Fifty Years of Ethnic Conflict and Cohesion: 1945-94

Author(s): Noël Bonneuil and Nadia Auriat

Source: Journal of Peace Research, Vol. 37, No. 5 (Sep., 2000), pp. 563-581
Published by: Sage Publications, Ltd.
Stable URL: https://www.jstor.org/stable/425279
Accessed: 24-07-2019 22:02 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

Sage Publications, Ltd. is collaborating with JSTOR to digitize, preserve and extend access to
Journal of Peace Research

This content downloaded from 200.27.72.243 on Wed, 24 Jul 2019 22:02:24 UTC
All use subject to https://about.jstor.org/terms
f^^^^BSS ? 2000 Journal of Peace Research,
vol. 37, no. 5, 2000, pp. 563-581
Sage Publications (London, Thousand Oaks,

f^ CA and New Delhi)

^f^frA<e'i^- [0022-3433(200009)37:5; 563-581; 014112]

Fifty Years of Ethnic Conflict and Cohesion:

1945-94

NOEL BONNEUIL

Institut national des etudes ddmographiques e& Scole des hautes 6tud
sciences sociales

NADIA AURIAT

UNESCO, Division of Social Science, Research and Policy, MOST

Programme

The diversity of conflict sequences observed among 163 ethnic groups ranging from 1945 to 1994 is
portrayed in correspondence with political and socio-economic factors. This diversity comes first from
the strong association between mobilization, slight discrimination, migration distress, religion and
repression. On a finer detail, discrimination appears to be associated with resistance if it is related to
land and power, with war if it concerns social mobility, or with insurgency if it has to do with social
customs. Migration from rural to urban and abroad is accompanied by high mobilization and rioting
or war when social mobility is at stake. Declining state power and democratization can open the door
to violent action. Finally, there is also international diffusion of protest. Locating minority groups in
this landscape of correspondence helps to compare and characterize the various particular histories.
These range from Lebanon's period of war, or the years of insurgency in Iran or Somalia, to the mixture
of verbal opposition and terrorism in Western democracies. Crystallizing protest in India or China is
differentiated against the deterioriation of group coherence in the Middle East; sporadic bursts of vio-
lence in Africa are contrasted against insurgency and rioting in East and Southeast Asia. The results
point to the need for conflict prevention policies to pay more attention to the promotion of equitable
social mobility.

Introduction and political actions' of communal groups. He

Gurr (1993a: 162) advocated the empirical
analysis of conflict from study of 'the status
* The views expressed, the designations employed and the
presentation of material throughout this article do not
imply the expression of any opinion whatsoever on the part
of the UNESCO Secretariat concerning the legal status of
any country, territory, city or area of its authorities, or the
delimitations of its frontiers or boundaries. The data used
in this article can be obtained from: http:llwww.
prio.no/jpr/datasets.asp. Correspondence should be
addressed to the authors by e-mail: bonneuil@ined.fr and
n.auriat@unesco.org.
collected the conflict histories of 233 minority
groups, documenting their major political and
socio-economic determinants in 1980, and of
268 groups in the 1990s. These data comprise
the 'Minorities at Risk' database.1
Gurr (1993a,b, 1994, 1997) has analyzed
1 UNESCO, Social and Human Sciences Sector (MOST
Programme), purchased this database from the Centre for
International Development and Conflict Management,
University of Maryland at College Park. We completed this
dataset for the period 1990-97 from the Minorities at Risk
(MAR) website: http:/lwww.bos.umd.edu/cidcm/mar.

563

This content downloaded from 200.27.72.243 on Wed, 24 Jul 2019 22:02:24 UTC
All use subject to https://about.jstor.org/terms
564 journal of PEACE RESEARCH volume 37/ number 5 / september 2000

this dataset extensively, employing multiple of their relationships', whereas we only have
regression analysis and statistical modelling, macro-level variables. Aminzade (1993), in a
assessing which aggregated variables correlate case-study of the revolutionary communes in
with ethnic conflict. We suggest a comple- France in 1870-71, also highlighted the
mentary approach based on two key premises: importance and role of sequences of
we focus on sequences of conflict rather than motivated actions in understanding the con-
on conflicts at a given period; instead of flict. Mahoney (1999) provided an illumi-
modelling causal patterns through economet- nating discussion on macro-causal analysis in
ric techniques, we use correspondence analy- history.
sis so as to visualize the structure of the dataset Our main purpose here is to offer com-
in terms of dispersion and multiple corre- prehensive displays of histories of ethnic con-
lation among the most frequent sequences of flict since 1945, highlighting the richness
ethnic conflicts from 1945 to 1994, and in and the complexity of this qualitative dataset
terms of their statistical associations with and addressing competing explanations for
socio-political factors. We then position each why ethnic conflict occurs. We can anchor
minority group in this landscape. our argument in three contrasting theoretical
By doing this, we are not building a model schools - called intrumentalism, primordial-
in the usual sense, where we would test the ism and constructivism. Respecting the diver-
effect of a given determinant on a given sity of the data and their temporal structure,
response. Instead, we consider all statistical we offer an empirical backdrop for some of
correlations, even for collinear variables, thusthese theories. According to the instrumen-
helping to reveal relevant variables, to pin- talist approach, 'echoing rational choice
point misleading covariations, to detect theory' (Vaiyrynen, 1994: 8), the ethnic
potential outliers, which also enables assess- endowment of an individual or a group con-
ment of the quality of the data. tains a repertory of social resources and roles
Our proposal to rework Gurr's data on that can be drawn upon to pursue material
ethnic mobilization and conflict by means ofgain. This approach 'considers violence a
correspondence analysis and textual rep- means to a variety of ends' (Bates, 1983;
resentation of local histories comes close to Fearon & Laitin, 1996; Hardin, 1995;
the ideas of Abbott (1998: 176). This latter Vayrynen, 1994: 8). Primordialism, on the
scholar acknowledged that 'interactional contrary, 'stresses kinship and blood ties, and
fields are probably too complex for us to associated myths and rituals that cement an
predict', but that 'they do show various inter- ethnic group into a Gemeinschafr. This theory
nal patterns; they do sketch the "rules of thealso views ethnic conflict as a consequence of
game"; they do portray the limits and possi-group identity, itself viewed as an integral
bilities of action'. Franzosi & Mohr (1997: part of personal identity, and it maintains
150-151) and Franzosi (1997, 1998), in that ethnic attachment is a natural kind of
exploring the origins of Italian fascism,affiliation'
were (Geertz, 1993; Stavenhagen,
1990; Vayrynen, 1994: 9). Finally, the con-
similarly concerned with 'processes of mobil-
isation (and counter-mobilisation) unfolding
structivist approach 'stresses the manufacture
of ethnic identity' and 'the crystallisation of
over time', and with turning 'textual infor-
mation into a usable data source, with the
the group for a particular societal purpose',
collection, organisation, and analysis of notably conflict (Thompson, 1991; Ander-
narrative'. In fact, Franzosi (1997) went son, 1991; Hobsbawm, 1990; Schaeffer,
further than ourselves, for he was able to 1990; Vayrynen, 1994: 10).
address 'actors and actions, and the network Tilly (1995, 1997a,b), drawing on human

