Poverty, Exclusion, and Dissent - Support for Regimes in Developing Countries

Poverty, Exclusion, and Dissent*
Support for Regimes in Developing Countries
Raj M. Desai†
Anders Olofsgård‡
Tarik M. Yousef§
August 2008
Abstract
Poverty is associated with political conflict in developing countries, but evidence of individual grievances
translating into dissent among the poor is mixed. We analyze survey data from 40 developing nations to
understand the determinants radicalism, support for violence, and participation in legal anti-regime actions
as petitions, demonstrations, and strikes. In particular, we examine the role of perceived political and
economic inequities. Our findings suggest that individuals who feel marginalized tend to harbor extremist
resentments against the government, but they are generally less likely to join collective political movements
that aim to instigate regime changes. This potentially explains the commonly-observed pattern in low- and
middle-income countries whereby marginalized groups, despite their political attitudes and high-levels of
community engagement, are more difficult to mobilize in nation-wide movements. We also find that
arenas for active political participation (beyond voting) are more likely to be dominated by upper-middle
income groups who are committed, ultimately, to preserving the status quo. Suppressing these forms of
political action may thus be counterproductive, if it pushes these groups towards more radical preferences.
Finally, our findings suggest that the poor, in developing nations, may be caught in a vicious circle of self-
exclusion and greater marginalization.
Prepared for delivery at the annual meeting of the American Political Science Association, August 28 – 31,
2008, Boston. The authors are grateful for invaluable research assistance from Joshua Hermias.
†
Edmund A. Walsh School of Foreign Service, Georgetown University, and Wolfensohn Center for
Development, Brookings Institution. E-mail: desair@georgetown.edu.
‡
Department of Economics and Edmund A. Walsh School of Foreign Service, Georgetown University, and
Stockholm Institute for Transition Economics. E-mail: afo2@georgetown.edu.
§
Dubai School of Government. E-mail: Tarik.Yousef@dsg.ae.

1
Introduction
A central assumption of research on state fragility is that poverty is associated
with violence. Most of the world’s poorest live in conflict-ridden societies. While cross-
country evidence shows a strong relationship between income levels and conflict, the
micro-level evidence is mixed. In this paper we examine the determinants of three
categories of attitudinal and behavioral outcomes in the developing world: (1) radicalism
and support for revolution; (3) support for violence against the state; and (3) participation
in (legal) anti-governmental actions. We focus primarily on the effects of wealth and
poverty and on perceptions of political and economic inequities.
The problem of legitimacy is at the heart of effective government. Governments
are better able to function where its citizens are motivated by a ―willingness to cooperate
. . . backed by coercion‖ (Levi 2006). Naturally, that willingness is part of a social
contract whereby citizens consent to be governed in exchange for the provision of public
goods. For governments in poor countries, moreover, the challenge of good government
is often placed alongside the imperative of reducing poverty. But if it is generally
accepted that good government depends crucially on what Levi has called the ―quasi-
voluntary compliance‖ of the governed, there is much less agreement on the determinants
of political engagement, attitudes, and behaviors of individuals in low- and middle-
income countries. Understanding individual- or household-level determinants of protest
and dissent requires a better understanding of the political lives of citizens in low- and
middle-income countries.
Using World Values Survey data we examine the sources of anti-regime behavior
among approximately 44,000 individuals in over 40 lower- and middle-income countries,

2
identifying whether the factors that drive radicalism and revolutionary preferences also
determine other modes of political action. We focus, in particular, on the role of
economic and political grievances. Intuitively, these grievances should generate anti-
regime attitudes. The question, however, is the manner in which those attitudes map to
concrete actions. Citizens have, generally speaking, several options between two poles:
at one end, actions within the existing legal order; at the other, open revolt. We
investigate the types of responses that perceived injustices produce.
Our results suggest that economic and political grievances breed both radicalism
and exclusion, particularly among the poorest. We find that individuals who are
dissatisfied with the political and economic inequities are more likely to nurture
revolutionary preferences, and are more likely to support violent action against
incumbent governments. But they are also less likely to be politically engaged. If true,
this would suggest that the poor in developing nations may be caught in a vicious cycle of
increasing alienation, whereby the gap between their allegiances to their government and
that of their better-off counterparts grows ever wider. Our analysis is organized as
follows. In Section 2 we discuss related literature, and in Section 3 we present the data,
and motivate our empirical strategy. Section 4 presents our results, and we conclude in
Section 5.
The Political Economy of Dissent
Over 70% of the more than one billion persons living on less than $1 per day are
currently in a civil war or have recently been through one (Collier 2007). Statistically, a
low-per capita income is one of the more robust determinants of the risk of internal

3
conflict (Elbadawi and Sambanis 2001; Collier and Hoeffler 1998; Rice 2006). Conflicts,
once begun, perpetuate destitution by perpetuating hunger, destroying assets (Humphreys
and Weinstein 2006), spreading disease (Pedersen 2002) and degrading the environment
(Dudley et al. 2002).
Yet sub-national, household, and individual-level evidence of the connection
between poverty and violence is often mixed, or at odds with cross-national evidence.
District-level evidence, for example, shows a positive connection between community
poverty and the likelihood of conflict in Uganda (Deininger 2003) and in the Philippines
(Malapit et al. 2003); evidence from Colombia (Sánchez and Chacón 2003) and from
Indonesia (Barron et al. 2004), however, show no such connection. An events-analysis of
Indonesia, by contrast, finds an inverted-U shaped relationship between income levels
and violence, with rising incomes increasing the likelihood of violence but lowering the
likelihood as communities increase income levels beyond a certain threshold (Tadjoeddin
and Murshed 2007). Analyses of terrorist events, finally, show little connection between
individual poverty and participation in terrorism (Krueger and Maleckova 2003; Piazza
2006).
Meanwhile, when it comes to more common (legal, non-violent) forms of
political participation—voting, petitioning, joining interest groups, etc.—the evidence is
equally mixed. Cross-national studies of developing countries once found support for the
―modernization thesis‖ that those of higher socioeconomic status are more likely to be
politically engaged (Powell 1982; Huntington and Nelson 1976; Lipset et al. 1993).
More recent evidence finds that many developing countries still conform to this
expectation (Blais 2000). Other surveys of participation in developing countries, by

