The Many Mexicos:
Income Inequality and Polarization in Urban Mexico During the 90s *
Mabel Andalón L.
Center of Statistical Resources
Presidency of Mexico
and
Luis F. López-Calva
Universidad de las Américas-Puebla and El Colegio de México
First version
May, 2002
Do not quote
Abstract
The Zapatista revolt in the Mexican state of Chiapas in 1994, which started the same day in which
the NAFTA was officially implemented in the North American Region, made clear the Mexico was
not one homogeneous country, but one characterized by important differences in development
across regions. This paper analyzes the evolution of income inequality and polarization in 34 urban
areas grouped into regions with similar characteristics. The richer regions, North and Center, also
show higher degrees of inequality and lower polarization. The most polarized region is the South,
also the poorest. Even more interesting, the polarization between regions has increased during the
nineties. This may explain, at least partially, the increasing level of conflict in the south and the
tension between regions during the negotiation of federal budget decentralization. The polarization
measures used are those developed by Foster and Wolfson (1993) and Wolfson (1994), Esteban
and Ray (1994), Kanbur and Zhang (1999), Tsui and Wang (1998), and Esteban, Gradín and Ray
(1999). It is shown that polarization measures provide information that inequality measures do not,
adding important insights for the analysis of regional income dynamics. Despite the fact that total
income inequality has decreased in Mexico during the nineties, other distributional dimensions open
important policy questions. First, labor income inequality has increased, showing a skill bias after
trade liberalization. Also, even though inequality is lower in the southern part of the country,
polarization measures consistently show that the south has moved farther away from the rest of the
country.
Keywords : Mexico, Inequality, Polarization, Regional Development.
JEL codes:
*
Version prepared for the Cornell-LSE-WIDER Conference on Spatial Inequality, London, June 28, 29, and
30th , 2002.
I.
Introduction: The Many Mexicos
The Zapatista revolt in the Mexican state of Chiapas in 1994, which started the same day in
which the NAFTA was officially implemented in the North-American trade block, made clear that
Mexico was not one homogeneous country, but one characterized by important differences in
development across regions. Historically, Mexico has been characterized by the coexistence of
highly competitive, outward-oriented industrial sectors, and backward agricultural, poorly
developed regions. Somehow public policy decisions and the political economy behind them has
not promoted stronger linkages between those sectors and regions. This paper analyzes the
evolution of income inequality and polarization in the Mexican urban areas grouped into regions
with similar characteristics.
During the late eighties and the nineties, Mexico went through a process of structural reform
which included privatization of state-owned enterprises, trade liberalization, and decentralization of
fiscal expenditures. The macroeconomic results have been overall positive, though in 1994 a major
financial crisis brought the process to a halt. Mexican GDP fell 6.5% and inflation reached 52%
during 1995. The recovery, however, was relatively fast (graph 1). As opposed to the debt crisis of
the eighties, in which Mexico was still a closed economy with heavy state intervention, the recovery
of positive growth rates took only 15 months, instead of several years.
Graph 1
Annual Growth Rate, GDP
(1994=100)
Debt crisis
1994 crisis
15
5
0
-5
-10
-15
Years
2000
1998
1996
1994
1992
1990
1988
1986
1984
1982
-20
1980
Percentage
10
Overall inequality in the country, considering either total income or total expenditures, has
shown a slightly downward trend during the period (graph 2 and table 1).
Inequality in Mexico
1989-2000
0.6
0.55
0.5
Ingreso Gini
Ingreso Theil
0.45
Gasto Gini
Gasto Theil
0.4
0.35
0.3
1
2
3
4
5
6
Source: Own calculations using data from the Income -Expenditure Surveys.
This paper will show that such trend is not sustained when we take only labor income in
urban areas (table 2). Several papers have shown that the skill premium in Mexico rose after trade
liberalization, causing an increase in labor income inequality during the nineties (World Bank,
2000, Hanson, 2002). The question analyzed here, however, is whether the reform process during
that period has also an impact on the regional inequality and polarization pattern, as has been stated
by analysts who suggest that this phenomenon would be the main cause of the political problems in
the southern states, mainly the states of Chiapas and Guerrero (De La Torre, 1996).
Poverty, on the other hand, stayed basically constant between 1992 and 2000, mainly
because of the increase in the FGT index after the 1995 crisis, which was not offset by the recovery
of economic growth in the following years.
Year
Poverty Gap
FGT(2) Index
Moderate
Extreme
Moderate
Extreme
1992
42.1
32.0
22.9
14.7
1994
41.3
32.3
22.3
14.7
1996
47.3
36.5
27.9
18.3
1998
47.3
39.0
28.3
20.6
2000
43.2
34.0
24.1
16.6
Source: Authors´calculations from household survey data.
