Content uploaded by Benjamin Vaughn Allison
Author content
All content in this area was uploaded by Benjamin Vaughn Allison on Feb 28, 2018
Content may be subject to copyright.
Proxy War as Strategic Avoidance
A Quantitative Study of Great Power Intervention in Intrastate Wars, 1816-2010
Benjamin V. Allison
Abstract: States engage in proxy war for numerous reasons, the most important of which is the avoidance
of direct conflict, especially with a state of greater or equal strength. This strategic rationale has gone
unquestioned for some time, as the logic of its underlying assumption seems sound. As a result, few analyses
of the costs and benefits of proxy war exist, and none directly address what is perhaps the most important
cost-benefit question one can ask about such conflicts: Does the strategic use of proxy war actually decrease
the chances of direct kinetic action between two states? Additionally, it is commonly assumed that the Cold
War remained cold because of the use of proxies; but is this actually the case? To answer these questions,
this study uses quantitative methods to analyze an original dataset on relations between the Great Powers
from 1816 to 2010. This study finds that proxy wars do not have a statistically significant effect on direct
wars, but that the possession of nuclear weapons increases the chances of proxy war. These findings cast
doubt on the notion that proxy wars decrease the likelihood of direct wars between two Great Powers. In
addition, I present an original argument for including both state and non-state actors in definitions of
proxies.
Allison 1
States engage in proxy war for numerous reasons, the most important of which is the
avoidance of direct conflict, especially with a state of greater or equal strength. This strategic
rationale has gone unquestioned for some time, as the logic of its underlying assumption seems
sound. As a result, few analyses of the costs and benefits of proxy war exist, and none directly
address what is perhaps the most important cost-benefit question one can ask about such conflicts:
Does the strategic use of proxy war actually decrease the chances of direct kinetic action between
two states? Additionally, it is commonly assumed that the Cold War remained cold because of the
use of proxies, and that these were employed because of the fear of nuclear war (e.g. Towle 1981).
To answer these questions, this study uses quantitative methods to analyze data on relations
between the Great Powers from 1816–2010.
LITERATURE REVIEW & THEORY
There seem to have been two principal periods of academic interest in the phenomenon of
proxy war: the 1980s and the 2010s. This is unsurprising in light of the state of international affairs
in these two periods, as one saw the last decade of the Cold War, and the other – broadly speaking
– witnessed the continuation of the interventionism associated with the so-called “Bush Doctrine”
under the Obama administration.
1
While the latter period entailed greater direct intervention than
the former, it still saw the use of proxies, most clearly seen in the examples of Libya and Syria.
Before assessing whether proxy wars between Great Powers actually decrease the chances
of direct war, proxy war must be defined. Unfortunately, as with so many terms in conflict studies,
2
the definition of “proxy war” is the subject of a seemingly interminable and quite indeterminate
* Special thanks to Dr. Samuel S. Stanton, Jr. (Grove City College) for his assistance with the data analysis for this
paper.
1
See Goldsmith 2016.
2
One of the most infamous examples being “terrorism.”
Allison 2
debate in both the academic and policymaking communities. Perhaps the most useful initial
distinction that may be drawn is between those academics who hold that proxy wars can entail the
use of states or non-states as proxy agents, and those who hold that only non-state actors may be
considered proxies. Each side of this debate is considered below.
Defining Proxies and Proxy War
One of the early influential works arguing that only non-state actors can be used as proxies
is Bertil Dunér’s “Proxy Intervention in Civil Wars” (1981). In his article, Dunér (1981, 359) goes
so far as to posit that “it is impossible to demonstrate a single example of a state acting as a proxy
for some other state.” His argument rests upon the assumption that the would-be proxy (e.g. Cuba
in its direct intervention in Angola) would have to initially decide to intervene as a result of
“positive or negative sanctions (or threats of sanctions) […] directed against it” (Dunér 1981, 357).
If it could be shown that state B had directly intervened at least partly because of the influence of
state A – whether due to A’s power or the compatibility of interests between A and B – Dunér
would presumably count such an event as proxy war by state.
In his book My Enemy’s Enemy, Geraint Hughes (2012, 13-14) holds forth a proposition
similar to that of Dunér, arguing that labeling states as proxies “underestimates the degree to which
a state can follow its own national interests and make strategic choices which may be in accordance
with that of a more powerful ally, but which are not actually influenced by the latter.” The Dunér-
Hughes argument may be thought of as the “agency thesis.” The basic premise of this argument is
that states cannot be used as proxies, because the historical record demonstrates states will only
intervene if they have some interest in doing so. In light of this, Dunér (1981, 359) concludes that
“[i]t may be that proxy intervention serves as the absolute zero. It is never reached, but its existence
is important for a certain theory.”
