Article

Journal of Conflict Resolution

2020, Vol. 64(2-3) 402-431
ª The Author(s) 2019
Polygynous Neighbors,
Excess Men, and Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/0022002719859636
Intergroup Conflict journals.sagepub.com/home/jcr

in Rural Africa
Carlo Koos1,2,3 , and Clara Neupert-Wentz2

Abstract
We argue that polygyny creates a social imbalance where few, economically well-off
men marry many wives and many poor men marry late or never. By definition,
polygyny produces what we refer to as “excess men.” In order to gain material
wealth, excess men are likely to raid, plunder, and rob neighboring ethnic groups.
We test this hypothesis with georeferenced data on polygyny and intergroup conflict
in rural Africa and find strong support. Drawing on Afrobarometer survey data, we
explore the underlying mechanisms and find that young men who belong to poly-
gynous groups feel that they are treated more unequally and are readier to use
violence in comparison to those belonging to monogamous groups. Our article
makes an important contribution to the peace, conflict, and development literature
by emphasizing a fundamental aspect of human life: marriage and family.

Keywords
intergroup conflict, polygyny, Africa, GIS, family, ethnicity

Social institutions have long been a focal point in the analysis and explanation of
intrastate peace and conflict. While state institutions have been studied to a large
extent in this context, the internal norms, traditional institutions, and customary laws

1
Chr. Michelsen Institute, Bergen, Norway
2
Department of Politics and Public Administration, University of Konstanz, Konstanz, Germany
3
School of Public Policy, University College London, London, UK

Corresponding Author:
Carlo Koos, Chr. Michelsen Institute, Jekteviksbakken 31, 5006 Bergen, Norway.
Email: [email protected]
Koos and Neupert-Wentz 403

of ethnic groups have not yet received much attention. Key institutions that influence
the social order of ethnic groups are marriage and family. In most societies, the
family is the smallest social entity that shapes the everyday life of people (Weber
[1922] 1980, xvii). The family typically fulfills reproductive, social, economic, and
prestige functions (Becker 1993; Murdock 1949; Hudson and Matfess 2017).
We argue in this article that the type of marriage institution practiced by an ethnic
group, monogamy or polygyny, affects the likelihood of members of that group
attacking neighboring groups. By definition, polygyny creates a social imbalance:
while some men marry several wives, rear many children, and have large families,
other men marry late in life or not at all. A common pattern is that marriage is
confined to economically well-off men in the highest tiers of society, leaving eco-
nomically deprived men unwed (Irons 1983; Mesquida and Wiener 1999; McDer-
mott 2018). We refer to the latter as excess men. In traditional rural societies where
social norms make a man’s reputation dependent on, among other things, the size of
his family, excess men fail to meet basic criteria for attaining social prestige (Hen-
rich, Boyd, and Richerson 2012, 657; Hudson and Matfess 2017, 12).
However, in our understanding, excess men will not accept the fate of remaining
bachelors. According to Hans Morgenthau, propagation is one of the main drivers of
any political action ([1948] 1985, 39). Since economic resources are key to getting
married and starting a family, excess men have incentives to acquire these resources.
When legitimate sources of income are unavailable or insufficient, excess men
become “risk-takers” (Barash 2016, 30): crime, theft, violence, and raids become
viable options. Excess men in rural areas who strive to conform to the social norms
that derive from marriage and family therefore have two basic choices: to steal from,
plunder, and raid one’s own group or to do the same to another group.
Since ethnic groups often function as extended families and have established
mechanisms to monitor and sanction misbehavior (cf. Fearon and Laitin 1996),
excess men will be more likely to raid other groups than their own. Following this
reasoning, we expect that polygyny does not necessarily increase intragroup vio-
lence but rather heightens the risk of violence for neighboring ethnic groups.
In our analysis, we examine whether the extent of borders shared with polygy-
nous ethnic groups increases a group’s risk of experiencing intergroup violence.
Specifically, we create a risk profile for each ethnic group that measures the per-
centage of total border shared with polygynous neighbors. Building on the growing
literature that analyses the long-term effect of historical institutions and politics
(e.g., De Juan and Koos 2019; Wig 2016; Michalopoulos and Papaioannou 2016;
Nunn and Wantchekon 2011; Nunn 2008), we rely on precolonial data on ethnic
groups’ mode of marriage—which has been shown to correlate with current poly-
gyny rates (Dalton and Leung 2014)—to predict contemporary violent conflict
events between ethnic groups in rural Africa. Using a set of pretreatment exogenous
geographical and historical variables that could have affected both the prevalence of
polygyny and intergroup conflict (e.g., ancient wars, slave trade, and malaria pre-
valence), we show robust evidence that for groups with higher percentages of shared
404 Journal of Conflict Resolution 64(2-3)

boundaries with polygynous groups, the number of violent events increases substan-
tively, a finding which supports our hypothesis.
In a second step, we employ a pooled sample of Afrobarometer survey data to
better understand the underlying mechanisms of this relationship. We are able to
demonstrate that childless young men who belong to polygynous ethnic groups feel
that they are treated more unequally and regard violence more frequently as a
justified means to achieve their goals in comparison to their peers in monogamous
groups. This lends support for our proposed mechanism, which suggests that excess
men are the linkage between polygyny and intergroup violence.
In addition to our contribution to the literature on the long-term effects of his-
torical institutions, we provide a substantially refined theoretical argument and
improved empirical test to the literature on family institutions and violent conflict,
an aspect that has not received adequate attention. Additionally, we complement the
literature on local-level and communal conflicts (e.g., Eck 2014; Fjelde and von
Uexkull 2012; Varshney 2003; Tajima 2013; Fearon and Laitin 2000).

Polygyny and Conflict: Concepts and Gaps

We argue in this article that polygyny is a group-specific institution, which should
affect violent conflict between ethnic groups. To justify our approach, we discuss the
literature on polygyny and conflict, identify critical shortcomings, and in the process
describe how we address them in this article.
The only two existing quantitative studies on polygyny and armed conflict have
focused on state-based violence, which has led to contradictory findings. While
Kanazawa (2009) finds that higher rates of polygyny measured at the state level
increase the risk of large-scale civil wars, Gleditsch et al. (2011) cannot replicate this
finding either at the state or the group level. We see two major reasons for the
nonreplicability of Kanazawa’s (2009) finding, one relates to the theory and
mechanisms and the other to the data sources used to operationalize polygyny.1
The first concern relates to the theoretical foundation behind the type of violence
chosen as well as the level of analysis. Both articles analyze the effect of polygyny
on conflicts in which one party represents the state. However, it remains unclear how
polygyny features in these kinds of conflicts—that is, why the state would be the
appropriate target of organized violence by excess men. In fact, the only evidence we
find suggests that polygyny—or a skewed sex ratio toward young single males—
increases nonpolitical violence such as homicides (Wilson and Daly 1985), violent
crime and property crime (Edlund et al. 2007), societal violence (Hudson and Den
Boer 2002), or violence against women (McDermott and Cowden 2018). Therefore,
we do not expect polygyny to affect large-scale mobilization against the state.
In this vein, we also suggest an analysis that takes this variation at the meso-level
into account instead of aggregating to the state level. Our understanding of polygyny
is that of a group-specific institution, with local effects on conflict. Marriage,
throughout the world, is frequently regulated not only by state law but also by
Koos and Neupert-Wentz 405

customary law and local social norms and practices. Especially in Africa, the mode
of marriage is essentially group-specific since many ethnic groups have their own set
of customary law and traditional institutions (Holzinger et al. 2018; Holzinger, Kern,
and Kromrey 2016). The ethnic group therefore lends itself as a natural unit of
observation with which to study the effect of polygyny on conflict in Africa.
This notion is underpinned by anthropological, archaeological, and psychological
research suggesting that polygyny may rather be associated with intergroup violence
than with civil wars. Henrich, Boyd, and Richerson (2012) argue that due to the
negative effects of polygyny, such as crime and violence, polygyny creates a com-
parative disadvantage in intergroup competition leading to intergroup clashes. In his
study of the Yanomamö tribes in southern Venezuela and northern Brazil, Chagnon
(1988) finds that more successful warriors are married to more wives than their less
successful counterparts, providing a link between polygyny and violent raids of
other communities. Raffield, Price, and Collard (2017) use archeological evidence
to show how intergroup Viking raids in Scandinavia during the Late Iron Age were
related to an increase in unmarried men, which was caused by polygyny, concubi-
nage, and increasing social inequality. Using three case studies, Hudson and Matfess
(2017) identify polygyny, alongside inflationary bride-prices, as a marriage market
barrier “predisposing young men to become involved in organized group violence”
(2017, 8). While these findings come from diverse disciplines, they support our
claim that the subnational variance in marital institutions between ethnic groups
should rather play out locally and affect conflicts between groups rather than
between groups and the state.
The second issue with the existing research concerns the data and the potential of
reverse causality. Although most of the—qualitative and quantitative—studies
hypothesize that polygyny is a cause of violence, it may be the case that the rela-
tionship runs the other way. This is due to explanations as to why polygyny emerged
in the first place. One widely accepted explanation is demographic: polygyny is
likely to be the result of actual skewed sex ratios tipped toward women, due to
external influences that change the gender ratio (White and Burton 1988).2 Changes
in this ratio can have many causes such as labor migration, slave trade (Dalton and
Leung 2014), and male fatalities during wars, leading to excess women (cf. Gle-
ditsch et al. 2011, 267; Goldstein 2001, 226; White and Burton 1988). It follows that
polygyny could also be caused by deadly conflict. This would lead polygyny to at
least partially fluctuate over generations in response to demographic changes and
war. However, with contemporary measures of polygyny as coded by Kanazawa and
Still (1999) and replicated by Gleditsch et al. (2011), this potential reverse causal
direction cannot be accounted for (see Online Appendix Section 1 for data validity
and persistence).
To address this problem, we argue that historical evidence on marriage institu-
tions helps to circumvent concerns about reverse causality. Albeit possible time
variance and societal changes, there is reason to believe that polygyny as a family
institution stays intact once sex ratios are even again, that is, a generation after
406 Journal of Conflict Resolution 64(2-3)

