How the Weak Win Wars

No one had given Muhammad Ali a chance against George Foreman in the World Heavyweight
Championship Žght of October 30, 1974. Foreman, none of whose opponents
had lasted more than three rounds in the ring, was the strongest, hardest hitting
boxer of his generation. Ali, though not as powerful as Foreman, had a
slightly faster punch and was lighter on his feet. In the weeks leading up to the
Žght, however, Foreman had practiced against nimble sparring partners. He
was ready. But when the bell rang just after 4:00 a.m. in Kinshasa, something
completely unexpected happened. In round two, instead of moving into the
ring to meet Foreman, Ali appeared to cower against the ropes. Foreman, now
conŽdent of victory, pounded him again and again, while Ali whispered
hoarse taunts: “George, you’re not hittin’,” “George, you disappoint me.” Foreman
lost his temper, and his punches became a furious blur. To spectators, unaware
that the elastic ring ropes were absorbing much of the force of
Foreman’s blows, it looked as if Ali would surely fall. By the Žfth round, however,
Foreman was worn out. And in round eight, as stunned commentators
and a delirious crowd looked on, Muhammad Ali knocked George Foreman to
the canvas, and the Žght was over.

The outcome of that now-famous “rumble in the jungle” was completely unexpected.
The two Žghters were equally motivated to win: Both had boasted of
victory, and both had enormous egos. Yet in the end, a Žght that should have
been over in three rounds went eight, and Foreman’s prodigious punches
proved useless against Ali’s rope-a-dope strategy.
This Žght illustrates an important yet relatively unexplored feature of interstate
conict: how a weak actor’s strategy can make a strong actor’s power ir-
How theWeak
WinWars
Ivan Arreguín-Toft
A Theory of Asymmetric Conict
International Security, Vol. 26, No. 1 (Summer 2001), pp. 93–128
© 2001 by the President and Fellows of Harvard College and the Massachusetts Institute of Technology.
Ivan Arreguín-Toft is a postdoctoral fellow in the International Security Program at the Belfer Center for
Science and International Affairs at Harvard University’s John F. Kennedy School of Government.
The author wishes to thank Stephen Biddle, David Edelstein, John Garafano, Charles Glaser, John
Mearsheimer, SharonMorris, Jordan Seng, Jessica Stern, Ward Thomas, Monica Toft, Stephen Walt,
participants in the Program on International Security Policy at the University of Chicago, participants
in the 1999–2000 National Security Seminar at the John M. Olin Institute for Strategic Studies
at Harvard University, and three anonymous reviewers for their helpful comments. He also gratefully
acknowledges support from the Institute for the Study of World Politics and the Smith Richardson
Foundation.
relevant.1 If power implies victory in war, then weak actors should almost
never win against stronger opponents, especially when the gap in relative
power is very large.2 Yet history suggests otherwise: Weak actors sometimes
do win.3 The question is how.
Understanding the conditions under which weak actors win wars is important
for two reasons. First, if there are dynamics unique to asymmetric
conicts—or if their analysis provides fresh insights into symmetrical
conicts—a general explanation of asymmetric conict outcomes is not only
desirable but necessary, both to reduce the likelihood of unwinnable wars and
to increase the chances of U.S. success when a resort to arms is necessary. Second,
because asymmetric conicts ranging from catastrophic terrorism to military
intervention in interstate, ethnic, and civil wars are the most likely threat
to U.S. security and interests, only a general theory of asymmetric conict outcomes
can guide U.S. policymakers in their efforts to build the kinds of armed
and other forces necessary to implement an effective U.S. strategic response.
Thus far, only one scholar has advanced a strong general explanation of
asymmetric conict outcomes.4 In “Why Big Nations Lose Small Wars,” Andrew
Mack argues that an actor’s relative resolve or interest explains success
or failure in asymmetric conicts.5 In essence, the actor with the most resolve
International Security 26:1 94
1. “Actors” here mean states or coalitions of states, although the same dynamics would apply to
governments Žghting against rebels. “Conicts” in this analysis are restricted to wars (1,000 battle
deaths per year on average), although again, similar dynamics may apply in conicts that are not
wars (e.g., terrorism, trade wars, and labor disputes). Because this analysis focuses on explaining
asymmetric conicts, it excludes those few wars in which the ratio of forces changed dramatically
(toward symmetry) between the conict’s beginning and end.
2. In this article I follow established practice by introducing an imperfect but quantiŽable proxy
for power. A strong actor is one whose material power exceeds that of its adversary or adversaries
by at least ten to one. Material power is the product of a given state’s population and armed forces.
“Strong” and “weak” therefore have meaning only in a particular dyadic context: Italy in 1935 is
weak compared with Nazi Germany, but strong compared with Ethiopia. For an introduction to
the literature on empirical and quantiŽable measures of relative power, see John Jacob Nutter,
“Unpacking Threat: AConceptual and Formal Analysis,” in Norman A. Graham, ed., Seeking Security
and Development: The Impact of Military Spending and Arms Transfers (Boulder, Colo.: Lynne
Rienner, 1994), pp. 29–49.
3. T.V. Paul, Asymmetric Conicts: War Initiation by Weaker Powers (New York: Cambridge University
Press, 1994), p. 4. Paul’s analysis focuses on why weak states initiate wars against stronger
ones. His threshold of analysis for asymmetry is a power ratio of 1:2, where power is measured in
traditional—that is, material—terms. On Paul’s deŽnition of asymmetry, see ibid., p. 20. On the
problems of reducing power to material net assessments, see ibid., p. 22; Nutter, “Unpacking
Threat”; and Michael P. Fischerkeller, “David versus Goliath: Cultural Judgments in Asymmetric
Wars,” Security Studies, Vol. 7, No. 4 (Summer 1998), pp. 1–43.
4. See Andrew J.R. Mack, “Why Big Nations Lose Small Wars: The Politics of Asymmetric
Conict,” World Politics, Vol. 27, No. 2 (January 1975), pp. 175–200.
5. Robert A. Pape considers Mack’s explanation as part of a family of balance-of-interest arguments,
which closely approach balance-of-resolve arguments. I agree that they are analytically diswins,
regardless of material power resources. Mack contends that this resolve
can be derived a priori by assessing the structure of the conict relationship.
Power asymmetry explains interest asymmetry: The greater the gap in relative
power, the less resolute and hence more politically vulnerable strong actors
are, and the more resolute and less politically vulnerable weak actors are. Big
nations therefore lose small wars because frustrated publics (in democratic regimes)
or countervailing elites (in authoritarian regimes) force a withdrawal
short of military victory. This seems true of some conicts, but not of others.
In this article I argue that the best predictor of asymmetric conict outcomes
is strategic interaction.6 According to this thesis, the interaction of actor strategies
during a conict predicts conict outcomes better than do competing explanations.
7 The Žrst section lays out the puzzle of strong-actor defeat in
asymmetric conicts and Mack’s interest asymmetry argument more fully. The
second section introduces the strategic interaction thesis, which holds that
strong actors will lose asymmetric conicts when they use the wrong strategy
vis-à-vis their opponents’ strategy. The next two sections offer quantitative and
qualitative tests of the argument. The article concludes by drawing out theoretical
and policy implications of the strategic interaction thesis and suggests avenues
for further research.
How the Weak Win Wars 95
tinct, but this distinction does not matter in practice because the two motives invariably move in
the same direction: High interest implies high resolve, and vice versa. I therefore use the two terms
interchangeably. See Pape, Bombing to Win: Air Power and Coercion in War (Ithaca, N.Y.: Cornell University
Press, 1996), p. 4, n. 9.
6. My use of the term strategic interaction should not be confused with its usage in formal models
in which strategy is invoked as both a proxy for actor rationality (i.e., strategic actors calculate
risks and beneŽts) and an attribute of states, especially great powers. See, for example, Curtis S.
Signorino, “Strategic Interaction and the Statistical Analysis of International Conict,” American
Political Science Review, Vol. 93, No. 2 (June 1999), p. 279. On the distinction between rational choice
treatments of strategy and strategy as traditionally used in strategic studies, see Pape, Bombing to
Win, pp. 8–9.
7. Among these are nature of actor, arms diffusion, and rise of nationalism arguments. The logic of
the nature of actor argument is that because (1) democratic actors do not Žght as well as authoritarian
actors, and (2) an increase in strong-actor failures in asymmetric conicts correlates with an
increase in the total number of asymmetric conicts involving democratic strong actors, actor regime
type explains asymmetric conict outcomes. See Michael C. Desch, “Democracy and Victory:
Why Regime Type Does Not Matter (Much),” unpublished manuscript, University of Kentucky,
September 2000. The arms diffusion argument is a strand of offense-defense theory that holds that
the diffusion of relatively advanced small arms to the developing world caused the costs of conquest
or occupation to rise sufŽciently to offset expected beneŽts. On offense-defense theory in
general, see Sean M. Lynn-Jones, “Offense-Defense Theory and Its Critics,” Security Studies, Vol. 4,
No. 4 (Summer 1995), pp. 660–691. The rise of nationalism argument holds that after World War II,
nations came to believe that only through self-determination could they prevent colonialist destruction
of their social fabric. In this view, the costs of conquest and occupation rose because nationalists
are generally more stubborn, and more willing than others, to risk death in pursuit of
autonomy. See, for example, Eric R. Wolf, Peasant Wars of the Twentieth Century (New York: Harper
and Row, 1973).
Explaining Asymmetric Conict Outcomes
Since Thucydides, the root principle of international relations theory has been
that power implies victory in war.8 Thus, in asymmetric conicts9 the strong
actor should almost always win. Indeed this expectation is on balance supported
(see Figure 1). Yet if one divides the roughly 200-year period covered in
the Correlates of War data set, two related puzzles emerge (see Figure 2). First,
weak actors were victorious in nearly 30 percent of all asymmetric wars, which
seems high given the ³ 5:1 asymmetry represented here. Second, weak actors
have won with increasing frequency over time. If relative power explains outcomes,
and structure of the conict is held constant as in Figure 2, conict outcomes
should not shift over time as they have.What explains both strong-actor
defeat in asymmetric wars and the trend toward increasing weak-actor victories
over time?
interest asymmetry, or relative power revisited
Andrew Mack’s explanation for how weak states win asymmetric wars comprises
three key elements: (1) relative power explains relative interests; (2) relative
interests explain relative political vulnerability; and (3) relative vulnerability
explains why strong actors lose. According to the logic of this argument,
strong actors have a lower interest in winning because their survival is not at
stake. Weak actors, on the other hand, have a high interest in winning because
only victory ensures their survival.10 Mack introduces the concept of political
vulnerability to describe the likelihood that an actor’s public (in democratic
International Security 26:1 96
8. Throughout this article the term international relations theory refers to a simple version of realist
theory with three key elements: (1) there is no authority above states that is capable of regulating
their interactions; (2) all states have some capacity to harm other states; and (3) states therefore
seek to increase their relative power, which can deter other states from launching attacks, intimidate
them into making concessions, or defeat them in war.