Allison 3
Between the two sides of the debate over which types of entities may serve as proxies is
another school of thought, best represented by Yaacov Bar-Siman-Tov’s (1984) important essay
on defining proxy wars. While Bar-Siman-Tov (1984, 265-266) holds that either a state or non-
state actor may be considered a proxy, his argument is similar to the “agency thesis,” in that “[t]he
test for defining a war by proxy is not where the war is fought, and not only by whom, but if we
can prove that one actor has been asked by another to fight for him.” He holds fast to the notion
that state A must first ask state B to act as its proxy, and that “if not for the request, or coercion,
[state B] would not act on his own accord” (Bar-Siman-Tov 1984, 265). As such, Bar-Siman-Tov
(1984, 267) posits that two kinds of proxy exist: those who are coerced, and those who voluntarily
intervene due to what Dunér would refer to as “compatibility of interests.” Conceiving of the
relationship between the sponsor state (i.e. the “activator”) and the proxy as “a special type of
patron-client relationship, i.e. a special case of formal or informal alliance between small and big
states,” Bar-Siman-Tov focuses upon the interests of the potential proxy. If the smaller entity
engages in a conflict purely in pursuit of its own interests, the aid of an activator or sponsor is
irrelevant, and the war is not a war by proxy; if, however, the smaller state goes to war because of
the influence of the sponsor, such a conflict is a war by proxy. As a result, conflicts like the Arab-
Israeli wars of the late 1960s and early 1970s would not count as proxy wars, because these
struggles began “because of the small states’ interests, and not as a result of the superpowers[’]
interests, and sometimes even in opposition to them” (Bar-Siman-Tov 1984, 266).
Intriguingly, there is a noticeable absence of arguments in favor of attributing the “proxy”
label to state actors. Rather than attempting to justify such labeling, they assume that states may
operate in a proxy function, and (typically) set parameters as to when a state counts as a proxy.
One extreme example is Yaacov Bar-Siman-Tov’s aforementioned work. Another, less extreme
Allison 4
case is Chris Loveman’s (2002) article on “Assessing the phenomenon of proxy intervention.”
Loveman (2002, 32) considers Dunér’s “compatibility of interests” to be one of the key attributes
of proxy intervention; in his conception, the “principal” (sponsor/activator) and the proxy must
“share a mutual desire to oppose a common enemy,” even if “[t]heir ideologies, motives and
concerns may be different, even antithetical.”
In this study, both state and non-state actors may serve as proxies; however, this is based
on more than a simple assumption. I take issue with the aforementioned “agency thesis.” While of
course I agree that states big or small decide their course of action, this does not necessarily mean
that a definition of proxy war must rest upon the perception of a potential proxy state. On the other
hand, the definition of proxy war used in this study does not rest on a simplistic – some might say
crass – realpolitik view of international relations, perhaps best seen in realist interpretations of
Thucydides’ famous Melian Dialogue (i.e. “the strong do what they can and the weak suffer what
they must”).
3
Rather, I acknowledge the agency of smaller states, but argue that agency alone does
not qualify as a legitimate criterion in determining who may be a proxy.
First, the “agency thesis” is inconsistent in its appeal to the agency of state actors, because
it assumes that non-state actors are more willing to abandon their own interests to pursue those of
a more powerful state. While coercing a non-state actor is ostensibly easier than coercing a state,
for one to apply the “agency thesis” to states, but not to non-state actors – to borrow from Geraint
Hughes’ (2012, 13-14) aforementioned criticism – “underestimates the degree to which a [non-
]state [actor] can follow its own [sub-]national interests.”
Second, proxy war is all about the way the senior partner uses the junior partner, not the
other way around. Generally speaking, the very idea of proxy war rests upon the assumption that
3
For more on the debate over interpretations of the Melian Dialogue, see Alekseeva 2015, Chance 2013, Dobransky
2015, and Kyriakos 2015.
Allison 5
one entity (the sponsor) is more powerful than the other (the proxy), or at least that the proxy needs
or benefits from the aid of the sponsor. While certainly true that some proxies outfox their sponsors
– sometimes even using their own materiel against them, as seen in the US experience in
Afghanistan since the early 1980s – the label of “proxy war” is applied with the understanding that
the proxy is in effect furthering the interests of the sponsor, whether that is the proxy’s primary
objective or not.