In this paper we analyze the trends in regional income differences, emphasizing the
polarization and inequality between the southern part and the rest of the country. An important
constraint to be considered in the analysis is that the income-expenditure survey in Mexico
(ENIGH) does not allow for regional comparisons because it is statistically representative only to
make inferences at the rural-urban level of disaggregation. That has forced us to use the Urban
Employment Survey (ENEU), which is representative at the level of urban labor markets for
between 33 and 43 different areas.
Table 1
Inequality Using Total Income and Total Expenditure
Year
Income
Gini
Expenditure
Theil
Gini
Theil
1989
0.515
0.584
0.452
0.389
1992
0.526
0.568
0.447
0.366
1994
0.498
0.493
0.431
0.346
1996
0.500
0.501
0.423
0.337
1998
0.524
0.567
0.438
0.360
2000
0.503
0.495
0.431
0.341
Table 2
Inequality Using only Labor Income
Theil mean
year
Gini
Theil
log
entropy
deviation
measure
measure
1987
0.369
0.260
0.255
1988
0.395
0.380
0.289
1989
0.398
0.334
0.275
1992
0.442
0.448
0.341
1994
0.444
0.399
0.345
1996
0.469
0.458
0.385
1998
0.451
0.418
0.358
2000
0.441
0.389
0.341
If one takes the Human development Index by state in Mexico, the most developed entity is
Mexico City, whose HDI compares with that of a rather dynamic European economy, namely,
Ireland. The least developed region, however, is Chiapas, whose HDI is at the level of Albania, the
poorest Eastern European country (Esquivel and Lopez-Calva, 2002). Such is the magnitude of the
development gap within Mexico.
The next section reviews briefly the polarization and inequality measures used in the paper.
The third section reviews the evolution of regional inequality and polarization in Mexico during the
nineties, whereas the fourth section looks at evidence of convergence since the sixties. Finally, we
offer some conclusions and policy recommendatio ns.
II. Brief Discussion on Inequality and Polarization Measures
In this section we present a brief overview of the inequality and polarization measures used
throughout the paper. This is especially useful in the case of the polarization measures, which are
not as common as the inequality measures and whose discussion is scattered in the literature
without any comprehensive analysis.
II.1 Inequality Measures
Labor income inequality is defined as the dispersion of salaries with respect to the mean. As
it is well-known in the literature, inequality measures can be divided into two categories: normative
and positive measures (Chakravarty and Majumdar, 2001). The former are indicators that take as a
reference the notion of social welfare, for example, Atkinson’s measures. Positive measures, which
will be used here, only require consistency with the Lorenz criteria. Examples of those measures are
the Gini Coefficient and the Theil Index.
II.1.1 Gini Coefficient
The Gini coefficient calculates the differences between all the pairs of individuals and adds
them up. If the Gini coefficient is normalized, the expression can be written as:
G =
1
µ N ( N − 1)
∑
∑j
i> j
yi − y j
(1)
Where,
µ Is the mean income of the distribution
N is the number of observations
yi , yj is the ith (jth ) income
This index is more sensitive to transfers in the middle of the distribution. It is well known
that when all the indiv iduals of the population share the same income µ, the Gini coefficient is
equal to zero and there is no evidence of inequality in this distribution. The highest inequality
occurs when there is one person that has income Nµ and everybody else has no income. Therefore,
there are N-1 absolute differences each one equal to Nµ, (The Gini coefficient is equal to one).
Deaton (1997) has proposed another expression to be used for big samples. It is expressed
as follows:
N
N +1
2
G=
−
∑ ρi yi
N − 1 N ( N − 1)µ i =1
Where ,
(2)
ρi is the range of the ith individual in the distribution. The richest person gets a range equal to one
meanwhile the poorest gets a range of n. The latter is the Gini formulation used in this paper.
II.1.2 Theil Index
When the parameter c of the generalized entropy index is equal to 1 or 0, we have the Theil
index. For the natural logarithm of incomes it is expressed as:
1 N y y 
TH =  ∑ i ln i 
 N  i=1 µ  µ 
( 3)
Where,
N is the population
yi is the ith income
µ Is the mean income of the population
Just as a reminder, the upper bound of the Theil index rises as N does. The main advantage
of this measure is that when the individuals are clustered in different groups it can be decomposed
into their “between” and “within” components. 1 This index allows us to identify the contribution of
different components to total inequality.
II. 2 Polarization Measures
II.2.1 Esteban and Ray (1994) Measure
The concept of polarization has to do with the weakening of the middle class. Several
measures have been proposed thus far, all of them trying to capture the concept in a more robust
manner. Most of the indices proposed are positive.