9. In this analysis the threshold of asymmetry that matters is ³ 5:1 in favor of strong actors, where
power is the halved product of a strong actor’s armed forces and population at the start of a
conict versus the simple product of the weak actor’s armed forces and population. Halving
strong-actor power simulates the tendency of strong actors to have diverse security interests and
commitments. Data for this survey come primarily from J. David Singer and Melvin Small, Resort
to Arms: International and Civil Wars, 1816–1980 (Beverly Hills, Calif.: Sage, 1992). Additional data
are from Walter Laqueur, Guerrilla: A Historical and Critical Study (Boston: Little, Brown, 1976); and
John Ellis, From the Barrel of a Gun: A History of Guerrilla, Revolutionary, and Counter-Insurgency Warfare,
from the Romans to the Present (London: Greenhill, 1995). Cases used in this analysis are listed
in the appendix.
10. The classic work on this point is Glenn H. Snyder and Paul Diesing. See Snyder and Diesing,
Conict among Nations: Bargaining, Decision Making, and System Structure in International Crises
(Princeton, N.J.: Princeton University Press, 1977), p. 190. See also Mack, “Why Big Nations Lose
Small Wars,” p. 181.
regimes) or competing elites (in authoritarian regimes) will force its leaders
to halt the war short of achieving its initial objectives.11 A strong actor’s low
interests imply high political vulnerability. In contrast, a weak actor’s high
interests imply low political vulnerability. Mack argues that this political
vulnerability explains why the strong lose to the weak:12 Delays and reverses
on the battle-Želd will eventually encourage war-weary publics or greedy
elites to force the strong actor’s leaders to abandon the Žght. Mack’s argument
therefore reduces to the claim that relative power explains strong-actor
defeat in asymmetric wars: Power asymmetry determines interest asymmetry
(high power equals low interest). Interest asymmetry is the key
How the Weak Win Wars 97
11. Ibid., pp. 180–182.
12. Ibid., pp. 194–195.
Figure 1. Percentage of Asymmetric Conflict Victories by Type of Actor, 1800± 1998.
Figure 2. Percentage of Asymmetric Conflict Victories by Type of Actor in Four Fifty-
Year Periods.
causal mechanism, and Mack’s thesis is in this sense an interest asymmetry
argument.
Mack applies this logic to the case of U.S. intervention in Vietnam, where it
appears to provide a strong explanation of that war’s unexpected outcome.
According to Mack, the United States lost the war because it had less at stake
than did North Vietnam. Over time the United States failed to coerce North
Vietnam and was eventually forced by an angry and frustrated American public
to withdraw short of achieving its main political objective: a viable, independent,
noncommunist South Vietnam.
Mack’s interest asymmetry thesis has at least three problems. First, relative
power is a poor predictor of relative interest or resolve in peace or war. In
peacetime, a strong state may act as if its survival is at stake when it is not. A
state that imagines itself “leader of the free world,” for example, might rationally
calculate that although the defeat of an ally in a distant civil war would
be materially insigniŽcant, its own survival as a free-world leader depends on
a favorable outcome. These calculations are often intensiŽed by domino logic,
in which a series of individually insigniŽcant interests are linked so that their
cumulative loss constitutes a material threat to survival. Prior to the South African
War (1899–1902),13 for example, Great Britain had calculated that the fate
of its empire hinged on protecting India, which demanded that Britain secure
the sea-lanes of communication passing the Cape of Good Hope. This in turn
required control of Cape Colony, which made it imperative to resist with force
the independence demands of two tiny republics in the hinterlands of the
southernmost region of Africa. Similarly, both identity survival and domino
rationales inuenced the U.S. decision to intervene in the civil war in Vietnam.
14 Once strong actors enter into a conict—even one acknowledged to
have been initially peripheral to their interests—their resolve to win may increase
dramatically. This was as true of Soviet calculations in Afghanistan as it
was of U.S. calculations in Vietnam.15 Second, the operation of political vulnerability,
which Mack uses to explain weak-actor success, presupposes a span of
International Security 26:1 98
13. The South African War is what historians formerly called the Boer War. The new name more
accurately reects the important role played by black South Africans in the war, and for this reason
I refer to the Boer War as the South African War throughout this article.
14. See, for example, George C. Herring, America’s Longest War: The United States and Vietnam,
1950–1975, 2d ed. (New York: McGraw-Hill, 1986), p. 170; and Stanley Karnow, Vietnam: A History
(New York: Viking, 1983), pp. 169–170, 377–378, 399, 423.
15. On Soviet calculations in Afghanistan, see Ralph H.Magnus and Eden Naby, Afghanistan: Mullah,
Marx, and Mujahid (Boulder, Colo.: Westview, 1998), pp. 68, 122. For particularly apt examples
from the U.S. intervention in Vietnam, see Herring, America’s Longest War, p. 222.
time. But nothing in the interest asymmetry thesis explains why some asymmetric
conicts end quickly, yet others drag on.16 Third, if the interest asymmetry
thesis is right, there should be little or no variation over time in the
distribution of asymmetric conict outcomes when relative power is held constant.
But as shown in Figure 2 (in which relative power is held constant),
weak actors are increasingly winning asymmetric conicts.
In sum, Mack’s interest asymmetry thesis is weakest when explaining actor
interests as a function of relative power, and strongest when explaining strongactor
failure as a consequence of political vulnerability. In the next section I
present a theory of asymmetric conict outcomes that subsumes Mack’s thesis
by bracketing the conditions under which political vulnerability causes strong
actors to lose asymmetric wars.
The Strategic Interaction Thesis: A Theory of Asymmetric Conict
This section introduces the strategic interaction thesis as a general explanation
of asymmetric conict outcomes. It begins with deŽnitions of key terms, followed
by an exploration of the theory’s logic, and concludes with several derived
hypotheses.
strategy
Strategy, as deŽned here, refers to an actor’s plan for using armed forces to
achieve military or political objectives.17 Strategies incorporate an actor’s understanding
(rarely explicit) about the relative values of these objectives.18 In
How the Weak Win Wars 99
16. Mack recognizes this problem explicitly, by suggesting that guerrilla warfare strategy explains
the longer duration of asymmetric conicts. See Mack, “Why Big Nations Lose SmallWars,” p. 195.
But this argument suffers from two related problems: (1) weak actors do not always defend with
guerrilla warfare (this limits the generality of Mack’s theory); and (2) some defenders using guerrilla
warfare strategy are defeated quickly (this limits his theory’s explanatory power).
17. The meaning of strategy is both complicated and constantly evolving. John J. Mearsheimer
uses perhaps the simplest deŽnition—”the plan of attack.” See Mearsheimer, Conventional Deterrence
(Ithaca, N.Y.: Cornell University Press, 1983), pp. 28–29. For a discussion of strategy and its
evolution, see B.H. Liddell Hart, Strategy, 2d ed. (New York: Praeger, 1967), pp. 333–346; and J.P.
Charnay, “Strategy,” in André Corvisier and John Childs, eds., Chris Turner, trans., A Dictionary of
Military History and the Art of War (Cambridge, Mass.: Blackwell, 1994), pp. 768–774.
18. This lack of explicitness is an important component of strategy because guessing wrong about
how an adversary values its objectives can lead to unexpected outcomes. U.S. strategy in Vietnam,
for example, assumed that after sustaining a certain level of casualties, North Vietnam would no
longer be willing to support the insurgency in the South. The search for this breaking point, and
uncertainty over whether it would have any political utility, bitterly divided the Johnson administration.
There may in fact have been a breaking point in Vietnam, but as U.S. Secretary of Defense
Robert McNamara concluded in 1967, reaching that point would have resulted in virtual genocide.
this sense, strategy should be distinguished from two closely related terms:
grand strategy and tactics. Grand strategy refers to the totality of an actor’s resources
directed toward military, political, economic, or other objectives. Tactics
refers to the art of Žghting battles and of using the various arms of the
military—for example, infantry, armor, and artillery—on terrain and in positions
that are favorable to them.19 Grand strategy, strategy, and tactics all describe
different points on a continuum of a given actor’s means toward a single
end: compelling another to do its will.
The following typology of ideal-type strategies is a useful starting point for
analysis:
Attack (strong actor) strategies:
1. direct attack
2. barbarism
Defense (weak actor) strategies:
1. direct defense
2. guerrilla warfare strategy
This typology includes two assumptions: (1) strong actors initiated the asymmetric
conict in question, and therefore “strong actor” and “attacker” are synonymous;
20 and (2) these ideal-type strategies are war-winning rather than
war-termination strategies.21
direct attack. Direct attack means the use of the military to capture or
eliminate an adversary’s armed forces, thereby gaining control of that opponent’s
values. The main goal is to win the war by destroying the adversary’s
capacity to resist with armed forces. Both attrition and blitzkrieg are direct-
International Security 26:1 100
See Steven Rosen, “War Power and the Willingness to Suffer,” in Bruce M. Russett, ed., Peace, War,
and Numbers (Beverly Hills, Calif.: Sage, 1972), pp. 167–168; John Mueller, “The Search for the
‘Breaking Point’ in Vietnam: The Statistics of a Deadly Quarrel,” International Studies Quarterly, Vol.
24, No. 4 (December 1980), pp. 497–519; Frederick Z. Brown, “Comment on Mueller: American
Misperceptions,” ibid., pp. 525–529; and Karnow, Vietnam: A History, pp. 454, 596.
19. This deŽnition of tactics is a paraphrase of one from the Littré Dictionary as quoted by Charnay,
“Strategy,” p. 770.
20. Of course, empirically, strong actors are not always the initiators in asymmetric conicts. Paul
counts twenty weak-actor-initiated conicts from 394 b.c. to 1993, of which eleven are included
here. See Paul, Asymmetric Conicts, pp. 3–4.
21. In eight asymmetric conicts (4.1 percent), the outcome was affected by a war termination or
conciliation strategy. Conciliation strategies include the use of bribes, offers of amnesty, power
sharing, and political reforms, and do not require armed forces to implement. Examples of asymmetric
conicts that ended as the result of a conciliation strategy include the Murid War (1830–59),
the Third Seminole War (1855–58), the Malayan emergency (1948–57), the British-Cypriot conict
(1954–59), and the Philippine-Moro conict (1972–80).
attack strategies.22 Some readers may Žnd inclusion of the blitzkrieg in this
deŽnition puzzling, because it seems the very deŽnition of an indirect attack
strategy.23 But because armored formations target enemy armed forces (the adversary’s
capacity to resist) in a blitzkrieg, it counts as an indirect tactic but a direct
strategy.24
Historically, the most common pattern of a direct-attack strategy has been
one in which an attacker’s forces advance to capture a defender ’s values (a
capital city, an industrial or communications center, or a bridge) or strategic assets
(defensible terrain or a fort) and the defender moves to thwart that effort.
A battle or series of battles ensue, sometimes marked by lulls lasting entire seasons,
until one side admits defeat.
barbarism. Barbarism is the systematic violation of the laws of war in pursuit
of a military or political objective.25 Although this deŽnition includes the
use of prohibited weapons such as chemical and biological agents, its most important
element is depredations against noncombatants (viz., rape, murder,
and torture).26 Unlike other strategies, barbarism has been used to destroy an
adversary’s will and capacity to Žght. When will is the target in a strategic
How the Weak Win Wars 101
22. Mearsheimer’s deŽnitions of three main-attack strategies—blitzkrieg, attrition, and limited
aims—provide a useful starting point for the variations I introduce here. See Mearsheimer, Conventional
Deterrence, pp. 29–30, 33–43, 53–56.