4
contrast, find that participation is often greater among poorer households. Survey
evidence from several countries in Latin America, for example, finds that poverty is
positively related to partisan activism (Booth and Seligson 2006 ). Similarly, in India
civic participation among disadvantaged households is greater than that of richer
households (Diwakar 2008; Besley et al. 2005).
What then, are the factors that shape individuals’ willingness to participate in
collective acts of protest or violence? The distinction between ―preferences‖ and
―opportunities‖ is crucial since that which determines attitudes may not motivate action
against regimes ((Hirshleifer 1993). Anti-governmental political actions—which are
generally risky—do not necessarily follow discontent. First, grievances and perceived
inequities or ―deprivation‖ alone, of course, are unlikely to destabilize or threaten
governments. Second, evidence from studies of regime durability and transition shows
the myriad ways in which groups that are more likely to act upon their discontent can be
selectively pacified (Bueno de Mesquita et al. 2003). Anti-poverty programs have often
lavished benefits on politically relevant groups rather than the poorest (Schady 2000;
Diaz-Cayeros and Magaloni 2003)). Rentier states have used natural-resource revenues
to fund social programs targeted to key constituencies (Aslaksen and Torvik 2006;
Chaudhry 1997). And these generous social expenditures and public employment
opportunities have tended to create and reinforce allegiances to the state that have proven
deeply resilient (Yousef 2004).
All governmental behavior takes place in the shadow of conflict, that is, against
the prospect of widespread public non-compliance, resistance, or revolt. Consequently,
the ability to overcome collective-action problems is at the core of the credible threat of

5
dissent, since actual non-compliance ultimately requires coordination. And different
collective-action capabilities across groups ultimately make the difference between what
Georg Simmel famously referred to as ―sterile excitations‖ and actual mobilization.
In sum, the cross-national evidence of strong, causal links between economic
deprivation and political turmoil are at odds with some community- and individual-level
data. In particular, little is known about the link between attitudes and actions, and the
circumstances by which marginalization leads to legal or extra-legal (and violent)
responses. Opinion/values surveys can illuminate the conditions under which endemic
discontent produces anti-regime sentiments, and when these sentiments lead to individual
acts of resistance.
Data and Methods
We rely on data from the World Values Survey (WVS), an international survey of
attitudes towards political and cultural issues across 80 countries conducted in four waves
between 1981 and 2004. The WVS is constructed of representative national samples of
the populations in these countries, but our primary interest is not in the country-level
means or other indices. Rather, we are interested in individual-specific motivations, and
therefore use individual responses in our analysis. We focus on three main dependent
variables: radicalism (mainly, the preference for ―revolution,‖ but also self-positioning
on a left-right ideological spectrum), the support for violence against the government, and
a range of actual (self-reported) political activities. In addition, we generally restrict
(with some exceptions) our sample to approximately 40 lower- and middle-income
countries.

6
The problems of comparability when respondents are asked to use ordinal
response categories are well known. Different respondents may interpret subjective
questions in different ways based on unobservable characteristics. Ordinal scales,
moreover, may mean different things to different respondents based on idiosyncratic
factors such as mood or overall optimism. Individual-level perceptions would similarly
be affected by measurement error where identical individuals may have unequal
probabilities of answering questions about their own political preferences in the same
way. Additionally, the measurement error in subjective responses may be correlated with
a wide variety of individual characteristics and behaviors (King and Wand 2007;
Bertrand and Mullainathan 2001).
―Anchoring vignettes‖ or other hypothetical questions to establish baselines that
could normally correct survey responses for inter-personal incomparability, however, are
not included in the WVS. Consequently, we attempt to correct our specifications in the
following manner: in all estimations, we include a ―bias‖ proxy. We derive that proxy
by regressing a ―life satisfaction‖ response variable (coded from 1, ―dissatisfied‖ to 10,
―satisfied‖) against a ―freedom of choice and control‖ response variable (also coded from
1 to 10 in terms of degree of control respondents report to have over their lives), along
with country- and time-fixed effects, and taking into account population-sampling
weights. The residuals from this equation reasonably approximate an individual’s
personal ―bias‖—i.e., the level of dissatisfaction that cannot be explained by one’s
feeling of autonomy.5
5
We estimated results using a variety of different derivations of bias—for example, including income
deciles, responses to questions on financial satisfaction, etc. Our results are unaffected by the inclusion of
these economic-condition variables. We feel that the satisfaction-agency/autonomy gap is a better proxy
for bias than a life satisfaction-financial satisfaction gap (the correlation between life satisfaction and

7
Our basic specifications take the following form:
 ctiictxictwictyict yFQ   ,,ˆ,ˆ,ˆ,ˆ xw ,
where Q is the political attitude or action of interest for individual i living in country c in
year t, y is a measure of individual income, w is the set of individual ―grievances‖ based
on a rating of the policy-making effectiveness of the incumbent government, x is a vector
of individual demographic characteristics, θ is the individual-specific systematic bias, η
and μ are country- and time-fixed effects, respectively.
For the first of our dependent variables, the individual preference for revolution,
we rely on the following WVS question:
On this card are three basic kinds of attitudes concerning the society we live in. Please choose the
one which best describes your own opinion: (1) The entire way our society is organized must be
radically changed by revolutionary action; (2) Our society must be gradually improved by
reforms; (3) Our present society must be valiantly defended against all subversive forces.
We code this 1 if the respondent chooses the first answer, 0 otherwise. For the second
dependent variable—individual support for violence—WVS asks:
Could you please tell me if you agree strongly, agree, disagree, or disagree strongly with the
following statement: ―Using violence to pursue political goals is never justified."
Responses range from 1 (agree strongly) to 4 (disagree strongly). We use the full range
of responses in ordered logit regressions (rescaled so that the highest value represents
maximum support for violence).6
Finally, for the third outcome of interest we construct
an index of political action based on responses to five different sub-questions asked as
follows:
financial satisfaction is over 0.60 in our sample). The derivation of bias also includes sampling weights
(see below).
6
We also use logit regressions as a robustness check, coding this variable 1 if the respondent chooses 4, 0
otherwise.