The polarization measure proposed by Esteban and Ray (1994) (ER hereafter) has a more
normative flavor to it, for it claims to capture some behavioral implications through the axiomatic
derivation. The idea starts by defining “effective antagonism” of the society, based on the addition
of two behavioral functions: identification and alienation. The first is an attitude towards the people
of the same group or income class. Polarization rises as within-group homogeneity, given that in
Esteban and Ray (1994) the identification is an increasing function of the number of individuals in
each group. The other effect, alienation, is what a person feels for somebody who is “far away” in
the distribution. Polarization rises when the heterogeneity among groups rises. Indeed, their axioms
1
Such decomposition is only possible in the case of the Gini coefficient when the income of those groups do
not overlap.
imply that regressive transfers which cross the median of the distribution increase polarization,
whereas it decreases if progressive transfers take place in each half of the distribution, divided by
the median. The latter principle would be also captured, in a simpler manner, by Foster and
Wolfson (1992).
The main principle s in which the axioms of Esteban and Ray’s measure are based are as
follows:
1. The polarization rises when an extreme class gets away from the central class only if the other
extreme does not get any closer. The polarization rises when the heterogeneity between groups
rises.
2. When the middle class gains mass polarization declines.
3. Maximum polarization is obtained when the distribution is partitioned into two internally
homogeneous groups, each one with half of the population, located at the extreme of the
distribution.
The polarization measure obtained from Esteban and Ray is expressed in equation 4:
ki
kj
ER = A∑∑ f ( y i )1+α f ( y j ) y i − y j
( 4)
i =1 j =1
Where,
A is a normalization scalar
k is the number of groups
f(y i) represents the population share of the ith group
y i ,j is the natural logarithm of the mean income of the ith (jth)
α Is the degree of polarization sensitivity an is in the range of [1,1.6]
A and α are determined arbitrarily. The greater the value of α the greater deviation
of the ER index from the Gini coefficient because the concentration of the groups becomes
more important. When α=0, ER is equal to the Gini coefficient except for the income
transformation used (natural logarithm). Using this measure, polarization is based on the
distance between the mean income of the groups, the size of the clusters, and the degree of
sensitivity to polarization.
II.2.2 Extension in Esteban, Gradín and Ray (1999)
Esteban, Gradin , and Ray (1999) proposed and extension of the original ER measure (EGR
hereafter). The extension deals with the following problems of the ER measure:
1. Sensitivity to the formation of groups, which almost always depends on the availability of the
data.
2. Lack of cross identification between groups (there is no identification between individuals of
different groups although they can be located together on the distribution).
3. Also, when the size of each group is normalized, i.e. when we use deciles as groups, the α
parameter is neutralized.
The new measure is the ER measure corrected by an error term based on grouping. EGR
has the same advantages of ER but it is used when a density function of the income distribution can
be previously calculated.
An F distribution composed by k groups is assumed. A simplified representation of F is set
as the partition ρ=(zo , z1, z2 ,...,zk ; y 1 , y 2 ,... , y k ; p1 ,p2 ,...,pk) which delimits k groups. The ith group
is defined as a proportion p j of individuals. Their salary is in the [zi-1,zi] interval, with y i as mean
salary. When ρ is used to represent F an approximation error emerges, ε(F;ρ). The error is defined
as the income dispersion degree in groups measured with the Gini Coefficient.2 :
ε (F ; ρ ) = G( F ) − G( ρ )
( 5)
Where,
G(F) is the Gini coefficient
G(ρ) is the Gini coefficient assuming that all the groups are internally homogeneous
Thus, the error term is the difference between inequality in the society and a hypothetical
inequality that would result by assuming that the individuals within each group all have the mean
income of the cluster. The error term can be interpreted as the lack of internal identification. The
2
An approximation error exists because, when we form income classes, it is impossible to have individuals
totally identified (alienated) with others of the same (different) class.
greater the internal dispersion in a group, the less the identification and, in turn, the lower the
polarization.
The EGR measure calculates the polarization of the F distribution as follows:
EGR( F ;α , β , ρ ) = ER (α , ρ ) − βε (F , ρ )
(6)
Where,
β Represents the lack of internal identification
ER represents the Esteban and Ray (1994) measure applied in ρ.
The EGR index has a maximum value of two and a minimum value that is not less than -β.
It depends on the ρ chosen, as well as on β itself.
II.2.3 Foster and Wolfson´s measure
Foster and Wolfson (1992) and Wolfson (1994) proposed a polarization index (W) while
discussing the conceptual differences between polarization and inequa lity. The proposed measure is
as follows:
W =
2(2T − Gini)
mtan
( 7)
Where,
T = 0.5 − L( 0.5)
(8)
L(0.5) is the income share of the bottom half of the population (the population is divided into two
groups based on median income), and mtan is the ratio of the median to the mean. Wolfson (1994)
normalizes the index arbitrarily so that it takes values between 0 and 1. Graph 1 shows the intuition.
The light gray area is the polarization index, once normalized.