This content downloaded from 200.27.72.243 on Wed, 24 Jul 2019 22:02:24 UTC
All use subject to https://about.jstor.org/terms
 Noel Bonneuil 6' Nadia Auriat ETHNIC CONFLICT AND COHESION, 1945-94 565

psychology much in the spirit of Elster

variable describing rebellion from 1945-49
(1999), clarified the very specific historical
to 1985-89. Given this methodological diffi-
nature of every conflict and noted how any we constructed a single variable 'rebel-
culty,
lion' ranging from 1945-49 to 1990-94 (the
one causal theory was insufficient to explain
the different outcomes over time of thethird
same item of rebellion for 1990-94, 'local
determinant. Our treatment of temporal
rebellion - armed attempts to seize power in
a locale' (E), is then assimilated to 'small-scale
sequences, based on an original use of textual
analysis (Lebart et al., 1998), should help us
insurgency' (u)). We proceeded similarly for
nonviolent
to portray the diversity of the historical tra- protest in assimilating items of
jectories as recorded from 1945 to 1994, protest
and in 1990-94 (Gurr, 1997: 1089) to
to identify 'verifiable causal stories nestled in items of the variable 'nonviolent
specific
differing chains of cause-effect relations'
protest' as follows: 'a few demonstrations,
(Tilly, 1997b: 50). Our purpose, much strikes,
in the rallies; limited rioting; other' to the
spirit of factor analysis, is to provide item d; 'a number of demonstrations, strikes,
a clear
view of types of individuals, and israllies;
thus substantial rioting' to the item e; and
different from usual causal regression 'major
analy- demonstrations, serious and wide-
sis, which emphasizes the interaction ofspread
vari- rioting' to the item f This approxi-
ables. We favour correspondence analysismation is a little imperfect, but it has the
considerable advantage of increasing the
over other techniques because we acknow-
ledge Tilly's recommendation to consider the
time-series for nonviolent protest. The time-
series of the variable 'violent protest' was not
specificity of each minority's history. Repre-
senting patterns of conflict sequences in completed for the 1990-94 period and is
association with covariates enables us to therefore limited to 1945-49 to 1985-89.
discuss contending theories of ethnic conflict We retained these 163 'minorities at risk',
in the light of the empirical mapping. because they can be followed from 1945-49
to 1990-94. We note that Gurr (1993a)
Methods tackled the testing of a complex causal model
showing the dynamics of conflict between
The Minorities at Risk Dataset ethnic groups and states. The database
Gurr (1993a) worked around the premise includes political, ethnic and demo-econ-
omic factors in the 1980s and 1990s, charac-
that there were essentially three forms of con-
flict: nonviolent protest, violent protest and terizing minorities and their relationships to
rebellion. In his original Minorities at Risk dominant groups (Gurr, 1994: 349: MAR
dataset (MAR), Gurr (1993a) created these website). Gurr compared living conditions of
three conflict variables by aggregating a series minorities to those of the dominant group,
of indicators of protest which are assumed to thereby producing a set of economic,
be correlates of one of the three categories of
conflict.2 Of the 233 groups of the original 2 Nonviolent protest: a, 'none reported'; b, 'verbal opposi-
dataset, 163 were completed for the 1990s: tion'; c, 'political organizing activity on a substantial scale';
d, 'a few demonstrations or strikes'; e, 'a number of demon-
new groups and new variables were added for strations or strikes'; f, 'major demonstrations'; m, 'missing
this period, and the three variables of conflict value'. Violent protest: A, 'none reported'; B, 'scattered acts
of sabotage'; C, 'limited rioting'; D, 'substantial rioting'; E,
specified between 1945-49 and 1985-89
'serious and widespread rioting'; F, 'local rebellions - armed
were merged into two after 1990, namely attempts to seize power in a locale'; M, 'missing value'.
'protest' and 'rebellion'. The variable describ- Rebellion: r, 'none reported'; s, 'political banditry'; t, 'cam-
paigns of terrorism'; u, 'small-scale insurgency'; v, 'large-
ing rebellion for 1990-94 (Gurr, 1997: scale insurgency'; w, 'protracted civil war'; z, 'missing value'
1089) is unequivocally connected with the
(Gurr, 1993a: 169).

This content downloaded from 200.27.72.243 on Wed, 24 Jul 2019 22:02:24 UTC
All use subject to https://about.jstor.org/terms
566 journal of PEACE RESEARCH volume 37 / number 5 / september 2000

political and cultural covariates. He com-

groups in the countries in which they reside'
mented and graphically portrayed global and
and they 'were the focus of political mobilisa-
tion and action in defense or promotion of
regional patterns and trends in ethnopolitical
conflict from 1945-49 to 1985-89 (Gurr, the group's self defined interests' (Gurr,
1993a). Generally, the data show that there1993a:
is 5). Further criteria for a minority
a rise in forms of action undertaken group by to be considered in the study were that
the communal groups at risk resided in a
minority groups and that the trends in con-
flict vary markedly from region to region.country
A that had a population of at least 1
million in 1985; and that the group itself had
global observation of the data shows a general
rise in all forms of communal conflict, over
a population of at least 100,000 or, if less, in
time, with an acceleration of groups entering
excess of 1% of the population of at least one
into conflict in the 1960s and 1970s com- country in which the group resided (for
pared to earlier periods. Gurr commenteddetailed information on the conceptual
comprehensively upon regional trends ofclarification of a communal group, see Gurr,
conflict, for ethno-nationalists, indigenous 1993a, 1994).
peoples, ethnoclasses, communal contenders As noted by Tilly (1978), the legal defi-
and militant sects. He concluded that the nitions separating strikes from other forms of
grievances of contemporary minorities were conflict (such as demonstrations and 'riots')
driven mainly by political and economic themselves result from previous struggles
dynamics, and reasoned that 'political and among workers, employers and government
economic disadvantages motivate communal officials. They vary from one country and era
groups to demand greater access to the politi- to another. Other complications follow:
cal system and greater economic oppor- workers in different countries sometimes
tunities, whereas a history of political pursue similar ends by different means, and in
autonomy leads groups to attempt secession' moments of deep political division observers
(Gurr, 1993a: 87). and participants assign different meaning to
strikes than in moments of relative tranquil-
lity. We agree with Tilly's observation. But
Building Sequences of Conflicts
when using the MAR dataset, our analysis is
inevitably based on the conceptual foun-
Each of the 163 minority groups documented
from 1945-49 to 1990-94 was scored for the
dations used by Gurr in his data collection.
most widespread and intense event reported Our original methodological contribution
during each quinquennial period. With consists of regarding minority trajectories as
respect to Gurr's original dataset, we chose words
to containing ten letters (or nine in the
disregard summary or aggregated indices and
case of violent protest), that we have assigned
on the basis of one letter for each of the ten
to work directly with the primary variables,
which are listed in the Appendix. Gurr quinquennial periods ranging from 1945-49
(1993a) defined minorities as communal to 1990-94. For example, if a minority's
history in nonviolent protest is: 1945-49:
groups of people ethnically distinct from the
rest of the population and numerically in- none reported ('a'), 1950-54: verbal opposi-
ferior. The selected minorities are 'non state tion ('b'); 1955-59: none reported ('a');
communal groups that were politically salient1960-64: a few demonstrations ('d');
during the post-World War II era, that is, to1965-69: a number of demonstrations ('e');
say politicized communal groups'; they 'col-1970-74: a number of demonstrations ('e');
lectively suffer, or benefit, from systematic1975-80: a number of demonstrations ('e');
discriminatory treatment vis-a-vis other 1980-85: a few demonstrations ('d');

This content downloaded from 200.27.72.243 on Wed, 24 Jul 2019 22:02:24 UTC
All use subject to https://about.jstor.org/terms
 Noel Bonneuil e& Nadia Auriat ETHNIC CONFLICT AND COHESION, 1945-94 567