warfare has ceased. First, theories on institutional path dependence and the so-called
stickiness of institutions (cf. Fukuyama 2011, 450) claim that institutions are change
resistant. Pierson (2000, 252) formalizes this claim by arguing that “relative benefits
of the current activity compared with other possible options increase over time,”
which increases the probability of institutional persistence. Institutions, such as
polygyny, can then be described as endogenous in themselves. Second, socialization
processes through cultural transmission can amplify path dependence when it comes
to family institutions (Bisin and Verdier 2000). In their theoretical model, Bisin and
Verdier (2000) show that the intergenerational transmission of traits—and thereby
the adaption of patterns by children—is especially resilient for ethnic and religious
minorities. Anthropological research underpins this theory, showing that polygyny
as a marital institution is “self-sustaining” (Dalton and Leung 2014, 607). Further-
more, McDermott and Cowden (2018) argue that the persistence of polygyny is
sustained by those who practice it, as these are beneficiaries of the system, for
instance through male kin networks (see also Hudson 2018).
There is also empirical evidence that polygyny persists. Dalton and Leung (2014,
613) test the persistence of polygyny as coded in George Murdock’s (1969) Ethno-
graphic Atlas (EA). The EA systematically codes the sociocultural traits of more
than 386 ethnic groups in Africa at the time before their first encounter with Eur-
opeans, including the dominant mode of marriage. Dalton and Leung use a pooled
sample of 238,075 respondents from recent Demographic and Health Survey (DHS)
data in Africa and find that contemporary polygyny rates are five times as high in
ethnic groups that were coded as polygynous (10 percent practiced polygyny) in the
EA relative to those coded monogamous (2 percent practiced polygyny).3 This lends
strong support to the idea that polygyny is indeed persistent over time and can be
conceptualized as a “sticky” institution.
Furthermore, Fenske (2015, 72) analyzes the effect of colonial education in
comparison to current expansion of education in Africa and finds that “ethnic insti-
tutions are shaped by history.” We also compare specific groups. For instance, the
Bashi in South Kivu, DR Congo, are categorized by Murdock as a polygynous
group. According to a representative household survey conducted in March 2017
by one of the authors, 13 percent of Bashi men reported having more than one wife.
If these 13 percent of polygynous men marry only two wives, they marry 26 percent
of the potential brides and thereby leave 13 percent of the excess men without
prospects of marrying. Again, this evidence supports the idea that polygyny exhibits
a substantial degree of continuity over long periods of time.
The long-term effect of social institutions has furthermore been documented in
numerous studies (e.g., Michalopoulos and Papaioannou 2013, 2016; Wig 2016;
Nunn 2008; Alesina, Giuliano, and Nunn 2013). Given the potential endogeneity
of polygyny and conflict, we believe that employing historical data such as the EA is
more appropriate to the study of the effect of polygyny on conflict. By that it is
possible to circumvent the problem of reverse causality and rely on those family
institutions that are deeply entrenched.
Koos and Neupert-Wentz 407

The literature reviewed here highlights the conceptual and empirical gaps in the
research on polygyny and violent conflict. Most of the findings of these diverse
studies show that polygyny most likely results in violence between individuals and
groups, not with the state. We therefore present a new theoretical account, which
points out how the group-internal norms and geographical dimensions of polygyny
affect intergroup conflict, rather than state-based conflict. We suggest departing
from civil war definitions of conflict and turning to more local forms of violence.
That said, we expect polygyny to have a geographical effect: excess men—the
product of polygyny—should be more likely to attack neighboring groups, not
government forces. We thus believe that our article is more sensitive—theoretically
and empirically—to the dynamics on the ground.

Theorizing the Link between Polygyny and

Intergroup Conflict
We argue that the practice of polygynous marriage within ethnic groups leads to an
increased likelihood of conflict between neighboring ethnic groups. We start by
discussing why marriage and family are valuable objectives for males to achieve,
particularly in developing countries. We then make the case that polygyny by
definition creates a social imbalance that produces young, frustrated, poor, uninte-
grated men who become a risk—not so much for their own groups but for neighbor-
ing ethnic groups.

The Benefits of Marriage and the Value of Family in Africa

The family is the smallest social unit shaping the everyday lives of most humans
(Weber [1922] 1980). In many societies, the family is defined by the institution of
marriage, the purpose of which is to have and bring up children (Borgia 1980;
Becker 1993). Being married—and thus secured reproduction—is therefore one of
the most prevalent social norms, which brings about the fulfillment of basic needs.
Not only is the family the natural unit for the feeling of belonging for many people,
but being married is usually thought to bring about social and economic benefits. On
the one hand, being married and having children leads to social standing and accep-
tance within a community (Mealey 1985; Wilson and Daly 1985, 61). For instance,
Levinson (1996, 5) states that “childlessness is one of the most serious misfortunes
imaginable” for many people in Africa.
On the other hand, starting a family and with it a common household is tied to
economic benefits. These can stem from official support for family unions, such as
tax relief and access to social benefits in functioning states with a social welfare
system, but they are especially pronounced in the context of rural communities
without welfare systems that apply the principle of compensation. In such contexts,
a family and children can compensate for the absence of the welfare state by pro-
viding social benefits such as retirement provisions (e.g., Shanas 1979). In sum, for
408 Journal of Conflict Resolution 64(2-3)

men—particularly in Africa—the value of having a family, especially a large family,

is immense.

Polygyny Deprives Young, Poor Men of the Benefits of Marriage

The social standing and economic prospects related to marriage apply equally to
females and males. However, in a community where polygyny is practiced, women’s
chances of marrying remain unchanged, whereas men’s chances decrease.4 By def-
inition, polygyny creates a social imbalance among males: a few men marry several
women, and some men cannot marry or marry late in life.
The ability to marry is tied to preexisting social and economic status. As Mes-
quida and Wiener note, “there should be an association between a man’s socio-
economic status and his reproductive success” (1999, 182). When polygyny is
practiced within a community, the ability to marry several wives will thus be
“confined to higher-status males” (Irons 1983, 196).5 With reference to Dalton and
Leung’s (2014) finding of an average of 10 percent polygyny rates in polygynous
groups, a simple example shows that a stark social imbalance can result from poly-
gyny: a society in which 10 percent of the highest-status men live in polygynous
marriages with only one second wife each and 10 percent of the male population
remains unmarried. Under these conditions, a society with a generation of 10,000
young males would be expected to produce 1,000 excess men. We can suspect that
excess men are therefore torn between the societal expectations of starting a family
and their perceived inability of living up to these norms.
Polygyny can therefore be understood as “the functional equivalent of a high sex
ratio” (Hudson and Den Boer 2002, 25) tipped toward young males. It reduces the
prospect of marriage drastically among those who do not belong to the top tier of
society. This, in turn, creates competition within the cohort of young males of
marriageable age to acquire the resources necessary to marry, start a family, and
form a household (Yair and Miodownik 2016, 26).
We argue that this nurtures a frustration–aggression mechanism among those
with poor chances of marrying due to a perception of relative deprivation (Gurr
1970, 24; Dollard et al. 1939). These individuals perceive a gap between the values
to which they “believe they are rightfully entitled” and what “they think they are
capable of getting and keeping” (Gurr 1970, 24). This relative deprivation can then
lead to increased frustration, which can turn into aggression and a readiness to exert
violence in an attempt to achieve the goals they feel deprived of (cf. Homer-Dixon
1999, 136).