Graph 1
Figure 1 Wolfson’s Measure
Esteban, Gradín and Ray (1999) note that the Foster and Wolfson´s measure is a
transformation of the EGR index if the adjacent groups are of the same size.
An alternative way of expressing W according to Zhang y Kanbur (1999) is:
2 (µ * − µ L )
W =
m
(9)
where,
µ* Is the corrected mean income: µ multiplied by (1-Gini),
µ L is the mean income of the first half of the population,
m is the median of the population.
Maximum polarization would thus occur when half the population has zero income and the
other half has twice the mean.
II.2.4 Tsui and Wang (1998) Measure
Tsui and Wang (1998) generalized a new class of indices based on the Wolfson measure.
They use the two axioms of the partial ordering: increasing in the polarization and increasing on the
spread. Their measure is expressed as:
y −m
θ  k
TW =  ∑ f ( y j ) j
m
 N  j =1
r
(10)
Where,
y
j
Is the mean income of ith (jth ) group,
θ Is a positive scalar,
N is the total population,
k is the number of groups,
f(y j) is the population share in the jth group,
m is the median income,
r is a coefficient in the range of (0,1).
II.2.5 The index of Zhang and Kanbur
Zhang and Kanbur (1999) suggested a new way to look at polarization and applied it to the
evolution of regional disparities in China, mainly between inland and coastal areas. One of the main
motivations to propose a new index was the fact that the most popular polarization indices moved
very closely with inequality indicators. They considered a different way of thinking about
polarization, arguing that this new way of measurement came closer to capturing the spirit of many
of the concerns on this area.
This index (ZK hereafter) is based on the decomposition of the inequality measures,
specifically the Theil Index, discussed above. On the expressions below the Theil index is presented
as a function of its components.
k
k
T = T B + T w = ∑ Pj R j ln R j + ∑ Pj R j T j
j =1
(11)
j =1
n
1 j yi
y
T j = ∑ ln i
n j i =1 µ j µ j
(12)
where
T B Represents the inequality between groups which indicates the distance between their mean
value,
T w is the intra-group component, it means, the spread of the sub-group distributions k is the
number of groups, each one with n j individuals j=1...k ,
Pj is the population share in each group pj =nj /n,
Rj is the ratio of the mean income of each group to the distribution mean income Rj = µj /µ,
Yi is the individual income, and
µj is the mean income of each group.
The polarization index of Zhang and Kanbur (1999) is the ratio of inter-group inequality to
intra-group inequality. It captures the average distance between the groups in relation to income
differences seen within groups.
TB
ZK =
Tw
(13)
One problem with this index is its variability. The fact that between inequality components
tend to be relatively stable, while within components vary widely, makes the index very unstable. In
principle, the ZK index is more suitable for regional polarization in cases where the between
component is more important.
III. Mexico During the 90s
Mexico went through a deep structural reform process during the nineties. Privatization,
trade opening, financial liberalization, and the elimination of community-property schemes in the
rural sector are among the main components of such reform. The process was a result of important
political changes and led to political adjustments between groups, transforming the configuration of
the political forces in Mexico (Esquivel and Tornell, 1998). In January 1994 a new guerrilla
movement in Chiapas forced the government and public opinion to re-focus some policy priorities
towards the southern part of Mexico, the least developed region of the country. An imperfect,
though objective, indicator of development, namely the HDI, shows an enormous difference in
development between the northern, central, and southern parts of Mexico (see annex 1). Arguably,
the latter differences are not the result of the reform process, but the secular result of differentiated
policies in terms of public services provision and pricing, educational policies, local politic s, ethnic
differences, and historical factors. Given those initial conditions in the early nineties, a relevant
question is whether the new development strategy proposed during the nineties considered such
differences in order to promote the strengthening of the linkages between sectors and regions in
order to narrow the development gap in the country.
In order to answer the question, we use labor income data for more than thirty urban areas
throughout Mexico. The main reason to use labor data from the urban employment survey, instead
of total income or expenditure data from the income-expenditure survey, is the impossibility to
divide the country into regions in a statistically valid manner in the latter. On the other hand,
Hanson (2002), Lustig (2000), and others have shown that labor income inequality is an important
component of total inequality in Mexico and is directly linked to the reform process. The gap
between skilled and unskilled wages has increased during the last decade in Mexico (see graph 2).
Also, regional differences in investment and growth are related to transport costs, poor
infrastructure development, and educational levels (Davila, et. al., 2002).
Graph 2
Index (1984 = 100)
Unskilled
Skilled
130
120
110
100
90
80
70
19
19
90
84
Source: Lustig (1998).
For the purpose of this study, the urban areas are grouped into five regions as shown in table 3.
