Está en la página 1de 34

D

O
C
U
ME
N
T
OD
ET
R
A
B
A
J
O
.
S
2
0
0
8/
0
3

EXPECT
AT
I
ONSOFSOCI
ALMOBI
LI
T
Y
,
MERI
T
OCRACYANDT
HEDEMANDFOR
REDI
ST
RI
BUT
I
ONI
NSPAI
N
A
N
T
O
N
I
OM.
J
A
I
MEC
A
S
T
I
L
L
O

Centro de Estudios Andaluces

El Centro de Estudios Andaluces es


una entidad de carcter cientfico y
cultural, sin nimo de lucro, adscrita
a la Consejera de la Presidencia
de la Junta de Andaluca.
El objetivo esencial de esta institucin
es fomentar cuantitativa y cualitativamente
una lnea de estudios e investigaciones
cientficas que contribuyan a un ms
preciso y detallado conocimiento de
Andaluca, y difundir sus resultados
a travs de varias lneas estratgicas.
El Centro de Estudios Andaluces desea
generar un marco estable de relaciones
con la comunidad cientfica e intelectual
y con movimientos culturales en
Andaluca desde el que crear verdaderos
canales de comunicacin para dar
cobertura a las inquietudes intelectuales y culturales.

Las opiniones publicadas por los autores en


esta coleccin son de su exclusiva responsabilidad

2008. Fundacin Centro de Estudios Andaluces


Depsito Legal: SE- SE-4862-08
Ejemplar gratuito. Prohibida su venta.

Centro de Estudios Andaluces

S2008/03

Expectations of Social Mobility, Meritocracy and


the Demand for Redistribution in Spain1
Antonio M. Jaime-Castillo
FCEA & Universidad de Granada
Resumen: Los enfoques econmicos tpicos sostienen que la democracia
producira una redistribucin de la riqueza desde los ricos hacia los pobres.
Dado que el votante mediano es ms pobre que el ciudadano de renta media,
la mayora de la poblacin votar por la redistribucin de la renta. No obstante,
la relacin positiva entre democracia e igualdad ha sido ampliamente criticada,
tanto desde el punto de vista terico como emprico. La pregunta clave es
porqu los pobres no votan por una mayor redistribucin. Algunos
investigadores argumentan que las expectativas de movilidad social pueden
inducir a los pobres a creer que tendrn ingresos ms elevados en el futuro y
por eso se opondran a la redistribucin en el presente. Sin embargo, la
Sociologa ha mostrado que la gente no conoce con certeza su posicin relativa
en la distribucin de la renta, porque los individuos tienden a evaluar sus
propios ingresos a travs de comparaciones con otras personas. En este
trabajo, se analiza hasta qu punto las preferencias por la redistribucin a nivel
individual dependen de las expectativas sobre los ingresos futuros y la
meritocracia. Sin embargo, a diferencia de estudios previos, se investiga si las
expectativas de movilidad estn relacionadas con las autnticas probabilidades
de movilidad o con las expectativas subjetivas de movilidad, las cuales
dependen de la posicin relativa que cada individuo piensa que tiene en la
distribucin de la renta. Los resultados empricos muestran que las
preferencias por la redistribucin dependen de las expectativas de movilidad y
de las ideas sobre la meritocracia, aunque las expectativas subjetivas de
movilidad tienen un mayor impacto sobre las preferencias que las preferencias
objetivas.

A previous version of this paper was presented at the Conference Inequality Beyond Globalization
(Neuchtel, 2008). I am in debt to all the participants in the panel Attitudes towards Inequality and
Redistribution for their helpful comments. The author thankfully acknowledges financial support from
the Centro de Estudios Andaluces (PRY126/08).

Centro de Estudios Andaluces

Palabras clave: desigualdad, movilidad social, redistribucin, preferencias,


expectativas.
Abstract: Meltzer and Richard (1981) state that democracy will boost
redistribution of income from the rich to the poor. Given that the median voter is
poorer than the average-income voter, a majority of voters will vote for
redistribution. However, the positive relationship between democracy and
equalization of income has been highly criticized on both theoretical and
empirical grounds. From a rational point of view, why do not the poor vote for a
higher level of redistribution? Some scholars have argued that expectations of
social mobility may induce the poor to believe they will have a higher income in
the future and thus to vote against redistribution. However, the sociological
literature has shown that people are not well aware of their relative position on
the income ladder, because they evaluate their relative income through
comparison to others. In this paper, I analyze to what extent preferences for
redistribution at the individual level depend on expectations of future income
and meritocracy. Departing from previous studies, I analyze whether
expectations of social mobility are related to true probability of mobility or
whether they are based on subjective expectations of mobility, which depend on
perceived position on the income ladder. Empirical results show that
preferences for redistribution depend on both expectations of social mobility and
meritocracy, although subjective expectations of social mobility have a stronger
impact on preferences for redistribution than objective expectations.
Keywords: inequality, social mobility, redistribution, preferences, expectations.
JEL Classification: D63; D84; I38; J62

Centro de Estudios Andaluces

1. Introduction

According to Meltzer and Richards (1981) model, there must be a positive


correlation between inequality and demand for redistribution at the aggregate level
(more recently, Franzese, 2002). This will be true if preferences for redistribution
depend on relative income at the individual level. Poor voters will prefer a higher level
of redistribution, while those who earn above the mean income will oppose
redistribution, since they will become net losers from redistribution. However, empirical
evidence linking income levels to preferences for redistribution is not conclusive at the
individual level. Nor is the link between inequality and demand for redistribution at the
aggregate level. From a rational point of view, one explanation for the reason why the
poor do not expropriate the rich in a democracy is that individuals care not only about
their current relative income but also about their expected relative income in the future.
This brings the social structure of the society and the expectations of social mobility in a
given context into the model. Those who expect to be better off in the future may not
want to vote for redistribution today. Moreover, according to a normative argument
about opportunities for social mobility, if people think they live in a meritocratic
society, in which inequalities are the outcome of ones merit and effort, they may not
want to vote for redistribution, because hard-working individuals have the opportunity
to climb in the social hierarchy by their own efforts. Therefore, we would expect that in
open societies, where social mobility is relatively high, the demand for redistribution
will be lower, even if inequality is relatively high (Alesina and La Ferrara, 2005;
Benabou and Tirole, 2006).
The idea behind the prospect of upward mobility (the POUM hypothesis) was
formalized by Benabou and Ok (2001) and empirically tested by Alesina and La Ferrara
(2005) for the United States, and by Checchi and Filippin (2004) within an experimental
setting. The POUM hypothesis assumes that individuals have almost perfect knowledge
of the social structure. Specifically, it assumes that they know the shape of the income
distribution and the true probability of social mobility. These assumptions seem to be
somewhat unrealistic, since they do not take into account the fact that people may hold
incorrect beliefs about the fluidity of the social structure (Evans and Kelley, 2004).
More recently, Rainer and Siedler (2008) have addressed the issue of the subjective
probabilities of income mobility. Rainer and Siedler found that subjective expectations
1

about future income, as measured by the subjective expectation of a pay increase, will
shape preferences for redistribution at the individual level. In this paper I analyze a

Centro de Estudios Andaluces

different aspect of the problem, the subjective perception of ones position on the
income distribution and its impact on preferences for redistribution. In the sociological
literature on relative deprivation, it has long been known that individuals systematically
underestimate the real extent of inequality, since they do not know their true position in
the social hierarchy. Individuals view themselves through social comparisons with other
people, but the range of these comparisons is generally limited to the group to which
individuals belong (Davis, 1959; Lockwood, 1966; Pollis, 1968; Runciman, 1966; Rose,
2006). Given that social groups are relatively homogeneous in terms of status and
income, individuals believe incorrectly that they are very close to the average income
earner, as further empirical research has found (Evans and Kelley, 2004).
The main goal of this paper is to show how self-perceived position on the
income ladder affects preferences for redistribution at the individual level, even though
these subjective beliefs about relative position do not accurately represent the
individuals real position. I will show that expectations of social mobility can explain
preferences for redistribution. However, these expectations are based on the subjective
rather than objective probability of upward mobility. These subjective probabilities are,
in fact, the product of the individuals self-perceived position on the income ladder.
Statistical evidence presented here will show that subjective perception about relative
income is highly biased toward the middle of the income distribution. To test this
hypothesis, I will focus on the case of Spain. The rest of the paper is organized as
follows. The next section presents a brief overview of the literature on preferences for
redistribution and social mobility. In the two subsequent sections, I will present the
Spanish case and explain the methodology and the data used in this research. The
following section discusses the major statistical findings. The final section contains the
main conclusions of the paper.

