RECONCEPTUALIZING SALIENCE

1
RECONCEPTUALIZING SALIENCE IN TERRITORIAL
DISPUTES A CROSS SECTIONAL TIME SERIES ANALYSIS
NAZLI AVDAN nazli.avdan_at_duke.edu Presentation
March 5/05
KEYWORDS Territorial Disputes Contiguity Salience
Directional Dyads
2
INTRODUCTION
Problem in IR Consistent empirical findings on
territorial stakes as most conflict-prone for
interstate relations among all contentious issues
(Hensel 2000 2001 Hensel Laughlin 2001,
Vasquez 1993) General Puzzle Do identifiable
attributes of territories lead to an increased
propensity of conflict? Salience as primary
attribute How does territorial salience lead to
increased conflict propensity?
3
THE LITERATURE

• primacy of territorial issues (Hensel et al 1996,
  Vasquez 1993, Holsti 1991)
• gtterritorial issues enduring disputes (Goertz
  and Diehl 1992)
• Territorial issues as zero sum in conflict
  resolution (Huth 2000, Hensel 2004)
• II. Role of Salience in Interstate Disputes
• Conflicting hypothesis and findings
• -Vasquez (1993) posits a positive relationship.
• -Deutsch (1991) posits a negative relationship
• -Starr and Thomas (2001, 2002) posit a parabolic
  relationship.

4
VASQUEZ AND DEUTSCH HYPOTHESIS ON DISPUTE COSTS
BORDER SALIENCE
5
THEORITCAL CONTRIBUTIONSALIENCE REDEFINED

• I. Salience Reconceptualized as Relative Concept
• Incorporate tangible intangible assets (Huth,
  Hensel 2000)
• Implicit assumption of directionality
• Salience A places upon Bs territory ? Salience B
  places upon As territory
• II. Salience as Time-Dependent
• Through the conscious efforts of states,
  objective value of assets may change
• Ex More efficient use of natural resources
• Subjective value of assets may change
• Ex. diversionary conflict literature
  rally-around flag effect. (Levy 92, Gelpi 97)

6
HYPOTHESES

• Hypothesis I.i. Given low leader popularity,
  territorial salience will have a positive effect
  on the likelihood that a state will make a
  territorial challenge.
• Hypothesis .I. ii. Given a moderate level of
  leader popularity, territorial salience will have
  a modest positive impact on the likelihood of
  territorial challenge.
• Hypothesis I.iii. Given high leader popularity,
  territorial salience will exert a negative effect
  on the likelihood of territorial challenge.

7
TYPOLOGY OF ROLE OF SALIENCE
Effect of Salience on Likelihood of Territorial
Challenge
domestic context
8
RESEARCH DESIGN

• Domain of Study Territorial Conflicts in Western
  Hemisphere, 1816-1994
• Unit of Analysis Directional Dyad (84
  cross-sectional units)
• Dataset Composite of Hensel ICOW Territorial
  Claims, Reinhardt Domestic Vulnerability, Huth
  measures of intangible assets
• Dependent Variable ICOW territorial challenge
  dummy
• Primary Independent Variable
• Territorial Salience
• Leader Popularity
• Model BTSCS Methods necessary (Beck and Katz 98)

9
Random Effects Logit Results, 1945-1994
N994 Prob gt Chisquared 0.000
10
Random Effects Logit Results, Temporal Count
Included, 1945-1994
11
Random Effects Logit Results, 1816-1994, Temporal
Count Included
12
IMPLICATIONS

• Effect of salience significant
• Sign of controls in predicted directions
• Direction of effect depends on domestic context
• Salience increases the probability of territorial
  challenge when leaders face low popularity
• Salience decreases the probability of territorial
  challenge when leaders enjoy high popularity
• Salience exerts a less decisive impact at the
  moderate popularity range

13
CAVEATS

• Several caveats arising from structure of BCSTS
  data
• Not enough info contained in binary DV to use AR1
  /lagged dependents choices
  are limited
• Problem of time dependence
• caveats specific to data used in this research
  project
• Data unbalanced, missing data in earlier periods
• Existence of time-invariant variables
• Alternate measures of domestic context necessary

14
FUTURE RESEARCH
Where To Go From Here?

• Ideal is to construct a time variant salience
  variable
• In the meantime several possibilities
• Use alternative measures of salience
• Extend theory beyond territorial issues to all
  contentious issues
• Build on Hensels recent (2001) work on
  settlement of territorial disputes.

15
APPENDIX RANDOM EFFECTS AND FIXED EFFECTS
EQUATIONS

• FIXED EFFECTS MODEL

• Requires fewer assumptions than Random Effects
• Assumes Vi is fixed
• Similar to LSDV
• Coefficients not constant across cases
• Less efficient than R E but more consistent

16
(No Transcript)
17
RANDOM EFFECTS MODEL

• Assumes unit-specific error process more
  restrictive
• Weighted average of within and between estimators
• requires within unit variation
• no correlation between Vi and expected mean of X
• More efficient than FE

18
SPECIFICATION TESTS AND DIAGNOSTICS USED

• Chow Test for Structural Breaks
• 1914 WWI
• 1973 Oil Shock
• 1945 beg of Cold War
• Hausman Test for Specification
• gtgtshowed no systematic difference in coefficients
  between fe and re models
• Breusch and Pagan L-M Test for Specification
• gtcompares RE and FE Vi must not be significantly
  different from zero for RE to be justified.
• gtgtonly available after xtreg
• Test for autocorrelation regressed residuals
  against lagged residuals no significicant
  relationship

19
(No Transcript)