The Political Salience of Cultural Difference: Why Chewas and Tumbukas Are Allies in
Zambia and Adversaries in Malawi
Author(s): Daniel N. Posner
Source: The American Political Science Review , Nov., 2004, Vol. 98, No. 4 (Nov., 2004),
pp. 529-545
Published by: American Political Science Association
Stable URL: https://www.jstor.org/stable/4145323
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
American Political Science Association is collaborating with JSTOR to digitize, preserve and
extend access to The American Political Science Review
This content downloaded from
86.49.253.47 on Thu, 23 Sep 2021 20:15:09 UTC
All use subject to https://about.jstor.org/terms
American Political Science Review Vol. 98, No. 4 November 2004
The Political Salience of Cultural Difference:
Tumbukas Are Allies in Zambia and Adversari
DANIEL N. POSNER University of California, Los Angele
his paper explores the conditions under which cultural cleavages be
does so by taking advantage of the natural experiment afforded b
and Tumbuka peoples by the border between Zambia and Malawi.
objective cultural differences between Chewas and Tumbukas on both s
the political salience of the division between these communities is altogeth
difference stems from the different sizes of the Chewa and Tumbuka com
to each country's national political arena. In Malawi, Chewas and Tumb
a-vis the country as a whole and, thus, serve as viable bases for politic
Chewas and Tumbukas are small relative to the country as a whole and
bases of political support. The analysis suggests that the political salien
not on the nature of the cleavage itself (since it is identical in both countrie
it defines and whether or not they will be useful vehicles for political com
ultural differences are claimed to be at the
root of many of the world's conflicts, both
within states (Gurr 2000; Horowitz 1985, 2001;
Lake and Rothchild 1998) and among them
(Huntington 1996). Yet the mere presence of cultural
differences cannot possibly be a sufficient condition
for the emergence of political or social strife, for there
are far more cultural cleavages in the world than there
are conflicts. A reasonably competent ethnographer,
dispatched to almost any country in the world, could
probably enumerate dozens of differences among the
peoples he or she was sent to study. These might include
the color of their skin, the religions they practice, the
dialects they speak, the places from which they migrated,
the foods they eat, the clothes they wear, and
the marriage rituals they practice, among many other
attributes. Yet the vast majority of these differences are
likely to play no role as axes of political competition
or social conflict. Members of the society in question
might readily acknowledge these differences if asked
about them, but they will almost certainly view only a
tiny fraction of them as relevant bases of social identification
or political division.' This presents a puzzle:
Why do some cultural differences matter for p
and others not?
To pose this question is not, of course, to deny that
political differences sometimes do follow cultural lines.
In settings as diverse as Belgium, Sri Lanka, Sudan,
Cyprus, and New York City, political cleavages and
cultural differences correspond closely. Indeed, it is
frequently argued that the political divisions in these
places are made deeper precisely because they hew
to ethnic boundaries as much as they do (Horowitz
1985; Lijphart 1977). But for every cultural cleavage
that serves as a basis of political division there are numerous
others that have no political import at all. The
cultural division between Catholics and Protestants in
Northern Ireland may be highly salient, but not the one
between Presbyterians and Episcopalians. The cleavage
between citizens of Ile-Ife and citizens of Oyo may
be central to the politics of southern Nigeria, but not
the cleavage between Catholics and Muslims (Laitin
1986). Even Rwanda, a country usually assumed to
be divided by a single cleavage between Hutus and
Tutsis, also contains other bases of social division in
its midst, such as that between Hutus of the south and
northwest and, among Tutsis, between members of the
repatriated diaspora and those who have been longtime
Rwandan residents (Straus 2004). In principle,
any of these cultural cleavages might have emerged as
the central axes of political division in these countries.
Why, then, did some of them become politically salient
while the others remained politically irrelevant?
This paper proposes a simple answer to this question.
I argue that the political salience of a cultural
cleavage will depend on the sizes of the groups that
it defines relative to the size of the arena in which
political competition is taking place. If the cultural
cleavage defines groups that are large enough to constitute
viable coalitions in the competition for political
power, then politicians will mobilize these groups and
the cleavage that divides them will become politically
Daniel N. Posner is Assistant Professor, Department of Political
Science, Box 951472, University of California, Los Angeles, Los
Angeles, CA 90095-1472 (dposner@polisci.ucla.edu).
The author thanks Lovemore Hauya, Nahomi Ichino, Joseph
Musukwa, Grace Kamusikili, Charles Kasambala, Mzee Nkhwachi
Nyirongo, and Ian Shaba for their research assistance; Kimberly
Smiddy for her hospitality in Lilongwe; and Carles Boix, Wiseman
Chirwa, Matthew Kocher, Michael Thies, Matthew Turner, James
Vreeland, Jennifer Widner, and two anonymous reviewers for their
comments. I also received helpful feedback from participants at the
University of Chicago Comparative Politics Workshop, the Working
Group in African Political Economy, the Laboratory in Comparative
Ethnic Processes, and the Hoover Institution U.S. and World Affairs
Seminar. I am grateful to the Center for Social Research in Zomba,
Malawi and to the Institute for Economic and Social Research in
Lusaka, Zambia for facilitating this research, and to the James
S. Coleman African Studies Center at UCLA and the UCLA International
Institute for their financial support. All errors are my
own.
1 Fearon and Laitin (1996) provide the striking estimate that, in
Africa, there has only been one instance of ethnic violence for every
2,000 cases that would have been predicted on the basis
differences alone.
529
This content downloaded from
86.49.253.47 on Thu, 23 Sep 2021 20:15:09 UTC
All use subject to https://about.jstor.org/terms
Political Salience of Cultural Difference November 2004
salient. If the cultural cleavage defines groups that are
too small to serve as viable bases of political support,
then these groups will go unmobilized and the cleavage
that separates them will remain politically irrelevant.
The cultural differences between the groups will still
exist, but there will be no political importance attached
to them.
The idea that the political salience of a cultural cleavage
lies in nothing more than the sizes of the groups
that it defines contrasts with traditional approaches
to the question of why some cleavages matter rather
than others. For example, it is often assumed that what
matters for the salience of a cleavage is the degree of
cultural differences between the groups (e.g., Caselli
and Coleman 2002).2 By this logic, a cleavage between
Christians and Muslims would be expected to be more
salient than one between Anglicans and Lutherans. A
slightly different approach emphasizes not the degree
of difference between the groups but the nature of the
difference. By this logic, certain kinds of social cleavages-for
example, those based on race-are simply
assumed to have more power than others in generating
salient social and political divisions.3 A third approach,
exemplified in the classic work of Lipset and Rokkan
(1967), emphasizes the historical emergence of social
cleavages over time and the progressive displacement
of one by the next. By this logic, the salience of one
cleavage over another is a function of the particular
stage of historical development in which the political
system happens to be located at the time. Yet a fourth
approach emphasizes how the experience of colonialism
led to the reification of some social cleavages over
others (e.g., Young 1994). For example, Laitin's (1986)
investigation into the nonpoliticization of religious divisions
in Yorubaland shows how the experience of
British colonialism endowed social identities revolving
around connections with one's ancestral city-state with
hegemonic status vis-A-vis other kinds of identities. In
his account, it is the marriage between the local legitimacy
of city-state elites and the power of the colonial
state that makes the city-state identity category
salient.
In all of these approaches, either the specific characteristics
of the groups in question, the history through
which they were constructed, or the particular kind of
cleavage that divides them is central to the explanation
for why the cleavage comes to matter politically. In the
structural approach advanced in this paper, predictions
about which axis of social division will emerge as politically
relevant depend on none of these factors. The
origins of the cleavage (emphasized by constructivists)
and its cultural content (emphasized by primordialists)
are irrelevant. All that matters is cultural demography:
the sizes of the groups that the cleavage defines relative
to the political and social arenas in which they are operating.
Let me stress that my goal is not to suggest that
group size is always the most relevant factor. Rather it
is to demonstrate the plausibility of this spare and powerful
hypothesis by drawing on a natural experiment
that permits us to rule out competing explanations for
why one cultural cleavage came to emerge as the axis of
political competition and conflict rather than another.
A NATURAL EXPERIMENT
The natural experiment I draw upon is made possible
by the division of the Chewa and Tumbuka peoples by
the border between the African countries of Zambia
and Malawi. The Zambia-Malawi boundary follows
the watershed of a set of low hills that run from
southwestern Tanzania to northwestern Mozambique,
roughly parallel to the shoreline of Lake Malawi. The
boundary was originally demarcated by the British
South African Company in 1891 to distinguish the
territories of what were then called Northeastern and
Northwestern Rhodesia.4 Like many African borders,
the one that separates Zambia and Malawi was drawn
purely for administrative purposes, with no attention
to the distribution of groups on the ground. As a consequence,
it partitions the Chewa and Tumbuka peoples,
leaving about two-thirds of each group on the eastern
side of the border in Malawi and about one-fourth
of the Chewas and one-third of the Tumbukas on the
western side in Zambia.5
The division of the Chewa and Tumbuka communities
by the Zambia-Malawi border provides a
laboratory-like opportunity for comparing the salience
of an identical cultural cleavage in different settings
In taking advantage of this natural experiment, I join a
number of other researchers (e.g., Asiwaju 1985; Laiti
1986; Miles 1994; Miles and Rochefort 1991) who have
also exploited the partitioning of ethnic groups by na
tional boundaries. The difference between my study
and theirs is that mine emphasizes the effects of th
partition on intergroup relations. For these other au
thors, the unit of analysis is the ethnic group, and the
partitioning of the group by the national boundary i
employed to study how members of the same commu
nity respond to different social and political environ
ments. In this study, the unit of analysis is not the ethnic
group but the cultural dyad. Rather than explore how
members of the same community respond to different
national "treatments," I explore how the relations
between a pair of groups vary as a consequence o
the different environments in which those relations are
taking place.
It turns out that the relations between Chewas
and Tumbukas in each country are quite different
In Malawi, interactions between members of the tw
2 This assumption is the motivation for Fearon's (2003) index of
cultural diversity, which distinguishes among groups based on the
structural similarities in the languages spoken by their members.