8
I'm going to read out some different forms of political action that people can take, and I'd
like you to tell me, for each one, whether you have actually done any of these things,
whether you might do it or would never, under any circumstances, do it: (A) signing a
petition; (B) joining in boycotts; (C) attending lawful demonstrations; (D) joining
unofficial strikes; (E) occupying buildings or factories.
Two approaches to the construction of this variable are used. The methodology
followed by some other researchers is simply to sum up the number of acts
respondents said they have actually committed (Inglehart 1997; Norris 2002).
Given that these actions differ in terms of the effort and risks involved, and to put
extra weight on actions that signal stronger convictions, we construct a simple
―political action index‖ as A + 2×B + 3×C + 4×D + 5×E, yielding a value that
ranges between 0 and 15. Second, we construct a weighted index based on
principal components analysis. In reality, both approaches yield similar results,
and both are closely related to a simple sum of the five components.7
We use the WVS approximation of individual wealth, based on country-
specific correspondence between reported income and income deciles. There are
no questions in the WVS, unfortunately, that can measure individual or household
wealth in other ways—for example, questions about consumption patterns, food,
7
Analysis of the WVS has shown that the five components of political action form a single dimension.
Previous factor analysis has shown that only one factor emerges with an eigenvalue greater than 1, and all
five components are loaded strongly on that single factor (Dalton and van Sickle 2005). We combine the
five components across four survey rounds into a single factor:
Factor
Loadings
Signed a petition 0.414
Joined a boycott 0.481
Lawful demonstration 0.479
Unofficial strikes 0.453
Occupying buildings 0.403
Eigenvalue 2.744
% Variance 0.549

9
shelter, access to services, etc., making it impossible to generate a ―lived poverty‖
index as others have done (e.g., see Bratton 2006).
To construct indices of individual ratings of governmental policymaking
effectiveness, we rely on two separate variables. First, we construct a ―Change in
political rating‖ score based on respondents’ rating of the political system on a scale of 1
(worst) to 10 (best). We subtract the rating of the system ―as it was ten years before‖
from the current rating. Second, we use respondent answers on inequality of incomes as
an economic performance rating—where individuals are asked to place their views on a
scale of 1 (agreement with the statement ―incomes should be made more equal‖) to 10
(agreement with the statement ―we need larger income differences as incentives for
individual effort‖). The perception of economic inequality, is the most comprehensive
indicator of the perceptions regarding overall economic management.8
Among the demographic characteristics we include are: employment status, size
of town in which the respondent lives, gender (male = 1), age, and education. In addition
to country- and year-fixed effects, we include population-sampling weights in all
estimations. Country-specific stratification methods, over- and under-representation of
sub-samples, and non-response all require that sample weights be included in all
estimations to derive unbiased population means and correct standard errors.9
8
The WVS does include questions on ―opportunities to escape from poverty,‖ but these questions were
included in very few survey rounds in developing countries. Other questions focus more explicitly on
individual preferences for economic policy (e.g., whether the respondent believes in greater or lesser
government intervention in the economy) than on systemic economic performance.
9
Note that WVS response rates vary greatly. In addition, the WVS tended to over-sample upper-middle
income groups in developing countries. To correct for the sample design and the response rate, sample
weights were constructed with specific criteria for each country—essentially, the inverse of the probability
that an individual is a member of the sample population (see, e.g., Inglehart et al. 2004).

10
Results
Ideology and Radicalism
We begin with an analysis of preferences for political change and ideology. Table
1 and table 2 present results from estimations of political ideology. Table 1 estimates
radicalism, or the preference for revolution. WVS data on revolutionary ―tastes,‖ in other
research, has been considered a useful proxy for instability (MacCulloch and Pezzini
2002). Alongside radicalism, we also estimate the opposite, namely, the view that
society must be defended against ―subversive‖ elements. Perceptions of worsening
governmental performance, and of income inequality, increase the preference for revolt.
Although income levels do not have a significant impact on radicalism, richer cohorts are
more likely to be reactionary. Employment status and location have no effect.
Interestingly, the more educated are also more likely to support radical solutions. Those
who perceive the performance of the government improving, as expected, are less radical,
while those who perceive widening income gaps are more disposed toward radicalism.
We also compare, in this table, results from regressions for developing and high-income
countries, and observe two sign shifts: better-educated individuals in richer countries are
less likely to be radical (but are also less likely to be ―reactionary‖), and older individuals
in richer countries are more likely to be reactionary.
Figure 1 plots the coefficients of the political-rating and inequality-perception
covariates when restricting the sample to different income deciles. Analyses of Europe,
the US, and Latin America have shown that the impacts of inequality on attitudes such as
life satisfaction or happiness, for example, depend crucially on income levels (Alesina et
al. 2004; Graham and Felton 2005). Decile regressions indicate that, aside from a sharp

11
increase in the correlation between inequality perceptions and radicalism among middle-
income groups, the overall effect is essentially flat and barely above zero. By contrast,
the negative effect of political ratings declines slightly across deciles, suggesting that a
poorer view of governance is likely to push for more radicalism among richer individuals.
Table 2 uses a slightly different approach, examining how these factors influence
self-positioning on a left-right scale. Wealthier individuals, those who perceive
improvements in governmental effectiveness, and those who do not see greater inequality
gravitate rightward. Reflecting the finding in table 1, education increases left-wing
tendencies in both rich and poor countries. But in poor countries, the unemployed are
more right-wing, unlike the case in richer countries. Moreover, in rich nations, those
happy with the political performance of their country are more liberal. In addition, in
developing nations it is the urban resident who is more conservative—again, the reverse
of the situation in richer countries. Tables 1 and 2 (as well as tables 3 and 4 below) use
both nonparametric bootstrap re-sampling and parametric clustering to derive coefficient
errors. Equations with bootstrapped errors use 1,000 replications with fixed regressors.
We also cluster errors at the sub-national level in the last columns in both tables. Overall
the picture that emerges from developing nations is different than the conventional
wisdom from richer countries, where the presumption is that urban-dwellers and the
unemployed are the core constituencies of left-wing political movements. The situation
is more complicated in poorer countries, possibly due to the strong ties of populist parties
to rural areas and labor movements in these countries (see, e.g., Stokes 2001; Murillo
2000).

