Polarization and Social Cohesion in Latin America

1
Polarization and Social Cohesion in Latin America

• Luis F. Lopez-Calva
• Chief Economist, RBLAC
• UNDP

2
Main Idea

• Beyond inequality and poverty, the concept of
  social cohesion addresses the issue of whether a
  social group is able to define a common objective
  and whether the institutional context is
  conducive to the collective pursuit of such
  objective
• Polarization is proposed as a measure that
  recovers such feature in society

3
Conceptual Foundations

• Social cohesion has been discussed in several
  academic disciplines, from economics and
  political science to anthropology and sociology.
• However, a specific definition of social cohesion
  is often lacking

4
Example

• Social capital versus social cohesion
• Bonding vs. Bridging social capital

5
Literature

• In a recent study, ECLAC defines social cohesion
  as the dialectical relationship between
  mechanisms of social inclusion/exclusion and
  people's responses, perceptions and attitudes to
  the ways these operate in producing a sense of
  belonging to society.
• Proposal of more than thirty indicators to
  measure the distinct dimensions of social
  cohesion.

6
A Simple Definition of Social Cohesion

• Social cohesion could be simply defined by
• a group of individuals that potentially make up a
  single body, a politically constituted community
• and the force which drives them together, which
  unifies a group towards a common goal
• (taking the definition of cohesion and adding the
  social part to it)

7
Normative vs Functional Notions
8
The polarization concept
Unimodal dist
Poverty
F
Inequality
9
The polarization concept
Bimodal
F
Middle class
10
Polarization measures

• Esteban, Gradín and Ray (2006) introduced an
  extension
• of the Esteban and Ray 1994 (ER) measure of
  polarization
• that can be applied to density Functions. This
  derivation has
• the virtue of casting both measures in the
  context of a unified fremework.
• The ER formulation relies on what they called
  the identity-
• alienation framework. This means that individuals
  are identified with others who are close to
  them, while they are alienated from other who
  are far away.

11
Polarization measures

• The ER measure assume that the data arrives
  pre-grouped into appropiate population clusters,
  within which there are bonds of identification
  and across which are the tensions of alienation.
  But, the statistical classes into which the
  distributional data may be grouped may have
  nothing to do with the former conceptual
  grouping.
• ER suggest an extension of the measure to deal
  with them as follows

12
Polarization measures, example

• Suppose that a fixed number of income cutoffs
  are given to the researcher (income groups).
• These define the number of groupings but not
  their locations
• They propose to pin down group locations by
  minimizing the dispersions within the clusters
  created by any number of income cutoffs.
• Then, they apply the ER measure to the discrete
  groupings with a correction for intra-group
  dispersions.

13
Polarization measures

• This yields an extended measure of polarization
  which can be applied to all sorts of income
  distributions, especially when they are in the
  form of densities.
• As byproduct they derive the W measure as a
  particular special case of this formulation to
  cast both the ER and W measures in the context of
  a (statically) unified framework.

14
The identity-alianation framework Conceptual
issues
Suppose that an individual with income x feels
group identification I(x, F) under the
distribution F, and alienation r(x, y) with
respect to some individual with income y. As in
ER, they take the effective antagonism that
individual x feels towards y as some function
T(I, r) strictly increasing in r. Effective
antagonism increases with alienation, but this
alienation is taken to be fueled by some sense of
identification as well. Polarization is the
sum of all effective antagonisms
15
The identity-alienation framework Conceptual
issues

• The approach taken in ER is to combine this
  relatively broad starting point with a set of
  intuitive axioms that might compare polarization
  across distributions.
• The ER characterization is restricted, however,
  to distributions that are pre-arranged in groups
  so that for an individual with income x belonging
  to some group i, I(x, F) simply equals pi, the
  proportion of individuals in that group (under
  the distribution F).
• But there is no reason to believe that the
  grouping of income distribution data will
  conveniently conform to the psychological demands
  of group identification.

16
The identity-alianation framework Conceptual
issues

• Take alienation to be simply the linear distance
  between x and y, with the identification zone
  netted out
• r(x,y) max x - y - D, 0. (i)
• Then, a natural generalization of the ER measure
  is

(ii)
where a is some positive constant capturing the
importance of group Identification in the
determination of interpersonal effective
antagonism.
17
The identity-alianation framework Conceptual
issues

• Observe that if group identification is
  unimportant, then we can take D 0 and a 0 as
  well, in which case the measure in (i) reduces to
  a measure of inequality, the Gini coefficient.
  Thus it is the presence of identification that
  makes a measure of polarization fundamentally
  different from one of inequality (for more on
  this, see ER).
• There are some features of (ii). First, if the
  distribution is clustered entirely within a
  support of D, then polarization is zero. Second,
  and more problematic, is the fact that the
  measure is still not operational.

18
A statistical approach

• The extension of the ER measure of polarization
  can be summarized as follows
• The ER polarization measure for discrete groups
  should be used only after the population has been
  regrouped in a way that captures the group
  identification structure of society.
• This regrouping or clustering will lose some of
  the initial information that concerns the
  dispersion of the population around the clusters
  that we are treating as single groups.

19
Polarization in Latin America

• Main Findings from Gasparini, et al. studies
• RBLAC commissioned two papers on polarization in
• Latin America one on the dynamics and
  characteristics of
• polarization, the other on polarization,
  inequality and
• social conflict.
• 2) As a region, Latin America is highly polarized
  44
• more so than Europe.

20
Polarization in Latin America
21
Polarization in Latin America

• 3) Most of the indicators find for income
  polarization
• A slight increase in polarization between circa
  1990 and circa 2004
• A convergence across countries high
  polarization countries saw reductions in
  polarization (Chile, Brazil) Low polarization
  countries saw increases

22
Polarization in Latin America

• (CR, Uruguay, Venezuela) with no exceptions
  Bolivia and Colombia, which started high and saw
  increases.
• 4) For polarization by characteristics, the
  findings are Groups delineated by educational
  attainment show the highest income polarization,
  followed by groups denominated by
• Labor status (formal/informal)
• Region (urban, rural) or race in Paraguay,
  Bolivia and Brazil

23
Polarization in Latin America

• 5) Changes in the size of the middle-income
  groups (in terms of population and income) have
  been similar to those reported for polarization
  and inequality
• The middle-class seems to have been shrinking
  in most of South America, with the exceptions of
  Brazil and Chile.
• Changes in Central America and Mexico have been
  milder, without clear signs of a significant
  reduction in the middle class.

24
Polarization in Latin America

• 6) On the relationship between income
  distribution, institutions and conflict, the
  findings are
• Negative correlation between indicators of
  income distribution and indicators of the
  quality of broad- based institutions
• The most polarized countries are on average the
  ones that have higher levels of conflict
• (See Fajnzylber, et al. )

25
Discussion

• Is polarization a good sufficient statistic for
  social cohesion?
• It does have interesting properties
• When is polarization different from Poverty and
  Inequality
• This is what the work in progress is moving
  towards