1985-89: a few demonstrations ('d'); conflict histories and a set of covariates. The
1990-94: none reported ('a'), for purposes of
contingency table in which we are interested
our analysis, its chronological history appears
is defined by the proportion of cases sharing
as a 'word', which is 'abadeeedda'. a given sequence with any item of a given
This 'word' is then segmented into covariate. This table can be treated through
sequences of all possible lengths, ranging correspondence analysis.
from two to ten letters, which gives the
following 44 patterns:
Correspondence Analysis
ab, ba, ad, de, ee, ed, dd, da; aba, bad, ade, dee,
eee, eed, edd, dda; abad, bade, adee, deee, Correspondence analysis (Benzecri, 1992;
eeed, eedd, edda; abade, badee, adeee, deeed, Greenacre, 1984; Volle, 1985; Bry, 1996; des
eeedd, eedda; abadee, badeee, adeeed, deeedd, Netumieres, 1997) is an appropriate tool for
eeedda; abadeee, badeeed, adeeedd, deeedda;
representing contingency tables. In order to
abadeeed, badeeedd, adeeedda; abadeeedd,
badeeedda; abadeeedda. sketch its principle for those unfamiliar with
this technique, we consider a contingency
Since our purpose in this particular studytable between two variables A and B with I
does not reside in studying the length of any
and/items respectively, with kij observations
given state of conflict, we collapsed any
at item i for the first variable and j for the
repeated segment into its correspondingsecond. Then, the empirical frequency in the
single letter. To return to our example, our
cell (i,j) becomesfy = k/X=,k; and a point
list is then reduced to 23 elements:
Xi is defined as having the coordinatesB :=f
ab, ba, ad, de, e, ed, d, da; aba, bad, ade, eda,{f :=fij/ Xk-kJfk,j = 1, ...J}. The cloud of
ded; abad, bade, aded, deda; abade, baded, points Xi, I = 1, ..., Iis scattered alongJ- 1
adeda; abaded, badeda; abadeda.
dimensions. Correspondence analysis con-
A simple conflict history would thus be sists of finding the axes - those principal axes
- that give most of the variance of this cloud,
the succession of two letters, corresponding
to the succession of two types of differentvariance representing information. Thus, the
conflicts over the period; a word made of orthogonal
a projection onto the two-dimen-
single letter represents a stationary state sional first principal plane (made of Axis 1,
during at least two periods; the presence oftaking the largest part of the variance) and
the second-largest (Axis 2, orthogonal to Axis
long sequences testifies a history of signifi-
cant change. We adopted the following 1) provide the best two-dimensional approxi-
guidelines for coding the types of conflict: mation of the data. However, it is obvious
that part of the information is lost in passing
from a multi- to a two-dimensional space. In
* Nonviolent forms of protest: small letters
this study, information is completed by
ranging from 'a' to t ; missing values are
coded 'i'. examining higher order axes, such as the
* Violent forms of protest: capital letters third and fourth.
ranging from 'A' to 'F'; missing values are The intersection of the principal axes rep-
coded 'M'. resents the marginal distribution {t=, j, j =
* Forms of rebellion: small letters ranging1,... J}. The distance between two points X
from 'r' to 'w' ; missing values are codedand Xi' on the two-dimensional projection is
z. the projection onto this plane of the distance

^-1 6(ifj)2/X'I f. between the distri-

Each minority group is then defined bybutionsfB
a and/B; this distance must be inter-
preted in taking into account distorting
series of 'word' segments reflecting their

This content downloaded from 200.27.72.243 on Wed, 24 Jul 2019 22:02:24 UTC
All use subject to https://about.jstor.org/terms
568 journal of PEACE RESEARCH volume 37 / number 5 / september 2000

perspective effects that arise from any projec- different in terms of their patterns of conflict
tion. Notably, the proximity between points sequences, and how these patterns can be
can be interpreted only when these projected associated with covariates. The qualitative
points are located at the periphery of the nature of the data is respected, providing a
cloud; the closer to the origin of the axes, the more detailed picture than what could be pro-
less interpretable they are. Points, and hence vided by aggregated indices. It is particularly
variables, contributing too little to the prin- interesting since the aggregated indices con-
cipal axes are erased from the graphs. The structed by Gurr (1993a: 101-112) emerge
weight with which a point contributes to a from qualitative indicators of conflict which
given axis is an essential indication, and statistically must not be placed on a uni-
points appearing far away on the axis but dimensional quantitative scale. For example,
with a weak contribution must be excluded Gurr (1993a) regrouped, under the indicator
from the interpretation. Only those variables of'Non-Violent Protest', the variables verbal
giving the biggest contributions to a given opposition and large demonstrations. Although
axis are retained. Here, we do not publish the we agree that they are both nonviolent forms
weights of contributions, nor the various of protest, they are different phenomena that,
qualities of each representation simply for one could argue, deserve treatment per se.
lack of room, but these are available upon Similarly, Gurr regrouped under a single scale
request. Moreover, the principal axes for the of rebellion measures ofscatteredacts ofsabotage
cloud of points YJ with coordinates and protracted civil war.
{fJi :=flk=lfklX i = 1,. . I} are the same, so Correspondence analysis is usually
that points YJ and Xi can be put on the same applied to the contingency table of a given set
representation, and proximity between theirof variables crossed with itself; however, it is
projections can be interpreted, provided the possible to work with a contingency table
latter are located sufficiently at the periphery.that has different variables in its rows and
Finally, supplementary points can be added,columns (for example, with the procedure
representing distributions playing no role inCORRESP, option TABLE in the SAS soft-
the making of the axes: notably here, desig-ware package). This is one of the principles of
nations of minority groups are supplemen- 'textual analysis' (Lebart et al., 1998), which
tary variables that enable us to locate the is an adaptation of correspondence analysis
groups on the graphs. and which has similar features to our
Thus, instead of searching for causal linksapproach. Here, we build our own 'text' by
controlling for 'everything else equal', as in considering the historical sequences of con-
usual regression analysis, this correspondenceflict as words. As usual in correspondence
analysis permits visualizing the diversity of allanalysis, we select those items and those
factors as they are taken together, so that sequences of conflict appearing with 'suf-
different types of minorities can be identifiedficient' frequency (in this case 20, obtained
and characterized by socio-economic and after trial and error), to give enough weight
political variables. The objective of grouping to points so as to avoid serious distortions in
the results (Volle, 1985; Cibois, 1997).
minority groups according to their sequences
of conflict and their conflict determinants
prevails, in our analysis, over the notion of Results
causality between variables (des Netumieres,
1997) (notably, there is no difficulty in repre-The Four StructuringAxes
We begin by interpreting Figures 1 and 3,
senting collinear variables); we are able to
clarify how minority groups are similar or which provide information on sequences of

This content downloaded from 200.27.72.243 on Wed, 24 Jul 2019 22:02:24 UTC
All use subject to https://about.jstor.org/terms
 Noel Bonneuil & Nadia Auriat ETHNIC CONFLICT AND COHESION, 1945-94 569

Figure 1. First Principal Plane of the Correspondence Analysis of Sequences of Conflict Between
1945-49 and 1990-94 and Covariates

rstr
0.4 - rst
d
0. ACA
tr

b db
t
0 mbiL'
rebe for
llon 1 CA s
AD
West. Democ.
rut
od
o
ru odo AC
3e
moderate ieC? e
be dc cb
birth r. power minor rur. to
0
voluntary catholic urb. migration
migration
pop. g.<]
IV ~ UAD slight diff.
medium high
democracy 7 In income
ectil in profession
po icy autonomy highly high (.
mobil. 3 salient feature i.t diff- 90s ome. diff.
medium (3) mobil. 5, 6 slight diff. mn resilence
int. diff. 70s 90s slight diff. .i comnerce DA
0.0 iccVs5 in n n FA ADAec AK

weak identity 90s

service muslim nr ur mob. tor ce

dem ocracy 6 mult. sects moderate ac no, territorial shi'ite protest 4

migratiQn collectivlity sliht diff eo
mobil. 0 abroad restric.ted rvu ln offiual
animist
migration
repressive colectmty budhist PoS1itons W
matrial mosaic 90s policy restrict weak (1) M!
neea sunni eco. opportunities int. diff. 90s c
diffuse pol.grievances light diff. seek change all migrations
m lesser sanent Issue
m

no distinctive po :ice salient Issuerurs Ouer5oolsb m bib fo c rue

status migration reie ion ruw
M
.bi AM
r seriours migration mobil. 7
am wr ru us

democracy 2 mobil. for rn

a)
rebellion 3 v AF
hardship
migration ma
rlr AFA
a)
bI

-0\.
MA

-0.4 -

-0.2 0.0 0.2 Axis 1 0.4

G- Missing values Rise of war, violent conflict,

and mobilization

Total variance for axes 1 and 2, 12.4% and 9.0%, respectively. Int diff. means internatio
group protest. Democracy and state power are scored from 0-9; international diffusion
ization, mobilization for protest and mobilization for rebellion from 0-7. Scores appear on
parentheses. Nonviolent protest: a, 'none reported'; b, 'verbal opposition'; c, 'political organ
on a substantial scale'; d, 'a few demonstrations or strikes'; e, 'a number of demonstrations
'major demonstrations'; m, 'missing value'. Violent protest: A, 'none reported'; B, 'scattered
tage'; C, 'limited rioting'; D, 'substantial rioting'; E, 'serious and widespread rioting'; F,
armed attempts to seize power in a locale'; M, 'missing value'. Rebellion: r, 'none reporte
banditry'; t, 'campaigns of terrorism'; u, 'small-scale insurgency'; v, 'large-scale insurgency
civil war'; z, 'missing value'.