12
Support for Political Violence
Similar to our findings on radical tendencies above, in the ordered logit
regressions in table 3 we see that the politically and economically alienated are more
disposed towards the use of violence against the government in developing nations
(alienation, while pushing individuals towards radicalism in high-income countries,
seems to have no effect on individuals’ support for violence in these countries). The
unemployed and urban-dwellers—again—are natural groups to be mobilized into
violence (and again, this does not appear to be the case in rich countries).
Income levels do not appear to have any significant effect among individuals in
developing countries; in richer countries, wealthier individuals are less prone to support
violence. Simple nonparametric analysis of this relationship, however, reveals a roughly
quadratic, but overall positive, relationship between income and support for violence in
developing countries. Figure 2 shows partial-residual plots of support for violence
against income deciles, generated as follows. We estimate the full ordered-logit
specification as in column (1) in table 3 to generate standard (non-bootstrapped) errors of
the linear prediction. We then run local-polynomial regressions of these residuals on
income deciles. The results quantify the isolated effect of income levels on support for
violence. The effect increases through the bottom five income deciles and diminishes
thereafter.
Table 4 segments the sample by income quintile, to allow a more careful
examination of the conditional effect of income level on other covariates. Overall, the

13
effects are consistent across quintiles. We see, for example, that the dampening effect of
a positive political rating on support for violence is relatively larger for the 4th
quintile.
For the lower and upper quintiles, the effect is of slightly lower magnitude. Economic
inequities, on the other hand, have their greatest effect on increasing support for violence
among the middle quintile.
Political Behavior
We examine, finally, the propensity to engage in (non-violent) political action in
tables 5-6. Table 5 replicates our basic estimations with the two different political action
indices. The results are generally consistent between the two outcome variables. We find
strong support for the link between socioeconomic status and political engagement.
Richer, better-educated, employed individuals in urban settings are more likely to partake
in explicitly political acts. Those who are happy with the direction of governmental
performance are more, not less, likely to participate in political actions. Those who are
unhappy with economic inequities are less likely to do so (note that this is not the case in
richer countries). In developing countries, it seems, the aggrieved stay out of everyday
political life.
For a more complete look at the effects of the covariates on political action we
rely on quantile regression—estimating a series of functions where percentiles of the
conditional distribution of the political action index are expressed as functions of
observed covariates. Quantile regression affords a more complete picture of the shape of
the distribution of the outcome of interest (Koenker and Hallock 2001). We replicate the
equation from column (1) in table 5. The results are shown in figure 3. Each graph plots

14
the estimates of different coefficients for percentiles ranging from 0.5 to 0.95. For figure
3, the dependent variable is the principal-component weighted political action score. For
each covariate, these estimates may be interpreted as the estimated impact of a one-unit
change in the covariate on political action, holding all other covariates fixed. The shaded
areas represent 95% confidence intervals for the quantile regression estimates. The solid
lines show the ordinary least-squares (OLS) estimate of the conditional mean effect,
while the dotted lines represent the usual 95% confidence band for OLS estimates. Most
of the covariate plots show greater dispersion of the effects among the highest
percentiles—i.e., among the most politically active individuals.
Income decile has an increasing effect on political action in the lower end of the
distribution, but becomes negative after the 75th
percentile. Poverty actually increases
political activity among those in the higher end of the distribution. The effect of
perceptions of governmental effectiveness increase then falls to zero towards the higher
percentiles. Perceived economic inequities are a mirror-image to governmental
performance, suggesting that economic injustices galvanize the already-mobilized even
more. The effects of employment and urban location are positive but widely dispersed
among the better-mobilized. Age reduces proclivities towards political activity, but the
effect diminishes among the upper end of the distribution. Gender and education have
consistent, increasing effects. Our quantile regression shows diminishing effects of most
covariates on the upper percentiles of the politically active. For the more active, personal
wealth and grievances matter less than for the less mobilized.
In table 6, finally, we decompose the political action index into its separate parts,
enabling a more detailed view of the individual activities. We code each variable 0 if the

15
respondent claims to have never done these things, 1 if they ―might‖ do it; and 2 if they
have done it. Ordered logit regression results do reveal some differences across
activities. Unlike earlier results, poorer individuals are more likely to demonstrate (but
less likely to engage in other actions). Those who face economic inequities are more
likely to join boycotts or demonstrations but less likely to sign petitions, participate in
strikes or more extreme acts. All other covariates are consistent with earlier results.
Discussion
Radicalism in developing countries seems largely motivated by discontent with
the political system and a lack of economic equity. Unlike the case in richer countries,
wealth produced defenders of the status quo in developing nations. Similarly, poverty
breeds leftism. From our examination of political action—both violent and non-violent—
however, a different pattern emerges. The poor generally shy away from most political
engagements. Violence for political ends does have its supporters among those who
perceive malevolence on the part of their governments, or among those who complain
about economic inequities. But ―support‖ is not action. Controlling for other factors, the
economically and politically marginalized, while susceptible to the appeals of radicalism,
also avoid anti-governmental politics.
Political action against the state depends not only merely on preferences but also
on perceived opportunities. What can be won? What are the chances of success? And
what are the risks involved in the event of failure? The political lives of the marginalized
depend not simply on perceived or actual grievances, but on the possession of resources
needed to convert those demands into concrete action: income, education, and urban

16
residency seem to be critical factors in acquiring these resources. Acts of dissent and
protest, ironically, appear to be the province of those who support the overall polity, its
government, and its modes of economic management, rather than of citizens at the
periphery. Indeed, this is consistent with the common bias towards more politically-
engaged citizens found in matters of government spending and public service-delivery
(Banerjee et al. 2007). Better-off citizens not only tend to engage more in ordinary
political activities, they are also more likely to threaten its collapse.
Conclusions
Overall, we find evidence from survey data that economic and political discontent
breeds radicalism and, at the same time, exclusion. Thus we have the vicious cycle of
legitimation: alienation leads to self-exclusion, self-exclusion further marginalizes
vulnerable groups from access to and representation in policymaking, disenfranchisement
stacks the deck in favor of influential (better off) groups, which reinforces perceptions of
inequality and political failure leading to even more radical political attitudes. It is thus
the better-off who see the opportunities in political action to initiate change, even if it is
the less well-off and marginalized citizens who have the strongest preferences for change.
How can these findings be squared with evidence that poorer communities (e.g.,
in India, and in some Latin American countries) are more politically active? One
possibility is that political action in poor communities is oriented primarily towards the
maintenance of patron-client ties. In communities where public service delivery is
spotty, local brokers, caudillos, zamindars, etc., may provide certain services to local
residents in return for securing support for particular parties. Indeed much of the