2. Social mobility and preferences for redistribution: an overview

Empirical evidence shows that preferences for redistribution are related in some
way to income but that they cannot be fully explained by this variable alone (Benabou
and Tirole, 2006; Corneo and Grner, 2002; Svallfors, 2006). The self-interest approach
has developed two main sources of explanation to tackle this issue, both of which
2

attempt to incorporate actors beliefs about social mobility into the model. The first
explanation does not depart from self-centered motivations and extends the original

Centro de Estudios Andaluces

model to take into account expectations of future income. Given that taxes and
redistributive schemes are in place for a lengthy time period, individuals will care not
only about todays income after taxes, but also about future income. In contrast, the
second approach is group-oriented and takes into account peoples beliefs about social
structure and how these beliefs shape their preferences for redistribution. If individuals
think they live in a relatively open society where wealth is distributed according to fair
distributive principles (mostly meritocratic ones), they may not favor redistribution,
since the original distribution seems to be fair enough. Both the pure self-interest
approach and the group-oriented one focus primarily on beliefs about social mobility. In
the former, expectations of individual upward mobility may discourage people from
demanding redistribution, since present benefits may become future costs if they
achieve a high income position. From the group-oriented perspective, preferences for
redistribution depend on the degree of openness in the society. The main motivation is
to achieve a fair collective outcome. Assuming a shared meritocratic ideology,
individuals will compare social outcomes with outcomes in a totally open society, in
which only individual merit and effort determine income. The greater the difference
between the two, the greater the demand for redistribution we should expect.
The approach based on self-interest states that, from a dynamic point of view,
preferences for redistribution are to be linked to expectations about future income.
Extending Meltzer and Richards (1981) model to a multi-period setting, we would
expect people to maximize a multi-period utility function in which income at further
stages will be properly discounted. There are two main accounts of expectations for the
future in the literature on preferences for redistribution at the individual level: the
tunnel effect (Hirschman, 1973) and the prospect of upward mobility (POUM)
(Benabou and Ok, 2001).
Hirschman (1973) coined the term tunnel effect to indicate expectations about
future well-being, according to the following metaphor. Suppose a group of people is
driving through a two-lane tunnel in which both lanes move in the same direction.
Assume, furthermore, that there is an obstacle in the road beyond the tunnel, causing a
traffic jam that prevents any car from moving in either lane, although no driver can see
what is happening outside the tunnel. After a while, the cars in one of the lanes start to
move, but no movement occurs in the other lane. Hirschman argues that those in the
3

blocked lane will not feel angry. Instead they will believe that the obstacle on the road
has been removed and that it will be their turn to move sometime soon. We can learn

Centro de Estudios Andaluces

from this metaphor that people may be tolerant to growing inequalities for a while (one
group of people getting richer and the other remaining poor). The key idea is that when
we see others doing well we may expect that our well-being will also improve in the
foreseeable future. This would explain why some societies are able to suffer high
inequality levels without increasing pressures for redistribution.
However, people have a limited amount of patience. After being blocked in the
tunnel for a long time, they may become disappointed if those in the other lane keep
moving. Those who cannot move may suspect foul play and become furious. This
would open the door to complaints, perhaps even direct action. Hirschman (1973)
argued that if the expectations of improvement are not fulfilled, the consequences can
be dramatic. The relative tolerance for inequalities in the first stage may become
frustration in further stages if no improvement occurs. The main conclusion is that
increasing inequalities will lead to strong pressures for redistribution in the long-run.
Nevertheless, we should expect low demand for redistribution in the short-run if the
members of the poor group believe their fortune will improve in the future, even if their
relative position has not changed. Ravallion and Lokshin (2000) tested the tunnel
effect hypothesis in Russia during the highly volatile decade of the nineties. They
concluded that support for redistribution was higher among those who expected their
well-being to diminish and vice-versa. More interestingly, those who were experiencing
an improvement were less inclined to favor redistribution, even after controlling for
other variables such as their real income.
Benabou and Ok (2001) explain things differently. They formalize the history
behind the prospect of upward mobility (POUM), assuming that people care only about
their own income. In their model, people are supposed to maximize an inter-temporal
utility function in which expected future incomes are taken into account. The main
finding of Benabou and Ok (2001) is that a range of individuals with below-average
income will oppose redistribution since they expect that their future income will be
above average. It is clear that this cannot be true for all people with below-average
income. Even if some fraction of the poor of today increases its income above current
average income, those who are richer than average today will have even higher incomes
in the future. Therefore, the poor of today will be still poor in relative terms tomorrow.
Nevertheless, Benabou and Ok (2001) conclude that the prospect of upward mobility
4

could be rational in some circumstances. Specifically, the POUM hypothesis depends


on three critical conditions.

Centro de Estudios Andaluces

The first condition is that redistribution schemes chosen today will last for some
period of time. The second is that people are not too risk averse. In such a context,
people are therefore not too worried about changes in their relative position, since they
may be tempted to vote for redistribution as a means to assuring their current income.
Finally, some fraction of the population that is currently poorer than average must
expect that it will become above average in the future. Benabou and Ok (2001) made
the critical assumption that the income transition function is concave, which basically
means that the income will grow at a higher rate among low income levels. From these
assumptions, they derive two main results. Firstly, the more concave the income
transition function, the smaller the share of people with below-average income
supporting redistribution. Secondly, this fraction will be smaller as the time horizon for
the chosen taxation increases. Benabou and Ok (2001) argue that these are rather
plausible assumptions, given a negative marginal return of income. Under these
assumptions, it is still possible that a majority of the population might be both poorer
than the average now and richer than the average in the future. However, this depends
critically on a second assumption, which is that people have relatively low risk aversion.
As stated above, those who are risk averse may prefer to vote for redistribution in order
to insure against misfortune. Nevertheless, empirical evidence provided by the authors
suggests instead that motivation to risk aversion is stronger than the prospect of upward
mobility. In a different vein, Checchi and Filippin (2004) designed an experiment to test
the main implications of the POUM hypothesis and found strong support for the POUM
hypothesis in different settings.
From a group-oriented point of view, beliefs about meritocracy must play a role
in shaping preferences for redistribution. For instance, Piketty (1995) concluded that
individual preferences for redistribution depend on beliefs about the sources of social
mobility. According to Pikettys model, income is the result of both effort and inherited
factors. However, individual agents do not fully know to what extent effort is important
in determining their own income. They must learn about the true parameters of the
model using a Bayesian rule to incorporate information about their own experience. As
they exert different levels of effort and rise and fall in the social structure, they will gain
information about the effect of effort on income. Because they can only learn from their
own experience, however, each generations knowledge cannot be transmitted to its
5

offspring, a key assumption in Pikettys model. Since learning the true role of effort in
determining income level is costly (it requires a lot of costly experimentation), each