This content downloaded from 200.27.72.243 on Wed, 24 Jul 2019 22:02:24 UTC
All use subject to https://about.jstor.org/terms
570 journal of/PEACE RESEARCH volume 37 / number 5 / september 2000

strikes
conflict and their covariates. Figures 2 and 4 ('d') on the upper side of Axis 2 on
give supplementary information onFigure
the 1 contrast with missing information
minority groups. As a basic principle of 'm') and with large-scale insurgency
('M',
interpretation, the greater the variance
('rvr') situated on the lower side of this axis;
accumulated by the selected axes, the better
similarly, moderate or voluntary migration,
the perspective given by the graph. Here, Catholics,
the Western democracies and weak
first two axes accumulate 21.4%, the firstmobilization
four for rebellion (score of 1) on the
axes 37.0%. For such an important mass of side of Axis 2 are opposed to weak
upper
data gathering 82 variables (or 360 items) and
democracy (score of 2), hardship migration
andthe
87 retained sequences, the 37% testifies to medium level of mobilization for rebel-
heterogeneous nature of the sequences lion
and (score of 3) on this axis's lower side.
In Figure 3, the right-hand side of Axis
provides us with a good (considering the size
of the contingency table) summary of 3the
(8.1%) 'crystallizes' protest: political
sample on the first four axes. banditry ('s'), political organizing ('c')
Axis 1 (12.4% of the total variance)turning
on into demonstrations and verbal
opposition ('cd', 'cb'); it can also be seen that
Figure 1 contrasts the rise of war and violent
it opposes high international diffusion of
conflict on the right-hand side with missing
values on the left-hand side, together with
group protest (score 4 on a 0-5 scale) and dis-
crimination associated with land on the
rising mobilization (for protest, for rebellion,
in general), slight discrimination, restriction
right-hand side of Axis 3, to the deterioration
to collectivity and to economic oppor- of the group's coherence: advantaged minor-
tunities, serious constrained migration ity
andchallenged, serious emigration abroad and
specific religions - Buddhist, Muslim, Shi'ite
high mobilization (score of 7). Axis 4 (7.5%)
- on the right-hand side opposed to the on Figure 3 contrasts a deteriorating situation
absence of mobilization, of distinctive status,
of sporadic bursts of violence (substantial
rioting, 'D'; serious and widespread rioting,
of identity, to animism and multiple sects on
the left-hand side. It should be noted that 'E'; insurgency, 'u', 'v'; and protracted civil
war, 'w') to groups in a permanent situation
Figure 1 indicates missing values are not ran-
domly distributed in this dataset and that of insurgency and rioting ('vr', 'vu', v, 'ruv');
however, never degenerating into war. This
they explain a large part of the variance. The
lower left-hand quadrant in Figure 1 shows opposition is concomitant with weak democ-
racy (score of 2), slight discrimination in
that missing values are associated with 'hard-
ship migration' and 'no distinctive status'. state-related services (higher education,
One could infer that these missing values are official position, income, recruitment to
associated with specific ethnic groups that police) on the upper side, opposed to high
also appear in the lower left-hand quadrant ofstate power, high democracy (score of 8),
Figure 2. medium mobilization, share of public funds
Mobilization scores are gradually located on the lower side. The percentage of the vari-
along Axis 1: 0 and 3 on the left-hand side, ance falls to 4.9% for the fifth principal axis.
then 5 and 6 on the right-hand side near Axis As previously mentioned, variables
2, then 7 at the extreme right-hand side. Axis located at the origin of the graph do not con-
2 (9.0%) shows the rise of political banditry tribute significantly to the axes and their
('s'), scattered acts of sabotage ('B') and proximity to any other variable cannot be
interpreted. For the sake of clarity, we have
terrorism ('t'), contrasted with large-scale
erased them from the graph (but they are
insurgency ('rv', 'vr', 'v'); a mixture of verbal
opposition ('b'), and a few demonstrations listed
or in the Appendix).

This content downloaded from 200.27.72.243 on Wed, 24 Jul 2019 22:02:24 UTC
All use subject to https://about.jstor.org/terms
This content downloaded from 200.27.72.243 on Wed, 24 Jul 2019 22:02:24 UTC
All use subject to https://about.jstor.org/terms
572 journal of PEACE RESEARCH volume 37 / number 5 / september 2000

Figure 3. Axes 3 and 4 of the Correspondence Analysis of Segments and Covariates

rwr AM
0.3 -
M
c)
rVw
a)
rur
I cbm democracy 2
o cm
'4-4
ADA DA AE
0
o
low birth r.
cb
slight diff.
c; C rw w ln income
ar
t
Jiihd.li ~~~int diff.

II
W slight diff. AD
in official positions . dif oisteeh of protest7lOs
8U

o
un.
se n o us S ucl~~dff h snt issue lost ecades ago vz

M medium rs Trb. comm a sliigr s.ight ff democracy 7 r Al CA

aie
int. dff mlgratilon con ers t n d it {
toiht
of protest senouremgrat on pomic state
avuy ^ ftr 29 ma MproDerty
mb CA
po weOr 7 ree .1 1 m ICnf a enTl

90s a^dvJaedjnority erecession chanrern.nolF' i psEses?ion p0e psiions

8

nonterritonal
collectivty
Xsueans medianformal (4) t. difE
recomition ,01 pro1 o8s
3 0.0
o,o
mobil. 7 mobil 4 5 lad, job, resources serious gas
state power 3
seiolus high
sono1is90-91 bm county p
all migrations g
eteacfJn decrease in st
a yp'oti rest icteco. opp. some diff, in social
state power 4 democracy 4 db
eco-classes diffuse
lesserasol.
en fancelosaic neglectremedial policy
state power 8 san.
.i medium int diff(2)aosiie
Pgntic denied recognition
on sispnarticip PA
in fecion- making
of group protest 90-91

m bil for.
democracy 8 catholic re Bel lon 3

I1
serious share of budhist
U
public funds 90-9]
aca
r aod
rt'
bA
state power 9

-' 'U V

-03 -

Axis 3 04
0.0
-0.3
Deterioriation of
the group's coherence Crystallizing protest

The segments represent temporal sequences of conflicts from 1945-49 and 1990-94, 8.1% and 7.5% of
total variance, respectively. Int diff. means international diffusion of group protest. Democracy and state
power are scored from 0-9, mobilization (all types), for protest and for rebellion from 0-7. Scores appear
on the graph in parentheses. Nonviolent protest: a, 'none reported'; b, 'verbal opposition'; c, 'political
organizing activity on a substantial scale'; d, 'a few demonstrations or strikes'; e, 'a number of demon-
strations or strikes'; f, 'major demonstrations'; m, 'missing value'. Violent protest: A, 'none reported'; B,
'scattered acts of sabotage'; C, 'limited rioting'; D, 'substantial rioting'; E, 'serious and widespread riotin
F, 'local rebellions - armed attempts to seize power in a locale'; M, 'missing value'. Rebellion: r, 'none
reported'; s, 'political banditry'; t, 'campaigns of terrorism'; u, 'small-scale insurgency'; v, 'large-scale insu
gency'; w, 'protracted civil war'; z, 'missing value'.

