class levels. A
good part of the reason for this may be the availability of government benefits in Canada
that support married couples. We now turn to this issue.
4. GOVERNMENT POLICY AND THE MIDDLE CLASS
Pressman (2007) argued that government fiscal policy is an important determinant of the
size of the middle class in all developed nations; Pressman (2010) showed that two
specific policies (family allowances and family leave policies) significantly increase the
size of the middle class in these nations and that this is especially true of households with
children. Family allowances are annual payments to households for their children and are
meant help households support their children. Virtually every country in the world has
some sort of child or family allowance policy (Macinol 1980; Vadakin 1958, 1968).
Family leave policies include payments to prospective mothers while on leave from their
regular job, birth payments, and payments to parents of newborns so they can stay home
and care for their child. Most developed nations have some sort of family leave policy;
these policies are less prevalent in less developed nations and also less generous.
These are just two of the many fiscal policies employed throughout the developed
world to aid families and so that they can enjoy the benefits of a middle-class lifestyle.
Unemployment benefits and disability insurance make sure that short-run economic
problems or long-run health problem do not result in abject poverty for the household.
Government pension programs were developed to keep elderly households out of poverty
by forcing them to save during their working years and providing them with a fixed
income during retirement years. These payments allow older households to continue
living a middle-class existence after retirement. Finally, a progressive tax system,
especially one where low-income households pay negative taxes, also helps to support
middle-class incomes and lifestyles during difficult times. This section examines these
government policies in the context of less developed American countries.
To do this, we follow the procedure employed when examining the impact of
wives’ earnings on the size of the middle class. First, we look first at the size of the
middle class for a selected sub-group of the population. Then we subtract government
income sources from household income and recalculate the percentage of households that
are middle class.
There is, of course, some question here as to whether additional income from the
government affects individual behavior. For example, child allowances may encourage
households to have more children. Since larger households are less likely to be middle
class, by ignoring behavioral changes we have overestimated the size of the middle class.
For developed countries, several studies have found very small and insignificant
behavioral changes from this policy (Ermisch 1988a, 1988b; Gauthier & Hatzius 1997).
In addition, there may be macroeconomic effects of government policies that counteract
any behavioral changes. If child allowances increase the size of the middle class and
thereby household expenditures, it should expand the economy, which will further
increase the size of the middle class. For this reason, the computations below assume no
behavioral changes. To the extent that government policies do lead to behavioral changes,
and to the extent that these changes exceed the income effects of these policies, our
estimates of the impact these government policies have on the middle class will be biased
What is true of child allowances is also true of government pensions and
unemployment insurance. These benefits may encourage people to collect benefits rather
than work. Again, we ignore these possible effects, assuming them to be close to zero.
Such an assumption is quite reasonable for less developed nations since benefit levels are
generally smaller and the conditions for receiving benefits are more restrictive. And, as
with child allowances, unemployment insurance and government pensions should
increase aggregate expenditures, economic growth, and thus the size of the middle class.
This macroeconomic effect or income effect should counter any microeconomic
substitution effects of these government policies.
Table #4 shows that in most Latin American countries government policies are
rather ineffective in developing a national middle class among its targeted beneficiaries.
Child allowances and family leave policies increase the percentage of households with
children that are middle class by just a fraction of a percentage point in Brazil and
Uruguay, the only two countries for which we have such data in the LIS. In contrast,
Pressman (2010) finds that, in developed nations child allowances increase the fraction of
households with children that are middle class in developed nations by close to 5
percentage points in those countries with such a policy. But in the Americas, these
policies provide little support and are relatively ineffective in helping families achieve
middle-class status. In Canada they increase the size of the middle class by 2.1
percentage points; in the US, the increase is zero since the US has no family allowance
program. The bottom of Table #4 takes into account family leave policies as well as child
allowances in Canada and the US. This addition does little to change the results from
family allowances alone. From the first columns of Table #4, and from similar data for
developed countries, we can conclude that one reason Latin America has a smaller
middle class than developed countries is their meager child allowance programs.
Unemployment insurance and disability insurance do no better than family
allowances and family leave policies in the three nations for which we have LIS data. In
general, unemployment and disability payments are small in Latin American nations and
many households (especially those working in the informal sector) are ineligible for such
payments (ECLAC 2005; Lindert et al. 2006). As a result, the increase in the fraction of
middle-class households due to these policies is just a small fraction of a percentage
The bottom rows of Table #4 compare our Latin American nations to the
developed American nations. Unemployment and disability insurance in Canada (a
relatively stingy country when it comes to government benefits) increases the size of the
middle class by 2 percentage points. The US, the stingiest of all major developed nations
when it comes to these issues, does only a bit better than the three Latin American
nations we have data for. Thus, it seems that a second reason for a small middle class in
Latin America is their meager unemployment and disability programs.