Table 3
Regions
South-Southeast
8 Mérida
11 Orizaba
12 Veracruz
Central Region
1 Mexico
4 Puebla
Central-Western
2 Guadalajara
5 León
7 San Luis Potosí
Nothwest
21 Tijuana
1990
13 Acapulco
18 Villahermosa
19 Tuxtla Gutiérrez
1994 28 Campeche
30 Coatzacoalcos
31 Oaxaca
16 Toluca
29 Cuernavaca
1998 41 Cancún
39 Tlaxcala
43 Pachuca
42 Cd. del Cármen
2000
14 Aguascalientes
15 Morelia
27 Tepic
32 Zacatecas
33 Colima
34 Manzanillo
36 Querétaro
37 Celaya
38 Irapuato
24 Culiacán
25 Hermosillo
Northeast
3 Monterrey
6 Torreón
9 Chihuahua
10 Tampico
20 Ciudad Juárez
22 Matamoros
23 Nuevo Laredo
17 Saltillo
26 Durango
35 Monclova
40 La Paz
44 Mexicali
45 Salamanca
In order to analyze the changes in the regional development patterns, as measured by
regional disparities, we look at overall, inequality in the south versus inequality in the rest of the
country, between and within components across regions, and polarization measures. A first
interesting result is the fact that overall inequality, using both the Gini and Theil indices, is higher
in the rest of the country when compared to the southern part of the country. The common belief
has been that the southern part is typically more unequal, which is not the case in terms of labor
income, according to the data. It is important to mention that such labor income includes some
workers reporting income from agricultural activities, though they live in urban areas.
Graph 3
Inequality in Mexico
0.49
0.47
0.45
0.43
0.41
Gini Rest
0.39
Gini SS
Theil Rest
0.37
Theil SS
0.35
1990
1994
1998
2000
Also, as seen in the graph, the southern part is the only one with a decreasing trend between
1998 and 2000. Systematically, within inequality is the one driving the trend of overall inequality
(Table 5). Taking, however, different measures of polarization, all of them show an increase in
polarization when one takes two groups: southern region versus the rest of the country. Polarization,
indeed, also increased for the overall distribution (Tables 6 and 7). Previous literature has
emphasized the trends in inequality between groups of different skills. Also, there has been work on
the inequality trends between rural and urban sectors. Here, we want to call the attention to a
different dimension: regional trends in polarization, whic h demand policies that consider the
strengthening of linkages across regions. Not surprisingly, the southern part of the country is the
one also showing the highest indices of conflict: strikes, guerrilla, post-election struggles, and
dissident groups within the national teachers union.
IV. Secular Convergence
According to Hanson and Harrison (1999), during the 1980s in Mexico the wage gap
between skilled and unskilled workers widened. The authors assess the extent to which this
increased wage inequality was associated with Mexico’s sweeping trade reform in 1985. Examining
data on 2,354 Mexican manufacturing plants for 1984-90 and Mexican Industrial Census data for
1965-88, they find that the reduction in tariff protection in 1985 disproportionately affected lowskilled industries. Goods from that sector, the authors suggest, may have fallen in price because of
increased competition from economies with reserves of cheap unskilled labor larger than Mexico’s.
The consequent increase in the relative pric e of skill-intensive goods could explain the increase in
wage inequality. The authors discuss the fact that regional differences in income have not been
followed by migration flows, which prevent regional income differences from falling. There is,
however, robust evidence in terms of convergence in human development across states in Mexico.
The interesting result is that convergence in income across states is relatively weak (Esquivel,
1999). When one runs the same convergence regression in human development index, incorporating
schooling and health indicators, the evidence of long-run convergence is robust (see graph 4). Also,
the coefficient of variation of the HDI between 1950 and 2000 has decreased monotonically.
Taking, however, the HDI of the southern part and comparing it with a weighted HDI for the other
states, the pace of convergence is about a third of overall convergence. That is, the south has been
slow in catching up with the rest of the country. All this notwithstanding, from the conclusion of
convergence in HDI important implications for federal interventions can be derived: education and
health policies have been historically controlled by the federal government and have had a
progressive effect in regional development. Given the evidence in Harrison and Hanson (1992) and
Hanson (2002), one of the key obstacles for the southern regions to improve their income
possibilities and catch up is the level and quality of schooling.
Graph 4
HDI Growth Rate 1940-2000
Convergence in Human Development
3.50
3.00
2.50
2.00
1.50
1.00
2
R = 0.9511
0.50
0.00
0.00
0.10
0.20
0.30
0.40
HDI 1940
0.50
0.60
0.70
Graph 5
Coefficient of Variation, HDI 1950-2000
0.35
0.30
0.25
0.20
0.15
0.10
0.05
0.00
1950
1960
1970
1980
1990
1995
2000
The analysis of the potential explanations for the increase in regional polarization in
Mexico leads us to the conclusion that the process is multidimensional. In order to understand such
process, the microeconomic analysis based on the “assets approach” is useful. A household´s
income generation capability at time t is given by:
IGC = ∑i =1
n
∑
k
j =1
Aij γ ij wij + ∑i =1 Ti
n
Where IGC stands for the income generation capability of the household, n is the number of
family members, k is the number of assets, Aij is the endowment of asset j owned by member i, γ ij
is the intensity of use agent i makes of asset j, and wij is the price individual i can get in the market
from selling one unit of asset j. Finally, Ti are the transfers received by member i. Let us consider,
within this framework, two assets: education and health. The prices paid for the combination of
such assets, which determine productivity, is just the hourly wage for certain level of skills. Two
measures of the intensity of use of the asset are participation in the labor market and average hours
worked.