17
evidence from Latin America suggests that the poor are more active in political parties,
rather than in protest participation. A second, related possibility is that poor communities
are characterized by greater fragmentation or polarization than richer communities.
Consequently, political action is directed in support of politicians more likely to provide
goods to particular groups. There is evidence from India, for example, of politicians
distributing public goods towards the social groups to which they belong (Banerjee and
Pande 2007). In both cases, increased political activity at the community or village-level
would not necessarily translate into a willingness to engage in anti-governmental dissent.
Understanding better the relationship between alienation and action requires a
more complete picture of the effects household income and demographic factors, as well
as the dynamics of group relations, on political behavior beyond voting—something that
cannot, at the present, be done for more than a handful of developing countries given the
data requirements. Complementing cross-country analyses to understand how economic
conditions and different political preferences influence conflict is a potentially fruitful
agenda for future data-collection and research.
Finally, of particular importance would be to identify the role of political
institutions and economic policies for fostering inclusion. Common approaches are not
without their pitfalls. Free and periodic multi-party elections give marginalized groups a
potential voice, risk becoming captured by better-organized groups. Public policies in the
form of public good provision, food subsidies, and cash transfers can also foster inclusion
both directly (if groups observe that they can benefit from political participation) as well
as indirectly by providing resources for education, but these types of programs require
certain governmental capacities for effective targeting and for relatively high returns to

18
education. Decentralization of policy making, minority quotas, and community-based
developmental interventions carry the potential to increase the incentives to limit self-
exclusion, but these outcomes can be heterogeneous across different groups. These
interactions between the poor, non-poor, and public officials as well as the calculations
involved in the provision of policies aimed at fostering inclusion require further
investigation and analysis. From our perspective a central obstacle to these approaches is
not that program designers lack knowledge, managerial skill, or fiscal resources, but that
politicians face strong incentives to use redistributive programs for partisan purposes,
many of which require the continued political marginalization of the poor.

19
References
Alesina, Alberto, Rafael Di Tella, and Robert MacCulloch. 2004. ―Inequality and
Happiness: Are Europeans and Americans Different?‖ Journal of Public
Economics 88 (9-10): 2009-42.
Aslaksen, Silje , and Ragnar Torvik. 2006. ―A Theory of Civil Conflict and Democracy
in Rentier States*.‖ Scandinavian Journal of Economics 108 (4): 571-85.
Banerjee, Abhijit, Lakshmi Iyer, and Rohini Somanathan. 2007. ―Public Action for
Public Goods.‖ Working Paper 12911. National Bureau of Economic Research,
Cambridge, MA.
Banerjee, Abhijit, and Rohini Pande. 2007. ―Parochial Politics: Ethnic Preferences and
Politician Corruption.‖ CEPR Discussion Paper 6381. Centre for Economic
Policy Research, London.
Barron, Patrick, Kai-Alexander Kaiser, and Menno Pradhan. 2004. ―Local Conflict in
Indonesia : Measuring Incidence and Identifying Patterns.‖ Policy Research
Working Paper 3384. World Bank, Washington, D.C.
Bertrand, Marianne, and Sendhil Mullainathan. 2001. ―Do People Mean What They Say?
Implications for Subjective Survey Data.‖ American Economic Review 91 (2): 67-
72.
Besley, Timothy, Rohini Pande, and Vijayendra Rao. 2005. ―Participatory Democracy in
Action: Survey Evidence from South India.‖ Journal of the European Economic
Association 3 (2-3): 648-57.
Blais, André. 2000. To Vote or Not to Vote? The Merits and Limits of Rational Choice
Theory. Pittsburgh, PA: University of Pittsburgh Press.
Booth, John A., and Mitchell Seligson. 2006. ―Inequality and Democracy in Latin
America: Individual and Contextual Effects of Wealth on Political Participation.‖
Paper presented at Prepared for presentation at the Duke University Workshop on
Poverty and Democracy, Durham, NC.
Bratton, Michael. 2006. ―Poor People and Democratic Citizenship in Africa.‖
Afrobarometer Working Paper 56. IDASA, Cape Town.
Bueno de Mesquita, Bruce, Alastair Smith, Randolph M. Siverson, and James D.
Morrow. 2003. The Logic of Political Survival. Cambridge, MA: MIT Press.
Chaudhry, Kiren Aziz. 1997. The Price of Wealth: Economies and Institutions in the
Middle East. Ithaca: Cornell University Press.