Excess Men and the Risk of Intergroup Conflict

Since polygyny acts as a barrier to fulfilling the desire and expectation to marry,
young males will opt for alternative strategies to attain the resources that enable
them to reach their societal and economic goals. Anthropological research suggests
Koos and Neupert-Wentz 409

that polygyny-induced reproductive competition will lead to conflicts between

groups (e.g., Borgia 1980; Raffield, Price, and Collard 2017; Henrich, Boyd, and
Richerson 2012). We therefore expect theft and raids to take place in neighboring
communities and not in excess men’s own groups. Hence, we argue that polygyny
makes local forms of intergroup violent conflict more likely.
Our intuition is that the institutions of ethnic groups, as well as the dense social
networks within groups, should deter young males from opting to use violence
against members of their own group to obtain critical resources that would allow
them to marry. Fearon and Laitin (1996) show that intragroup punishment and a high
likelihood of detection prevent group members from disturbing the peace within
their communities. Hence, we do not expect raids to take place within young males’
own ethnic communities but rather outside in neighboring ethnic groups where the
identification and punishment of perpetrators is less likely. Gaining access to
resources in this way is less costly as the violence is directed toward another group
and group-internal reputation and prestige is not jeopardized, perhaps even
rewarded.
We further hold that violence intended to secure access to resources in order to
marry is largely spontaneous, individualistic, and local. Hence, in comparison to
political violence against the state, it does not require sophisticated mobilization
strategies that involve ideology, elite leadership, and resource mobilization (cf.
Fjelde and von Uexkull 2012, 446; Fearon and Laitin 2000). The raiding of another
community by individuals or groups may then result in further violence between the
groups (Fearon and Laitin 1996). Due to the feasibility of the violence and small-
scale mobilization, we expect raids to have a geographical dimension: they are most
likely to take place within close proximity of the raiding group.
To derive our hypothesis, we conceptualize a risk profile for each ethnic group,
which expresses the percentage of their border shared with polygynous neighboring
groups. Hence, ethnic groups—independent of their own marriage institutions—that
are surrounded by polygynous neighboring groups—and thus excess men—should
be at greater risk of being attacked. When groups share a higher percentage of their
borders with polygynous groups, there is a larger pool of excess men likely to attack
the respective community.

Hypothesis 1a: The greater an ethnic group’s share of common borders with
polygynous neighboring groups, the higher their risk of intergroup conflict.

We furthermore hold that the risk of intergroup conflict should be higher in

border regions, particularly when the homelands of ethnic groups are large. The
argument is straightforward: raiding is a costly and risky endeavor. Geographical
distance should therefore play a role in the opportunity structure that leads excess
men to choose to raid neighboring ethnic groups. Excess men will refrain from
intruding deeply into the territory of other ethnic groups and instead choose villages
close to the border regions for tactical reasons. We therefore expect that the effect of
410 Journal of Conflict Resolution 64(2-3)

Figure 1. Example of border proximity and conflict events. This figure shows an illustrative
section of the borders of ethnic group homelands in Chad, Sudan, the Central African
Republic, and South Sudan according to Murdock’s Ethnographic Atlas. The black lines indi-
cate the borders of ethnic group homelands, the shaded buffers around these border lines
extend twenty-five kilometers into each side. The small squares indicate intergroup violent
events of the African Conflict Location and Event Data Project (ACLED) within the twenty-
five kilometer buffer along the border, and the small circles indicate violence deeper within
group territory. For Hypothesis 1a, we use all of these events and for Hypothesis 1b only the
subset of violent events near the border regions.

our explanatory variable, the share of common borders with polygynous neighboring
groups, to be stronger for intergroup conflict events close to group borders. Figure 1
illustrates this conceptualization.
Based on this notion, we derive our second hypothesis:

Hypothesis 1b: The risk of intergroup conflict should be higher in the border
regions between polygynous groups and their neighbors.

Data
We examine our hypothesis in the context of Africa. Africa is well suited for
studying this question because it is home to thousands of ethnic groups that practice
monogamy or polygyny. Since our analysis examines the relationship between
polygyny and contemporary intergroup violence between neighboring groups, we
rely on spatial information. Nunn and Wantchekon (2011) provide a georeferenced
map of the ethnic groups identified in Murdock’s (1959) EA. This map covers 815
groups which serve as our unit of analysis. The EA documents more than eight
sociocultural, political, and economic features of ethnic groups in Africa before
their first contact with Europeans (Fenske 2013; Michalopoulos and Papaioannou
Koos and Neupert-Wentz 411

2013). While Murdock’s EA surely does not measure all group features accurately,
something which Murdock (1969, 3) himself admits, it is the most comprehensive
and coherent source on the sociocultural characteristics of ethnic groups in Africa
and is explicitly meant to facilitate comparative cross-cultural research. For our
purposes, the EA is superior to other data—for instance, the Encyclopedia of World
Cultures (EWC; Levinson 1996) used by Kanazawa (2009) and in the subsequent
response by Gleditsch et al. (2011)—for three reasons. First, the EWC only covers 91
ethnic groups, as compared to the 837 covered in the EA. Second, the EA provides a
complete and systematic coding of polygyny, while the EWC only sporadically
provides information on marriage patterns for better-documented groups, not for all
groups. Third, as discussed above, by utilizing historical data, we can circumvent the
problem of reverse causality.
Our cautious confidence in Murdock’s EA is further supported by a number of
already-seminal studies which examine the long-term effects of precolonial traditional
institutions, agriculture, the slave trade, and colonial borders on contemporary political
and economic development (e.g., Wig 2016; Nunn and Wantchekon 2011; Michalo-
poulos and Papaioannou 2013; Alesina, Giuliano, and Nunn 2013; Fenske 2013; Nunn
2008; Michalopoulos and Papaioannou 2016). Our initial sample is composed of 815
ethnic groups on mainland Africa and Madagascar georeferenced by Nunn and Wantch-
ekon (2011). Of these 815, we lose 9 groups for which no polygyny data are available
and one duplicate entry, leaving us with 805 ethnic groups in our sample.6

Explanatory Variable: Polygynous Neighbors

To operationalize our explanatory variable, we rely on the variable family organi-
zation of the EA. This variable documents the prevailing form of domestic and
familial organization of an ethnic group. It can take up to sixteen different values
and indicates not only monogamy or polygyny but also the living arrangements of
extended families. We recode all ethnic groups according to Murdock’s coding
scheme to create a variable that distinguishes between “monogamy” (44 groups),
“limited or occasional polygyny” (118 groups), and “general polygyny” (643
groups; Murdock 1969, 47).7 Since Murdock does not provide information about
when groups are considered “limited or occasionally polygynous,” we focus on the
effect of general polygyny.8 This corresponds to our theoretical argument because it
will be general polygyny that produces a critical mass of excess men and increases
competition for brides among males.
Panels (A) and (B) in Figure 2 show the spatial distribution of polygyny among
ethnic groups in Africa.9 We can see that in all regions of Africa, both monogamous
and polygynous groups exist. However, there is a certain degree of clustering, which
we take into account with country fixed effects in the statistical analysis and a quasi-
experimental matching approach in the robustness checks.
According to our argument, the social pressure that polygyny creates for unmar-
ried men has little effect on violent conflict within a group but rather affects violence
412 Journal of Conflict Resolution 64(2-3)

Figure 2. Ethnic groups, polygyny, and intergroup conflict events. (A): ACLED events. (B):
UCDP-GED events.

toward other neighboring groups independent of whether these are monogamous or

polygynous. We use the QGIS software version 3.4 to calculate the percentage of the
border that is shared with neighboring polygynous groups (polygynous neighbors)
for each observed group.10 This measure of relative exposure to polygynous neigh-
boring ethnic groups ranges from 0 to 100 percent.
For instance, if a group has a total border length of hundred kilometers and fifty
kilometers thereof are shared with polygynous groups and the remaining fifty kilo-
meters are shared with monogamous groups, the explanatory variable polygynous
neighbors has a value of 0.5. Figure 3 shows the density of this variable. We can
observe that about 60 percent of groups are completely surrounded by polygynous
groups (value 1), which represents the highest risk profile according to our hypothesis.
The remaining 40 percent are evenly distributed between no polygynous neighbors at
all (value 0) and being largely surrounded by polygynous neighbors (value 0.9).
In addition to our main explanatory variable polygynous neighbors, we include a
binary measure that captures whether the observed group itself practices polygyny. While
our theoretical argument is silent about the risk other polygynous groups pose, we believe
it is important to control for potential group-internal confounding effects of polygyny.

Outcome Variables: Intergroup Conflict

Our outcome is intergroup conflict. Intergroup conflicts refer to violent conflict events
between members of ethnic groups including armed civilians and ethnic militias. We
draw our conflict data for Africa from the Armed Conflict Location and Event Data
(ACLED, version 7, 1997–2016) project (Raleigh et al. 2010) and the Uppsala
Koos and Neupert-Wentz 413

Figure 3. Distribution of explanatory variable (percentage of shared border with neighboring

polygynous groups).

