lllllllllllllllllllllllllllllllllllllllllllllllllllllll

31293 01812 7260

This is to certify that the
dissertation entitled

Signaling Commitments by Investing Costs:
The Provision of Extended Deterrence via Alliance

presented by

Christopher Matthew Sprecher

has been accepted towards fulfillment
of the requirements for

 

Ph.D. degree in Political Science

 

Majorp ffessor ;

Date June 30, 1999

MSU i: an Affirmative Action /Equal Opportunity Institution 0-12771

 

 

 

LIBRARY
Michigan State
University

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINFS return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

~~x

   
  
  
 
 
 
   
  

- .31: i' .
i s - ‘ i» i; Costs: The Provision of Extended Deterrence
via Alliance

I m lx'i’in
' ,. '.-‘.' .»‘_’ Christopher Matthew Sprecher
' ‘- _ tdk‘ui'

mugs: A

’ awn: iii‘x‘. ,
‘ ADISSERTATION
- U .

pm Submittedto
“9:, ‘ . Michigan State University
' j fulfilhnentofthe requirements
’ i ' forthe degree of

'- ‘v' . ice-(n

wir a,» 1,; 3 n Q§OPHY
5- ~ - .5 '1: n.7,” ‘
. .

  

 

   

 

      
    

 

By

     

6. Christopher Matthew Sprecher
"i l ‘ . ‘

thgain little in terms of security from allying with minor powers, while
White in fear of being abandoned by their major power patrons. However. as
”41991) theorized, major states gain some autonomy benefits from such alliances,
mgthminor power(s) in the alliance gain security. In this dissertation I conceptualize
WMenence as a benefit that great power defenders proffer to minor power allies.
Mamas are able to receive benefits from their minor power allies that permit the
Memories to maintain a global presence. Minor powers gain security from this
Win, with the major power’s capabilities enhancing the deterrent aspect of the
“my,“lmintain that the credibility of such a deterrence commitment can be developed
W defender uses costly signals as a means of demonstrating their
WM ally’ 3 defense These costly signals serve as investments made by

' .1; . 1;!“ minor power. These investments provide cues to potential

  
  
  
  

a {~- - as indicators of how committed a defender is to protecting an ally.
' . -. in this dissertation as security benefits provided by

I m the great power defender must receive some benefit

 

   
  
   
    

~ . I maintain that the autonomy benefits the protégé is willing
’ .. gaining security also send signals to a potential attacker, and help
. I. M a defender has for its protégé.
’ 4 - general framework of security/autonomy 1 construct a signaling model
« ' via alliance. I argue that perceptions about the strength of the
' .. Maurine whether an aggressor state makes a demand of a great
‘ . . - ~ lempirically test hypotheses derived from the signaling model on a set
m dawn from the nineteenth and twentieth centuries. Each of these alliances
bum component explicated in the reason for its existence. I utilize a
m [l'obitwith selection to analyze data pertaining to signal strength, benefits
Wfiflhe major power, the balance of forces, domestic costs, and international
reputation costs. I analyze both extended general and extended immediate deterrence,
bother individual equations and as part of a larger, sequential process. I find that
changes in the strategic environment play a role in determining whether extended general
(Harm succwds or fails. This is largely a function of the investment of time and
money that the defender has made in the relationship. At the level of extended immediate
We I demonstrate that great power defenders are more likely to intervene if major
Mommy benefits are threatened. This holds even if the defender is weaker than the
mm. The findings of the dissertation suggest that alliances are better at

“Womban has previously been assumed, and that if enough costs are

 

 

 

 

,Mmisr....

M round <--: r‘w

Ml Hulk 1“ ’
. 3
my: In m‘
.1!
id.
.dtheu tinny
.t" . '
=~i

'Jynufrrw'
' . denier: .,

l
I .

“'i‘n) .L' " f
ottlww

. them. f
To Jill, with love

 

thirty yaw 4":

I ,.
,v'w
ml H ._ »

my”! in my shirt Mara L'A-

 

this dissertation, I am struck by all of

  
   
 
 
  
 
 
  
 
 
 
 

~ m. Myfirstmundofthanksmustgotomy
k . I ' ' W, Paul Huth, Mark Jones, and Michael

- many ways to my education here at Michigan

- . .., of their time, providing me with feedback and

' ' » if to my numerous e—mail messages, and allowing me

My ejecting me when I had overstayed my welcome!
. .. 5 . . -.-- ,. . ; who are no longer at Michigan State: Gretchen
fixBoth of them taught me much about international relations

' " model them. I hope that they are able to see their influence
‘ alllagttes, both at Michigan State and elsewhere, played a substantial
. ' of the ideas presented here. At Michigan State l would like to
WW Chris Butler, Erick Duchesene, Ada Finifter, Mark Hurwitz,
WW Kuenzi, Sara McLaughlin (now Mitchell), David Lektzian, Alan
Poi-mm Power, Brandon Prins, and Mark Souva for comments on these and other
WW matters. At other institutions I would like to thank John Freeman, Erik

(twat-tum, Bob Muncaster, Bill Reed, Michael Simon, Dina Zinnes and the

~ atthc 1999 Junior Masters Class at the University of Illinois.

 

 

 

 

  
  
 
 
  
  
     
 
  
  
  
  
   
     
 

3,: i);
" :‘W “piper” is finally completed. My mother-in-

«- - sinus lave also endured this process and

., ‘ ii. I’d like to thank them, as well as Sally, for all

p w. :4 . 7 «u v remain. The first is to Blackie, Tigger, and
.l’

., . entertained me for the last two years as I have labored on
.‘ l.~l‘
. My ..

.v.‘v- ‘1-

‘WW

dismay, they have at times served as actors w1th1n my
a. :0 ~.
-‘ wane"

.WI 4

Str (well, but I believe that they have achieved a form of
ob -.

M of mere humans.

. to explain my theoretical ideas to others. Unfortunately they
andP *

eitpress my love and appreciation to my wife Jill. She has
fir as long as I have, and at times with considerably more
UM'. ’ J'. i "
Milieu I became frustrated, and helped me work through my
A5: .

, m'her'leve and support, this dissertation would have never been

 

W Romano m ..

‘ moon-av um

...... . u... .. ....
7‘ ‘9 i ' «or... «00”.. «u "up. n.” u .
. , Q
YE“ -~ : . .
. ,._ . I .

7:13 111'

 
 

 

f
f

 

  
 
 
 

 

 

'lhePuzzleandResearchQ "

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

ll 1
1.2 Contributions of This Study 4
. 111:3 Modeling Strategy 6
1.4 Plan of the Dieeerratinn 1 1
Chapter-2 Literatm'eandF r‘ " “ 13
2.1 ‘ ‘ ‘ " 18
22 Concepmalizing Deterrpnrp 19
2.3 Alliances in World Politics 20
2.4 Linking Alliances to Extended Deterrence 38
2.5 The Rational Deterrence Debate 42
2.6 Conclusion 49
Chapter 3 A Theory of Extended Deterrence via Alliance:
Signaling Commitments by Investing Costs 51
3.1 ‘ “ ’ " 51
3.2 The Level of Analysis Problem in Deterrence Situations and
The Agent-Structure Debate 54
3.3 A Conceptual Framework: Opportunity and Willingness as Ordering
Concepts in the Study of Extended Deterrence and Alliances ...................... 58
3.4 Willingness and the Pursuit of National Goals 6';
3.5 Signaling Commitments, or how to establish a Credible Reputation ............ 76
My.” _‘.(" ' ' an
Wa’l‘m Formal Models of Extended Deterrence via Alliance .......................... 82
9.} lat! * " 27
'lk'flleaetieal Model with Complete Information in
, (.‘aeetsJ’ayofisand'l‘y“M 01
: “Mk TheGameflsEquilibfin 113

 

 

  

W7

7.1
7.2
7.3
7.4

 
  
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

, 1 '. , - . - Uncertainty to the Model: Perceptions and Signals ....................... 1 17
. ‘ ,. .1 , Deterrence via Alliance with Incomplete Information
,- M M18 120
7. .conclusion 133
, ’ _' ‘hepositions, Cases, and Variables 137
I a A. A: 137
Propositions 139
The Problem of Selection Effects in Analyzing Deterrence Situations ...... 157
The Set of Alliances 162
The Set of Crises: Measuring Deterrence Failure 166
Operationalization of the Independent Variables:
Measuring Influences on Willingness 171
Descriptive Statistics: A First Cut at Analysis 18 5
Conclusion 187
Empirical Tests 190
' ‘ ‘ " 190
The Probit Framework 191
The Bivariate Probit Model with Selection:
Accounting for Selection Bias om:
Conclusion 713
f‘ ‘ 771
Conclusion 771
Implications from the Formal Analysis: Integrative Cumulation ............... 221
Implications from the Empirical Analysis: Additive Cumulation .............. 225
Concluding Remarks and Directions for Future Research ......................... 228

Appendix A Proof of the Equilibria from the Model with Complete Information ...... 234

Appendix B Derivation of Beliefs via Bayes’ Rule

Appendix C Proofs of the Equilibria of the Signaling Model with Incomplete

Appendix D Deterrence Alliances, 1870—1984
Wile-phi

 

Information

742

 

743

754

 

771

 

 

mm:
were!“
refms ‘ 7

l'U’l ’_

Table :14:
Table 4.5:
Table 4.6:
Table 5.1:

Table 5.2:
Table 5.3:
Table 5.4:
Table 5.5:

Table 6.1:
Table 6.2:
Table 6.3:
Table 6.4:
Table 6.5:

Table 6.6:

Table 6.7:

 

LIST OF TABLES

Terms in the Model ....................................................... 96
Payoff: for the Model .................................................. 102
Preference Orderings of Each State .

(Ordinal Outcomes ...................................................... 109
Defining Actions in the Signaling Model ........................... 126
Equilibria under Complete Information .............................. 131
Equilibria under Incomplete Information ........................... 131

Alliances in Which Extended Deterrence

Is Attempted ............................................................. 166
Frequency Distributions of EGDSUCC .............................. 169
Frequency Distributions of EIDSUCC .............................. 171
Descriptive Statistics ................................................... 186
Distribution of the Model’s Outcomes ............................... 188

Parameters and Variables Influencing
Extended General Deterrence Success ............................... 195

Parameters and Variables Influencing
Extended Immediate Deterrence Success ........................... 196

Probit Estimates of Extended General
Deterrence Success ..................................................... 199

Probit Estimates of Extended Immediate
Deterrence Success ..................................................... 200

Substantive Effects of Independent Variables
On Extended General Deterrence Success ........................... 203

A Unified Model of Extended General Deterrence
And Crisis Escalation Using Bivariate Probit Analysis ............ 212

Substantive Effects of Independent Variables on Extended
Immediate Deterrence Success ........................................ 215

 

ION

Escalation ..... ............. .62
’ ' “Aban- \' a
‘ :1“ ~ ~ Framework of Costly
. , limimorExrended
alumnae ...... ....... ..........80

-‘med. .1
viaAlliance with

e. ‘ .~ viaAlliancewith
Information ................ 136

I I"
also seen ;:, ~
1; 1

flexing: ..

atlmti.x ' II

,Ulr “

:l" .
, We buzznu 1, . '

N- 12111}: -1 111.1 I

. has
M .WIscnwlu.v. 13 “1.1.31.3,“

-’--1“R '
W
H- -R
'b (“A v‘fi’

 

1"

mm CHAPTER 1

 

WI “03 I" INTRODUCTION

l‘Ll'. ‘~ '1

1.1 The puzzle and research question

i i I Inthe aftermath of the Cold War, there was much euphoria surrounding the North
Atlantic Treaty Organization’s (NATO) success at “defeating” the Soviet Union. The
Cold War ended with, it was maintained, a victory for the West. without the specter of
superpower conflagration. This victory emerged in spite of various crises that had
erupted between the superpowers during the Cold War era, particularly on the issue of

divided Germany.

 

This victory was also seen as a triumph for the theory of deterrence, which was
thedriving force behind the existence of the NATO alliance. In effect. NATO became
the symbol for successful alliance policies. However, such success is not commonplace
within the realm of international politics, particularly if one examines the functioning of
international alliances. In fact, much evidence suggests that alliances are quite
ineffective at deterring aggression and preventing war.

Take, for example, the case of Poland prior to the Second World War. Poland and
the United Kingdom were bound together by a common alliance, in which the British
pledged to come to Poland’s defense in the case of an attack. Of course, as history
demonstrated, the deterrent capacity of the British-Polish alliance did not have the
desired effect, and Germany invaded Poland, launching the Second World War.

What explains such discrepancies in the impact of existing alliances? In the case
of NATO, the United States’ presence in the alliance is credited with preventing the

comm of war. In the British-Polish case, the existence of a defense pact between the

 

 

 

   
  
 
 
  
  
  
  
  
  
  
  
 

1 _‘~ c.» aWhile-these are but two cases
"1 ~ 1 w the use, it is indicative of a larger alliance
' beingformedifthey are not successful at

.1 ,1. .

-‘ 2 'on, lies in the underlying logic of why

dyeing tools of capability aggregation, I argue that
I i ‘ .1? 3- benefitsto their members. The context in which they
1 i ., . if we are to understand why and under what conditions
" - :z i s : deterrence effective. I maintain that is through the
' - . can be viewed as investments made by a major power in a
. 1mm deterrence.
.. this study on a set of twenty alliances that were formed in the
. ' , 1. a. A have the deterrence of an adversary as their common goal. I argue
W of alliances is much more reliable than many earlier studies have
WWW] tend to be successful at deterring aggression. I maintain that
m which contradicts many of the other findings within the alliance/war
Wt of an inadequate understanding of why certain alliances form.
two OW twofold problem is often encountered in the conduct and study of
maturity and conflict issues, and the study of deterrence and alliances is no

*1 . Moduli, information is often at a premium to decision-makers. Potential

.e--

\f.‘ 15°“ .1 ,

   
     

.3 Wthe true strength of the foe they are facing, and may be
V '_ Fa can: 3 -
"i 3 “chased on the perceptions they have about their

. . talents: ,

 

   
  
 
 
  
 
 
  
  
   
 
 

”invention that costs in both the
= tenders; The pmsuit of international
11» -- orto attack states that arejoined by a common

'- may ameliorate such endeavors.

 

impact that an existing alliance has on the
m .escalme a militarized crisis. If alliances are

1' . ..-, 1- capabilities, as most power-based theories of
. - Ind their members are supposed to be reliable, then why
:4 1m by adversaries? This puzzle has not been adequately
. bmpirical studies of crisis initiation and escalation. or by
' '1~ . . inspect which alliances have on the occurrence of war.

1.1 . of earlier empirical studies lies, I contend, first in their ;
1 the sequential strategic interplay that is inherent in international
WWO! actions that any state takes is conditioned by the potential
W in the international system. Interactions tend to be of a more
M, tether than static. To address this dynamic role of an alliance’s
m we need to account for the sequence of moves that takes place between
MW‘ r
walk it ..1. 11
“h WI! tau :‘ .

    
  
   
     

.3 ~1 of alliances and their deterrent role are static Nor are all
1'. 1 For example, game-theoretic literature on alliances and deterrence
" , .9: ' . (1995:, 1995b); and Sorokin (1994). In a strictly
. . alliances and war has been examined from a more dynamic
‘ ; ocean and Hoole (1978); Wayman (1990); and Zinnes et.

 

 

 

 

 

wdmgnethat the context within which alliances are formed must be
mm don’t understand why the alliance exists. we cannot hope to understand
why it “to war instead of peace.

. Mi 1 AW note regarding what this dissertation does not concern. While much of the
Won alliances and war focuses on alliance reliability. this is not the central focus
oftherelearch being conducted here. While the focus on reliability does inform this
sully to anextent, the contradictory results found within the literature make the setting
lipeforu different focus on the impact of alliances on the initiation and escalation of
other.

1.2 Contributions of this Study

To address these issues, I construct and empirically test two game-theoretic models of
crisis initiation and escalation. Through the use of these models I am able to answer four
questions of interest:

I. What role do alliances play in ensuring effective extended deterrence?

2. What impact does an existing alliance have on crisis initiation and escalation?
3. What role does limited information play in crisis initiation and escalation?
4. What domestic and international factors influence alliance behavior?

This dissertation builds upon and expands earlier work in a variety of ways. First, it
focuses on the processes that lead to war. Most empirical studies relating the existence of
alliances to war have concentrated merely on the war variable, and paid scant attention to
the events that precede the conflict. Secondly, it emphasizes the costs and limited

information that states face in a crisis situation. Third, the dissertation extends the

.‘I‘
‘. .-

.

;

, .

 

 

....
‘.
, .
.
s
\

 
 
  
 
  
  
  
  
  
  
  
 

.‘.
I1

V ’ phwlntheeontextofthisthird
'f‘lf“: ‘edeterrence alliance.
J ‘ . commitments credible, great powers
‘7‘ w . Greatpowersareable toindicate,

, i - ' (costly signaling), the strength of commitment

' .- : As developed in chapter 3 in a theoretical
. .. 3-. formal models, the investment of costs by a
. to a state’s commitment to its ally.

. '. states, both at home and abroad, the theory and
w 'tly incorporate the costs that states face in a crisis
u ,, . n on the costs that an alliance imposes upon its members.
' * 3' t such an alliance provides. I argue that, in alliances formed for
’ Win an inherent trade-off in the benefits that each party attains.

' j ., . x (1991), I argue that major powers in an alliance receive autonomy
“all powers gain in security. Autonomy, as described by Morrow, is the
We: mue its goals within the international system. Security, in contrast.

W of threats conveyed by an external enemy. By focusing on what

W alliances, those comprised of one major power and one or

W I argue that we can learn much about the deterrent effects of

We: rev. .

 

 

 

 

" .ufl

 

 

 

 

v
w--
r-"
u
. .. . ~
‘ _
o
I
~, ‘
-
a
‘-
' u
..
s.
u
4
-\
v
\
.
.
p
\ .
.~
.
b
h
,\

 

. 3M. Quantitative Analysis, and

   
  
 
 
 
 
 
 
  

.. above, I utilize two distinct methodological
. . » , . via alliance. [use game-theoretic modeling to
1' " “Which extended deterrence via alliance will succeed,
. - of crisis escalation. From the formal models
We propositions that are amenable to quantitative
' . at some measure of generalizability for my theory, and
, « T an artifact of the modeling enterprise. Finally, the use of
' v- throughout the dissertation permits abstract theory and
“to be grounded in historical reality.
W this dissertation is to improve our scientific understanding of
Wm. In order to make scientific progress, we need to be able
MW Why certain events will occur. The research strategy being used
“Whom the means of improving our strategy through the use of modeling
Maui“ and substantiate theory, and using statistical methods and historical
W the validity of the theory and model. Prior to delving into the
w pintisuof the dissertation it would be wise to heed the words of Bueno de

W935; 128), “T o the extent that logical consistency is accepted as an elemental

Wmh, formal, explicit theorizing takes intellectual, if not temporal

Wm While formal theorizing need not necessarily be

 

 

.
‘r

.l

v

-

u‘
0

1

 

 

.F . accused of being merely positivism with no

.mnmv .

