Literature Review: Arms Race and Michael Horn

• Horn, Michael Dean. 1987. "Arms Races and the International System." PhD dissertation. Rochester, NY: Department of Political Science, University of Rochester.
• Richardson, Lewis F. 1960. Arms and Insecurity. Pacific Grove, CA: Boxwood Press.
• Huntington, Samuel P. 1958. "Arms races: prerequisites and results." Public Policy 8: 41-86.
• Wallace, Michael D. 1979. "Arms races and escalation: some new evidence." Journal of Conflict Resolution 23: 3-16.
• Morrow, James. 1989. "A Twist of Truth: A Reexamination of the Effects of Arms Races on the Occurrence of War." Journal of Conflict Resolution 33(3): 500-529.

Horn (1987) offers an independent means of determining when two countries are engaged in an arms race. His relatively straightforward measure has two qualifications for definition as an arms race. First, that the growth rates of a country's military expenditures are on average higher in the period preceding a dispute than in the whole period under study (for him, the Correlates of War period from 1816 to 1980). Second, in order to be classified as an arms race, the growth rate must be higher in the second half of the period than in the first. In this, he attempts to include the criterion of acceleration which is so vital to the Richardsonian distinction between stable and unstable arms races (Richardson, 1960, pp. 74-75). Only when growth is speeding up or seemingly out of control is there an arms race.

He finds that of longer arms races (he looks at Huntington's twelve-year period, cf. Huntington, 1958) over one-half end in war (Horn, 1987, p. 56). Experimenting with a shorter time period, Richardson's suggested six years (Richardson, 1960) yielded no significant relationship. He concludes from this that an arms race over the longer period is indicative of a continuing conflict of interests and/or hostile relationship between countries, but the shorter measure could easily catch incidental or shortterm conflicts that never reach the level of war.

He determines that arms races are considerably rarer than others have thought, but they do have some relation to the escalation of disputes to war, though his findings are less robust than Wallace's (Horn, 1987, p. 60; Morrow, 1989, p.502).

From: Sample, Susan G. 1997. "Arms Races and Dispute Escalation: Resolving the Debate." Journal of Peace Research 34: 7-22.

Literature Review: Arms Race, Michael D. Wallace, Erich Weede, Michael F. Altfield, and Paul F. Diehl

• Wallace, Michael D. 1979. "Arms races and escalation: some new evidence." Journal of Conflict Resolution 23: 3-16.
• Weede, Erich. 1980. "Arms Races and Escalation: Some Persisting Doubts." Journal of Conflict Resolution 24(2): 285-287.
• Wallace, Michael D. 1980. "Some Persisting Findings: A Reply to Professor Weede." Journal of Conflict Resolution 24(2): 289-292.
• Altfield, Michael F. 1983. "Arms Races? — And Escalation? A Comment on Wallace." International Studies Quarterly 27(2): 25-231.
• Diehl, Paul F. 1983a. "Arms Races and Escalation: A Closer Look." Journal of Peace Research 20(3): 205-212.
• Diehl, Paul F. 1983b. "Arms Races and the Outbreak of War, 1816-1980." PhD dissertation. Ann Arbor, Ml: Department of Political Science, University of Michigan.
• Diehl, Paul F. 1985. "Arms Races to War: Testing Some Empirical Linkages." Sociological Quarterly 26(3):  331-349.

In 1979, Michael Wallace published a pathbreaking study in which he found a very strong relationship between the existence of an ongoing arms race between two countries and the escalation of a militarized dispute between them to war. Looking at 99 militarized disputes among the major states between 1816 and 1965, he found that 23 of the 26 disputes that escalated to war were characterized by an ongoing arms race (Q = 0.98 and phi = 0.80). In other words, if an arms race was going on between two countries, there was an 82% chance that a militarized dispute between them went to war. On the other hand, only about 4% of the disputes that occurred when both countries were not arming at abnormally high rates escalated to war (Wallace, 1979).

Wallace's conclusions were an empirical challenge to many long-held assumptions about politics among nations, including some of the foundations of deterrence theory. However, as Weede asserted, Wallace's results were not inconsistent with the para bellum hypothesis: 'according according to most realist or conservative thinking in the West... the most dangerous occurrence is not a runaway arms race but the status quo powers losing the arms race' (Weede, 1980, pp. 286-287). In other words, if a revisionist power strives to overturn the balance of capabilities, it is likely that both the revisionist and the status quo states will show high growth rates, but the relevant change is that of the power balance, not the rate of arming. In answer, Wallace used the ratio of expenditure growth rates between revisionist and status quo states to determine who 'won' the race.

There was some small difference in escalation between a 'victory' on the part of the status quo state or the revisionist one, but the relationship is much weaker than Wallace's original test (Wallace, 1980). A further test, using a static measure of comparative capabilities (the ratio of absolute military expenditures between the revisionist and status quo states) rather than the dynamic ratio of growth rates, achieved even weaker results for the para bellum hypothesis (Wallace, 1981). These findings lend support to Wallace's conclusion that it is the arms race itself, rather than the power balance, which is dangerous.