With that said, however, it is not possible for a state to unwittingly act as a proxy without
support from the sponsor. For example, Andrew Mumford (2013, 17) posits that by invading Iraq
in 2003, the United States “fulfilled a long-term Iranian ambition by cementing Tehran as a pre-
eminent regional power despite the American’s [sic.] lack of desire for this outcome.” In this case,
I agree with Bar-Siman-Tov’s (1984, 266) assessment that “[t]he fact that an external power may
profit from that war is not enough to define such a war as a war by proxy.” Though in a sense true,
I hold that Mumford’s (2013, 17) assertion that “[t]he fulfilment of a strategic goal by proxy does
not necessarily have to be a conscious or deliberate act” should not mean that if state B effects a
change benefiting state A, then state B is automatically the proxy used in state A’s proxy war. On
the contrary, Iranian interference in Iraq post-2003 is actually an example of an entirely different
(and perfectly legitimate) strategy altogether, similar to what John J. Mearsheimer (2014, 153)
describes as “bloodletting.” Bloodletting entails state A keeping a conflict between its rivals, B and
C, going. When necessary, A supports the losing side, so as to allow the bleeding to continue.
Iran’s meddling in Iraq via the sponsorship of Shia militias during the sectarian violence of 2005-
2009 was an attempt both to keep the Iraqi state unstable and to weaken the United States by
forcing it to keep sending in ground troops; indeed, it is far more accurate to say that Iran’s
Allison 6
sponsorship of Shia militias like the Supreme Council for Islamic Revolution in Iraq (SCIRI) or
the Jaysh al-Mahdī (JAM) constituted proxy war.
Third, subscribers to the “agency thesis” hold that at least part of the reason a proxy initially
goes to war on behalf of a sponsor is the same reason why the sponsor propositions the proxy; in
short, compatibility of interests. This is incorrect for several reasons. First, it ignores the prospect
that A might coerce B, despite the fact that the “agency thesis” also acknowledges the possibility
of A’s use of coercion. Second, even if one argues that such a compatibility of interests is only
important in cases of voluntary proxies, Bar-Siman-Tov’s (1984, 265) assumption that “if not for
[A’s] request, […] the proxy would not act on his own accord,” is illogical, because it dismisses
the notion that A could indirectly support and directly benefit from the actions of B. Of course, this
illogic stems from one of the basic assumptions of the “agency thesis” – that what matters is not
how the sponsor uses the proxy, but how the proxy uses the sponsor – disproven in the preceding.
Another point at which Bar-Siman-Tov’s assumption is problematic, however, is that it presumes
that any benefit A receives from B’s actions can only be gained if B’s initial decision to act is at
least partly predicated on the influence of A, whether through compatibility of interests or coercion.
Proxy war, then, is an indirect form of conflict in which a sponsor state uses another state
or non-state actor to achieve a strategic objective by engaging in conflict with either a state or the
proxy of a state.
4
The sponsor state aids the proxy by supplying financial support; arms and/or
materiel; military advisors and/or intelligence; sanctuary; and, in some cases, “manpower (via
4
Note that this definition does not necessarily adhere to typical political science definitions of war as requiring
1,000 uniformed battle deaths within a calendar year. The terms “proxy war” and “proxy conflict” are used
interchangeably in this paper.
Allison 7
surrogates)” (Mumford 2013, 61).
5
Philip Towle’s (1981, 21) succinct description of proxy war as
“vicarious belligerency” captures the essence of the definition I use here.
Thus, in a dyad of sponsor states A and C, with B acting as A’s proxy, and D acting as C’s
proxy, two types of proxy conflict are possible: one-sided and two-sided. In a one-sided proxy
conflict, the proxy directly attacks its sponsor’s rival (e.g. B attacks C on behalf of A). An historical
example of a one-sided proxy conflict is the USSR’s and China’s support for North Vietnam
(DRV) and the Vietcong against the United States after the passage of the Tonkin Gulf Resolution
in 1964. In a two-sided proxy conflict, the proxies attack one another on behalf of their sponsors
(e.g. B and D fight on behalf of A and C, respectively). US support for South Vietnam (GVN)
against the Soviet and Chinese proxies of the DRV and Vietcong before the Tonkin Gulf
Resolution is an example of two-sided proxy war.
Earlier in this study, it was argued that the label of “proxy war” cannot be applied to a
conflict in which actor B unwittingly serves the interests of actor A without the latter providing
some form of material aid. This phenomenon is better described as an “accidental proxy war,” or
a strategy of “proxy by coincidence,” and should not be conflated with proxy war.
6
Another type
of relationship which does not count as proxy war under the present definition is that of active
military alliance. Both parties in an active military alliance are engaged in combatting the same
enemy or group of enemies, and therefore, to borrow from Towle’s (1981, 21) aforementioned
phrase, the vicarious nature of the belligerency is lost. For example, contrary to what Mumford
(2013, 69ff) asserts, the now-famous Anbar (or Sunni) Awakening of 2006 in Iraq was not an
5
Not all of these come from Mumford (2013), and some components of his list are left out (i.e. soft power, which I
consider somewhat redundant). Rather, they are a list of means of support that I have determined based on (1) the
literature and (2) my own thinking. Mumford is cited here because I use his turn of phrase.
6
This does not contradict my claim that a proxy can be unsure as to the intentions of its sponsor, nor does it mean
that a sponsor who supports a proxy whose conflict happens to coincide with its own strategic goals.