Centro de Estudios Andaluces

generation remains unable to learn the real chances of social mobility. Even if agents
are truly rational Bayesian learners, their initial beliefs will not converge with the true
parameters for the whole population. In Pikettys model, final beliefs about social
mobility depend on initial beliefs, whether or not they are true. Thus individuals holding
different initial beliefs will interpret an experience of upward (or downward) mobility in
different ways. And therefore, they will vote for different redistribution schemes,
although all of them care about social justice. Those who perceive their society as quite
open will vote for more generous redistribution and vice-versa.
Benabou and Tirole (2006) have developed a somewhat different model
intended to explain cross-national differences in support of redistribution. Their main
argument is cultural. Individuals have different beliefs about the origin of social
inequalities. Due to imperfect will-power, individuals strive to motivate themselves to
make a great effort. To make sense of their own actions, people attempt to convince
themselves that effort will bring wealth. This creates a vicious circle: people need to
believe in meritocracy to commit themselves to a level of high effort. At the same time,
the need to avoid cognitive dissonance requires that people who exert a high level of
effort need to believe that their effort will have a proper reward. Let us imagine that in a
given society the majority of the people are committed to the ideology of meritocracy.
This would generate an equilibrium in which the majority will vote for low taxes. At the
same time, low taxes will motivate people to work hard, since they already know that
redistribution is low. Let us now imagine the reverse situation, in which the majority
holds the view that inequalities are caused by external factors and do not depend on
personal effort. In this situation, the majority will vote for generous redistribution.
Furthermore, it will be difficult to find a motivation for higher levels of exertion. Now,
because the belief against meritocracy prevents people from taking risks, people will
vote for higher redistribution. According to Benabou and Tirole (2006), while the
former equilibrium describes the American case, the latter is in place in most European
countries. Interestingly, both of these are self-maintained and self-reinforcing equilibria,
making it very difficult to move from one to another.
Both Pikettys and Benabou and Tiroles approaches explicitly take into account
beliefs about meritocracy and social mobility in explaining preferences for
redistribution. For Piketty (1995), individuals are concerned about a fair model of
6

society, while for Benabou and Tirole (2006), individuals adapt their beliefs about the
opportunities for social mobility to the dominant beliefs represented by the equilibrium

Centro de Estudios Andaluces

in place. Empirically, Linos and West (2003) have found that ideas about the
determinants of social mobility play a role in shaping preferences toward redistribution,
even though there are important differences between countries, depending on the
welfare regime. This conclusion is fundamentally consistent with Benabou and Tiroles
model.
From this short review of the literature, it is clear that there are many sources of
expectations concerning prospects of social mobility, of which three are particularly
important here: a) objective expectations: individuals compute probabilities of mobility
based on the aggregate flux in the occupational structure or in the income distribution
(Alesina and La Ferrara, 2005); b) subjective expectations: individuals make their own
assessment of the probabilities of mobility based on their idiosyncratic circumstances,
such as expectations of promotion at workplace (Rainer and Seidler, 2008); c)
meritocracy: individuals expect different chances of social mobility depending on their
beliefs about the social structure (if they believe that wealth is the result of talent and
hard work) (Benabou and Tirole, 2006; Piketty, 1995).
None of the previous expectations of social mobility takes into account how
people view themselves on the income ladder. Some of the approaches assume that
people already know where they are (such as the objective probabilities of income
mobility). The others do not consider this issue (such as the beliefs about meritocracy,
which are neutral regarding the individuals position on the income ladder). In the late
sixties, Runciman (1966) stated that perceptions of inequality in society as a whole are
shaped by feelings of relative deprivation at the individual level. Deprivation theory is
based on the idea that people view their own position in relation to others (Davis, 1959;
Pollis, 1968). This assessment is parallel to the tunnel effect hypothesis proposed by
Hirschman (1973). However, contrary to Hirschmans prediction, Runciman concluded
that social order is possible in the presence of persistent high inequality, because people
underestimate the extent of inequality in the society as a whole. The tendency to
underestimate social inequalities has been demonstrated in several empirical studies
(Lockwood, 1966; Rose, 2006). According to Runciman (1966), one possible
explanation derives from reference group theory. Though the determination of the
relevant reference group for a given individual is ultimately an empirical question (Fehr
and Schmidt, 1999), most people have a very limited group of reference within which
7

they make comparisons about their own position in the social hierarchy, given that
social groups are relatively homogenous in status and income. Thus people tend to view

Centro de Estudios Andaluces

themselves as very close to the average individual, because they are not so different
from the representative member of the group to which they belong. This belief is, of
course, incorrect, but people behave as if it were true.
Rose (2006) re-examined the conclusions posited by Runciman forty years after
his seminal work. The social structures of contemporary societies changed remarkably
during this period, and mass media now have a greater influence on how people
evaluate social reality. However, despite these important changes, Rose concluded that
individuals continue to make very limited social comparisons in everyday life. As a
result, they do not appreciate the full extent of inequality. Because they tolerate
inequality, it does not become a source of social conflict. In a similar vein, Evans and
Kelley (2004) found that most people tend to place themselves systematically in the
middle of the income ladder. Evans and Kelley interpret this interesting fact as a special
case of the availability heuristic, which suggests that any given individual builds his
own image of the society at large by generalizing from his experience within familiar
groups and information received from the media. Therefore, peoples perceptions of
their place in the social hierarchy are largely formed by the circle of their close
acquaintance (Evans and Kelley, 2004: 4).
As previously argued, we should expect that preferences for redistribution will
depend on expectations about future income and beliefs about mobility. However,
measuring expectations of social mobility is not an easy task. Expectations of future
income depend on the relative position on the income ladder and a transition matrix of
income between periods. While we can compute objective relative positions and
transitions matrices, these objective measures are not necessarily identical to subjective
beliefs, since it is very unrealistic to assume that people have accurate knowledge of the
social structure. As previous studies have found, people usually locate themselves in the
middle of the income ladder (Evans and Kelley, 2004). Therefore, it is more plausible to
assume that people make their expectations about future income based on their
perception of their current position. Even if they have full knowledge of the transition
matrix of income, their expected future income will be biased, since they do not really
know where they are now on the income ladder. Nevertheless, even if we have
information about where people think they are on the income ladder, we cannot know
the probability that people assign to being in each of the positions on the income ladder
8

in the foreseeable future. The next two sections are devoted to explaining the way in
which expectations of social mobility at the individual level were computed in this

Centro de Estudios Andaluces

context and the empirical strategic decisions undertaken.

3. Expectations of social mobility in Spain

From a comparative point of view, income inequality in Spain is relatively high


within the European context. Inequality of income distribution at the end of the nineties,
as measured by Gini index, was 0.33 (Eurostat, 2000). At this time, it was similar to
Greece and Great Britain and only lower than that of Portugal. Consequently, Spaniards
believe they live in a very unequal society (Jaime-Castillo, 2000). Data from the ISSP
(1999) show that, 89.3 % of Spaniards agree (to some extent) with the statement that
differences in income in Spain are too large. At the same time, demand for
redistribution in Spain was higher than the European average and similar to that in other
Southern European countries. On a five-point scale, the average demand for
redistribution was 4.01 with a standard deviation of 0.93. Only Portugal had a higher
average among the Western European countries included in this study. As to social
mobility, Spain had intermediate levels of social mobility (Carabaa, 1999) and
relatively high income mobility (Cant, 2000). These facts seem to translate into
popular beliefs about social structure. Within the European context, Spaniards believe
that they live in a relatively open society: 38.2 % of them agree (to some extent) with
the statement that people get rewarded for their effort. This proportion was lower than
in the US (64.7 %) and in West Germany (57.0 %) but higher than in most of the
European countries, including those with less inequality.
The preceding figures seem to reflect some degree of congruence between
objective data and subjective perceptions of the reality. However, as occurs in other
countries (Evans and Kelley, 2004), Spaniards lack accurate information about their
position in the social hierarchy. Figure 1 shows the position where people locate
themselves on the income ladder in Spain. On a ten-point scale, 42.2 % of the people
place themselves at Point 6, which is the modal value. The mean of the distribution is
5.17 with a standard deviation of 1.45. An analysis of the relationship between real
position on the income ladder and subjective position reveals two different patterns that
would explain the inflation around the middle of the subjective scale. First, low-income
earners overestimate their position. At the same time, those at top underestimate their
9

relative position, resulting in a regression to the mean of the income distribution. This
would explain why the correlation between real income decile and subjective

confidence level.

10

Percent
20

30

40

Figure 1: Perceived position on the income distribution

Centro de Estudios Andaluces

evaluation, measured by Spearmans rho, is only 0.338, while significant at 95.5 %

4
6
8
Income decile (subjective evaluation)

10

Source: International Social Survey Program.