23. A blitzkrieg features mobile armored formations—tanks supported by motorized infantry, mobile
artillery, and tactical air support—concentrating their attack against a narrow portion of a defender’s
extended position (usually thought of as a line). Once the enemy’s line is penetrated,
instead of turning immediately to engage anked forces, armored columns penetrate deep into the
defender’s rear areas, cutting communications, capturing supplies, and generally making it impossible
for now-isolated forward-deployed defenders to coordinate a defense. As a result, “an opponent
can be disarmed without numerous battles of annihilation.” See ibid., p. 30.
24. Mearsheimer notes that a successful blitzkrieg encirclement often results in the destruction of a
surrounded unit’s will to resist. See ibid., p. 38. Although in the case of blitzkrieg the impact is restricted
to the unit level (tactical), the observation supports the association between indirectness
and will proposed here.
25. The laws of war have evolved considerably over time. As used here they include the Hague
Conventions of 1899 and 1907, the Genocide Convention of 1948, the Four Geneva Conventions of
1949, and the Two Additional Protocols to the Geneva Conventions of 1977. On the laws of war
and their historical development, see André Corvisier and Barry Paskins, “Laws of War,” in
Corvisier and Childs, A Dictionary of Military History and the Art of War, pp. 443–453; Michael
Walzer, Just and Unjust Wars: A Moral Argument with Historical Illustrations, 2d ed. (New York: Basic
Books, 1992); andMichael Howard, George J. Andreopoulos, and Mark R. Shulman, eds., The Laws
of War: Constraints on Warfare in the Western World (New Haven, Conn.: Yale University Press,
1994). For a concise collection of the documents themselves, see Adam Roberts and Richard Guelff,
Documents on the Laws of War, 3d ed. (New York: Oxford University Press, 2000).
26. See, for example, Walzer, Just and Unjust Wars, p. 151. Chemical and biological weapons have
been traditionally included in this category because they are inherently indiscriminate. Deliberate
destruction of a defender’s natural environment (through deforestation, draining of swamps, etc.)
is also a violation of the laws of war for the same reason. See Mark Perry and Ed Miles, “Environbombing
campaign, for example, the strong actor seeks to coerce its weaker
opponent into changing its behavior by inicting pain (destroying its values).27
When will is the target in a counterinsurgency (COIN) campaign, the strong
actor attempts to deter would-be insurgents through, for instance, a policy of
reprisals against noncombatants.28 Strong actors can also target a weak actor’s
capacity to sustain an insurgency by, for example, the use of concentration
camps.29 Historically, the most common forms of barbarism include the murder
of noncombatants (e.g., prisoners of war or civilians during combat operations);
the use of concentration camps;30 and since 1939, strategic bombing
against targets of no military value.31
International Security 26:1 102
mentalWarfare,” in Roy Gutman and David Rieff, Crimes of War: What the Public Should Know (New
York: W.W. Norton, 1999), pp. 132–135.
27. The classic work here is Thomas C. Schelling, Arms and Inuence (New Haven, Conn.: Yale University
Press, 1966), especially chaps. 1 and 4. See also Robert A. Pape, Jr., “Coercive Air Power in
the Vietnam War,” International Security, Vol. 15, No. 2 (Fall 1990), pp. 103–146; and Pape, Bombing
to Win. In theory, it is possible to employ strategic air power to target an adversary’s capacity to
Žght by using air forces to destroy or interdict supplies, demolish key communications points
(railroad junctions, bridges, and airŽelds), or level arms factories. If it were possible to do so without
killing noncombatants, this would count as a direct-attack strategy. But in practice—even taking
into account advances in precision-guided munitions—strategic air power is too a blunt
weapon, and noncombatants are killed disproportionate to the military necessity of destroying the
targets. NATO’s strategic air campaign in Kosovo in 1999 is a case in point. See, for example, Independent
International Commission on Kosovo, Kosovo Report: Conict, International Response, Lessons
Learned (New York: Oxford University Press, 2000), pp. 92–94.
28. Such reprisals typically include executing randomly selected civilians in retaliation for the killing
of an occupying soldier. See, for example, Robert Asprey, War in the Shadows: The Classic History
of Guerrilla Warfare from Ancient Persia to the Present (New York: Little, Brown, 1994), p. 108.
29. Insurgent intelligence and support networks depend on the participation of sympathetic noncombatants,
and concentration camps disrupt these networks. See, for example, Donald W. Hamilton,
The Art of Insurgency: American Military Policy and the Failure of Strategy in Southeast Asia
(Westport, Conn.: Praeger, 1998), p. 59; and Paula M. Krebs, “‘The Last of the Gentlemen’s Wars’:
Women in the Boer War Concentration Camp Controversy,” History Workshop Journal, No. 33
(Spring 1992), pp. 41–42.
30. The British used concentration camps as a COIN strategy during the South African War. Although
not intended by the British, as many as 28,000 Boer women and children died in these
camps—more than the combined total of combatant casualties from both sides. On the use of concentration
camps as a COIN strategy, see John Ellis, A Short History of Guerrilla Warfare (New York:
St. Martin’s, 1976), p. 111. On their use and consequences in the South African War, see Christiaan
De Wet, The Three Years’ War (New York: Charles Scribner’s Sons, 1902), pp. 192–193; Thomas
Pakenham, The Boer War (New York: Random House, 1979), chap. 29 and pp. 607–608; and Krebs,
“‘The Last of the Gentlemen’s Wars.’”
31. Allied bombing of Dresden is a common example. See Ronald Schaffer, Wings of Judgment:
American Bombing in World War II (New York: Oxford University Press, 1985), pp. 97–99. On the
subject of strategic bombing as coercion against Nazi Germany more generally, see Pape, Bombing
to Win, pp. 260–262. In terms of Pape’s argument, strategic bombing that targets noncombatants
would count as barbarism.When air power is used to target enemy forces, it would count as a direct-
attack strategy. Attacks on infrastructure and industry are more problematic: Noncombatants
are not targets, but those who employ this strategy know beforehand that noncombatants will be
killed in such attacks.
direct defense. Direct defense refers to the use of armed forces to thwart
an adversary’s attempt to capture or destroy values such as territory, population,
and strategic resources. Like direct-attack strategies, these strategies target
an opponent’s military. The aim is to damage an adversary’s capacity to
attack by crippling its advancing or proximate armed forces. Examples include
limited aims strategies, static defense, forward defense, defense in depth, and
mobile defense.32
The inclusion here of limited aims strategies may seem counterintuitive.
Like preemptive or preventive attack strategies, these strategies begin with an
initial offensive—say, attacking concentrations of enemy armed forces across
an international border—but their ultimate aims are defensive.33 Limited aims
strategies target an adversary’s capacity to attack by destroying vital strategic
forces or by seizing key strategic assets (territory, bridges, promontories, etc.).
They are most often employed by weak actors that have initiated wars against
strong actors.34 Examples include Japan’s air assault on Pearl Harbor in 1941
and Egypt’s attack on Israel in 1973.
guerrilla warfare. Guerrilla warfare strategy (GWS) is the organization
of a portion of society for the purpose of imposing costs on an adversary using
armed forces trained to avoid direct confrontation.35 These costs include the
loss of soldiers, supplies, infrastructure, peace of mind, and most important,
time.36 Although GWS primarily targets opposing armed forces and their
support resources, its goal is to destroy not the capacity but the will of the
attacker.37
How the Weak Win Wars 103
32. For full summary descriptions, see Mearsheimer, Conventional Deterrence, pp. 48–50.
33. Ibid., pp. 53–56.
34. Ibid., pp. 24–26, 168.
35. A related strategy is terrorism, which commonly has political objectives similar to GWS. The
logic of most terrorism mirrors that of coercive strategic bombing. A largely urban phenomenon,
terrorism seeks either to inict pain on noncombatants so they will pressure their government to
accede to the terrorists’ political demands, or to delegitimize a government as a means to replace
or coerce it. This implies that the strategy will be most effective when citizens have a say in their
government’s policies.
36. On this point especially, see Samuel Huntington’s remarks in Stanley Hoffmann, Samuel P.
Huntington, Ernest R. May, Richard N. Neustadt, and Thomas C. Schelling, “Vietnam Reappraised,”
International Security, Vol. 6, No. 1 (Summer 1981), p. 7. See also Eliot A. Cohen, “Constraints
on America’s Conduct of SmallWars,” International Security, Vol. 9, No. 2 (Fall 1984), p. 157.
37. For general introductions to GWS, see Laqueur, Guerrilla; Ellis, From the Barrel of a Gun; Asprey,
War in the Shadows; and Anthony James Joes, Guerrilla Warfare: A Historical, Biographical, and Bibliographical
Sourcebook (Westport, Conn.: Greenwood, 1996). On Chinese and Cuban variations of
GWS, see Mao Tse-tung, On Guerrilla Warfare, trans. Samuel B. GrifŽth (New York: Praeger, 1961);
Edward L. Katzenbach and Gene Z. Hanrahan, “The Revolutionary Strategy of Mao Tse-Tung,”
Political Science Quarterly, Vol. 70,No. 3 (September 1955), pp. 321–340; Che Guevara, Guerrilla Warfare
(New York: Monthly Review, 1961); and Regis Debray, Revolution in the Revolution? (New York:
Penguin, 1968).
GWS requires two elements: (1) physical sanctuary (e.g., swamps, mountains,
thick forest, or jungle) or political sanctuary (e.g., weakly defended border
areas or border areas controlled by sympathetic states), and (2) a
supportive population (to supply Žghters with intelligence and logistical support,
as well as replacements). The method of GWS is well summarized by perhaps
its most famous practitioner, Mao Tse-tung: “In guerrilla warfare, select
the tactic of seeming to come from the east and attacking from the west; avoid
the solid, attack the hollow; attack; withdraw; deliver a lightning blow, seek a
lightning decision. When guerrillas engage a stronger enemy, they withdraw
when he advances; harass him when he stops; strike him when he is weary;
pursue him when he withdraws. In guerrilla strategy, the enemy’s rear, anks,
and other vulnerable spots are his vital points, and there he must be harassed,
attacked, dispersed, exhausted, and annihilated.”38
GWS is not a strategy for obtaining a quick, decisive defeat of invading or
occupying forces.39 Moreover, because guerrillas cannot hold or defend particular
areas, they do not provide security for their families while on operations
or when demobilized to await new missions. GWS is therefore a strategy that
requires placing key values (e.g., farms, family, religious or cultural sites, and
towns) directly in the hands of the adversary. Logically then, important costs
of adopting a GWS depend on both the purpose and the restraint of the adversary.
40 When invading or occupying forces do not exercise restraint in the use
of force, or when their purpose is the destruction of a weak actor’s people,
GWS can become a prohibitively expensive defensive strategy.
the logic of strategic interaction
Every strategy has an ideal counterstrategy. Actors able to predict their adversary’s
strategy can therefore dramatically improve their chances of victory by
choosing and implementing that counterstrategy. Mao, for example, argued
that “defeat is the invariable outcome where native forces Žght with inferior
weapons against modernized forces on the latter’s terms.”41 Mao’s maxim suggests
that when the weak Žght the strong, the interaction of some strategies
will favor the weak, while others will favor the strong.