Only in the case of pensions does government policy in Latin America help
somewhat. Pensions increase the size of the elderly middle class in several countries, and
in some cases (Brazil and Uruguay) significantly so. However, it should be noted that in a
few cases (Columbia and Guatemala), pensions actually reduce the size of the middle
class. In addition, the differences from country to country are quite large. A good part of
these cross-national differences probably stem from national differences in the percentage
of the labor force employed in informal sectors that are not covered by old-age pensions.
Overall these results are consistent with a large literature that finds that the bulk of social
expenditures in Latin America, especially pensions, are ineffective in protecting the poor
and favor those households in the two upper income quintiles (ECLAC 2005; Lindert et
al. 2006). For example, according to this literature, most social security expenditures go
to the top two quintiles—80 percent in Columbia, and 50 percent in Brazil and Uruguay
(ECLAC 2005: 144-6).
Again, the contrast with developed nations is large and significant. For the two
other American nations in the LIS database (Canada and the US), pensions are
significantly more effective in increasing the fraction of middle-class elderly households.
In contrast, to an average increase of 6.2 percentage points in our less developed
American countries, pensions increase the fraction of the elderly middle class by almost
16 percentage points in the US and by nearly 26 percentage points in Canada. And it
should be remembered that government policy in Canada and the US does relatively
worse in promoting and sustaining a middle class compared to other Western developed
nations (see Pressman 2010).
Finally, we look at the role of tax policy on the middle class. Again, there are
important limitations to performing such an analysis. Half the LIS datasets for our less
developed American nations are in net terms (with taxes already taken out), and so we
have no information about the taxes paid by different households. Table #5 provides
information on the impact of taxes on the size of the middle class for all datasets with
information on tax payments by households along with comparable estimates for Canada
and the US.
As this table shows, tax policy in Brazil, Columbia and Guatemala does little to
change the size of the middle class in these nations. The change in the size of the middle
class is just a tenth or so of a percentage point, essentially zero. The contrast with Canada
and the US here is large. In the US, tax policy increases the size of the middle class by
more than 5 percentage points. In Canada, the increase is more than 8 percentage points.
For a larger sample of mainly developed nations, Pressman (2010) found that tax policy
(income and mandatory social insurance taxes) increased the size of the middle class by
7.5 percentage points. Mainly, tax policy worked by reducing the income of wealthy
households and bringing them into the middle class. But, it also needs to be remembered
that the money obtained through a progressive tax system was used to raise the income of
other households and provide middle-class incomes to many households. So the overall
impact of taxes on the size of the middle class for developed nations is much greater than
what is reported in Table #5.
5. LABOR MARKET FACTORS AND THE SIZE OF THE MIDDLE CLASS
Just as socio-demographic factors and government policy might affect the size of the
middle class, so too can labor market forces impact the size of the middle class. This
section examines how a few labor market variables affect the size of the middle class.
Several studies carried out with US data have shown the size of the middle class
to vary with the business cycle (e.g., Horrigan & Haugen 1988). If unemployment is a
key determinant of the size of the middle class, we would expect that rising
unemployment in one nation would lead to a fall in the size of the middle class as many
households lose employment and lose income. Pressman (2007: 190-91) examined this
question at the macroeconomic level and found that changes in the national
unemployment rate had little correlation with changes in the size of the national middle
class. One possible reason for this result is that unemployment benefits make up for
income lost due to unemployment and keep households in the middle class. Another
possible reason is that broad macroeconomic variables, such as the national
unemployment rate, cannot capture what is going on at the microeconomic level.
Pressman (2010) examined this issue at a more microeconomic level. Using LIS
microdata on the number of earners, he employed shift-share analysis to ask how much
the middle class changes as a result of changes in the number of earners per household
between Wave #5 and Wave #6. The number of earners per household should change
with macroeconomic conditions, as more adults find employment and older children are
able to find part-time jobs; both changes will augment household income. This analysis
found a very small increase in the size of the middle class in Luxembourg and a small
decline in the middle class in Taiwan due to increases in the number of earners over time,
but little change overall due to the number of earners per household in each country.