The other issue we must consider in order to estimate subjective expectations


about future income is the transition matrix between periods. Unfortunately, this is
difficult to compute, since survey data do not provide us with enough information to
make such calculations. There are two different basic approaches to estimating this
value, both of which have strengths and weaknesses. While Rainer and Siedler (2008)
use subjective expectations of receiving an increase (decrease) in pay (in work salaries)
in the coming year, Alesina and La Ferrara (2005) use objective probabilities of income
mobility between periods of one and five years. These objective probabilities are
derived from a transition matrix similar to that presented in Table 1. Alesina and La
Ferrara (2005) also include a subjective prospect of mobility: agreement with the
statement that people like me and my family have a good chance of improving our
standard of living. However, it is unclear whether this variable truly reflects subjective
10

prospects of upward mobility at the individual level. For instance, one may contend that
it is more likely to measure the degree of openness in a given society.

Centro de Estudios Andaluces

Subjective prospects of increase in pay are problematic since they do not enable
one to infer a quantifiable expected future income. Hence, they do not necessarily
reflect the subjective probability of being above the mean of income distribution (which
is the cut-off point in the POUM hypothesis) or any other reference point. Alesina and
La Ferrara (2005) have found that not all measures of mobility work as predictors of
preferences for redistribution. Interestingly, only those that are related to the probability
of earning an above average income seem to have a significant effect. This would
suggest that expectations of social mobility must be defined in such a way that they
make it possible to distinguish between net losers and net winners from redistribution.

Table 1: Transition matrix of income mobility in Spain (1993-1998)

1st
2nd
3rd
4th
5th
6th
7th
8th
9th
10th

1st
34.9
22.4
10.4
9.3
7.3
4.6
6.3
4.0
0.9
1.1

2nd
17.9
26.6
21.9
14.8
8.4
7.2
6.3
2.1
2.0
1.5

3rd
10.1
14.1
24.6
24.6
13.1
8.7
7.2
4.5
2.2
0.9

4th
8.7
11.0
11.7
17.8
21.0
12.4
8.5
6.4
4.4
1.5

5th
8.3
7.8
9.3
9.1
17.8
19.1
14.8
9.4
3.8
2.4

6th
6.7
9.8
6.1
8.0
14.3
18.7
17.3
12.0
7.1
2.8

7th
4.6
3.4
8.6
7.3
8.1
12.1
14.6
21.5
14.0
6.3

8th
3.9
2.2
3.8
4.3
3.2
11.1
11.7
18.6
21.5
8.3

9th
3.0
1.6
2.9
3.9
4.5
3.9
8.7
15.8
28.2
19.2

10th
1.8
1.1
0.7
0.9
2.4
2.4
4.7
5.7
16.0
55.9

Source: European Community Household Panel (Eurostat, 2000).

At the same time, objective probabilities of income mobility may not truly
reflect subjective prospects. As has been shown, individuals have a biased perception of
their position on the income ladder. Therefore, it is difficult to assume that they will
have reliable information about their chances of being mobile. As in the previous case,
they may lack important information about the social structure of the society in which
they live, or they may deceive themselves into wishful thinking in the guise of
predictions. In both cases, transition matrices of income mobility are not fully reliable.
However, the goal of this paper is to compare the effect that expectations of social
mobility based on subjective position on the income ladder have on preferences for
redistribution with the effect of expectations of social mobility based on objective
income. This requires comparable transition matrices to perform the empirical analysis.
I have therefore used a transition matrix of income mobility in Spain similar to that used
11

by Alesina and La Ferrara (2005). The cells of the matrix have been computed using
data from the European Community Household Panel (ECHP). Household income is

Centro de Estudios Andaluces

defined according to the modified OECD scale. The total net household income is
divided by the number of members in the household. However, the scale assigns a
weight of 1.0 to the first household member aged 14 or over, 0.5 to each additional
member aged 14 or more and 0.3 to each member under 14 years old.
Even if this transition is not an appropriate measure of expected future income, it
will allow us to compare the effect of objective and subjective income on preferences
for redistribution, under the assumption that individuals have full knowledge of chances
for mobility within the social structure. Thus, if the expected future income based on
perceived position on the income ladder has a stronger effect than the expected future
income based on the real income position, we will be able to conclude that subjective
expectations matter in a way different to that shown in previous studies.
Table 1 shows the transition matrix of income mobility in Spain between 1993
and 1998. The number in each cell represents transition probabilities between income
deciles. Thus, pij in Row i and Column j is the probability that an individual whose
income is in the ith decile in 1993 moved to the jth decile in 1998. The elements in the
diagonal represent the probability of remaining unchanged along a period of five years.
Cells below the diagonal represent the probability of experiencing downward mobility,
while cells above the diagonal represent the probability of upward mobility. The time
interval chosen is five years, as in Alesina and La Ferrara (2005), although their study
includes an additional interval of one year. As they pointed out, the probability of
remaining in the same income decile increases as the length of the temporal interval
increases. Small variations in relative position in the short run (primarily between
adjacent deciles) are cancelled out in the long-run. Although there is no clear theoretical
reason to prefer any particular length of time to compute the transition matrix, it seems
to be more appropriate to choose a relatively stable measure such as the five year
interval. In analyzing the empirical results, Alesina and La Ferrara (2005) experimented
with different definitions and time horizons, obtaining fairly similar results with both
the one-year and the five-year transition probabilities1.

Although the results are not reported in tables, I have performed a similar analysis using both one-year
and five-year time horizons. The magnitude and the significance of the coefficients in the ordered probits
remained mostly unchanged.

12

A simple inspection of the transition matrix in Table 1, using a simple absolute


mobility measure such as the proportion of individuals outside the main diagonal of the

Centro de Estudios Andaluces

matrix, reveals that 74.2 % of the population experiences some degree of mobility,
while the rest remains immobile. On the other hand, the probability of income mobility
(upward or downward) is higher in the middle of the income ladder than at the
extremes. For instance, the probability of being mobile for those in the 5th decile is
0.822, while the same probability is 0.651 for those in the bottom income decile and
0.441 for those in the top income decile. The data also suggest that mobility is higher at
the bottom of the income distribution. This would indicate that low income positions are
more volatile than high income positions. A closer look at the table shows that mobility
is limited primarily to adjacent deciles; the higher mobility at the bottom of the income
ladder is mostly between similar positions. In fact, this is a very general pattern. The
probability of mobility between any two given deciles is inversely proportional to the
distance between them. For instance, the probability of changing more than one income
decile within a given period is 0.431 for the whole table, and the probability of changing
more than two deciles is 0.260. According to the data in the table, the objective
probability of earning an above-average income in the next five year period is very low
at the bottom of the income distribution. Further analysis will show that if people were
truly aware of their relative position, only those who are currently above average would
expect to be above the average in the next period2. Thus, if people are using objective
probabilities of mobility, expected future incomes will have a very limited impact on
preferences for redistribution under the assumption of risk-neutrality.
Transition probabilities are not the same for every individual, however. Factors
such as gender, age and education affect probabilities of income mobility. Different
income transition matrices have therefore been computed for different groups. Three
factors have been taken into account in computing these specific income transition
matrices: gender (male and female), age (18 to 34 years, 35 to 44 years, 45 to 64 years,
and 65 and older) and education level (no education, primary school, secondary school,
and university). Note that, as we divide the sample into more homogeneous clusters of
individuals, we obtain more accurate information about their real probabilities of
income mobility. At the same time, the subsamples become smaller as we divide the
2

Calculations are straightforward. Given that only the 7th to 10th deciles have a higher income than
average, and after computing the probability of being in the 7th to 10th deciles for each decile, it can be
shown that only those in the range from the 7th to the 10th deciles have probabilities larger than 0.50 of
being in the same range.

13

general sample into subgroups, and thus transition probabilities may become
statistically unreliable. For this reason, we can only test the effect of one variable at a

Centro de Estudios Andaluces

time. In the analysis below, I present and compare the results of different specifications
of the income transition matrix.