International Security 26:1 104
38. Tse-tung, On Guerrilla Warfare, p. 46.
39. See, for example, Hamilton, The Art of Insurgency, p. 27.
40. In March 1900 the British captured the Žrst Boer capital, Bloemfontein. The surviving Boer
commanders gathered to decide whether to surrender or keep Žghting. The commanders were
closely divided, but tipping the balance in favor of continued—guerrilla—resistance was faith in
British civility. The Boer ultimately found their faith unjustiŽed. See De Wet, The Three Years’ War,
pp. 192–193.
41. Quoted in Mack, “Why Big Nations Lose Small Wars,” p. 176 (emphasis in original).
Building on Mao’s insight, I argue that the universe of potential strategies
and counterstrategies can be reduced to two distinct ideal-type strategic approaches:
direct and indirect.42 Direct approaches target an adversary’s armed
forces in order to destroy that adversary’s capacity to Žght. Indirect approaches
seek to destroy an adversary’s will to Žght: Toward this end, a GWS
targets enemy sodiers, and barbarism targets enemy noncombatants.43 Sameapproach
interactions (direct-direct or indirect-indirect) imply defeat for weak
actors because there is nothing to mediate or deect a strong actor’s power advantage.
These interactions will therefore be resolved quickly. By contrast, opposite-
approach interactions (direct-indirect or indirect-direct) imply victory
for weak actors because the strong actor’s power advantage is deected or
dodged.44 These therefore tend to be protracted, with time favoring the weak.45
In asymmetric conicts when strategic interaction causes an unexpected delay
between the commitment of armed forces and the attainment of military or
political objectives, strong actors tend to lose for two reasons. First, although
all combatants tend to have inated expectations of victory,46 strong actors in
asymmetric conicts are particularly susceptible to this problem.47 If power
implies victory, then an overwhelming power advantage implies an overwhelming—
and rapid—victory. As war against a Lilliputian opponent drags
on, however, dramatic overestimates of success force political and military
elites in the strong state to escalate the use of force to meet expectations (thus
increasing the costs of a conict) or risk looking increasingly incompetent. Either
way, domestic pressure to end the conict is likely to result. And as Mack
highlights in his discussion of political vulnerability, the longer a war drags on,
the greater the chances are that the strong actor will simply abandon the war
effort, regardless of the military state of affairs on the ground. Strong actors
also lose asymmetric wars when, in attempting to avoid increasing costs—such
as declaring war, mobilizing reserves, raising taxes, or sustaining additional
How the Weak Win Wars 105
42. This reduction of strategies to two mutually exclusive types is well established in the strategic
studies literature. See, for example, André Corvisier and John Childs, “Indirect Warfare,” in
Corvisier and Childs, A Dictionary of Military History and the Art of War, p. 378; and Liddell Hart,
Strategy, pp. 197, 361–364.
43. For a similar deŽnition, see Pape, “Coercive Air Power in the Vietnam War,” pp. 106–107.
44. In GWS, an attacker’s armed forces are physically avoided or engaged only on favorable
terms. In a blockade or strategic bombing campaign against a direct defense in a limited war, the
strong actor’s destructive power is deected, because such attacks invariably place the noncombatant
population between attackers and political elites.
45. On the importance of conict duration as a cost of conict, see Mearsheimer, Conventional Deterrence,
p. 24; and Katzenbach and Hanrahan, “The Revolutionary Strategy of Mao Tse-Tung,”
pp. 324–326.
46. See Geoffrey Blainey, The Causes of War, 3d ed. (New York: Free Press, 1988), p. 53.
47. See, for example, Mack, “Why Big Nations Lose Small Wars,” pp. 181–182.
casualties—they yield to the temptation to employ barbarism. Barbarism conserves
friendly forces, but even when militarily effective it is risky:48 Barbarism
carries the possibility of domestic political discovery (and opposition) as well
as external intervention.
strategic interaction: explaining the trend. My explanation for the
trend toward increasing strong-actor failure is suggested both by the timing of
the biggest shift in outcomes favoring weak actors (1950–98) and by the logic
of Kenneth Waltz’s argument that actors in a competitive international system
“socialize” to similar policies and strategies. As Waltz argues, “The fate of each
state depends on its responses to what other states do. The possibility that
conict will be conducted by force leads to competition in the arts and the instruments
of force. Competition produces a tendency toward the sameness of
the competitors.”49
This said, what is the appropriate spatial context for socialization? I argue
that socialization works regionally, and that after World War II two different
patterns of socialization emerged in two different regions of the world. In
the blitzkrieg pattern, success was measured by the capacity to produce and
deploy large mechanized and combined-armed forces designed to destroy
an adversary’s armed forces and capture its values without costly battles of
annihilation. This model was imitated by the United States, its European allies,
the Soviet Union, and to some extent Japan. In the guerrilla warfare pattern,
success was measured by the ability to prosecute a protracted conict against
a technologically superior foe. Mao’s long Žght for, and eventual conquest
of, China was a model copied by Algerian rebels, the Vietminh,
the Hukbalahap, Cuban insurgents, Malayan communists, and to a large extent
the Mujahideen.50 The blitzkrieg model emphasizes direct strategic approaches;
the guerrilla warfare model, indirect strategic approaches. When the
two interact systematically, strong actors should lose more often.
These patterns of socialization suggest that actors on the threshold of armed
conict are not entirely free to choose an ideal strategy for two reasons. First,
forces, equipment, and training—all closely integrated—are not fungible.
Moreover, the development and prosecution of an actor’s ideal strategy may
International Security 26:1 106
48. Mack correctly emphasizes that barbarism is judged in proportion to the relative power of the
actors: Weak actors will be forgiven abuses for which strong actors will be hanged. See ibid.,
pp. 186–187.
49. Kenneth N. Waltz, Theory of International Politics (New York: McGraw-Hill, 1979), p. 127.
50. On Mao’s revolutionary guerrilla warfare as template, see Hamilton, The Art of Insurgency,
p. 18; and Katzenbach and Hanrahan, “The Revolutionary Strategy of Mao Tse-Tung,” p. 322.
be blocked by contrary organizational interests or traditions.51 Second, actors
prioritize threats: If the United States and Soviet Union, for example, identiŽed
each other as the primary threat, and both calculated that the most likely area
of direct confrontation would be the heart of Europe, then adopting warwinning
strategies, forces, equipment, and doctrines favorable to winning that
sort of war would be a sound strategy.
hypotheses: strategic interaction and conflict outcomes
This section explores the logic of four distinct strategic interactions and explains
how hypotheses derived from each can be reduced to a single hypothesis.
The expected relationship of strategic approach interactions to outcomes in
asymmetric conicts is summarized in Figure 3.
direct attack versus direct defense. In this interaction both actorsmake
similar assumptions about the priority of values over which they will Žght.
Both can therefore be expected to agree about the implications of a catastrophic
loss in battle, the rules of war, or the capture of a capital city. Because in this interaction
nothing mediates between relative material power and outcomes,
strong actors should win quickly and decisively.
Hypothesis 1: When strong actors attack using a direct strategy and weak actors defend
using a direct strategy, all other things being equal, strong actors should win
quickly and decisively.
direct attack versus indirect defense. Unlike direct strategies, which
involve the use of forces trained and equipped to Žght as organized units
against other similarly trained and equipped forces, indirect defense strategies
typically rely on irregular armed forces (i.e., forces difŽcult to distinguish from
noncombatants when not in actual combat). As a result, an attacker’s forces
tend to kill or injure noncombatants during operations, which tends to stimulate
weak-actor resistance. Most important, because indirect defense strategies
sacriŽce values for time,52 they necessarily take longer to resolve so long as
weak actors continue to have access to sanctuary and social support.53 In
asymmetric conicts, delay favors the weak.
How the Weak Win Wars 107
51. See, for example, Andrew F. Krepinevich, Jr., The Army and Vietnam (Baltimore, Md.: Johns
Hopkins University Press, 1986); Susan L. Marquis, Unconventional Warfare: Rebuilding U.S. Special
Operations Forces (Washington, D.C.: Brookings, 1997); Cohen, “Constraints on America’s Conduct
of Small Wars,” pp. 152–154; and Asprey, War in the Shadows, p. 836.
52. See Katzenbach and Hanrahan, “The Revolutionary Strategy of Mao Tse-Tung,” pp. 325–326.
53. Although a GWS requires sanctuary and social support, mere access to them in no way mandates
its adoption. French forces, for example, had access to both as they faced defeat in the
Hypothesis 2: When strong actors attack with a direct strategy and weak actors defend
using an indirect strategy, all other things being equal, weak actors should win.
indirect attack versus direct defense. Because the overwhelming force
available to the strong actor implies success against a weak adversary that attempts
a direct defense, an attacker’s use of an indirect strategy in this context
targets the defender’s will to resist. Prior to the advent of strategic air power
and long-range artillery (e.g., the V-1 and V-2 rockets in World War II), blockades
and sieges were the only means of coercing adversaries in this way. Today
strategic bombing campaigns54 are the most common form of indirect attack
against direct defense.
As coercive strategies intended to destroy an adversary’s will to resist, strategic
bombing campaigns tend to backŽre, stimulating precisely the sort of resolve
they aim to break: German bombing of London did not cow the British
into surrender, as Adolf Hitler and Hermann Göring expected. Instead it stiffened
British resolve. Strong actors will lose these interactions because they are
time-consuming and tend toward barbarism.55
International Security 26:1 108
Figure 3. Expected Effects of Strategic Interaction on Conflict Outcomes (expected
winners in cells).
Franco-Prussian War in 1870. France actively considered adopting a GWS after its disastrous defeat
at Sedan. Yet threatened with the loss of Paris, France surrendered instead. See Michael
Howard, The Franco-Prussian War (New York: Methuen, 1961), pp. 249–250.
54. Pape, Bombing to Win, chap. 6, has shown that strategic bombing or “punishment” strategies
rarely work (and that they cannot work against indirect defense strategies, such as GWS). See also
Mark Clodfelter, The Limits of Air Power: The American Bombing of North Vietnam (New York: Free
Press, 1989). If Pape is right and tactical air power is highly effective as a means to coerce an adversary,
then tactical air support that accepts collateral damage should become increasingly common;
and “human shield” defense of, say, armored or transport columns would become an increasingly
common countermeasure.
55. Strategic bombing campaigns usually start out with the intent to spare noncombatants—often
to the point of putting pilots and their crews at considerable risk (e.g., by ying lower or slower)—
Hypothesis 3: When strong actors attack using an indirect strategy and weak actors
defend using a direct strategy, all other things being equal, strong actors should lose.
indirect attack versus indirect defense. Indirect defense strategies presuppose
a certain level of moral restraint on the part of attackers.When strong
actors employ a strategy that ignores such restraint, weak actors are unlikely
to win—both because there would be no one left to win for, and because
GWS depends directly on a network of social support for intelligence, logistical
assistance, and replacements.56 Barbarism works as a COIN strategy because
by attacking either or both of the essential elements of a GWS—
sanctuary and social assistance—it destroys an adversary’s capacity to Žght.