3 For a response to the position that racial divisions are somehow
different from other kinds of cultural cleavages, see Horowitz (1985,
42-46).
4 Nothwestern Rhodesia later became Northern Rhodesia and
Northeastern Rhodesia later became Nyasaland. After independence
in 1964, these territories became Zambia and Malawi, re-
spectively.
5 Figures are from Grimes and Grimes 1996. The Chewa population
does not sum to 100% because the Chewa community extends into
neighboring areas of Mozambique and Zimbabwe.
530
This content downloaded from
86.49.253.47 on Thu, 23 Sep 2021 20:15:09 UTC
All use subject to https://about.jstor.org/terms
American Political Science Review Vol. 98, No. 4
communities are often antagonistic (Chirwa 1998).
Members of the two groups tend to see each another
as political adversaries. Since the advent of competitive
multiparty elections in 1994, each group has been associated
with its own political party and voters from each
community have rarely crossed party lines (Kaspin
1995; Posner 1995). Although the hostility between
Chewas and Tumbukas in Malawi has never generated
outright violence, informed observers have suggested
that it easily could (e.g., Kaspin 1995; Vail and White
1989). In Zambia, by contrast, Chewas and Tumbukas
tend to view each other as ethnic brethren and political
allies. Since independence, they have overwhelmingly
supported the same political party. The prospect of
violence between members of these groups is almost
unthinkable. When asked what makes Chewas and
Tumbukas different from one another, most Zambian
members of these communities will tell you, "We are
the same."
The border that separates Zambia and Malawi
thus does more than simply partition the Chewa and
Tumbuka peoples. It also demarcates two completely
different zones of intercultural relations. Whereas the
cultural cleavage between Chewas and Tumbukas is
highly politically salient in Malawi, it has almost no political
salience in Zambia. The objective cultural differences
between members of these two communities may
be the same on both sides of the border, but the salience
of the cultural cleavage that divides them is different.
What explains this striking cross-border contrast? Why
are Chewas and Tumbukas allies in Zambia and adversaries
in Malawi? Why is the cultural cleavage between
these two communities so much more politically and
socially important in one country than the other?
CHEWA-TUMBUKA RELATIONS
IN ZAMBIA AND MALAWI
Experts on Zambian and Malawian politics would
probably be willing to stipulate that relations between
Chewas and Tumbukas are more amicable in Zambia
than in Malawi. The secondary source literature on
these countries is full of references to the political and
social divisions between these groups in Malawi but
entirely silent on the nature of their relations in Zambia
(which one would have to take as evidence that cleavage
between them is not important).6 Nevertheless,
to confirm that Chewa-Tumbuka divisions are in fact
deeper in Malawi than in Zambia, I collected original
data in the border area of both countries during July
and August 2001.
Research Design
To document the degree of social distance between
Chewas and Tumbukas in each country, I administered
a questionnaire in four villages: two Chewa
villages (Mkanda, Zambia, and Mkanda, Malawi),7 and
two Tumbuka villages (Mwase-Lundazi, Zambia, and
Kapopo, Malawi). Each pair of villages was located
directly across the border from each other. Their locations
are depicted in Figure 1.
I selected the four research villages with two criteria
in mind. The first was the homogeneity of their Chewa
and Tumbuka populations. In the midnineteenth century,
Ngoni invaders from South Africa settled in what
was to become the Zambia-Malawi border area and
became interspersed with the region's longstanding
Chewa and Tumbuka inhabitants. Because the Ngoni
practice was to marry local women, adopt local laws
and customs, and otherwise incorporate themselves
into the communities they conquered, the Ngoni invasion
did little to alter the preexisting Chewa and
Tumbuka cultures (Barnes 1954; Tew 1950). The presence
of Ngonis in any of the four research villages
therefore probably would not have affected the analysis.
Nonetheless, to guard against the possibility that it
might, I specifically selected villages that were homogeneously
Chewa or Tumbuka. To prevent the possible
contamination of villages by exposure to Ngonis living
nearby, I also chose each Chewa and Tumbuka village
pair so as to be equidistant from the nearest significant
concentration of Ngonis.
My second criterion was similarity across each
Chewa and Tumbuka village pair. Given that the purpose
of the exercise was to be able to rule out competing
explanations for whatever variation in intergroup
relations that I might find, it was important to control
for as many potentially confounding factors as possible.
I therefore choose pairs of villages that were very close
together, on the logic that this would provide a natural
control for geographic and ecological factors that
might affect villagers' welfare or modes of agricultural
production and, thus, potentially their attitudes toward
outgroup members. Mkanda (Zambia) and Mkanda
(Malawi) are within 15 kilometers of each other and occupy
nearly identical ecological niches. The same is true
of Mwase-Lundazi and Kapopo. Indeed, both pairs of
villages are so close to each other that several respondents
reported regularly visiting friends and relatives
across the border in the other village. Agricultural production
in all four villages is also very similar, revolving
around the production of maize and groundnuts for
local consumption and small amounts of tobacco for
domestic and international markets.
I was also careful to select villages that were equally
exposed to national political affairs in their respective
countries, as well as to domestic markets, and to
members of the other cultural community. I proxied
for exposure to politics and markets by distance from
paved roads (which was approximately eight kilometers
in all cases) and for exposure to members of the
other community by distance from the nearest Chewa
or Tumbuka village (which was equal in each pair).
The only meaningful difference between the village
6 In addition to the sources cited above, see also Bates 1976, Dresang
1974, Gertzel et al. 1984, Kishindo 1998, and Vail 1981.
7 Both villages take their name from Chief Mkanda, who lives in
Malawi but exercises his traditional authority on both sides of the
border.
531
This content downloaded from
86.49.253.47 on Thu, 23 Sep 2021 20:15:09 UTC
All use subject to https://about.jstor.org/terms
Political Salience of Cultural Difference November 2004
FIGURE 1. Research Sites
Mwase Lundazi, Zambia Kapopo, Malawi
Mkanda, Zambia Mkanda, Malawi
pairs was the fact that Mkanda (Zambia) and MwaseLundazi
were on the Zambian side of the border and
Mkanda (Malawi) and Kapopo were on the Malawian
side.
I began my work at each of the four research sites
by visiting the local chief to present my credentials
and request his permission to carry out interviews in
the surrounding area. Having granted his permission,
the chief would then provide a messenger to accompany
my research team to the neighboring subvillages,
where we would repeat this process with the local
headmen before actually beginning the survey work.
Though tedious and time-consuming, such formalities
are essential for village-level work of the sort we were
undertaking. The time it took for introductions to
be made and permissions to be granted allowed for
news of our business in the area-and, critically, word
that this business had been approved by the chief-to
filter through the community. This was indispensable
for securing the willing cooperation of our respondents.
It was also particularly important given that the
survey we were administering required respondents
to make candid statements about potentially sensitive
subjects.
We interviewed 42 respondents in each of the two
Zambian villages and 48 respondents in each of the
two Malawian villages, for a total sample of 180. We
selected respondents through a random stratified quota
sampling procedure (with stratification by gender and
age) from every third unrelated household.8 The surveys
were conducted in the respondent's local language
8 Age categories were 18-26, 26-44, and 45+ years. Each category
contains approximately one-third of the voting-age population.
532
This content downloaded from
86.49.253.47 on Thu, 23 Sep 2021 20:15:09 UTC
All use subject to https://about.jstor.org/terms
American Political Science Review Vol. 98, No. 4
(either Chichewa or Chitumbuka). We asked five q
tions:
* What, if anything, makes Chewas different fr
Tumbukas?
* Do you think Chewas (Tumbukas) here would vote
for a Tumbuka (Chewa) if he were standing for pres-
ident?
* What about you? Would you vote for a Chewa
(Tumbuka) if he were standing for president?
* Are you married?
[If married] Are you married to a Chewa
(Tumbuka)? [If no] Would you have considered marrying
a Chewa (Tumbuka)?
[If not married] Would you marry a Chewa
(Tumbuka)?
* In general, how do people here feel about Chewas
marrying Tumbukas?
In addition to the formal surveys, I also conducted focus
groups in each of the four survey sites to gather further
information about the characteristics of each village
and the history of Chewa-Tumbuka relations in the
area.
The Differing Salience of an Identical
Cultural Cleavage
The purpose of the first open-ended survey question
was to confirm that the objective cultural differences
between Chewas and Tumbukas-or, more accurately,
respondents' perceptions of these differences-were
the same in both countries. This was necessary to establish
to rule out the possibility that the difference in
the salience of the Chewa-Tumbuka cleavage was a
product of greater objective differences between these
groups on one side of the border than on the other.
Again and again, respondents in all four villages
identified the same attributes as distinguishing members
of the two communities. Many began by pointing
to differences in the two groups' languages and
traditional dances. Others pointed to the communities'
different norms regarding lobola (brideprice) payments:
Whereas Tumbuka parents customarily demand
payment of seven cows from their daughters' suitors,
Chewa parents require only a chicken. Several respondents
pointed out that Tumbukas are patrilocal,
whereas Chewas are matrilocal. Others mentioned that
Tumbukas insist on the ritual cleansing of the wife after
the death of her husband, while Chewas do not. One
respondent noted that Tumbukas use long hoes and
Chewas use short hoes. Another noted that whereas
Tumbukas do not make ridges when planting groundnuts,
Chewas do.
While the specific cultural practices that respondents
mentioned varied somewhat from person to person,
nearly every interviewee on both sides of the border
(83% overall) was able to identify at least one attribute,
tradition, or custom-and often several-that
made Chewas and Tumbukas culturally different from
one another. On average, respondents mentioned 1.36
differences (SD = 0.97), although 23% of Zambian
respondents and 15% of Malawian respondents mentioned
none. Among those who mentioned at least one
difference, the average number of differences mentioned
was nearly identical on both sides of the border:
1.66 for Zambians (SD = 1.0) and 1.70 for Malawians
(SD = 0.91).
Zambian and Malawian respondents did differ in one
important respect. Malawians occasionally supplemented
their list of objective differences with negative
statements about members of the other community.