Conflict Data Program (UCDP) Georeferenced Event Dataset (UCDP-GED, version

17.1, 1989–2016; Sundberg and Melander 2013; Croicu and Sundberg 2017). As both
data sets have their merits, we use ACLED and UCDP-GED as separate samples to
assess our hypotheses.11 We project violent events of the respective data sets onto the
map of the ethnic homelands (see panels [A] and [B] in Figure 2).
As already discussed, we theorize that polygyny pushes men into criminal or violent
activities. Since men won’t turn against their own group, they will direct violence
against other groups, most likely neighboring groups, to accumulate assets and—at
times—women and children. We therefore restrict our analysis to intergroup violent
events within an ethnic group’s territory and exclude any events of political violence that
include state actors.12 We assume that the institution of polygyny is most “sticky” in
rural areas where modernization processes are slower to occur and traditional norms and
practices can withstand the influence of market economies, education, and moderniza-
tion processes. Therefore, we exclude events of urban violence by placing a buffer of
twenty kilometers around urban centers (population > 100,000) and excluding events
that fall within these boundaries. This leaves our ACLED sample with 4,919 events from
1997 to 2016 and our UCDP-GED sample with 2,933 events from 1989 to 2016. We are
left with events of intergroup violence in rural areas and explicitly exclude ethnic urban
violence (e.g., the ethnic postelection clashes in Nairobi in 2008). We rely on this
outcome measure to examine Hypothesis 1a. To test Hypothesis 1b, we add a fifty
kilometer buffer zone to all group boundaries (twenty-five kilometers into each group’s
territory, see Figure 1) and then count only those violent events from our specification
that have occurred within this buffer zone inside an ethnic group’s territory. This reduces
the ACLED sample to 3,724 events and the UCDP-GED sample to 2,134 events.
Figure 4 shows the density plots of these two outcome variables for the ACLED
sample (panels [A] and [B]) and the UCDP-GED sample (panels [C] and [D]). All
414 Journal of Conflict Resolution 64(2-3)

Figure 4. Distribution of outcome variable. (A): ACLED events. (B): ACLED events, fifty
kilometer buffer. (C): UCDP-GED events. (D): UCDP-GED events, fifty kilometer buffer.

variables show a negative binomial distribution—that is, almost 60 and more than 70
percent of ethnic groups have not experienced intergroup conflict in their territory
according to ACLED and UCDP-GED, respectively. We want to emphasize here
that ACLED and UCDP-GED reflect certain biases. First, there are no data for
violent events before 1996 in ACLED and 1989 for UCDP-GED. Second, since
ACLED and UCDP rely on news reports, reporting bias is at work. Media bias is
likely to particularly underestimate rural and nonlethal violent events, which are less
likely to be covered by international media outlets than urban and large-scale vio-
lence (cf. Weidmann 2016). Nevertheless, these are the most established subnational
violent-event data sets that span dozens of countries and are therefore appropriate for
cross-country analysis.

Control Variables
As shown by Dalton and Leung (2014), the social institution of polygyny as coded
by Murdock’s EA has persisted over time and is highly correlated with contemporary
Koos and Neupert-Wentz 415

levels of polygyny in Africa. Since our main goal is to estimate a causal relationship
and not primarily to explain as much variance in our outcome as possible, our
conditioning strategy and the selection of control variables aim to block the back-
door paths, meaning noncausal associations, between our explanatory variable poly-
gyny—the shared border with polygynous neighbors—and intergroup conflict
(Morgan and Winship 2014, Part III; Pearl 2009).
Since Murdock’s coding of polygyny relates to the time before colonialization,
we need to consider only those factors that may have contributed to the emergence of
polygyny in the first place and may affect our outcome intergroup conflict through
mechanisms other than polygyny.
Using the most parsimonious setup, we control for the size of the territory of an
ethnic group (log of land area) and the group size (log of population). Furthermore,
there are several geographic conditions that can confound the relationship between
polygyny and conflict. High agricultural fertility, value of land, and intensive agri-
culture may have resulted in more economic prosperity and affected social class
stratification, thereby allowing wealthier men to marry more women (Fenske 2013).
Furthermore, there may be a connection between peasant violence and marriage
institutions (Mokuwa et al. 2011). At the same time, valuable land may have led
to more competition and cycles of conflict between groups over time, independent of
polygyny. We therefore include measures that express land value, including distance
to coast, mean elevation, agricultural suitability, and a malaria stability index.
Apart from these geographic features, several historical pretreatment conditions
could have similarly affected both our explanatory and outcome variables through
backdoor paths. Therefore, we further include an indicator for distance to empires,
the nearest precolonial conflict, and a dummy for the existence of ancient cities
(Besley and Reynal-Querol 2014).
We also control for slave exports, as the extraction and death of men may well
contribute to the emergence of polygyny and, independent of that, keep fueling
contestation and conflict between groups over time. We draw the above control
variables from a data set by Michalopoulos and Papaioannou (2013). Lastly, we
include a variable, which measures the share of Muslims per country since the Koran
allows marrying several wives (Pew Research Center 2011).
Broadly speaking, our model specification is consistent with other works that
examine the long-term effects of historical institutions (e.g., Nunn and Wantchekon
2011; Michalopoulos and Papaioannou 2013; Alesina, Giuliano, and Nunn 2013).
We provide the descriptive summary statistics for all variables in Table A1 of the
Online Appendix.

Analysis
Our outcome variable counts the number of intergroup conflict events per total
ethnic group territory (Hypothesis 1a) and those within a twenty-five kilometer
buffer along the group’s border to its neighboring group (Hypothesis 1b). As Figure
416 Journal of Conflict Resolution 64(2-3)

4 has shown, these count variables are overdispersed. Therefore, we use a negative
binomial model with robust standard errors clustered at the country level and we
include country fixed effects to account for unobserved heterogeneity.
In Table 1, we test Hypothesis 1a using both ACLED and UCDP-GED. We focus
first on the discussion of the main variables and then highlight some control vari-
ables. Columns 1 and 3 show the parsimonious models including our main expla-
natory variable polygynous neighbors, the polygyny status of the observed group,
and land area and population size of the ethnic group.13 As expected by Hypothesis
1a, a higher share of common borders with polygynous neighbors has a positive and
statistically significant effect on the likelihood of intergroup conflict. Columns 2 and
4 show our main specification, in which we control for pretreatment exogenous
geographic and historical variables. Essentially, the effect of polygynous neighbors
remains robust to the parsimonious model. These results lend support to Hypothesis
1a, which suggests that a higher percentage of shared borders with polygynous
groups increase an observed group’s conflict risk.
To assess Hypothesis 1b, Table 2 shows the results when we limit our outcome
variable to conflicts within a twenty-five kilometer distance from group boundaries.
We argue that excess men on raids have a tactical advantage in villages in the outer
regions of neighboring groups’ territories, which lie closer to their own homelands.
As in Table 1, columns 1 and 3 show the parsimonious models and columns 3 and 4
our main specification. We see an essentially similar pattern at work: sharing more
borders with polygynous neighboring groups increases the number of conflicts for an
observed group. While the results are supportive of our general theoretical argument,
the regression coefficients of polygynous neighbors are similar to Table 1. We
examine the substantive effects below.
Apart from our main variable, a few other variables are noteworthy. For instance,
in all models using the ACLED sample, we see that when an observed group is
polygynous, the effect is negative and statistically significant. Monogamous groups
are therefore more likely than polygynous groups to experience violent events on
their territories. We did not formulate prior expectations on the effect of an observed
group’s mode of marriage and can therefore only propose a tentative explanation. As
an addition to our main theoretical argument, an observed group’s mode of marriage
may proxy a target selection mechanism. Excess men that result from polygynous
groups are easily mobilized for offensive acts but should also increase defensive
capabilities of their group. However, monogamous groups should not produce
excess men and thereby the pool of mobilized defenders should be smaller. These
fighting capabilities may matter greatly for excess men’s strategic considerations of
whom to attack. When the pool of defenders in monogamous groups is smaller than
in polygynous ones, attacking excess men should rationally choose the easier target:
monogamous groups. Consequently, monogamous groups should be attacked more
often and thus experience higher levels of violence. Nevertheless, we believe that
this aspect requires more in-depth research to be understood more comprehensively.
Koos and Neupert-Wentz 417

Table 1. Polygynous Neighboring Groups and Intergroup Conflict Events.

(1) (2) (3) (4)

ACLED ACLED UCDP-GED UCDP-GED

þ
Polygynous neighbors 1.55** (0.56) 1.66*** (0.34) 1.65 (0.96) 1.80** (0.66)
Observed group: 0.76* (0.38) 0.72** (0.26) 0.42 (0.44) 0.25 (0.42)
Polygynous
Land area (log) 0.45*** (0.11) 0.41*** (0.06) 0.48* (0.24) 0.51*** (0.12)
Population (log) 0.43*** (0.09) 0.56*** (0.06) 0.42þ (0.22) 0.59*** (0.13)
Precolonial conflict 0.31 (0.55) 0.15 (1.22)
Distance to coast 0.00þ (0.00) 0.00* (0.00)
Mean elevation 0.24 (0.63) 0.01 (0.77)
Agricultural suitability 0.28 (0.62) 1.34 (0.87)
Malaria stability index 0.54 (0.73) 2.16* (1.05)
Precolonial kingdom 0.59** (0.20) 1.31*** (0.27)
Distance to empires 0.35 (0.60) 0.19 (0.94)
Major city in AD 1400 0.64* (0.28) 0.65 (0.61)
Slave exports by land (log) 0.10** (0.04) 0.10 (0.07)
Muslims (%) 0.21*** (0.04) 0.54** (0.20)
Intense agriculture 0.42** (0.15) 0.46* (0.22)
Country FE Yes Yes Yes Yes
Pseudo R2 .146 .164 .158 .189
AIC 2,966.20 2,923.44 1,851.33 1,781.38
BIC 3,008.42 3,012.57 2,015.51 1,936.18
Observations 805 805 805 805

Note: Outcome variable: number of intergroup conflict events per ethnic group territory. Robust
standard errors clustered by country. ACLED ¼ Armed Conflict Location and Event Data; UCDP-GED ¼
UCDP Georeferenced Event Dataset; FE ¼ fixed effects; AIC ¼ Akaike information criterion; BIC ¼
Bayesian information criterion.
þ
p < .10.
*p < .05.
**p < .01.
***p < .001.