Allison 8
example of proxy war, because the Sunni tribes who switched their loyalties from Al-Qaeda in
Iraq (AQI) to the United States were fighting the same enemy as US forces. On the other hand, US
support for the United Kingdom, Soviet Union, and others through its Lend-Lease policy
constituted a sponsor-proxy relationship, because it was not until late 1941 that the United States
was at war with the Axis Powers.
RESEARCH DESIGN, MODELING & HYPOTHESES
Although most studies of proxy war tend to use qualitative methods to draw conclusions
based on case studies, this paper utilizes quantitative methods, in hopes of detecting general
patterns of conflict. I employ two separate time-series panel-corrected standard error (PCSE)
models to test a number of hypotheses related to the primary query of this study: does engaging in
proxy war actually decrease the likelihood of resorting to direct conflict? Before discussing the
findings of the model, however, I specify the timeframe, unit of observation, variables, and data,
presenting my hypotheses throughout.
Timeframe and Unit of Observation
This study uses an original dataset of proxy wars that took place in intra-state conflicts and
insurgencies from 1816 to 2010.
7
The starting point, 1816, was a logical one, given that (1) it
marked the beginning of the Concert of Europe system initiated with the Congress of Vienna
(1815), and (2) the data for many of the variables I test in this study only go back to 1816 at the
earliest. To measure the year-to-year effects of engaging in proxy wars, I created 32 dyadic pairs
7
Given the lack of coded data on proxy wars, and of time and resources with which to complete this study, it was
unrealistic to compile a dataset including interstate proxy wars, not least because of the changing nature of state
involvement in wars. For example, the US was engaged in proxy war against the Central Powers in World War I,
until it joined the war in 1917, at which point it became a direct war. The Dragon War between China and Vietnam
(1979), in which Vietnam was supported by the USSR, is another example of interstate proxy war. Although
interstate proxy wars are excluded from this study, direct interstate wars are included, since they are
Allison 9
out of the nine Great Powers under study. The dates for each state’s status as a Great Power are
outlined in Figure 1 below.
8
Figure 1: Great Power Status
Austria/Austro-Hungarian Empire: 1816–1918
France: 1816–1940, 1946–1962
Great Britain/United Kingdom: 1816–1968
Prussia/Germany: 1816–1945
Italy: 1861–1943
Russia: 1816–1991, 2007–2010
United States: 1898–2010
Japan: 1905–1945
China: 1946–2010
The unit of observation in both models is the Great Power Dyad-Year. In the dataset, each
Great Power dyad is active for the period(s) of time in which they simultaneously hold Great Power
status. For example, the dyad “United States–Austro-Hungarian Empire” spans the period 1898–
1918, while “United Kingdom–Russia” covers 1816–1968. Because no two dyads have the same
dates, the panel variable (Dyad-Year) is unbalanced, necessitating the use of a time-series panel-
corrected standard error (PCSE) regression model (Beck and Katz 1995; Bailey and Katz 2011).
This model standardizes the analysis’ results based on the standard errors.
8
I assigned Great Power status using two flexible criteria. First, a state had to possess at least 10% of the major
powers’ CINC over a period. Second, after 1945, states also had to possess nuclear weapons to even be considered
for inclusion in my list of Great Powers. In addition to these two criteria, however, I incorporated the overriding
criterion of history; in other words, some states who matched the economic criterion in the pre-1945 world, or both
criteria in the postwar world were excluded, and some who did not match the economic requirement of the pre-
nuclear era were included. For example: Until 1973, there was not a single year in which Japan held the requisite
percentage of the major powers’ CINC score, yet because of its meteoric rise after the Russo-Japanese War (1904-
1905), it was included until its collapse in 1945. Similarly, although Russia had the requisite nuclear and economic
qualifications after 1991 (the latter qualification ended in 2004), they were excluded from 1992-2006, because of the
historical reality of the collapse of the Soviet Union.
Allison 10
Variables
DEPENDENT VARIABLES:
ACTIVE DIRECT WARS IN A DYAD & ACTIVE PROXY WARS IN A DYAD
The primary purpose of this study is to investigate whether the strategic logic of proxy war
is sound; that is, does waging proxy wars really allow a state to avoid going to war with its rival?
To answer this question, two broad types of conflict were entered into the dataset on a Dyad-Year
basis: direct conflicts in a dyad (i.e. between the dyad’s two states) and proxy wars in a dyad.
9
Data on intrastate wars came from a variety of sources, including Cunningham et al.’s (2009 and
2012) Non-State Actor Data, Högladh et al.’s (2011) Uppsala Conflict Data Program (UCDP)
External Support Dataset, and general reference works like Wikipedia and Encyclopedia
Britannica.
10
I derived data from direct interstate conflicts from general reference works.