Education and health indicators are considerably lower in the southern region when
compared to other regions and the rest of the country in general. Prices paid, i.e., wages, are also
lower, on average, for similar characteristics. Participation of women in the labor market is 40%
lower in the south when compared to the rest of the country, though participation of men is about
the same. Average hours worked are slightly higher in the southern region, on average. The
hypothesis in the literature is that, given lower wages, poor workers have to work longer hours to
obtain subsistence incomes, which results in a positively-sloped labor supply curve below
subsistence levels (Hernandez-Licona, 2001).
The latter distributional features of the southern region explain why the overall distribution
of income has polarized regionally between the southern part and the rest. The same framework is
helpful to derive certain policy implications: improve provision of education and health, and
promote investment in the region. In terms of the former, inducing the demand for services is as
important as the supply side, using programs like PROGRESA (conditional cash transfers). The
latter cannot be achieved without creating conditions for profitable investment in the region:
infrastructure and credible government. Also, a policy to enhance people ´s migration capabilities
can be helpful to redistribute opportunities.
Final Remarks
Despite the fact that total income inequality has decreased in Mexico during the nineties,
other distributional dimensions open important policy questions. First, labor income inequality has
increased, showing a skill bias after trade liberalization. Also, even though inequality is lower in the
southern part of the country, polarization measures consistently show that the south has moved
farther away from the rest of the country. Regional linkages seem to be weak and have not become
stronger after the structural reform. However, there seems to be evidence of long-term convergence
in human development, using a version of the HDI by states. Educational and health policies seem
to be key in fostering a faster catch up of the south, as well as infrastructure investment. On the
agenda in terms of further studying the causes and effects of regional polarization in Mexico,
Andalon and Lopez-Calva (2001) show robust evidence of the impact of polarization on violent
crime in urban Mexico.
References
Andalón, M. and Lopez-Calva, L. F. (2001), “Labor Income Inequality, Polarization, and Crime in
Mexico”, mimeo, El Colegio de Mexico.
Chakravarty, S. and M. Majumdar (2001), , AEP.
Cowell, F. (1995). Measuring Inequality, Prentice Hall/Harvester.
Cowell, F. y Champernowne, D (1998). Economic inequality and income distribution. Cambridge
University Press.
De la Torre, R. (1998). “Economic Polarisation and Governability in Mexico”. Mimeo, UIA-Santa
Fe.
Esquivel, G. and L. F. Lopez-Calva (2002), “Human Development in Mexico: Regional
Characteristics and Limits to Convergence,” El Colegio de México.
Esquivel, G. and A. Tornell (1998), “The Political Economy of Mexico´s Entry to NAFTA”, NBER
Working Paper.
Esteban, J.M. (1995). “Desigualdad y polarización en la distribución interregional de la renta”.
Mimeo,. Instituto de Análisis Económico, CSIC. Barcelona, España.
Esteban, J.M y Ray, D. (1994). “On the Measurement of Polarization”. Econometrica. Vol (62).
Esteban, J.M y Ray D. (1999). “Conflict and Distribution”. Journal of Economic Theory. Vol. (87)
Pp 379-415.
Esteban, J.M., Gradín, C y Ray D. (1999). “Extensions of a Measure of Polarization with an
Application to the Income Distribution of Five OECD Countries”. Mimeo. Syracuse University.
Foster, J. and R. Wolfson (1993), “Measuring the Middle ,” mimeo, Vanderbilt University.
Gradín, C. (1994). “Aproximación a la aplicación de una medida de polarización de rentas”, mimeo,
Universitat Autónoma de Barcelona. Barcelona, España.
----------- (2000ª). “Polarization and Inequality in Spain: 1973-1991”, mimeo, Universidad de
Vigo. Por publicarse en el Journal of Economic Literature.
----------- (2000b). “Polarization by Subpopulations in Spain, 1973-1991”, Review of Income and
Wealth. Vol. (46).
Gradín, C. y Rossi, M. (2000). “Polarización y Desigualdad Salarial en Uruguay”, 1986-1997” El
trimestre económico.
Hanson, G., (2002), “Regional Differences in Inequality in Mexico”, UCSD, mimeo.