  
  
 
  
  
  
 

' ~ the desire here is to improve both our additive and
abbot the impact of alliances on the provision of
-.t 9%, 1981) makes the distinction between these two forms
m ; that additive cumulation involves the use of new

.. : techniques, and a greater awareness in the
' ihataresults have been discovered. Integrative cumulation
m of new theory, or explanations of old theories in a new light.2
.. ddress both of these issues surrounding both types of cumulation.
new dataset especially designed to test when extended deterrence via
mm. In applying these data to statistical methods, I utilize some new
W, notably binary cross sectional time series techniques along with
awheh'que, a bivariate probit model, that allows me to control for the selection
Will “occurs in studies of deterrence.

film hmdsto integrative curnulation, it is the hope that this dissertation will allow
“Incubus to understand more clearly the influences that shape deterrence
m. Through the use of game-theoretic modeling I uncover some interesting
lllllthta the role that information plays in deterrence situations. However, if we are
battlinefihetlevel of knowledge of international conflict, we must take care not to

I' ll . . ' I' I1 ”caches. Inthis spiritI combine two different forms of analysis,

   
   

'yg cumulation corresponds loosely with Kuhn’s (1962) notion of
| ‘ » g, "1910) concept of progressive research programs. For good
‘ v 'rnrnternatronal politics, albeit a bit dated, see the selections in

"* ZimamdSinser(l98l)-

‘ - '_l ‘ _ _._ _ .
-“ \k‘vm‘i.’ ., '_ .' . uln.‘.4'.r‘r_‘1_'- .rL

 

 

 

n

 

   
 
  
 
  
  
  
  
  
  
  
 

usefulnessofthe theory. [argue

- -howledgeofextendeddeterrence via

'- h OUll‘r"? .'

t ricaIRigor
1“"

. 1,5 1' - t «as atheory of political events. I specify a theory in

ll': .

"me can be used as a metaphor, an analogy,

. , -. i pressures and international ambitions into a theory of
. ~ . In this vein, therefore, I use my game-theoretic model
- theory. The combination of theoretical thinking
I » .1 . »- modeling meets the above-mentioned criteria of Bueno de
“precede empirics. Proceeding in this fashion allows us to
‘ ' .- . y' ‘ form of cumulation for understanding why alliances often
= worried deterrence.
Waite heart of my theory suggests that states behave in a rational manner,
1% theretic modeling is the best approach to the questions being
Wdissertation. I first construct two game-theoretic models. The models
“Immortal structures, but both capture the strategic interaction among
humanistic, inwhich A is a common threat to both B and C, and B and C

”W ' ' acrisis as being a set of sequential moves, with the severity of

      

rthe hostilities escalate. Much of the inability of earlier empirical

., ‘ ' between alliance reliability and war, I argue, is that these
. ~ .. I- “Sills .

 

 

 

  
  
 
 
  
  
  
 
 
 

u , crisis. However, not all crises lead to war.
’ ‘ h oflen successful. To ignore the interactions
" '1 1b the outbreak of war is to ignore the sequential processes
A Mortal arena, be it among adversaries or among
. interactions that I feel that the role of alliances on
ibest understood.
. - 'c modeling has an additional advantage. It allows me to
. to one or more of the players. Since successful deterrence
~-- -' w r of an adversary, it is necessary to model issues of limited
‘« - and demonstrates the effects of uncertainty.3 This uncertainty
‘5’ ~‘. . ... : u when an aggressor is unaware of how committed a defending state
EMQMWHletmgible commitments can be observed, such as arms transfers
Whit military bases abroad, there are issues that are not so easily measured
Maghreb intmgibles is the value that each state places on the issue at stake in
mom through the use of these formal models that unique insights can be
What-cannot be uncovered by a cursory examination of the historical record
1.11 mm Analysis: The Search for General Patterns
or course, abstract theory alone cannot tell us whether alliances are successful at
Won or not. Abstract theory that is not grounded empirically is merely
W theory. Only an examination of a large number of cases can let us know

W. , ifthe timer-y can be falsified. Quantitative analysis helps us in our

 
 
 
 
  

‘ _ .,. information models in international relations' is a relatively new
I, ' , hawk, Bueno de Mesquita, Morrow and Zorick (I997);
‘ {.17 g .-. (1981) as representative pieces in this genre.

, .. .4
- _‘_"u‘ r-

 

4
c
1
¢
‘
n
-.

 

 

mw‘.aueansoftestingthe

  
  
   
 
   
  
  
  
  
 
  
  

Inducertainifthetheoryhasany
”Va": .
~ ’Bexhortation above notes, we should take care in

 

' ~ data analysis. Barefoot empiricism serves no

. m to be heaped upon those who endeavor to study
' . « :mner. But, in this light, the converse is true as well.
. ,u : models that are elegant theoretically, but have no
he, formal models need to be linked to the empirical world if
A ‘ i ' . . .- ., . . - and aid us in the scientific quest for understanding in

. _ quantitative analysis also provides additive cumulation as well. The

.‘l
q.-
l

i ‘ - w . dissertation represents new measures of alliance commitments, and

. I.

., . “logic between domestic and international goals. Additionally, new

  

".'..‘ s,

W for time-dependent variables that are binary in nature allow us to
m ' ’ ‘ ' .nephisticated methods to study issues surrounding deterrence.
WM“: The Quest for Detail '

v «to “many would argue that mere reliance upon abstract theory and cold
W ignores much of the nuance that comprises international

   
  

«r to an extent are correct, and are even warranted. If we are to
in a scientific manner, we need to make certain that we are aware

the context in which these interactions occur.
'. mien ,j

.....

 

' n. "' '\
u ' a .. . , . ’
. . y . .' ".

w steam

 

  
  
  
  
  
 
 
  
 
     

throughout the dissertation help provide
g emerges from the modeling enterprise.4 It allows
' . . particularly to those who neither understand nor
~ . Theory and statistics allow us to pursue rigor, and
‘ novel conclusions. However, we must never
‘ ' -- goals tend to be driven by events we have witnessed

‘ _. ii: is to these events that our endeavors must speak.

I ‘ I ofthis chapter I discuss the remaining sections of the

‘; " "address much of the literature here, but leave it for its relevant place

. a n

W chapter I outline the conceptualizations and definitions which
“My of extended deterrence via alliance. In particular I concentrate on the
mm of importance to my theory: deterrence and alliances. I draw a
“Won-between deterrence and defense, and reconceptualize the concept of an
‘ " ”alliances that form for deterrence purposes are between great powers and
mail: This forces us to clarify what states expect to gain when they do enter into

MW it is the thrust of this dissertation to demonstrate that great powers

   
 
 

; Rather than seeking to test my theory of extended deterrence via
‘ 3 . . 've case study approach, 1 am using individual cases to
._ -.. many of the theoretical assumptions and empirical findings.

Ii

 

.
_.'. '
t.
.u “
in
p...

 

 

 

    

 

   
  
  
  
  
  
  
  
  
 

Siam variety of reasons. However, the smaller
- m fiaomtherelationship.
,. ‘ Meal chapter is to distinguish between
~ dimes. I posit that deterrence alliances are a special
. - fled to be examined within the context of their
' v: . ' - . ~ have a deterrent nature, but continue once the threat
' ' ~ hadisappeared. The continued debate over NATO expansion
« v is just one case in which the original threat (the Soviet Union)
' ._ continues to exist. Hence, I denote a new type of

alliance, which has as its specific purpose the active attempt to

»» mm as minor power.

‘w‘m that I do acknowledge that many of these alliances serve dual
Winn functioning solely as a deterrence entity. Again, NATO provides a
W The purpose behind NATO’s inception was threefold: keep the United
“halved in European affairs; keep Germany under control; and deter Soviet
WW8 Western Europe (Osgood, 1962). Now that the Soviet threat has
Vim-NATO still persists. However, its overall deterrent role ended with the collapse
of tinSovietEmpire in 1991.

:- H» x' MI ugue that traditional means of examining alliance commitments are

Mininflyfing the linkages between alliances and deterrence. I argue that a

%

 

T r " J as a tangible item that is transferred between allies to
W's-he. for each other.

 

k

 

   

w. within the context of the rational

   
  
  
  
   
 
  
   
 

All status and applicability of rational models in
I ,. -:. «may ofdeterrence in particular. While many
7 , rationality are flawed, and have a tendency to

f '» the‘assumptions of a rational choice approach to

,' "l " ’ '- Midas us with a great amount of insight into the
A ‘

, . - -.. play.

.r:

I i“ l meptual underpinnings of deterrence and alliances, I

.‘q.

"9

. ' 'll ' v- W framework of extended deterrence via alliance. I place

  

h

ll 4,..- ‘ ‘ . ' . , - within the agent-structure debate within international relations,
mm: two levels of analysis are interrelated. l utilize the conceptual
We and willingness (Starr, 1978: see also Most and Starr, 1989;
MIMM 1991) to theoretically link macro structural attributes of the
W system and the micro-decisions that take place within individual states.
hmiatainthat deterrence is effective if a great power defender is able to transfer

Mm into its minor power protege to make its commitment credible. This
mm such as the provision of weaponry, the establishment of bases, or
“Min, to all costly endeavors that must be borne by the defender. Such
Who-viewed as akin to investments that individuals make in the market; one
W in order to make it.

William: context, I argue that providing such investments are costly

r"

' m; N. v.1 a." *1. - .‘A

 

4.2a}

 

 

 

 

 

  
  
  
  
  
  
  
  
  
  
  
 

a W15» the higher the level of

.. , 'Ndetbnder, the more likely extended general

‘ . “benefits that exist between alliance partners and
n r in a crisis: aggressor; target; and defender.

H fiat are derived from the alliance relationship, and
.9. <- . situation, I then demonstrate how costly signaling
. ‘ '-U=¢vmhlor power protege provides information about the

‘ mitment to its ally. I explore the impact of alliance

~ of extended general deterrence, and frame the discussion
. ‘ honomy and security. I argue that asymmetric deterrence

. . m .. security benefits to the smaller powers in the alliance, at the

. ‘ ' ' '- 'fllc value each state places on these benefits, and the costs that an
“W engenders, explain crisis initiation and escalation, even though an
‘ W ”WEI Me:- I present two formal models of crisis initiation and escalation. 1
Wow of the basic model, and I then discuss the sequence of moves within
Emmanuel, information is complete and perfect. This allows me to specify
WM~strategies available to each actor in a simplified environment. 1
WM underlie the model, and provide examples from nineteenth

mm diplomatic history to illustrate the outcomes of the model in a

 

 

 

 

  
  
  
 
  
  
  
  
  
  
  
 

.- - an aidescription of the costs and benefits that
" , , i "and derive the payoffs that each player holds for

. ‘ .. " _‘. 1 flow the derivation of payoffs, I discuss the types of
‘ . i a »- In regards to the aggressor, there exist three types
. " . . rain with the defender and protege. In regards to the

based upon the value it holds for its protége’. From this

.'- « ..~ the preference orderings that each player has for each
‘ " ~ that are imposed. Following that, I then proceed to solve

' ‘ - tlleresults.

. 4» fie model under conditions of complete and perfect information, I

. ., . .. of limited information. By making the information

'25- .u ~, A:

"t

« w, I am able to interject a greater measure of reality into the

mm the types of defender that were examined in the first model (strongly
W and weakly committed defender) with the different types of
w, md introduce the issues of perception and credibility. Credibility, I argue, is
why laminating deterrence behavior within the context of alliance bonds. The
mm tint the aggressor is uncertain about how committed the defender is to its
mmpotégé. I solve for the equilibria using the concept of Perfect Bayesian
Wand discuss the implications of the model. Notably, l discern that conflict

Win“ information conditions and conditions in which there is limited

I, . . .1, However. extendcd deterrence succeeds only when conditions that

 

 

 

-o,i

  
  
 
 
   
  
 
   
  
  
  
  
  
   
   

' -- then hachldown if it witnesses resistance or

I anti ‘,.'1-':
W , W

"' L " r ’ - know turn to an explication of the propositions

 

' , ~' n After outlining and discussing these propositions, I
~ i .. of asymmetric alliances from 1870-1984 that can
. I their central purpose. I identify a set of twenty

with a total of 302 alliance year observations.

data sources, notably the MID and ICE datasets, to determine
. - w ... general deterrence was attempted, when it failed, and .

l

‘ « - ' deterrence succeeded or failed. I code the outcome of each I

,. ' , » . . if my, thedefending state took. I then examine whether the crisis I

'.. , , F ‘ “not,“ what the final resolution of the crisis was. In this context I discuss

a' .
l -- " r" ' ,o. .hias 'mherent in such testing, and I provide a means of avoiding

“Mil...

mm the population of cases to be examined, I operationalize
“My investment signals (security benefits), autonomy benefits, balance of
Walton term and long term), domestic political costs, and value of the

“hindefm. Having completed this task, I turn to the empirical testing

  
 
    
   

‘ ‘ -- derived fiorn the model.

‘ 1' Ipmvide empirical tests of my formal models. I outline two

" _ ‘ gene!!! deterrence success/failure, and one for the possibility

 

 

 

 

 

l"

 

of“ immediate deterrence success/failure. In order to test my propositions, l
ulilinebivariate probit model (Brehm and Gates, 1997; Dubin and Rivers. 1989; Reed.
Mfltlflhelps capture the selection process that is inherent in the quantitative study of
deterrence. Additionally, I utilize binary cross sectional time series techniques (BCSTS).
indicated by Beck, Katz and Tucker (1998) to account for the issues of
admonition that often arise when pooled time-series data is used. I then run estimates
of my equations, and perform goodness of fit tests. To add validity to the analysis. I
return to the cases I have traced throughout the dissertation, and provide a detailed
discussion of the factors at work in these crises. The combination of statistical results
and historical case study helps to increase the internal and external validity of the theory
and formal models I proposed earlier. I close with a discussion of what the theoretical
and substantive results mean, and what alternative explanations could be at work.
Chapter 7

In the final chapter of this dissertation I tie together the findings of the formal
models and their empirical tests, and discuss what has been learned about the role of
deterrence alliances in the initiation and escalation of militarized international crises. I
relate the findings back to the earlier discussions on cumulation, and maintain that the
findings of the model, notably the role that certainty plays, has greatly increased our
understanding of how and why alliances function as successful agents of deterrence. I

close by pointing out directions for filrther research.

 

or CHAPTER2

 

  
 
  
  
   
 

AND CONCEPWALIZATIONS

' , ., formation of military alliances and their deterrent capacities
. .‘ my undies of alliances tend to emphasize their ability to
' . ' deter aggression, not all alliances perform these functions. While
. . dwintroduction to this dissertation, was formed with the explicit
‘Soviet aggression toward Western Europe, how does one explain an
(Withdraw typified by the Nazi-Soviet non-aggression pact of 1939? This
mm was to witness, was offensive in nature, and secretly provided for
W of Poland, which launched the Second World War.
3.? i .Stihzdiscrepancics are not novel observations on the part of the author. Ward
(maximum), and Gibler and Vasquez (1998) have all expressed dissatisfaction
Mmeetablished manners of identifying alliances and their specific functions.
Wilma dissertation it is maintained that if we are to understand the conditions
Warhol! alliances serve as successful instruments of deterrence, we need to clarify
Militia First of all, what is meant by deterrence in international politics? How does

fix “‘73 all J.

W? Secondly, what makes an alliance a deterrence alliance, rather
\Wmml Iii—t». i‘. . .
' body?

”‘0. in the 'm'ru -
Wet of dis chapterI develop the concept of a deterrence alliance. I
' -War. so: ‘ .t‘:
1“ wiofdeterrenee difi‘erentiate it from the concept of
.it so“ a '-
«‘4 harmoul politics. I move fiomthis
thin focus cl t an em
“7‘

 
  
     
   

 
 

 

 

 

 

 

I .
‘U‘u

 

 

phalantwo areas: the reason for alliance formation and the impact of alliances on the
“stand spread of war. These are the two areas that greatly emphasize the linkages
between the role that alliances play in international politics and their deterrent effects. 1
discussthe concept of a deterrence alliance within the context of extended deterrence.
and place this conceptualization squarely within the rational deterrence debate in the
literature.
2.2 Conceptualizing Deterrence
As was argued above, alliances often exist for purposes of deterrence. However.
what does this exactly mean? Is there a distinction between deterrence and defense? In
the following sections I discuss the conceptual underpinnings of deterrence, and
distinguish deterrence from defense.
2.2.1 A Typology of Deterrence
Deterrence has been one of the cornerstones of superpower security policies since
the end of the Second World War. However, it is not merely an artifact of the nuclear
age.5 In fact, as Mearsheimer (1983) discusses, it was a feature in both World Wars. as

well as the modern day Arab-Israeli conflicts.6 And Huth (1988) finds evidence of

 

5The literature on nuclear strategy and nuclear deterrence is voluminous. especially
given its centrality in superpower thinking after 1945. The focus in this dissertation is on
conventional deterrence, as defined by Mearsheimer (1983, 15) “is s a function of
denying an aggressor his battlefield objectives with conventional forces”. Conventional
forces are, in the simplest sense, non-nuclear forces. For further discussions on nuclear
deterrence and its role in superpower (particularly American) foreign policy after the
Second World War, see George and Smoke (1974); Huth (1990); Kahn (1965); Kugler
(1984); Morgan (1984); Powell (1985,1990);Schelling (1960,1966). While this list is
smely not comprehensive, it does provide a good overview of the nuclear deterrence
“termite. The conventional deterrence literature will be discussed in more detail in the
body of the text, given that it is the focus of this study.

6 See-also Shimhoni (1988) for a discussion of deterrence in the context of Israeli
may policy.

_ '- . i 19

  
  
  
  
  
  
  
  
  

"" W ~fiemportionofflienineteenthcentury,as

9: ~ ' “of the policy implications of this strategy,

» . -- camera. In this vein, therefore, I out1ine a basic

, ,' . 'W formulations of deterrence.

i i ‘ basic level refers to a policy of attempting to use threats to
i ' -.' political costs of using military force will be greater than

. t the military force (Huth, 1988, 15). Deterrence in this

m '- : to deny an adversary any advantage that may come

h of crisis. Thus, deterrence is used here as a denial strategy.7 A

, , ... .~ . and makes it clear to an adversary that it will respond in kind if

A, , ‘ -' 3W: »'

mm, in contrast, tends to be regarded as a punishment strategy.
mesued such a policy during the Cold War in regards to the Soviet
mm nuclear weapons that were able to withstand a Soviet first strike,
m M was able to convince the Soviets that any attempt to launch an attack on
WM would be met in kind, and the retaliation would be extreme, in order to
mm a lesson”. The focus here is on deterrence as a denial strategy, rather
than apoliey pursued to punish an aggressor extensively if a hostile action is
“mimosa strategy of denial, therefore, assumes two distinct types. The first

Wm Such a scenario implies that an adversarial relationship exists

 
  
  

. Wentheroleofdeterrenceasbomadenialanda
' SeedsoMearsheimer (1983, 15).

 

 

 

 

-u..

 

  
  
   
  
  
  
  
  
  
  
  
   
 
 

. inset least one state would consider resorting to

. .4 ' , Wmfld be willing to consider a resort to arms
‘ _ . ‘dilimsthreats of its own in order to prevent such a

.. 1988, 15-18; Morgan, 1984, 42-43). ~

" 3‘ , .. «an scenario that involves a pair ofstates is that of

' J 1 ’ . -:~ . a :a case three conditions must be met: 1) oflicials in at least
'. -— .. : an attack on the other; 2) the leaders of the second state

.. 33) the leaders in the second state issue threats to retaliate if the

. lath its intentions (Morgan, 1984, 38).

- however, if a third state is involved? In particular, what if a state

" I rn'~ with a third state and desires to protect it? This concept is referred

Meme. This simply implies to a situation in which one state threatens

J v
t-‘

s‘ , w ' m against another state to prevent the second state from using military
Molly of the first state (Huth, 1988, 16). Following from the earlier
Wanna] and immediate deterrence, one can have cases in which extended
Meme is profi‘ered, and cases in which extended general deterrence falters
WW immediate deterrence emerges.
Cum to understand the circumstances under which extended deterrence will be
WW need to be defined. The first is deterrence credibility, and the

g .: ,l . ‘ l 'l'ty. Both are important to understanding the influence that a

   
  
  

. . . ,- ,- r‘ . 7 by: on shaping the perceptions of an aggressor. Deterrence
I i A wand, refers to the ability of a defender to communicate their desire

,, .. Huh (1988, 4-10) discusses this problem within the

 

 

 

 

   
  
  
  
  
  
  
  
  
  
 

‘ I mdibifity is largely a function of the military
' ”willingness (resolve) to actually use them in a crisis

, hour costs demonstrate more resolve, and therefore

. ‘ l.
in“ >

G W involved in conflict to protect their “vital” interests.
- '-~ 3, states are able to demonstrate their credibility

' . ~ ‘ . This is accomplished by investing tangible benefits into

. . I“! on costly signaling shifts the discussion regarding credibility

« . ~~ upon capability ratios, and maintains that states can make

..m 3: other methods. This is not, however, meant to slight the role

iplays in international politics and in the success of deterrence in

I ’, Md above, deterrence, as a punishment strategy requires the ability to

4

_ theta defender has made. Military might is the means to such an end.