Since the new LIS datasets from Latin Americas provide us with only one data
point, we lack the time-series data necessary to repeat such an analysis. Nonetheless, we
can examine how the size of the middle class changes according to changes in the
number of earners in the household within one country at one point in time. Again,
macroeconomic conditions should be one reason that the number of earners varies from
country to country, and more earners should increase the size of the middle class,
especially since (as we saw earlier) government tax and spending policy has little impact
on the size of the middle class.
Table #6 calculates the size of the middle class in less developed American
nations broken down by the number of earners in the household. In developed nations,
such variation is small. Some of our six Latin American countries follow this pattern. For
example, there is little trend for Uruguay or Guatemala (until we reach a point where
households have more than three earners, which is possibly due to the small sample size).
On the other hand, in Brazil and Mexico it is pretty clear that an increase in the number
of earners in the household brings with it a greater probability of middle-class incomes,
and so it appears that increased labor force participation by the household is important to
generate middle-class status there. The results for Peru also indicate that the number of
earners is likely a factor affecting the size of the middle class there.
These results are more remarkable because of the relationship between household
size and middle-class status. Larger households are less likely to be middle class than
smaller households because of the additional income needs of more household members.
To control for this, Table #7 looks at two fixed household types and examines how the
probability of being middle class changes with changes in the number of earners. It
shows how the fraction of middle-class households changes with the number of earners
for non-elderly, married couples with two children and with three children. Data for
households with no earners are not included because the small number of such
households in this category.
Again, our results are somewhat mixed, but they are somewhat different from the
result of Table #6. For Uruguay now, as the number of earners increases from 1 to 3
household members, the probability of being middle class increases sharply. For
Guatemala there is little change; if anything, there is a small decrease in the probability of
being middle class as the number of earners increases. Peru also seems to follow this
pattern. Finally, for Brazil and Mexico, more earners still increase the probability of
being middle class. Thus, for these two countries it appears that labor market variables
(such as labor force participation rates and employment) do matter for attaining middle-
class status. For the other Latin American countries, the relationship is ambiguous at best.
We can also examine this question by looking at some labor force variables in LIS
datasets—for example, the number of hours worked by household heads and by spouses,
the number of weeks worked full-time and part-time by household heads and spouses, or
the number of weeks that the head and spouse were unemployed during the year. During
times of unemployment, we would expect to see declines in the number of hours worked
and the number of weeks worked full-time during the year. As a result, the middle class
should shrink. Alas, the available labor force data is limited for developed nations; data is
even more limited for developing ones.
Nonetheless it is possible to begin such an analysis. Table #8 examines the size of
the middle class by the employment status of the household head and spouse. For obvious
reasons we look only at households headed by someone under 60 years old and restrict
the analysis to married couples. We calculate and report the probability of a household
being middle class when the head is employed, when the spouse is employed, when either
is employed and when both are employed.
Table #8 indicates that employment seems to make little difference for the less
developed nations of the Americas. The probability of being middle class does not seem
to depend on which spouse is employed. It is pretty much the same if either one works. In
addition, the probability of being middle class does not change very much when the
spouse works in addition to the household head. Going from column 1 (where only the
household head is employed) to column 4 (where both spouses are employed), we see
that the size of the middle class falls in Columbia, Guatemala and Peru, and increases just
a bit in Uruguay. Only in Brazil and Mexico does the addition of an employed spouse
increase the size of the national middle class. This supports our earlier result about the
importance of employment for middle-class status in Brazil and Mexico.
Finally, we add comparable figures for the two developed American nations at the
bottom of Table #8. Following the patterns of Latin American nations other than Brazil
and Mexico, the employment situation of the spouse of the household head does not
matter very much in terms of providing middle-class incomes in Canada and the US.
6. SUMMARY AND CONCLUSION
This study has used the LIS to examine the size of the middle class across several less
developed American nations. One main finding is that in the mid 2000s the size of the
middle class in our sample of Latin American countries does not seem to depend on
demographic factors. A second finding is that, in contrast to most developed nations,
government tax and spending policies do little to increase the size of the middle class in
less developed America. Finally, as with the developed world, labor market factors do not
have much impact the size of the middle class. The main exceptions here seem to be
Brazil and Mexico, where employment appears to increase the size of the middle class.
