Alliance Proximity and Effectiveness of Extended Deterrence

ABSTRACT

This article examines whether geographical proximity between defensive allies and a protégé increases the likelihood of successful extended general deterrence. I argue that proximate allies are better at making a credible deterrent threat because proximate allies are more willing and able to help alliance partners in times of crisis than distant ones. This claim is theorized by examining how geographical distance influences the international and domestic costs of alliance commitments. The empirical tests reveal that a potential aggressor is less likely to initiate a militarized dispute against a target with proximate defensive allies. Furthermore, I find that the power of proximate defensive allies is a more significant determinant of successful extended general deterrence than that of distance allies. Interestingly, I also find that the constraining effects of geographical distance on the effectiveness of extended general deterrence have been alleviated by advancements in military technologies throughout history.

KEYWORDS:

• alliance commitment
• extended deterrence
• geographical proximity

Acknowledgments

I thank Ashley Leeds for her helpful comments on the initial research idea of this article, and the editor and anonymous reviewers for constructive suggestions. I also thank Matthew Fuhrmann for useful conversations on this research.

Supplemental data

Supplemental data for this article can be accessed on the publisher’s website.

Notes

¹ Siverson and Starr (1991) show that the likelihood of war diffusion increases as potential joiners have warring borders (vis-à-vis opportunity) or/and alliance commitments (vis-à-vis willingness). Joyce and Braithwaite (2013) supplement Siverson and Starr’s (1991) findings by demonstrating that states are more likely to join ongoing conflicts that occur at a location close to their territories.

² The term “defender” does not necessarily mean that a defensive ally is stronger than the protégé (that is, the potential target). The terms “aggressor” and “challenger” are interchangeably used.

³ While Sabrosky (1980) found that about 27% of alliance commitments were honored in times of conflicts, a more recent study with improved alliance data found that almost 75% of alliance commitments were honored (Leeds et al. 2000).

⁴ Huth and Russett, 1988) found that formal alliance is positively related to immediate deterrence failure. Fearon (1994b) offered an explanation for this finding. A potential aggressor decides to issue a military threat conditioned upon ex ante observable factors that convey information about the credibility and power of a target’s defensive allies. Thus, less resolved types of challenger are screened by the existence of (strong) defensive allies at the stage of general extended deterrence, while initiating a dispute (or making a military threat) is likely only for highly resolute types that are also likely to fail immediate deterrence.

⁵ See, for example, Murray (1994); Gray (1999).

⁶ I take into account this possibility later in this section.

⁷ Regarding the effect of “tyranny of distance” on the grand strategy of the United States, see Mearsheimer (2001) and Webb (2007). For an excellent literature review and various examples on the relationship between distance and military operations, see Sakaguchi (2011).

⁸ For example, the cost of Gulf War in 1991, one of the shortest wars fought by the United States, was about 102 billion dollars in 2011 constant US dollar (Daggett 2010). For more realistic estimates, an hour flight of F-22A Rapter costs $68,362, and B-2A Stealth Bomber $169,313 (Thompson 2013). The estimated operating cost of an aircraft carrier strike group is about $6.5 million per day (Hendrix 2013:5).

⁹ “The U.S. Air Force’s peacetime posture in East Asia includes only eight fighter squadrons spread across a handful of regional airbases, all under threat from ballistic and cruise missile attack. Chinese aircraft, on the other hand, operates from dozens of far more secure locations. US forces would necessarily rely on space-based communications, while China could employ more robust ground-based systems. US reinforcements would flow from bases located anywhere from 5,000 to 9,000 miles distant, while Chinese units could deploy directly into battle. Such are the advantages of geography, or the “tyranny of distance” (Heginbotham and Heim 2015:188).

¹⁰ Gleditsch (2002) and Joyce and Braithwaite (2013) also demonstrate that states tend to show a greater willingness to intervene in disputes near their territories than those occurring at a distant location. This claim is also consistent with the findings of Clark and Regan (2003).

¹¹ A similar line of reasoning is also found in Clare (2013) and Huth (1988).

¹² Credibility and power of alliances are not mutually exclusive because alliance credibility is likely to decrease as alliance power declines due to geographical distance. However, this interdependent relationship does not counteract to my theoretical expectation, but strengthens it further.

¹³ Also cited in Diehl (1999) and Vasquez (1993).

¹⁴ I also test if the relatively greater deterrent effect of proximate alliances has significantly decreased since World War I in the appendix.

¹⁵ “World War II accelerated advances in aviation technology that saw production of faster, larger, higher-flying, and longer-range airplanes. Japan’s surrender shortly after the first atomic bombs fell on Hiroshima and Nagasaki demonstrated that air power could be decisive in the outcome of wars. Jet aircraft, ballistic and cruise missiles, pressurized cabins, and radar were all legacies of the war. So too were the introductions of airborne operations, the helicopter as a military vehicle, and global air transport” (Haulman 2003:iii).