3"? ‘

l r T. ’l

.. i :« _.. stability, on the other hand, refers to the defender taking such actions
‘W attacker does not fear an offensive strike (Huth, 1988, 11). Most states
“attentive their conflicts peacefully. The purpose behind making a
“Wt stable is to convince potential attackers that the commitment is
m Witwill not be utilized in an offensive manner, nor will force be used
hkwwe before continuing. There exists a basic distinction between

“SW above, and compellence. Deterrence is an attempt to prevent an

   
  

n - before it begins. American troops stationed overseas in West

. 7 . -7 he Cold War, and American behavior in this region of the world

 

 

 

 

 

 

   
 
 
 
   
 
  
  
  
 
 
 

.' ._ ‘ '. m make an adversary stop doing something
' . ~ ”.mof action (Schelling 1966, 69-
- - mm of attack, the target must decide if it
. to force the aggressor to back down. If one
.1 .. taking place in the former Yugoslavia, one witnesses
. NATO demanded that Serb President Slobodan
' “ v w a x of ethnic Albanians in the Kosovo province of Serbia.
r of the demands made upon him by NATO. In an attempt to
‘ " “‘ ;_ ‘4. , NATO, under the direction of the United States, began an air war

9' "‘ . r“
. ' ' ' ‘ n attempt to halt, or compel, their persecution of the Albanian

'l‘l"'

W’o

 
 
  
 
 

gun in

ant pain in this dissertation to ensure that cases of extended general
at txt

(both cases that were attempted and those that failed) were actually cases of
III the in

m and not of compellence. More on the coding rules regarding cases of

of on ’-

deterrcnce can be found in chapter 5.
lulu in M, l

‘ Bled upon this discussion, a typology of deterrence can be constructed. This
«a an: lywd :-
W

borrows from the work of Ruth (1988, 17), and helps clarify the different
hi: I“ I'r-l

of actors and immediacy of threat that exist in deterrence situations. I then
“Mil . (his; 1 5
“in lubctmtive terms what each cell of the following table means. Figure 2.1
w in

MM diflbrent combinations of deterrence that exist, given an adversary, a
(Mi) 2 1‘
We.

 

 

 

 

 

1 Direct 2
General
Deterrence

 

9:

Extended 4
General
Deterrence

 

 

 

 

   

1 “In first cell above, we have a case of direct immediate deterrence. A good
Milt: ' . I

" QT _' 10‘ direct immediate deterrence can be found in the conflict involving China and

0'00! firm in... = .

”II 1962. India and China have both levied a variety of claims to disputed territory

*0:- zt? ;. ‘

mwbtl'ders. In 1962 China responded to Indian provocation in the region by making

”men‘- -‘

“of their own to prevent Indian acquisition of the disputed areas. China responded
luau worn: in .

 

Ilium of military action of its own if India continued to pursue a policy of
w But all; -

v , . mitorial acquisition. Through the use of these threats China attempted
NU _' in puma.

     
 
 

a , II it turned out) to deny India any advantages it may have incurred on the

 

_ an ~ 1
' 5 .~ ,Me-cn'naeonnicteanbefoundinwmnnu1975).

 

 

 

   

l ,
_ a
,.
'V‘
.o
.u
“h. gr.
1
r a .

 

 

 

 

 

   
  
  
  
  
  
  
  
  
   

= “of direct general deterrence, which is best
W8 one another during the Cold War. Once the
' - each obtained nuclear weapons, a stalemate occurred
-' .. -: to strike first at each in a nuclear capacity.
, v't 7. ~- of each other, and publicly claimed that the other
"fl. I‘m weapons led to peace, albeit uneasy at times, between the

F

J A it; ‘ ‘ we have a case of extended immediate deterrence, in a case
M century. In this case, the deterrence commitment is
‘talliance. Austria-Hungary was allied with Serbia, and promised aid
»;. rtbe case of an attack (real or threatened) upon Serbian territory. In

\i , ' H ‘ ,- ~ , x 'brselfatwar with Bulgaria, and soon found herself being threatened
r; w. . . .
a , , - . troops. Austria threatened to intervene militarily, and was able to prevent

Wis actions on the behalf of the Bulgarians against its Serbian protege.”
it magnum cell, we have extended general deterrence operating. It involves yet
WNW deterrence via alliance, and involves the case of American
W a presence in West Germany after 1961. In the aftermath of the
W War, the division of Germany signified the ideological divide the
mumps to dislodge the American presence from West Germany, and

, flame“ , the Soviet Union and the United States settled into a state of

m.» w- .

 
      
  
   
 
  

‘ a». - z. p ‘ g Hylflnd (1989) on the rivalry between the United States and the
a , my

_ ., . :(i 973); and Langer (1950) all present excellent depictions of
. "bf-1885, and the Austrian response.

 

 

 

 

  
   
  
  
  
  
  
  
  
  
  
 
 

V1
1“. .-
’ . Wmaintainedtroopsandothermilitarymeans

. ‘ i .11..» mightwas there to deter hostile actions from the
with... atthe general level.”
1 ~ ' ysis of deterrence in the academic literature has tended
. . : the strategic (nuclear) level and the conventional level.
' t , - mending deterrence in an alliance context, both of the levels
' ' .. .. nuclem' capabilities lie at the core of much of the post ~1945
.. :vcin, therefore, I discuss the findings that emerge from
extended deterrence in the following paragraphs. I focus on the
. . by 1 forces play, but also acknowledge that nuclear weapons also play a
. ~ distinction with a brief discussion of the role of alliances in extended

,L. and establish the groundwork for a discussion of alliances in world

WM distinction should be drawn, however, before proceeding further. What
khm “between deterrence and defense? These terms are often bandied about
Mutilation by academics and policy-makers alike, with little regard for the
WW the two. Defense, according to Snyder (1961 , 3), “means reducing
“Native com and risks in the event that deterrence fails”. The distinction,

W defense occurs after deterrence has failed. It is an attempt to prevent
it u.___ A, u

M r ‘plmes American relations with West Germany in a context of

. . .
. .( -' '
¥ ‘ |-‘ 1 fill“.

    
 
  
  

-... - to ignore the vast literature in deterrence theory that utilizes garne-
~ Many of these references will be addressed below, when our

. ‘ .. deterrence debate. Because most ofthe extended deterrence
'\ . ‘ fromthe quantitative study ofthese phenomena, this

 

 

 

any further damage, either through the use of compellence techniques or. in the nuclear
case, by using some form of punishment strategy. The focus of this dissertation on
deterrence, therefore, restricts our attention to cases in which potential threats are
confronted with counter-threats. '3 In the case that deterrence breaks down totally, and
conflict erupts, the focus would shift to defensive tactics. Although interesting, they are
beyond the scope of the work here.
2. 2. 3 Existing Findings Regarding Extended Deterrence

Quantitative studies of extended deterrence can largely be traced back to Russett
(1963), and his seminal treatment of the “calculus” of deterrence. '4 Such studies tend to
find themselves largely grounded in the framework of realism, with a focus on the role
that military capabilities play in deterring hostile actions. Additional realist variables that
matter include alliance ties, level of trade, and strategic importance of the protege to the
defender. Huth and Russett (1984, 1988) and Huth (1988) focus on crises at the level of
extended immediate deterrence, with specific focus on realist factors. '5

Studies in general deterrence are not as common as studies in extended immediate
deterrence. Much of this dearth of rigorous study can be attributed to the difficulty in
identifying cases of extended general deterrence. As the discussion of selection effects in

chapters 4 and 5 demonstrate, it is difficult to identify if deterrence was operating, or if

 

13 . . . . .
However, it should be noted that havrng a strong defenswe posture, especrally in
regards to capabilities, plays a great role in determining extended deterrence success.

'4 Of course, many have noted that the article could more properly be called the “algebra”
of deterrence.

'5 Reviews of the empirical literature on extended deterrence can be found in Levy (1988,

1989). See also the edited volume by Stern et. a1. (1989) for a number of perspectives on
deterrence, both nuclear and conventional.

27

some other factor was an issue in preventing an outbreak of hostilities. For example, why
do we witness such a long period of peace between Bulgaria and Romania in the latter
part of the nineteenth century? We know that Romania is allied with Austria-Hungary

for the explicit purpose of deterring Bulgarian designs for a greater role in the Balkans.
This presumes that Bulgaria has hostile intentions regarding the Romanian state;

however, no hostility, with the minor exception of the Second Balkan War, erupts

between the two minor powers during the almost thirty-five years that the Austro-
Romanian alliance exists. Is the threat of Austrian intervention enough to deter the
Bulgarians, or are they in truth satisfied with the status quo situation during this time
period? Such issues plague studies of general deterrence, either direct or extended.

Most of the empirical studies regarding deterrence have focused on the immediate
level. Much of this is due to the fact that cases of immediate deterrence are easier to
identify. Two major exceptions to the paucity of studies in general deterrence are Huth
and Russett (1993) and Sorokin (1994). Both studies examine general deterrence within
the context of enduring rivalries. In the case of Huth and Russett, the study is conducted
regarding direct general deterrence, and does not address issues of strategic interaction.
Sorokin uses a game-theoretic interpretation to demonstrate that a defender will intervene
when it is in the defender’s interest, and discusses the role that alliance formation plays in
enforcing stable deterrence situations in the Middle East.

Since extended deterrence, as evidenced from the above discussion, is meant to
project a defender’s power onto a minor power, it seems natural that alliances would be
the means that great powers would use to extend their military capabilities for the

protection of their protége'. However, in the 58 cases of extended immediate deterrence

——_

28

that Huth (1988) identifies during the period 1885-1984, only 24 of the cases involve an
alliance. Hence, there seems to be something that distinguishes extended deterrence
relationships that function via alliance mechanisms from those that do not formally
cement their relationship with an alliance treaty. In this vein, therefore, I turn the
discussion from one of extended deterrence to one regarding the role of alliances in
international politics.

What is of importance here is that in many instances, states undertake specific
deterrence commitments to protect protégés, and in some of these cases, the commitment
comes in the explicit form of a security alliance. This area of inquiry, unfortunately, is
understudied in the deterrence literature, and evidence regarding the effectiveness of
deterrence via alliance is mixed at best.

For example, the impact of alliances on extended immediate deterrence have been
studied to some extent, albeit not great. Within this context, Huth and Russett (1984,
1988), Ruth (1988) investigate the impact that a defender has on 1) deterring aggression
against a protégé and 2) what ensures that the defender will intervene if deterrence fails.
They find that the existence of a formal alliance between the protégé and defender is one
of the best predictors that the defender will intervene on its protégé’s behalf. However,
as Huth (1988) notes, the existence of an alliance has a negative, statistically insignificant
relationship with the prospect of deterrence success. An alliance relationship, therefore,
appears to increase the possibility of intervention in a conflict if deterrence breaks down,

but does not necessarily prevent a demand from being made in the first place'6

l6Others, notably Morrow (1994a) and Smith (1995), examine the deterrence impact of
alliances in their formal models of alliance behavior, but their focus is on alliances more
generally, rather than on deterrence per se.

29

 

.
u. -
."f
:..
I
..

 

 

»
II.
.
x
\
‘K
.
.
-.
.N
-
7

 

 

 

Alternatively, in a formal exposition and empirical corroboration, Sorokin (1994)
demonstrates that alliances can and do play a deterrent role, and that they can be effective
at ensuring stability among adversarial states. These tenuous findings suggest that
further study regarding the linkages between alliances and deterrence ‘is warranted. Since
alliances play such a major role in the discourse surrounding international affairs, I now
turn to a discussion of the alliance literature, and the implications an alliance has for

providing mechanisms of extended deterrence between great and minor powers.
2.3 Alliances in World Politics
In this section I discuss the role that alliances play in relations among states. In
particular] address the type of alliance that is most readily classified as a deterrent
structure; namely, the asymmetric alliance. I review two strands of international
relations literature that have implications for extended deterrence via alliance: alliance
formation and the alliance and war linkage. This allows me to present my
conceptualization of deterrence alliances as a distinctive class of alliance, and link
deterrence policies more tightly to a number of military alliances throughout the years.
2.3.] The Role of Alliances in International Politics: Issues of C onceptualization and
Definition
That alliances are central actors in the conduct of international relations is neither
a novel nor unique assertion. Indeed, the literature is replete with discussions about the
purposes and actions of such organizations, particularly as elucidated in the prominent
realist and neorealist paradigms (Morgenthau, 1985; Walt, 1987; Waltz, 1979).
Since the focus of this work is on an existing alliance’s impact on the occurrence

of conflict, it is imperative that we understand what an alliance is, and is not. There

30

exists some confusion in the literature about the exact nature of alliances, so definitional
clarity is an absolute necessity. Ward (1982, 5-10) makes a good first cut in

distinguishing between alliances, alignment and coalitions. Alignment differs from
alliance in that it is not signified by formal treaties, and is not focused only on a military
dimension. Coalitions tend to be short-term alliances, established in the face of a specific
threat. Once the threat passes, the coalition ceases to exist.

More recent work by Walt provides an expanded definition of alliances.

According to his study, "an alliance is a formal or informal arrangement for security
cooperation between two or more sovereign states" (Walt, 1987, 12). Unfortunately, this
classification blurs the distinction between alliances and alignments. To prevent such a
problem, the definition of alliance used in this paper is one adopted from Kegley and
Raymond (1990, 52). An alliance is a formal agreement between sovereign states for the
putative purpose of coordinating their behavior in the event of specified contingencies of
a military nature. Although it was mentioned above that alliances oftentimes form for
reasons other than security, a majority of alliances also entail some security component
within their agreement. The existence of a formalized treaty is what makes the

distinction between an alliance and other forms of security cooperation.

Within the empirical literature, alliances tend to be categorized into three distinct
categories, depending upon the strength of their commitment. The point of departure for
discussions of alliance commitment is the pioneering work of Singer and Small (1966a,
1966b, 1968). In these studies, they categorized alliances as defense pacts, non-
aggression pacts, or ententes. In descending order of commitment level are defense

pacts, which require a signatory to come to the aid of a beleaguered ally; non-aggression

31

pacts, which stipulate that the signatories will not assist a third party who may attack any
member of the alliance; and ententes, which merely state that the nations involved will
consult each other about possible coordinated action if any of them are attacked by an
external party.

While this classification scheme for alliances provides some distinction between
the different levels of commitment that exist, it is problematic in that it does not account
for why alliances form in the first place. The following section addresses this central
issue in the alliance literature.

2.3.2 Why Alliances Form

Traditional realist (and neorealist) interpretations of alliance formation View them
as forming in order to balance against power inequalities in the international system. In
more recent work, Walt (1987, chapter 5) has reformulated balance of power theory
somewhat. According to his view, states ally in order to counter a threat to their
sovereignty (this is known as balance of threat theory). Instead of forming to balance
power capabilities, he argues that states ally in order to counter states which they view as
threats. Threat, in Walt’s conceptualization, is represented by four factors: aggregate
power, geographic proximity, offensive power, and aggressive intentions (Walt. 198 7,
21-26).

In a world characterized by an anarchic international system, states are
responsible for their own survival. Thus, as the discussion above makes clear, security is
an important component of why alliances form. However, is this the only reason? Why
would a state such as the United States, with an overwhelming abundance of power

capabilities, ally with a smaller state with limited capabilities? Such an arrangement

32

would, in traditional views of alliances, only serve to entangle the United States, without
providing any extra security for the more powerful state.

The existence of alliances between unequal powers is puzzling to students of
international politics, since minor powers have very little to offer to major powers in
regards to security. However, such alliances persist and continue to be formed. For
example, Midlarksy’s (1988, 1989) work on internal alliance structure determined that
these alliances between unequal powers are more likely to endure. According to his
studies, two forms of alliance structures predominate: hierarchical and non-hierarchical.
According to his definitions, hierarchical alliances are those that have one major power
and at least one other smaller power (Midlarsky, 1989, 56). A good example would be
the United States’ alliance relationship with South Korea from 1954 onwards. In
contrast, a nonhierarchical alliance is one composed of two or members possessing
relatively equal strength. The Triple Alliance of the latter portion joining of the
nineteenth century, Austria-Hungary, Germany and Italy is an archetypal type of this
alliance structure.

As an attempt to explain why such alliances exist, I build upon the pioneering
work of Morrow. In a hierarchical alliance, the power is distributed unequally, and
corresponds to what Morrow (1991, 914) terms an asymmetric alliance. In such a
structure, the major power is generally assumed to be able to set the agenda for the
alliance as a whole. In a non-hierarchical alliance (or symmetric one, to use Morrow's
distinction), power is distributed relatively equally among all the members. It stands to
reason from this that if power is equally dispersed, then all members will have an equal

share in making alliance decisions.

33

.—
, u

a

he! ..

— .—

n- '*
. >

g.”

.‘ ‘.

to -

 

- -§
u» . .
'-
‘ v
0
~.
."
1
u
h
1‘.
.§
.x .
,‘ ‘
r
. .
.,
'b

 

\
's

 

As Altfeld (1984) and Morrow (1991, 1993) describe it, a trade-off exists between
forming an alliance and building up arms as a means of ameliorating security concerns of
a state. However, these are not the only concerns that a state has when it determines its
alliance policies. Larger states do not necessarily acquire security benefits from smaller
states. Smaller allies receive benefits in the form of increased security (through the
policy of extended deterrence), while the larger ally receives autonomy benefits, in terms
of military bases, or increased advantages in terms of trade.

If one looks at alliances from this angle, then one can surmise that they form for
security purposes, as Lalman and Newman (1991) argue, or for benefits of trade (Gowa,
1994).l7 Either way, I conceptualize states as forming alliances for the purpose of
advancing national policy goals, which may be ideological, security related, or related to
trade.

When nations enter into an alliance, they tend to sacrifice either autonomy or
security. According to Morrow (1991, 908-9), autonomy is the degree to which a nation
pursues desired changes in the status quo. In contrast, security is its ability to maintain
the current resolution of the issues that it wishes to preserve. A dominant power within
an alliance has the ability to constrain both the autonomy and security of its other
partners. The question that remains is the extent to which it will act in such a manner.

Of course, there are inherent dangers to both members of such an alliance treaty.
Minor powers, as Christensen and Snyder (1990), and Snyder (1984), fear being
abandoned by their major power allies in times of crisis. Such fears are borne out in the

findings of Saborsky (1980), who notes that 76% of all alliances are unreliable in times of

34

..
\ “
.
. 'v-
I
a
v
_ :-

 

.--.»

 

 

 

 

 

 

 

war. On the other hand. major powers can be fearful that their minor power partners.
reveling in the knowledge that their alliance partner will intervene on their behalf if
threatened, may undertake precarious foreign policy ventures. This “moral hazard"
(Snyder, 1984), is often attributed to the onset of war. and the major power defender will
be dragged into conflict to protect its interests in its protege.

In determining how to act, an alliance member must determine if it is more
important to have greater security, or greater autonomy. In this sense. it develops an
ideal point. This is defined as its preferred resolution of various issues (Morrow. 1991.
908). Rational actors will seek out an ideal point that will maximize the benefits of both
security and autonomy. If a situation arises which disrupts the actor's equilibrium. it
must determine its best strategy in order to resolve the crisis.