For example, several Malawian Chewas volunteered
that, in addition to having different cultural practices,
Tumbukas are odzikonda (literally, they "love
themselves"). Others referred to Tumbukas as nepotistic,
selfish, or boastful. One Chewa summarized the
relations between the two communities by saying
that, between the groups, there is mpatukano (loosely,
"great division"; literally, "going different ways").
Some Malawian Tumbukas also had negative things to
say about Chewas. More than one Malawian Tumbuka
respondent characterized Chewas as "crooks" or as
being "too clever." Others referred to them as "lazy."
While the number of Malawian respondents who volunteered
such negative characterizations was fairly
small (just 13 of the 93 who answered the question), this
stood in marked contrast to the Zambian respondents,
only two (of 83) of whom supplemented their objective
characterizations of Chewa-Tumbuka differences
with negative comments about members of the other
group. In fact, the modal Zambian response to our
request for a list of differences between Chewas and
Tumbukas was to emphasize that "We are one" or "We
are united."
The different views held by Malawians and
Zambians about members of the other group were
demonstrated even more clearly in the responses to
the next four questions. The first asked respondents
whether they thought that people in their village
would be willing to vote for a member of the other
group if such a person were running for President.
Whereas only 21% of Zambians said they thought
that others would be unwilling to support a candidate
from the other community, fully 61% of Malawians
said they thought their fellow villagers would be
unwilling to vote across group lines. When asked
what they themselves would do (a formulation that
usually leads to more inhibited answers and lower
professed levels of intergroup antipathy), 45% of
Malawians still insisted that they would not support
a member of the other group, compared to just 24% of
Zambians.
These response patterns carried over from the political
sphere into the realm of private social relations.
In response to the question about whether they
would marry a member of the other group, just 24%
of Zambians said they would not, compared to 55% of
Malawians.9 When asked more generally how people
9 Given the ethnically homogeneous nature of the survey villages,
individuals would have needed to go to some lengths even to find a
spouse from outside of their own group. Hence, the vast majority of
respondents were answering the intermarriage question in terms of
533
This content downloaded from
86.49.253.47 on Thu, 23 Sep 2021 20:15:09 UTC
All use subject to https://about.jstor.org/terms
Political Salience of Cultural Difference November 2004
FIGURE 2. Chewa-Tumbuka Relations in Zamb
70
60.9
60
55.2
50
45.3
40 - 37.6
30
23.8 23.8
21.4
20
14.0
2.4 .
percent
stateme
group when asked what, if Presidential candidate from candidate from the other other group is frowned upon
anything, makes Chewas the other group group
different from Tumbukas
SZambia U Malawi
in their community felt about members of one group
marrying members of the other, only 6% of Zambians
said it was frowned upon, whereas 38% of Malawians
said so. Note that these differences cannot be attributed
to the possibility that respondents on one side of the
border had greater exposure to members of the other
group than respondents on the other side: Both Chewa
villages were equidistant from the nearest concentration
of Tumbukas, and both Tumbuka villages were
equidistant from the nearest concentration of Chewas.
Also, all villages were located equally far from paved
roads that might have promoted exposure to outsiders
or facilitated out-migration by those who were more
open to marrying outside the group.
The gulf between the Zambian and the Malawian
respondents on all of these questions is made clear
in Figure 2. For all four questions, the observed difference
between the Zambian and the Malawian response
patterns is significant at p = .01 in a two-tailed test.
These differences are all the more remarkable when
we remember that the proximity of the village pairs
(less than 15 kilometers in both cases), the porosity
of the border, and the peripheral location of the four
research sites from the center of each country's political
affairs almost certainly generated a bias against finding
a difference between the Zambian and the Malawian
respondents. If anything, the findings must be viewed
as a lower bound on the true difference in the salience
of the Chewa-Tumbuka cleavage in each country.
In Table 1, I revisit these results in a series of logit
regressions in which I control for respondents' tribal
affiliations, gender, and age and the number of cultural
differences they mentioned in response to the openended
question about Chewa-Tumbuka differences. I
include this last variable to control for the possibility
that the difference in the salience of the ChewaTumbuka
cleavage in Zambia and Malawi might be
a product of different perceptions about the objective
differences between these communities in each
country. If the degree of perceived cultural difference
between Chewas and Tumbukas matters in shaping intergroup
relations, then we might expect respondents
who mentioned more differences to be more hostile
to members of the other community. This would be
reflected in a significant positive coefficient on the
Number of Differences variable.
The analysis yields several interesting results. First,
the coefficient on the Number of Differences variable
is insignificant in all four regressions, signaling that
the perceived degree of cultural difference between
the two communities has no effect on respondents'
attitudes toward members of the other group. Second,
in both countries, Tumbuka respondents demonstrate
more hostility to Chewas than Chewas do to Tumbukas,
particularly when it comes to the issue of intermarriage.
Third, women are significantly less willing to marry
the hypothetical (i.e., whether they would have considered marrying
a member of the other community). Only 7.1% of Zambian respondents
(6 of 84) and 5.2% of Malawian respondents (5 of 96) reported
that they actually were married to a person from the other ethnic
group.
534
This content downloaded from
86.49.253.47 on Thu, 23 Sep 2021 20:15:09 UTC
All use subject to https://about.jstor.org/terms
American Political Science Review Vol. 98, No. 4
TABLE 1. The Determinants of Chewa-Tumbuka Relations
Number of
Dependent Variable Countrya Tribeb Genderc Age Differences Constant
Believes others in the area would not 1.98*** 0.77* -0.60 0.31 0.07 -1.92***
vote for a presidential candidate (0.370) (0.360) (0.360) (0.219) (0.187) (0.510)
from the other group
Say they would not vote for a 1.16** 0.91 ** -0.78* 0.04 -0.07 -1.33***
presidential candidate from the (0.353) (0.348) (0.349) (0.208) (0.190) (0.478)
other group
Say they would not have married 1.89*** 2.05*** -1.57*** 0.16 0.07 -2.11***
(have considered marrying) a (0.410) (0.416) (0.405) (0.231) (0.208) (0.557)
member of the other group
Say that, in general, marriage to a 2.43*** 0.86* -0.91* 0.37 -0.03 -3.24***
person from the other group is (0.533) (0.428) (0.427) (0.255) (0.238) (0.718)
frowned upon
Note: Standard errors in parentheses. Ns= 172, 175, 176, 172. * p= .05; ** p= .01; *** p= .001.
a Country coded 0 for Zambia, 1 for Malawi.
b Tribe coded 0 for Chewa, 1 for Tumbuka.
c Gender coded 0 for female, 1 for male.
outside their group than men. This is perhaps not surprising
given the generally low status of women in both
Chewa and Tumbuka societies and the preference of
women to maintain links to ethnic kin who might offer
them protection from abusive husbands or in-laws.
The most important finding, however, is that, controlling
for all of these other factors, the most powerful
determinant of a respondent's attitudes toward
members of the other community is his or her physical-and,
it bears underscoring, from an historical
perspective, entirely accidental-location on one side
of the Zambia-Malawi border or the other. Malawian
respondents are significantly less willing to vote for a
presidential candidate from the other group and significantly
less willing to marry across group lines than are
their otherwise identical Zambian counterparts. These
findings accord with the received wisdom from the secondary
source literature: Both in the political realm and
in the sphere of social relations, the divisions between
Chewas and Tumbukas run much deeper in Malawi
than in Zambia. The objective cultural cleavage that
separates members of one community from the other
may be the same, but the salience of that cleavage as
a marker of political and social division is altogether
different.
WHY CHEWAS AND TUMBUKAS ARE
ALLIES IN ZAMBIA AND ADVERSARIES
IN MALAWI
Why are Chewa-Tumbuka relations so different in the
Zambian and Malawian survey villages? The literature
on intergroup relations offers a number of potential
hypotheses, nearly all of which can be ruled out on the
basis of research design.
Modernization explanations (e.g., Deutsch 1961;
Huntington 1968) provide no leverage in accounting
for the differing salience of the Chewa-Tumbuka
cleavage, since there are no discernible differences in
the level of development of Chewa and Tumbuka respondents
in each country, no differences in resource
scarcity, and, by design, no differences in the levels of
the respondents' exposure to national political affairs
across the village pairs. Nor does the so-called "contact
hypothesis" (Forbes 1997) offer an explanation, since
the Chewa and Tumbuka villages on each side of the
border were selected so that Zambian and Malawian
respondents would have similar levels of exposure to
people from the other ethnic group.
The effects on ethnic relations of political institutions
such as electoral rules and federalism have received
much attention in recent years (e.g., Horowitz 1991;
Reilly 2001; Reynolds 1999). Yet their impact can also
be ruled out, since both Zambia and Malawi have identical
single-member plurality electoral rules and very
similar, highly centralized, candidate selection procedures.
They also have parallel histories of alternating
one-party and multiparty governance since independence
(which both countries achieved in 1964). Both
countries also share a unitary political system, with
very little effective power devolved to local authorities.
Colonial history also offers few bases of differentiation,
since Zambia (formerly Northern Rhodesia)
and Malawi (formerly Nyasaland) were both not only
British colonies but, between 1953 and 1963, also partner
states in the British-controlled Central African
Federation.
Another possible explanation for the different findings
in each country is that the timing of the survey
work might have caught the two countries at different
points in their respective electoral cycles and that this
might have predisposed respondents in one country
to have been more conscious of intergroup tensions
than their peers across the border. Elections tend to
be a time when cultural differences are exploited by
vote-seeking politicians, so proximity to an electoral
contest might plausibly bias the salience of ethnic divisions
upward.'0 If anything, however, the timing of the
10 Bannon, Miguel, and Posner (2004) find a strong positive relationship
between the proximity of an election and the salience of
ethnicity in an analysis of nine African countries.
