![Sample our Politics & International Relations journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5947/TOC-Subject-Sample-2021-Politics-International-Relations.jpg"})
ABSTRACT
In counterinsurgency, U.S. officials often feel trapped by a localally who appears unable to survive the departure of U.S. forces. Advocates for withdrawal argue that only a deadline to depart will induce local governments to accept greater burdens, while critics of this position argue that plans to withdraw embolden insurgents. We argue instead that American leaders gain leverage from U.S. public opinion favouring withdrawal. Analysis of 200+ U.S. demands of local allied governments in Vietnam and Iraq suggests that public pressure for withdrawal is associated with greater local compliance, but that formal U.S. withdrawal plans does not motivate compliance.
Disclosure Statement
No potential conflict of interest was reported by the author(s).
Notes
1. There are exceptions: Christia (Citation2012) is a foundational study of alliances in civil wars, while several studies examine decisions to withdraw from interstate coalitions during wars (Choi Citation2012, Tago Citation2009; Author, Citation2016). None of this work, however, examines changes in the political influence of external allies arising from (threats of) withdrawal from military intervention.
2. Some readers may wonder about the extent to which the insurgent strategy can influence cooperation between the local regime and the external counterinsurgents. The insurgents obviously are a critical strategic actor in the war, and in principle could attempt to woo the local regime away from the foreign intervener. In practice, however, the central dispute in counterinsurgencies concerns whether the local government or the insurgents will ultimately hold power; as such, the insurgents have comparatively few options for driving a wedge between the local regime and the external counterinsurgent short of continuing to fight an effective insurgency in order to exhaust the alliance between local and external counterinsurgents. As such, we do not examine the insurgents’ strategy closely in this article.
3. Indeed, as Goldstein (Citation1995) points out, there is good reason not to free ride on existential issues such as regime survival or, in his particular focus, nuclear conflict.
4. This point in particular was the reason for not including Afghanistan. Data on U.S. demands of the Afghan government ends in 2010 (the end of the period covered by the Wikileaks Cablegate data release), prior to serious moves to withdraw from Afghanistan. Moreover, public opinion data, which we use to capture credible pressure to withdraw, is surprisingly limited. In particular, there were no polls prior to 2008 that asked respondents about withdrawal from Afghanistan. Even the alternative of using data on whether the war was a mistake – which we use in robustness checks – is unhelpful here because the question was not asked between 2004 and 2007, and because opinion did not vary substantially over the period we are able to analyse.
5. For a more on the data collection process, see [Elias Citation2020].
6. Some readers may wonder whether we could develop a more fine-grained measure of withdrawal based on troop commitments. There are two problems with such an approach, however. First, the argument of withdrawal advocates suggests that it is the commitment to withdraw more than the stage in the withdrawal process that concentrates minds in the local government. Second, and more importantly, this approach conflates low troop commitments late in the withdrawal process with low troop commitments early in the conflict, when the external ally is politically committed but either underestimates the scale of the challenge or is still ramping up the intervention.
7. American officials acknowledged public pressure for withdrawal following Tet, yet assured their Vietnamese counterparts that the U.S. was ‘determined and unflappable’ (U.S. Department of State, Office of the Historian Citation1968).
8. Minor variations in question wording exist across surveys. In roughly half the surveys, the initial ‘In your opinion’ was omitted, while in four surveys the question refers to ‘the Vietnam situation’ instead of simply to Vietnam. There is no reason to think that these wording changes substantially influence responses. By contrast, we omit one survey from December 1968 that asks, ‘What would you like to see [President-elect] Nixon do about the Vietnam situation?’ as the reference to Nixon is likely to activate partisan identities and to cue different timeframes, with respondents frequently considering Nixon’s entire term in office rather than the more immediate future.
9. The authors coded these categories independently, with information about survey date and response frequency removed. We then had an independent coder perform the same codings for the small number of cases where initial decisions differed, and adopted the third codings for these cases. We thank Jessica Stanton for providing the third codings.
10. This question was not asked in three years in our sample. For years in the middle of the conflict, we use interpolated values for public opinion – we experimented with several interpolation techniques, with no substantive change to the results. For 1973, the primary measure simply uses the value from 1972. Substituting an extrapolated value produced analogous results.
11. ‘Which comes closest to your view about what the U.S. (United States) should now do about the number of U.S. troops in Iraq – the U.S. should send more troops to Iraq, the U.S. should keep the number of troops as it is now, the U.S. should begin to withdraw some troops from Iraq, or the U.S. should withdraw all of its troops from Iraq?’.
12. The correlation coefficient between support for withdrawal and with agreement that intervention was a mistake is 0.94 in Vietnam and 0.83 in Iraq.
13. For more information on this coding process, see [Author Citation2017].
14. Note that this measure is prospective, reflecting assessments that capacity limitations were likely to limit the local government’s ability to comply, though in some cases the government managed to comply despite the limitations.
15. Results are robust to controlling for years in which the local government conducted a national election.
16. While in principle it would be preferable to use narrower time windows, such as monthly observations, in practice limitations arising from both American tracking of demands and the difficulty of determining the exact timing of compliance for some demands means that we are confident of the timing of compliance only at the level of the year.
17. Checks of the proportional hazard assumption revealed no evidence of violations.
Additional information
Notes on contributors
Barbara Elias
Barbara Elias is Assistant Professor of Government at Bowdoin College specializing in international relations, alliance politics, counterinsurgency, national security and U.S. foreign policy. Her book, Why Allies Rebel: Defiant Local Partners in Counterinsurgency Wars was published by Cambridge University Press in 2020.
Alex Weisiger
Alex Weisiger is an associate professor of politicalscience at the University of Pennsylvania. His published work examines war termination, reputation in international politics, alliance politics, and the democratic peace. His book, Logics of War: Explanations for Limited and Unlimited Conflicts, was published by Cornell University Press in 2013.