20
Collier, Paul, and Anke Hoeffler. 1998. ―On Economic Causes of Civil War.‖ Oxford
Economic Papers 50 (4): 563-73.
Dalton, Russell J., and Alix van Sickle. 2005. ―The Resource, Structural, and Cultural
Bases of Protest.‖ Paper 05-11. Center for the Study of Democracy, University of
California, Irvine, CA.
Deininger, Klaus. 2003. ―Causes and Consequences of Civil Strife: Micro-Level
Evidence from Uganda.‖ Oxford Economic Papers 55 (4): 579-606.
Diaz-Cayeros, Alberto, and Beatriz Magaloni. 2003. ―The Politics of Public Spending:
The Programa Nacional de Solidaridad (PRONASOL) in Mexico.‖ Background
paper to the World Development Report 2004. World Bank, Washington, DC.
Diwakar, Rekha. 2008. ―Voter Turnout in the Indian States: An Empirical Analysis.‖
Journal of Elections, Public Opinion & Parties 18 (1): 75 - 100.
Dudley, Joseph P, Joshua R Ginsberg, Andrew J. Plumptre, John A. Hart, and Liliana C.
Campos. 2002. ―Effects of War and Civil Strife on Wildlife and Wildlife
Habitats.‖ Conservation Biology 16 (2): 319-29.
Elbadawi, Ibrahim, and Nicholas Sambanis. 2001. ―How Much War Will We See?
Estimating the Incidence of Civil War in 161 Countries.‖ Policy Research
Working Paper 2533. World Bank, Washington, DC.
Graham, Carol, and Andrew Felton. 2005. ―Does Inequality Matter to Individual
Welfare? An Initial Exploration Based on Happiness Surveys from Latin
America.‖ Center on Social and Economic Dynamics Working Paper 38.
Brooking Institution/Johns Hopkins University, Washington, DC.
Hirshleifer, Jack. 1993. ―Cooperation, Conflict, and All That.‖ Working Paper 695.
UCLA Department of Economics, Los Angeles, CA.
Humphreys, Macartan, and Jeremy M. Weinstein. 2006. ―Handling and Manhandling
Civilians in Civil War.‖ American Political Science Review 100 (3): 429-47.
Huntington, Samuel P., and Joan M. Nelson. 1976. No Easy Choice: Political
Participation in Developing Countries. Cambridge, Mass.: Harvard University
Press.
Inglehart, Ronald. 1997. Modernization and Postmodernization : Cultural, Economic,
and Political Change in 43 Societies. Princeton, N.J.: Princeton University Press.
Inglehart, Ronald, Miguel Basáñez, Jaime Diéz-Medrano, Loek Halman, and Ruud
Luikx, eds. 2004. Human Beliefs and Values : A Cross-Cultural Sourcebook
Based on the 1999-2002 Values Surveys. 1st ed. México: Siglo XXI.

21
King, Gary, and Jonathan Wand. 2007. ―Comparing Incomparable Survey Responses:
Evaluating and Selecting Anchoring Vignettes.‖ Political Analysis (15): 46-66.
Koenker, Roger, and Kevin F. Hallock. 2001. ―Quantile Regression.‖ Journal of
Economic Perspectives 15 (4): 143-56.
Krueger, Alan B., and Jitka Maleckova. 2003. ―Education, Poverty and Terrorism: Is
There a Causal Connection?‖ Journal of Economic Perspectives 17 (4): 119-44.
Levi, Margaret. 2006. ―Why We Need a New Theory of Government.‖ Perspectives on
Politics 4 (1): 5-19.
Lipset, Seymour Martin, Kyoung-Ryung Seong, and John Charles Torres Torres. 1993.
―A Comparative Analysis of the Social Requisites of Democracy.‖ International
Social Science Journal 136 155-75.
MacCulloch, Robert, and Silvia Pezzini. 2002. ―The Role of Freedom, Growth and
Religion in the Taste for Revolution.‖ Suntory and Toyota International Centres
for Economics and Related Disciplines, London.
Malapit, Hazel Jean L., Tina S. Clemente, and Cristina Yunzal. 2003. ―Does Violent
Conflict Make Chronic Poverty More Likely? The Mindanao Experience.‖ Paper
presented at ―Staying Poor: Chronic Poverty and Development Policy‖
Conference, Manchester.
Murillo, M. Victoria. 2000. ―From Populism to Neoliberalism: Labor Unions and Market
Reforms in Latin America.‖ World Politics 52 (2): 135-74.
Norris, Pippa. 2002. Democratic Phoenix : Reinventing Political Activism. New York,
NY: Cambridge University Press.
Pedersen, Duncan. 2002. ―Political Violence, Ethnic Conflict, and Contemporary Wars:
Broad Implications for Health and Social Well-Being.‖ Social Science &
Medicine 55 (2): 175-90.
Piazza, James A. 2006. ―Rooted in Poverty?: Terrorism, Poor Economic Development,
and Social Cleavages.‖ Terrorism and Political Violence 18 (1): 159 - 77.
Powell, G. Bingham. 1982. Contemporary Democracies: Participation, Stability, and
Violence. Cambridge, Mass.: Harvard University Press.
Rice, Susan E 2006. ―The Threat of Global Poverty.‖ The National Interest 83: 76-82.

22
Sánchez, Fabio, and Mario Chacón. 2003. ―Conflict, State and Decentralisation: From
Social Progress to an Armed Dispute for Local Control, 1974-2002.‖ Crisis States
Programme Working Paper 70. Development Studies Institute, London.
Schady, Norbert Rüdiger. 2000. ―Political Economy of Expenditures by the Peruvian
Social Fund (FONCODES) 1991-1995.‖ American Political Science Review 94
(2): 289-304.
Stokes, Susan C. 2001. ―Economic Reform and Public Opinion in Fujimori's Peru.‖ In
Public Support for Market Reforms in New Democracies, ed. S. C. Stokes.
Cambridge: Cambridge University Press.
Tadjoeddin, Mohammad Zulfan, and Syed Mansoob Murshed. 2007. ―Socio-Economic
Determinants of Everyday Violence in Indonesia: An Empirical Investigation of
Javanese Districts, 1994-2003.‖ Journal of Peace Research 44 (6): 689-709.
Yousef, Tarik M. 2004. ―Development, Growth and Policy Reform in the Middle East
and North Africa since 1950.‖ Journal of Economic Perspectives 18 (3): 91-116.