This content downloaded from 200.27.72.243 on Wed, 24 Jul 2019 22:02:24 UTC
All use subject to https://about.jstor.org/terms
 Noel Bonneuil & Nadia Auriat ETHNIC CONFLICT AND COHESION, 1945-94 573

Figures 1 and 3 are to be superimposed

continuous struggle ofPolisario Front against
onto Figures 2 and 4, respectively. Weoccupying
begin Moroccans since 1975 and against
by superimposing Figure 1 on Figure Mauritanian
2 while forces until 1978 (Litvinoff,
commenting upon some peripheral 1997:
groups437). Similarly, Kachins from Burma
led an armed resistance from the 1960s until
which may help validate our visual represen-
the ceasefire of 1994 (Litvinoff, 1997: 533).
tation. By superimposing these two figures,
we can see that Lebanon - Sunni, Maronites,
Several groups from Iran - Bakhtiaris,
Palestinians, Druzes, Shi'is - which experi-
Kurds, Turkmen, Arabs, Baluchis - are gath-
ered on the lower side of Axis 2, toward
enced war in the period 1975-90, is located
at the extreme right-hand side of missing
Axis 1,values. On the lower side of Axis 2,
which is the axis of war and mobilization ('r',rising insurgency, mobilization and
toward
weak democracy, Isaaqs from Somalia, close
'u, w, 'uw'), discrimination and migration.
We can visualize this pattern as well fortoPales-
and equidistant between 'ruw' (no rebel-
tinians from Jordan and from Israel. The
lion reported then small insurgency then war)
and are
analysis also shows that Kurds from Iraq 'MA' ('missing value', followed by 'no
violent
similar in this respect, and we know that theyprotest'), tried to liberate their clan
have fought for their autonomyterritory
since against the state in the 1980s at the
Kurdish nationalism in the 1930s, and have of tens of thousands of civilians killed
expense
experienced successive defeats against (Litvinoff, 1997: 454). Although Isaaqs did
Saddam Hussein and his use of chemical not experience the sequence 'AFA', they are
warfare. The same can be said for Shi'is of located close to it: this comes from the statis-
Iraq and their continued resistance and tical association of 'AFA' ('none reported' fol-
defeat against Saddam Hussein. Continu- lowed by 'local rebellions, armed attempts to
ously superimposing Figures 1 and 2, we can seize power in a locale' followed by 'none
draw similar observations on the Kurds fromreported') with the other sequences contain-
Turkey, who owe their location at the upper ing insurgency, through the other minority
right-hand corner, close to the sequence of groups. Crushed tribal uprisings and constant
sustained terrorism, to the guerrilla move- rural agitation in the 1950s and continuing
ment led by the PKK since 1984. At the resentment associated with constant migra-
right-hand side of Axis 1, Tibetans from tion to cities led Berbers from Morocco to be
China have been suffering from armed located at the lower right, close to 'AFA' and
oppression, from restrictions to higher edu-'ru', which these Berbers experienced.
cation and from economic discrimination to Along the upper side of Axis 2, toward sus-
the advantage of Han Chinese (Litvinoff, tained verbal opposition (b) and demon-
1997: 605). The location of indigenous strations, (d) accompanied by rising political
banditry and terrorism, we find Western
peoples from Guatemala at the lower right-
democracies: USA, France, Spain, Italy and
hand corner of the first principal plane refers
to the social movements in the 1 960s Switzerland. For example, the Jurassians
demanding land and fair wages, to their formed a militant separatist movement in the
repression and the following armed insur- 1960s and 1970s which resorted to arms to
gency, to the massacres in the 1980s by protest against the Swiss-German-speaking
General Rios Montt, to the discrimination Protestant bureaucracy (Litvinoff, 1997:182).
from the legal, political, economic and social The Hispanics from the USA joined in a
systems of the country (Litvinoff, 1997: 93). protest movement in the 1960s and, more
Saharawis from Morocco, also located at recently, against urban poverty and discrimi-
the lower right-hand corner, reflect the nation (Litvinoff, 1997: 34). Close to these

This content downloaded from 200.27.72.243 on Wed, 24 Jul 2019 22:02:24 UTC
All use subject to https://about.jstor.org/terms
574 journal of PEACE RESEARCH volume 37 / number 5 / september 2000

Figure 4. Minorities in Axes 3 and 4

0.4 -
Sunni
(Leh .) Fulani/FJilbe
ma. (SfSia) lA Bakhtiaris
Baganda ' - (Iran)
__ Mene Mende (Guinea) (Uganda) Turkmens
Tamils AMMlmba
rw /fielrmL. (Iran) Kurds
(Sri L.) (Iran)
Merina (ISierr L.)
(Madag.) Palesfinian , Alawis
Luba -Kasai (Uordan) Ewe ay^
(Zaire) Togo)
R (Maurita.) Af s Bamileke Pushtuns
1. (Cameroon) (Pakistan)
Shi'ite ADA Azeref
(I raq) A wr (In n)M~fossi^ Jfj^ar Sikhs
(n)ana) r
rw d;a (SAZ'~~~d (Nigerr (India)
Ashanti ( eri d schedule (I Pa I. .(Intala)
(Ghana) AD Zan
T r5 ds a Irales i ACA Tipuras
ec WesternersMgs,
ia)s [Sut AF(sael) A. (India)
eec ^reoJCamero rna
cec T0i.traeains( prouls (apa1 rt Sardiniafs V
ru)nt^%
KabrA6n11a)1 A ask.). Co (Italy) Corsicans
(Israel) (Nlgera) nralss * s
(Togo)
Palestinians B( ko nativ Irdians Kas,figjims
0.0 (Leh.) b Lruzes lasques
Marnites Ovimbudu (Leh.)
b (Spain) Lozi
cL#bh.) (Angola) ( ) b ,cdc . (Zambia) native Am.
Ewe ,(Lfeosp
(Ghana) Inli nous ppl,Maoris Hungarian(USA) db
Indigenous ppis a)u Vew Za.) (Romania) Tibfetans
(C?h l R)] (Rwand)' Berbers Sami (China)
Morocco) (Norway) A m
Ct,nye (Colombia) Mons AlA
(BSurma)
Quebecois (tugeriaS) (g) Shans Pap qans
(Can.) aca (Burma) (Imdo.) u
wr native ppls acd
I idigenous PP^ .?s Tgis cs (Can.i
gGuat^ P^di7n anese (India) v
Creoles (Zim.)
(Sierra L.) ChIne itr s Cordillera people
e (Morocco) P ii .
Kadqzans (Indo.
(Malay.)
East Timorese
(Indo.)
Malay - iuslims
(Thai.i

Montagnards
(Viet.)
Hmong
(Lao.sf
-0.5 -

- 04 (Costa
Antillean Blacks
Bica) I Hui muslims
0.0 (China) AXiS 3.
Aborigines Haitian Blacks Bouganvilleans Tiyiwanese 0.5
South A) (Dominica) (Papua N. Guinea) noelwan Ah1tadis
fjfuA.)il) Blacks (Efuado0 Indigenous ppls c hainese (Pah1tani
H 7h a sCopts (Egypt) I,Paragueay (Thailand) Chinese
Bannladsh)en Sunnis (Iraq) Afro-Peruvlans (Peru) Blacks (Malaysia)
sPan 7tusflmen
mostwana) Irnzea
Vietnamese anal
Europeans hnsAf) Venzuela) Mainland Chinese
(South
Chin/Zomls (Kampuehea) Asians (South Af:) /chanol/ (Taiwan)
(Burma) Luo (Kenya foreign workers Northe il TribaI LundaiYeke
17rks(Bulgaria) Maasais (cenya) (Switzerland) ffif (Zaire)

A few sequences of conflict are left to help the superimposition with Figure 3. Nonviolent protest: a,
'none reported'; b, 'verbal opposition'; c, 'political organizing activity on a substantial scale'; d, 'a few
demonstrations or strikes'; e, 'a number of demonstrations or strikes'; f, 'major demonstrations'; m,
'missing value'. Violent protest: A, 'none reported'; B, 'scattered acts of sabotage'; C, 'limited rioting
'substantial rioting'; E, 'serious and widespread rioting'; F, 'local rebellions - armed attempts to seize
power in a locale'; M, 'missing value'. Rebellion: r, 'none reported'; s, 'political banditry'; t, 'campaign
terrorism'; u, 'small-scale insurgency'; v, 'large-scale insurgency'; w, 'protracted civil war'; z, 'missing v