Such an approach is not an uncommon one in the alliance literature. Many
additional studies address the relationship between small powers and great ones in
alliance relationships. Most of these. in fact. tend to shift the focus from "balance of
power" or “balance of threat” issues to ones in which the great power defender trades
some security benefit for something other than capability aggregation. Over thirty years
ago Rothstein ( I 968) noted those small powers are not merely great powers writ small.
Rather, they play a unique role in the international system, and should be analyzed

differently than their major power counterparts.18 More recently. Morgan and Palmer

 

17 Gowa (1994) and Mansfield and Bronson (1997) argue that alliances that form for
trade purposes tend to provide security extemalities, which are unintended security
benefits of the trading relationship.

is Similar arguments are put forth by Keohane (1971). in his discussion of the large
influence of small allies.

 

 

s
o
kw
.
's
.~
‘.
K

 

 

(1997) utilize the terminology preservation and proaction to demonstrate the trade-offs
that exist between unequal allies.

In a more recent research project, Reiter (1994, 1996) discusses learning among
small powers. He finds that they tend to acquire knowledge from formative events (the
World Wars in his study), and apply this knowledge to their decisions on when to ally
and with whom. These studies, while not exhaustive, are indicative of the fact that
various benefits are traded between small powers and their defenders. As I shall
demonstrate in chapter 3 when a full theory is developed, small powers play a distinct
role in the deterrence alliance, largely due to the fact that they provide their great power
partners with benefits that differ from security.

2.3.3 Alliances and War: Random Chance or Reliability?

Secondly I turn to an examination in which the evidence is mixed. Specifically,
what is the relationship between alliances and the occurrence of international war? If
alliances are supposed to be organizations which are formed to deter aggression against
one or member states, then the incidence of conflict between nations is quite puzzling. ‘9
Why form alliances if they are ineffective in deterring attackers? An interesting quote
concerning alliances and war is indicative of the puzzle surrounding this question:

First, alliances do not prevent war or promote peace;
instead they are associated with war, although they
are probably not a cause of war. Second, the major
consequence of alliances is to expand war once it has

started; in the war alliances are important in accounting
for the magnitude and severity of war. (Vasquez, 1987, 119)

—__

'9 It should be noted that while many (but not all) alliances are formed with some
deterrent component in mind (N ATO in the post World War 11 era is the best example),
not all deterrence relationships are alliance relationships (the United States and Israel
being a case in point).

36

A majority of the work that examines the relationship between alliances and the
occurrence of war emanates out of the Correlates of War (COW) project. It is their
findings that Vasquez is discussing above. It is in an early work conducted by members
of the project (Singer and Small, 1966), that the relationship between alliance
membership and the occurrence of war is first empirically examined. Their initial
findings suggested that alliances are particularly correlated with certain factors pertaining
to war, including number of wars a state is involved in and the number of years of war a
state is involved in.

After this initial study, a number of other writings sought to clarify this
relationship. In particular, much of the focus changed to an examination of alliance
reliability; namely, under what conditions would states actually honor the pledges they
had made to other states. However, different studies produce dramatically different
results, depending upon how they conceptualize and treat alliance reliability.20 For
example, Holsti, Hopmann and Sullivan (1973), Singer and Small (1966a, 966b, 1968)
found that alliances tend to be honored, especially if their casus foederis is invoked. In
contrast, Sabrosky (1980) finds that 73 percent of all alliances are unreliable, and Bueno
de Mesquita (1981) finds that allies are as likely to attack each other as they are to come
to each other's aid. In addition, Siverson and King (1979, 1980) examined the effects
that an existing alliance had on the expansion of war. They find that a state is five times

more likely to become involved in a war if it is allied with one of the disputants than if

20 Reliability is never explicitly defined in either of these works, but it is inherently linked
with the notion of commitment, For purposes being outlined in this dissertation, I use the
terms honor commitment and reliability interchangeably. I define reliability as a

situation in which a state honors a pledge to an ally. This is similar to Smith’s (1996, 17)
definition of alliance reliability.

37

the war involves states with which it has no alliance ties. And, within a context of
extended immediate deterrence, Huth (1988, 1994, 1998) discems that defenders who are
allied with beleaguered states are more likely to militarily intervene when a crisis erupts.

These contradictory findings are quite disconcerting when it Comes to determining
when an ally will honor its commitments and the causal link between alliances and war is
never fully established. I argue that much of the problem surrounding this question is
that the reason for alliance formation has never been adequately addressed in these
studies. If we are to understand how and why alliances have an impact on the occurrence
of war in some way, then we need to theoretically specify why they exist in the first
place. I argue that it is in the realm of deterrence that we can understand, at least
partially, why some alliances exist, and what benefits they provide to their member states.
It is now to an examination of deterrence via alliance that this discussion now turns.
2.4 Linking Alliances to Extended Deterrence: Conceptualizing Deterrence
Alliances

Having seen that great powers pursue alliances with minor powers for purposes
other than capability aggregation, I think it imperative to move beyond some of the
earlier conceptualizations of alliances that have driven empirical research in the past. It
appears that many asymmetric alliances form for reasons other than balancing power, as
many studies, recent and otherwise, have demonstrated. Gibler (1996), for example,
defines certain alliances as territorial settlement alliances, while Gibler and Vasquez
(1998) focus on the determinants of what types of alliances will lead to war. And

Schroeder (1976) argues that alliances serve as mechanisms for managing the internal

38

politics of member states. From these attempts to reconceptualize the alliance variable
then, I posit a new type of alliance: the deterrence alliance.

I define a deterrence alliance along three dimensions. The first dimension
stipulates that the alliance must be formed for (potentially) some defensive purpose,
should deterrence fail. Therefore, the treaty that is signed must make some stipulation for
military support in the event that deterrence fails. This rules out alliances such as the
Nazi—Soviet non-aggression pact and the alliances formed in 1866 by Prussia with a
number of the smaller German states. All of these alliances were formed to aggregate
capabilities in order to wage offensive war.

Secondly, the identification of a threatening state must be evident. This helps
distinguish a deterrence alliance from one that is purely defensive in nature. An alliance
can form with purely defensive intentions, such as the American relationship with
Australia and New Zealand after the Second World War (the ANZUS alliance).
However, there is not a state that is actively threatening the vital interests or national
security of these smaller states. In contrast, we see American relations with Taiwan and
South Korea being cemented by formal alliances after hostilities involving both of these
smaller states erupts in the post 1945 era:

Third, the great power defender must be able to make its commitment to its ally
known. “Commitment is the adhesive of human interaction. . .Of great importance for the
stability of the international system are the processes by which alliance commitments are
terminated” (Kegley and Raymond, 1990, 1). As indicated by this statement, the failure
to honor a pledge to an ally can lead to a restructuring of the international order, as

nations break their old alliance bonds and forge others anew. Unfortunately, the concept

39

 

 

b

 

 

 

 

r... -.
..
‘. .
.,
l
~.
>1. ‘
.
-.
V I
..,.
‘5
~‘-
~c
' h
'h
' r
b
«.1
.
a. ,
5"
u.
7 I
u
‘-
.
\
.

 

of alliance commitment is not overly clear. In order to comprehend why states in an
alliance will adhere to their promises or break them; one needs to know precisely what
these commitments are. Hence, the decision to proffer security within a context of
extended general deterrence indicates that deterrence is a benefit that can be provided by
the defender, in exchange for some other good.

Other than the work emanating out of the COW project, very little research has
been conducted that discusses the honoring of alliance commitments in any rigorous
fashion. Two more promising avenues of explaining commitment among allies have
been witnessed in recent years, but they are a distinct minority. A first branch has
focused on members honoring their commitments based upon tangible resources
promised to the alliance's other members from a public goods perspective (the basic point
of departure is Olson and Zeckhauser, 1966). The second strand emerges specifically
from the literature on deterrence, as exemplified in Huth ( I 988); Huth and Russett
(1988); and Huth (1994).

Recent work by Gates and Terasawa (1992, 104) has extended the discussion of
the public goods problem among defensive alliances. They make a distinction between
resources (viewed as troops) fully committed to an alliance (which are public benefits),
partially committed to an alliance (which are both public and private benefits), and those
Which are used for out of alliance purposes (strictly private benefits). While this is not the

place for a discussion of the entire public goods literature, one thing emanates from this

40

 

 

'. '~ .I'
t- I
to}.
s. ‘

n“'

 

..--~--
.- - ..-
v. --
a. .,r
H“. .,
A.--I..
.
1‘ n
v-
. _ .
I... 1
-V k
x
< _ ..
t.
..._
v ..
’t‘
.
.
~‘l
n
1-
.
in
f n
\
‘9
. .
I

 

u
'\
n
. V.
n --
H
, y
‘.
'v
..
a
a
‘t.
I
\
‘\
u
a.
.2.
'\
'--.
'v .
‘y
a.
“ c
,. . _
\
u
'-

 

 

 

 

 

 

' work. Commitments can be measured as tangible resources, rather than as vague
promises or alliance configurations.21

Operating from a perspective on deterrence, Huth (1994) discusses instances in
which states will assume extended deterrence commitments. He finds, in keeping within
the realist fiamework, that states will undertake such commitments when the benefits to
their national security are served. The notion of commitment implicit in such a
framework is that of a promise to aid an ally. This keeps the focus within our discussion
of defense pacts, and extends the notion of commitment as resources.

Based upon the above discussion of commitment, I feel that it is necessary to
move beyond the notion under which Sabrosky (1980) operated in his study of alliance
reliability. Commitment implies constraints on the actions of an individual state, as
witnessed by the discussion of security versus autonomy above. We therefore need to
focus our attention on individual states, rather than on the alliance as a whole. Thus, I
conceptualize Commitment here as a promise, backed up by resources, to the other
members of the alliance. As will be outlined in the following chapter, the great power

must provide some tangible security benefit to its prote'gé in an attempt to prevent

 

2' Olson and Zeckhauser’s (1966) article on deterrence as a non-excludable good among
allies generated a spate of work in the realm of the economic theory of alliances. Sandler
( 1993) provides an excellent overview of this literature.

Palmer (1992) extends the purely theoretical work done in this tradition by
statistically testing hypotheses derived from the original theory of Olson and Zeckhauser.
Coneybeare and Sandler (199]) apply their public goods model to the Triple Alliance and
Triple Entente of the late nineteenth and early twentieth centuries. Much of this
literature, however, is rooted in the discussion of burden sharing within NATO.

Although an interesting literature, I maintain that deterrence benefits can be withheld by a

great power in an alliance. Unfortunately such an analysis is beyond the purview of this
dissertation.

4]

aggression on the part of a hostile adversary. These commitments are revealed through
the use of costly signals sent by the defender to the protégé.
2.5 The Rational Deterrence Debate

Having provided linkages between a policy of extended deterrence and alliances.
it would be appropriate to outline a complete theory of extended deterrence via alliance at
this juncture. However, there does exist a debate, both within the international relations
community and the political science discipline as a whole, which must be addressed prior
to continuing on this quest towards theory.

This debate is the one that surrounds the use of rational choice methodologies in
analyzing political phenomenon. This debate is currently quite central in the discipline,
and is exceptionally critical of the work that is done within international politics. In
regards to analyses conducted of deterrence, it is especially critical of what constitutes a
deterrence encounter, and how they are measured. In many cases these critics call into
question the efficacy of deterrence as both a social-science theory and a policy
prescription for statesmen to adopt. In the following two sections 1 present the criticisms
of rational deterrence theory, at both the formal and the quantitative levels, and discuss
why these criticisms are flawed.22
2.5.1 Criticisms based on Formal Modeling

Those that challenge rational choice models tend to fall into two distinct factions.
The first, as exemplified by the writings of Green and Shapiro (1994), maintains that

rational theories of politics have led to no interesting conclusions. Such critics argue that

22 The recent review by Harvey (1998) reexamines the rational deterrence debate within
the broader methodological concerns of necessity and sufficiency. Many of Harvey’s
cnthues of critics of rational deterrence theory are echoed below.

42

, .pvr

I '

 

 

 

 

rational choice approaches are dependent upon arcane jargon, understood by few, and
that empirical tests of rational choice models are poorly conducted (if conducted at all).23
Others, notably Lebow and Stein (1989), Lowi (1992) and Walt (1991) have decried the
use of rational choice modeling techniques, both in political science as a discipline and in
international politics in particular. As most of them argue, the political world is too rich.
too complex to be distilled into a mere mathematical formula.24

The second group of critics approaches rational choice modeling from a
psychological perspective, and argues that individuals do not necessarily attempt to
maximize gains. Rather, in many instances they attempt to minimize losses.25 Such
adherents to this method of explanation tend to assert the inadequacy of rational choice in
determining actors’ preferences.

While I could attempt to refute all of these charges levied at the rational choice
enterprise, I allow more able hands (Lalman et. al., 1993; O’Neill, 1995) to defend the

virtues of rational choice modeling at all levels of the discipline. In the following

 

23 O’Neill (1995) presents a thoughtful review and critique of Green and Shapiro’s work.
In it he takes special issue with how Green and Shapiro advocate testing of formal
rational choice models, and the phiIOSOphy of science underpinnings that they use to
strengthen their claims. See also the edited volume by Friedman (1996) for additional
responses to Green and Shapiro’s commentary.

2’ For an early defense of rational models that addresses many ofthese criticisms, see
Moe (1979).

25 Critics that exemplify the psychological approach to deterrence more generally include
Lebow (1981, 1987), Lebow and Stein (1989, 1990), Lebow, Stein and Jervis (1985).
Levy (1997) presents a more thoughtful and constructive approach to using psychological
approaches to studying international politics generally.

43

 

 

 

i
'l’. l .

.-
. _,.
0-._
' ..
O .

.

 

 

 

 

 

paragraphs 1 address these charges as they pertain to the modeling of international
politics in general, and the modeling of deterrence in particular.26

I begin with the charges that rational choice models are arcane, jargon-laden and
merely skeletal representations of actual events. The rationale behind using mathematical
techniques for modeling is to ensure that results can be replicated, and that all
assumptions are laid bare and easily identified. The assumption of rationality is just that,
an assumption. In many instances assuming rationality may be appropriate. In others, it
may not.

The formal rational choice literature, contrary to many popular conceptions, has
advanced our understanding of international politics in general, and has additionally
influenced our understanding of deterrence processes. Bueno de Mesquita (1981), for
example, utilizes a straightforward expected utility framework to explain international
conflict on a general level. The early work of Schelling (1960, 1966) is largely based
upon rational choice assumptions, and can be said to serve as the basis for much of
American strategic nuclear policy during the Cold War.

Additional work by F earon (1994a, 1994b), Kilgour and Zagare (1991), Powell
(1987, 1990), and Zagare (1987, 1990), to mention just a few, has taken the study of
deterrence, formalized it, and provided us with insights into how states behave under
different conditions. Keeping in mind that these are just a few examples from a vast

literature that applies game-theoretic rational models to the study of deterrence, I think it

2’ Nicholson (1992) also does this, and in a much more eloquent style. Suffice it to say
my defenses are not wholly original, and I do stand on the shoulders of those who have
gone before me in defense of rational deterrence theory.

44

premature to assert that rational choice has contributed little substantively to the study of
international politics and deterrence.

Psychological critiques of rational choice models, on the other hand, tend to focus
on the specification of the utility function. Specifically, many of these criticisms (Lebow
and Stein, 1989) maintain that leaders do not all act as risk-acceptant utility maximizers.
Rather, they argue that leaders have a tendency towards risk aversion, at least when it
comes to terms of losses. Additionally, psychological opponents to rational choice
theorizing argue that controlled experiments belie the assumptions that rationality makes.

The charge that rationality must assume risk-acceptance is a quite simple one to
counter. While many models do assume risk-acceptance on the part of the actors, there is
no hard and fast rule that states that this assumption must be an integral component of
rational models. One can assume differing attitudes towards risk in constructing rational
models; as long as the assumptions are made explicit this is not problematic.27 One can
subsume a variety of behaviors under the rubric of rational thinking. All that is necessary
is that preferences for actors be clearly specified, and that these preferences be sensible.

To demonstrate this point, I point to two contributions to the study of rationality
that are outside of the international politics field. The work of the Nobel laureate
Amatrya Sen (1970), for example, demonstrates that social welfare can be advanced in a
rational manner, even if it does not necessarily maximize profits for those providing this
benefit. And Monroe (1994) demonstrates in her analysis of German behavior during the
Holocaust that altruism can be a rational response to events. In both instances rationality

M

27 This argument is made by Kahneman and Tversk)’ (1979), and Levy (1 997)’ in regards
to prospect theory.

45

 

prevails, but the preferences are not necessarily risk-acceptant, or strict maximization of
benefits.

The charge that rational choice models perform poorly in controlled laboratory
experiments is a bit more serious. Levy (1997), in his discussion of prospect theory in
international relations, notes that this is one of the shortcomings of the rational choice
approach.28 Alternatively, Zinnes and Muncaster (I997) determine that in a simulation
study, a rational expected utility model performs better as a predictor of behavior than
does a psychological model based on prospect theory.

Once again, the evidence is mixed on which approach performs better. The moral
of the story can best be expressed by stating that modeling is admittedly an attempt to
distill a set of interactions in political life down to its simplest components. These
models help us understand a variety of dynamics that occur in political life, albeit at the
expense of the richness found in much work that is done on a case by case basis. As long
as we properly specify our theories, clearly articulate our theoretical assumptions, and use
our models to illuminate evidence, then the use of rational choice modeling techniques is
no worse than alternative approaches.

2.5.2 Criticisms based on Quantitative Studies

Moving from the abstractions of theory and model building, I now turn my
attention to the portion of the rational deterrence debate that is concerned with the
empirical identification and testing of deterrence theory. The concerns advanced by

those who critique rational deterrence theory from an empirical standpoint also tend to

 

28 O’Neill (1995, 735-6) addresses some of the issues that plague psychological
CXperimental studies of rational choice in the laboratory.

46

 

 

 

 

t...

v.'~

 

focus on two distinct areas. The first camp, typified by the writings of Lebow and Stein
(1989, 1990), calls into question how cases of deterrence are initially identified.

In a series of exchanges with Huth and Russett,29 Lebow and Stein (1989, 1990)
call into question the coding rules used by the first pair of authors, and the overall
relevance of the rational deterrence model in the social sciences. Much of their criticism
focuses on the fact that psychological pathologies of leaders cannot be accounted for by
rational deterrence theory. This was noted above, so our attention turns to their other
major criticism of rational deterrence theory; namely, that deterrence doesn’t work, and
that deterrence encounters have tended to be conflated with attempts at compellence.

The majority of this criticism is aimed at the level of extended immediate
deterrence. The concern, as elucidated by Lebow and Stein (1990, 481) is that both pairs
of authors agree on a basic definition of extended immediate deterrence. The
disagreement is in how the definition is to be applied. Huth and Russett rely largely upon
behavioral indicators. They examine instances in which threats were verbally
communicated from an aggressor towards a target, when troops were mobilized, when
counter-threats were issued on the part of the defender. Lebow and Stein, keeping with
their psychological orientation, maintain that in order to understand and identify
deterrence encounters, one must look at the psychological pathologies of the leaders
involved. By doing so, they dismiss 41 of the 51 cases that Huth and Russett code as
being extended immediate deterrence encounters.

Much of the issue being debated here surrounds intentions. Under what

conditions do we witness challenges to proteges, and how do we ascertain with clarity

29 When I refer to the Huth and Russett studies, I am focusing on the following works:
Huth and Russett (1984, 1988); Huth (1988).

47

 

v‘ ,.‘

 

 

   

.
\
. n.
\ D-
‘I
\
.
‘H
‘.
\ 'K i
\X
I
\ x

 

that a challenge was truly a threat? Lebow and Stein tend to be vague on their codings.
remarking that they relied upon a vast array of historical evidence, and expert advice
from distinguished diplomatic historians. Huth and Russett (1990) respond to these
charges by clearly explicating their coding decisions, and noting that attempting to
determine actual intentions is plagued with methodological difficulties, and that
attempting to ascertain how serious an attacker was about using force is not consistent
with good social science (Huth and Russett, 1990, 482).

