Document Sample

Appendix: Data Measurements and Sources

Area. This variable (AREA) measures the square-miles terrestrial surface of countries. Source:
Africa South of the Sahara.

Average democratic score. This variable (CIVLAVG) averages, for each boundary, the Gastil
index score of the countries on both sides over the period 1960-1989. Source: Average of the
decade figures in Easterly and Levine (1997).

Average political culture and spread of political cultures. These variables (MEANF3 and
SDF3) measure the central tendency and the spread of pre-colonial political cultures in
contemporary African states. Pre-colonial political systems are assigned to one of three
categories: stateless societies (scoring 1), chiefdoms (2), and kingdoms or larger centralized states
(3). We then measure for each country the percentage of its population which belongs to an
identity group in either one of these categories. From this we derive a weighted average political
culture for each country, which takes on higher values the more state-like the country’s
component groups are, and a frequency-based standard deviation which captures the “distance”
between the different cultures comprised in the state. A country can score 2 on the weighted
average because it only comprises chiefdom societies or because it is evenly split between groups
scoring 1 and 3. In the first case, the standard deviation will be 0, whereas in the second it will be
1. Note that the ordinal nature of the original data is converted to continuous by the process of
frequency- based weighing. The “distance” between the cultures remains arbitrary, however, and
the findings can better be described as capturing trends rather than actually meaningful
magnitudes. Source: Computed by the authors from multiple sources including Murdock (1967),
Morrison et al. (1989), and Olson (1996). For methodology, see Wonacott and Wonacott (1985).

Boundary length. In the boundary data set, this variable (LENGTH) measures the length of each
bilateral boundary in kilometers. In the country data set, it measures the total length of each
country's land boundaries. Source: CIA Factbook 1999. The Tanzania-DRC border is an
estimate using a map and dental floss because no entry was reported in the CIA factbook.

Capabilities differential. For each boundary, this variable (CAPDIF) is the absolute value of the
difference between the average extractive capacity scores of the countries on both sides for all
available years. Source: Kugler (199_).

Civil Wars. A variable (CIVWAR) which measures the number of years between independence
and 1999 during which a country experienced episode(s) of civil violence, including irredenta,
secession and resistance, adjusted for the number of years during which the country was
independent between 1960 and 1999. Source: calculated by the authors from appendix in Hoslti
(1996), expanded to 1999. Mauritania, mentioned as having had 20 year of civil violence was
downgraded to 7 years. The 20 years in Holsti (1996) cover mostly irredentist claims of Morocco
over Mauritania and Mauritania over Western Sahara, which are not properly domestic in nature.
We substituted 2 years for linguistic/ethnic violence in 1966 and 1967, 3 years of Saharaoui
guerilla actions while Mauritania included southern Western Sahara (1976-1979), and two years
of ethnic violence between Moors and Toucouleur/Pulaar populations in 1989-1990. See
Morrison et al. (1989), and Africa South of the Sahara (1995).
Colonization by Belgium. Dummy variable (BELGIUM) which takes on a value of 1 if a
country was originally colonized by Belgium.

Colonization by France. Dummy variable (FRANCE) which takes on a value of 1 if a country
was originally colonized by France.

Dismemberment or percentage population partitioned. In the boundary data set, this variable
(ASTRPROP) is the unweighted average proportion of groups astride a border. It equals one half
the sum of the percentages of national populations that these groups represent on both sides of
border. This variable is unweighted so as to reflect the proportion of ethnic groups in small
countries. For example, if one country of 10,000 people contains 5,000 split nationals (50%) and,
next door, a country of 100,000 also contains 5,000 split nationals (5%), the weighted average
would be less than 10%, whereas the unweighted average will be 22.5%. Hence, the simple
average gives a better reflection of the average national burden of partition between these two
countries. Interpretation of the variable is difficult. It is better thought of as an index of
dismemberment than as an actual proportion. In the country data set, the variable (SPLIT) simply
represents the percentage of the country’s population that belongs to an ethnic group that was
partitioned by colonial borders. Source: Asiwaju (1985) and Englebert (2000). In both data sets,
Burundi- Rwanda and Burundi-Tanzania have been changed from Asiwaju (1985), with Rundi
populations no longer counted as partitioned, as per Griffiths (1986).