Hanson, G. and A. Harrison (1999), “Trade Liberalization and Wage Inequality in Mexico,”
Industrial and Labor Relations Review, Vo. 52, No. 2, January.
Hernandez-Licona, G. (2001), “Re-shaping the Labor Supply Curve for the Poor”, mimeo, ITAM,
Mexico.
Wolfson, M. (1994). “When Inequalities Diverge?” American Economic Review. Vol. (84), no. 2
Pp. 353-358
Tsui, K. y Y. Wang (1998). “Polarization Ordering and New Classes of Polarization Indices”.
Mimeo.
Zhang, X. Y Kanbur, R (1999). “What Difference Do Polarization Measures Make? An Application
to China”, mimeo, Cornell University.
Table 4
All regions between 1990 and 2000
Region
Year
%
Population
Mean Real
Income
(pesos)
% of Total
National
Income
Log Income
Gini
Theil
Sur sureste
Centro País
Centro Occidente
Noroeste
Noreste
Sur sureste
Centro País
Centro Occidente
Noroeste
Noreste
Sur sureste
Centro País
Centro Occidente
Noroeste
Noreste
Sur sureste
Centro País
Centro Occidente
Noroeste
Noreste
2000
2000
2000
2000
2000
1998
1998
1998
1998
1998
1994
1994
1994
1994
1994
1990
1990
1990
1990
1990
5.29%
55.41%
14.91%
3.54%
20.84%
5.32%
56.80%
14.11%
3.39%
20.38%
5.20%
57.42%
13.87%
2.96%
20.55%
5.50%
56.67%
15.04%
2.59%
20.20%
522.941
620.183
564.138
1036.471
731.514
479.591
548.093
482.900
866.606
610.824
609.179
682.896
590.107
806.388
726.401
503.338
559.520
520.244
1012.416
624.260
4.29%
53.31%
13.05%
5.70%
23.65%
4.56%
55.71%
12.19%
5.25%
22.28%
4.66%
57.77%
12.06%
3.52%
21.99%
4.81%
55.12%
13.60%
4.55%
21.92%
6.259
6.430
6.335
6.944
6.595
6.173
6.306
6.180
6.765
6.415
6.412
6.526
6.380
6.693
6.588
6.221
6.327
6.254
6.920
6.437
0.46364
0.50852
0.43798
0.43837
0.44481
0.46924
0.48744
0.45820
0.43333
0.44257
0.46723
0.50185
0.46077
0.41086
0.45720
0.44147
0.46647
0.43133
0.45173
0.44171
0.42461
0.53964
0.36546
0.36540
0.38710
0.42088
0.46938
0.43195
0.36650
0.37687
0.41108
0.51202
0.37451
0.32267
0.39587
0.38100
0.44413
0.38998
0.39583
0.38219
Source: own calculation using data from ENEU.
Table 5
Constant regions throughout 1990-2000
Region
Sur sureste
Centro País
Centro
Occidente
Noroeste
Noreste
Sur sureste
Centro País
Centro
Occidente
Noroeste
Noreste
Sur sureste
Centro País
Centro
Occidente
Noroeste
Noreste
Sur sureste
Centro País
Centro
Occidente
Noroeste
Noreste
Year
%
Population
Mean Real
Income
(pesos)
% of Total
National
Income
Log Income
Gini
Theil
2000
2000
10.71%
45.40%
545.060
608.886
9.20%
43.57%
6.301
6.412
0.45564
0.50998
0.39944
0.54506
2000
2000
2000
1998
1998
18.45%
6.75%
18.69%
10.85%
46.54%
576.285
886.875
714.566
501.116
538.046
16.76%
9.44%
21.05%
9.85%
45.35%
6.357
6.788
6.572
6.217
6.288
0.42569
0.41992
0.44267
0.47112
0.48776
0.34436
0.32578
0.38048
0.42996
0.47105
1998
1998
1998
1994
1994
17.53%
6.71%
18.36%
9.91%
47.95%
488.360
771.905
599.179
591.970
680.473
15.50%
9.39%
19.92%
8.74%
48.62%
6.191
6.649
6.396
6.383
6.523
0.45800
0.38911
0.43889
0.46903
0.50151
0.44192
0.27945
0.37102
0.41183
0.51154
1994
1994
1994
1990
1990
17.55%
4.82%
19.77%
5.50%
56.67%
620.971
780.381
705.601
503.338
559.520
16.24%
5.60%
20.79%
4.81%
55.12%
6.431
6.660
6.559
6.221
6.327
0.45726
0.38150
0.45382
0.44147
0.46647
0.43154
0.26335
0.39094
0.38100
0.44413
1990
1990
1990
15.04%
2.59%
20.20%
520.244
1012.416
624.260
13.60%
4.55%
21.92%
6.254
6.920
6.437
0.43133
0.45173
0.44171
0.38998
0.39583
0.38219
Source: own calculation using data from ENEU.