The second group, exemplified by the writings of Alexander George and his
colleagues (George, 1979; George and Smoke, 1974; 1989; see also Lebow, 1981), calls
into question the use of statistical methodologies to study cases of deterrence. These
arguments assert that statistical models that deal in large numbers of cases tend to gloss
over the inherent richness of detail that exist within individual crises.

As part of a larger forum on the study of rational deterrence, Achen and Snidal
(1989, 143-4) point out that the case study literature has provided tremendous insight into
the dynamics of various historical cases of deterrence. Such studies are needed in order
to place deterrence within a proper historical context. However, case studies are plagued
with a variety of problems, as are formal models and statistical techniques. As Lijphart
(1971, 684) notes, "The comparative method resembles the statistical method in all
respects except one. The crucial difference is that the number of cases it deals with is to
small to permit systematic control by means of partial correlations".

This comparative method is what George (1979) advocates in his assertion that

studies of deterrence should rely upon a method of structured, focused comparison of a

small number of studies. He maintains that tracing common threads through a number of

48

 

V
--
s
.
s
.u-
, n
:1.
c..
l
1

 

 

 

cases will provide researchers with the richness of a single case study, with an increased
ability to generalize across a number of cases.30

Such a method may appear to be of some interest to researchers, but it is still
lacking. It is difficult to derive hypotheses and full theories from analyses of only a few
cases. A quote from Achen and Snidal (1989, 147) summarizes this quite nicely; “Only
when yoked clearly to deductive theory and to statistical inference, and made to serve this
end, can case studies provide genuine theoretical contributions”.
2.6 Conclusion

In this chapter I have surveyed the literature on deterrence, and linked it to the
study of alliances. I have argued that a certain class of alliances, the deterrence alliance,
exists for the sole purpose of deterring hostile aggression. A definition of a deterrence
alliance, therefore, is one in which the major power forms an alliance with the minor
power with the express intention of deterring aggression against the smaller state, and the
major power defender is able to undertake some action to demonstrate the value it places
in protecting its protége'.

Following from this discussion, I turn in the next chapter to a formal explication
of a theory of extended deterrence via alliance. 1 link the twin concepts of security and
autonomy with the rationale behind asymmetric alliance formation. I then discuss how
great powers are able to represent, through the process of costly signaling, the value they
hold for their minor power allies. These costly signals can be viewed as investments in a
protege. I argue that such an alliance arrangement will deter aggression and, if that fails,

be much more reliable than prior research on alliance commitments has suggested.

L;

30 Dion (1998) discusses the comparative case study method in some detail, and
ascertains that the flaws often associated with the method are not as grave as are often

49

\-

 

The theory and forrnalizations set forth in the next two chapters pay close
attention to the issues of conceptualization and theory construction that have been
addressed in this chapter. I am cognizant of the debates surrounding much of these
modeling debates, and recognize the inherent limitations of relying merely on individual
cases. I also acknowledge that formal mathematical models and statistical significance
tests also mean nothing without relevant empirical evidence to flesh out the theory. I
concur with both the statements of Schrodt (1982) and Zinnes (1991), who maintain that
mathematical modeling for the sake of math is merely an exercise in esoteric modeling.
Models, formal and statistical, must be able to tell us something about the empirical

world that surrounds us. It is in this vein that I proceed into the remaining chapters of

this dissertation.

assumed.

50

CHAPTER 3

A THEORY OF EXTENDED DETERRENCE VIA ALLIANCE: SIGNALING
COMMITMENTS BY INVESTING COSTS

3.1 Introduction

In this chapter I develop a theory of extended deterrence via alliance. The theory
developed herein expands upon Morrow’s (1991) conceptual framework of autonomy
and security within the context of extended deterrence. I argue that great powers utilize
deterrence alliances, as defined in the previous chapter, as a means to pursue international
goals and protect those issues that they hold vital to national security. In the case of
alliances with minor powers, the focus of this dissertation, the great power gains
concessions from its smaller ally, and in return provides deterrence benefits to its partner.
NATO can be construed as a good example of this type of security arrangement. The
United States was able to utilize its alliance with the states of Western Europe to maintain
a presence on the European continent and confront and contain the Soviet Union. The

smaller European partners to this alliance were able to gain security from the American

nuclear umbrella."

 

3' While this dissertation is not specifically about NATO, the North Atlantic alliance
demonstrates the concept of a great power/minor power alliance quite well. Discussions
regarding NATO’s formation, and the American role within, can be found in Osgood
(1968). Chemoff (1995) also discusses the functioning of NATO and its changing role
after the Cold War. On a more theoretical level, Olson and Zeckhauser (1966) use
NATO as their example of alliances as collective goods.

This propensity within the literature to focus on NATO has its roots in the
difficulty of how to exactly analyze alliances. As Liska (1962, 3) has written, "It is
impossible to speak of international relations without referring to alliances: the two often
merge in all but name. For the same reason, it has always been difficult to say much that
is peculiar to alliances on the plane of general analysis". In responding to the above
citation from Liska, Glenn Snyder (1991, 121) notes that "would-be alliance theorists.

51

The existence of such alliances, I maintain, shapes the perceptions of a potential
adversary. How will an adversary know if a defending state will intervene? Under what
conditions can it expect that a crisis could escalate to a three party war? The actions a
great power defender undertakes on behalf of its protége’ shape the actions that a potential
aggressor will take, and is essential to understanding conditions under which extended
deterrence via alliance will be successfill.

What emerges, then, is the issue of credibility. How does the larger state make its
commitment to its smaller ally known? To phrase it another way, if “alliances can
prevent war through deterrence, or, if deterrence fails, help nations win whatever war
comes along” (Wayman 1990, 94), then how and when do alliances function as credible
deterrents? While, as I argued in the last chapter, there is some problem with this
assertion of Wayman’s, the issue of ensuring credibility and deterrence effectiveness still
exists. I argue that great powers use signals to indicate their value for the alliance, and
that through investing costs in the alliance, they are able to make their alliance
commitments credible. Thus, alliance ties function as signals to adversaries. I argue
that the greater the costs incurred on the part of a defender, the greater the certainty that
exists concerning the likelihood that it will defend its protégé in times of conflict.

Studies of deterrence regularly use the presence of an alliance as a signal of
common interests among states (Huth I988; Huth and Russett, 1993; Jervis, 1970;
Sorokin, 1994). These studies, however, tend to focus on the existence of an alliance as
the signal, rather than in any specific transfer of tangible benefits as the signal of

commitment. The theory developed herein expands upon these earlier ideas, and argues

realizing that they do not really want to write a general theory of international relations,
retreat to something more manageable-another analysis of NATO, perhaps."

52

that in order to maintain credibility, great powers must use tangible benefits as signals of
their commitment to their minor power allies. This public exchange of benefits can be
witnessed by potential aggressors, who must then ascertain whether the benefits indicate
that a defender has a willingness to intervene on behalf of a protege in a crisis.

In the remainder of this chapter I develop a theory of extended deterrence via
alliance, with a focus on the costly signaling actions of a defending state. In the preceding
two chapters I surveyed the current state of the literature on alliances and deterrence, and
placed this research project squarely within the camp of the rational deterrence theorists.
I now turn to a full explication of my theory. The theory developed below examines and
specifies the linkages that exist between structural conditions and national level
attributes. Particular focus is placed on the impact that the existing alliance and its
actions has on shaping the perceptions of a potential aggressor. In this vein, therefore, I
begin with a discussion of the level of analysis issues that emerge in my theory, and then
turn to a conceptual framework that helps focus and guide my theoretical concerns.

One final note before proceeding. Because I conceptualized deterrence in the
previous chapter as a policy that is extended by a defender to a protege, the focus in the
theory developed below is on the actions of a great power defender in its relationship
with its minor power protégé. However, it should be noted that the actions taken by the
defender influence greatly the actions that an aggressor will take. Therefore, while my
discussion below is often couched in regards to the actions a great power defender takes
in regards to its protégé, the impact of its actions are of great importance to the potential

aggressor as it considers making a demand of the minor power target.

53

'1
t
..

 

 

 

‘\
>—
t_-_ '
.. . .
~
, ,, .
. n
, n
.. _
~
‘.
s..."
I D
an
' r
t, N ~
‘ I
"~
I
l'
Au
. -7
\
w
‘ i
..
x
‘
‘
o . ‘.
.
,.‘
c
‘-
\MN
. ‘1
m
'\

 

 

 

 

 

.t,
.\‘ .-
~1

\
"«
‘l a

3.2 The Levels of Analysis Problem in Deterrence Situations and the Agent-

Structure Debate

As issues of deterrence and alliances are examined, one is struck by a seemingly
perplexing problem. Alliances tend to be systemic factors, as evidenced by a large
amount of the literature (Midlarsky, 1988, 1989; Morgenthau, 1985; Schweller, 1998;
Snyder, 1997; Christensen and Snyder, 1990; Walt, 1987; Waltz, 1979). Rightly or
wrongly, most of this literature, firmly grounded within the realist/neorealist camp, views
alliances in terms of capability aggregation. Much of the debate on polarity and
international stability (Deutsch and Singer, 1964; Waltz, I964) centers on the role of
alliance blocs and their role in maintaining peace. The terms of such discussions tend to
be couched within a framework of relative power and its distribution throughout the
international system.

In contrast, decisions on whether to provide extended deterrence through an
alliance relationship, or whether or not to initiate or escalate a crisis, are made at the
national or individual level. In fact, a rational choice approach to international affairs
assumes that decisions are made by a unitary actor (Morrow, 1997).32 How can such
disparate levels of analysis be reconciled and still provide valid, meaningful insights into
the functioning of deterrence relationships? This is one of the central concerns of the

32 Much recent literature attempts to move away from this unitary actor assumption,
notably Schultz (1998) and Smith (1998). See also the essays in the volume edited by
Siverson (1997). Much of this recent spate of work focuses on inter—party competition
and pressure from domestic interests. However, the final decision on whether to
undertake any foreign policy venture tends to rest in the hands of one individual. Thus, I
retain the assertion that the ultimate decision-making process follows from the unitary
actor assumption. I do, however, maintain that such competition is public to a large
extent, and can be witnessed by outside parties.

54

level of analysis problem, and finds itself manifested in the debate on agency and
structure.
3.1.1 Agency and Structure

The agency-structure debate in international relations is largely concerned with
providing explanations of how states act in the international system. The definition of agent.
following the writing of Wendt (1987) is a single actor, which in the international system is
simply a state. The question that emerges from the debate is quite similar to the level of
analysis problem, and is concerned with the proper unit of analysis in regards to explaining
intemational phenomenon. Two issues emerge from this debate, one theoretical and one
substantive. The first is the definition of structure, and the second is how to transcend the
gap between theorizing about international relations and actually empirically analyzing
them.

The concept of "structure" is an oft-maligned one in the literature on international
relations. Waltz (1979), the leading neorealist. identifies it as "the arrangements of units
in the system". Dessler (1989) defines it, according to a scientific realist vieWpoint, as
"the social forms that preexist action." In turn, Morrow (1988) states that structure is "the
processes which create transitive collective orders". Regardless of how it is defined, it is
apparent that structure provides some sort of restraining framework that inhibits how
states act.

The problem that arises is one of causation. It becomes difficult to determine if
agent shapes structure, or if the reverse is true. Dessler and Carlsnaes (1992) tend to
favor a more dynamic model between structure and agency, while Morrow advocates a

focus on preferences of the individual agents. Morrow supposes that game theory serve

55

 

as a bridge between agent and structure, with the preferences of agents clearly outlined.
Dessler, and to a lesser extent, Carlsnaes, suggest that agent and structure can assume
different forms, and these changes will influence how each of these phenomenon behave.

What implications for empirical research can be drawn from this debate? Two
examples will suffice, in order to demonstrate the feasibility (and pitfalls) of extending the
debate from the theoretical domain to the empirical one. A researcher can examine, for
example, the alliance-war relationship. Is the polarity of the system (structure) a
constraining factor, or is it the attributes of the individual alliance (agent) which are

responsible for the presence of conflict? Another area, which is currently one of the "hot"
topics in international relations, is why democracies do not engage in wars with each other.
Is it qualities of the agents or structure that cause such phenomenon to occur? Of utmost
importance here is the fact that the two interact to some degree: they do not exist solely in a
vacuum. Uneven power distribution or alliance formation (both structural factors) may
deter an aggressive state, due to the fact that state leaders may not wish to be defeated in a
military contest. The costs are not worth the risks of gaining some benefit through
asymmetric conflict initiation.

Of course, the goals of a leader (agency factors) can transcend or mitigate the
influence that structural factors have on choice processes. Pursuit of a desired goal may lead
to conflictual outcomes, as evidenced by the onset of the Second World War. It was quite
apparent that the Germans could not emerge victorious from a long two-front war,
especially if the United States became involved. However, Hitler was determined to pursue
his policy of conquest, and embarked upon a murderous course of action. We witness the

failure of structural explanations here in this example, as they are ameliorated by the goals

56

 

 

 

 

 

 

 

vg‘

 

‘,

7.. r’

n..--

and desires of an agent. Regardless, the philosophical underpinnings of the debate suggest
that a scientific approach to international relations is possible, even in areas such as
comparative foreign policy, where such rigor has ofien been lacking. As a discipline grows
there are certain to be disagreements surrounding how research should be conducted, and
international relations is no exception.

The theory developed here is unabashedly rational in its outlook, and focuses on
the individual state acting within the constraints of the international system. Although
various authors (Singer, 1961; Waltz, 1959) have made distinctions between the various
levels of analysis, none have posited linkages between the various levels, as the agent
structure debate would have us do. Singer’s (1961) seminal piece demarcates the level of
analysis problem between systemic and national factors. Waltz (1959), in contrast,
focuses his attention on the causes of war emanating from the individual, the state, or the
international system.

The belief in structural factors as the cause of war culminated in Waltz’s (1979)
Theory of International Politics, and has effectively stunted much of the research being
conducted in the realist paradigm.”3 Those who are adherents to Waltzian neorealism
argue that any attempt to theorize at lower levels is reductionist in nature, and should be
avoided at all costs. As developed below, a theory of extended deterrence via alliance

requires that national and international factors must be taken into consideration. Agency

33 Recent work by Jervis (1997) on system effects attempts to move beyond mere
structural theorizing and incorporate issues of perception. See also the review by
Hopmann (1998) on Jervis’ new work. Additional debates on the degenerative nature of
(neo) realism can be found in the December 1997 issue of the American Political Science
Review. See the comments by Vasquez, and responses by Christensen, Elman and

Elman, Schweller, Walt, and Waltz. Additional challenges to realism can be found in
Vasquez (I998).

57

., .l‘ -
17%,. ‘
J. . -.

L-
-..

 

 

and structure do interact, and cannot be viewed as separate processes functioning
independently.34
3.3 A Conceptual Framework: Opportunity and Willingness as Ordering

Concepts in the Study of Extended Deterrence via Alliance

In an attempt to bridge this gap, at least in the study of extended deterrence within
an alliance context, I utilize a framework and terminology developed by Starr (1978).35 I
conceptualize a state's behavior as responding to various opportunities within the
international system. An opportunity, according to this framework, "means that
interaction exists between individuals of one nation state and those of another so that it is
possible for conflicts to arise-and to arise over values potentially important enough to
warrant the utilization of violent coercive action by one or both" (Starr, 1978, 368).
Opportunity is the possibility of acting, given a set of circumstances within the
international system.

Let us pause for a moment and take an example from the empirical world. The
United States established a treaty of alliance with Taiwan in 1954, largely in response to
mainland China’s shelling of the offshore islands. However, the United States had just
concluded its involvement in the Korean War, and was interested in maintaining an active
role in Southeast Asia. The establishment of a military alliance with Taiwan permitted

the United States to establish bases on Taiwan, and confront communist China under the

—___

34 The recent book by Friedman and Starr (1997) is the most recent attempt to posit
linkages between agency and structure.

35 The Opportunity and willingness fi'amework of Starr borrows heavily from the work of
the Sprouts (1956, 1965,1968) in which they posit relationships between men (or states)
and their environment. Their focus is on the constraining influences of the environment
surrounding individual actors.

SR

 

guise of containment. Thus, the United States was able to pursue an active policy of
containment in the Far East, and continue to encircle what it viewed as the communist
menace.36

However, this relationship engendered some risks as well. Being allied with
Taiwan increased the likelihood that the United States would be drawn into conflict with
communist China, as China maintained that Taiwan was merely a rebellious province and
sought to subjugate it. Debate over the nature of the treaty to be signed between the
United States and Taiwan, and public debate within the United States regarding foreign
policy towards China indicated to the communist regime that pressure could be exerted
on Taiwan with impunity. In 195 8, therefore, the mainland Chinese regime once again
began bombarding the offshore islands of Taiwan. It sensed that a shift in its strategic
environment (American concern over the treaty between the United States and Taiwan)
presented China with a good opportunity to make a demand of the Nationalist
government on Taiwan.

If we assume that favorable conditions exist for the initiation of a crisis on the
part of a challenger, we then require an explanation of how and why states will take
advantage of these opportunities and potentially exploit them. Willingness, in the
framework of Starr and his colleagues, is another way of perceiving a state’s decision
process. “It is through willingness that decision makers recognize opportunities and then

translate these opportunities into alternatives that are, in some manner, weighed”

(Siverson and Starr, 1991, 25). As Most and Starr (1989) point out, by focusing on

h

36 See George and Smoke (1974) for a good history of the establishment of the American
deterrence commitment to Taiwan. See Christensen (1995) for a more detailed

discussion on the role of American domestic politics and their interaction with the pursuit
Of American foreign policy goals at the beginning of the Cold War.

59

 

opportunity and willingness, one is engaged in a study of political processes. An
approach that envisions politics as a process is better equipped to account for the
sequential nature of interactions, and account for any dynamics that may exist in the
course of some event.

Having discussed the role of alliances and their linkages to extended deterrence. it
is only natural that we must now turn our attention to the arena in which deterrence
comes into play. International crises are viewed here as the ideal arena in which to
examine alliance behavior in the context of extended deterrence. As Morgan (1994, ix)
notes, "it has been argued that while some crises do not end in war, all wars are preceded
by crises; thus, our ability to explain war rests with our ability to explain crisis
outcomes". And, as Lebow (1981) notes, crises and deterrence tend to go hand in hand.
According to this logic, therefore, crises are necessary, but not sufficient conditions for
the occurrence of war. Of course, in order to prevent any obfuscation, a definition of an
international crisis is needed.

All interstate crisis is defined by Snyder and Diesing as "a sequence of
interactions between governments of two or more sovereign states in severe conflict,
short of actual war, but involving the perception of a dangerously high probability of
war" (1977, 6). In a crisis situation, there is the likelihood that an alliance will be

required to act in some manner, either to salvage the peace or go to war to maintain its
international position. Alliance action (or inaction) requires that some or all of the
members of the alliance honor their commitments to the alliance's purpose.

In this study, I conceptualize a change in the status quo between an aggressor and

a target state as an opportunity. By opportunity I mean a situation in which the status quo

60

.7 ,, ,
-05 ‘
11.11-.
I
-I
i..".'..
41'
l .7-

‘e'v a.

 

 

“- as
v—
I
I.
l.
' a
-~.,,
.
V .
o‘_‘
. ,—
w
t"
n
.
-‘_
‘
a
»4
-.
'-
x
c
u.

   

 

.‘. u
x
n
' ,‘A‘
.
‘_‘~
.
x.
\" ,
t. .
l

 

 

has been altered in such a manner that a potential aggressor feels that the time is ripe to
initiate a crisis in an attempt to gain a favorable outcome.

