Transcript NUMBERS AT WORK

DATA AT WORK:
NEGOTIATING CIVIL WARS
Again: Summary Measures
for Cross-Tabulations
• Lambda-b
PRE, ranges from zero to unity;
measures strength only
• Gamma
Form and strength (-1 to +1); PRE,
based on “pairs” of observations
• Chi-square
Significance, based on
deviation from “null hypothesis”
Questions and Themes
• Theme: Outcomes of civil wars
• General questions: How do civil wars end?
Does it make any difference?
• Specific questions: Why are there negotiated
settlements in some instances and not in
others? Do negotiations produce different
(better) outcomes than military victories?
Defining the Unit of Analysis
• What is a “civil war”?
• 1. Leaders concerned about living in same political
unit with current enemies
• 2. Multiple sovereignty: people within area obey
more than one institution—each side has troops made
up of local residents
• 3. Large-scale violence: (a) at least 1,000 battle
deaths per year and (b) effective resistance
Characteristics of the Key Variable
• N civil wars (1945-1993) = 91
• Of these, 57 had been ended for at least 5 years
by 1994
• Of these, 14 ended by negotiation, 43 by
military victory
• Among these 57, 11 were followed by largescale violence that was not a civil war
• Therefore effective N = 46
Selecting Levels of Measurement
• Dichotomous variables:
–
–
–
–
Negotiation or victory?
War after settlement? (yes/no)
Genocide after settlement? (yes/no)
Type of war: identity or political-economic
• Ordered nominal variable:
– Casualties: 1,000-10,000 deaths, 10,000 to
100,000, 1000,000 to 1 million, over one million
On Measuring Association
“This data set is fairly primitive. The number of cases is
quite small, even when the entire population rather than the
sample is analyzed. Many of the variables are nominal, and
the reliability of some of the figures is questionable. I do not
believe this data [sic] warrants the use of sophisticated
statistical techniques. Instead, I simply report Pearson’s chisquared, the probability that the two variables are independent
of one another, and associated statistics. Even these are
handicapped by the small expected values in some cells, but at
least they give us some sense of the likelihood that the
relationships have occurred by chance.”
Central Hypothesis
• Wagner hypothesis: “Negotiated settlements of
civil wars are more likely to break down than
settlements based on military victories;
consequently, the long-term casualties of
negotiated settlements are likely to be greater
than those of military victory.”
Testing the Hypothesis
• Among the civil wars under study, threequarters (34 of 46) “were not renewed” after
the settlement. “Such a renewal occurred in
only 15% of the victories, as opposed to 50%
of the negotiated settlements. Several different
chi-squared measurements show this
relationship to be significant at the .05 level or
better.”
Empirical Models Y = f (X)
•
•
•
•
•
WARS = f (TERMINATION) [Table 1]
CASUALTIES = f (IDENTITY)
GENOCIDE = f (IDENTITY)
TERMINATION = f (IDENTITY)
WARS = f (IDENTITY, TERMINATION)
[Table 2]
• GENOCIDE = f (IDENTITY,
TERMINATION) [Table 3]
Table 1. Method of Termination and
Same Wars after Settlement
_Y: Wars? (%)_
X: Termination__ None
Same __N__
Military Victory
Negotiation
85
50
15
50
34
12
Table 2. Issue, Termination, and More
Wars after Settlement
___Y: Wars? (%)__
None
Same
X: Termination__
Z: Identity Wars:
Military
79
Negotiation
33
Z: Political/Economic Wars:
Military
100
Negotiation
100
_N_
21
67
24
9
0
0
10
3
Table 3. Issue, Termination, and
Genocide
__Y: Genocide? (%)_
X: Termination__ Yes
No
Z: Identity Wars:
Military
19
81
Negotiation
0
100
Z: Political/Economic Wars:
Military
7
93
Negotiation
0
100
__N__
27
10
16
4
Major Findings
• 1. Negotiated settlements of civil wars are less
likely to endure than results of military victories
[Table 1]
• 2. Identity vs. political-economic wars do not
have clearly different casualty patterns
• 3. Identity wars are not more “intense” than
political-economic wars
• 4. Identity and nonidentity wars are equally
likely to end in negotiated settlements
Major Findings (cont.)
• 5. Negotiated settlements of identity wars
are less stable than military victories
[Table 2]
• 6. Military victories in identity wars may be
more likely to be followed by genocide
than negotiated settlements [Table 3]
• 7. Genocide is less likely to follow
termination of nonidentity conflicts in
general