This content downloaded from 200.27.72.243 on Wed, 24 Jul 2019 22:02:24 UTC
All use subject to https://about.jstor.org/terms
 Noel Bonneuil e& Nadia Auriat ETHNIC CONFLICT AND COHESION, 1945-94 575

groups, we find for example the Lozi fromnotably after the 1965 coup, but 'their
assimilation into the local communities in
Zambia, resentful of the active discrimination
which they live has been government policy
practised by central authorities, which resulted
in a Lozi secessionist movement in the mid-
since the 1970s, and 'anti-Chinese policies
1990s (Litvinoff, 1997: 528); and Kashmiris
have been muted or relaxed' (Litvinoff, 1997:
615-616).
from India, resentful of the Indian govern-
ment for lack of economic development
(Litvinoff, 1997: 566); and Merina from
Roads to Conflict
Madagascar, where educational and language
The right-hand side of Axis 1, showing the
policy issues are made 'explosive' by polarities
in social class (Litvinoff, 1997: 495). path toward war or rioting, shows closeness
We also notice that the right-hand corner
between five sets of variables: high mobiliz-
ation (scoring 7 on a 0-7 scale) and mobiliz-
of Axis 3, the axis 'crystallizing' resistance
ation for rebellion (scoring 4 on a 0-7 scale)
against oppression related to discriminatory
are linked to migration stress (important
land and ownership practices, gathers several
groups from India and Pakistan: Kashmiris, migration from rural to urban and abroad),
Tripuras, Nagas, Sikhs, Pushtuns, as well repression
as (restricted collectivity, restricted
Kazakhs and Tibetans from China. On the economic opportunities, repressive policy),
left-hand side of Axis 3, toward mobilization,specific religions (Muslims, Shi'is and Bud-
are groups from Lebanon for example. Mostdhists) and slight discrimination (by income,
of the groups from the African continent are profession, residence, commerce, official pos-
located along the upper side of Axis 4, towarditions and access to civil service). This
weak democracy and slight discrimination. arrangement of these five classes of determi-
Bates (1983) showed that competition for nants along the axis of war and rioting echoes
the loop mechanism presented by Gurr &
land, for markets and for jobs stratifies con-
temporary African society, and that 'the Moore (1997: 1081). According to these
spatial diffusion of modernisation makes authors, repression and grievances activate
inevitable' that one ethnic group is more mobilization which, together with grievances,
advanced than another (Bates, 1983: 159), triggers rebellion and this, in turn, brings
and that frustration generated by compe- about repression. Our analysis is incapable of
tition creates fertile soil for conflict. On the revealing such a dynamic mechanism;
opposite side, moving toward the lower end however, our study does show the close statis-
of Axis 4, groups from East and Southeast tical associations between these factors.
Asia - Hmong from Laos, Montagnards from Notably, protest and rebellion do not occur
Vietnam, Malay-Muslims from Thailand, without mobilization. To state that there is a
East Timorese and Chinese from Indonesia, causal relationship between them is another
Dayaks, Indians and Kadazans from matter; we prefer to limit our observation by
saying that, from the observed minorities,
Malaysia, Cordillera People from the Philip-
pines, Arakenese, Shans and Mons from these factors are likely to occur jointly, and
Burma, Chinese from Vietnam - are gath- this pattern is highly salient because it reflects
ered at this lower end of Axis 4, toward high most of the variance of the dataset. Religion
state power and medium mobilization for plays an important role as observed by its
rebellion, significant linguistic assimilation location on the far right of Axis 1. This loca-
and participation in decisionmaking. For tion probably is due to the fact that religion is
example, Chinese from Indonesia in the past one important determinant of group coher-
'have suffered significant discrimination', ence, itself a factor in group mobilization

This content downloaded from 200.27.72.243 on Wed, 24 Jul 2019 22:02:24 UTC
All use subject to https://about.jstor.org/terms
576 journal of PEACE RESEARCH volume 37 / number 5 / september 2000

(Gurr & Moore, 1997: 1083; Fox, 1999a: suggest that protest turns to war when auton-
304; 1999b). omy, income, higher education or access to
Axes 3 and 4 give a more fine-grained official positions are at stake.
version of discrimination, which we describe The closeness of migration distress -
as a three-pronged star: (1) the right-hand serious migration abroad and from rural to
side of Axis 3, which crystallizes rising action urban - to slight differences in access to
of the group, captures slight differences in official positions both on the first principal
land and property, moderate dispossession plane (lower right-hand quadrant) and on
from land and from positions of power in the the plane of Axes 3 and 4 (upper left-hand
1980s and serious discrimination related to quadrant) and the proximity to slight differ-
land and resources in 1990-91; (2) the upper ences in higher education on this plane
side of Axis 4 is characterized by slight dis- echoes the joint effect of geographic and
crimination related to other economic and social mobility toward mobilization and war
political factors, namely income, access to or rioting. In the particular case of 19th-
higher education, organizing political century France, Bonneuil & Rosental (1999)
activity and access to official positions; (3) showed a strong correlation between these
cultural discrimination appears along the two forms of mobility: migrants are very
lower side of Axis 4, close to medium levels likely to attain a higher-valued occupation
of state power and of democracy and to than those who are sedentary. In the present
decreasing state power. Thus, discrimination context, we could suggest that truncated
plays a two-stage role: in the first order (Axis social mobility (through discriminatory
1), it is strongly associated with mobilization access to official positions and unequal access
and rioting and war; in the second order to higher education) is a form of injustice
(Axis 3 with variance nearly equal to that ofagainst which people who are robbed of any
Axis 2), discrimination related to land and tohope for a better future for themselves, and
positions of power, as well as lost autonomy,particularly for their children, are sensitive to
is associated with rising action; in the third the possibility of migrating abroad or mobil-
order (Axis 4), discrimination concerning izing into rioting and war. Burguiere (1991:
income, higher education and access to 493) similarly remarks on the relative abun-
dance of the comparatively deprived younger
official positions is associated with war, and
discrimination more specifically of socialgeneration in the outbreak of violence during
the French Revolution of 1789.
customs is clearly linked neither to rioting
nor to verbal opposition. These results are Gurr & Moore (1997: 1083) concluded
complementary to the view that different that 'democracy is simply a proxy for repres-
forms of discrimination have different sive behaviour' and stressed the relationship
effects, notably that 'economic and social between state repression toward a minority
grievances and demands for greater politicalgroup and the extent of ethno-rebellion.
rights . . . were weakly but consistently corre-Violent ethno-political conflict was thus
lated with magnitudes and communal unlikely to develop in democracies, contrary
protest' while 'resentments about restricted to autocratic states or to democratic states
access to political positions and a collective that have not yet developed a 'dense network
history of lost autonomy drive separatist of institutions to channel and respond to
demands and rebellion generally' (Gurr, protest'. The upper side of Axis 2 confirms
1993b: 189). Our results distinguish that Western democracies are associated with
between different types of economic nonviolent protest mixed with outbursts of
dis-
crimination (i.e. related to land or not)political
and banditry and terrorism, excluding

This content downloaded from 200.27.72.243 on Wed, 24 Jul 2019 22:02:24 UTC
All use subject to https://about.jstor.org/terms
 NoIel Bonneuil e& Nadia Auriat ETHNIC CONFLICT AND COHESION, 1945-94 577