These findings support my previous work (Pressman 2007, 2010) on the
determinants of the size of the middle class, mainly in developed nations. There it was
found that neither socio-demographic nor labor market variables were important in
determining the relative size of the middle class across nations. Rather, the thing that is
important for a large middle class is government policy—a progressive tax system and a
set of generous government spending programs that benefit low-income and middle-
income households. Lacking such policies is a good part of the reason Latin American
nations do not have a large middle class compared to developed nations. If the middle
class is going to expand in Latin America, it will be necessary for the state to develop and
expand policies that support a large middle class—child allowances and family leave,
unemployment and disability insurance, and a more inclusive and more generous
The conclusion of this paper also supports the results of numerous studies on
income inequality throughout the world. One main finding of this extensive literature is
that generous public safety nets and social services result in greater income equality
(Bradly et al. 2003; Moller et al. 2003; Western & Healy 1993). Although this work has
mainly focused on developed nations, it seems as though the same conclusions hold for
less developed countries.
Adelman, Irma & Morris, Cynthia, Society, Politics, and Economic Development: A
Quantitative Approach, Baltimore, Johns Hopkins University Press, 1967.
Aristotle, Politics, trans. H. Rackham, Cambridge, MA, Harvard University Press, 1932.
Barro, Robert, “Determinants of democracy,” Journal of Political Economy, vol. 107,
December 1999, pp. S158-S183.
Birdsall, Nancy, Graham, Carol & Pettinato, Stefano, “Stuck in the tunnel: is
globalization muddling the middle class?, Brookings Institution, Center on Social and
Economic Dynamics, Working Paper #14, 2000.
Bradley, David, Huber, Evelyne, Moller, Stephanie, Nielsen, François & Stephens, John,
“Distribution and redistribution in postindustrial democracies,” World Politics,” vol. 55,
January 2003, pp. 193-228.
Burtless, Gary, “Effect of growing wage disparities and family composition shifts on the
distribution of U.S. income,” European Economic Review, vol. 43, April 1999, pp. 853-
Cancian, Maria & Reed, Deborah, “Assessing the effect of wives’ earnings on family
income inequality,” Review of Economics and Statistics, vol. 80, February 1998, pp. 73-
Cancian, Maria & Reed, Deborah, “Changes in family structure: implications for poverty
and related policy,” in Danziger, Sheldon & Haveman, Robert (Eds) Understanding
Poverty, New York, Russell Sage, 2001, pp. 69-96.
Cancian, Maria & Schoeni, Robert, “Wives’ earnings and the level and distribution of
married couples’ earnings in developed countries,” Journal of Income Distribution, vol.
8, Summer 1998, pp. 45-61.
Cashell, Brian, “Who are the ‘Middle Class’?,” Congressional Research Service Report
#RS22627, October 22, 2008.
Easterlin, Richard, “Income and happiness: towards a unified theory, Economic Journal,
vol. 111, July 2001, pp. 465-484.
Easterly, William, “The middle class consensus and economic development,” Journal of
Economic Growth, vol. 6, December 2001, pp. 317-335.
ECLAC (Economic Commission for Latin America and the Caribbean), Social
Panorama of Latin America, 2005, Santiago, United Nations, 2005.
Ermisch, John, “Econometric analysis of birth rate dynamics in Britain,” Journal of
Human Resources, vol. 23, Fall 1988a, pp. 53-76.
Ermisch, John, “Economic influences on birth rate,” National Institute Economic Review,
November 1988b, pp. 71-81.
Estache, Antonio & Leipziger, Danny, (Eds) Stuck in the Middle, Washington, DC,
Brookings Institution, 2009.
Esping-Andersen, Gøsta, “Sociological explanations of changing income distributions,”
American Behavioral Scientist, vol. 50, January 2007, pp. 639-658.
Frank, Robert, Falling Behind: How Rising Inequality Harms the Middle Class,
Berkeley, CA, University of California Press, 2007.
Fuchs, Victor, “Redefining poverty and redistributing income,” The Public Interest, #8,
Summer 1967, pp. 88-94.
Galbraith, James, “A perfect crime: global inequality,” Deadalus, vol. 131, Winter 2002,
Gauthier, Anne-Helene & Hatzius, Jan, “Family benefits and fertility: an econometric
analysis,” Population Studies, vol. 51, November 1997, pp. 295-306.
Goldstein, Markus & Estache, Antonio, “The scope and limits of subsidies,” in Estache,
Antonio & Leipziger, Danny (Eds), Stuck in the Middle, Washington, DC, Brookings
Institution, 2009, pp. 75-96.