A crisis evolves along two dimensions. First of all, I maintain that such disputes
occur because of disagreement on some issue in the status quo. I assume that crises arise
due to contentious issues between the parties involved. Each state has a specific value
that it attaches to the issue in question. The value that is placed on the issue determines
in part if a state initiates and escalates a crisis, resists or acquiesces, or intervenes. While
the issues at stake are exogenous to the model, they are important empirically.

The aggressor has deemed the issue to be of enough importance that he is willing
to challenge the prevailing status quo in order to rectify the situation. Such a challenge
provides a defending state with the choice of whether or not to intervene on behalf of its
ally. I view the initiation of a crisis as the start of a sequential process, which can result
in war, but does not necessarily have war as the sole outcome. A breakdown in extended
general deterrence, therefore, can be linked to a shift in the international environment that
provides an aggressor with an opportunity to make gains on some issue it holds dear.

This breakdown in extended general deterrence leads to a situation of extended
immediate deterrence. In case such circumstances arise, a defender must determine if the
value of the issue at stake is one that is worth fighting for. Additionally, the aggressor
must determine whether escalating the crisis, and possibly not gaining its demands, is
worth the effort. In cases of extended immediate deterrence, therefore, both defender and
aggressor must determine how much they are willing to absorb in order to gain their
objectives. Will an aggressor exploit a momentary weakness in the deterrence alliance it

faces. or will it recognize that the military power it faces is too great? If extended general

61

 

 

.
I.
,‘-
I‘ ,
s”,-
> .
t
u
--
\
a. I.
"»
4'.
'.
-
, o
>-

.z-2
.,.¢

 

deterrence encounters are noted as attempts to exploit a structural opportunity, then
extended immediate deterrence encounters revolve around the (potential) use of military
capabilities. Figure 3.1 depicts the relationship between the two levels of deterrence, and

the relationship between the defender and challenger, within the possibility of crisis

 

 

 

escalation.
FIGURE 3.1
STAGES OF CRISIS ESCALATION I
Challenger’s Action Defemler’s Respopg State of Relations
No action taken General Deterrence Status Quo i
Challenge * *
Immediate Deterrence Crisis

Escalation
t l l
Large Scale Use of Force Deterrence Failure War

 

 

 

 

In this vein, then, I argue that an existing alliance has an impact on the possibility
of a crisis escalating to war. Previous studies of alliances and the occurrence of war have
neglected to account for the sequential nature of disputes. It is through the notion of a
crisis being a process, with decisions being made at various points during the sequence,
that we can understand the impact which alliances have on deterring aggression, and, if
extended general deterrence fails, then the influence they have on inhibiting the spread of
the dispute. The possibility of a third party intervention raises the potential costs that an

aggressor must absorb in order to obtain its desired outcomes.

62

3.4 Willingness and the Pursuit of National Goals

Having addressed the context in which allies and adversaries interact, I now turn
to a discussion of the reasons why deterrence alliances form, and what factors influence
the decisions of an aggressor and defender in the context of an adversarial relationship.
Since deterrence is a benefit extended by a defender to a protége’, the theory deveIOped
here focuses on the influences that shape an aggressor’s behavior, given the existence of
an alliance prior to crisis initiation. Intra-alliance behavior, I argue, influences the
willingness of a potential aggressor. One cannot assume that any of the states involved
acts in a vacuum. Thus, while the discussion below is largely couched in terms of the
interactions transpiring within an alliance, it is the impact that these interactions have on
the behavior of the aggressor that is of utmost importance for understanding extended
deterrence via alliance.

I maintain that four factors shape a defender’s behavior and the perceptions of
the potential aggressor: rational behavior, the issues surrounding the relationship of these
three states; military capabilities; and domestic politics. I examine each of these
foundations in some detail, and then turn to a signaling theory that explains how
defending states communicate their commitments in a credible manner, and in doing so
provide successful extended deterrence to their allies.

3.4.1 Rationality
Before progressing into a detailed elaboration of a signaling theory of extended
deterrence via alliance, one initial assumption is required at the onset. States, whether

they are initiators of crises, are the targets, or are the defenders, are rational actors.

63

 

 

 

 

 

 

 

 

 

 

'9
Av

' .

V \
.. "-.

‘u.
t- ‘
' .
‘5.
ts

5;.

 

"Being rational simply implies that the decision-maker uses a maximizing strategy in
calculating how best to achieve his goals" (Bueno de Mesquita, 1981, 31).37

The assumption of a unitary, rational actor in international relations is quite
widespread (e.g., Bueno de Mesquita, 1981; Bueno de Mesquita and Lalman. I992;
Morgenthau; 1985; Morrow, 1997; Waltz, 1979; Zagare, 1987), although often
misunderstood. Assumptions of rationality stipulate that actors have preferences, and that
they can be ranked in order of greatest preference to least favorite preference. These
preferences are determined by the various costs and benefits associated with them, and
are assumed to be fixed for the period in which a decision is to be made.

It should be noted that decision-makers make choices behave in a manner denoted
as instrumental rationality. This merely stipulates that rationality is a simple ends-means
calculation (Zagare, 1987, 9). This contrasts with a more detailed definition of
rationality, known as procedural rationality. Procedural rationality requires a decision-
maker to “properly” define his goals and consider all possible alternatives (Zagare. 1987.
8). Procedural rationality places much greater emphasis on the subject nature of the goals
of the decision-maker. Instrumental rationality, as adopted here, assumes that, in the
context of some decision scenario, a decision-maker can order his preferences and choose
the strategy that greatest affords him the ability to arrive at his most preferred outcome.

It says nothing about the relative worth of the decision-maker’s goals; only that he attains

them in the manner that incurs the least amount of cost.38

 

37 Discussions of the debate surrounding rationality 311d its applicability can be found in
chapter 2.

38 A recent paper by Bennett and Stam (1998) addresses the issue of whether instrumental
rationality is applicable across a variety of cases and geographical regions, or if it is
confined to such scenarios as European conflicts or Great Power behavior. With some

64

 

 

 

 

 

 

.~ -
.3...
o
_....t..
u-_ I
‘. 8
-.. i

. t

-§... 1
‘ r
o

" :4
\a
~v
”A..A

S“

 

 

 

 

 

 

With such a focus on decision-makers, then, the model is of necessity state-
centric, for the focus is on the choices which individual states make. The state in this
conceptualization is represented by a unitary decision-maker, which makes decisions
based upon the interests of the state.39 This holds true if we are examining the decision to
initiate a crisis by an aggressor, or a choice made by the target or defender in how to
respond to a specific use of force. Although a BC alliance exists, it is an aggregation of
states, bound together for some common purpose. Each state within the alliance must fl "

make a conscious choice when confronted with an opportunity to act.40 l

3.4.2 Pursuit of Goals

 

Secondly, I assume that deterrence alliances form for specified reasons. If we are

to understand such structures, we must-understand the context within which they are

 

caveats, they do find that instrumental rationality (in the guise of the expected utility
theory of war) is applicable across regions and across time.

39 Students of bureaucratic politics and prospect theorists often take issue with this
depiction of a unitary actor taking actions on behalf of a state’s interest. As Bueno de
Mesquita and Lalman (1992, 26) note, when it comes to instances when military force is
likely to be used, the ultimate decision rests in the hands of a single individual. As they
further explicate (27), “So long as decisions to negotiate or to use force are made
rationally (or as if instrumental rationality were operative), it does not matter, in the
context of our model, whether the decision is made by a single actor or by a group”. For

an elaboration and discussion of theories that differ from rational choice, see the essays in
Geva and Mintz (1997).

40 Given that the focus of the model is on an existing alliance’s impact on the initiation
and escalation of a crisis, it makes sense to focus on B and C as individual actors.
However, in many cases, there exist allies on A’s side that are also involved in the
initiation of hostilities. One need only look at the origins of the First World War to
witness Germany and Austria-Hungary coordinating their actions prior to commencing
hostilities against Serbia to realize that an initiator can be more than one state. (See Joll,
1992, for an in-depth discussion of this phenomenon). In order to cope with this
potentially complicating issue, I assume that A behaves as if it is a unitary rational actor,

regardless if it is a threatening alliance of states (as in the case of July 1914), or actually
is a single state.

65

formed.“ As I argued in the previous chapter, deterrence alliances exist to prevent
hostile aggression from a defined foe. In the context being examined here, I am looking
at alliances in which power is distributed asymmetrically.

Minor powers can offer very little to their major power edunterparts in terms of
military capability. However, to refer back to Rothstein and Keohane from the previous
chapter, such states can have a tremendous amount of influence on their great power
partners. I maintain that such relationships exist is due to the trading of security and
autonomy benefits within such alliances, as advanced by Morrow (1991).

Minor powers are unable to provide much security to their great power partners.
For example, as Marlowe (1965, 300-1 ), points out, Egypt was able to contribute little
militarily to the British. Therefore, the British role in Egypt was primarily of providing
security to the Egyptian peOpIe. In exchange, the British were able to maintain control
over vital trade and strategic routes in the Middle East.

Given that minor powers tend to be weak, they cannot hope to provide direct
security benefits to their alliance partners that are great powers. Hence they must
provide some additional benefit that the great power desires. For example, in the post
1945 era, the United States used many of its alliances as means to contain the Soviet
Union. By offering protection to its allies in Western Europe, for example, the United
States was able to gain a military presence in Europe. In addition, the United States was
able to gain access to European markets for its exports. Such autonomy benefits are not

limited to the nuclear era. For example, the British allied with Turkey in 1878 to deter

 

4‘ The best study of context and international relations remains Goertz ( l 994).

66

 

Russian aggression in the Balkans. In return, the British were able to gain territorial
control over Cyprus, and gain access to the Dardanelles (Jelavich, 1973; Langer, 1950).
Additionally, great powers can gain autonomy through the importation of
strategic raw materials from their protégé, or through increased trade. Regardless of what

benefits are gained by both parties, the alliance always provides the possibility of
entanglement on the part of the great power (Snyder, 1984), in which the defender faces
the possibility of becoming dragged into a conflict it did not desire. If such a crisis
occurs, then the defender must weigh the benefits it receives from the alliance versus the
costs that will be incurred if its promise of assistance is broken.

When examining such deterrence alliances, however, a more basic question
emerges. What types of states join together in such structures? Differing opinions exist
regarding this question. Siverson and Emmons (1991) ascertain that democracies are
more likely to ally, in comparison to a baseline probability model. Given the spread of
democracies in the world, particularly after the Second World War, this finding has a
certain intuitive appeal.

However, Gartzke and Simon (1996) and Simon and Gartzke (1996) mount a
serious challenge to this view. They maintain that “friends” do not need to sign treaties
of alliance in order to cement their security concerns. Democracies, as is frequently
argued in the democratic peace literature, tend to maintain cordial relations.42
Adversaries, on the other hand, have something to gain by allying. The argument made
by Gartzke and Simon (1996) is that non-democracies can provide goods and services to

democracies that other democracies cannot provide. This includes such goods as

42 . . . . .
The current literature on the democratic peace 18 volumrnous. For a good introduction
to this current topic of interest in international relations, see Ray (1995).

67

 

allowing the defender to establish bases on a protégé’s soil, or coordination of foreign
policy goals. With fewer constraints on the part of a non-democracy. concessions are
easier made in forming alliance relationships.

Hence, we should expect, as Gartzke and Simon (1996) suggest, to see more non-
democratic alliances formed for purposes of deterrence than democratic ones. If we
examine the data collected, of the 20 deterrence alliances formed between 1870 and
1984, 9 of them (45%) have a democratic defender and a non-democratic protége’. While
not an overwhelming number, it does call into question the logic espoused by adherents

. . . . . 4
to a VlSlon of democratic alllances belng most common. 3

 

Of course, the underlying assumption surrounding why deterrence alliances form
in the first place, and why their members are willing to trade security and autonomy in
spite of the fear of abandonment and entrapment, is the fact that some contentious issue
exists between a potential aggressor and the prote'gé of the defender. These issues are
often obscured in much of international relations scholarship, with its traditional focus on
power. However, as a number of authors (Diehl, 1992; Holsti, 1992; Vasquez, 1993)
have noted, the role that issues play in international conflict are often more important
than military capability ratios or power distribution in a systemic setting.

If we are to understand why deterrence alliances form, we need to examine what
issue is at stake between the protege, who desires protection from some potential
attacker, and the attacker itself. Most states have, at some point in their history,

squabbled with adversaries over the issue of territory. Other states attempt to impose

—_¥

43 It should be noted of the remaining 11 alliances, 8 of them are comprised of two non-
democratic states, and only 3 are comprised of two democratic states.

68

4..

.4 1”
u
- v.
-
. . \l
"x-“-
. s...
‘ . I .

Q.- ,su .. \

II .
I‘. ' l .
.3 ll
.

 

 

 

 

 

certain regime types on a smaller state, or attempt to control a target economically.
Regardless of the issue, it is important to note that the issue affects two sides and, in the
case of extended deterrence, potentially three sides. These issues must be examined and
analyzed if extended deterrence success and failure is to be explained.

3.4.3 The Role of Power and Capabilities

Of all the concepts utilized in the social sciences, power is probably the most
maligned. In spite of this, it is of central importance to both deterrence and alliance
theory. While power, according to Morgenthau (1985) can be defined as the ability of A
to get B to undertake some action C that B would not ordinarily like to do, in the case of
alliances, power is much more than the ability to influence. Instead, when power is
discussed in this dissertation, it is in regards to the military capabilities that a state
possesses.

As Huth (1988, 34) points out, the balance of military capabilities is largely a
structural feature of any deterrence relationship. Additionally, the use of military power,
and its buildup, tends to play a large role in deterrence success or failure. “A credible
deterrent depends upon whether the defender appears to possess both the military
capabilities to inflict substantial costs on an attacker and the will to use these capabilities
if necessary” (Huth, 1988, 33).

As noted above, and demonstrated in the formal and empirical analyses in the
following chapters, power is especially important at the level of extended immediate
deterrence. The argument made above asserts that aggressor states will initiate a crisis

when they view a favorable Opportunity in their relationship with their intended target

69

 

 

and its defender. It is at the level of immediate deterrence, when a crisis has the
possibility of escalating to conflict, that capabilities play the greatest role.

A potential attacker desiring to alter the status quo in its favor, and sensing a
positive environment for doing so, has three military strategies to choose from: limited
aims strategy; rapid offensive attack (blitzkrieg); and a war of attrition.44 Each strategy
assumes different goals for the attacker, and is impacted upon by the balance of military
forces in different ways.

If an attacker adopts a limited-aims strategy, it makes a rapid seizure of some
territory that belongs to the target, then build up its forces and be prepared to repel any
counterattacks on the part of the target and its defender. Such a strategy assumes that the
target can be overwhelmed quickly, before it has the opportunity to mobilize and
counterattack. Israeli raids into Jordan in the 19505 typify this sort of strategy. As
reprisals for Jordanian terrorist attacks within Israel, the Israeli army mounted numerous
short attacks into Jordan, where they were able to overwhelm Jordanian troops and inflict
punishment before troops could be mobilized (Blechman, 1972; Shimshoni, 1988).

The second strategy available to policymakers is that of rapid offensive attack, or
blitzkrieg. The underlying principles behind such a strategy is the rapid destruction of a
target’s military, and force a defeat due to collapse of the target’s economic and civilian
infrastructure before it or its allies have a chance to mobilize and retaliate. Such was the
case of German aggression towards Poland at the start of the Second World War.

Germany overwhelmed the Polish army with such speed and ferocity that there was no

N

44 Much of the work on different types of military strategy can be found in Huth ( 1 98 8);
Mearsheimer (1983); Posen (1984); Reiter (1998); and Stam (1996).

70

 

 

 

 

 

 

 

‘ .
.
‘eiv
.
I
I.
s
'c
x
. c
.‘ .
o
‘n
l

 

chance for the Poles to mobilize and ward off the German onslaught, nor time for the
British and French to intervene on behalf of their Polish ally.

The final strategy available to statesmen is a strategy of attrition. If a state adopts
such a policy, it is hoping, as in the case of rapid offensive attack, to decisively defeat the
military of its adversary. However, a strategy of attrition differs from a blitzkrieg
strategy in that the leaders of the attacking state realize that the war will not be quick, but
protracted. Witness Japanese involvement in the Pacific theater of the Second World
War. The Japanese leadership recognized that they would be unable to defeat the United
States in a prOtracted fight. However, the hem of the Japanese military was to inflict
enough damage on the United States that the Americans would capitulate to Japanese
demands rather than fight. Once the war became drawn out, the Japanese were incapable
of attaining victory, due to the overwhelming superiority of the American military base.

If we examine these three strategies we see that different balances of military
capabilities are at work. The balance of military capabilities can be viewed as operating

at three distinct levels: the immediate balance of forces; the short-term balance of forces;

and the long-term balance of forces.“

In the case of a limited-aims strategy, the immediate balance of forces is most
crucial. If a state wishes to inflict some sort of punishment on an adversary, it requires
the ability to win quickly and decisively. Hence, the troops that are on the ground at the

immediate point of contact are most critical in determining the success of the battle.

When it comes to a rapid attack strategy, the immediate balance of forces does not

play as large of a role. Rather, it is the forces that can be mobilized to repulse an invader

7l

 

 

 

r-‘ h
t' ‘
..,. n
t

n tau-I
t.’ r‘
a .u.
r

i “‘
" .-
._ ‘
Iv

 

“ l
"
- ., -.
n
' I
t.‘
.
5.
.
‘l '
‘I
F! U‘
M»
~l
t.
l
I.
w .
i
‘s l
a.
‘1
l V.
l‘
V - b
“I
.
\I
's
I ~ I

 

in a short span of time that are important. If an attacker wishes to defeat its target
rapidly, it needs to have the ability to decisively win an initial battle, and then reinforce
its troops to continue a rapid attack and eliminate the remainder of the target’s military
power before external powers have the ability to intervene.

In a war of attrition, the long-term military capability, coupled with a country’s
economic base, becomes critical to determining the victor. Knowing that it cannot hope
to win such a conflict in a rapid manner, a state must be prepared to mobilize every
portion of society in order to prevail. Given the enormous costs that such a conflict
inflicts upon a state, both attacker and target (as well as defenders that become involved),
it is not a strategy that is normally considered on the part of an attacker. Such wars do
occur, but not as a consequence of a leader deliberately initiating a war of attrition.

Military might, therefore, does appear to play a role in how crises are resolved,
and what level of force is actually used. Notably absent from such an analysis, however,
is the role that domestic factors play on influencing a leader’s decision on what strategy
to follow. I now turn to an examination of the role that domestic political considerations
play in extended deterrence situations.

3.4.4 Domestic Influences

Finally, I assume that domestic politics impinges on the conduct of international
affairs. Much recent scholarship has demonstrated that international politics do not take
place in a vacuum. Domestic politics plays a role in the decision by states to become
involved in conflict. However, this has not always been widely accepted in the academic

community.

__.

’5 Posen (1984-85) discusses these terms exclusively in regards to NATO and the
Warsaw Pact during the Cold War. In spite of his limited application, the terms are

72

 

In the study of international relations after the Second World War, realism
emerged as the dominant paradigm. Promulgated by such theorists as Carr (1949),
Morgenthau (1985), and Bull (1977), the trends within the discipline tended to focus on
the primacy of international politics and issues of power and scourity. Domestic politics
and economic issues were largely relegated to the background, and deemed not worthy of

inclusion in the realist mode of thought.

Recent work, however, has begun to challenge the fundamental assumptions of

realism. In particular, two issues have emerged from this challenging scholarship: the

 

linkage between domestic and international politics as addressed by Putnam (1993) in l
regards to two-level games, and the audience cost concept as advanced by Fearon (1994a,
1994b)
Putnam’s seminal work addressed a simple question. Why should a decision-
maker be concerned with the attitudes of his domestic constituency, especially if they are
in disagreement with him? Given the dominant status of realism in international
relations, one would expect that political leaders would not concern themselves with such
beliefs. Unfortunately, at least for officeholders, this is not necessarily the case. Political
leaders, as rational actors, wish to retain their privileged position in power. However, if a
leader ignores the desire of his constituency, whether it is an entrenched oppositional
force or a reluctant legislature, he is inhibiting his opportunity of remaining in office.
This can be referred to as "double edged diplomacy", for a leader is attempting to bargain
with and satisfy two constituencies; one international, the other domestic.
Therefore, in light of these complications of being forced to please two audiences, l

we can see that a policy leader is bounded in what is considered acceptable by his

A

 

useful in studying deterrence relationships and military strategy elsewhere.