more serious conflicts. Interestingly, in

rebellion', the causal mechanism, if any,
Figure 3, Axis 3 contrasts a pattern of nobetween
con- international diffusion and conflict
flict ('aca'), high democracy (score of 8ison
nota linear.
0-9 scale) and high state power (score of 8)
in the lower left-hand quadrant, to limited
Discussion
rioting and insurgency, with decrease in state
power, medium state power and democracyThrough this correspondence analysis, we
(both score 4 on a 0-9 scale), medium mobil-
have attempted to decipher the statistical
ization for rebellion, denied recognitionassociations
and between 50 years of successions
significant participation in decisionmaking.
of conflicts and their accompanying political
and
The coexistence of these two patterns on thesocio-economic factors. We notably
lower side of Axis 4, the axis of state power
highlighted four structuring axes in terms of
information (variance) provided by the
and democracy, and their confrontation
dataset. The first principal axis, reflecting
along Axis 3, the axis of awakening resistance,
can be interpreted in the argument of most
Gurr of the variance, helped characterize the
(1993a) or Gurr & Moore (1997), according
road to rioting and war, by providing a hier-
archy
to which, said simply, democracy inhibits too of the importance, first, of mobiliz-
much violent action in permitting protest,
ation, associated with (slight) discrimination,
whereas democratization accompanied by
migration stress and religion and, to a lesser
decreasing state power opens the door to with repression. The pattern delin-
extent,
insurgency and rebellion. eated along Axis 1 corroborates the multifac-
The relatively important contributions torial
to nature of the roots of violent conflict,
the principal axes of sequences like 'political
without however imposing a causal mechan-
ism such as was found in Gurr & Moore
banditry' followed by 'small-scale insurgency'
('su'), 'none reported' followed by 'small-scale
(1997: 1081). It intermingles primordialist
insurgency' followed by 'large-scale insur- notions, such as religion, and instrumentalist
gency' ('ruv'), or 'none reported' followed by ones, such as grievances, mobilization,
'small (or large)-scale insurgency' followed by migration stress and repression.
'war' ('ruw' and 'rvw') suggest that rebellion We also note that discrimination was split
is a process that builds up over time. The into power and land-related issues along Axis
same can be noticed for 'political banditry - 3 with limited rioting and demonstrations; it
campaigns of terrorism' ('rstr'). was felt through unequal social mobility, up
Axis 1 opposes medium to high and weak along Axis 4, with substantial rioting and
international diffusion of group protest, Axis 3war; and discrimination through social
medium to high, Axis 4 the highest to mediumcustom, located down Axis 4, with verbal
low. Thus, international diffusion is generallyopposition and local rebellions. We can thus
present in all kinds of conflict structuring theagree with Fisher (1993: 120), according to
dispersion of the dataset. Gurr (1993a: 91) alsowhom 'distributive injustice or structural
noted the transnational dimensions to most inequality combined with historically signifi-
kinds of intrastate conflicts. However, the roadcant ethnic cleavages provides a particular
from insurgency to war (below the right-handfocus for the frustration of basic needs and
side of Axis 1) is associated only with weak dif-the development of protracted social con-
fusion in the 1990s, so that according to Gurrflict'. Our (sample-dependent) results, more-
and Moore (1997:1083), if'the occurrence of over, distinguish a finer structure combining
similar rebellions elsewhere in a region is different types of discrimination together
hypothesized to have a positive effect on with the outbreak of forms of conflict.

This content downloaded from 200.27.72.243 on Wed, 24 Jul 2019 22:02:24 UTC
All use subject to https://about.jstor.org/terms
578 journal of PEACE RESEARCH volume 37 / number 5 / september 2000

Moreover, the relationship between dis- References

crimination and conflict is not linear, as 'sub-
Abbott, Andrew, 1998. 'The Causal Devolution',
stantial differences' will contribute to none of
Sociological Methods and Research 27(2):
the first four axes. This lack of mechanism 148-181.
between substantial discrimination and con- Aminzade, Ronald, 1993. Ballots and Barricades:
flict has already been noted by Hobsbawm Class Formation and Republican Politics in
(1990: 167) in the case of Poland. France, 1830-1871. Princeton, NJ: Princeton
The association between geographic and University Press.
social mobility together with rioting and war Anderson, Benedict, 1991. Imagined Communi-
ties, 2nd edn. London: Verso.
emerges from the upper right-hand quadrant
of the plane made of Axes 3 and 4. Although Barkey, Karen & Sunita Parikh, 1991. 'Compara-
tive Perspectives on the State', Annual Review
this is not a first-order pattern, we strongly
of Sociology 17: 523-549.
believe it deserves attention, for this kind of
Bates, Robert H., 1983. 'Modernization, Ethnic
discrimination involves the future of indi-
Competition, and the Rationality of Politics in
viduals and that of their children. Social
Contemporary Africa', in Donald Rothchild
mobility has been a sensitive matter for & Victor A. Olorunsola, ed., State Versus
developed countries; it appears as a legitimate Ethnic Claims: African Policy Dilemmas.
claim for emerging countries as well as a Boulder, CO: Westview (152-172).
strong incentive for migration or mobiliz-Benzecri, Jean-Paul, 1992. Correspondence Analy-
ation and conflict. We therefore recommend sis Handbook. New York: Marcel Dekker.
that conflict prevention policies shouldBonneuil, Noel & Paul-Andre Rosental, 1999.
'Changing Social Mobility in 19th Century
devote more attention to the promotion of
France', Historical Methods 32(2): 53-73.
equitable social mobility.
Bry, Xavier, 1996. Analyses factorielles multiples
Finally, we confirm the presence of inter-
[Multiple Factor Analyses]. Paris: Economica.
national diffusion of protest in a variety of
Burguiere, Andre, 1991. 'La destabilisation de la
different types of conflict. Several variables,
societe francaise' [The Destabilization of
such as settlement patterns (Toft, 1998), for French Society] in Histoire de la population
example, did not appear to contribute to any francaise [History of the French Population],
of the four axes and thus they convey very vol. 2. Paris: Quadrige, Presse Universitaire de
little of the variance (we recall that all vari- France (475-493).
ables not contributing enough to the princi- Cibois, Philippe, 1997. 'Les pieges de l'analyse
pal axes were erased from the graphs). des correspondances' [Traps in Correspon-

Our use of correspondence analysis has dence Analysis], Histoire et Mesure, special
issue, 'Penser et mesurer la structure' [Think-
enabled us to situate each minority group on
ing and Measuring Structure] 12(3/4):
the panorama of 50 years of sequences of
299-320.
conflicts from 1945 to 1994. This results less
Elster, Jon, 1999. Alchemies of the Mind. Cam-
in the building of a general theory, where uni-
bridge: Cambridge University Press.
versals would emerge (Tilly, 1 997b) and fromFearon, James D. & David D. Laitin, 1996.
which predictions could be made, than in 'Explaining Interethnic Cooperation', Ameri-
contributing toward a statistical description can Political Science Review 90(4): 715-735.
of the diversity of connections between Fisher, Ronald J., 1993. 'Toward a Social-psycho-
sequences of conflicts and their historical logical Model of Intergroup Conflict', in Knud
determinants. S. Larsen, ed., Conflict and Social Psychology.
London: Sage, for PRIO (109-112).
Fox, Jonathan, 1999a. 'The Influence of Religious
Legitimacy on Grievance Formation by

This content downloaded from 200.27.72.243 on Wed, 24 Jul 2019 22:02:24 UTC
All use subject to https://about.jstor.org/terms
 Noel Bonneuil e6 Nadia Auriat ETHNIC CONFLICT AND COHESION, 1945-94 579