535
This content downloaded from
86.49.253.47 on Thu, 23 Sep 2021 20:15:09 UTC
All use subject to https://about.jstor.org/terms
Political Salience of Cultural Difference November 2004
electoral calendar should have heightened the salience
of group divisions in Zambia rather than in Malawi. The
survey work for the project was conducted in July and
August of 2001, at a time when Zambia was preparing
for a general election to be held four months later,
in December of that year.1l Malawi, by contrast, had
most recently held a national election in 1999 and was
not scheduled to hold its next one until 2004. To the
extent that preelection ethnic mobilization affected the
survey results, it should have made Chewa-Tumbuka
relations more adversarial in Zambia than in Malawi.
The bias, if there is one, therefore runs against the
reported findings.
Two additional case-specific explanations also merit
consideration. General similarities in colonial history
aside, a well-known explanation for the animosity between
Chewas and Tumbukas on the Malawian side
of the border traces it to the impact of the Christian
missionary societies that settled in the central and
northern regions of the country during the late nineteenth
and early twentieth centuries (Vail 1981; Vail
and White 1989). There are two variants of this account.
The first of them stresses the different emphasis
placed on African education by the missionary societies
that came to dominate in each region. According
to this account, Tumbukas were fortunate to be settled
by missionaries associated with the Livingstonia
Mission, a branch of the Free Church of Scotland that
put great stress on education. Chewas, meanwhile, had
the misfortune of being settled by missionaries from the
White Fathers (a French Catholic group) and the South
African Dutch Reformed Church-societies that, for
reasons of ideology and racism, invested little energy
in promoting African literacy (Vail and White 1989).12
Thus, when the British colonial administration began
hiring African clerks and teachers to staff its growing
bureaucracy and schools, it focused its recruiting
among the Tumbuka, since they were the most qualified.
According to Vail (1981, 145), Tumbukas comprised
more than 50% of the highest-ranking Malawian
civil servants in 1969, despite constituting only 12% of
the Malawian population. The educationally disadvantaged
Chewas, by contrast, were comparatively underrepresented
in these positions. Not surprisingly, this
skewed distribution of civil service and teaching jobs,
along with the attempts made by the government after
independence to redress the imbalance by favoring
Chewa job applicants over Tumbukas, was a source of
conflict between the two communities. Thus one plausible
explanation for the high salience of the ChewaTumbuka
cleavage in present-day Malawi is that it is
a legacy of the lopsided educational endowments of
each community caused by their different missionary
experiences during the colonial period.
The problem with this explanation is that the same
missionary societies that settled among the Tumbukas
and Chewas on the Nyasaland (Malawian) side of the
border also set up outposts across the border among the
Tumbukas and Chewas in Northern Rhodesia (Henkel
1985). By 1925, the Free Church of Scotland was
running 313 mission schools in the eastern part of
Northern Rhodesia, including a large number in the
Tumbuka-speaking areas (Snelson 1974). In the same
year, the Dutch Reformed Church and White Fathers
were running nearly 750 schools between them, largely
in Chewa-speaking areas (Snelson 1974). Yet while
the Free Church of Scotland schools were committed
to promoting literacy, the White Fathers and Dutch
Reformed Church schools provided little more than
a "smattering of what passed as education" (Snelson
1974, 80). The disparity in the educational commitment
of these societies is reflected in the gap in the
educational attainment of Chewas and Tumbukas in
the areas in which these groups worked. Drawing on
data from the 1992 Zambia Demographic and Health
Survey (Republic of Zambia 1992), I estimate that the
average number of years of education in predominantly
Tumbuka districts was more than double that in predominantly
Chewa districts during the colonial era: 3.5
versus 1.6 years.13 This gap is almost identical to what
I estimate for the Tumbuka- and Chewa-dominated
districts across the border in Malawi.14 The differences
in the intergroup resentments among the Zambian and
Malawian respondents in my 2001 survey thus cannot
be attributed to differences in the colonial era educational
endowments of Tumbukas and Chewas on either
side of the border.
A slightly different explanation for the high salience
of the Chewa-Tumbuka cleavage in Malawi emphasizes
not the amount of missionary education that
Chewas or Tumbukas received during the colonial era
but its content. As Vail and White (1989) document,
the very idea of "Tumbukaness" was the creation of a
small group of missionary-educated intellectuals who,
in cooperation with their missionary teachers, sought
" Although the election date had not yet been announced at the
time of the fieldwork, the Zambian constitution required that the
election be held before the end of the year, so survey respondents
knew that it would take place within the next several months.
12 At Independence, primary school enrollment rates in the
Tumbuka-speaking districts of Karonga, Rumphi, and Mzimba were
more than double those in the Chewa-dominated districts of Mchinji,
Dowa, Salima, and Chikwakwa (Vail 1981, 144).
13 I restricted the analysis to men who were 45 years old or older at
the time of the survey, whose schooling therefore took place entirely
during the colonial period. The Tumbuka-dominated districts were
Lundazi and Chama (N = 46); the Chewa-dominated districts were
Chadiza and Katete (N = 44). All four districts are almost entirely
rural, so it is unlikely that the differences in average educational
attainment were caused by in-migration of better educated people
in search of employment opportunities. Indeed, the differences between
the Tumbuka and the Chewa districts are probably attenuated
by the tendency of the best-educated individuals to migrate out of
these districts to seek employment in one of the major towns.
14 I did, however, find Malawians (both Chewas and Tumbukas) to
have slightly more years of education than Zambians. In Tumbukadominated
districts of Malawi (Rumphi and Mzimba; N= 216), the
average number of years of schooling for men 45 years of age or
older was 4.7; in Chewa-dominated districts (Mchinji and Lilongwe;
N = 151), it was 2.4. Because Lilongwe and Mzimba districts contain
large numbers of urban respondents, the analysis includes rural respondents
from these districts only. Again, the tendency of highly
educated people to migrate out of these rural areas probably attenuates,
rather than increases, the cross-district variation. Data come
from Malawi's 1992 Demographic and Health Survey (Republic of
Malawi 1992).
536
This content downloaded from
86.49.253.47 on Thu, 23 Sep 2021 20:15:09 UTC
All use subject to https://about.jstor.org/terms
American Political Science Review Vol. 98, No. 4
to unify the northern tribes under a common ethni
banner to increase their political leverage with the colonial
administration. Once constructed for this purpos
Tumbuka ethnic consciousness was then propagated
through the curriculum of the mission schools. By this
process, a Tumbuka identity took hold in the norther
part of Malawi that had no parallel elsewhere in the
territory. It is plausible, then, that the high salience
of the Chewa-Tumbuka cleavage in Malawi might be
traced to the success of these early culture brokers in
creating a Tumbuka group consciousness that could b
deployed in opposition to other cultural communities
The comparative weakness of the Chewa-Tumbuk
cleavage in Zambia, meanwhile, could be attributed t
the absence in that country of a comparable group o
missionary-supported intellectuals who were capable
of spearheading a similar identity-building project o
that side of the border.
Although a constructivist explanation along these
lines cannot be ruled out entirely, it does a much
better job of accounting for the greater awareness of
Tumbuka ethnic identity on the Malawian side of the
border than for why differences between Chewas and
Tumbukas became a source of political division in one
country but not in the other. Viewing one's group as
a unified cultural entity may be a prerequisite for the
development of a politically salient cleavage between
one's own group and one's neighbor. But it in no way
guarantees that the cleavage between the two groups
will become salient. In this vein, it is useful to recall
that what differentiated the Chewa and Tumbuka survey
respondents in Zambia and Malawi was not their
awareness of the differences between their two communities
but whether or not they attached political or
social meaning to these differences. The possibility that
Malawian Tumbukas might, as a legacy of early twentieth
century missionary-promoted culture brokering,
be more self-conscious of their cultural identities than
their fellow Tumbukas across the border might lead us
to expect to find Malawian Tumbukas identifying more
differences than Zambian Tumbukas between themselves
and their respective Chewa countrymen.15 But
it provides no clear prediction about how they would
respond to questions about their willingness to vote for
or marry members of the other community. Nor does
it provide any explanation for why the cleavage would
develop between Tumbukas and Chewas, rather than
between Tumbukas and other cultural communities.
Indeed, if anything, Tumbuka identity was constructed
in opposition to the Ngonis rather than to the Chewas
(Vail and White 1989). So if colonial era identity construction
were the key, then we would have expected
the Tumbuka-Ngoni cleavage to have been more important
than the one between Tumbukas and Chewas.
A final potential explanation for the observed differences
in Chewa-Tumbuka relations across the two
countries is that ethnic relations per se-not just those
between Chewas and Tumbukas-are simply more
conflictual in Malawi than in Zambia. To the extent
that this is the case, the greater salience of the ChewaTumbuka
cleavage in Malawi could be viewed as an
artifact of that country's deeper level of interethnic
discord more generally. Indeed, there is much evidence
to suggest the plausibility of this hypothesis.
From 1964 until 1991, Zambia was ruled by Kenneth
Kaunda, a president noteworthy for his commitment to
nonracialism and national unity (Sklar 2003). Malawi,
meanwhile, was ruled from 1964 to 1994 by Hastings
Kamuzu Banda, a leader notorious for championing
his own Chewa culture, channeling investment and
development projects to his Chewa kin, and actively
discriminating against members of other ethnic communities-Tumbukas
in particular (Kaspin 1995; Vail
and White 1989). This is not to say that Kaunda did not
sometimes "play the ethnic card" or that Zambia under
his rule (or that of his successors, Fredrick Chiluba and
Levy Mwanawasa) has been a place where ethnicity
has played no role in politics. Kaunda was, at times,
a skillful manipulator of ethnic divisions (Bates 1976;
Dresang 1974; Gertzel et al. 1984), as have been his successors
(Posner, forthcoming). But it is nonetheless the
case that ethnic divisions have historically been more
central to Malawi's politics than to Zambia's and that
this might account for why Chewa-Tumbuka divisions
are deeper in the former than the latter.
The problem with this explanation is that it conflates
two quite different issues. The first is whether Presidents
Kaunda or Banda chose to divide their countries
along ethnic lines (they both did, albeit to different
degrees). The second is why, when Kaunda or Banda
did use ethnicity as a political tool, they chose to exploit
the particular lines of ethnic division that they did.