For example, in the Murid War—the Russian empire’s struggle to conquer the
Muslim tribes of the Caucasus Mountains from 1830 to 1859—the Russians
found that they could not make any headway against the mountain Murids
because of severe attrition and supply problems associated with passing
through the heavy beech forests of Chechnya. The Murids raided Russian
forts and settlements from mountain fortresses that were virtually impregnable
to attack except by heavy artillery. Masters of marksmanship and hitand-
run tactics, the Chechens would strike heavily armed Russian columns
and then vanish into the forest before the Russians could mount a counterattack.
The Russians ultimately fought back by felling thousands of trees, virtually
deforesting Chechnya. By 1859 they were Žnally able to use their heavy
artillery to blast Murid mountain strongholds into rubble, and resistance soon
collapsed.
Hypothesis 4: When strong actors employ barbarism to attack weak actors defending
with a GWS, all other things being equal, strong actors should win.
Each of these hypotheses describes an interaction of either same-approach or
opposite-approach strategic interaction. It follows that all four may be tested as
a single hypothesis.
How the Weak Win Wars 109
but in most cases, these campaigns escalate until either noncombatant casualties are simply accepted
(as was the case with U.S. bombing of North Vietnam during the Rolling Thunder campaign),
or noncombatants become accepted targets (as in the case of the Žre bombings of Dresden
and Tokyo and the atomic bombing of Hiroshima and Nagasaki).A strict interpretation of the laws
of war may make a strategic air campaign that accepts “collateral damage” (viz., death or injury to
noncombatants) a war crime. The issue is tricky: The laws of war permit collateral damage so long
as that damage is proportional to the military value of the target itself. But if strategic air power
has no military utility, then collateral damage from strategic bombing would constitute a war
crime.
56. Mao Tse-tung once analogized the relationship between insurgents and citizens in a people’s
war by likening the Žghters to Žsh and the people to the sea. Effective counterinsurgency would
therefore require either altering the terrain (making the sea transparent) or killing, expelling, or
Hypothesis 5: Strong actors are more likely to win same-approach interactions and lose
opposite-approach interactions.
A First Test: Strategic Interaction and Asymmetric War Outcomes,
1800–1998
The aim of this section is to determine whether a statistically signiŽcant correlation
exists between strategic interaction and asymmetric conict outcomes. It
begins with a discussion of how cases were coded and then analyzes three key
relationships: (1) strategic interaction and conict outcomes, (2) strategic interaction
and conict duration, and (3) strategic interaction and the trend toward
increasing strong-actor failure over time.
coding and cases
The basic method of coding cases was to examine the history of each war in the
Correlates of War data set. A conict was coded asymmetric if the halved product
of one actor’s armed forces and population exceeded the simple product of
its adversary’s armed forces and population by ³ 5:1. If the strong actor used
armed forces to attempt to destroy a weak actor’s forces or capture values, it
was coded as a direct attack. If the weak actor used armed forces to attempt to
thwart these attacks, it was coded as a direct defense. A coding of barbarism
was reserved for strong actors that systematically targeted noncombatants,
employed illicit weapons, or accepted collateral damage in a strategic bombing
campaign after bomb damage assessments cast considerable doubt on the
efŽcacy of the campaign as a whole. A weak actorwas coded as using a GWS if
it sought to impose costs on the strong actor with armed force while avoiding
pitched battles. Each conict dyad was coded with one of four strategic interactions
(direct-direct, direct-indirect, indirect-direct, or indirect-indirect),57 be-
International Security 26:1 110
imprisoning the people (drying up the sea). Both would count as barbarism. See Mao Tse-tung,
“Primer on Guerrilla Warfare,” in Donald Robinson, ed., The Dirty Wars: Guerrilla Actions and Other
Forms of Unconventional Warfare (New York: Delacorte, 1968), p. 284.
57. Most asymmetric wars contain a single strategic interaction from start to Žnish, but a few—
such as the South African War and the U.S. intervention in Vietnam—contain multiple-sequential
or multiple-simultaneous interactions, respectively. In multiple-sequential interaction conicts,
strategies change, but in temporal sequence: One side’s strategic shift is quickly followed by another’s.
Multiple-simultaneous interaction conicts are those in which a single actor, or an actor
and its allies, pursue different strategies against the same adversary within a single theater of operations.
The empirical distribution of conict types is as follows: single, 134 (77.5 percent), multiple-
sequential, 29 (16.8 percent), and multiple-simultaneous, 10 (5.8 percent). In this analysis,
conict outcomes are explained by strategic interaction outcomes.
fore being reduced to one of two interaction types (same approach or opposite
approach).58
The key variable of analysis is strategic interaction (STRATINT) as compared
to conict outcome (OUTCOME). If strategic interaction causes change in
conict outcome, then a shift in the value of strategic interaction across the
case universe should be matched by a corresponding shift in outcome. The
STRATINT variable was coded 0 if the strategic interaction was same approach
(direct-direct or indirect-indirect), and 1 if it was opposite approach (directindirect
or indirect-direct). The OUTCOME variable was coded 0 if the strong
actor lost, and 1 if it won.59
strategic interaction and conflict outcomes
Running correlations established both that strategic interaction and asymmetric
conict outcomes are associated, and that the relationship is statistically
signiŽcant (see Figure 4).60 The results thus support hypothesis 5.61 Strong actors
won 76 percent of all same-approach interactions, and weak actors won 63
percent of all opposite-approach interactions.
strategic interaction and conflict duration
The key causal mechanism of the strategic interaction thesis is time: Sameapproach
interactions should be over quickly, whereas opposite-approach interactions
should be protracted (with weak actors tending to win drawn-out
wars). An analysis of the average duration of same-approach and opposite-
How the Weak Win Wars 111
58. The relatively few wars with multiple interactions were reduced to single interactions. In multiple-
sequential conicts, the Žnal interaction was used to represent the overall conict: The South
African War was coded same approach because it ended with an indirect-indirect interaction. In
multiple-simultaneous conicts, interactions were averaged: U.S. intervention in Vietnam was
coded opposite approach because although some interactions were same approach, on balance, the
contest was decided by a direct-indirect interaction. The chief consequence of these reductions is a
tougher test for the strategic interaction thesis, because collapsing interactions increases the impact
of relative material power on outcomes. Because strong actors have a greater material capacity to
adapt to failure than do weak actors, collapsing interactions hides strong-actor rather than weakactor
failures.
59. Stalemates and ongoing conicts were coded losses for the strong actor.
60. Pearson chi-squared (1), 14.56, p < .001.
61. An analysis of the relationship between strategic interaction and interaction outcomes (as opposed
to war outcomes) produces an even more striking Žnding: Weak actors win 23.1 percent of
same-approach and 78.4 percent of opposite-approach interactions. This relationship is statistically
signiŽcant: Pearson chi-squared (1), 40.95, p .001. An analysis of the impact of external
noncombat support for weak actors did not refute the strategic interaction thesis: Even when weak
actors received no support, they were still three times more likely to win opposite-approach interactions
than they were same-approach interactions: Pearson chi-squared (1), 11.38, p .001.
approach interactions supports this claim: Same-approach interactions lasted
2.69 years on average (2.98 years was the overall mean); opposite-approach interactions
lasted an average of 4.86 years.
strategic interaction and long-term trends
Both opposite-approach interactions and strong-actor failures have increased
over time: From 1800 to 1849, 5.9 percent of interactions in thirty-four asymmetric
conicts were opposite approach. From 1850 to 1899, 10.1 percent of
interactions in sixty-nine asymmetric conicts were opposite approach.
From 1900 to 1949, 16.1 percent of thirty-one asymmetric conicts were opposite,
and from 1950 to 1998, 22.2 percent of thirty-six asymmetric conicts were
opposite.
In sum, the data analysis supports three key hypotheses relating strategic interaction
to asymmetric conict outcomes. First, strong actors are more likely
to lose opposite-approach strategic interactions. Second, opposite-approach interaction
conicts take longer to resolve than do same-approach interactions.
Third, the frequency of opposite-approach interactions has increased in proportion
to strong-actor failure over time.
The analysis is limited, however, because some data are missing: Many civil
and colonial wars recorded neither the quantity of forces committed nor the
strategies employed. Although these defects are balanced by statistical controls,
even a perfect data set would support only a correlation between variables,
not causation. Thus, although the data analysis might have refuted the
strategic interaction thesis, only in combination with a careful comparison of
historical cases could the thesis be conŽrmed.62
International Security 26:1 112
Figure 4. Strategic Interaction and Asymmetric Conflict Outcomes, 1800± 1998.
62. See Ivan Arreguín-Toft, “Arts of Darkness: Guerrilla Warfare and Barbarism in Asymmetric
Conicts,” Ph.D. dissertation, University of Chicago, 1998.
U.S. Intervention in the Vietnam War
In this section I present a synoptic case study of U.S. intervention in Vietnam
(March 1965 to January 1973) as a preliminary test of the causal logic of the interest
asymmetry and strategic interaction theses. The Vietnam War represents
the strongest case study for Mack’s interest asymmetry thesis. If strategic interaction
can explain the war’s outcome better than can interest asymmetry, then
all other things being equal, it should be considered a better theory.
background
U.S. military intervention in Vietnam began soon after the defeat of France at
Dien Bien Phu in 1954. The commitment of U.S. combat troops, however, did
not occur until eleven years later. In 1965 the U.S. population was about 194
million, while North Vietnam’s stood at approximately 19 million.63 U.S. and
North Vietnamese armed forces totaled about 2.5 million and 256,000, respectively.
Adding in allies that contributed combat troops (negligible), multiplying
population and armed forces, and dividing the strong actor’s total by two
results in a relative force ratio of about 53:1. Even allowing for the fact
that the United States did not actually devote half of its armed forces and
half of its population to the conict,64 there is no question that (1) this was
an asymmetric conict, and (2) the United States and its allies were the strong
actor.65
u.s. military intervention, 1965–73
U.S. military intervention involved four distinct strategic interactions: (1) barbarism
(Rolling Thunder) against a direct defense, (2) a direct attack against a
direct defense (the main-force units war), (3) direct attacks against a GWS (the
guerrilla war in the South I), and (4) barbarism against GWS (the guerrilla war
in the South II).
interaction 1: rolling thunder, 1965–68. The Žrst strategic interaction of
the war began in March 1965 with a U.S. strategic bombing campaign, later
How the Weak Win Wars 113
63. These Žgures are taken from Singer and Small, Correlates of War, war number 163.
64. The logic of the relative power estimate does not require calculating how much of a given
actor’s power resources were applied to the Žght. What matters is the resources that an actor
could have applied to win the war. On this point, see Schelling, Arms and Inuence, pp. 142–143,
172.
65. Even if we assume, for the sake of simplicity, that the contest was between only the United
States and North Vietnam, the ratio shifts to 49:1 in favor of the United States.
named Rolling Thunder.66 Its main goal was to destroy the willingness of
North Vietnam to support the guerrilla war campaign in the South, and as its
name implied, the campaign was expected to take time: “Instead of a coordinated
air campaign . . . which would destroy the enemy’s ability to wage war
and break their will to resist, air operations over the North were designed as a
diplomatic ‘slow squeeze’ signaling device. As Secretary of Defense Robert S.