4. Data and methods

As argued previously, preferences for redistribution are shaped by probabilities


of social mobility and meritocracy, although probabilities of social mobility are not
necessarily accurate. Even assuming that people know what these probabilities are, their
expected income will depend on where they place themselves on the income ladder. To
solve this problem, two different probabilities of income mobility have been computed.
The objective probability refers to the real probability of being mobile, derived from the
current relative income and the chances of mobility between deciles; subjective
probability refers to the self-assessed probability of being mobile, taking into account
the place where people locate themselves on the income distribution.

Data and variables

In order to test whether objective or subjective probabilities are the most


important factor shaping preferences for redistribution at the individual level, two
different sources of data have been used. Attitudinal variables come from the
International Social Survey Program (ISSP): Social Inequality III (1999). Data about
income and income mobility are computed using the European Community Household
Panel (ECHP, 1999, 6th wave) in Spain. The dependent variable used in this research is
the individual preference for redistribution. Respondents were asked about their level of
agreement with the following statement: It is responsibility of the government to
reduce the differences in income between people with high incomes and those with low
incomes. The ordered categorical responses were: (1) strongly disagree, (2)
disagree, (3) neither agree nor disagree, (4) agree, and (5) strongly agree (the
response categories followed the opposite order in the questionnaire, but I reversed
them for easier interpretation).

Table 2: Descriptive statistics

14

Centro de Estudios Andaluces

Variable
Government must reduce differences in
income
Real income
Expected income (objective probability)
Expected income (subjective probability)
Probability of being in the 7th-10th deciles
(objective)
Probability of being in the 7th-10th deciles
(subjective)
People get rewarded for their effort
People get rewarded for their skills
Gender
Age
Education level
Work status
Self-employed
Public worker
Union member
Differences in income are necessary

Mean
4.008

Std. Dev.
0.934

Min.
1

Max.
5

6905.653
7113.308
7345.719
0.240

2162.174
1450.657
1066.788
0.153

2313
5328.146
5328.146
0.083

13767
17168.540
17168.540
0.897

0.252

0.106

0.083

0.897

2.893
3.057
1.509
45.187
1.212
2.055
0.126
0.123
0.069
2.667

1.089
1.033
0.500
18.315
0.810
0.964
0.332
0.329
0.254
1.121

1
1
1
18
0
1
0
0
0
1

5
5
2
91
3
3
1
1
1
5

Source: International Social Survey Program and European Community Household Panel.

Four types of explanatory variables are included in the analysis: household net
equivalised income, expectations of mobility and future income, perception of
meritocracy in society, and socio-demographic variables. Given that the ISSP survey
does not have data about real income, household net equivalised income is computed as
the mean income of the decile to which the individual belongs. The data of mean
income by decile are taken from the ECHP 6th wave (1999). Natural logarithm of
annual net equivalised household income is taken as an explanatory variable in
statistical models. Expected future income is measured using the method followed by
Alesina and La Ferrara (2005). However, I distinguish between objective and subjective
expected future income, although I assume that probabilities of mobility are common
knowledge. The objective expected future income is computed as if individuals already
know their position on the income ladder:
10

OEI d ,t = pd j y j ,t +1
i= j

This expression represents the income that an individual who is in decile d at


time t will expect to have at time t+1. It is a weighted average of the mean income of all
deciles in time t, where the weights are the probabilities that the individual has of
moving to those deciles from t to t+1, leaving the income decile d to which he belongs
at time t. In addition to the general expected income, specific expected incomes have
15

been computed by group, as previously discussed (groups are defined by gender, age,
and education level).

Centro de Estudios Andaluces

The subjective expected future income is computed as a function of the income


decile to which individuals think they belong:
10

SEI s ,t = ps j y j ,t +1
i= j

This expression represents the income that an individual who thinks he is in


decile s at time t will expect to have at time t+1. It is a weighted average of the mean
income of all deciles in time t, where the weights are the probabilities that the individual
has of moving to those deciles during the period from t to t+1, leaving the income decile
s to which he thinks he belongs at time t. Natural logarithms of objective and subjective
expected future income are taken as explanatory variables. In both cases, the
probabilities of changing from one income decile to another are those contained in the
transition matrix reported in Table 1.
The second measure of expectations about position on the income ladder is the
probability of being above the mean income in the next period. Given that the mean net
equivalised household income for the whole population (using the modified OECD
scale) in the period analyzed is Euro 8,905, and the mean income for the sixth and
seventh deciles is Euro 8,058 and 9,412, respectively, we can safely assume that those
in the seventh decile are above the mean income (a result similar to those used by
Alesina and La Ferrara, 2005). For the case of the expected future income, objective and
subjective expectations are computed. Objective probability of being above the mean
income is computed according to the following expression:
10

Pr( J 10 decile) d = pdi


j =1

This expression represents the probability that an individual whose income is in


decile d in time t will move to deciles greater than or equal to J in time t+1. In addition
to these general probabilities, specific probabilities by group have also been computed,
as previously discussed (groups are defined by gender, age, and education level). The
subjective probability of being above the mean income is computed using the following
expression:
10

Pr( J 10 decile) s = psi


j =1

16

This expression represents the probability that an individual who thinks he is in


decile s at time t will move to deciles greater than or equal to J at time t+1.

Centro de Estudios Andaluces

Perception of meritocracy is measured through two variables, reflecting the


degree of agreement with the two following statements: people get rewarded for their
effort and people get rewarded for their skills. Ordered categorical responses are the
same as those of the dependent variable. Socio-demographic variables include those that
reflect differences in sources of income: gender (0 = Male, and 1 = Female), age
and age squared, education level3 (0 = No formal education, 1 = Primary School, 2
= Secondary School, and 3 = University), work status4 (1 = Employed, 2 =
Unemployed, and 3 = Not in the labor force), and self-employed (0 = No and 1 =
Yes). Two additional variables have been included: private or public sector in which
the person works (0 = Private sector and 1 = Public sector), and trade union
membership at the present time (0 = Not member and 1 = Member). I also consider
individuals beliefs about inequality as a control variable. Respondents were asked about
their level of agreement with the following statement: Differences in income are
necessary. The ordered categorical responses were: (1) strongly disagree, (2)
disagree, (3) neither agree nor disagree, (4) agree, and (5) strongly agree (the
response categories followed the opposite order in the questionnaire but are reversed
here for easier interpretation). A descriptive analysis of these variables is reported in
Table 2.

Statistical methodology

Given that the dependent variable is ordered categorical, I have used an ordered
probit model to estimate the effect of explanatory variables (Greene, 2008; McKelvey
and Zavoina, 1975). I assume that support for redistribution of individual i can be
defined by a latent variable yi*, which is a function of a vector of individual
characteristics xi:

yi* = ' xi + i
3

This variable has been recoded from original values. No formal education category includes those who
have no education and those who have not finished primary education. In the same vein, those who have
not completed an educational level have been assigned to the highest level that they have completed.
4
This variable has been recoded from original values. The category employed includes full-time and
part-time workers. The category not in the labor force includes those who are helping a family member,
students, retired people, housewife or houseman, the permanently disabled and others not in the labor
force.

17

We do not observe yi*, but a variable yi taking values 1 to 5 increasing in


individual support for redistribution. The probability that an individual observed that his

Centro de Estudios Andaluces

preference for redistribution yi is m can be expressed as the probability of yi* being


between the cut-off points m-1 and m:
P ( yi = m | xi ) = P (m-1 yi* < m |xi )

We can now compute the probability of being in the m category as:


P ( yi = m | xi ) = (m xi ) ( m-1 xi )

Assuming that the distribution of the error term i is normal, I estimate an


ordered probit model. For purposes of testing the robustness of the estimates an ordered
logit model has been estimated for each of the probits reported, even though the
coefficients are not displayed in Table 3. The sign, the significance level and the
magnitude of both estimates were quite similar.