Whether Banda was more of a "divider" than Kaunda
is not the central issue. The relevant question is why
Banda chose to divide Chewas and Tumbukas, rather
than other groups, and why, to the extent that Kaunda
also engaged in ethnic political mobilization, he chose
to stress ethnic divisions other than those between
Chewas and Tumbukas. Simply arguing that Banda
exploited ethnicity more than Kaunda is not enough. In
order to explain the differing salience of the ChewaTumbuka
cleavage in Zambia and Malawi, one must
account for why Banda and Kaunda chose to exploit
the particular axes of cultural division that they did.
Group Size and Cleavage Salience
One additional, quite obvious, way in which Zambia
and Malawi differ is in their respective sizes-Zambia's
land area is almost eight times larger than Malawi'sand
in the sizes of the Chewa and Tumbuka communities
in each country relative to the country as a whole.
As Figure 3 makes clear, the Chewa and Tumbuka communities
are both large in Malawi (28% and 12% of
the national population, respectively) but peripheral
in Zambia (7% and 4%, respectively).16 Much more
15 In fact, they identify fewer: 1.2, on average, compared to 1.27 for
Zambian Tumbukas.
16 Malawi's figures come from Fearon 2003. Tumbukas include
Tonga- and Nkhonde-speakers. Zambia's figures are from that country's
1990 Census of Population and Housing based on answers to
the question "What is your Zambian tribe?"
537
This content downloaded from
86.49.253.47 on Thu, 23 Sep 2021 20:15:09 UTC
All use subject to https://about.jstor.org/terms
Political Salience of Cultural Difference November 2004
FIGURE 3. Chewas and Tumbukas in Zambia and Malawi
I2~/
than the differential impact of missionary education or
the contrasting styles of the two countries' leaders, this
simple structural difference critically affects the relative
political salience of the Chewa-Tumbuka cleavage
on either side of the border.
Politics in Zambia and Malawi, as in other developing
nations, revolves around the competition for
control over scarce patronage resources-things like
jobs, schools, clinics, roads, import and export licenses,
tariff exemptions, and access to credit. Because these
resources are controlled by the national government,
the key to gaining access to them is to build a political
coalition that can either capture political power outright
(which, in the highly centralized political systems
of these countries, means capturing and retaining the
presidency) or become a strong enough political force
to exert pressure on the person who holds that office.17
In the course of building these coalitions, politicians
find it useful to mobilize their supporters by exploiting
cultural divisions. However, there are numerous cultural
divisions that they might exploit. Which should
they choose? If the purpose of mobilizing the cultural
cleavage is to build a coalition that can help them
achieve political power, then it is natural for them to
emphasize the cleavage that defines the most usefully
sized coalitional building blocks and to ignore those
that define groups that are too small to be politically
viable. This logic provides a powerful explanation for
why Chewa-Tumbuka differences are so much more
politically salient in Malawi than in Zambia.18
In Malawi, both the Chewa and the Tumbuka communities
are sufficiently large as to constitute viable
political coalitions in the competition over national
power. Thus, when Malawian politicians from the
Northern and Central Regions seek to build political
support bases, they find the Chewa and Tumbuka
groups to be useful building-blocks, and their mobilization
of these communities renders the cultural cleavage
between them politically salient. There are other cultural
cleavages that politicians from these parts of the
country might mobilize: those between the Ngonde and
Henga, between the Lambya and Sukwa, or between
the Ngoni and Tonga, for example. However, none of
these other groups are large enough to be useful political
vehicles-they are all less than 5% of the national
population-so these cultural cleavages remain unexploited
and politically irrelevant. Only the Chewa and
Tumbuka are big enough groups in this part of the country
to represent viable political support bases for the
national political stage, so, among the several cultural
cleavages that might be mobilized, only the cleavage
between the Chewa and the Tumbuka is seized upon
by politicians, and only this cleavage comes to matter
politically.
In Zambia, neither the Chewa nor Tumbuka community
alone has the size to be an effective political
vehicle. These groups therefore go unmobilized by
Zambian politicians and the cultural differences between
them remain politically inconsequential. This17 Note that while capturing the presidency may require mobilizing
one's coalition in the context of a competitive election, securing
resources by exerting pressure does not. Positioning oneself as the
leader of a weighty political coalition is thus a central ingredient for
political success even in undemocratic contexts.
18 A formalization of this argument, along with an application to
other cases, is provided in Posner, forthcoming.
538
This content downloaded from
86.49.253.47 on Thu, 23 Sep 2021 20:15:09 UTC
All use subject to https://about.jstor.org/terms
American Political Science Review Vol. 98, No. 4
is not to say that Zambian politicians do not exploit
cultural differences in their coalition-building efforts,
for they do. Rather it is to say that, when they do "play
the ethnic card," they choose not to emphasize the
cleavage between Chewas and Tumbukas. The logic
of ethnic coalition-building in Zambia leads them instead
to mobilize Chewas and Tumbukas as part of
a common coalition of "Easterners." This "Eastern"
coalition, which includes Chewas, Tumbukas, and other
Nyanja-speaking groups from the Eastern part of the
country (and for this reason is sometimes referred to
as the coalition of "Nyanjas" or "Nyanja-speakers"),
constitutes just over 20% of the national population.
This is a size sufficiently large to permit it to compete
formidably in the national political arena, and this accounts
for why it is mobilized and for why the cultural
differences among its ethnic subunits (including the
Chewas and Tumbukas) are ignored.
Thus, in 1991, when President Kaunda embraced an
overtly ethnic appeal in that year's founding multiparty
election, he chose to mobilize Chewas and Tumbukas
together as Easterners rather than as distinct, competing
coalitions (Bratton 1994). Contrast this with
Banda's strategy in Malawi's first competitive multiparty
election three years later. His response was to
mobilize Chewas as Chewas, and to do so in large part
by invoking the threat posed to the community by its
Tumbuka rivals (Kaspin 1995; Posner 1995). In playing
the Chewa card, Banda was mobilizing a coalition that
he had assiduously cultivated over the past 30 years.
Banda had chosen to build up the Chewa community in
part because he was a Chewa himself, but also because
he recognized that if he could bring a handful of other
groups under the Chewa umbrella he could construct
a coalition that would "provide him with a very large
constituency among the common-folk" (Kaspin 1995,
604-5). Through a combination of strategic cultural redefinition19
and policies designed to force non-Chewas
to adopt the Chewa language (Kishindo 1998), Banda
slowly augmented the size of the Chewa coalition until
it comprised, on paper at least, more than 50% of
Malawi's population. So obsessed was he with the size
of the Chewa block that he reportedly ordered the
University of Malawi to fire a distinguished British linguist
after the linguist observed in a report that official
government estimates greatly exaggerated the number
of Chewa-speakers in the country (Vail and White 1989,
191). The fact that the size of the Chewa coalition was
so clearly more important to Banda than the homogeneity
of its cultural makeup suggests that coalitionbuilding
calculations were the motivating factor in his
invocation of the Chewa-Tumbuka cleavage.
Banda sought to build up the Chewa coalition, and
played Chewas off against Tumbukas as a means to
this end, because he was convinced that emphasizing
this cleavage would be politically useful to him. Meanwhile,
Kaunda ignored the Chewa-Tumbuka cleavage
not because, as he liked to claim, he was "above tribe"
but because he did not believe it was a useful line of
social division to exploit. Both politicians found it advantageous
to mobilize their supporters along ethnic
lines, and the cultural cleavage between Chewas and
Tumbukas was available for both of them to employ.
Yet while Banda chose to emphasize it, Kaunda did
not. The reason stemmed from the different sizes-and
thus political usefulness-of the Chewa and Tumbuka
communities in each setting. The contrasting cultural
demographies of the two countries cannot explain why
Banda played the ethnic card more vigorously than
Kanuda. But it can explain why, when they each embraced
ethnicity as a tool for political mobilization, they
chose to mobilize the particular cultural cleavages that
they did.
Politicians are not the only ones attuned to the logic
of ethnic arithmetic. Just as politicians seek to build
coalitions of viable sizes, voters also seek to gain entry
into coalitions that will permit one of their own to
win political power. Thus Chewa and Tumbuka voters
in Zambia were responsive to appeals made in
terms of their common identity as Easterners, and
Chewa and Tumbuka voters in Malawi were responsive
to appeals made in terms of their mutual differences.
Although Kaunda ultimately lost the 1991
election in a landslide, his strategy of trying to
build a unified support coalition among Chewas and
Tumbukas worked. He won 74% of the vote in
Eastern Province, compared to just 16% in the rest
the country. More importantly, within Eastern Provin
itself, he received comparably strong support fro
Chewa- and Tumbuka-dominated districts: 75% of
the vote in the Chewa-dominated districts of Chadiza
and Katete and 81% in the Tumbuka-dominated districts
of Chama and Lundazi.20 In the 1994 contest in
Malawi, meanwhile, the country split its vote almost
exactly along ethnic lines, with Chewa voters from the
Central Region overwhelmingly supporting Banda's
Malawi Congress Party (MCP), Tumbuka voters from
the Northern Region overwhelmingly supporting the
Tumbuka-led Alliance for Democracy (AFORD), and
almost no crossover voting taking place between the
two communities (Kaspin 1995; Posner 1995).21 The
union of Chewas and Tumbukas behind a common political
banner that took place in Zambia was entirely
absent in Malawi.
19 Banda would regularly visit non-Chewa regions of the country
and tell people that, contrary to their prior beliefs about who they
were, they were really Chewa (Vail and White 1989).
20 According to 1990 census data, Chadiza and Katete districts are
together 90% Chewa and Chama and Lundazi districts are together
75% Tumbuka. The United National Independence Party, Kaunda's
long-time political organization, boycotted the 1996 election, so the
unanimity of Chewa and Tumbuka support is difficult to assess in that
contest. However, clear patterns of intraregional unity were evident
in Eastern Province in the election of 2001 (Burnell 2003).