In each Dyad-Year, I entered data on each conflict using a range of variables. Those
variables that are critical to this study, however, are direct_active_dyad (number of active direct
conflicts in that Dyad-Year; hereafter DAD) and proxy_active_dyad (number of active proxy
conflicts in that Dyad-Year; hereafter PAD). Both are count variables. In the first model, DAD is
the dependent variable. PAD is the dependent variable in the second model.
INDEPENDENT VARIABLE:
ACTIVE PROXY WARS IN A DYAD
The first model, testing DAD as the dependent variable, uses PAD as the independent
variable, with a host of control variables. In the second model, however, I run a number of control
variables against PAD, technically without a sole independent variable. I outline the control
variables for both models below.
9
See ADP and ADE in the “Control Variables” section for more details on how different types of direct conflicts are
differentiated.
10
I used Wikipedia as a starting point for entering basic data (participants and years) for conflicts not covered in
Cunningham et al. (2009 and 2012) or Högbladh et al. (2011), and I cross-checked the Wikipedia information with
other sources when necessary. This is hardly the first time that Wikipedia has been used as a general reference in
creating a dataset; indeed, Gleditsch et al. (2013) used it several times in coding their data.
Allison 11
Hypothesis 1: The more that Great Powers engage in proxy war against one another, the
less likely they are to engage each other in direct war.
11
CONTROL VARIABLES
This study makes use of a number of control variables. In addition to using lags of DAD
and PAD, use “pure” (active_directpure; hereafter ADP) and “external” (active_directext;
hereafter ADE) variants of DAD to determine the types of direct conflicts in a dyad. For example,
an ADP is a conflict in which two states in a dyad engage in direct warfare on the territory (original
or conquered) of one of those two states; consider France and Germany in World War I. Likewise,
an ADE is a conflict in which both states engage in direct warfare, but that warfare takes place in
the territory of another state outside of the dyad, and is often not a declared war; consider the direct
conflict between Chinese and American troops during the Korean War.
I assess whether one or both sides possessed nuclear weapons (nukebin_dyad) and the
degree of (dis)parity between their nuclear forces (nuke_parity). I also make use of the Composite
Index of National Capability (CINC) scores of each state to measure their National Material
Capabilities; these CINC scores are compared in the variable cinc_parity, which measures the level
of (dis)parity between two state’s scores. Additionally, the level of (dis)parity between two states’
military personnel (milper_parity and lags) are taken into account. I also include a pair of
geographic variables measuring whether the countries in a dyad were separated by a significant
body of water (water_dyad) and whether they shared a border (border_dyad).
12
Data sources for
the control variables are summarized in Table 1 as needed.
Hypothesis 2: If either or both states in a dyad possess nuclear weapons, that dyad will
experience more proxy wars than direct wars.
Hypothesis 3: The greater the disparity in two states’ Composite Index of National
Capability (CINC) scores, the more proxy wars that dyad will engage in.
11
This is the null hypothesis, as it is a fairly common assumption, held by policymakers and academics alike.
12
On the “stopping power of water,” see Mearsheimer 2014.
Allison 12
Hypothesis 4: If a significant body of water exists between the two states in a dyad, there
will be less direct wars in that dyad.
Hypothesis 5: If the states in a dyad share a border, they will engage each other in direct
war more often than they engage each other in proxy war.
Hypothesis 6: The greater the disparity between two state’s military forces, the greater the
chances of that dyad experiencing direct war.
Table 1: Control Variables and their Sources
Control Variable
Sources
Nuclear Weapons
Norris and Kristensen 2010
CINC Scores
National Material Capabilities Dataset13
Military Personnel
Ibid.
Modeling
To test the above hypotheses, I run two time-series PCSE models. In the first, I run DAD
as the dependent variable, PAD as the independent variable, with the following controls: 1-, 2-,
and 3-year lags of PAD; cinc_parity; milper_parity (with 1-, 2-, and 3-year lags); nukebin_dyad;
nuke_parity; water_dyad; and border_dyad. This first model tests hypotheses 1, 2, 4, 5, and 6.
In the second model, PAD is the dependent variable, with control variables of cinc_parity,
nukebin_dyad, nuke_parity, and border_dyad. It also runs 1-, 2-, and 3-year lags of DAD, ADP,
and ADE. This second model tests hypotheses 2, 3, and 5.
FINDINGS
Both models are strong and well-specified, as indicated by their Wald chi2 and Prob > chi2
scores (Model 1: Wald chi2(13) = 22.32, Prob > chi2 = 0.0506; Model 2: Wald chi2(16) = 540.49,
Prob > chi2 = 0.0000). Additionally, there is a fairly strong relationship between the variables in
both models. Given the strength and fit of the model, I now assess the validity of my hypotheses.
Hypothesis 1: The more that Great Powers engage in proxy war against one another, the
less likely they are to engage each other in direct war.