The other control variables have the expected effect direction and significance.
Land area and population size increase the number of conflict events. The indicator
for slave exports has a consistent negative effect in the ACLED sample, which
reflects previous findings arguing that areas affected by the slave trade and the
resulting reduction of men in these societies reduced the pressure on the marriage
market (Dalton and Leung 2014).
418 Journal of Conflict Resolution 64(2-3)

Table 2. Polygynous Neighboring Groups and Intergroup Conflict Events in 50 km Buffer

Zone.

(1) (2) (3) (4)

ACLED ACLED UCDP-GED UCDP-GED

50 km Buffer 50 km Buffer 50 km Buffer 50 km Buffer

Polygynous neighbors 1.50** (0.57) 1.59*** (0.34) 1.78þ (0.98) 1.84** (0.66)
Observed group: 0.81* (0.34) 0.79*** (0.23) 0.54 (0.43) 0.36 (0.44)
Polygynous
Land area (log) 0.27* (0.11) 0.23*** (0.07) 0.35 (0.24) 0.39** (0.14)
Population (log) 0.48*** (0.10) 0.61*** (0.07) 0.42þ (0.23) 0.59*** (0.13)
Precolonial conflict 0.27 (0.59) 0.21 (1.16)
Distance to coast 0.00þ (0.00) 0.00* (0.00)
Mean elevation 0.32 (0.65) 0.10 (0.73)
Agricultural suitability 0.24 (0.62) 1.33 (0.82)
Malaria stability index 0.37 (0.73) 2.18* (1.02)
Precolonial kingdom 0.58** (0.18) 1.30*** (0.27)
Distance to empires 0.48 (0.69) 0.28 (0.93)
Major city in AD 1400 0.53þ (0.29) 0.60 (0.60)
Slave exports by 0.11** (0.04) 0.08 (0.07)
land (log)
Muslims (%) 0.09** (0.03) 0.07 (0.20)
Intense agriculture 0.40** (0.14) 0.44þ (0.23)
Country FE Yes Yes Yes Yes
Pseudo R2 .141 .158 .150 .181
AIC 2,790.47 2,752.80 1,787.75 1,718.82
BIC 2,837.38 2,841.93 1,956.62 1,873.61
Observations 805 805 805 805

Note: Outcome variable: number of intergroup conflict events per ethnic group territory. Robust
standard errors clustered by country. ACLED ¼ Armed Conflict Location and Event Data; UCDP-GED ¼
UCDP Georeferenced Event Dataset; FE ¼ fixed effects; AIC ¼ Akaike information criterion; BIC ¼
Bayesian information criterion.
þ
p < .10.
*p < .05.
**p < .01.
***p < .001.

Apart from the statistical significance, the results of our main models 2 and 4 in
Tables 1 and 2 are substantively meaningful. We use these four models to compute
the predicted number of events by varying the values of our explanatory variable
polygynous neighbors from its minimum 0 to its maximum 1. Each of the panels in
Figure 5 shows that an increase in shared borders with polygynous groups increases
the predicted number of intergroup conflict events significantly. Moving from 0
percent to 100 percent shared border with polygynous groups increases the predicted
number of intergroup conflict events by about 300% from less than two events to
Koos and Neupert-Wentz 419

Figure 5. Predicted number of conflict events. The black line represents the predicted
number of conflict events based on varying percentages of the border shared with polygynous
neighboring groups. The gray shaded areas indicate a 95 percent confidence interval. The
predictions are based on models 2 (panel A) and 4 (panel C) of Table 1 and models 2 (panel B)
and 4 (panel D) of Table 2. (A): ACLED events. (B): ACLED events, 50 km buffer. (C): UCDP-
GED events. (D): UCDP-GED events, 50 km buffer.

almost eight in panel (A). The pattern is similar in the other three panels, albeit the
effect strength is somewhat smaller, in particular when using the UCDP-GED data.
While the pattern holds for conflicts in border regions (Hypothesis 1b), we do not
find a stronger effect and thus do not find support for Hypothesis 1b. Although the
absolute number of predicted events appears small, the effects are massive in mag-
nitude for both hypotheses. Remember that violent-event data particularly for rural
areas—not so much for cities—typically suffer from underreporting bias (cf. Eck
2012; Weidmann 2016). We therefore assume to underestimate the effect. It is
furthermore important to note that the substantive effect occurs not only when we
compare the extremes of having no (0) to only (1) polygynous neighbors but also
with more moderate in-between ranges.
420 Journal of Conflict Resolution 64(2-3)

In sum, we find robust support for Hypothesis 1a, which suggests that a larger
share of polygynous neighbors increases the number of conflict events between
ethnic groups. The effect of polygynous neighbors on conflicts in the border regions
(Hypothesis 1b) is of comparable magnitude, but not stronger. Bringing this result
back to our research question, we can say that polygyny is associated with higher
levels of intergroup conflict.

Robustness Tests
Next, we perform a number of robustness checks to minimize the risk of model and
specification dependence. First, we exchange country-level fixed effects against
region-level fixed effects and add spatial lags of the outcome variable on the left-
hand side of the model. This model reflects our main intuition that states (i.e., country
fixed effects) can be conceptualized as posttreatment variables (see King, Keohane,
and Verba 1994, 182). Table A2 shows that the results remain almost identical.
Importantly, our explanatory variable polygynous neighbors retains its positive effect
and is highly significant both in the UCDP-GED and in the ACLED sample.
Second, we take a closer look at our outcome variable and exclude events that
could potentially reflect intragroup fighting and not intergroup conflict. To do so, we
manually exclude UCDP-GED events in which ethnic subgroups are mentioned to
fight against each other. The number of UCDP-GED events decreases by 111 from
3,085 to 2,974. In the ACLED sample, we excluded 378 events in which both
conflict actors have the same name. This reduced the number of events from
4,985 to 4,607 events. We employ the same specification as in the main results
presented above. The coefficient of the polygynous neighbors variable remains
robust in both samples UCDP-GED and ACLED (Table A3).
Third, we use different specifications for our explanatory variable. In a first step,
we rely on the EA’s initial coding of the polygyny variable, which includes an
intermediate category for “limited or occasional polygyny.” Table A4 in the Online
Appendix shows the results when using the ACLED data (model 1) and the UCDP-
GED data (model 2). There are now two border variables. The variable polygynous
neighbors: limited expresses the percentage of border shared with groups that prac-
tice “limited or occasional” polygyny. The variable polygynous neighbors: general
is identical to the one used in the main analysis. Model 1 shows that while the limited
polygyny variable is positive and insignificant, the general polygyny variable has a
stronger effect on the number of conflict events than in the main model. Model 2
uses the UCDP-GED data and shows a similar pattern. This result supports our
suspicion that it is general and widely practiced polygyny that results in intergroup
conflict.
Fourth, in Table A5, we use state-based conflict events on the group’s territory as
the outcome variable. We include these two models to demonstrate that polygyny is
not related to state-based conflict as argued in our theory. We exclude the polygy-
nous neighbors variable since this measure has no theoretical relationship or
Koos and Neupert-Wentz 421

conceptual relevance to violence between an ethnic group and the state. If anything,
in the spirit of Kanazawa (2009) and Gleditsch et al.’s (2011) work, we believe that
polygyny can additionally create youth bulges that serve as a recruitment pool for
rebel groups or government forces in civil wars. However, this does not seem to be
the case since the variable observed group: polygynous is not significant.
Fifth, we preprocess the data with the coarsened exact matching (CEM) algorithm
to remove observations without common empirical support. Blackwell et al. (2009)
have shown that CEM reduces model and specification dependence and thereby
improves causal inference. The detailed approach of applying the CEM procedure
is described in the Online Appendix before the presentation of the results in Table
A6. The results confirm our main findings, although the CEM procedure signifi-
cantly reduces our sample size. The effect of our explanatory variable polygynous
neighbors is statistically significant at 1 percent.
Sixth, we additionally test political (polity level and change) and economic
variables (GDP growth) used in standard civil war models as well as measures for
legal polygamy, polygamy provisions in customary law, and a women’s rights
indicator from the data used by Gleditsch et al. (2011) on the basis of the Woman-
Stats Project (Caprioli et al. 2009). In Table A7, we present the results which show
that our explanatory variable polygynous neighbors remains statistically significant.
Note that our sample is reduced from 805 to 761 observations due to missing data.
These tests demonstrate that even with the inclusion of posttreatment political,
economic, and legal variables, our hypothesis holds.
Finally, we use the number of polygynous neighboring groups instead of the
percentage of shared border with polygynous neighboring groups as explanatory
variable. The number of polygynous neighbors ranges from 0 to 19 with a mean
value of 4.2 and a standard deviation of 2.5. Table A8 reports the results in the same
regression setup as in the main table. The results strongly support our previous
findings. The number of polygynous neighbors is statistically significant at 1 percent
and 5 percent (models 1 and 2) and 1 percent (models 3 and 4), respectively. The
other variables retain their effect direction and statistical significance.