23
Table 1: Preferences for Revolution, Logit Regressions
(1) (2) (3) (4) (5) (6)
Radical (1) – (3) Reactionary (4) – (6)
Dev
Countries
High
Income
Countries
Clustered
Errors
Dev
Countries
High
Income
Countries
Clustered
Errors
Income decile -0.0150
(0.0108)
-0.0615**
(0.0275)
-0.0162
(0.0175)
0.0470***
(0.0092)
-0.0120
(0.0151)
0.0488
(0.0693)
Change in political rating -0.0710***
(0.0077)
-0.0360
(0.0265)
-0.0707***
(0.0157)
0.0037
(0.0067)
-0.0588***
(0.0153)
0.0031
(0.0131)
Inequality perception 0.0678***
(0.0077)
0.0477
(0.0291)
0.0669***
(0.0178)
-0.0291***
(0.0070)
-0.0648***
(0.0153)
-0.0289
(0.0339)
Unemployed -0.0710
(0.0817)
0.7775***
(0.2297)
-0.0744
(0.0895)
-0.0057
(0.0731)
0.0835
(0.2076)
-0.0014
(0.1023)
Location 0.0088
(0.0104)
-0.0305
(0.0307)
0.0098
(0.0142)
0.0169*
(0.0099)
0.0202
(0.0169)
0.0167
(0.0187)
Gender 0.3383***
(0.0485)
0.0413
(0.1296)
0.3458***
(0.0769)
-0.0680
(0.0418)
0.0583
(0.0751)
-0.0659
(0.0525)
Age -0.0135***
(0.0017)
-0.0315***
(0.0045)
-0.0133***
(0.0034)
-0.0223***
(0.0016)
0.0150***
(0.0024)
-0.0224*
(0.0131)
Education 0.0468***
(0.0119)
-0.0614*
(0.0361)
0.0470**
(0.0205)
0.0093
(0.0101)
-0.1782***
(0.0202)
0.0083
(0.0522)
Bias -0.0257**
(0.0116)
-0.0249
(0.0381)
-0.0264
(0.0180)
0.0140
(0.0104)
0.0133
(0.0229)
0.0139
(0.0154)
N 40,939 8,616 40,772 40,939 8,616 40,772
R2
0.3514 0.0835 0.3528 0.2752 0.1043 0.2767
Notes: All estimations include intercepts, country and time dummies (not reported), and are weighted by
population-sample weights. Standard errors are based on bootstrap re-sampling of 1,000 replications. Errors in (6)
are clustered by region.
*** p < 0.01
** p < 0.05
* p < 0.10

24
Figure 1: Determinants of Radicalism, Decile Regressions
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
1 2 3 4 5 6 7 8 9 10
Deciles
Coefficient
Inequality
Political
Rating

25
Table 2: Left-Right Ideology, Logit Regressions
(1) (2) (3)
Developing
Countries
High
Income
Countries
Clustered
Errors
Income decile 0.0373***
(0.0073)
0.0227***
(0.0088)
0.0384***
(0.0143)
Change in political rating 0.0795***
(0.0055)
-0.0972***
(0.0100)
0.0801***
(0.0295)
Inequality perception -0.1208***
(0.0059)
-0.1382***
(0.0093)
-0.1224*
(0.0672)
Unemployed 0.1713***
(0.0606)
-0.3367***
(0.1017)
0.1744**
(0.0854)
Location 0.0147**
(0.0074)
-0.0394***
(0.0099)
0.0152
(0.0353)
Gender 0.2238***
(0.0330)
0.1066**
(0.0432)
0.2270***
(0.0537)
Age -0.0011
(0.0011)
0.0100***
(0.0014)
-0.0008
(0.0067)
Education -0.0845***
(0.0083)
-0.0432***
(0.0111)
-0.0857*
(0.0485)
Bias 0.1146***
(0.0090)
0.0519***
(0.0141)
0.1152***
(0.0261)
N 36,777 11,875 36,523
R2
0.4540 0.1240 0.4561
*** p < 0.01
** p < 0.05
* p < 0.10

26
Table 3: Violence, Ordered Logit Regressions
(1) (2) (3)
Developing
Countries
High-Income
Countries
Clustered
Errors
Income decile -0.0048
(0.0057)
-0.0199*
(0.0111)
-0.0048
(0.0109)
(0.0041)
0.0095
(0.0109)
-0.0153***
(0.0035)
(0.0048)
-0.0066
(0.0109)
0.0128**
(0.0052)
Unemployed 0.0897**
(0.0426)
0.0760
(0.1367)
0.0897*
(0.0478)
Location 0.0125**
(0.0056)
-0.0016
(0.0128)
0.0125**
(0.0058)
Gender 0.1224***
(0.0261)
0.1514***
(0.0548)
0.1224***
(0.0339)
Age -0.0038***
(0.0009)
-0.0067***
(0.0017)
-0.0038***
(0.0010)
Education -0.0269***
(0.0069)
-0.0825***
(0.0141)
-0.0269***
(0.0091)
Bias 0.0095
(0.0061)
-0.0054
(0.0168)
0.0095
(0.0089)
N 23,056 6,610 23,056
R2
0.0144 0.0313 0.0144
*** p < 0.01
** p < 0.05
* p < 0.10

27
Figure 2: Support for Violence by Income Decile, Local-Polynomial Regression.102.103.104.105.106
1 2 3 4 5 6 7 8 9 10
Income Decile
95% CI lpoly smooth
Note: Graph shows isolated effect of income decile on individual support for political violence, developing
countries only, based on local-polynomial regression with biweight (quartic) kernels and bandwidths of 2.0.
The shaded band represents 95% confidence.

28
Table 4: Violence and Political Action, Income-Quintile Regressions
(1) (2) (3) (4) (5)
Violence is Justified (1) – (5)
Quintile: 1st
2nd
3rd
4th
5th
(0.0142)
-0.0453***
(0.0150)
-0.0663***
(0.0158)
-0.0722***
(0.0178)
-0.0488**
(0.0233)
(0.0162)
0.0453***
(0.0159)
0.0729***
(0.0179)
0.0176
(0.0204)
0.0579**
(0.0284)
Unemployed 0.1118
(0.1186)
0.0649
(0.1484)
-0.2989
(0.1890)
-0.3136
(0.2599)
-0.6238
(0.4264)
Location -0.0141
(0.0187)
-0.0009
(0.0213)
0.0226
(0.0222)
0.0280
(0.0255)
0.0008
(0.0368)
Gender 0.2343***
(0.0896)
0.2190**
(0.0972)
0.1406
(0.1033)
0.3869***
(0.1100)
0.5798***
(0.1546)
Age -0.0140***
(0.0034)
-0.0191***
(0.0038)
-0.0115***
(0.0039)
-0.0109**
(0.0044)
-0.0090
(0.0056)
Education 0.0227
(0.0241)
0.0713***
(0.0262)
0.0372
(0.0262)
0.0050
(0.0291)
0.0196
(0.0409)
Bias 0.0018
(0.0194)
-0.0586***
(0.0213)
-0.0006
(0.0242)
-0.0789***
(0.0294)
-0.1066***
(0.0391)
N 8,589 12,763 8,569 6,901 2,330
R2
0.7777 0.1576 0.5860 0.2604 0.0993
population-sample weights. Standard errors are based on bootstrap re-sampling of 1,000 replications.
*** p < 0.01
** p < 0.05
* p < 0.10