21 Voters from the ethnically heterogeneous Southern Region rallied
behind a third party, the United Democratic Front (UDF),
whose presidential candidate, Bakili Muluzi, won the election. In
1999, AFORD and MCP formed an electoral alliance and ran a
single presidential candidate. Patterns of party support in the simultaneously
held parliamentary contests nevertheless confirm that
Tumbuka-speaking Northerners were still overwhelmingly committed
to AFORD and that Chewas from the Central Region still overwhelmingly
supported the MCP (Ott, Phiri, and Patel 2000; Wiseman
2000). As in 1994, very little crossover voting took place between
Chewas and Tumbukas.
539
This content downloaded from
86.49.253.47 on Thu, 23 Sep 2021 20:15:09 UTC
All use subject to https://about.jstor.org/terms
Political Salience of Cultural Difference November 2004
TABLE 2. Trust among Chewas and
Tumbukas in Eastern Province, Zambia
(Percentage)
How Much Someone
Do You from Your Fellow NonTrust...
Own Tribe? Easterners? Easterners?a
Distrust them 5.6 8.4 18.3
at least
somewhat
Feel neutral 2.8 4.2 20.6
about them
Trust them 91.5 87.4 61.1
at least
somewhat
Source: Bratton (1998).
Note: N= 71.
a Average of responses for Southern, Northern, Western,
Luapula, Northwestern, and Central Provinces.
While the logic of ethnic coalition-building might
help us account for the greater willingness of Zambian
Chewas and Tumbukas to vote for a presidential candidate
from the other ethnic community, it provides a
much weaker explanation for the differences we find in
the two countries in the nature of social relations between
Chewas and Tumbukas more generally. Indeed,
one of the most striking findings in my survey results
is the much greater willingness of Chewas and Tumbukas
to marry across group lines in Zambia than in
Malawi. The Chewa-Tumbuka cleavage would appear
to be not only less politically salient in Zambia, but
also less socially salient. This finding is reinforced by
the results of a democratic governance survey administered
in Zambia in 1996 (Bratton 1998). The survey
asked respondents how much they trust members of
their own tribes, people from their province, and people
from other provinces. Table 2 reports the levels of
trust for people in each of these categories, as reported
by the survey's 71 Chewa and Tumbuka respondents
from Eastern Province. The relevant finding is that
while reported levels of trust for fellow tribespeople
(i.e., fellow Chewas and/or Tumbukas) were roughly
equivalent to those for fellow Easterners, both were
significantly higher than reported levels of trust for
non-Easterners (compare Table 2, columns 2 and 3 with
column 4). The survey results suggest that the relevant
line of division for Chewas and Tumbukas in Zambia
is not between each other but between themselves and
people from other regions of the country.22
What, then, explains the different social salience of
the Chewa-Tumbuka cleavage in each country? One
possibility is that social divisions are products of political
divisions. Over time, political mobilization and
counter-mobilization might breed distrust and animosity
that carries over from the political to the social
realm. Eventually such sentiments might come to be
reflected in beliefs about who can be trusted and who
is a suitable marriage partner. While this hypothesis is
plausible, testing it would require data (currently unavailable)
on changes over time in the depth of political
divisions, levels of trust, and rates of intermarriage.
Another possibility is that a person's sense of who
they are (and, by extension, who they trust and are willing
to marry) is shaped by the boundaries of the social
arena they inhabit. It is well documented that human
beings have a natural tendency to want to organize their
environments into a relatively small (and thus manageable)
set of categories (Brown 1986; Mitchell 1974).
It is also well known that changes in the situation in
which persons find themselves will alter how they think
about who they are (Hobsbawm 1996; Kasfir 1979). A
possible explanation for the different responses to the
intermarriage question among Zambian and Malawian
respondents (and for the different levels of trust that
Chewas and Tumbukas from Eastern Province exhibit
toward Easterners and other Zambians) is that while
"Chewa" and "Tumbuka" are large enough categories
to be included in Malawians' mental mappings of their
country's social landscape, they are too small to play
that role in Zambia, where they get lumped together
as part of the broader category of "Easterners." People
in both Zambia and Malawi might be more trusting of
and prefer to marry "one of their own." But because
the scales of the national social arenas they inhabit are
different, so too is their understanding of who is, and
is not, a member of their own group.23
The implied counterfactual is that if the Zambian
political and social arena were restricted to Eastern
Province alone, then the Chewa-Tumbuka cleavage
would be as salient as it is across the border in Malawi.24
But as long as the boundaries of the political and social
arena are congruent with the boundaries of the country
22 Unfortunately, the otherwise similar survey administered in
Malawi three years later did not include questions about trust, so
a comparison of the Zambian findings with findings from Chewa and
Tumbuka respondents in Malawi is not possible.
23 This may explain why, when well-educated Tumbukas were recruited
to work as clerks and bookkeepers in the Northern Rhodesian
copper mines during the colonial era, they were commonly
referred to not as "Thmbukas" but as "Nyasas" (i.e., those from
the far east of the country, near the border with Nyasaland), a designation
that applied equally to Chewas (Epstein 1958). The category
"Tumbuka" was based on too fine-grained a categorization scheme to
be part of peoples' colonywide mental mappings. Of course some of
these "Nyasas" actually were from Nyasaland. But no distinction was
generally made between those "Nyasas" who were from Nyasaland
and those who were from Northern Rhodesia. More importantly, no
distinction was made between the Tumbuka "Nyasas," who made up
the bulk of the managerial labor force, and the Chewa ones.
24 I tried to test this hypothesis by asking my Zambian survey respondents
to indicate how they would feel if the president were to
appoint a member of the other group as the deputy provincial minister
for Eastern Province. Unfortunately, there turned out to be two
problems with this question. First, the deputy provincial minister has
very little effective power over the distribution of resources, so the
question was one with very low stakes. Second, and more importantly,
President Kaunda, as part of his efforts to build national unity, had
historically gone out of his way to appoint provincial ministers from
outside of their home areas. Southerners were appointed in Eastern
Province, westerners were appointed in Northern Province, and so
forth. So when respondents reflected on how they would feel if a
Chewa or Tumbuka were appointed as deputy provincial minister,
their implicit point of comparison was the appointment of someone
from a completely different part of the country. Compared to this
possibility, either a Chewa or a Tumbuka appointee was considered
quite satisfactory.
540
This content downloaded from
86.49.253.47 on Thu, 23 Sep 2021 20:15:09 UTC
All use subject to https://about.jstor.org/terms
American Political Science Review Vol. 98, No. 4
as a whole, the cultural distinction between Chewas
and Tumbukas in Zambia will be ignored. The national
frame dictates a political strategy (and conjures up
a mental mapping of the country's relevant groups)
that overlooks the differences between Chewas and
Tumbukas and aggregates them as a single entity.
THE POWER OF ADMINISTRATIVE
BOUNDARIES
To explain why the Chewa-Tumbuka cleavage is so
much more salient in Malawi than in Zambia, I have
emphasized the differences in the sizes of the Chewa
and Tumbuka communities vis-a-vis the sizes of the
larger Zambian and Malawian political arenas. An
additional potentially relevant factor that I have yet
to consider is the correspondence in Malawi of the
Chewa-Tumbuka cleavage with the administrative
boundary between that country's Northern and Central
Regions. Indeed, for most Malawians, the labels
"Tumbuka" and "Northerner" (and, to a somewhat
lesser degree, "Chewa" and "Central Region") mean
the same thing. Could the regional boundary be doing
some of the work in reifying the cultural division?
Might the correspondence between the administrative
and the cultural boundary in Malawi be making the
cleavage between Chewas and Tumbukas salient?
The situation across the border in Zambia provides
useful perspective. The Zambian Chewa and Tumbuka
communities are also separated by an administrative
boundary-that between the districts of Chipata and
Lundazi. Yet the coincidence between the cultural and
the administrative dividing lines in Zambia appears
to do nothing to make the Chewa-Tumbuka cleavage
politically salient. This is because the administrative
units-the districts-that it defines are themselves
too small relative to the national political arena to
become useful building blocks for political coalitionbuilding.25
Moreover, the administrative boundary that
does matter-the provincial one-lumps Chewas and
Tumbukas together as part of a common entity, and this
probably undermines whatever cleavage-reinforcing
effect the Chipata-Lundazi district boundary might
have.26 The lesson for Malawi seems to be that if
the correspondence between the Northern-Central
Region boundary and the cultural boundary between
the Chewa and the Tumbuka peoples plays any role
in rendering the Chewa-Tumbuka cleavage salient,
it is only because the communities that the administrative
boundary demarcates are already sufficiently
large to be useful political vehicles and/or categories for
countrywide social categorizations. It is not the administrative
boundary as such that reinforces the cultural
cleavage but the fact that it happens to coincide with
the dividing line between a pair of big, politically viable
units.27
But can administrative boundaries do more than simply
reinforce existing cultural cleavages? Providing
that they define groups of the right size, might they also
have the power to create social and political cleavages
that have no cultural foundation? To answer this question,
I collected additional data among a third Malawian
group, the Lakeshore Tonga. The Lakeshore Tonga
straddle the Northern Region-Central Region border
along the shore of Lake Malawi. One of the most
studied ethnic groups in all of Malawi, the Lakeshore
Tonga are regarded as having a particularly strong and
autonomous culture (Tew 1950; van Velsen 1964). They
are thus a particularly unlikely group to have their loyalties
divided by the imposition of an administrative
boundary within their midst. By taking advantage of
the natural experiment afforded by the division of this
community by the Northern-Central Region boundary,
I sought to test whether they had been.
I administered a survey to 30 randomly selected
respondents in villages outside of Nkhata Bay, the
largest Lakeshore Tonga population center.28 Nkhata
Bay is located approximately 95 kilometers north of
the Northern Region-Central Region border. Nearly
everyone in Nkhata Bay (and everyone in my sample)
is Lakeshore Tonga by tribe and "Northerner" by region.
The surveys asked two questions:
* Suppose there were two people standing for president:
one person was from Rumphi and the other was
from Liwaladzi. Which one would you vote for?29
* Which kind of person would you feel more comfortable
marrying: a person from Rumphi or a person
from Liwaladzi?30
Rumphi, which is located approximately 125 kilometers
northwest of Nkhata Bay, is the unofficial capital
of Tumbukaland and a major Northern Region town.