Mechanism: Frustration and Aggression among Excess Men

In the previous section, we have performed an analysis at the ethnic group level and
established a link between polygyny and intergroup conflict. Next, we provide
evidence on the suggested mechanism linking polygyny and conflict between ethnic
groups: the role of excess men. Specifically, we rely on individual-level evidence
from a pooled sample of Afrobarometer surveys, which we matched with our poly-
gyny measure.
We have argued that polygyny produces excess men who are disadvantaged in
competing as viable partners on the marriage “market,” largely because they lack the
financial means to compete with better-off men. We further argued that this inability
leads to frustration because the expectations and social pressure on men to start a
422 Journal of Conflict Resolution 64(2-3)

family and produce offspring are supposedly particularly high in such traditional
polygynous social environments. We should therefore observe that young men
without family in polygynous ethnic groups perceive more inequality than their
peers in monogamous ethnic groups where the competition on the marriage market
is by definition much less fierce.
Since gaining resources through stealing, raiding, and plundering in neighboring
ethnic groups allows excess men to increase the financial competitiveness on the
marriage “market,” we should also observe that men in polygynous societies are
somewhat more inclined to accept violence as a legitimate means. We test the
mechanism using data from the 2005 Afrobarometer. In this version of the Afroba-
rometer, Nunn and Wantchekon (2011) have matched the ethnic group names of
Murdock’s EA with those respondents reported in the Afrobarometer. We match our
polygyny dummy variable to the ethnic groups to distinguish between respondents
who belong to monogamous or polygynous ethnic groups.
The fully merged sample of the Afrobarometer includes 25,397 respondents from
eighteen countries. These countries reflect a bias in the sense that these are rather
stable countries (e.g., Benin, Botswana, Ghana, Kenya, Mozambique) and exclude
African conflict hotspots such as the DR Congo, Sudan, South Sudan, or Chad.
However, we believe that traces of our mechanisms should be independent of the
political context. In order to test our assumptions, we limit the Afrobarometer
sample to those we believe are most severely affected, young men below 40 years
without children (excess men).14 Furthermore, we employed placebo tests on two
further samples where we would not expect a similar effect: (a) for men above 40
years and (b) for women.
We have identified two questions in the Afrobarometer that tap into aspects that
we believe are indicative of a frustration–aggression mechanism, which we propose
to link to polygyny and conflict. These questions reflect perceptions of inequality
and a justification to use violence.
In Figure 6, we show the predicted values for the two questions depending on
whether a respondent belongs to a polygynous group (1) or not (0).15 The underlying
regression models are presented in Table A9 in the Online Appendix.16 The left
panels show the results for men below 40 years without children (excess men), the
central panels for men above 40 years, and the right panels for women. The upper
row shows the predicted values for the degree to which people feel treated unequally
under the law (Q53D).17 Higher values indicate higher perceptions of inequality.
Young childless men in polygynous societies are significantly more likely to report
higher perceived inequality than their peers belonging to monogamous groups. For
men above 40 years, we do not find this effect. Women in polygynous groups also
report perceptions of inequality, which we believe resonates with the notion of
gender inequality associated with polygyny (cf. McDermott and Cowden 2018;
Hudson et al. 2010).
The lower row shows the results for the question on whether people see the
use of violence as a justified means for their cause (Q51), where higher values
Koos and Neupert-Wentz 423

Figure 6. Individual-level evidence from matched Afrobarometer surveys. Bars indicate 95

percent confidence intervals. The regression model controls for age, age2, education level,
assets, and an urban/rural dummy with robust standard errors. The regression table under-
lying these predicted responses is available in Table A9. Figure A1 in the Online Appendix
shows the first differences of these estimates which are robust to the results. (A) and (B)
denote men below 40 years without children. (C) and (D) denote men above 40 years. (E) and
(F) denote women.

indicate agreement. While young men below 40 years without children (excess
men) are significantly more likely to view violence as justified, our placebo
groups of men above 40 years and women do not share this perspective. This
result supports our initial argument that excess men are an important link
between polygyny and conflict and not a broader societal disposition toward
violence.
For males above the age of 40, we do not find any statistically significant difference
in the two survey items between polygynous and monogamous groups. Next to other
possible explanations, such as age, this could be argued to add to the self-sustainability
of polygyny as a sticky institution. Thereby, one could speculate that polygynous
males will likely be more supportive of transmitting the marriage institution of poly-
gyny to the next generation (McDermott and Cowden 2018; Hudson 2018).
In sum, the results provide individual-level evidence that excess men hold views
compatible with a disposition to theft, crime, and raids. These results provide sug-
gestive support for our mechanisms that young, childless men in polygynous soci-
eties are under social pressure to perform. They perceive this pressure as unequal and
424 Journal of Conflict Resolution 64(2-3)

unfair and that, more broadly, they evince a greater readiness to exert to violence
than their peers in monogamous groups. We acknowledge that these questions are
very general and do not really point toward violence against neighboring ethnic
groups. However, given the data constraints and the usual noise in survey data,
we believe these findings lend additional powerful evidence to our theoretical argu-
ment that polygyny produces excess men which in turn contribute to intergroup
violence experienced between neighboring ethnic groups.

Conclusion
We have argued that, by definition, polygyny creates a social imbalance where a
few, usually well-off, men marry many wives and many, usually poor, men
marry late or never. Polygyny therefore systematically creates a surplus of
young, poor, unmarried men: excess men. Since marriage, family, and offspring
are often the social metrics according to which the value of a man is assessed in
traditional societies, excess men seek alternatives to become viable mates. In
traditional and particularly rural settings, the ethnic group is perceived as the
extended family, which leads excess men to abstain from turning against their
kin; however, they have an incentive to pursue violent economic ventures (e.g.,
theft, crime, raids) in neighboring groups. From a security perspective, polygyny
is not a problem for the polygynous group itself but rather for its neighbors.
Being surrounded by many polygynous groups increases the risk of intergroup
violence.
To examine this theory, we have applied georeferenced data on polygyny for
more than 800 African ethnic groups and combined these with violent-event data.
We have used a set of exogenous geographic and precolonial controls and report a
strong effect of polygyny on conflict—that is, we find robust support for this theory.
By exploiting Afrobarometer survey data, we find additional individual-level evi-
dence for our proposed underlying mechanism that respondents who belong to
polygynous ethnic groups hold more problematic views on perceived fairness, the
obedience to the rule of law, and the readiness to use violence.
Our article contributes to several research strands and should therefore be of
interest to a wide audience of scholars. First of all, it emphasizes the importance
of the institutions of marriage and family and their role in social order, peace, and
conflict. Our study substantially improves the only two existing quantitative studies
with conflicting findings on polygyny and conflict in several ways. We view poly-
gyny as a local dynamic that affects local violence. We also use more coherent,
expansive, and reliable data and engage in elaborate reliability tests using alternative
data sources. We provide a more sensible model specification, taking into account
the geographic and historical determinants that have affected the emergence of
polygyny in the first place.
Second, our article contributes to the small but growing literature on low-
intensity community-level conflict (e.g., Eck 2014; Varshney 2003; Fjelde and von
Koos and Neupert-Wentz 425

Uexkull 2012; Raleigh 2010). This research strand has significantly increased our
understanding of conflict processes worldwide, also because local violent events
frequently spur larger conflicts (Brass 1997). Due to this, we stress that the type of
conflict analyzed should be carefully chosen on the basis of theoretical considera-
tions. We disagree, for instance, that large-scale civil wars are driven by polygyny as
a “first law” (Kanazawa 2009).
Third, this study contributes to the literature on the long-term effects of historical
institutions. As we have theorized and demonstrated, polygyny can be understood as
a “sticky” institution that is still at work in today’s societies in Africa (Dalton and
Leung 2014). In this sense, our article adheres to the spirit of recent seminal studies
examining the long-term effects of traditional institutions (Wig 2016; Michalopou-
los and Papaioannou 2013), precolonial nation states (De Juan and Koos 2019), the
slave trade (Nunn 2008; Nunn and Wantchekon 2011), the Berlin Conference
(Michalopoulos and Papaioannou 2016), and ancient wars (Besley and Reynal-
Querol 2014).
With regard to its practical relevance, our article provides insights for humani-
tarian and development agencies. We are aware that polygyny does not explain all
conflicts, but we believe that our study provides systematic support for anecdotal
examples where polygyny and conflict are widespread—for instance, in South
Sudan, the DR Congo (Verweijen 2017), Nigeria, and even Western countries
(Rauch 2006). While our analysis focuses on Africa, we believe that the operating
principles and societal implications of polygyny are—with few exceptions—uni-
versally problematic as they create a cohort of society that has always been associ-
ated with trouble around the world: excess men.

Acknowledgments
We would like to thank Kristian Skrede Gleditsch, Katharina Holzinger, Valerie M. Hudson,
Tobias Ide, Alexander de Juan, and Stephen Nemeth, four anonymous reviewers, as well as
participants at the annual meetings of American Political Science Association (APSA 2017),
the European Political Science Association, the German Political Science Association
(DVPW 2017), the International Studies Association (ISA 2018), and a research colloquium
at the University of Essex for their helpful comments. All errors are our own.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, author-
ship, and/or publication of this article.