29
Table 5: Propensity to Engage in Political Action
(1) (2) (3) (4) (5) (6)
Political Action Index
(principal components)
Political Action Index
(5-component sum)
Dev
Countries
High
Income
Countries
Clustered
Errors
Dev
Countries
High
Income
Countries
Clustered
Errors
(0.0048)
0.0498***
(0.0079)
0.0183**
(0.0084)
0.0532***
(0.0188)
0.0549
(0.0389)
0.1559***
(0.0586)
(0.0032)
0.0266***
(0.0079)
0.0274***
(0.0099)
0.1246***
(0.0119)
0.1228**
(0.0492)
0.1062**
(0.0439)
(0.0033)
0.0607***
(0.0078)
-0.0183
(0.0226)
-0.0820***
(0.0129)
-0.0831
(0.1032)
0.3174***
(0.0740)
Unemployed -0.0984**
(0.0384)
-0.0143
(0.1078)
-0.0903**
(0.0446)
-0.3525**
(0.1431)
-0.3201**
(0.1626)
-0.0434
(0.4278)
Location 0.0238***
(0.0048)
0.0765***
(0.0092)
0.0251
(0.0158)
0.1257***
(0.0181)
0.1303*
(0.0745)
0.3343***
(0.0830)
Gender 0.4946***
(0.0214)
0.2424***
(0.0392)
0.4963***
(0.0268)
2.0557***
(0.0820)
2.0574***
(0.1219)
1.1597***
(0.2480)
Age -0.0104***
(0.0007)
-0.0212***
(0.0011)
-0.0103***
(0.0025)
-0.0387***
(0.0027)
-0.0380***
(0.0084)
-0.0799***
(0.0063)
Education 0.0924***
(0.0053)
0.1805***
(0.0103)
0.0935***
(0.0145)
0.2525***
(0.0208)
0.2542**
(0.1037)
0.6874***
(0.0604)
Bias -0.0241***
(0.0050)
0.0210*
(0.0112)
-0.0239**
(0.0097)
-0.0669***
(0.0193)
-0.0659
(0.0422)
0.2530**
(0.1234)
N 35,595 13,592 35,332 44,655 44,366 13,054
R2 0.3450 0.4179 0.3462 0.3080 0.3091 0.4430
and in (6) are clustered by region.
*** p < 0.01
** p < 0.05
* p < 0.10

30
Figure 3: Determinants of Political Action, Quantile Regressions
-0.050
-0.040
-0.030
-0.020
-0.010
0.000
0.010
0.020
0.030
0.040
0.050
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95
Percentile
Decile
-0.030
-0.020
-0.010
0.000
0.010
0.020
0.030
0.040
0.050
0.060
0.070
0.080
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95
Percentile
PoliticalRating
-0.150
-0.100
-0.050
0.000
0.050
0.100
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95
Percentile
Inequality
-0.400
-0.350
-0.300
-0.250
-0.200
-0.150
-0.100
-0.050
0.000
0.050
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95
Percentile
Unemployment
-0.010
0.000
0.010
0.020
0.030
0.040
0.050
0.060
0.070
0.080
0.090
0.100
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95
Percentile
Location
-0.200
0.000
0.200
0.400
0.600
0.800
1.000
1.200
1.400
1.600
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95
Percentile
Gender
-0.030
-0.025
-0.020
-0.015
-0.010
-0.005
0.000
0.005
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
Percentile
Age
-0.050
0.000
0.050
0.100
0.150
0.200
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95
Percentile
Education
Notes: Graphs show coefficients for covariates in regression from table 5, column (1) by sample percentiles based
on a quantile regression with bootstrapped confidence intervals.

31
Table 6: Types of Political Action, Ordered Logit Regressions
(1) (2) (3) (4) (5)
Petition Boycott Demonstrate Strike Occupy
(0.0069)
0.0024
(0.0072)
-0.0175**
(0.0072)
-0.0060
(0.0081)
0.0429***
(0.0097)
(0.0046)
0.0384***
(0.0049)
0.0368***
(0.0047)
0.0411***
(0.0055)
0.0241***
(0.0069)
(0.0049)
0.0304***
(0.0050)
0.0216***
(0.0049)
-0.0500***
(0.0057)
-0.0276***
(0.0069)
Unemployed -0.3782***
(0.0522)
-0.0745
(0.0580)
-0.0517
(0.0514)
-0.1329**
(0.0651)
-0.0014
(0.0769)
Location 0.0242***
(0.0064)
0.0143**
(0.0070)
0.0155**
(0.0068)
0.0667***
(0.0080)
-0.0046
(0.0099)
Gender 0.5052***
(0.0298)
0.6902***
(0.0321)
0.5142***
(0.0314)
0.7122***
(0.0365)
0.7025***
(0.0454)
Age -0.0069***
(0.0010)
-0.0087***
(0.0011)
-0.0111***
(0.0010)
-0.0245***
(0.0012)
-0.0367***
(0.0016)
Education 0.1643***
(0.0075)
0.1293***
(0.0078)
0.1653***
(0.0079)
0.0191**
(0.0090)
0.0986***
(0.0111)
Bias 0.0027
(0.0071)
-0.0621***
(0.0079)
-0.0093
(0.0075)
-0.0364***
(0.0090)
-0.0582***
(0.0108)
N 41,384 39,575 40,002 41,104 37,989
R2
0.2545 0.2303 0.1547 0.2334 0.2037
population-sample weights. Standard errors are based on bootstrap re-sampling of 1,000 replications.
*** p < 0.01
** p < 0.05
* p < 0.10

Poverty, Exclusion, and Dissent - Support for Regimes in Developing Countries

More Related Content

What's hot

Similar to Poverty, Exclusion, and Dissent - Support for Regimes in Developing Countries

More from Stockholm Institute of Transition Economics

Recently uploaded

Poverty, Exclusion, and Dissent - Support for Regimes in Developing Countries