Gottschalk, Peter & Danziger, Sheldon, “Inequality of wage rates, earnings and family
income in the United States, 1975-2002,” Review of Income and Wealth, vol. 51, June
2005, pp. 231-254.
Horrigan, Michael & Haugen, Steven, “The declining middle class thesis: a sensitivity
analysis, Monthly Labor Review, vol. 111, #5, May 1988, pp. 3-13.
Karoly, Lynn & Burtless, Gary, “The effects of rising earning inequality on the
distribution of US income,” Demography, vol. 32, August 1995, pp. 379-405.
Korzeniewicz, Roberto & Smith, William, “Poverty, inequality and growth in Latin
America: searching for the high road to globalization,” Latin America Research Review,
vol. 35, October 2000, pp. 7-54.
Landes, David, The Wealth and Poverty of Nations, New York, W.W. Norton, 1998.
Layard, Richard, Happiness: Lessons from a New Science, New York, Penguin Press,
Lerman, Robert & Yitzhaki, Shlomo, “Income inequality effects by income sources: a
new approach and applications to the United States”, Review of Economics and Statistics,
vol. 67, February 1985, pp. 151-156.
Lindert, Kathy, Skoufias, Emmanuel & Shapiro, Joseph, Redistributing Income to the
Poor and the Rich: Public Transfers in Latin America and the Caribbean, Washington,
DC, World Bank, 2006.
Luttmer, Erzo, “Neighbors as negatives: relative earnings and well-being,” Quarterly
Journal of Economics, vol. 120, August 2005, pp. 963-1002.
Luxembourg Income Study (LIS) Database, http://www.lisproject.org/techdoc.htm
(multiple countries; accessed 28 August 2009 through 28 August 2010).
Macinol, John, The Movement for Family Allowances, 1918-45: A Study in
Social Policy Development, London, Heinemann, 1980.
Malthus, Thomas Robert, An Essay on the Principles of Population, 2
Moller, Stephanie, Huber, Evelyne, Stephens, John & Bradley, David & Nielsen,
François, “Determinants of relative poverty in advanced capitalist democracies,”
American Sociological Review, vol. 68, February 2003, pp. 22-51.
OECD, The OECD List of Social Indicators, Paris, OECD, 1982.
Orshansky, Mollie, “Consumption, Work and Poverty,” in Seligman, B.B. (Ed.) Poverty
as a Public Issue, New York, Free Press, 1965, pp. 52-84.
Orshansky, Mollie, “How Poverty is Measured,” Monthly Labor Review, vol. 92, #2,
February 1969, pp. 26-41.
Pew Research Center, Inside the Middle Class: Bad Times Hit the Good Life,
Washington, DC, Pew Research Center, 2008.
Pressman, Steven, “The decline of the middle class: an international perspective,”
Journal of Economic Issues, vol. 41, March 2007, pp. 181-200.
Pressman, Steven, “Public policy and the middle class in the mid 2000s,” Journal of
Economic Issues, vol. 44, March 2010, pp. 243-262.
Sen, Amartya, Development as Freedom, New York, Random House, 1999.
Shorrocks, Anthony, “The impact of income components on the distribution of family
Income,” Quarterly Journal of Economics, vol. 98, May 1983, pp. 311-326.
Smeeding, Timothy, Buhmann, Brigitte & Rainwater, Lee, “Equivalence Scales, Well-
Being, Inequality and Poverty: International Comparisons Across Ten Countries Using
the Luxembourg Income Study (LIS) Database,” LIS-CEPS Working Paper # 17, 1988.
Smeeding, Timothy, Rainwater, Lee, Buhmann, B. & Schmaus, Gunther, “Luxembourg
Income Study (LIS) Information Guide,” LIS-CEPS Working Paper #8, 1988.
Smeeding, Timothy, Schmaus, Gunther & Allegra, Serge, “An Introduction to LIS,” LIS-
CEPS Working Paper #1, 1985.
Thurow, Lester, “The Disappearance of the Middle Class,” New York Times, Feb. 5,
1985, p. F3.
Vadakin, James, Family Allowances, Miami, University of Miami Press, 1958.