Funding
The author(s) disclosed receipt of the following financial support for the research, authorship,
and/or publication of this article: Financial support from the German Research Foundation
(DFG Grants HO 1811/10-1 and KO 5170/1-1) is gratefully acknowledged.

ORCID iD
Carlo Koos https://orcid.org/0000-0002-1259-7495
426 Journal of Conflict Resolution 64(2-3)

Supplemental Material
Supplemental material for this article is available online.

Notes
1. Gleditsch et al. (2011) furthermore substantiate the nonreplicability by arguing that
misogyny, rather than polygyny, is the mechanism driving political violence, which they
test at the state level. Although a broader set of gender-based discrimination may be a
source of political violence (see, e.g., Melander 2005; Hudson et al. 2010), we do not
believe that polygyny—which can be regarded as part of misogynistic practices—and a
general concept of misogyny adhere to the same underlying mechanism with regard to
local conflict. In other words, misogyny and polygyny cannot be seen as competing
hypotheses when studying intergroup conflicts at the local level.
2. Another explanation of the emergence of polygyny is economic, that is, as the result of
income inequality and female subsistence contributions (see White and Burton 1988, 872,
for an overview).
3. Dalton and Leung (2014) arrive at their results by analyzing forty-five DHS surveys in
twenty-five African countries between 1990 and 2010. They use a question which asks
married female respondents whether they have a co-wife to calculate current polygyny
rates at the ethnic group level specified by Nunn’s and Wantchekon’s (2011) EA map.
4. This does not mean that these marriages are voluntary unions. For a discussion on female
choice in polygynous societies, see the third and fourth chapter of Barash (2016).
5. For a comparative assessment of socioeconomic inequality, hierarchy, and polygyny, see
Betzig (1986).
6. For the Kisama and Bomvana, there are two entries (name¼KISAMA and
v107¼BOMVANA).
7. For a description of the coding of the variable, see Online Appendix Section 2.
8. We provide robustness checks using the three-scale variable (monogamy, limited poly-
gyny, and general polygyny) in Table A4 in the Online Appendix. The results are robust.
9. To produce the maps, we used the georeferenced group borders used in Nunn and
Wantchekon (2011). These are available at http://scholar.harvard.edu/files/nunn/files/
murdock_shapefile.zip.
10. If a group borders the sea, a lake, or uninhabited territory (i.e., West Saharan Desert,
Libyan Desert), we exclude these border segments from the denominator.
11. One favorable feature of ACLED is that its categories—in particular communal militia
activity—resonate well with our hypothesized effect on local intergroup violence. Relat-
edly, ACLED also includes nonlethal violent events, which also speaks to our theory.
Adversely, ACLED has been criticized to incorporate an urban bias (Eck 2012, 132), a
problem we can address as we exclude events in urban areas. Conversely, UCDP-GED is
argued to be superior to ACLED because the media sources of UCDP-GED are more
consistent by focusing on major international media outlets.
12. Specifically, for the ACLED sample, we use the INTERACTION variable, which clas-
sifies each event according to prespecified actor interactions. We are interested only in
Koos and Neupert-Wentz 427

those events which involve ethnic militias but exclude any political or rebel-based orga-
nizations. We use only the following interaction values: 40—sole communal militia
action, 44—communal militia versus communal militia, 47—communal militia versus
civilians. Furthermore, we restrict our sample to events where (1) the spatial location
quality (GEO_PRECIS) is exact or (2) part of a region. We exclude (3) less precise
events. From the UCDP-GED, we include only events which have been categorized as
nonstate conflict (type_of_violence ¼ 2) and events whose location (where_precise) was
either (1) exactly identified or (2) identified within a twenty-five kilometer radius
[“region” ¼ “Africa” AND “type_of_violence” ¼ 2 AND (“where_precise” ¼ 1 OR
“where_precise” ¼ 2)].
13. All tables are produced using estout (Jann 2005, 2007).
14. Unfortunately, these Afrobarometer surveys do not include questions on whether a
respondent is married, but we believe that having no children is a reasonable indicator
of being married or not in many African societies.
15. Figure A1 shows the first differences between members of polygynous and monogamous
groups, and the results are robust to those in Figure 6.
16. We use a linear model with robust standard errors and control for age, age2, education,
assets, and a dummy for urban/rural residence.
17. Afrobarometer Merged Round 3 Codebook, 2005, http://www.afrobarometer.org/data/
merged-round-3-codebook-18-countries-2005.

References
Alesina, Alberto, Paola Giuliano, and Nathan Nunn. 2013. “On the Origins of Gender
Roles: Women and the Plough.” The Quarterly Journal of Economics 128 (2):
469-530.
Barash, David P. 2016. Out of Eden: The Surprising Consequences of Polygamy. New York:
Oxford University Press.
Becker, Gary Stanley. 1993. A Treatise on the Family. Cambridge, MA: Harvard University
Press.
Besley, Timothy, and Marta Reynal-Querol. 2014. “The Legacy of Historical Conflict: Evi-
dence from Africa.” American Political Science Review 108 (2): 319-36.
Betzig, Laura L. 1986. Despotism and Differential Reproduction: A Darwinian View of
History. New York: Aldine.
Bisin, Alberto, and Thierry Verdier. 2000. ““Beyond the Melting Pot”: Cultural Transmission,
Marriage, and the Evolution of Ethnic and Religious Traits.” The Quarterly Journal of
Economics 115 (3): 955-88.
Blackwell, Matthew, Stefano M. Iacus, Gary King, and Giuseppe Porro. 2009. “cem: Coar-
sened Exact Matching in Stata.” The Stata Journal 9 (4): 524-46.
Borgia, Gerald. 1980. “Human Aggression as a Biological Adaptation.” In The Evolution of
Human Social Behavior, edited by Joan S. Lockard, 165-91. New York: Elsevier.
Brass, Paul R. 1997. Theft of an Idol: Text and Context in the Representation of Collective
Violence. NJ: Princeton University Press.
428 Journal of Conflict Resolution 64(2-3)

Caprioli, Mary, Valerie M. Hudson, Rose McDermott, Bonnie Ballif-Spanvill, Chad F.

Emmett, and S. Matthew Stearmer. 2009. “The Womanstats Project Database: Advancing
an Empirical Research Agenda.” Journal of Peace Research 46 (6): 839-51.
Chagnon, Napoleon A. 1988. “Life Histories, Blood Revenge, and Warfare in a Tribal Pop-
ulation.” Science 239 (4843): 985-92.
Croicu, Mihai, and Ralph Sundberg. 2017. “UCDP GED Codebook Version 17.1.” Uppsala,
Sweden: Department of Peace and Conflict Research, Uppsala University.
Dalton, John T., and Tin Cheuk Leung. 2014. “Why Is Polygyny More Prevalent in Western
Africa? An African Slave Trade Perspective.” Economic Development and Cultural
Change 62 (4): 599-632.
De Juan, Alexander, and Carlo Koos. 2019. “The Historical Roots of Cooperative Behavior—
Evidence from Eastern Congo.” World Development 116:100-12.
Dollard, John, Neal E. Miller, Leonard W. Doob, Orval H. Mowrer, and Robert R. Sears.
1939. Frustration and Aggression. New Haven, CT: Yale University Press.
Eck, Kristine. 2012. “In Data We Trust? A Comparison of UCDP GED and ACLED Conflict
Events Datasets.” Cooperation and Conflict 47 (1): 124-41.
Eck, Kristine. 2014. “The Law of the Land: Communal Conflict and Legal Authority.”
Journal of Peace Research 51 (4): 441-54.
Edlund, Lena, Hongbin Li, Junjian Yi, and Junsen Zhang. 2007. “Sex Ratios and Crime:
Evidence from China’s One-child Policy.” (No. 3214): IZA Discussion Papers. Bonn,
Germany: Institute for the Study of Labor.
Fenske, James D. 2013. “Does Land Abundance Explain African Institutions?” The Economic
Journal 123 (573): 1363-90.
Fenske, James D. 2015. “African Polygamy: Past and Present.” Journal of Development
Economics 117:58-73.
Fearon, James D., and David D. Laitin. 1996. “Explaining Interethnic Cooperation.” Amer-
ican Political Science Review 90 (04): 715-35.
Fearon, James D., and David D. Laitin. 2000. “Violence and the Social Construction of Ethnic
Identity.” International Organization 54 (4): 845-77.
Fjelde, Hanne, and Nina von Uexkull. 2012. “Climate Triggers: Rainfall Anomalies,
Vulnerability and Communal Conflict in Sub-Saharan Africa.” Political Geography
31 (7): 444-53.
Fukuyama, Francis. 2011. The Origins of Political Order: From Prehuman Times to the
French Revolution. London, UK: Profile Books.
Gleditsch, Kristian Skrede, Julian Wucherpfennig, Simon Hug, and Karina Garnes Reigstad.
2011. “Polygyny or Misogyny? Reexamining the “First Law of Intergroup Conflict”.” The
Journal of Politics 73 (1): 265-70.
Goldstein, Joshua S. 2001. War and Gender: How Gender Shapes the War System and Vice
Versa. Cambridge, MA: Cambridge University Press.
Gurr, Ted Robert. 1970. Why Men Rebel. Princeton, NJ: Princeton University Press.
Henrich, Joseph, Robert Boyd, and Peter J Richerson. 2012. “The Puzzle of Monoga-
mous Marriage.” Philosophical Transactions of the Royal Society 367 (1589):
657-69.
Koos and Neupert-Wentz 429