¹⁶ MID incident-level data might provide useful information on the distinction between general and immediate deterrence cases, but the data are available only since 1993.

¹⁷ In the sample, about 80% of dyad years include a target with at least one contiguous defensive ally.

¹⁸ For about 19% of dyad years in the sample, a potential target has more than three contiguous allies, and the maximum number of contiguous defensive allies is 10 (Saudi Arabia in 1990).

¹⁹ The data are downloaded using the EUGene program (Bennett and Stam 2000). Inter-capital distance is coded as zero for contiguous allies.

²⁰ Adjusted power = CINC^{log[(distance/distance per day)+(10−e)]}, in which the value of distance is selected by per-day transportation ranges: 250 miles (402 km) for years from 1816 through 1918, 375 miles (604 km) from 1919 to 1945, and 500 miles (805 km) years after 1945. For critiques on this adjusting method, see Diehl (1985), Moul (1988), and Lemke (2002).

²¹ In this particular analysis, I use the (unadjusted) sheer power of defensive alliances because the decomposition process is a method of adjusting the sheer power of defensive allies, which allows me to separate out the geographical effects on the deterrent values of alliance power.

²² The threshold year is 1945. However, the effect of wartime technological advancements might have appeared even during the war, so that I also use a different threshold (1939) to ensure the robustness of the results. Interestingly, I also find evidence that the level of geographical constraints on power projection is significantly different after and before World War I.

²³ Except peace-year polynomials, I expect that these control variables affect both the likelihood of dispute initiation and target states’ incentives to form defensive alliances. To see how control variables affect the coefficient estimates of the key independent variables, I also run a bivariate analysis while keeping peace-year polynomials without which bias driven by temporal dependence is expected to arise. I find similar results, as shown in Table I1 in the appendix.

²⁴ This variable is constructed by dividing the weaker state’s CINC score by the sum of CINC score for both challenger and target in a given dyad year. It ranges between 0 and 0.5, and higher values indicate greater power parity.

²⁵ A state is classified as democracy when the state’s polity score is higher than six, and a dyad is treated as joint democracy when both the challenger and the target in the given dyad are democracies.

²⁶ These peace-year polynomials are included to capture temporal dependence in the binary time-series cross-sectional data.

²⁷ These variables are taken from the replication data set of Johnson et al. (2015).

²⁸ This variable is found to have a significant positive effect on the likelihood of extended general deterrence success (Johnson et al. 2015). It is operationalized by (Target CINC score + Allies’ CINC scores)/(Target CINC score + Allies’ CINC scores + Challenger CINC score).

²⁹ This variable is measured by the weighted squared distances between the target’s and defensive allies’ alliance portfolio. Alternative measures that capture the same concept, using UN voting records, are also available (Häge 2011). Using these alternative measures does not change the main findings substantially, but restricts the temporal dimension of my analysis to the post-World War II period.

³⁰ The 95% confidence intervals range from 1.72 to 2.01 times.

³¹ It is equivalent to an increase from 287 miles to 2854 miles.

³² The 95% confidence intervals range from 14% to 23%.

³³ This result may be driven by the fact that strong targets may not need much support from allies for deterrence in the first place, which mitigates the deterrent effect of defensive alliance power. Thus, I reran the model including target’s sheer capabilities and found that the main findings in Table 2 do not change. See Table L1 in the appendix.

³⁴ The correlation between these two different power measures is −0.197.

³⁵ This odd finding is also shown in Johnson et al. (2015:330). As they also show, once I lower the democracy threshold to five or higher, joint democracy gains statistical significance (p = .004).

³⁶ A recent study by Kenwick et al. (2015) find that deterrent effects of defensive alliances differ in the pre- and postnuclear ear.

³⁷ The results are presented in Table E1 in the appendix.

³⁸ The results are presented in Tables A1–A3 in the appendix.

³⁹ The results are presented in Tables B1–B3 in the appendix.

⁴⁰ The results are presented in Table C1 in the appendix.

⁴¹ The results are presented in Table J1 in the appendix.

⁴² See Table H1 in the appendix. The deterrent effect of geographical proximity holds controlling for US troop deployment. Military bases may capture the sunk costs of the defender, which makes the defender more willing to help the protégé. Another measure that captures the sunk costs would be the level of peace time military coordination. I checked the robustness of the results including the “peacetime military coordination” measure that was created by Leeds and Anac (2005) and used by as Johnson et al. (2015) . Table M1 shows that the main findings about geographical proximity remain the same.

⁴³ The data on the number of defensive allies with ballistic missiles in a given target-country-year are drawn from Mettler and Reiter (2012). The results are presented in Table F1 in the appendix.