Vadakin, James, Children, Poverty, and Family Allowances, New York, Basic Books,
Western, Bruce & Healy, Kieran, “Explaining the OECD wage slowdown: recession or
labor decline?,” European Sociological Review, vol. 15, #3, September 1999, pp. 233-
CHILD POVERTY RATES ACROSS NATIONS AND OVER TIME
COUNTRY WAVE # 1 WAVE # 2 WAVE # 3 WAVE # 4 WAVE # 5 WAVE # 6
17,8% 17,4% 17,5% 12,8% 13,9% 14,7%
N.A. 5,7% 6,1% 9,1% 7,9% N.A.
18,0% 17,7% 17,0% 16,6% 18,3% 19,5%
N.A. 5,4% 5,0% 3,5% 3,9% 5,1%
N.A. 3,9% 3,1% 2,3% 4,8% 5,5%
11,8% 13,2% 12,3% 12,4% 14,0% N.A.
5,9% 10,0% 7,6% 12,6% 11,1% 12,9%
N.A. 14,1% 7,3% 22,2% 21,0% 24,5%
N.A. 7,5% 5,9% 8,9% 17,7% 17,1%
N.A. 4,8% 8,8% 9,1% 9,2% N.A.
4,9% 4,7% 5,2% 4,8% 4,2% 6,3%
6,3% 4,5% 3,9% 10,6% 6,3% 6,7%
10,0% 15,2% 20,9% 22,8% 21,1% 16,9%
24,8% 29,9% 31,0% 30,2% 27,0% 26,3%
12,4% 11,0% 10,8% 12,7% 12,9% 14,1%
Source: Author's calculations from the Luxembourg Income Study datasets
Note: See text for definition of poverty
COUNTRY WAVE # 1 WAVE # 2 WAVE # 3 WAVE # 4 WAVE # 5 WAVE # 6
20,6% 20,0% 19,8% 25,1% 26,0% 27,1%
N.A. 18,9% 17,5% 18,6% 15,3% N.A.
19,9% 19,8% 18,6% 21,3% 23,9% 25,1%
N.A. 7,7% 8,8% 7,4% 8,3% 9,3%
N.A. 8,3% 10,1% 14,1% 15,0% 13,3%
18,4% 24,9% 24,4% 22,2% 21,9% N.A.
11,7% 15,9% 13,7% 17,6% 18,7% 22,6%
N.A. 14,1% 16,5% 22,2% 21,0% 28,2%
N.A. 15,2% 17,8% 19,7% 29,2% 28,5%
N.A. 13,0% 13,9% 15,5% 14,5% N.A.
7,8% 7,1% 10,4% 11,3% 7,9% 11,3%
8,5% 7,7% 8,5% 13,5% 10,9% 11,0%
15,7% 23,2% 26,0% 29,7% 28,5% 23,4%
24,8% 29,9% 31,0% 30,2% 27,0% 26,4%
15,9% 16,1% 16,9% 19,2% 19,2% 20,6%
CHILD POVERTY RATES ACROSS NATIONS AND OVER TIME,
LESS CHILD ALLOWANCES
Source: Author's calculations from the Luxembourg Income Study datasets
Note: See text for definition of poverty
COUNTRY WAVE # 1 WAVE # 2 WAVE # 3 WAVE # 4 WAVE # 5 WAVE # 6
-2,8% -2,6% -2,3% -12,3% -12,1% -12,4%
N.A. -13,2% -11,4% -9,5% -7,4% N.A.
-1,9% -2,1% -1,6% -4,7% -5,6% -5,6%
N.A. -2,3% -3,8% -3,9% -4,4% -4,2%
N.A. -4,4% -7,0% -11,8% -10,2% -7,8%
-6,4% -11,7% -12,1% -9,8% -7,9% N.A.
-5,8% -5,9% -6,1% -5,0% -7,6% -9,7%
N.A. N.A, -9,4% N.A. N.A. -3,7%
N.A. -7,7% -11,9% -10,8% -11,5% -11,4%
N.A. -8,2% -5,1% -6,4% -5,3% N.A.
-2,9% -2,4% -5,2% -6,5% -3,7% -5,0%
-2,2% -3,2% -4,7% -2,9% -4,6% -4,3%
-5,7% -8,0% -5,1% -6,9% -7,4% -6,5%
0,0% 0,0% 0,0% 0,0% 0,0% -0,1%
-3,5% -5,5% -6,1% -7,0% -6,7% -6,4%
THE IMPACT OF CHILD ALLOWANCS ON CHILD POVERTY RATES
Source: Author's calculations from the Luxembourg Income Study datasets
Note: See text for definition of poverty