5. Findings and discussion

This section presents the results of two empirical analyses of preferences for
redistribution. First, I will focus on the effect of the prospect of upward mobility on
preferences. Specifically, I test the hypothesis that the probability of being above
average income will shape preferences for redistribution, using both objective and
subjective evaluations of relative position on the income ladder. Second, I focus on the
effect of expected future income on preferences. In this case, I test the hypothesis that
preferences for redistribution depend on expected future income. Both objective and
subjective expected income are used in statistical analysis. In this case, present income
is not incorporated in the equation for technical reasons. That is, present income is
highly correlated with future income, given that the latter depends strongly on the
former. This would amount to assuming that preferences for redistribution are
independent of present income, such that they are only shaped by prospects of future
incomes, even though this assumption is neutral regarding the relative effect of
subjective and objective expectations. Several statistical specifications have been run in
order to test each of these hypotheses. Socio-demographic variables are always present
as controls.

Prospect of upward mobility

18

To test the hypothesis of upward mobility, five ordered probits have been

Centro de Estudios Andaluces

estimated. The results are displayed in Table 3. The first model includes only sociodemographic characteristics and the logarithm of real income (Column 1 in Table 3). In
this model, real income has a negative and significant effect on preferences for
redistribution. Among the other variables, only union membership has a significant
effect on preferences. This supports the general hypothesis of self-interest, according to
which preferences for redistribution will be lower as the income increases.
The second model incorporates the probability of moving to the seventh decile
or higher on the income distribution in the next period, which is equivalent to the
probability of being above mean income (Column 2 in Table 3). In this specification, I
focus on objective probabilities. In this model, both the logarithm of real income and
union membership exercise a significant effect. However, objective probability does not
have a significant impact on the dependent variable, and the sign is the opposite of that
expected. This finding has two different implications. Firstly, it supports the idea that
wealth is important in explaining preferences for redistribution. Secondly, the negative
sign of the variable probability of moving to seventh or higher deciles is due primarily
to co-linearity with the log of the real income. That is, given the high correlation
between current and future incomes, the prospects of being above the mean income in
the next period do not have a significant impact on preferences for redistribution.

Table 3: Preferences for redistribution. Ordered probit

Real income (ln)

[1]
-0.341b
(0.153)

Prospect of upward mobility


[2]
[3]
[4]
-0.585b -0.582b -0.556b
(0.259) (0.261) (0.263)

Expected future income


[6]
[7]
[8]

-0.502b
(0.255)

Expected income
(objective) (ln)
Expected income
(subjective) (ln)
Prob. 7-10 deciles
(objective)
Prob. 7-10 deciles
(subjective)
Effort

0.566
(0.482)

0.648
(0.485)
-0.782b
(0.394)

Skills
Female

[5]

-0.075
(0.082)

-0.079
(0.082)

-0.096
(0.083)

0.628
(0.487)
-0.789b
(0.398)
-0.134a
(0.049)
0.040
(0.052)
-0.090
(0.084)

-0.200
(0.291)
-0.800b
(0.397)
-0.134a
(0.049)
0.036
(0.052)
-0.085
(0.084)

-0.073
(0.082)

-0.408
(0.260)
-0.671b
(0.303)

-0.370
(0.262)
-0.675b
(0.307)

-0.087
(0.083)

-0.132a
(0.049)
0.038
(0.052)
-0.082
(0.084)

19

Age
Age squared

Centro de Estudios Andaluces

Primary school
Secondary school
University
Unemployed
Not in the labor
force
Self-employed
Public worker
Union member
Differences are
necessary
Threshold 1
Threshold 2
Threshold 3
Threshold 4
No. Obs.
McFadden's
pseudo-R2
Log likelihood

0.009
(0.012)
0.000
(0.000)
-0.014
(0.129)
-0.191
(0.165)
-0.269
(0.193)
0.167
(0.161)
0.133
(0.104)
-0.039
(0.122)
0.067
(0.124)
0.431a
(0.154)
-0.018
(0.034)
-4.935
(1.369)
-4.230
(1.366)
-3.759
(1.365)
-2.423
(1.363)
833
0.015

0.009
(0.012)
0.000
(0.000)
-0.001
(0.129)
-0.197
(0.165)
-0.290
(0.194)
0.158
(0.161)
0.135
(0.104)
-0.037
(0.122)
0.073
(0.124)
0.420a
(0.154)
-0.021
(0.034)
-6.967
(2.212)
-6.264
(2.212)
-5.792
(2.211)
-4.455
(2.209)
833
0.016

0.013
(0.013)
0.000
(0.000)
-0.011
(0.133)
-0.198
(0.169)
-0.264
(0.199)
0.217
(0.163)
0.156
(0.104)
-0.023
(0.124)
0.094
(0.126)
0.411a
(0.154)
-0.017
(0.034)
-7.035
(2.232)
-6.333
(2.231)
-5.868
(2.231)
-4.521
(2.228)
817
0.019

0.014
(0.013)
0.000
(0.000)
0.043
(0.135)
-0.163
(0.171)
-0.203
(0.202)
0.200
(0.164)
0.143
(0.105)
-0.025
(0.125)
0.080
(0.127)
0.420a
(0.155)
-0.021
(0.035)
-7.033
(2.245)
-6.329
(2.244)
-5.864
(2.243)
-4.500
(2.240)
805
0.024

0.013
(0.013)
0.000
(0.000)
-0.014
(0.132)
-0.230
(0.168)
-0.251
(0.200)
0.244
(0.163)
0.166
(0.105)
-0.016
(0.124)
0.062
(0.127)
0.434a
(0.155)
-0.018
(0.035)
-2.346
(0.370)
-1.639
(0.362)
-1.175
(0.360)
0.183
(0.358)
805
0.021

0.009
(0.012)
0.000
(0.000)
-0.029
(0.128)
-0.204
(0.165)
-0.279
(0.194)
0.178
(0.161)
0.138
(0.103)
-0.036
(0.122)
0.061
(0.124)
0.437a
(0.154)
-0.017
(0.034)
-6.374
(2.262)
-5.667
(2.258)
-5.196
(2.257)
-3.861
(2.256)
833
0.014

0.013
(0.013)
0.000
(0.000)
-0.027
(0.132)
-0.191
(0.169)
-0.241
(0.199)
0.231
(0.163)
0.161
(0.104)
-0.018
(0.124)
0.086
(0.126)
0.434a
(0.154)
-0.014
(0.034)
-11.430
(3.2769
-10.726
(3.273)
-10.261
(3.272)
-8.916
(3.269)
817
0.018

0.013
(0.013)
0.000
(0.000)
0.027
(0.134)
-0.160
(0.172)
-0.184
(0.202)
0.215
(0.164)
0.150
(0.105)
-0.018
(0.125)
0.070
(0.127)
0.442a
(0.155)
-0.018
(0.035)
-11.339
(3.315)
-10.633
(3.312)
-10.168
(3.311)
-8.807
(3.309)
805
0.023

-1013.7

-1013.0

-989.9

-971.9

-974.2

-1014.3

-990.7

-972.7

Notes: a, b, and c indicate significance level at 1%, 5%, and 10%, respectively. Standard errors in
brackets.
Source: International Social Survey Program and European Community Household Panel.

The third model incorporates the subjective probability of moving to the seventh
decile or higher on the income distribution in the next period. This is equivalent to the
subjective probability of being above mean income, given the self-assessed position on
the income ladder (Column 3 in Table 3). We see that the subjective probability has a
significant and negative effect on preferences for redistribution. At the same time, no
changes occur in the magnitude or the sign of other coefficients.
The full model includes probabilities of being above mean income and
perceptions of meritocracy (Column 4 in Table 3). Two additional variables are
included at this step: the feeling that people are rewarded for their skills and the feeling

20

that they rewarded for their effort. The latter variable has a negative, significant and
strong effect on the preferences for redistribution, while the former has no significant