McNamara said on February 3, 1966, ‘U.S. objectives are not to destroy or to
overthrow the Communist government of North Vietnam. They are limited to
the destruction of the insurrection and aggression directed by North Vietnam
against the political institutions of South Vietnam.’”67
The United States wanted to inict enough pain on North Vietnam to compel
it to stop supporting the GWS in the South. North Vietnam’s defense
against Rolling Thunder was direct: It sought to thwart U.S. military attacks on
its infrastructure and forces by means of Žghter aircraft and an increasingly
dense radar and surface-to-air missile defense network.
U.S. Air Force generals and civilian leaders shared a theory about the general
effectiveness of strategic bombing. Strategic bombing should have both
hampered North Vietnam’s war effort and coerced its leadership into giving
up. When neither expectation was realized, military and civilian elites faced a
stark choice: either reject the theory or blame failure on some aw in implementation.
The Air Force chose to emphasize aws, while the Johnson administration
was increasingly split: Some agreed that the United States was hitting
the wrong targets—or not hitting the right targets hard enough. Others, including
eventually Defense Secretary McNamara, concluded that against North
Vietnam strategic bombing could not work. McNamara’s reports indicated
that the military value of Rolling Thunder’s destroyed targets was zero.68
International Security 26:1 114
66. On the origins of the Rolling Thunder campaign and its purpose, see Karnow, Vietnam: A History,
p. 397. On collateral damage to noncombatants in Rolling Thunder, see ibid., p. 458. On relative
U.S. moral restraint in Rolling Thunder, see ibid., p. 653. For a dedicated analysis of Rolling
Thunder and its failure, see James Clay Thompson, Rolling Thunder: Understanding Policy and Program
Failure (Chapel Hill: University of North Carolina Press, 1980); for a theoretical treatment that
anticipates the strategic interaction thesis, see Pape, Bombing to Win, especially chap. 6. Pape argues
that when air power is used to target an adversary’s armed forces, it generally wins, and
when used to target an adversary’s values (including infrastructure), it generally fails. In other
words, punishment and denial represent indirect and direct strategic approaches, respectively;
Pape’s argument, as Mearsheimer’s before it and mine after, therefore constitutes an argument for
the independent causal impact of strategy on conict outcomes.
67. Quoted in Harry G. Summers, Jr., Historical Atlas of the Vietnam War (New York: Houghton
Mifin, 1995), p. 96.
68. Its political utility was less than zero: It increased international and domestic opposition to the
U.S. war effort, and although the North Vietnamese feared and hated the bombing, they never considered
altering their war aims as a result of the pain it inicted.
Bombing that accepted collateral damage subsequent to this recognition was
therefore a war crime: barbarism.
Rolling Thunder continued until a week before the November 1968 U.S.
presidential election. It was an interaction in which a strong actor (the United
States) employed an indirect strategy against a weak actor (North Vietnam) using
a direct strategy, and lost.69
interaction 2: the main-force units war, 1965–69. This phase of the war
featured a series of pitched battles between North Vietnamese regular units
and those of the United States and the South Vietnamese Army (ARVN). In a
sequence of engagements that lasted throughout the war, U.S. forces proved
overwhelmingly successful at destroying North Vietnamese Army (NVA) and
Vietcong (VC) main-force units.
Examples of this interaction, in which a strong-actor direct-indirect opposed
a weak-actor direct strategy, include Operation Starlite (August 1965), the Battle
of Ia Drang (October–November 1965), Masher/White Wing (January–
March 1966), and Phase II of Operation Attleboro (October–November 1966).70
As the ground war continued into 1968, the frequency of these interactions
dwindled, and U.S. forces in the South focused increasingly on the problem of
counterinsurgency.
Part of the decreasing frequency of this direct-direct interaction reected actor
learning: North Vietnam’s leadership analyzed main-force unit engagements
carefully and concluded that U.S. forces were so adept at combining
maneuver and Žrepower that, unless the encounter took place between dramatically
mismatched forces (as at LZ Albany in November 1965), NVA and
VC main-force units would invariably be destroyed. In the context of Mao’s
warning above, native forces were Žghting with inferior weapons on U.S.
terms; and as predicted, they lost.
As the loser in this interaction, the North began to innovate tactically and
strategically. In tactical terms, NVA units attempted to “cling to the belt” of
U.S. and ARVN forces,71 getting so close so quickly that allied forces could not
How the Weak Win Wars 115
69. No one involved in the campaign on the U.S. side considers Rolling Thunder a success, although
it is fair to say that the reasons proposed to explain the campaign’s failure have been numerous
and controversial. From the military perspective, the consensus is that the campaign failed
because it misapplied military means to achieve a political end. For a more thorough discussion
and analysis of Rolling Thunder’s target and impact, see, for example, Thompson, Rolling Thunder,
Herring, America’s Longest War, p. 147; Clodfelter, The Limits of Air Power; and Pape, “Coercive Air
Power in the Vietnam War.”
70. These examples are drawn from Summers, Historical Atlas of the Vietnam War; and Marc
Leepson, ed., Webster’s New World Dictionary of the Vietnam War (New York: Simon and Schuster,
1999).
71. See Karnow, Vietnam: A History, p. 463.
beneŽt from close air or artillery support. In strategic terms, the North shifted
more of its resources into the guerrilla campaign in the South.
interaction 3: the guerrilla war in the south i, 1965–73. Guerrillas in
the South waged their campaign with considerable skill and were countered with a
bewildering mix of professionalism, passion, sadism, and incompetence. In this dimension
of the conict more than any other, U.S. efforts were heavily Žltered
through, and constrained by, the United States’ South Vietnamese allies.72
In U.S. Army areas of responsibility, the Žrst attempts to defeat the VC insurgency
involved search and destroy missions. In this type of mission, regular
army units, acting on intelligence that had located enemy unit concetrations,
would seek to make contact with these concentrations and destroy them.
Large-scale examples of this type of interaction include Phase I of Operation
Attleboro (September–October 1965), Operation Cedar Falls (January 1967),
and Operation Junction City (February–May 1967); but these interactions repeated
themselves on a smaller scale throughout the war. Forces were killed on
both sides, but the strategic balance slowly shifted to favor the VC, largely because
U.S. forces relied heavily on indirect Žrepower—tactical air and artillery
support—a consequence of which was considerable death or injury to noncombatants.
73
Not all U.S. COIN efforts failed outright or succeeded through barbarism.
U.S. Marine Corps forces in the mountainous northernmost area of South Vietnam,
for example, pursued a COIN strategy combining local and highly motivated
(but poorly trained and equipped) villagers with direct support from
U.S. Marine combat platoons. These combined action platoons (or CAPs) operated
on an “inkblot” principle: Secure a village or hamlet, then patrol out in
widening circles until intersecting with another CAP’s secured area. The strategy
had two major disadvantages, however. First, although it protected South
Vietnamese citizens from immediate danger of terror attacks by VC guerrillas,
it could achieve success only in the long term—and time favored the VC.74 Second,
although militarily effective, CAP success only highlighted the inability of
the South Vietnamese government to protect its own citizens.
Ultimately, the United States lost this interaction. Its combat forces had been
trained and equipped to Žght a uniformed regular adversary using massive
International Security 26:1 116
72. The record of performance by ARVN units is mixed. Some units and their commanders are extolled
for their bravery, skill, and loyalty, while many are remembered only for their cowardice, incompetence,
and venality. See, for example, Karnow, Vietnam: A History, p. 441; and Summers,
Historical Atlas of the Vietnam War, p. 80.
73. See Asprey, War in the Shadows, p. 881.
74. See ibid., p. 848.
Žrepower, not an invisible enemy that refused to meet it in battle.75 The indiscriminate
impact of the U.S. Army’s heavy reliance on artillery and air support
progressively alienated potential allies among South Vietnam’s people. As losers,
however, U.S. forces were not slow to innovate a strategic response.
interaction 4: the guerrilla war in the south ii, 1965–73. U.S. strategic
innovations that aimed at seriously undermining the VC guerrilla campaign in
the South took two forms. The Žrst was the Strategic Hamlets program, and
the second was the Phoenix program.76
The U.S. Strategic Hamlets program was modeled after a French program in
which South Vietnamese villagers were forced from their homes and relocated
to fortiŽed hamlets.77 Where implemented effectively, the U.S. program had
the military beneŽt of severely damaging VC intelligence and supply networks,
but it also extracted a signiŽcant political cost: As nightly news broadcasts
ashed images of more and more wailing peasants being forced to leave
their villages, U.S. public opinion began to turn against the war. The program,
however, was rarely implemented effectively. In most cases, corrupt ofŽcials
failed to deliver weapons and embezzled funds and supplies intended to turn
the hamlets into functioning communities.78 As a result, the program alienated
the people whose good will the United States and South Vietnam needed to
win the war: Forced to leave their homes and then abandoned, many South
Vietnamese turned against their government and became active supporters of
the VC. By decreasing the program’s COIN beneŽts and increasing its political
costs, the South Vietnamese government’s corruption and incompetence eventually
rendered the Strategic Hamlets program a disaster.
The second U.S. innovation was the Phoenix program, whose aims and legitimacy
continue to provoke sharp debate.79 The overall military view is that
Phoenix was essentially a legitimate military operation. It relied on special intelligence
to target and destroy VC leadership, and it proved to be the single
most successful strategic initiative pursued by U.S. forces during the war.80
How the Weak Win Wars 117
75. This is the central thesis in Krepinevich, The Army and Vietnam, p. 4.
76. “Phoenix” was the code name for a U.S. assassination program that targeted VC leadership.
77. On the Strategic Hamlets program in Vietnam, its logic, and its successes and failures, see
Karnow, Vietnam: A History, pp. 255–257; and Herring, America’s Longest War, pp. 85–86, 88–90.
78. Karnow, Vietnam: A History, p. 257.
79. Asprey, War in the Shadows, pp. 910–911.
80. There is little question that Phoenix effectively disrupted the capacity of the VC to continue
their GWS in the South. Even the North Vietnamese admit this: “Nguyen Co Thach, Vietnam’s foreign
minister from 1975, admitted that the Phoenix effort ‘wiped out many of our bases’ in South
Vietnam, compelling numbers of North Vietnamese and Viet Cong troops to retreat to sanctuaries
in Cambodia.” See Summers, Historical Atlas of the Vietnam War, p. 148. See also Karnow, Vietnam: A
History, p. 534; and Herring, America’s Longest War, p. 232.
To most observers, participants, and historians, however, the sustained effort
to kill noncombatants raised troubling questions about the program’s legitimacy
as an extension of U.S. policy, or as a COIN strategy, regardless of its
effectiveness.
Overall, the United States won this interaction. The Strategic Hamlets program
was never implemented properly, so its contribution to U.S. success in
this interaction was negative. By contrast, the Phoenix program, which eviscerated
the VC command infrastructure in the South, may have even provoked
the North into its premature and disastrous direct confrontation with U.S. regular
forces during the 1968 Tet Offensive. Because both strategies systematically
and deliberately targeted noncombatants, both must be counted as
barbarism—albeit barbarism at the mildest end of the violations spectrum.
politics
Two political aspects of the war deserve special attention. First, why did the
United States think it necessary to send combat troops into South Vietnam?
Was Vietnam a vital or peripheral U.S. interest? Second, why did the United
States withdraw? Was it fought to a stalemate on the battleŽeld or forced to
abandon the war by U.S. domestic opposition?
