Testing Alternative Models of Alliance Duration, 1816-1984

At least four models of alliance duration exist, namely the capability-aggregation, security-autonomy, domestic politics, and institutionalization models of alliances. The independent tests of these models have been underspecified, however, as no simultaneous comparative testing of the hypotheses fr...

Full description

Saved in:
Bibliographic Details
Published in:American journal of political science 1997-07, Vol.41 (3), p.846-878
Main Author: Bennett, D. Scott
Format: Article
Language:English
Subjects:
Aggregation
Alliances
Balance of power
Cold War
Datasets
Hypotheses
Hypothesis testing
Institutionalization
International alliances
International politics
International relations
Military alliances
Modeling
Political science
Politics
Threats
Treaties
Variables
War
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:At least four models of alliance duration exist, namely the capability-aggregation, security-autonomy, domestic politics, and institutionalization models of alliances. The independent tests of these models have been underspecified, however, as no simultaneous comparative testing of the hypotheses from these models has been conducted. Alliance duration may be a function of security threats, changes in national capabilities, the security-autonomy tradeoffs between larger and smaller states in an alliance, regime type and changes in regime, and the time the alliance has had to become institutionalized. A hazard model employing time-varying covariates applied to 207 alliances from 1816 to 1984. Key elements of the security-autonomy and domestic politics models have effects on alliance duration as expected. It does not appear, however, that institutionalization over time has an effect on alliance duration, as alliances appear to become more likely to dissolve the longer they last. Additionally, while one hypothesis advanced by the capability-aggregation model receives some support, a second variable is statistically significant in the opposite direction of the model's expectation, suggesting possible problems with the model's validity.
ISSN:0092-5853
1540-5907
DOI:10.2307/2111677