Centro de Estudios Andaluces

effect. Those who believe that people get rich in proportion to their effort are inclined to
accept inequalities and therefore reject redistribution. Interestingly, the belief that innate
skills are the source of income differentials has no significant effect at all. This means
that people accept inequalities caused by circumstances that depend on individuals will
(such as the effort they exert), but not on innate circumstances, which are beyond their
control. Thus, it can be concluded that the perceived degree of the societys openness is
important in explaining preferences for redistribution. People usually think that
openness makes it possible for individuals with different backgrounds and different
levels of abilities to be rewarded for their personal effort.
In Columns 2 to 4 in Table 3, a counterintuitive result appears to hold: the effect
of the objective probability of being above the mean income in the next period on
preferences for redistribution seems to be positive, though not significant. Is this
logical? Statistical tests suggest that this positive effect is primarily the result of colinearity between this variable and the log of the real income. Since we do not have real
income data in the ISSP survey, but only the income decile to which each household
belongs, we must infer household income using the mean of income decile. Given the
high correlation between origins and destinations in the income transition matrix,
prospects of being above the mean income and real income are closely correlated. This
would explain the positive sign of the first variable in Columns 2 to 4. In Column 5, I
checked a new specification in which the log of real income is not included. Not
surprisingly, the coefficient of the prospect of being above the mean income changes to
a negative effect, given that the effect of current income is not taken into account.
However, the coefficient is not statistically significant. Note that the rest of the
coefficients do not change (either sign or magnitude) from Column 4 to Column 5,
given that co-linearity affects only these two variables.
These results are consistent with those of Alesina and La Ferrara (2005). As in
their study, both of the probability of being above the mean income in the next period
and the feeling that people are rewarded for their effort have a significant impact on
preferences for redistribution (rewards based on innate ability remain nonsignificant).
However, only subjective probabilities seem to have a significant impact. While one can
argue about the effect of co-linearity between objective expectations and current income
in these results, it is clear that subjective expectations have a consistent effect on
21

preferences for redistribution. Moreover, removing either current income or objective


probability from the equation only influences the effect of these two variables. Only

probabilities remain nonsignificant. These findings support the hypothesis that


preferences for redistribution are shaped by expectations of upward social mobility but
that the subjective probabilities of being upwardly mobile are the most significant
factor. Statistical results explain well why some rich people still favor redistribution,
given that they view themselves as very close to the average income earner. As
Runciman (1966) argued, people tend to evaluate their position in the social structure
through the social group to which they belong. And since social networks of
acquaintances are relatively homogeneous in terms of social status and income, people
may think they are close to the average for the society as a whole as long as they are
close to their networks average.

Figure 2: Preferences for redistribution and the probability of being above the mean income

Preferences for redistribution (strongly agree)


.15
.2
.25
.3
.35

Centro de Estudios Andaluces

current income seems to have a significant effect in the full model, while objective

.2
.4
.6
Probability of being above mean income
Objective expectation

.8

Subjective expectation

Source: International Social Survey Program.

Further analysis is required to show the influence of subjective expectations of


social mobility on preferences for redistribution. Figure 2 graphs the probability of
being above the mean income in the next period (both objective and subjective
22

probabilities) against the probability of strongly agreeing with the statement that the
government must reduce differences in income (probabilities have been computed

Centro de Estudios Andaluces

from estimates in Column 5 of Table 3). The graph shows a negative relationship
between probabilities of upward social mobility and preferences for redistribution.
However, the relationship between subjective probability and preference is much
sharper. As we can see, the variation in preference for redistribution is sharper for
subjective expectations than for objective expectations. If we focus on subjective
expectations, we see that, for those who have a less than 10 % probability of being
above the mean income in the next period, the probability of strongly agreeing with the
statement is 36 %. At the other extreme, those whose probability of being above the
mean income is close to 80 %, the probability of strongly agreeing is approximately 17
%. If we focus on objective expectations, variation in probabilities is much lower, from
32 % to 27 %.
The socio-demographic characteristics of the individuals do not seem to have a
significant impact on preferences for redistribution in the estimated models, although
the direction of the coefficients is as expected. For instance, the relationship between
education level and demand for redistribution is negative but not significant. Further,
public workers prefer a higher redistribution, but the effect of working status is
nonsignificant. Only union membership has a significant impact on preferences. Union
members demand a higher level of redistribution. This seems to reflect an ideological
predisposition in favor of redistribution of those who belong to workers organizations.
It also reflects the influence of these organizations in shaping a class consciousness,
expressed in political preferences.

Expected future income

In the second step, I will explore the relationship between expected future
income and preferences for redistribution. Although we should expect to obtain similar
results, it is important to test whether the previous findings are robust enough to be
found using different model specifications. Three ordered probits have been estimated.
The first model includes the natural log of the objective expected future income as
explanatory variable as well as controls (Column 6 in Table 3). The estimates show that
objective expectations have a significant impact on preferences in this model. Those
who expect to have a lower income (based upon objective probabilities) will demand
23

higher redistribution. This finding reinforces the hypothesis of self-interest and the idea
that people care about the future.

subjective expected future income (Column 7 in Table 3). Both variables have a
significant impact on preferences for redistribution, showing that both expectations
matter, although the coefficient for subjective expectations is higher. The last model
includes expected future income (both objective and subjective) and meritocracy as
explanatory variables (Column 8 in Table 3). The results show that, when I control for
meritocracy, objective expected future income is no longer significant, although the
impact of meritocracy is similar to the results referred to in the first step. Those who
think that people are rewarded for their effort oppose redistribution. However, peoples
opinion about the relationship between rewards and innate skills does not generate
significant differences in preferences for redistribution. The socio-demographic
characteristics of the individuals remain nonsignificant, although union membership is
significant.

Figure 3: Preferences for redistribution and expected income

Preferences for redistribution (strongly agree)


.2
.25
.3
.35
.4

Centro de Estudios Andaluces

The second probit includes the natural log of both the objective and the

8.6

8.8

9
Expected income (ln)

Objective expectation

9.2

9.4

Subjective expectation

Source: International Social Survey Program.

24

These findings agree fundamentally with those found in the first step, and they
support the idea that preference for redistribution depends on expectations about future

Centro de Estudios Andaluces

income. Nevertheless, when I take into account the combined effect of subjective and
objective expectations, as well as meritocracy, only subjective expectations remain
statistically significant. Figure 3 shows the relationship between the natural log of
expected future income (both objective and subjective expectations) and the probability
of strongly agreeing with the statement that the government must reduce differences in
income. As we can see, the variation in preferences for redistribution is sharper for
subjective expectations than for objective ones. If we focus on subjective expectations,
for those who expect to have an income below Euro 5,500 in the next period, the
probability of strongly agreeing with the statement is 38 %. At the other extreme, for
those who expect to have an income above Euro 12,000 the probability is around 19 %.
If we focus on objective expectations, variation in probabilities is lower: from 35 % to
24 %.

Table 4: Preferences for redistribution for different transition matrices

Prob. 7-10 deciles


(objective)
Prob. 7-10 deciles
(subjective)
Expected income
(objective) (ln)
Expected income
(subjective) (ln)

Prospect of upward mobility


Gender
Age
Education
-0.105
-0.361
-0.326
(0.292)
(0.288)
(0.315)
-0.814b
-0.790b
-0.783b
(0.397)
(0.396)
(0.398)

Expected future income


Gender
Age
Education

-0.324
(0.257)
-0.679b
(0.307)

-0.438c
(0.229)
-0.680b
(0.305)

-0.662b
(0.291)
-0.633b
(0.307)

Notes: a, b, and c indicate significance level at 1%, 5%, and 10%, respectively. Standard errors in
brackets.
Source: International Social Survey Program and European Community Household Panel.

The results previously found are very robust for different specifications. As in
the case of Alesina and La Ferrara (2005), different specifications of the income
transition matrix have been tested. Table 4 presents these results. For the sake of clarity,
only relevant coefficients are depicted, given that the rest of the coefficients are roughly
the same as those in Table 3. The first three columns in Table 4 can be compared to
Column 5 in Table 3, and the last three columns in Table 4 can be compared to Column
8 in Table 3. The columns in Table 4 reflect the impact of using different income
transition matrices by groups. In the first column, objective probabilities of being above
25

mean income in the next period have been computed, taking into account differences by
gender. In the fourth column, expected future income has been computed, taking into

Centro de Estudios Andaluces

account differences by gender. Columns 2 and 5 reflect differences in age, and Columns
3 and 6 reflect differences by education level.
As we can see in Table 4, the differences between the specific income transition
matrices and the general matrix are not sufficiently large. As we would expect, the
subjective probability of being above mean income and subjective expected income
have roughly the same effect on different specifications. This is true, since subjective
expectations are not affected by the computation of objective expectations. However,
changes in the effect of objective probabilities are also small when compared through
different transition matrices. We must also note that we cannot fully trust transition
matrices by gender, since the proportion of households headed by women is very small
in the Spanish ECHP sample. On the other hand, coefficients by age and education are
roughly the same. The only major difference is that objective expected income becomes
statistically significant when I compute income transition matrices by age or education
level.