73

domestic constituency. Given these parameters he is able to function rationally, and
maximize his opportunity to remain in office.

Putnam's contribution, therefore, is to move beyond systemic explanations and
unitary actor assumptions. Instead, he brings the individual baCk into international
politics, and demonstrates how he or she makes decisions given international and
domestic constraints. This work sparked an intense debate within the international

relations field about the proper level of analysis for studying international political

 

phenomena. As can be seen from the discussion above, the two-level game framework

 

differs greatly from a realist/neorealist perspective, with their foci on structural power.
Other scholars such as Iida (1993) and Mo (1994) seized upon this idea and formalized it
to examine conditions in which domestic constraints matter, or situations when the
negotiator may have a different desired outcome than his or her constituents.

What happens if a leader ignores the desires of his constituents? As F earon
(1994a, 1994b) demonstrates, there are domestic audience costs associated with
international interactions.46 These costs can sway public opinion, and, in the case of
democratic societies, cost a politician his or her office.47 This is not an uncommon
argument, at least pertaining to the United States. Many arguments have been made

ascertaining that American presidents tie foreign policy ventures to reelection campaigns.

—__

4" Morrow (1994a) places the audience cost concept specifically within an alliance
context.

‘7 In regards to the United States, Mueller (1973), Ostrom and Job (1986), and Nincic

(1992) are just some of the more common studies which demonstrate that public opinion l
can and does play a role in the fortunes of politicians. Hagan (1993) provides a
comparative perspective on the influence of domestic politics and opposition on the
conduct of international affairs. Bueno de Mesquita, Woller and Siverson ( 1992) and
Bueno de Mesquita and Siverson (1995) discuss in more general terms the impact that
war involvement can have on the fates of political leaders.

74

Ostrom and Job (1986) discuss the use of force by the president in regards to election
cycles, and Aldrich et al. (1989) address the implications of foreign policy issues for
presidential electoral outcomes. In summarizing the implications of foreign policy on
presidential elections, Nincic remarks (1992, 122), "When foreign policy violates the
limits of what interests and expectations have defined as acceptable...it will eventually be
pounced upon by the opposition and, sooner or later, denounced by the electorate".
Some recent work (i.e., Raknerud and Hegre, 1997) suggests that democratic

states are less likely to become involved in conflicts, but, once they do, they are more

 

likely to win. This finding emphasizes the fact that it is more difficult for democratic
states to make commitments within the international system, but that they have a
tendency to honor their commitments when they are challenged.

The implication I draw here differs from many popular conceptions regarding the
role that domestic politics play in the conduct of foreign policy. I argue, as does Fearon
(1994a, 1994b), that as a crisis escalates, there is a cost to be paid if a leader backs down.
I take this one step further and assume that domestic costs are paid if a leader fails to
intervene on behalf of an embattled protégé. The logic for this lies in an argument that
could be termed guns vs. butter (Norpoth, 1987; Powell, 1993). In such a
conceptualization, as a defender supplies a protége’ with guns (security), it is not spending
this money at home on butter for its citizens. As long as the guns are going to protect
some vital national interest, pe0ple do not seem to mind. However, if the costs are sent

abroad and then they are not protected when a protége’ comes under attack, these invested
costs are in essence thrown away. It is under these conditions that the electorate becomes

incensed over foreign affairs, and that leaders should take care.

——___

7S

 

l
., 4»-
. r n
K
.» a
I ,

. .t
.- .
...
L. l‘
..

.1.
I...
~—-
. -.._
v
.F
ltl
. - 7.
u. _‘-

 

 

 

 

.
“a.
.L‘
.-
k.
1v
a
-‘~.
., l .l
--
_ c
‘-
~.
-
"—!
\)"
.\
c
‘..
- t
‘t.
I

 

~
'i
'I.
I
~
4
-.

 

3.5 Signaling Commitments, or how to establish a Credible Reputation“

We have seen the influences, both systemic and national level, which impinge
upon the decision of a great power to become involved in an extended deterrence
relationship with a minor power. I now turn my attention to the role that costly signaling
plays in ensuring successful deterrence. Specifically, how can a defender make it known
that it values its protege, and will intervene on its behalf if the protege comes under
attack? Given the variety of influences that affect a defender, it must find a means of
communicating its value for its protege.

In particular I argue that it is through the use of costly signaling that great power
defenders are able to make their commitments to their allies known, and ensure peace for
their alliance partners. Defenders use investments in their proteges as a means of
demonstrating their interest in their minor power allies. These investments, at least in
theory, commit the defender to come to the aid of an embattled ally. I argue that these
investments function as signals that tie the hands of state leaders. If the defender
abandons its ally, then it has in essence wasted these investments. As shall be discussed
in the remainder of this chapter, and formalized in chapter 4, the greater the investment,
the more likely it is that extended deterrence, both general and immediate, will be

successful

48 Modeling a reputation is not the focus of this section. Reputation implicitly (others
would say explicitly) assumes that repeated interaction occurs between states. As the
models in the next chapter are specified, iteration does not exist; a crisis is a one shot
interaction. However, past behavior does influence the belief structures that the models
utilize, so reputation does play a limited role in how aggressors perceive defenders. For
more on reputation in international politics, see Mercer (1996). Huth (1997) provides a

thoughtful critique of Mercer’s work, and Mercer (1997) responds to criticisms of his
work.

76

3.5.1 Defining Costly Signaling

One of the central concerns within international politics is the question of how
state leaders, in the absence of any mechanism able to enforce agreements, make their
promises credible? I maintain that, within the context of extended deterrence, states can
make their promises credible through the use of costly signaling. Costly signaling, as

defined by F earon (1997, 69) can be conceptualized in the following manner. Costly

‘r
I

signaling is the ability of a leader to undertake some action that creates or incurs some
cost that the leader would not be inclined to absorb it he or she were not willing to

actually carry out the promise or commitment that is made.

 

Understandably, these communications of signals do not come without a price.
Of particular importance in the study of international interactions, be they cooperative or
conflictual, is the notion of costs. All actionswhich states undertake entail some form of
expense, and many such actions place restrictions on a state’s range of choices. Entering
into an alliance is no exception to this rule. States within an alliance are faced with the
same dilemma. In an international system characterized by anarchy, there exists no
overarching authority capable of enforcing such agreements. Thus, states must determine
whether or not to honor their alliance pledges based upon the costs that such an
intervention would incur versus the value they place on the current resolution of the issue
at stake.49 In order to demonstrate credibility in their interactions with minor power

proteges, I argue that great power defenders must send some form of concrete signal to

M

49 .

Of course, costs are imposed upon an aggressor state as well. The consequences of
initiating a crisis and carrying it to an outcome of war implies that lives may be lost, or
reputations damaged. Leaders who lose wars which they initiate can suffer removal at
the hands of the victors, or from their own constituencies. Thus, leaders who initiate

77

indicate their value in their ally. It is through the use of costly signals that state leaders
can make their true intentions known.

Fearon (1997) distinguishes between two ideal types of costly signals. The first
type can be defined as “tying-hands” signals. Such signals mean taking an action that
increases the costs of backing down or reneging on a promise. Such signals should
influence the actions a defender takes when a crisis erupts surrounding a protégé, since
they can be viewed as investments made by the defender into the alliance relationship.
These “tying-hands” signals, I maintain, are represented by the investments that
defenders make in providing security benefits to their proteges.

Many great power defenders take some action once a crisis erupts. Oftentimes, as
in the British case prior to the Nazi invasion of Poland, public statements are made
concerning coming to an ally’s defense in times of crisis. Alternatively, other actions can
be undertaken once extended general deterrence falters and a crisis emerges. These tying
hand signals, according to Fearon (1994b) also generate costs, especially 0n the domestic
front. While they do not provide tangible benefits to the protege, they do impinge upon
the reputation of the defender, and potentially raise additionalconsiderations at home.

In contrast, a second type of costly signal can be referred to as a “sunk cost”
signal. Such signals are more difficult to identify in a pure case, but they are actions that
are costly to undertake in the first place. However, they do not come into consideration
in times of conflict. Sunk costs, such as troop mobilizations or arms production, are

costly. but they do not affect decisions made in times of crisis.

*—

disputes must consider the benefits of success versus the costs of failure. In doing so,
they must consider the credibility of the deterrent threat made by the defender.

‘79

‘ . .t
‘m

.- .
l "
| .
n. v ..
Cult. .. ..

 

o.

 

 

. v

 

.. ~-
t

n.
r t
u...

“N.

 

I advance here a concept of a signaled commitment that unifies Fearon’s two ideal
types of costly signal. In an alliance relationship that is asymmetric, as was discussed
above, the alliance partners trade security and autonomy. If a defender transfers goods
such as arms or aid to its prote’gé, it provides the minor power With increased security,
even if the great power fails to intervene.

I maintain that sunk costs and tying hands signals can be combined and viewed as
one form of costly signal; an invested cost. In regards to a sunk cost view of providing
extended deterrence benefits to a protege, the formation of an alliance is costly, both in
regards to materiel and to reputational effects. If a defender transfers goods to its
protege, such as arms or troop deployments abroad, these are impose costs on the

defender that could be spent elsewhere. It is the classic guns versus butter problem, with

the “guns” being sent abroad to assist a prote'ge' rather than providing “butter” at home for

domestic constituents.

Such investment costs also encompass tying-hands signals. This is because they
demonstrate that the defender is willing to invest some tangible, potentially costly, good
or service in its protege, in order to provide security to the smaller power in the alliance.
These investment costs allow a defender to signal its interest in its protege, and also
provide the prote'gé with a means of defense if a crisis escalates. Much of this signaling,
therefore, serves to enhance the credibility of the defender, at both the levels of extended
deterrence, immediate and general.

As Thomas Schelling (1960, 14) points out, the notion of credibility in regards to
a threat or a commitment by one state against an adversary is of utmost importance. If a

state believes that a threat or commitment by one state is credible (i.e., it is true), it is

79

Mr

 

 

 

 

 

Vi

 

 

 

likely that it will be deterred. However, credibility is often difficult to ascertain, and
states must rely on beliefs and witnessed actions in order to determine beliefs and choose
a course of action. Through costly signaling, a defender is able to signal at various stages
of a crisis its value in its ally. These signals can be interpreted and witnessed by an
adversary, who must somehow interpret these signals in order to determine how serious
the commitment among the allies is.

Figure 3.2 depicts the theoretical framework underlying a costly signaling

 

approach to understanding the provision of extended deterrence via alliance.

 

 

 

FIGURE 3.2

THE THEORETICAL FRAMEWORK OF COSTLY SIGNALING AND
THE PROVISION OF EXTENDED DETERRENCE VIA ALLIANCE

INVESTING COSTS INTO PROTEGE

l

CREDIBILI T Y 0F DEFENDER 'S COMMITMENT ENHANCED
l
EFFECTIVE EXTENDED DE T ERRENCE

3.6 Conclusion

In the past two chapters I have examined the impact that alliances have on war,
and the role that they play in establishing both deterrence credibility and stability. I have
maintained that international and domestic factors interact to influence the perceptions
that adversaries have of one another, and that structural and agency level variables need

to be considered if extended deterrence via alliance is to be comprehended.

80

 

 

In the next chapter I present two formalizations of the theoretical framework
advanced in this chapter. I demonstrate that the costly signals sent by a defender have a
great impact on ensuring deterrence effectiveness, and that more than structural factors
shape such relationships. In particular, I maintain that costly signaling helps create an air
of certainty surrounding the relationship between a defender and a protege. This
certainty, I argue, will deter many types of potential aggressors, and lead to the

preservation of the status quo.

 

This focus on certainty stresses the importance of the role that information and

 

 

 

 

 

beliefs play in shaping the outcomes of international interactions. In the following
models, and empirical tests, I capture this uncertainty by focusing on the benefits that are
exchanged between defender and protege. I maintain that the costly signaling that occurs
within extended deterrence relationships serves to shape the beliefs of potential
aggressors, at least concerning their feelings towards the credibility of the alliance
commitment. These beliefs play a great role in determining whether extended general
deterrence succeeds or fails. If extended general deterrence falters, then the costly

signals that are sent by a defender should assist in helping resolve the crisis.

It should not be forgotten, however, that the defender also gains benefits from the
alliance relationship with its protégé. These autonomy benefits are also visible to
potential adversaries, and must be considered within the dynamics of an extended
deterrence crisis. With these thoughts in mind, I now turn to a formal explication of the
costly signaling theory of extended deterrence via alliance that has been developed in this

chapter.

81

 

 

CHAPTER 4

TWO FORMAL MODELS OF EXTENDED DETERRENCE VIA ALLIANCE
4.1 Introduction
In the previous chapter I introduced a theory of extended deterrence via alliance.
I discussed in some detail how extended deterrence emerges as a benefit for a minor
power in an asymmetric alliance, and the international and domestic factors that shape
deterrence effectiveness. I also argued that the actions a defender undertakes influence
the perceptions of a potential aggressor. Signals that are sent between allies tend to relay
some modicum of information to an adversary.

In this chapter I present two formal models of extended deterrence via alliance.

In the first model there exists no uncertainty about the types of players or the signals sent.

In the second model, I introduce the concept of uncertainty, and the issues of perception
that arise due to this phenomenon.

In the next section of this chapter I outline the game theoretic model with
complete and perfect information, and discuss the sequence of moves. Following from
that discussion I then elaborate upon the outcomes of the model, and the costs and
benefits associated with these outcomes. I then turn to a discussion of the types of
players that are available, with type being determined by value for issue at stake and
relative military capabilities. I then delineate payoffs for each outcome for each type of
state, and discuss the preferences each type of state holds for the various outcomes. I

then derive the equilibria of this model, and discuss their implications.
Having presented a simple model of extended deterrence via alliance, I then

introduce the concept of uncertainty, and the role that it plays in informing perceptions. I

82

 

 

 

 

..
. .u’
‘k
I.
.
"t t.

l
..
. ,.
'\
u
c
l
"a
‘

restructure the game so that the defender C has the ability to signal its commitment to its
prote’gé B prior to the onset of a crisis, but the aggressor A does not know for certain
what type of defender C is. I then derive the equilibria from this more fully specified
model, and discuss its implications. I discuss the conditions under which these equilibria
occur, and the implications that can be drawn from them.

Throughout the chapter I draw upon a number of international incidents to help
illustrate the model. These are the Czech crisis of 1939; the Polish Crisis of 1939; the
Serbian-Bulgarian War of 1885; US. Soviet relations in Western Europe after 1961; and
the Sine-Vietnamese War of 1978-1979. In all of these cases an ally has proffered
extended deterrence, but different outcomes emerge from each crisis.

4.2 The Theoretical Model with Complete Information
I now present a formal model of extended deterrence via alliance with complete
and perfect information. Figure 4.1 depicts the model and the sequence of moves.
[FIGURE 4.] ABOUT HERE]
Below are the assumptions that structure the game and the moves that each player makes.
Assumption 1: The model involves three players, A, B, and C. A, the aggressor, is a
threat to B. B is the minor power prote'gé of C, and C is a great power defender of B. A

BC alliance is assumed to be established exogenously from the game, and is in existence
prior to any crisis that erupts.

. . . . so
Assumption 2: A crzszs concerns only one issue.

Assumption 3: Each actor in the model is assumed to be an expected utility maximizer.

Assumption 4: The game is played under conditions of perfect information.5 I This means
that every player in the game knows all previous moves that every other player has made.

—_

5° While this may be a controversial assumption, it is borne out by much of the historical
evidence. It also makes the model more tractable. Thus, I do not account for issue
linkages within the model, as Morrow (1992) does.

83

 

 

 

 

 

 

_ .

‘. .v.
'.
..

Irr-
rw ‘
nod-r. .

‘.""
r - ..n‘
. _ ,

, l
a. . .

o 'v'
t - y o
um .
I'O .

' 'It

'v ,

. ’9"

‘.,r
,l
l
t
l

 

 

 

 

 

Assumption 5: the game is played under conditions of common knowledge. Common
knowledge implies that everybody in a game knows everything, everyone knows that
everyone knows everything, into infinity. Nothing is hidden or secret regarding
preferences over outcomes or payoffs concerning outcomes.

Assumption 6. Information in the game is complete. Complete information is
characterized by having all payoflfs to all players as common kno wledge.”

Assumption 7: Movement in the game is sequential rather than simultaneous. Each
player moves in the prescribed manner laid out in the game. and then the other players
have the opportunity to make their moves.

Assumption 8: In a game of complete and perfect information, states play strategies that
maximize their expected utilities, and they are playing strategies that are subgame
perfect. Strategies that are subgame perfect indicate that the strategies restricted to a
proper subgame form a Nash equilibrium for that subgame.”

Assumption 9. Every outcome of the game has a specific cost and/or benefit associated
with it. All costs and benef ts are greater than 0.

Assumption [0: War is a risky venture. Neither side knows with certainty that it can win
in the event a conflict arises. 55

 

5'Discussions of all game-theoretic concepts used 111 this dissertation can be found 1n
Morrow (1994b) and Gates and Humes (1997).

52 In later sections of this chapter I relax this assumption and make information
incomplete for player A.

53 Subgame perfection is an equilibrium refinement on Nash equilibrium, the central
equilibrium property in game theory. A Nash equilibrium is defined as a pair of

strategies 8; and Sj that are best replies to each other (Morrow, 1994, 80). There exists no
incentive for either player to deviate from its equilibrium strategy. Subgame perfection
refines this concept of Nash equilibrium somewhat, and makes certain equilibria
incredible; that is, not possible given a specific equilibrium path.

54 See Table 4.1 below for a description of all the benefits and costs which comprise the
various payoffs. See Table 4.2 below for an explication of the specific payoffs for each
outcome.

5 War in this depiction, then, is viewed as the potential outcome of two lotteries, in
Which an actor must choose whether to accept some situation with a greater degree of
certainty, or take a gamble on improving his position by engaging in conflict. I capture
this risk in the payoff structure outlined later in this chapter.

84

or

Q

a- ‘
T
l
i
.

,. I

..
.. ,
t“
I -V\.
... ,

 

.
y
r
1 .,‘
'\ u.
.' ‘
‘ 9 m.
*\b
v to
l”-- o‘-
I.
.x‘
r
m»

‘ .
“a. A
' A
._..)‘\

.
a} u
v».

.,
t

‘V.
.t-.,

.1‘ I

".

Q.‘

. .
.5
s k

I‘
‘.
1.

 

 

 

4. 2.1 The Sequence of Moves

The model outlined above is not intended to be a literal depiction of international
crises. It does, however, capture the essence of sequence found in most international
crisess6 Below is a discussion of how the crisis unfolds, and the strategies that each
state can choose from as the crisis escalates.

Prior to the onset of the game, B and C form an asymmetric alliance (hereafter,
BC), in which B is a minor power and C is a great power defender. This decision to form
an alliance is not explicitly modeled in this game. Rather, I merely assume that the
alliance exists due to the common threat of A, and that each state entering into the
alliance gains security benefits (in the case of B) or autonomy benefits (in the case of C)
as discussed in chapters 2 and 3.57

Although the model is sparse, it captures the essential details of an international
crisis without becoming hopelessly immured in minute details of particular disputes. It

focuses on the opportunities, represented by the potential for a crisis given a contested

 

56 George and Smoke (1974); Fearon (1994a, 1994b); and Powell (1990) all depict
international crises as proceeding in a sequential manner. A notable exception to this

modeling trait is Snyder and Diesing (1977), who model international crises as games of
simultaneous choice.