The United States never entirely decided whether Vietnam was a vital or peripheral
interest. If there is a consensus view among historians, it is that the
United States got into Vietnam incrementally, conŽdent that after just one more
escalation U.S. forces could “stabilize” South Vietnam and exit. Eventually the
fate of South Vietnam became inextricably linked with perceptions of the
United States’ own credibility, and in that sense constituted a vital interest.
The United States withdrew from Vietnam when and how it did because
U.S. public opinion had shifted against the war. The United States did not Žght
as effectively as it could have. But even given the limitations of placing a force
trained and equipped to Žght and win a land battle in Europe in the mountains,
jungles, and marshy river deltas of Indochina, the United States proved
remarkably successful at innovating tactical and strategic responses to North
Vietnam’s strategic initiatives. As a result, by 1969 U.S. forces had achieved the
military defeat of the North. Nevertheless, the war dragged on for another four
years.
The U.S. withdrawal from Vietnam in 1973 was a consequence of two unrelated
problems. First, successive U.S. administrations incorrectly assumed that
the defeat of North Vietnam’s military forces would lead the North to accede
to U.S. demands. George Herring characterizes this problem well: “Nixon’s secret
diplomacy and implied military threats failed to wrench any concessions
International Security 26:1 118
from Hanoi. . . . Still hurting from those losses suffered in the Tet Offensive but
by no means ready to quit the Žght, Hanoi in 1969 shifted to a defensive, protracted
war strategy, sharply curtailing the level of military activity in the
south and withdrawing some of its troops back across the demilitarized zone.
Certain that American public opinion would eventually force Nixon to withdraw
from Vietnam, the North Vietnamese were prepared to wait him out, no
matter what additional suffering it might entail.”81
Second, the military defeat of North Vietnam and the VC insurgency in the
South took too long to accomplish.82 The American public had been led to believe
that victory would come quickly—not only because the United States
Želded the most powerful and technologically advanced armed force on earth,
but because as in past wars, Americans believed its cause to be just. Political
and military elites in the Kennedy, Johnson, and Nixon administrations did little
to temper this expectation, and in most cases encouraged it.83
summary
This synoptic presentation describes the Vietnam War in terms of four roughly
simultaneous strategic interactions.84 These interactions and their contributions
to the outcome of the war are summarized in Table 1.
In Vietnam, the weak actor had two entirely distinct militaries ready to oppose
U.S. forces: one trained and equipped to Žght an indirect war, the other
trained and equipped to Žght a direct war. Thus the North could be far more
nimble than the United States in shifting its strategic approach. As Andrew
Krepinevich, Donald Hamilton, and others have argued, with the possible exception
of the U.S. Marine Corps—which had considerable operational experience
with COIN operations—the U.S. military could never reconcile itself to
How the Weak Win Wars 119
81. Herring, America’s Longest War, p. 226.
82. On this point, see Karnow, Vietnam: A History, pp. 464, 480; and Herring, America’s Longest War,
p. 200.
83. This encouragement culminated in the infamous prediction by U.S. Comdr. William
Westmoreland—just prior to Tet—that “the enemy’s hopes are bankrupt.” See Karnow, Vietnam: A
History, p. 539.
84. Lewis Sorley argues that the war had yet another crucial phase after Tet, in which U.S. forces
shifted strategies under their new commander, Gen. Creighton W. Abrams. In terms of the strategic
interaction thesis, this would count as “Phase 5: Guerrilla War III/indirect-indirect/U.S. wins:
abandons Vietnam/war ends.” The difŽculty with Sorley’s argument is in measuring the effectiveness
of Abrams’s talented leadership in the context of a VC recently devastated by Phoenix and
Tet, and a similarly routed NVA. If, as many assert, the North had retreated to lick its wounds after
Tet, then U.S. military effectiveness would appear high regardless of its strategy. For this reason,
and because its inclusion would add little in the way of a test of competing explanations, I do not
include the interaction in this analysis. See Lewis Sorley, A Better War: The Unexamined Victories and
Final Tragedy of America’s Last Years in Vietnam (New York: Harcourt Brace, 1999).
the demands of a COIN war.85 These demands do not imply the necessity of
creating a force capable of barbarism: As the CAPs demonstrated, it was possible
to Žght a GWS in the South within the framework of the laws of war. What
it was manifestly not possible to do was defeat a people in arms quickly.
Thus the CAP example only underscores the importance of the key causal
mechanism of the strategic interaction thesis: When the power relationship implies
a speedy victory, and the interaction implies a delay, the way is clear for
the operation of what Mack calls political vulnerability. That is, even an ideal
COIN strategy—one that destroys enemy forces without destroying enemy
values—takes time. If such strategies are to become a model for future COIN
operations, this implies a counterintuitive policy: Strong-actor political and
military elites must prepare their publics for long-delayed victories against
even very weak adversaries when those adversaries employ indirect defense
strategies.
explaining vietnam: interest asymmetry and strategic interaction
Strategic interaction provides a powerful way to explain asymmetric conict
outcomes—not only individual wars but the observed trend toward increasing
strong-actor failures over time. But the interest asymmetry thesis has the same
goal, and it too appears to explain U.S. failure in the Vietnam War. Although
the United States ultimately quit Vietnam because of domestic political pressure,
as Mack’s thesis suggests, the strategic interaction thesis offers two important
qualiŽcations of his explanation of the war’s outcome.
International Security 26:1 120
85. See Krepinevich, The Army and Vietnam; and Hamilton, The Art of Insurgency, p. 155.
Table 1. Summary of Strategic Interactions and Effects in U.S. Intervention in Vietnam,
1965± 73.
Strategic
Interaction Innovation Effect
Duration
Effect
Rolling Thunder indirect-direct United States loses;
abandons strategy
lengthens war
Main-Force Units War direct-direct North loses; withdraws,
then escalates (Tet)
shortens war
Guerrilla War I direct-indirect United States loses;
switches to barbarism
(Phoenix)
lengthens war
Guerrilla War II indirect-indirect North loses; escalates,
then withdraws
shortens war
First, Mack argues that theoretically, at least, actor resolve can be derived
from relative power: That is, relative power and political vulnerability vary directly.
In Vietnam, however, U.S. resolve had nothing to do with relative
power. In fact, U.S. interests in the security and stability of South Vietnam were
far greater86 than those predicted by the interest asymmetry thesis. Vietnam
was a “limited” war not because South Vietnam’s fate was a peripheral U.S. interest,
but because U.S. political elites believed that the use of force in proportion
to U.S. interests might provoke Chinese military intervention and lead to a
third world war. In the case of North Vietnam, its legendary resolve was not a
consequence of its being a weak actor Žghting for survival. As many commentators
have since observed, the true sources of North Vietnam’s resolve were
nationalism87 and revenge for the suffering caused by the U.S. strategic bombing
campaign against the North Vietnamese people. According to Stanley
Karnow, “As a practical strategy, however, the bombing backŽred. American
planners had predicted that it would drive the enemy to capitulation, yet not
only did the North Vietnamese accept the sacriŽces, but the raids rekindled
their nationalistic zeal, so that many who may have disliked Communist rule
joined the resistance to alien attack.”88
Second, Mack is right that the United States was politically vulnerable, and
that this vulnerability ultimately drove it from Vietnam in defeat. But his thesis
assumes that political vulnerability will generally affect the ability of strong actors
to defeat weak actors. The strategic interaction thesis brackets the conditions
under which political vulnerability operates: It does so only when there is
an unanticipated delay between the commitment of armed forces and victory.
Conclusions
Strong actors lose asymmetric conicts when they adopt the wrong strategy
vis-à-vis their weaker adversaries. Same-approach interactions—whether direct-
direct or indirect-indirect—favor strong actors because they imply shared
values, aims, and victory conditions. Because nothing therefore intervenes between
raw power and goals, strong actors will win same-approach interactions
How the Weak Win Wars 121
86. Karnow, Vietnam: A History, pp. 377–378.
87. This is not to suggest that North Vietnam’s was a monolithic nationalism. South Vietnamese
communists had a national identity separate from that of Northerners. What united them strongly
during the war was the presence of foreigners on Vietnamese soil. SeeWalker Connor, “Ethnology
and the Peace of South Asia,” World Politics, Vol. 22, No. 1 (October 1969), pp. 51–86; Karnow, Vietnam:
A History, pp. 462, 534; and Herring, America’s Longest War, p. 271.
88. Karnow, Vietnam: A History, p. 458.
in proportion to their advantage in relative power. Opposite-approach interactions—
whether direct-indirect or indirect-direct—favor weak actors because
they sacriŽce values for time. This results in a signiŽcant delay between the
commitment of armed forces and the attainment of objectives. Time then becomes
the permissive condition for the operation of the political vulnerability
that Mack and others rightly identify as attaching to strong actors in asymmetric
conicts.
theory and policy implications
The strategic interaction thesis simultaneously supports, and in key ways mediates,
the importance of relative material power in explaining conict outcomes.
As has often been observed, material power is useful for theory
building because it is quantiŽable and measurable in a way that courage, leadership,
and dumb luck are not. This study has demonstrated empirically that
on balance, relative material power is more than simply a methodologically
useful concept; taken alone, it explains a majority of conict outcomes since
1800. The strategic interaction thesis makes clear, however, the limitations of
relative material power by highlighting the conditions under which it matters
more or less.
This analysis suggests key policy implications for both weak and strong actors.
For weak actors, successful defense against strong actors depends on an
indirect strategy. Because indirect strategies such as GWS rely on social support,
weak actors must work tirelessly to gain and maintain the sympathy or
acquiescence of a majority of the population in question. Given the risks involved
in either aiding or taking part in a guerrilla resistance, this in itself is no
mean feat. Additionally, weak actors must have or gain access to the physical
or political sanctuary necessary to make an indirect strategy a viable choice.
For strong actors, the strategic interaction thesis suggests that weak adversaries
employing an indirect defense will be difŽcult to defeat. Of course, not all
or even most asymmetric conicts need follow this pattern, but when they do,
and when a resort to arms is the only viable option, how should a strong actor
such as the United States react?
One response might be a resort to barbarism, which appears to be an effective
strategy for defeating an indirect defense.89 But even a cursory review of
International Security 26:1 122
89. This appears equally true of counterterrorism and terrorism. Laura K. Donohue analyzes the
impact of British counterterrorist legislation in Northern Ireland and concludes that Britain’s numerous
“temporary” and “emergency” measures—which were never temporary, and which violated
civil rights and due process—proved highly effective in the short term. Her analysis
postwar histories reveals that at best barbarism can be effective only as a military
strategy: If the desired objective is long-term political control, barbarism
invariably backŽres. The French, for example, used torture to quickly defeat
Algerian insurgents in the Battle of Algiers in 1957. But when French military
brutality became public knowledge, it catalyzed political opposition to the war
in France and stimulated renewed and intensiŽed resistance by the non-French
population of Algeria.90 Within four years, France abandoned its claims in Algeria
even though it had “won” the war. Barbarism thus sacriŽces victory in
peace for victory in war—a poor policy at best.91
An ideal U.S. strategic response in an asymmetric conict therefore demands
two central elements: (1) preparation of public expectations for a long war despite
U.S. technological and material advantages, and (2) the development and
deployment of armed forces speciŽcally equipped and trained for COIN operations.