13
See Singer et al. 1972; Singer 1987.
Allison 13
Model 1 indicates that the null hypothesis (1) is indeterminate. Engaging in proxy war does not
decrease (or increase) the chances of State A engaging in direct conflict with State B in a
statistically significant way. This unexpected finding seemingly turns the strategic logic of proxy
war on its head, as discussed in the conclusion.
Table 2: Summary of Time-Series Regression PCSE Models
Active Direct Wars
Active Proxy Wars
Coefficient
Std. Dev.
Coefficient
Std. Dev.
(Constant)
0.119476
0.028103
0.132891
0.027151
PAD
0.034024
0.043266
padlag
0.003316
0.056921
padlag2
-0.03686
0.05739
padlag3
0.007518
0.044033
water_dyad
0.000413
0.017442
cinc_parity
-0.01468**
0.005945
-0.04152***
0.008556
nukebin_dyad
-0.00243
0.039139
0.540266***
0.08693
nuke_parity
0.009892
0.021292
-0.11403***
0.043115
border_dyad
-0.02665*
0.01534
-0.04783***
0.010342
milper_parity
0.013713
0.011364
-0.00884
0.01222
mperlag
-0.0136
0.013808
-0.00867
0.013986
mperlag2
0.013249
0.013763
0.00492
0.014862
mperlag3
-0.01733
0.011354
0.017281
0.012718
DAD
0.232711***
0.084279
dadlag
0.128474
0.094917
dadlag3
-0.20087***
0.055508
ADP
-0.30599***
0.088335
adplag
-0.1287
0.095187
adplag2
-0.02374
0.059405
adelag2
0.079598
0.105972
adelag3
0.16127
0.105458
N = 2,248 (both models). *p < .1, **p < .05, ***p < .01. In Model 2, dadlag2, adplag3,
active_directext, and adelag were omitted due to collinearity. Linear Equation (Model
1): DAD = 0.119 – 0.015 cinc_parity + -0.027 border_dyad. Linear Equation (Model
2): PAD = 0.133 – 0.0415 cinc_parity + 0.54 nukebin_dyad – 0.114 nuke_parity -0.0478
border_dyad + 0.233 direct_active_dyad – 0.2 dadlag3 – 0.306 active_directpure.
Intriguingly, Model 2 suggests that a dyad that sees a direct conflict has a 23% chance of
experiencing a proxy conflict that same year. This is not surprising, given that it is not uncommon
for warring states to attempt to undermine their enemies elsewhere. Model 2 also indicates that if
a dyad has experienced direct conflict of any sort for four years, it results in a 20% decrease in the
Allison 14
likelihood of a proxy conflict occurring during that fourth year, ostensibly because the states
involved would likely refocus their efforts on the primary direct conflict, rather than “nibbling at
the edges.” Likewise, a “pure” direct conflict between two Great Powers decreases the likelihood
of proxy conflict in that year by almost one-third (-30.6%), which follows a similar strategic logic.
Hypothesis 2: If either or both states in a dyad possess nuclear weapons, that dyad will
experience more proxy wars than direct wars.
Results from Model 2 demonstrate that the possession of nuclear weapons increases the chances
of states engaging in proxy war. Dyads in which both states possessed nuclear weapons were more
likely to experience proxy war than dyads in which only one or neither state had such weapons. In
fact, when either or both states in a dyad possess nuclear weapons, there is a 54% increase in the
likelihood that they will engage in proxy war. In a seeming contradiction of that finding, however,
dyads with a greater disparity in nuclear arsenals are 11% less likely to experience proxy war.
14
Unfortunately, there were no statistically significant results relating to this hypothesis in Model 1,
and thus the present research cannot quantitatively assess the impact of nuclear weapons on direct
conflicts.
Hypothesis 3: The greater the disparity in two states’ Composite Index of National
Capability (CINC) scores, the more proxy wars that dyad will engage in.
Hypothesis 4: If a significant body of water exists between the two states in a dyad, there
will be less direct wars in that dyad.
Hypothesis 5: If the states in a dyad share a border, they will engage each other in direct
war more often than they engage each other in proxy war.
Hypothesis 6: The greater the disparity between two state’s military forces, the greater the
chances of that dyad experiencing direct war.
14
This finding may hold implications for the debate over the “nuclear taboo.” See Tannenwald 1999, 2005, and
2010; Atkinson 2010; Hanania 2017; Paul 1995 and 2010; Press et al. 2013; and Walker 2010.