Dispute. In its dummy version (as in Table 2), this variable (DISPDUM) takes on a value of 1
for the occurrence of any dispute over the border or territorial issue since independence, whether
full armed conflict or skirmish or lesser dispute. In its ordinal version (as in Table 3), the variable
(DISPUTE) takes on the following values: 0 = no dispute; 1 = boundary line is disputed but
dispute is not pursued; 2 = nonviolent claim by at least one state; 3 = claim and some fighting; 4
= full war or ICJ case. Source: computed by the authors from multiple sources (see dispute table
at the end of this appendix).

East Africa location. Dummy variable (EAST) which takes on a value of 1 if a country is
located in East Africa.

Economic instability. This variable (STDGROW) measures the standard deviation of a
country's annual growth rates over the years of available data. Source: World Bank Africa
Database CD-ROM, 1999.

French on one side. A dummy variable (FRENCH1) which takes on the value 1 if at least one
side of a boundary was colonized by the French. Source: Authors.

National design. A variable (HERBST) which measures the quality of a country's geographical
and demographic features in terms of ease of broadcasting of power. This is an ordinal variable
with values of 1,2,3 and 4 as the quality of the design increases. Source: Table 5.2 in Herbst

Number of ethno-cultural cluster groups. This variable (NUMBER) measures the number of
cluster or aggregated ethnic groups per country, collapsing similar ethnic/linguistic groups in a
same cluster. Source: Morrison et al. (1989).

Political culture of populations astride. This variable (ASTRMF3) measures the unweighted
average political culture of populations astride a given boundary. Stateless groups score 1,
"tribal" groups 2 and state-like groups 3. Source: Computed by the authors from multiple
sources, including Asiwaju (1985), CIA (1999), Grimes (1999), Library of Congress misc.
Country Studies, Murdock (1957), Morrison et al. (1989), and Olson (1996). The partitioned
groups were identified from Asiwaju (1985). If the size of a group identified by Asiwaju was
insignificant, the entry was deleted from the total of both countries.

Political instability. All coups, coup attempts, alleged coup plots between 1960 and 1990
in proportion of years of independence betwen 1960 and 1990 (PROPCOUP). Source:
Bates 2000.

Population growth rate. Average rate of population growth (POPGROW) over the 1960-1992
period or available years. Source: World Bank Africa Database CD-ROM, 1999.

Portuguese colonialism. Dummy variable (PORTUGAL) which takes on a value of 1 if a
country was colonized by Portugal.

Proportion of neighboring countries with same colonizer. A variable (NEIGSAME) which
computes the percentage of a country's neighbors which were colonized by the same power.

Quality of demarcation. Ordinal variable (MARC) capturing the quality of boundary
demarcation as of 1979. The following scale is used: No demarcation due to severe problems in
delimitation = 1; No or poor demarcation using unreliable natural features = 2; Partial
demarcation using some beacons and reliable features = 3; Good demarcation, but requires some
completion or repair = 4; Complete and clear demarcation = 5. Source: Coded by the authors
from Browlie (1979).

Religious polarization. This variable (HERETIC) multiplies the proportion of Muslims with the
proportion of Christians, both expressed in decimals, for each country. Its maximum theoretical
value is 0.25 for a country with 50% Christians and 50% Muslims. Source: Computed by the
authors from data in Kurian (1998).

Same colonizer. A dummy variable (SAMECOL) which takes on the value 1 if both sides of a
boundary were colonized by the same country. Only original colonizers are considered since they
determined the boundaries (eg, the Kenya-Tanzania border scores 0). Source: Authors.

Secessions. This variable (SECEDE) measures the natural logarithm of 1 + all instances of
secessionist movements from 1946 to 1998. Source: quantified by the authors from Baker

Southern Africa location. Dummy variable (SOUTH) which takes on a value of 1 if a country is
located in Southern Africa, as assessed by membership in SADC. Source: Africa South of the

Straight line. Dummy variable (LINEDUM) taking a value of 1 if at least part of a boundary is
composed of straight lines. Source: Authors, from National Geographic map.