When respondents were asked about "a person from
Rumphi," they therefore understood this to refer to a
fellow Northerner who was not a Tonga. Liwaladzi is
the southernmost discernibly Lakeshore Tonga town.
It is located about 130 kilometers south of Nkhata Bay
25 Zambia is divided into more than 60 administrative districts. Leaving
aside the district encompassing Lusaka, the capital city, they each
contain at most 5% of the national population.
26 Note that if the provincial boundary happened to correspond with
the Chewa-Tumbuka cultural boundary-for example, if Chewas
were considered "Easterners" but Tumbukas were considered
"Northerners"-than the provincial administrative boundary would
render the cultural boundary salient. It just happens that the cultural
unit that matters (being an "Easterner") includes both Chewas and
Tumbukas as constituent parts. I thank Michael Thies for pointing
this out.
27 Note that whether these units are viewed as cultural groups or as
geographic/administrative regions is irrelevant. What matters is the
size.
28 As with the Chewa-Tumbuka survey work, the sample was stratified
by age and gender, and respondents were selected from every
third unrelated household. The survey was conducted in Citonga.
29 If respondents hesitated to choose one or the other, they were
prompted further: "What if you had to choose one of the two? You
can assume that the two people are equally qualified. The only difference
is that one comes from Rumphi and the other comes from
Liwaladzi. If you had to choose one of them which would it be?"
30 If respondents hesitated to choose one or the other, they were
prompted further: "What if you had to choose one of the two? You
can assume that the two people are the same except that one comes
from Rumphi and the other comes from Liwaladzi. Would you feel
more comfortable with one than the other? Which one?"
541
This content downloaded from
86.49.253.47 on Thu, 23 Sep 2021 20:15Jan 1976 12:34:56 UTC
All use subject to https://about.jstor.org/terms
Political Salience of Cultural Difference November 2004
FIGURE 4. Lakeshore Tonga Research Sites
Rump
Nkhata Bay
Liwaladzi
and about 30 kilometers south of the Northern RegionCentral
Region border. When respondents were asked
about "a person from Liwaladzi," they therefore understood
this to refer to a fellow Tonga who was not a
Northerner. Figure 4 makes the research design clear.
As should be obvious, the questionnaire was designed
to force respondents to choose between their
tribal and their regional identities. If they responded
that they preferred to vote for or to marry someone
from Rumphi, they were indicating that, at least in the
admittedly narrow contexts defined by the questions
I asked, they valued their regional identity more than
their tribal identity. If they responded that they preferred
to vote for or to marry someone from Liwaladzi,
they were indicating that they valued their tribal identity
more than their regional identity. A purely culturalist
expectation would lead us to anticipate all respondents
to express a preference for Liwaladzi over
Rumphi, since their links to Liwaladzi candidates and
potential spouses are based on shared culture rather
than simply chance assignment to the same administrative
region-only the former being "real."31 The culturalist
expectation turns out to be wrong. Nearly half of
the respondents (46%) expressed a preference for the
fellow Northerner candidate over the fellow tribesperson32
and well over a third of the respondents (38%)
expressed a preference for a spouse from the same
31 Of course, Tumbukas and Tongas are not that dissimilar culturally.
But the cultural ties between fellow Tongas are still much closer than
between Tongas and Tumbukas.
32 These findings match Kaspin's claim regarding the 1994 election
that "Tonga voters in the north voted for AFORD [whose
presidential candidate, Chikufwa Chihana, was a Tumbuka], while
the contiguous Tonga population in the centre gave most of their [sic]
support to the UDF [whose presidential candidate, Bakili Muluzi,
was a southerner]. The cohesion of ethnicity qua ethnicity was thus
542
This content downloaded from
86.49.253.47 on Thu, 23 Sep 2021 20:15:09 UTC
All use subject to https://about.jstor.org/terms
American Political Science Review Vol. 98, No. 4
region but a different tribe.33 Although the sampl
is too small to be more than suggestive, the results
should nonetheless give pause to those who would
assign special status to cleavages built around "real"
cultural material.
In the introduction, I motivated my analysis by noting
that the presence of an identifiable cultural difference
is not a sufficient condition for the emergence of a
salient political cleavage. The results of the Lakeshore
Tonga analysis suggest that it may not be a necessary
condition either. Whether or not a cleavage matters
would seem to depend not at all on the material from
which it is built. That material can be as sturdy as the
traits, customs, norms, and practices that a professional
ethnographer might identify or as flimsy as an arbitrary
boundary drawn by an uninformed colonial officer.
Much more than the stuff from which the boundary is
made, what matters seems to be whether the cleavage
defines groups that, because of the scope of the larger
political and social arena in which they are located, are
of a useful size for political mobilization and/or social
categorization.
CONCLUSION
In this paper, I have drawn on a pair of natural experiments
to make two important points about the
relationship between culture and politics. The first is
that the political or social salience of a cultural cleavage
does not follow axiomatically from the fact that
the cultural cleavage exists. Some cultural cleavages
matter a lot for political competition and social categorization,
and others do not. The second key point is
that innate cultural differences do not necessarily have
greater power than noncultural differences to generate
political or social division. Administrative boundaries
with no cultural basis whatsoever-in the example explored
here, a boundary that partitioned an otherwise
homogeneous cultural community-can, under certain
conditions, have the power to create salient cleavages.
Taken together, these findings suggest that cultural differences
are neither sufficient nor necessary for the
emergence of political or social divisions.
Although the natural experiments discussed in this
paper provide especially compelling illustrations of
these two points, neither conclusion is particularly
new-or, perhaps, surprising. What is more novel is
the paper's argument about why the Chewa-Tumbuka
cleavage matters in Malawi but not in Zambia: namely,
that the political and social salience of the cleavage
depends on the sizes of the groups that the cleavage defines
relative to the sizes of the political and social arenas
in which the groups are located. In Malawi, Chewas
and Tumbukas constitute large coalitions whose mob
lization is politically useful in the national competiti
for state resources. As a consequence, the Chew
Tumbuka cleavage is highly politically salient in th
country. In Zambia, the Chewa and Tumbuka comm
nities are both too small to serve as viable politi
vehicles in and of themselves, so the cultural differ
ences between these groups go largely ignored. Indee
when cultural differences are mobilized in that regi
of Zambia, Chewas and Tumbukas tend to be mobiliz
together as part of the same broader cultural categor
This further reduces the salience of the cleavage th
exists between them.
Apart from their own theoretical import, the arguments
advanced in this paper have a number of
additional implications. One implication is for the literature
that employs indexes of ethnic fractionalization
to test for the effects of cultural diversity on outcomes
like economic growth, political stability, and civil
war. With few exceptions (Posner 2004; Scarritt and
Mozaffar 1999), these indexes are built from ethnographers'
listings of ethnically distinct communities, often
with language or religion as the determining criterion.
The problem is that consumers of these indexes implicitly
equate objective cultural differences with politically
or socially salient differences, and, as this paper
suggests, the two are not the same thing. Thus, for example,
Morrison et al. (1989) include the Chewa and
the Tumbuka in their enumeration of ethnic groups
in both Zambia and Malawi. While it is certainly true
that the Chewa and Tumbuka are culturally distinct
groups in both countries, including them as separate
entries in the list of relevant ethnic communities in
Zambia would be a mistake if one then used this list
to calculate a fractionalization value that purported to
measure the country's politically relevant ethnic diversity.
The objective cultural difference between these
groups is simply not reflected in the country's political
affairs.
A second implication is with respect to the longstanding
debate about the importance of Africa's arbitrary
boundaries (e.g., Englebert et al. 2002). Much of
this debate turns on the premise that, because they are
arbitrary, Africa's boundaries are weaker sources of social
identity than the "authentic" cultural communities
that these boundaries overwrote. One of this paper's
implications is that this premise is wrong: A lack of
cultural "authenticity" need not imply that a political
boundary is any less politically or socially meaningful
than one based on cultural difference. Moreover, the
paper suggests that perhaps the key effect of the Partition
was to alter the salience of the preexisting cultural
cleavages within African countries. Whatever salience
there may have been to the Chewa-Tumbuka cleavage
in Zambia prior to the drawing of the Zambia-Malawi
border has clearly been altered by the fact that Chewas
and Tumbukas now operate in a political and social
environment with a new, much broader, frame of reference.
The implication is that the impact of Africa's
colonial era boundary drawing (as well as the impact
of a redrawing of those boundaries) may lie as much
within states as at their borders.
relatively weak compared to the greater claims of regional affiliation"
(1995, 615).
33 In keeping with the results of the Chewa-Tumbuka analysis, female
respondents were much more likely to express a preference
for a spouse from the same tribe. Fully 12 of 15 said that they would
prefer to marry a man from Liwaladzi. Male respondents were almost
exactly evenly split: seven expressed a preference for a wife from
Rumphi, six expressed a preference for a wife from Liwaladzi, and
two insisted that the decision would depend on other factors.
543
This content downloaded from
86.49.253.47 on Thu, 23 Sep 2021 20:15:09 UTC
All use subject to https://about.jstor.org/terms
Political Salience of Cultural Difference November 2004
A third implication is with respect to the role of
political entrepreneurs in ethnic mobilization. It is n
commonplace to view ethnic political mobilization
a product of the strategic behavior of rational polit
cians (Bates 1983; Brass 1991; Chandra 2004; Kasf
1979). Yet I suggest that the behavior of such poli
cians must itself be viewed as a product of underlyi
ethnic demography. Political entrepreneurs are nec
sary agents in the story. They are the ones who do
mobilizing. But, if they hope to be successful, they mu
be attuned to the cultural demography that they ha
to work with. And that cultural demography crea
incentives for mobilizing the different cleavages in dif
ferent contexts.
The final implication of the paper is methodological.
Like the political actors we study, political scientists are
affected by the boundaries of states. Analysis tends to
stop where one country ends and another begins. Yet
the insights this paper provides were made possible by
focusing precisely on the seams between political systems.
Natural experiments of this sort are everywhere,
and much can be learned from taking fuller advantage
of them.