Allison 15
While results in both models for Hypothesis 3 were statistically significant, they lacked substantive
significance. The greatest change in the probability of proxy war in a dyad was found in Model 2,
which found that dyads with greater disparities between two states’ CINC scores were slightly less
likely to engage in proxy war (-4.15%); direct wars, on the other hand, were 1.46% less likely
under such circumstances. Unfortunately, the DAD model showed no statistical significance when
testing the “stopping power of water” (Hypothesis 4) in relation to direct conflict. Tests of
Hypothesis 5 yielded statistically significant results, but showed that shared borders between Great
Powers only marginally decreased the chances of direct (-2.67%) and proxy (-4.78%) war.
Unfortunately, neither model returned statistically significant results when testing for a
relationship between (dis)parity in military forces and the likelihood of engaging in direct or proxy
war (Hypothesis 6).
CONCLUSIONS
This study finds that there is no statistically significant relationship between proxy war and
direct war in Great Power dyad-pairs for the period 1816–2010, thereby calling conventional
wisdom into question. While the lack of statistically significant results does not mean that the
strategic logic of proxy war has been debunked here, it has most certainly not been reinforced.
Unsurprisingly, the present research also finds that there is a strong relationship between the
presence of nuclear weapons and the use of proxy war in a Great Power dyad. This finding,
however, does not necessarily validate the strategic logic of proxy wars, as there is no statistically
significant relationship between direct war and the presence of nuclear weapons, nor between
proxy war and direct war (i.e. proxy wars do not necessarily decrease the likelihood of direct wars),
though active direct wars in a dyad-year do increase the likelihood of a proxy war occurring in that
dyad-year. It is worth noting, however, that this increase in the use of proxy war during the
“nuclear age” ostensibly also has a connection to the general improvement in communication
Allison 16
technologies over time. For example, communication between an insurgency and its state sponsor
would have been much slower in 1918 than in 1999, simply because of the evolution in
telecommunications over that period. Future research should investigate this potential correlation.
The present paper provides a jumping-off point for future research. Several potential
avenues for such research exist. First, a dataset of intra- and interstate conflicts, built using the
definition of proxy war provided in this study, could provide more concrete answers to the
questions posed here. Second, data at the Country-Year level, considering all direct and proxy
conflicts in which that state participated that year, could also provide valuable insights into the
true efficacy of such a strategy.
If nothing else, my research should give Great Power policymakers pause before they
decide to engage in proxy wars with their rivals. Although true that the Cold War did not see any
open, direct “hot” wars between the Great Powers, they did still fight one another, though in a
covert manner, both alongside and through proxies. Though such conflicts may have stayed “cold”
due to the possession of nuclear weapons on both sides, the results generated by this study – or
rather the lack thereof – indicate that perhaps policymakers should be less prone to trust the “safety
net” of deterrence.
Allison 17
BIBLIOGRAPHY
Alekseeva, T. A. 2015. “Re-reading the ‘Classics’: Thucydides and Political Realism.”
Sravnitelʹnaâ Politika 6, Iss. 3, No. 20: 7-20.
Atkinson, Carol. 2010. “Using nuclear weapons.” Review of International Studies 36: 839-851.
Bailey, Delia and Jonathan Katz. 2011. “Implementing Panel-Corrected Standard Errors in R:
The psce Package.” Journal of Statistical Software 42, Coding Snippet 1: 1-11.
Bar-Siman-Tov, Yaacov. “The Strategy of War by Proxy.” Conflict and Cooperation 19, No. 4:
263-273.
Beck, Nathaniel and Jonathan N. Katz. September 1995. “What to do (and not to do) with Time-
Series Cross-Section Data.” The American Political Science Review 89, No. 3: 634-647.
Braumoeller, Bear F. February 2008. “Systemic Politics and the Origins of Great Power
Conflict.” American Political Science Review 102, No. 1: 77-93.
Brown, Seyom. “Purposes and pitfalls of war by proxy: A systemic analysis.” Small Wars &
Insurgencies 27, Issue 2: 243-257.
Castellano, Isaac. August 30, 2012. “Are Proxy Wars Cheaper?: Civil War Intervention and
Security Outcomes.” Paper presented at the American Political Science Association
Annual Meeting, New Orleans, Louisiana.
Chance, Alek. 2013. “Realpolitik, Punishment and Control: Thucydides on the Moralization of
Conflict.” Journal of Military Ethics 12, No. 3: 263-277.
Cunningham, David E., Kristian Skrede Gleditsch, and Idean Saleyan. August 2009. “It Takes
Two: A Dyadic Analysis of Civil War Duration and Outcome.” Journal of Conflict
Resolution 53, No. 4: 570-597.
————. January 24, 2012. Codebook for the Non-State Actor Data. Version 3.3.
http://privatewww.essex.ac.uk/~ksg/eacd.html (November 25, 2017).
Dobransky, Steve. 2015. “The Tragic Script of Thucydides in Political Science.” Peace Review
27, No. 2: 217-227.
Dunér, Bertil. 1981. “Proxy Intervention in Civil Wars.” Journal of Peace Research 18, No. 4:
353-361.