6. Conclusions

Meltzer and Richard (1981) predict that preference for redistribution will depend
on relative position on the income ladder. However, further empirical research has
shown that the link between income and preferences is rather weak, at both the
aggregate and the individual level. One strand of research has recently addressed this
problem by incorporating the expectations of future income (Alesina and La Ferrara,
2005, Benabou and Ok, 2001). Since relative income positions are dynamic and
redistributive schemes are assumed to be effective over a sufficiently long period,
people care not only about their present income but also about expected future income
levels. Another strand of research has proved that meritocracy and beliefs about social
structure have an impact on preferences for redistribution (Benabou and Tirole, 2006).
People may accept inequalities to some degree if they believe that these inequalities are
the product of different levels of effort. None of these previous works has considered
seriously the fact that people have rather limited knowledge of their own position in the
social structure. According to relative deprivation theorists, people deduce their relative
position from comparisons they make in everyday life (Runciman, 1966). Given that the
26

range of comparisons is limited to those with whom they interact frequently and that
social networks are relatively homogeneous in terms of social status and income, people

Centro de Estudios Andaluces

tend to underestimate the extent of the inequalities in their society as a whole. People
thus have a propensity to view themselves as close to the average income earner.
The main goal of this paper has been to confirm the findings of previous studies
about the relationship between expectations of social mobility and preference for
redistribution in the Spanish case, where income inequality and support for
redistribution are high (as compared to other European countries). Following the
framework proposed by Alesina and La Ferrara (2005) for the analysis of the
relationship between social mobility and preferences for redistribution, though in a
reduced scope, this research departs from previous studies of this topic by
differentiating between objective and subjective probabilities of mobility. The former
are computed as if people were aware of their relative income. The latter take into
account the place where people locate themselves on the income ladder. Similarly,
objective and subjective expected incomes have been computed using the same set of
assumptions. The difference between objective and subjective expectations takes into
account that individuals usually place themselves in the middle of the income
distribution, since rich people systematically underestimate their relative position and
poor people overestimate theirs.
Empirical results show meritocracy and income are the main forces shaping
preferences for redistribution in Spain. It has been also found that subjective
expectations of social mobility have a stronger impact on preferences for redistribution
than objective expectations. According to these findings, it can be said that is not where
you are (or where you should be in the future), but where you think you are (or where
you think you will be) that primarily determines attitudes towards redistribution. A
second analysis has found that both objective and subjective expected future income
have a significant impact on preferences. However, if both variables are incorporated
into the equation at the same time (along with meritocracy at the other controls), only
subjective expectations remain significant.
Taken as a whole, these results imply that expectations of social mobility are
crucial in explaining preferences for redistribution. However, they also reveal that
conventional measures of expectations are not accurate. What is most important, what
affects preferences for redistribution, is not the objective probability of being mobile,
but the subjective expectations formed through interactions in daily life. The most
27

plausible explanation for these findings comes from relative deprivation theory.
Because people tend to estimate their position on the income ladder from comparisons

Centro de Estudios Andaluces

made within a narrow range of income, they feel themselves to be similar to the average
income earner, although it is quite possible that that may overestimate their true
probability of upward mobility. This research has shown convincingly that subjective
expectations really matter and that the measurement of expectations for social mobility
must take subjective beliefs into account more carefully. Nevertheless, further research
must be undertaken in order to investigate the mechanisms through which these
expectations of social mobility are formed at the individual level, and what are the
relevant reference groups.

References

Alesina, A., and E. La Ferrara (2005). Preferences for redistribution in the land of
opportunities, Journal of Public Economics, 89(5-6), 897-931.
Arts, W. and Gelissen, J. (2001). Welfare states, solidarity and justice principles: Does
the type really matter? Acta Sociologica, 44(4), 283-299.
Benabou, R. and E. A. Ok (2001). Social Mobility and the Demand for Redistribution:
The POUM Hypothesis, Quarterly Journal of Economics, 116(2): pp. 447-487.
Benabou, R. and J. Tirole (2006). Belief in a Just World and Redistributive Politics,
Quarterly Journal of Economics, 121: 699-746.
Cant, O. (2000). Income Mobility in Spain: How Much is There?, Review of Income
and Wealth, 46(1): 85-102.
Carabaa, J. (1999). Dos Estudios sobre Movilidad Intergeneracional. Madrid:
Fundacin Argentaria.
Checchi, D. and A. Filippin (2004). An Experimental Study of the POUM
Hypothesis, Research on Economic Inequality, 11: 115-136.
Corneo, G. and H. P. Grner (2002). Individual Preferences for Political
Redistribution, Journal of Public Economics, 83: 83-107.
Davis, J. A. (1959). A Formal Interpretation of the Theory of Relative Deprivation,
Sociometry, 22: 280-296.
Eurostat (2000). Eurostat Yearbook. Luxembourg: Office for Official Publications of
the European Communities.

28

Evans, M. D. R. and J. Kelley (2004). Subjective Social Location: Data from 21


Nations, International Journal of Public Opinion Research, 16: 3-38.

Centro de Estudios Andaluces

Fehr, E. and K. M. Schmidt (1999). A Theory of Fairness, Competition and


Cooperation, Quarterly Journal of Economics, 114: 817-868.
Franzese, R. J. (2002). Macroeconomic Policies of Developed Democracies.
Cambridge: Cambridge University Press.
Greene, W. H. (2008). Econometric Analysis. New York: Prentice Hall.
Hirschman, A. O. (1973). The changing tolerance for income inequality in the course
of economic development, Quarterly Journal of Economics, 87: pp. 544-566.
International Social Survey Programme (1999). Social Inequality III (ISSP 1999).
Jaime-Castillo, A. M. (2000). Las Actitudes de los Espaoles ante las Polticas
Sociales, Revista Internacional de Sociologa, 26: 125-157.
Lane, R. E. (1959). The Fear of Equality, American Political Science Review, 53: pp.
35-51.
Linos, K., and West, M. (2003). Self-interest, Social Beliefs, and Attitudes to
Redistribution. Re-addressing the Issue of Cross-national Variation, European
Sociological Review, 19(4), 393-409.
Lockwood, D. (1966). Sources of variation in working class images of society,
Sociological Review, 14: 249-263.
McKelvey, R. D. and W. Zavoina (1975). A Statistical Model for the Analysis of
Ordinal Level Dependent Variables, Journal of Mathematical Sociology, 4:
103-120.
Meltzer, A. H. and S. F. Richard (1981). A Rational Theory of the Size of
Government, Journal of Political Economy, 89(5): 914-927.
Piketty, T. (1995). Social mobility and redistributive politics, Quarterly Journal of
Economics, 110: pp. 551-584.
Pollis, N. P. (1968). Reference Groups Re-examined, British Journal of Sociology,
29: 300-307.
Rainer, H. and T. Siedler (2008). Subjective Income and Employment Expectations
and Preferences for Redistribution, Economic Letters (forthcoming).
Ravallion, M. and M. Lokshin (2000). Who Wants to Redistribute? The tunnel effect
in 1990s Russia, Journal of Public Economics, 76: 87-104.

29

Rose, D. (2006). Social Comparisons and Social Order: Issues Relating to a Possible
Re-Study of W. G. Runcimans Relative Deprivation and Social Justice, ISER

Centro de Estudios Andaluces

Working Paper. Colchester: University of Essex.


Runciman, W. G. (1966). Relative Deprivation and Social Justice: A Study of Attitudes
to Social Inequality in 20th Century England. London: Routledge Kegan Paul.
Svallfors, S. (2006). The Moral Economy of Class. Class and Attitudes in Comparative
Perspective. Stanford: Stanford University Press.

30