REFERENCES
Asiwaju, A. I. 1985. Partitioned Africans: Ethnic Relations and
Africa's International Boundaries, 1884-1984. London: C. Hurst.
Bannon, Alicia, Edward Miguel, and Daniel N. Posner. 2004.
"Sources of Ethnic Identification in Africa." Typescript, UCLA.
Barnes, J. A. 1954. Politics in a Changing Society: The Political
History of the Fort Jameson Ngoni. Oxford: Oxford University
Press.
Bates, Robert H. 1976. Rural Responses to Industrialization: A Study
of Village Zambia. New Haven, CT: Yale University Press.
Bates, Robert H. 1983. "Modernization, Ethnic Competition and
the Rationality of Politics in Contemporary Africa." In State
versus Ethnic Claims: African Policy Dilemmas, ed. Donald
Rothchild and Victor A. Olorunsola. Boulder, CO: Westview, 152-
71.
Brass, Paul R. 1991. Ethnicity and Nationalism: Theory and Comparison.
New Delhi: Sage.
Bratton, Michael. 1994. "Economic Crisis and Political Realignment
in Zambia." In Economic Change and Political Liberalization in
Sub-Saharan Africa, ed. Jennifer A. Widner. Baltimore: Johns
Hopkins University Press, 101-28.
Bratton, Michael. 1998. "Zambia Democratic Governance Project
Political Attitudes Survey, 1993 and 1996" [computer file]. ICPSR
version. East Lansing, MI: Michigan State University, Department
of Political Science [producer], 1997. Ann Arbor, MI: Interuniversity
Consortium for Political and Social Research [distributor],
1996.
Brown, Roger. 1986. Social Psychology. 2nd ed. New York: Free
Press.
Burnell, P. 2003. "The Tripartite Elections in Zambia, December
2001." Electoral Studies 22 (June): 388-95.
Caselli, Francesco, and John Coleman. 2002. "On the Theory of Ethnic
Conflict." Typescript. Harvard University.
Chandra, Kanchan. 2004. Why Ethnic Parties Succeed: Patronage and
Ethnic Head Counts in India. Cambridge: Cambridge University
Press.
Chirwa, Wiseman C. 1998. "Democracy, Ethnicity and Regionalism:
The Malawian Experience, 1992-1996." In Democratization
in Malawi: A Stocktaking, ed. Kings M. Phiri and Kenneth R. Ross.
Blantyre: Kachere Books, 52-69.
Deutsch, Karl W. 1961. "Social Mobilization and Political Development."
American Political Science Review 55 (September): 493-
514.
Dresang, Dennis. 1974. "Ethnic Politics, Representative Bureaucracy
and Development Administration: The Zambian Case." American
Political Science Review 68 (December): 1605-17.
Englebert, Pierre, Stacy Tarango, and Matthew Carter. 2002.
"Dismemberment and Suffocation: A Contribution to the Debate
on African Boundaries." Comparative Political Studies 35
(December): 1093-118.
Epstein, A. L. 1958. Politics in an Urban African Community.
Manchester, UK: Manchester University Press.
Fearon, James D. 2003. "Ethnic Structure and Cultural Diversity
by Country." Journal of Economic Growth 8 (June): 195-
222.
Fearon, James D., and David D. Laitin. 1996. "Explaining Interethnic
Cooperation." American Political Science Review 90 (December):
715-35.
Forbes, H. D. 1997. Ethnic Conflict: Commerce, Culture and the Contact
Hypothesis. New Haven, CT: Yale University Press.
Gertzel, Cherry, Carolyn Baylies, and Morris Szeftel. 1984. The
Dynamics of the One-Party State in Zambia. Manchester, UK:
Manchester, University Press.
Grimes, Joseph E., and Barbara E Grimes. 1996. Ethnologue:
Languages of the World. Dallas: Summer Institute of
Linguistics.
Gurr, Ted Robert. 2000. Peoples Versus States: Minorities at Risk in
the New Century. Washington, DC: U.S. Institute of Peace.
Henkel, Reinhard. 1985. "Mission Stations in Zambia: Their Location
and Diffusion Patterns." Zambia Geographical Journal 35:
1-18.
Hobsbawm, Eric. 1996. "Are all Tongues Equal?" In Living as
Equals, ed. Paul Barker. New York: Oxford University Press, 85-
98.
Horowitz, Donald. 1985. Ethnic Groups in Conflict. Berkeley and
Los Angeles: University of California Press.
Horowitz, Donald. 1991. A Democratic South Africa? Constitutional
Engineering in a Divided Society. Berkeley and Los Angeles: University
of California Press.
Horowitz, Donald. 2001. The Deadly Ethnic Riot. Berkeley and Los
Angeles: University of California Press.
Huntington, Samuel P 1968. Political Order in Changing Societies.
New Haven, CT: Yale University Press.
Huntington, Samuel P. 1996. The Clash of Civilizations and the Remaking
of World Order. New York: Simon and Schuster.
Kasfir, Nelson. 1979. "Explaining Ethnic Political Participation."
World Politics 31 (April): 365-88.
Kaspin, Deborah. 1995. "The Politics of Ethnicity in Malawi's
Democratic Transition." Journal of Modern African Studies 33
(December): 595-620.
Kishindo, Pascal J. 1998. "Politics of Language in Contemporary
Malawi." In Democratization in Malawi: A Stocktaking, ed.
Kings M. Phiri and Kenneth R. Ross. Blantyre: Kachere Books,
252-80.
Laitin, David D. 1986. Hegemony and Culture: Politics and Religious
Change among the Yoruba. Chicago: University of Chicago
Press.
Lake, David, and Donald Rothchild. 1998. "Ethnic Fears and
Global Engagement." In The International Spread of Ethnic
Conflict: Fear, Diffusion and Escalation, ed. David Lake and
Donald Rothchild. Princeton, NJ: Princeton University Press, 339-
50.
Lijphart, Arend. 1977. Democracy in Plural Societies: A Comparative
Exploration. New Haven, CT: Yale University Press.
Lipset, Seymour Martin, and Stein Rokkan. 1967. "Cleavage Structures,
Party Systems and Voter Alignments: An Introduction." In
Party Systems and Voter Alignments: Cross-National Perspectives,
ed. Seymour M. Lipset and Stein Rokkan. New York: Free Press,
1-64.
Miles, William E S. 1994. Hausaland Divided: Colonialism and Independence
in Nigeria and Niger. Ithaca, NY: Cornell University
Press.
Miles, William E S., and David Rochefort. 1991. "Nationalism versus
Ethnic Identity in Sub-Saharan Africa." American Political Science
Review 85 (June): 393-403.
Mitchell, J. Clyde. 1974. "Perceptions of Ethnicity and Ethnic Behaviour:
An Empirical Exploration." In Urban Ethnicity, ed.
Abner Cohen. London: Tavistock, 1-35.
544
This content downloaded from
86.49.253.47 on Thu, 23 Sep 2021 20:15:09 UTC
All use subject to https://about.jstor.org/terms
American Political Science Review Vol. 98, No. 4
Morrison, Donald, Robert Mitchell, and John Paden. 1989. Blac
Africa: A Comparative Handbook. 2nd ed. New York: Parag
House.
Ott, Martin, Kings Phiri, and Nandini Patel. 2000. Malawi's Second
Democratic Elections: Process, Problems and Prospects. Blantyre:
Christian Literature Association in Malawi.
Posner, Daniel N. 1995. "Malawi's New Dawn." Journal of Democracy
6 (January): 131-45.
Posner, Daniel N. 2004. "Measuring Ethnic Fractionalization in
Africa." American Journal of Political Science 48 (October).
Posner, Daniel N. Forthcoming. Institutions and Ethnic Politics in
Africa. New York: Cambridge University Press.
Reilly, Benjamin. 2001. Democracy in Divided Societies: Electoral
Engineering for Conflict Management. New York: Cambridge University
Press.
Republic of Malawi. 1992. Malawi Demographic and Health Survey.
Lilongwe: Government Printer.
Republic of Zambia. 1992. Zambia Demographic and Health Survey.
Lusaka: Government Printer.
Reynolds, Andrew. 1999. Electoral Systems and Democratization in
Southern Africa. Oxford: Oxford University Press.
Scarritt, James, and Shaheen Mozaffar. 1999. "The Specification of
Ethnic Cleavages and Ethnopolitical Groups for the Analysis of
Democratic Competition in Contemporary Africa." Nationalism
and Ethnic Politics 5 (Spring): 82-117.
Sklar, Richard L. 2003. "Introduction of Kenneth David
Kaunda, Former President of Zambia, Speaking on 'HIV/AIDS
and Poverty in Africa' at UCLA." Los Angeles, February
27.
Snelson, Peter. 1974. Educational Development in Northern Rhodesia:
1883-1945. Lusaka: Kenneth Kaunda Foundation.
Straus, Scott. 2004. "The Order of Genocide: Race, Power, and War
in Rwanda." Ph.D. dissertation. University of California, Berkeley.
Tew, Mary. 1950. Ethnographic Survey of Africa: Peoples
of the Lake Nyasa Region. London: International Africa
Institute.
Vail, Leroy. 1981. "Ethnicity, Language and National Unity: The
Case of Malawi." In Working Papers in Southern African Studies,
Volume 2, ed. P. Bonner. Johannesburg: Ravan.
Vail, Leroy, and Landeg White. 1989. "Tribalism in the Political History
of Malawi." In The Creation of Tribalism in Southern Africa,
ed. Leroy Vail. Berkeley and Los Angeles: University of California
Press, 151-92.
van Velsen, Jap. 1964. The Politics of Kinship: A Study in Social
Manipulation among the Lakeside Tonga of Malawi. Manchester,
UK: Manchester University Press.
Wiseman, J. A. 2000. "Presidential and Parliamentary Elections in
Malawi, 1999." Electoral Studies 19 (December): 637-46.
Young, Crawtord. 1994. The African Colonial State in Comparative
Perspective. New Haven, CT: Yale University Press.
545
This content downloaded from
86.49.253.47 on Thu, 23 Sep 2021 20:15:09 UTC
All use subject to https://about.jstor.org/terms