Table 5
Inequality
Year
1990
1994
1998
2000
Gini
0.45907
0.47978
0.47248
0.47926
Theil
0.42612
0.45676
0.43794
0.45980
Polarization
Extended
Extended
Esteban, Esteban and
Esteban,
Esteban and Gradín and
Ray alpha
Gradín and
Tsui and
Ray alpha 1 Ray alpha 1
1.5
Ray alpha 1.5
Wang
0.02502
-0.38458
0.01620
-0.39340
0.68994
0.01664
-0.42957
0.00958
-0.43662
0.73239
0.02529
-0.38987
0.01420
-0.40096
0.71810
0.02801
-0.38724
0.01551
-0.39974
0.72079
Zhang and
Kanbur
0.01953
0.00469
0.01677
0.01877
Zhang and
Kanbur
(betw/(bet+with))
0.01916
0.00466
0.01649
0.01842
FosterWolfson
0.38536
0.39312
0.41885
0.41322
Polarization
Extended
Extended
Esteban, Esteban and
Esteban,
Esteban and Gradín and
Ray alpha
Gradín and
Tsui and
Ray alpha 1 Ray alpha 1
1.5
Ray alpha 1.5
Wang
0.02502
-0.38458
0.01620
-0.39340
0.68994
0.02078
-0.42665
0.01349
-0.43394
0.74378
0.02886
-0.39083
0.01869
-0.40100
0.71804
0.03433
-0.38405
0.02187
-0.39651
0.72654
Zhang and
Kanbur
0.01953
0.00546
0.01681
0.02159
Zhang and
Kanbur
(betw/(bet+with))
0.01916
0.00543
0.01653
0.02113
FosterWolfson
0.38536
0.39806
0.41519
0.39448
Constant regions 1990-2000
Inequality
Year
1990
1994
1998
2000
Gini
0.45907
0.48246
0.47357
0.48361
Theil
0.42612
0.46450
0.44055
0.47115
Table 6
South vs. Rest
Inequality
Year
1990
1994
1998
2000
Gini
0.45907
0.48246
0.47357
0.48361
Theil
0.42612
0.46450
0.44055
0.47115
Esteban and
Ray alpha 1
0.00732
0.00561
0.00810
0.01101
EGR α=1
-0.44488
-0.47153
-0.45794
-0.46261
Polarization
Esteban and
Ray alpha
1.5
EGR α=1.5
TW
0.00682
-0.44538
0.69871
0.00524
-0.47189
0.74719
0.00756
-0.45847
0.72642
0.01028
-0.46334
0.73710
Zhang and
Kanbur
0.00111
0.00064
0.00134
0.00226
ZK
0.00111
0.00064
0.00134
0.00225
FW
0.38536
0.39806
0.41519
0.39448
Annex 1
Human Development Index by State
1940 1970 1990 2000
Aguascalientes
0.53 0.67 0.73 0.78
Baja California Norte
0.66 0.77 0.78 0.80
Baja California Sur
0.54 0.78 0.78 0.78
Campeche
0.33 0.58 0.87 0.84
Coahuila de Zaragoza
0.46 0.70 0.73 0.83
Colima
0.39 0.61 0.69 0.74
Chiapas
0.20 0.37 0.50 0.53
Chihuahua
0.49 0.71 0.76 0.83
Distrito Federal
0.58 0.82 0.89 0.93
Durango
0.50 0.64 0.69 0.72
Guanajuato
0.26 0.51 0.61 0.66
Guerrero
0.16 0.42 0.50 0.59
Hidalgo
0.25 0.44 0.58 0.68
Jalisco
0.35 0.65 0.71 0.75
México
0.24 0.63 0.70 0.70
Michoacán
0.23 0.49 0.56 0.62
Morelos
0.34 0.60 0.68 0.71
Nayarit
0.35 0.58 0.63 0.64
Nuevo León
0.50 0.81 0.85 0.87
Oaxaca
0.08 0.33 0.46 0.56
Human Development Index by State (cont.)
Puebla
0.22
0.48 0.56 0.67
Querétaro
0.36
0.59 0.70 0.79
Quintana Roo
0.63
0.65 0.75 0.82
San Luis Potosí
0.28
0.54 0.65 0.69
Sinaloa
0.38
0.64 0.67 0.68
Sonora
0.45
0.74 0.76 0.81
Tabasco
0.22
0.53 0.65 0.64
Tamaulipas
0.49
0.70 0.72 0.78
Tlaxcala
0.28
0.46 0.57 0.63
Veracruz
0.31
0.55 0.61 0.64
Yucatán
0.43
0.58 0.63 0.68
Zacatecas
0.31
0.52 0.61 0.64
Max
0.66
0.82 0.89 0.93
Min
0.08
0.33 0.46 0.53
Source: Esquivel and López-Calva (2002).