92 Without a national consensus and realistic expectations, the United
States would be politically vulnerable in an asymmetric conict. Without more
special operations forces—the self-reliant and discriminate armed forces necessary
to implement an ideal COIN strategy—what begins as a military operation
against an isolated violent minority will tend to escalate into a war against
an entire people.93
The United States must be prepared to Žght and win both conventional and
asymmetric or “small” wars. The strategic interaction thesis shows why the
two missions demand two kinds of armed forces: one to defend U.S. interests
in conventional wars, and one to defend them in asymmetric wars. If the
United States, in other words, is to win future “boxing matches” against lightweight
opponents who use their own version of the rope-a-dope, it will need
Žghters with more initiative than discipline, and more endurance than punching
power.
How the Weak Win Wars 123
suggests, however, that insurgents always found a way around such measures, eventually
prompting yet another round of “emergency” restrictions. See Donohue, Counter-terrorist Law and
Emergency Powers in the United Kingdom, 1922–2000 (Portland, Or.: Irish Academic Press, 2001),
pp. 322–323.
90. See Mack, “Why Big Nations Lose Small Wars,” p. 180; and Asprey, War in the Shadows,
pp. 669–671.
91. See Liddell Hart, Strategy, p. 370.
92. See, for example, Ernest May in Hoffmann et al., “Vietnam Reappraised,” pp. 8, 9; Hoffmann,
ibid., p. 10; and Cohen, “Constraints on America’s Conduct of Small Wars,” pp. 166–167.
93. See Walzer, Just and Unjust Wars, p. 187.
War Name Start End OUTCOME1 STRATINT2
Spain-Peru 1809 1816 1 0
Russo-Georgian 1816 1825 1 0
Pindaro War 1817 1818 1 0
Kandyan Rebellion 1817 1818 1 0
Greek War of Independence 1821 1828 0 1
First Anglo-Burmese 1823 1826 1 0
First Ashanti 1824 1826 1 0
Javanese 1825 1830 1 0
Bharatpuran 1825 1826 1 0
Russo-Circassian 1829 1840 1 0
Albanian 1830 1831 1 0
Belgian Independence 1830 1831 1 0
Murid War 1830 1859 1 0
First Polish 1831 1831 1 0
First Syrian 1831 1832 0 0
Texan 1835 1836 0 0
Second Seminole War 1835 1842 1 0
First Zulu 1838 1840 1 0
First British-Afghan 1838 1842 0 1
Franco-Algerian 1839 1847 1 0
Bosnian-Turkish 1841 1841 1 0
Baluchi-British 1843 1843 1 0
First Maori 1843 1848 1 0
Franco-Moroccan 1844 1844 1 0
First British-Sikh 1845 1846 1 0
First KafŽr War 1846 1847 1 0
Cracow Revolt 1846 1846 1 0
Austro-Sardinian 1848 1849 1 0
First Schleswig-Holstein 1848 1849 1 0
Hungarian 1848 1849 1 0
Second British-Sikh 1848 1849 1 0
Roman Republic 1849 1849 1 0
Second KafŽr 1850 1853 1 0
La Plata 1851 1852 1 0
Second Anglo-Burmese 1852 1853 1 0
First Turko-Montenegran 1852 1853 0 0
Third Seminole War 1855 1858 1 0
Yakima War 1855 1858 0 1
Anglo-Persian 1856 1857 1 0
Second Opium War 1856 1860 0 0
Kabylia Uprising 1856 1857 1 .
Tukulor-French War 1857 1857 1 0
International Security 26:1 124
How the Weak Win Wars
Appendix 1: Cases and Key Variables
War Name Start End OUTCOME1 STRATINT2
French-Indochinese 1858 1863 1 0
Second Turko-Montenegran 1858 1859 0 0
Spanish-Moroccan 1859 1860 1 0
Italo-Roman 1860 1860 1 0
Second Maori 1860 1870 1 0
Apache and Navaho War 1860 1865 1 0
Taiping Rebellion 1860 1864 1 0
Nien Rebellion 1860 1868 1 0
Franco-Mexican 1862 1867 0 0
First Sioux War 1862 1864 1 0
Second Polish 1863 1864 1 0
Spanish-Santo Dominican 1863 1865 0 .
Second Schleswig-Holstein 1864 1864 1 0
Lopez War 1864 1870 1 1
Spanish-Chilean 1865 1866 1 0
British-Bhutanese 1865 1865 1 0
Second Sioux War 1865 1868 0 1
First Cretan 1866 1867 1 1
Ten Years’ War 1868 1878 1 0
Algerian 1871 1872 1 .
Second Apache War 1871 1873 1 0
Second Ashanti 1873 1874 1 0
Tonkin 1873 1885 1 0
Dutch-Achinese 1873 1878 1 0
Red River Indian War 1874 1875 1 0
Balkan 1875 1877 1 1
Third Apache War 1876 1886 1 1
Third Sioux War 1876 1877 1 0
Ninth KafŽr 1877 1878 1 0
Russo-Turkoman 1878 1881 1 0
Second British-Afghan 1878 1880 1 0
British-Zulu 1879 1879 1 0
Gun War 1880 1881 1 0
First Boer War 1880 1880 0 0
Tunisian 1881 1882 1 .
Franco-Indochinese 1882 1884 1 0
Mahdist 1882 1885 0 0
First Franco-Madagascan 1883 1885 1 0
Sino-French 1884 1885 0 0
Russo-Afghan 1885 1885 1 0
Third Anglo-Burmese 1885 1886 1 0
First Mandigo-French War 1885 1885 1 0
First Italo-Ethiopian 1887 1887 0 0
How the Weak Win Wars 125
War Name Start End OUTCOME1 STRATINT2
Second Cretan 1888 1889 1 .
Dehomey 1889 1892 1 0
Second Senegalese 1890 1891 1 0
Messiah War 1890 1891 1 0
Congo Arabs 1892 1892 1 .
Franco-Thai 1893 1893 1 .
Third Ashanti 1893 1894 1 0
Matabele-British War 1893 1893 1 0
Sino-Japanese 1894 1895 0 0
Franco-Madagascan 1894 1895 1 0
Balian 1894 1894 1 .
Cuban 1895 1898 1 0
Italo-Ethiopian 1895 1896 0 0
Fourth Ashanti 1895 1896 1 0
Third Cretan 1896 1897 0 0
Druze-Turkish 1896 1896 1 .
First Philippine 1896 1898 0 1
Sudanese War 1896 1899 1 0
Greco-Turkish 1897 1897 1 0
Indian Muslim 1897 1898 1 0
Nigerian 1897 1897 1 0
Hut Tax 1898 1898 1 0
Second Philippine 1899 1902 1 0
Second Boer War 1899 1902 1 0
Somali Rebellion 1899 1905 1 0
Russo-Manchurian 1900 1900 0 0
Ilinden 1903 1903 1 0
Russo-Japanese War 1904 1905 0 0
South West African Revolt 1904 1905 1 1
Maji-Maji Revolt 1905 1906 1 1
Second Zulu War 1906 1906 1 0
Spanish-Moroccan 1909 1910 1 0
First Moroccan 1911 1912 1 0
First Balkan War 1912 1913 0 0
Tibetan War of Independence 1912 1913 0 0
Second Moroccan 1916 1917 1 .
Arab Revolt 1916 1918 0 1
Irish Troubles 1916 1921 0 1
Yunnan 1917 1918 1 .
First Sino-Tibetan 1918 1918 0 .
Russo-Polish 1919 1920 0 0
Lithuanian-Polish 1919 1920 1 0
Hungarian-Allies 1919 1919 1 0
International Security 26:1 126
War Name Start End OUTCOME1 STRATINT2
Franco-Turkish 1919 1922 0 0
Third Afghan 1919 1919 1 0
Franco-Syrian 1920 1920 1 .
Iraqi-British 1920 1921 1 .
Sanusi 1920 1932 1 0
RifŽan 1921 1926 1 .
Druze Rebellion 1925 1927 1 0
U.S.-Nicaraguan 1927 1933 1 1
Chinese Muslims 1928 1928 1 .
Chinese Civil War 1930 1935 1 0
Manchurian 1931 1933 0 .
Soviet-Turkestani 1931 1934 1 .
Italo-Ethiopian 1935 1936 1 0
Sino-Japanese 1937 1941 1 0
Chankufeng 1938 1938 0 0
Winter War 1939 1940 1 0
Franco-Thai 1940 1941 0 .
Indonesian Independence 1945 1946 0 0
Indochinese 1945 1954 0 0
Madagascan 1947 1948 1 .
First Kashmir 1947 1949 1 0
Palestine 1948 1949 0 0
Malayan Rebellion 1948 1957 1 0
Hyderabad 1948 1948 1 .
Korean Conict 1950 1953 0 0
Sino-Tibetan 1950 1951 1 0
Philippines 1950 1952 1 0
Kenya 1952 1956 1 0
Tunisian Independence 1952 1954 0 .
Moroccan Independence 1953 1956 0 .
Algerian 1954 1962 0 0
British-Cypriot 1954 1959 0 0
Cameroon 1955 1960 0 .
Russo-Hungarian 1956 1956 1 0
Sinai 1956 1956 1 0
Tibetan 1956 1959 1 1
Cuba 1958 1959 0 1
South Vietnam 1960 1965 0 1
Congo 1960 1965 1 0
Kurdish 1961 1963 1 .
Angola-Portugal 1961 1975 0 .
Sino-Indian 1962 1962 1 0
Portugal-Guinea Bissau 1962 1974 0 1
Mozambique 1964 1975 0 1
How the Weak Win Wars 127
War Name Start End OUTCOME1 STRATINT2
Vietnam 1965 1975 0 0
Second Kashmir 1965 1965 0 0
Six-Day War 1967 1967 0 0
Israeli-Egyptian 1969 1970 0 0
Bangladesh 1971 1971 1 0
Philippine-Moro 1972 1980 1 0
Yom Kippur War 1973 1973 0 0
Turko-Cypriot 1974 1974 1 0
Eritrean 1974 1991 0 0
Kurdish Autonomy 1974 1975 1 0
East Timor 1974 1975 0 1
Vietnamese-Cambodian 1975 1979 1 0
Western Sahara 1975 1983 1 .
Chadian Civil War 1975 1988 1 0
Ethiopian-Somalian 1977 1978 1 0
Afghanistan 1978 1989 0 0
Sino-Vietnamese 1979 1979 1 0
Peruvian Civil War 1982 1992 1 0
Tamil Rebellion 1983 1990 0 1
Sino-Vietnamese 1985 1987 0 0
Gulf War 1990 1991 1 0
Iraq-Kuwait 1990 1990 1 0
Kurdish Rebellion 1991 . . 0
Serbian Rebellion 1991 1996 0 1
Russo-Chechen 1994 1996 0 0
Number of cases = 197
A value of . indicates missing data.
1 OUTCOME is the variable label for the conict outcome. Avalue of 1 means that the strong actor
won. Avalue of 0 means that the weak actor won, the war was a stalemate, or that it is ongoing.
2 STRATINT is the variable label for strategic interaction. A value of 0 means that it was a sameapproach
interaction (direct-direct or indirect-indirect). A value of 1 means that it was an opposite-
approach interaction (indirect-direct or direct-indirect).
International Security 26:1 128