edn. Cambridge & New York: Cambridge Uni-

Ethnoreligious Minorities', Journal of Peace
Research 36(3): 289-307. versity Press.
Jaggers, Keith & Ted Robert Gurr, 1995. 'Track-
Fox, Jonathan, 1999b. 'Do Religious Institutions
ing Democracy's Third Wave with the Polity
Support Violence or the Status Quo?', Studies
in Conflict and Terrorism 22(2): 119-139. III Data', Journal of Peace Research 32(4):
Franzosi, Roberto, 1997. 'Mobilization and 469-482.
Counter-Mobilization Processes: From the Lebart, Ludovic; Andre Salem & Lisette Berry,
"Red Years" (1919-1920) to the "Black Years" 1998. Exploring Textual Data. Dordrecht:
(1921-1922) in Italy', Theory and Society Kluwer Academic.
26(2-3): 275-304. Litvinoff, Miles, et al., 1997. World Directory of
Franzosi, Roberto, 1998. 'Narrative Analysis - or Minorities. London: Minority Rights Group
Why (and How) Sociologists Should be Inter- International.
Mahoney, James, 1999. 'Nominal, Ordinal, and
ested in Narrative', Annual Review of Sociology
24: 517-554. Narrative Appraisal in Macrocausal Analysis',
Franzosi, Roberto & John W. Mohr, 1997. 'New American Journal of Sociology 104(4):
Directions in Formalization and Historical 1154-1196.
Analysis', Theory and Society 26(2-3): des Netumieres, Felicite, 1997. 'Methodes de
133-160. regression et analyse factorielle' [Regression
Geertz, Clifford, 1993. The Interpretation of Cul-Methods and Factor Analysis], Histoire et
tures, 2nd edn. London: Fontana. Mesure, special issue 'Penser et mesurer la
Greenacre, Michael, 1984. The Theory and Appli-structure' [Thinking and Measuring Struc-
cation of Correspondence Analysis. London: ture] 12(3/4): 271-298.
Academic Press. Schaeffer, Robert, 1990. Warpaths, Politics of
Gurr, Ted Robert, 1993a. Minorities at Risk: A Partition. New York: Hill and Wang.
Global View of Ethnopolitical Conflicts.Skocpol, Theda, 1979. States and Social Revolu-
Washington, DC: United States Institute of tions. Cambridge: Cambridge University
Peace Press. Press.
Gurr, Ted Robert, 1993b. 'Why Minorities Rebel: Stavenhagen, Rodolfo, 1990. The Ethnic Ques-
a Global Analysis of Communal Mobilization tion: Conflicts, Development, and Human
and Conflict since 1945', International Politi- Rights. Tokyo: United Nations University
cal Science Review 14(2): 161-201. Press.
Gurr, Ted Robert, 1994. 'Peoples against States:Thompson, Edward Palmer, 1991. Customs in
Ethnopolitical Conflict and the Changing Common. New York: New Press, Norton.
World System', International Studies Quaterly Tilly, Charles, 1978. From Mobilization to Revol-
38: 347-377. ution. Reading, MA: Addison-Wesley.
Gurr, Ted Robert, 1997. 'Ethnopolitical Rebel-Tilly, Charles, 1995. 'To Explain Political Pro-
lion: A Cross-Sectional Analysis of the 1980s cesses', American Journal of Sociology 100(6):
with Risk Assessments for the 1990s', Ameri- 1594-1610.
can Journal of Political Science 41(4): Tilly, Charles, 1997a. 'James S. Coleman as a
1079-1103. Guide to Social Research', American Sociologist
Gurr, Ted Robert & Will H. Moore, 1997. 'Eth-2: 82-87.
nopolitical Rebellion: A Cross-sectional Tilly, Charles, 1997b. 'Means and Ends of Com-
Analysis of the 1980s with RiskAssessment for parison in Macrosociology', Comparative
the 1990s', American Journal of Political Science Social Research 16: 43-53.
41(4): 1079-1103. Toft, Monica D., 1998. 'Do Settlements Patterns
Hardin, Russell, 1995. One for All: The Logic ofMatter?'. Working paper, Department of
Group Conflict. Princeton, NJ: Princeton Uni- Political Science, University of Chicago.
versity Press. Vayrynen, Raimo, 1994. 'Towards a Theory of
Hobsbawm, EricJ., 1990. 'Nations and NationalismEthnic Conflicts and their Resolution'. Occa-
Since 1789: Programme, Myth, Reality', 5th sional Paper (1). Notre Dame, IN: Joan B.

This content downloaded from 200.27.72.243 on Wed, 24 Jul 2019 22:02:24 UTC
All use subject to https://about.jstor.org/terms
580 journal of PEACE RESEARCH volume 37 / number 5 / september 2000

Kroc Institute for International Peace Studies,

Different group language (LANG)
University ofNotre Dame. Group migration among regions
Volle, Michel, 1985. Analyse des donnees [Data
(MIGRANT)
Analysis]. Paris: Economica. Mobilization score (MOB80)
Score of mobilization for protest
(MOBPR080)
Appendix: Variables Used in the Score of mobilization for rebellion
Correspondence Analysis (MOBREB80)
Names in the Minorities at Risk datasetDemocracy
are score (NDEM86)
given in parentheses. National economic growth rate
(NECOGRO)
Values for 1980 Number of points change in democracy
Seek greater regional autonomy? (NDEM7586)
(AUTON4) Number of years with significant changes
Autonomy lost (AUTONEND) in national political Institutions
Population size (BESTPOP) (NICH7586)
Belief (BELIEF) Population growth rate (NPOPGRO)
Eco-classes? (CATEC) Change in state power (NSC06086)
Communal contenders (CATCC) State power (state control of economic and
Religious sects (CATMS) social life) (NSCOPE86)
Indigenous peoples? (CATIN) Number of adjoining countries with seg-
Group cohesion (COHEREX) ments of the group (NUMSEGX)
National people seeking autonomy? Seek greater political rights (POL2)
(CATNA) Seek greater central participation (POL3)
Differential social custom (CULDIFX5) Seek equal civil rights (POL4)
Different urban/rural distribution Seek change in officials, policies (POL5)
(CULDIFX6) Access to power positions (POLDIFX1)
Emigration abroad (DEMEMIG) Access to civil service (POLDIFX2)
Birth rate (DEMBIRTH) Recruitment to police (POLDIF3)
Rural-to-urban migration (DEMURB) Right to organized political activity
Dispossession from land (DEMEVICT) (POLDIFX5)
Inequalities in income (ECDIFX1) Differences in legal protection
Inequalities in land/property (ECDIFX2) (POLDIFX6)
Inequalities in higher/technical education Political discrimination (POLDISX)
(ECDIFX3) Political status (POLSTAT)
Presence in commercial activities Region (REGION)
(ECDIFX4) Group concentration (REG1)
Differential presence in profession Religion (RELIGS1)
(ECDIFX5)
Differential presence in official position Values for the 1990s
(ECDIFX6) Discrimination present? (ATRISK1)
Economic discrimination (ECDISX) Disadvantaged from past discrimination?
International diffusion of group protest in (ATRISK2)
the 1970s (ISEGPRO7) Advantaged minority being challenged
International diffusion of group protest in (ATRISK3)
the 1980s (ISEGPRO8) Group mobilized? (ATRISK4)

This content downloaded from 200.27.72.243 on Wed, 24 Jul 2019 22:02:24 UTC
All use subject to https://about.jstor.org/terms
 Noel Bonneuil e6 Nadia Auriat ETHNIC CONFLICT AND COHESION, 1945-94 581

Political discrimination (POLDIS90)

Group cohesion in 1990-91 (COHESX90)
Country population in 1990 (CPOP90)Diffuse political grievances, 1990-91
Group population in 1990 (POP90) (POLRIG11)
Birth rate, 1990-91 (DMBIRT90) Communal rights, 1990-91 (POLRIG21)
Emigration for political reasons, 1990-91 Participation in decision-making 1990-91
(DMEMPO90) (POLRIG31)
Disposession from land, 1990-91 Equal civil rights, 1990-91 (POLRIG41)
(DMEVIC90) Change in policy, 1990-91 (POLRIG51)
Deteriorating health conditions, 1990-91Other political grievances, 1990-91
(DMSICK90) (POLRIG61)
Economically advantaged, 1990-91
(ECADV90)
Economically discriminated, 1990-91 NOEL BONNEUIL, b. 1959, PhD in
(ECDIS90) Applied Mathematics (ficole des hautes etudes
en sciences sociales, Paris, 1991); Directeur de
Diffuse economic concerns, 1990-91
Recherche, Institut national d'etudes demo-
(ECRIG11)
graphiques, Paris (1995-); Directeur d'etudes,
Share of public funds, 1990-91 (ECRIG21) fcole des hautes etudes en sciences sociales,
Economic opportunities, 1990-91 Paris (1997- ). Most recent book in English:
(ECRIG31) Transformation of the French Demographic
Working conditions, 1990-91 (ECRIG41) Landscape, 1806-1906 (Oxford University
Land, jobs, resources, 1990-91 (ECRIG51) Press, 1997).
Groups spatial distribution (GROUPCON)
Intercommunal violence since 1990
NADIA AURIAT, b. 1965, PhD in Socio-
(INTERCON)
logical Methods (Universite Paris 5 Sorbonne,
Intracommunal violence since 1990
Paris, 1997); Programme Specialist, Division
(INTRACON)
of Social Science and Policy, MOST Pro-
International diffusion of group protest in
gramme (1992- ). Most recent book: Les
1990s (ISEGPRO9) defaillances de la memoire humaine [Cognitive
Political support from foreign governments, Aspects of Retrospective Surveys] (INED
1990-91 (ISPOL90) Presse Universitaire de France, 1997).

This content downloaded from 200.27.72.243 on Wed, 24 Jul 2019 22:02:24 UTC
All use subject to https://about.jstor.org/terms