Holzinger, Katharina, Roos Haer, Axel Bayer, Daniela M. Behr, and Clara Neupert-Wentz.
2018. “The Constitutionalization of Indigenous Group Rights, Traditional Political Institu-
tions, and Customary Law.” Comparative Political Studies 52 (12): 1775-1809. doi:10.1177/
0010414018774347.
Holzinger, Katharina, Florian G Kern, and Daniela Kromrey. 2016. “The Dualism of Con-
temporary Traditional Governance and the State: Institutional Setups and Political Con-
sequences.” Political Research Quarterly 69 (3): 469-81.
Homer-Dixon, Thomas F. 1999. Environment, Scarcity, and Violence. Princeton, NJ: Prince-
ton University Press.
Hudson, Valerie M. 2018. “The Deep Structure of Collective Security: Thoughts on McDer-
mott, Smuts, and Sanday.” In The Evils of Polygyny: Evidence of Its Harm to Women, Men,
and Society, edited by Rose McDermott and Kristen Renwick Monroe, 147-63. Ithaca,
NY: Cornell University Press.
Hudson, Valerie M., Mary Caprioli, Bonnie Ballif-Spanvill, Rose McDermott, and Chad F.
Emmett. 2010. “The Heart of the Matter: The Security of Women and the Security of
States.” International Security 33 (3): 7-45.
Hudson, Valerie M., and Andrea Den Boer. 2002. “A Surplus of Men, a Deficit of Peace:
Security and Sex Ratios in Asia’s Largest States.” International Security 26 (4): 5-38.
Hudson, Valerie M., and Hilary Matfess. 2017. “In Plain Sight: The Neglected Linkage
between Brideprice and Violent Conflict.” International Security 42 (1): 7-40.
Irons, William. 1983. “Human Female Reproductive Strategies.” In Social Behavior of
Female Vertebrates, edited by Samuel Wasser, 169-213. New York: Academic Press.
Jann, Ben. 2005. “Making Regression Tables from Stored Estimates.” Stata Journal 5 (3):
288-308.
Jann, Ben. 2007. “Making Regression Tables Simplified.” Stata Journal 7 (2): 277-44.
Kanazawa, Satoshi. 2009. “Evolutionary Psychological Foundations of Civil Wars.” The
Journal of Politics 71 (1): 25-34.
Kanazawa, Satoshi, and Mary C. Still. 1999. “Why Monogamy?” Social Forces 78 (1): 25-50.
King, Gary, Robert O Keohane, and Sidney Verba. 1994. Designing Social Inquiry: Scientific
Inference in Qualitative Research. Princeton, NJ: Princeton University Press.
Levinson, David. 1996. Encyclopedia of World Cultures: Africa and the Middle East. Boston,
MA: G.K. Hall.
McDermott, Rose. 2018. “The Meaning and Meanness of Polygyny.” In The Evils of Poly-
gyny: Evidence of Its Harm to Women, Men, and Society, edited by Rose McDermott and
Kristen Renwick Monroe, 8-32. Ithaca, NY: Cornell University Press.
McDermott, Rose, and Jonathan Cowden. 2018. “Polygyny and Violence against Women.” In
The Evils of Polygyny: Evidence of Its Harm to Women, Men, and Society, edited by Rose
McDermott and Kristen Renwick Monroe, 52-96. Ithaca, NY: Cornell University Press.
Mealey, Linda. 1985. “The Relationship Between Social Status and Biological Success: A
Case Study of the Mormon Religious Hierarchy.” Ethology and Sociobiology 6 (4):
249-57.
Melander, Erik. 2005. “Gender Equality and Intrastate Armed Conflict.” International Studies
Quarterly 49 (4): 695-714.
430 Journal of Conflict Resolution 64(2-3)

Mesquida, Christian G., and Neil I. Wiener. 1999. “Male Age Composition and Severity of
Conflicts.” Politics and the Life Sciences 18 (02): 181-89.
Michalopoulos, Stelios, and Elias Papaioannou. 2013. “Pre-colonial Ethnic Institutions and
Contemporary African Development.” Econometrica 81 (1): 113-52.
Michalopoulos, Stelios, and Elias Papaioannou. 2016. “The Long-run Effects of the Scramble
for Africa.” The American Economic Review 106 (7): 1802-48.
Mokuwa, Esther, Maarten Voors, Erwin Bulte, and Paul Richards. 2011. “Peasant Grievance
and Insurgency in Sierra Leone: Judicial Serfdom as a Driver of Conflict.” African Affairs
110 (440): 339-66.
Morgan, Stephan, and Christopher Winship. 2014. Counterfactuals and Causal Inference:
Methods and Principles for Social Research, 2nd ed. New York: Cambridge University
Press.
Morgenthau, Hans J. (1948) 1985. Politics Among Nations: The Struggle for Power and
Peace, 6th ed. revised by Kenneth W. Thompson. New York: Alfred A. Knopf.
Murdock, George Peter. 1949. Social Structure. New York: Macmillan.
Murdock, George Peter. 1959. Africa: Its Peoples and Their Culture History, 1st ed. New
York: McGraw-Hill, US.
Murdock, George Peter. 1969. Ethnographic Atlas. Pittsburgh, PA: University of Pittsburgh
Press.
Nunn, Nathan. 2008. “The Long-term Effects of Africa’s Slave Trades.” The Quarterly
Journal of Economics 123 (1): 139-76.
Nunn, Nathan, and Leonard Wantchekon. 2011. “The Slave Trade and the Origins of Mistrust
in Africa.” The American Economic Review 101 (7): 3221-52.
Pearl, Judea. 2009. Causality: Models, Reasoning, and Inference, 2nd ed. Cambridge, MA:
Cambridge University Press.
Pew Research Center. 2011. “The Future of the Global Muslim Population.” http://www.
pewforum.org/2011/01/27/the-future-of-the-global-muslim-population/ (accessed 19 June
2019).
Pierson, Paul. 2000. “Increasing Returns, Path Dependence, and the Study of Politics.” Amer-
ican Political Science Review 94 (2): 251-67.
Raffield, Ben, Neil Price, and Mark Collard. 2017. “Male-biased Operational Sex Ratios and
the Viking Phenomenon: An Evolutionary Anthropological Perspective on Late Iron Age
Scandinavian Raiding.” Evolution and Human Behavior 38 (3): 315-24.
Raleigh, Clionadh. 2010. “Political Marginalization, Climate Change, and Conflict in African
Sahel States.” International Studies Review 12 (1): 69-86.
Raleigh, Clionadh, Andrew Linke, Håvard Hegre, and Joakim Karlsen. 2010. “Introducing
ACLED: An Armed Conflict Location and Event Dataset.” Journal of Peace Research 47
(5): 651-60.
Rauch, Jonathan. 2006. “One Man, Many Wives, Big Problems.” The Atlantic, April. https://
www.theatlantic.com/magazine/archive/2006/04/one-man-many-wives-big-problems/3
04829/ (accessed 19 June 2019).
Shanas, Ethel. 1979. “The Family as a Social Support System in Old Age.” The Gerontologist
19 (2): 169-74.
Koos and Neupert-Wentz 431

Sundberg, Ralph, and Erik Melander. 2013. “Introducing the UCDP Georeferenced Event
Dataset.” Journal of Peace Research 50 (4): 523-32.
Tajima, Yuhki. 2013. “The Institutional Basis of Intercommunal Order: Evidence from Indo-
nesia’s Democratic Transition.” American Journal of Political Science 57 (1): 104-19.
Varshney, Ashutosh. 2003. Ethnic Conflict and Civic Life: Hindus and Muslims in India. New
Haven, CT: Yale University Press.
Verweijen, Judith. 2017. “Conflict Cows in the Congo?” Political Violence at a Glance.
http://politicalviolenceataglance.org/2017/02/16/judith/ (accessed 19 June 2019).
Weber, Max. (1922) 1980. Wirtschaft und Gesellschaft: Grundriß der verstehenden Soziolo-
gie, 5th ed. Tübingen: Mohr Siebeck.
Weidmann, Nils B. 2016. “A Closer Look at Reporting Bias in Conflict Event Data.” Amer-
ican Journal of Political Science 60 (1): 206-18.
White, Douglas R., and Michael L. Burton. 1988. “Causes of Polygyny: Ecology, Economy,
Kinship, and Warfare.” American Anthropologist 90 (4): 871-87.
Wig, Tore. 2016. “Peace from the Past: Pre-colonial Political Institutions and Civil Wars in
Africa.” Journal of Peace Research 53 (4): 509-24.
Wilson, Margo, and Martin Daly. 1985. “Competitiveness, Risk Taking, and Violence: The
Young Male Syndrome.” Ethology and Sociobiology 6 (1): 59-73.
Yair, Omer, and Dan Miodownik. 2016. “Youth Bulge and Civil War: Why a Country’s Share
of Young Adults Explains only Non-ethnic Wars.” Conflict Management and Peace
Science 33 (1): 25-44.