The two supports of Wallace's study, the disputes he used and the index he employed, both suffered massive criticism. The problem with the sample was that the strength of the findings seemed dependent on the division of the obviously multilateral World Wars into dispute dyads, hence overweighting the arms races that preceded those wars. If a small number of arms races make up the bulk of the supportive findings, it is evident that the certainty with which one is able to make general conclusions about the relationship between arms races and war is undermined.

The index that Wallace used to determine the existence of an arms race was also challenged. On theoretical grounds, it has been attacked as placing too great an emphasis on the last two years of military spending before a crisis, therefore capturing war preparations rather than an independent arming process. A further problem is that the index is multiplicative - the final index score for the dyad is arrived at through multiplying individual country scores together (Altfeld, 1983). It is conceivable that heavy unilateral buildups could be mistaken for arms races. Certainly in the former case, but possibly in the latter as well, conclusions drawn about the relationship between arms racing and dispute escalation might be overstated. A truly complete study would demand that the index be replicated, but that has yet to be done...

Diehl's work (1983a, b, 1985) on the subject of the relation of arms race and dispute escalation is an independent alternative to that of Wallace (1979, 1980, 1982, 1983, 1990). Although he criticizes Wallace at length, he does remain focused on the real issue by offering an alternative means of testing the identical proposition. Unlike Michael Wallace, he found little connection between arms races and dispute escalation. Fewer than 30% of the disputes that he finds occurring during an arms race escalate to war. The corresponding correlation coefficients are very different: Yule's Q drops from 0.98 to 0.42, and phi is an unimpressive 0.13 (Diehl, 1985, p. 338).

Diehl objects to Wallace's findings on both grounds: that his dispute sample was overweighting the World Wars, and that reliability and validity of his index were questionable. Therefore, he offers both a new index and a new set of disputes. In doing this, the criticisms of Wallace's work fail to be compelling because it is impossible to determine the extent of the effects of each weakness. We can certainly guess that both the dispute set and the index are flawed, but without being able to compare them to anything it is impossible to tell whether or not the assertion is true, or, more importantly, what particular influence these factors have had on the conclusions Wallace has drawn. Without such a comparison, it is impossible not to consider the possibility that Wallace has just found a superior means of dividing dangerous from non-dangerous arms buildups.

From: Sample, Susan G. 1997. "Arms Races and Dispute Escalation: Resolving the Debate." Journal of Peace Research 34: 7-22.

Literature Review: Tinderbox Hypothesis, Michael D. Wallace, Paul F. Diehl, and Mike Horn

• Wallace, Michael D. 1979. "Arms races and escalation: some new evidence." Journal of Conflict Resolution 23: 3-16.
• Wallace, Michael D. 1982. "Armaments and escalation." International Studies Quarterly 26: 37-56.
• Diehl, Paul F. 1983. "Arms races and escalation: a closer look." Journal of Peace Research 20: 205-212.
• Horn, Mike. 1984. "Arms races and the likelihood of war." Presented at the annual meeting of the International Studies Association, Atlanta, GA, March 27-31.

Wallace (1979, 1982) presented the first attempt to develop a comprehensive list of arms races and test the arms race hypothesis. He constructed his list of arms races by examining the military expenditures of participants in serious disputes, specifically those involving at least a threat to employ force prior to those disputes. Wallace's arms race index attempted to capture not only the growth in military expenditures of an arms race but also the underlying conflict between the racing parties by searching for arms races prior to serious disputes. However, this procedure would miss any arms race that did not end in a serious dispute, and consequently cannot provide a complete test of the arms race hypothesis. Instead, Wallace tested whether arms races made escalation to war more likely, or what could be termed the "tinderbox hypothesis." The tinderbox hypothesis states that arms races do not necessarily cause wars, but they do create an inflammable situation between the racing nations where even the slightest spark can

push a blaze to war.

Wallace found impressive results supporting the tinderbox hypothesis: 23 of the 28 disputes preceded by an arms race escalated to war. However, Wallace's determination of arms races and his accumulation of the results suffer from two methodological flaws: His arms race index may have been misleading, and he decomposed multilateral disputes into separate dyads (see Weede, 1980; Altfeld, 1983; and responses by

Wallace, 1980, 1983).

Diehl (1983) and Horn (1984) reanalyzed the tinderbox hypothesis using the same data as Wallace but different arms race indices, and without decomposing each dispute into separate dyads. Both required both sides' average growth in military expenditures to exceed certain thresholds over a set period of time prior to the dispute. Dichl (1983) found that only 3 of 12 disputes preceded by an arms race escalated to war. Horn (1984) found that at most 60% (9 out of 15) of 6-year arms races and 46% (6 out of 13) of 12-year arms races ended in war. These reanalyses suggest that the tinderbox hypothesis is only partially true; disputes preceded by arms races do escalate to war more often than other disputes, but they do not overwhelmingly escalate to war.

From: Morrow, James D. 1989. "A Twist of Truth:  A Reexamination of the Effects of Arms Races on the Occurrence of War."  Journal of Conflict Resolution 33(3): 500-529.