Gleditsch, Kristian Skrede, David Cunningham, and Idean Salehyan. November 23, 2013. Non-
State Actor Data: Version 3.4. Case Descriptions. http://privatewww.essex.ac.uk/~ksg/
eacd.html (November 25, 2017).
Allison 18
Gleditsch, Nils Petter, Peter Wallensteen, Mikael Eriksson Margareta Sollenberg and Håvard
Strand. 2002. “Armed Conflict 1946-2001: A New Dataset.” Journal of Peace Research
30, No. 5: 615-637.
Greig, J. Michael and Andrew J. Enterline. February 1, 2017. Correlates of War Project:
National Material Capabilities (NMC) Data Documentation. Version 5.0.
http://www.correlatesofwar.org/data-sets/national-material-capabilities/nmc-codebook-
v5-1 (November 25, 2017).
Groh, Tyrone L. February 23, 2010. “War on the Cheap? Assessing the Costs and Benefits of
Proxy War.” Ph.D. diss. Georgetown University.
Gurr, Ted Robert and Harr Eckstein. 1975. Patterns of Authority: A Structural Basis for Political
Inquiry. New York: John Wiley & Sons.
Hanania, Richard. Spring 2017. “Tracing the Development of the Nuclear Taboo: The
Eisenhower Administration and Four Crises in Asia.” Journal of Cold War Studies 19,
No. 2: 43- 83.
Högbladh, Stina, Therése Pettersson and Lotta Themnér. March 16-19, 2011. “External Support
in Armed Conflict 1975–2009. Presenting new data.” Paper presented at the 52nd Annual
International Studies Association Convention, Montreal, Canada.
Hughes, Geraint. 2012. My Enemy’s Enemy: Proxy Warfare in International Politics. Sussex,
UK: Sussex Academic Press.
Kyriakos, Milekis. 2015. “Realist Stronghold in the Land of Thucydides? Appraising and
Resisting a Realist Tradition in Greece.” European Quarterly Of Political Attitudes And
Mentalities 4, No. 4: 15-32.
Lebow, Richard Ned. 2011. Why Nations Fight: Past and Future Motives for War. Cambridge,
UK: Cambridge University Press.
Marshall, Monty G., Ted Robert Gurr, and Keith Jaggers. 2017. PolityTM IV Project: Political
Regime Characteristics and Transitions, 1800-2016. Dataset User’s Manual.
http://www.systemicpeace.org/inscr/p4manualv2016.pdf (November 25, 2017).
Mumford, Andrew. 2013. Proxy Warfare: War and Conflict in the Modern World. Cambridge,
UK: Polity Press.
Norris, Robbert S. and Hans M. Kristensen. 2010. “Global nuclear weapons inventories, 1945–
2010.” Bulletin of the Atomic Scientists 66, No. 4: 77-83.
Paul, T. V. December 1995. “Nuclear Taboo and War Initiation in Regional Conflicts.” The
Journal of Conflict Resolution 39, No. 4: 696-717.
Allison 19
————. October 2010. “Taboo or tradition? The non-use of nuclear weapons in world
politics.” Review of International Studies 36, No. 4: 853-863.
Potter, Joel and John L. Scott. 2010. “Issues in third-party intervention research and the role of
destruction in conflict.” The Economics of Peace and Security Journal 5, No. 1: 26-29.
Press, Daryl G., Scott D. Sagan, and Benjamin A. Valentino. February 2013. “Atomic Aversion:
Experimental Evidence on Taboos, Traditions, and the Non-Use of Nuclear Weapons.”
American Political Science Review 107, No. 1: 188-206.
Salehyan, Idean. June 2010. “The Delegation of War to Rebel Organizations.” The Journal of
Conflict Resolution 54, No. 3: 493-515.
Singer, J. David, Stuart Bremer, and John Stuckey. 1972. “Capability Distribution, Uncertainty,
and Major Power War, 1820-1965.” In Bruce Russett (ed) Peace, War, and Numbers.
Beverly Hills: Sage, 19-48.
Singer, J. David. 1987. “Reconstructing the Correlates of War Dataset on Material Capabilities
of States, 1816-1985.” International Interactions 14: 115-132.
Tannenwald, Nina. Summer 1999. “The Nuclear Taboo: The United States and the Normative
Basis of Nuclear Non-Use.” International Organization 53, No. 3: 433-468.
————. Spring 2005. “Stigmatizing the Bomb: Origins of the Nuclear Taboo.” International
Security 29, No. 4: 5-49.
Walker, William. October 2010. “The absence of a taboo on the possession of nuclear weapons.”
Review of International Studies 36, No. 4: 865-876.
Yeisley, Mark O. Winter 2011. “Bipolarity, Proxy Wars, and the Rise of China.” Strategic
Studies Quarterly: 75-91.