57 Others, notably Morrow (1994), Sorokin (1994), and Smith (1995), model the alliance
formation process prior to the onset of a crisis. However, if one is interested in
examinln' g the impact that alliances have on the initiation and escalation of crises, then
we need to have the existence of an alliance prior to the crisis’ onset. The pieces by
Morrow and Sorokin mentioned above provide such a structure within their models.
However, Smith’s piece models the crisis scenario as the aggressor making an initial
decision to attack or not, and then having a third party join either the attacker or the
target. While this piece, along with Morrow’s, re-specifies and more accurately captures
the strategic interactions inherent in earlier work of Altfeld and Bueno de Mesquita
(1979) and Wu (1990), it is a model that explains a decision to intervene by a previously
non-allied third party. Therefore, it is a more apt description of the formation of a
coalition rather than an examination of a pre-existing alliance on the occurrence of war.

85

 

 

“a...
. l

‘41.»...

 

-ob‘ ..
r i
'- I at

u
.t

R.
~-... H
..

-4, .L‘

n»,

 

 

 

 

 

 

status quo, and the willingness of a defender to seek an alteration in the status quo, given
that its adversary has formed an alliance with a great power.

Of course crises may last longer than the initial two-stage crisis model depicted
above.58 The essence of a crisis is preserved in the model presented, however, and thus is
much more tractable when it comes to purposes of analysis. Having said that, I proceed
into a description of the moves that are undertaken in the context of this model.

Having established the existence of a BC alliance, the game begins. A makes the

‘fi..- -.

initial move, choosing strategy 5 from a set of strategies S, s e S, where S = {initiate,

 

~initiate}.59 If A chooses not to initiate a dispute against B, the game ends, with the
outcome being the status quo (SQ). The possible success (or failure) of extended general
deterrence in this initial move can be regarded as the first stage in the game.

If A chooses to initiate, the next move in the game belongs to B, the target of A’s
aggression. Extended general deterrence has failed, and we enter the second stage of the
model, which is an extended immediate deterrence encounter. B must choose a strategy 3

from a set of strategies S, s e S, where S = {resist ~resist}. If B chooses not to resist,

 

58 Fearon (1994b) and Powell (1990) present crisis models with infinite horizons; that is,
models in which there is no foreseeable end in sight. Such crisis bargaining models
include discount parameters that influence payoffs as the models are iterated over time.
Future work on extended deterrence via alliance will account for these discount factors
by conceptualizing a crisis as a set of iterations. However, at the initial stage of this
tlleory’s development, I find it more useful to confine the model to a two-stage crisis, and
keep the findings more parsimonious.

59 A brief remark on notation. Tilde (~) is used throughout this dissertation as the symbol
for “not”. Thus, ~initiate is read as “not initiate”. For discussing strategy sets, I use the
symbol 6, which means, “is a member of the set”. Therefore, seS merely indicates that a
strategy sis a subset of a larger set of strategies that are denoted as S.

86

"or
4

_.....-I

.....
t\J\

. .h-

 

'rJ.‘\
. it

\
‘.Id--

 

 

 

 

‘.v
'.
‘a
fit
I.
. -
'1
‘A
I.
.‘.
‘v
I
.l
l.‘
\.

 

then it accedes to A’s demand, and the game ends with B’s acquiescence (ACQB). If B
chooses to resist, then the next move in the sequence is left to C.

C must then choose a strategy 5 from a set of strategies S, s e S, where S =
(intervene, ~intervene}. Regardless of what strategy C chooses, A makes the final move
in the game. A must choose a final strategy 5 from one final set of strategies S, s e S,
where S = (escalate, ~escalate}. If C chooses to intervene on behalf of B, and A chooses
to escalate the crisis, then the outcome is a trilateral conflict (CNALL). If C opts for a
strategy of non-intervention. and A chooses to escalate the crisis, then the outcome is a

bilateral conflict between A and B conflict (CNAB). If C chooses to intervene, and A
chooses not to escalate, then the outcome is back down by A (BD). If C chooses not to
intervene, and A chooses not to escalate, then the outcome is back down by A without
intervention by C (BD"‘).(’O Following from this are more detailed discussions of what
each strategy entails.
lam

States initiate a crisis when they feel it is advantageous to make some alteration in

the status quo. Initiation takes the form of a demand, coupled with the use or threat of

 

’0 In this model, negotiation is not a potential outcome of a crisis. While some may take
issue with the absence of this outcome, many demands made in international crises are
non-divisible. For example, many of the demands made by the Soviet Union against the
United States during the Cold War were take it or leave it demands. All of the crises that
erupted over the status of Berlin, for example, took this form, with the ultimate Soviet
goal being the withdrawal of Allied personnel from West Berlin. Or, in the Cuban
Missile Crisis, the possibility of using negotiation to force the Soviets to remove their
missiles from Cuba was not considered an option for the United States. For a discussion
and analysis of the various Berlin crises, see George and Smoke (1974) and Slusser
(I973). The definitive text on the Cuban Missile Crisis remains Allison (1971). See
Fearon (1995, 389-390) for a further discussion of the indivisibility of issues in
international politics.

87

“‘T
a

 

N”;

«ygwgr
”one.

-.
I. .,
‘. a
by. ‘
-._
‘4.
s
V

q

..

s._.:‘

- .

l \
- .
n:

. \

 

 

 

.,
c
J; .
.‘1‘-'\
..
u
a
'l \‘.
."\
l
I
- I.
‘ x
t
-

 

force. Initiation of a crisis by an adversary indicates the failure of extended general
deterrence. Hence, I focus on any demand made by an aggressor towards a target, be it
military, diplomatic, or otherwise.

At any point in time, states are confronted with the opportunity to make demands
of an adversary, and to couple these demands with force. However, as the data in
Chapter 5 demonstrate, demands coupled with the threat or actual use of force are a
relatively rare event in the international system. Conflict is not the normal state of affairs
among states. Extended general deterrence has a tendency to be successful, largely
because a challenger has no incentive (opportunity) to challenge the status quo. This is
not to say that the challenger is satisfied with the status quo. It may be more than willing
to desire a shift in the status quo in its preferred direction, but not have the opportunity to
do so. Rather, it indicates that changing the status quo is too costly for the challenger,
rather than assuming satisfaction with current arrangements.

Eggs;

Once an aggressor has initiated a dispute, the target must make a choice. It can
acquiesce to A’s demands, or it can choose to resist. If B chooses to accede to A’s
demand, then the dispute is resolved in A’s favor. Of more interest is what occurs if B
chooses to resist, and refuses to acquiesce to A’s demand.

Refusal to acquiesce means that B is actively engaging A in an attempt to prevent

A from attaining its goal. Resistance to A’s demand can range from B lodging a
diplomatic complaint to responding with force on its own part. If B has an alliance

partner in C, then it naturally hopes that C will intervene on its behalf. Such is the

88

 

 

~. ~--

 

 

 

~—
h- r.
I
,_ . o.
.‘ ','
i.-“\
‘ c .
I, 4 ~

 

 

   

 

 

 

 

\
u

. I
\ I-
c

n

l u

t

 

 

‘V I

.‘t

\

l

l V'~
‘x

Is.

. .

‘\.4

 

purpose of alliances. If they fail at deterring aggression, then they serve as units to
aggregate capabilities in the event of an armed conflict.

By explicitly incorporating B’s choice into the model, the traditional manner of
examining general deterrence has been expanded. As F earon (1993, 1994a) notes,
traditional rational deterrence models have focused on a challenger and a defender. They
don’t incorporate a move made by the target. This omitted move by B, I argue, is
imperative for understanding the role that alliances play in dispute escalation.6| If B
refuses to resist, then C has no opportunity to intervene and prolong or settle the
dispute.62
Intervention

It should be noted at the onset that intervention can assume various guises.
However, they all share one common trait. They provide an indication of the defender’s
value for its protege. This value, as was argued in chapter 3, is largely determined by the
benefits that the defender gains from its alliance with the protege.

Intervention can range from implementing diplomatic or economic pressure on
an adversary, to outright provision of troops and military resources to fight alongside a

threatened protege. If a state is to be viewed as valuing its ally, then it will intervene on

(’1 Of course, this tendency to focus on the major powers is largely a function of how
international politics tends to be conducted. As Kilgour and Zagare (1994) note, the
minor power protége’ in extended deterrence relationships takes on the role of a pawn,
which is what the major powers are fighting over. The pawn, as any chess player Will
acknowledge, is sacrificed if it is not of sufficient value to the player who commands it.
The same scenario emerges in extended deterrence. If the defender does not have a large
enough interest in the protege, then the protege will be sacrificed to the aggressor making
a demand. The fate of Czechoslovakia in 1938 and 1939 is a prime example of a prote’gé
sacrificed because its defenders (in this case, France and Great Britain) did not deem its
survival to be that important.

89

 

.4.-.- .

 

\
.
l
. ‘—
,K' ‘
~...,
\
._,__I

 

 

 

 

 

its ally’s behalf. In the context of the model developed here, I assume that intervention
means military intervention. A defending state can undertake any of the other
aforementioned actions on behalf of its protégé. However, given the definition in the last
chapter of alliances, with their focus on security affairs, and the rationale for issuing
deterrent threats to prevent an attack, I feel that the focus on military intervention is
justified, although I recognize that it may be a simplification.

Escalation

Escalation in this context means increasing the use of force to one in which large-
scale armed force is used by A against B (and potentially against C as well). If the
initiation of a crisis indicates the breakdown in extended general deterrence, then the
escalation of a crisis indicates the failure of extended immediate deterrence. As defined
earlier, a crisis in this theoretical framework is defined as a situation in which A makes a
demand of B accompanied by the use or threat to use force. By escalating a dispute to
war, A is indicating its willingness to change the status quo, presumably in its favor,
through the use of large-scale violence against another state.

As was mentioned above, and is elaborated in more detail below, war is a risky
venture. A realizes that if it decides to increase the level of force it uses against B, it runs
the risk of fighting a war against B and C combined. The decision to escalate a dispute
to war, therefore, indicates that A is more risk-acceptant than risk-averse. There always
exists the possibility that C will intervene in a dispute, and this is a risk which A faces
when it decides to use a greater level of force in an attempt to attain its goals, and is

crucial to understanding the impact that an existing alliance has on the possibility of war

h

 

62 If C chooses to engage A after B’s acquiescence, then that becomes another model,
Wthh is not the focus of this study

90

..,.....

 

 

 

 

 

 

.',‘ 0-
‘3.
. ,_
..
.- ~
’ ~
I
‘ n-
""o I.
‘ m»
-
,,-. _
'7».
..
.
-,
n
‘ n
,1 .-
-.
4-,
I
._ ,
'I
- D I
j. - .
’\
n,

 

occurring. A never knows for certain how much value C places on its relationship with

B, and must infer from C’s actions prior to and throughout the dispute process what the

range of possible outcomes are.63

4.3 Outcomes, Costs, Payoffs, and Types

Having outlined the assumptions that place constraints on actors within the model
being developed, I now focus on the outcomes of the game. Every outcome of the game-
theoretic model explicated above has a specific payoff associated with it. These payoffs

are comprised of the benefits (if any) that a state gains from the outcome less the costs

 

 

that the specific outcome incurs. Below I discuss in substantive terms what each
outcome means. I then discuss in some detail the various benefits and costs associated

with the game. I then calculate payoffs for each outcome of the game, with special

attention being paid to the type of player.

4.3.1 Outcomes of the Game

There exist six distinct outcomes of the game specified above. In turn I elaborate
upon what each one means, and provide a substantive example from the set of crises
mentioned in the introduction to this chapter. This helps provide a link between the

theoretical model and actual empirical events.

Status (SQ)

As elaborated upon in chapter 2, the essence of deterrence is to prevent hostile
aggression from an adversary through the use of threats. In the case of an asymmetric

alliance, it should be recalled that the great power in the alliance is extending its

*

63 . . . .
_ This uncertainty lS captured on the more complex model below, where incomplete
information is assumed.

9l

deterrence threat to a minor power. Successful deterrence implies that the potential
aggressor (A in the model) realizes that C has made a credible commitment to B via its
alliance agreement, and that the likelihood of C intervening is quite high. The possibility
of facing such a strengthened target may force a potential aggressor to not initiate any

aggressive actions, since the potential costs are viewed as being drastically high.

Deterrence succeeds, and the status quo prevails.

Such a scenario emerges after 1961 in Western Europe. The United States, acting
through NATO, had emerged from the Second World War as the defender of Western
Europe in general, and of West Germany specifically. The Soviet Union challenged this
position in the early days of the Cold War quite regularly. Of particular interest to the
Soviets was the issue of divided Germany, and especially that of divided Berlin. After a
number of attempts to force the United States and its allies from Berlin, the Soviets
erected the Berlin Wall in 1961. The Wall, in addition to becoming a symbol of a
divided Europe, also symbolized the end of Soviet aggression towards the West on the
European continent. This potential aggression was officially quelled in December 1972,
when the Soviet Union and the United States signed a treaty regarding the territorial

status of Berlin and recognizing its existing boundaries."4

Ac uiescence b B AC _)

Oftentimes, a target will know in advance if its ally will not intervene on its
behalf. Or, alternatively, the issue at stake is of such little importance that resistance,

and the possibility of conflict occurring because of it, is too costly for the target state to

M

64 See George and Smoke (1974) and Hanrieder (1989) for greater discussions of the
American role in West Germany after 1945.

92

 

 

 

~o'n
4
\
as”.
q-
n vn
& . l
. l
. .. l
v
- ‘. r
an H
w
'h r.
on
.. .
v" d
. . .

 

 

bear. In such circumstances, it is often better for B to accede to A’s demands. In such a
scenario, then, B acquiesces to A, and C never even has the opportunity to intervene.

The 1939 crisis surrounding Czechoslovakia prior to the outset of the Second
World War is a prime example of such a scenario. A bit of historical background sets the
stage. In 1938, as Nazi Germany was gathering its strength, Adolph Hitler made the
decision to annex the Sudetenland portion of Czechoslovakia. The issue at stake was the

fate of ethnic Germans residing in this territory. France was linked to the Czechs by a

 

treaty of alliance, with a pledge to intervene if Czechoslovakia’s sovereignty was ever

 

 

 

 

 

endangered. (Great Britain was also a defender of the Czech nation, although not as a
formal alliance partner). Unfortunately for the hapless Czechs, the Great Powers gave
away the Sudetenland at the Munich Conference of 193 8, in exchange for a German
promise of peace. Scarcely was the ink dry on the agreement than German tanks were
occupying the remainder of Czechoslovakia in early 1939.65

The Czechs appealed to their allies for assistance, but none was forthcoming.
The alliance failed, and the stage for the Second World War was set. In the face of
superior military strength, and without assistance from allies, the Czechs had no choice
but to acquiesce, first on the issue of the Sudetenland, and then on the rest of their
country. Germany saw an opportunity to act, given the appeasement strategy followed by
the British, and acted upon it. The willingness to intervene was lacking on the part of the

French and British, and Czechoslovakia became the first major victim of Nazi aggression.

M

65 See Schweller (1993) on the inability of the British and French to deter German
aggression prior to the onset of the Second World War.

93

i
u.»
u . t
, I
.. I
o» .
. .'
y.

kl: I,

up ~ ,_
Q
i.

t v
x

‘.
1;”

t
\.

 

 

I..
. ‘ ’
~.
t
1
."
-.
l
1
-
‘-‘
eh,“
. I.
n

 

 

BgclLDown by Aggiven C’s intervention (EDA)

If A makes a challenge against B, then extended general deterrence has failed.
Our situation then becomes one of extended-immediate deterrence. If B resists, then the
choice falls on C as to whether it should intervene or not. A good example of this
phenomenon comes from the Serbian-Bulgarian War of 1885.

As noted previously, Serbia was linked to Austria-Hungary by a defense pact. in
which Austria pledged to come to Serbia’s assistance if she were attacked by an outside
power. Serbia actually launched an offensive war against Bulgaria in the late autumn of
1885, but soon found itself on the losing side.66 Bulgaria, sensing the possibility of
territorial gains, massed on Serbia’s borders and threatened invasion. Serbia pleaded for
assistance to its Austrian ally. Given its vested nature in the stability of an independent
Serbia, Austria mobilized and threatened to intervene militarily. Through this action she
was able to spare Serbia from invasion and defeat at the hands of the Bulgars.67
Back Down by A, given C’s non-intervention (BD*_A)

Of all the potential outcomes in the model, this is the only one that does not have
a substantive example to describe it. If A is willing to initiate a dispute against B,
knowing prior to the crisis’ outset that a BC alliance is in existence, then it is not rational

for this outcome to occur. 1 demonstrate this mathematically below, in the section on

equilibrium analysis.

W

66 The parallels between the Serbian debacle in 1885 and the Turkish debacle during the
Crimean War in 1854 are remarkable, in that both states launched offensive wars and
were later obliged to appeal to outside assistance in order to survive.

67 See Blainey (1989) and Langer (1950) for more details on the Serbian-Bulgarian War,
Which, incidentally, is the only major war fought on the European continent during the
decade of the 18803.

94

.‘.‘
q
,..y.u
_.———-

 

"". at

,-
, 1"
' as.
o
'1
A
‘
t _u-
. “
u
-
4
a,

 

 

 

 

-,
..
‘ ,
A
s
.“
'\.,
. '\.
.
c
v
s

 

Bilateral Conflict (CONLB )

A dyadic conflict emerges when a challenger initiates a dispute against a target,
the target resists, and the defender does not come to the aid of its protégé. To reiterate
Saborsky’s ( 1980) claim, this is a common recurrence in international politics for the
period 1815—1969. In the following chapter this claim will be examined in the extended
deterrence context; right now it is sufficient to state that this assertion tends to be
accepted as conventional wisdom within the alliance research community.

A good example of abandonment by a major power ally comes from the Sim-
Vietnam War of 1978-1979. As Sino-Vietnamese relations began to deteriorate in the
19703, Vietnam began to look elsewhere for external security. Given the Sino-Soviet
split, and the attempts by the Soviet Union to “contain” Chinese influence in Southeast
Asia, an alliance between Vietnam and the Soviet Union was formed in November 1978.
Angered over Vietnam’s invasion of Cambodia, China invaded Vietnam on February 17,
1979. The Soviet Union, although allied with the Vietnamese, failed to intervene in any

significant manner, and the Vietnamese were left to conduct the war on their own.68

Trilateral Conflict (CONLLL)

A multiparty conflict emerges when a defending state decides to intervene
militarily on behalf of its prote'gé, and the aggressor continues to press its attack anyway.
This is the ideal for B, for if it is going to be attacked, because it can count on its ally’s
support. Such a scenario led to the onset of war in the autumn of 1939.

In the aftermath of the Czech debacle, France and Great Britain began to realize
the seriousness of the Nazi threat in the center of Europe. France strengthened her ties to

M

as See Horn (1987) and OrganSki and Kugler (1980) on the Sino-Vietnamese clashes.

95

and
.. ‘IV
5.- —N‘
. i 9.
.,. .4...
a u-
u .
n w
, ¢
--vv

-0...

///

//

 

Poland, and the British signed a treaty of alliance with the Poles as well. In addition to
this treaty, the British made public statements mentioning their intention to intervene
militarily if Poland’s sovereignty were compromised. In spite of this, and possibly
believing that the British and French commitments were not Credible, Germany attacked

Poland on September 1, 193 9. France and Great Britain declared war on Germany, and

the Second World War began.

4.3.2 Costs and Benefits

Having discussed the strategies and outcomes that comprise the basic model, I
now turn to a description of the costs and benefits that make-up the payoffs associated
with each potential outcome. I first outline the potential costs and benefits, and detail
their implications. From these discussions I formalize payoffs for the outcomes of the
model. Finally I delineate the different types of players that exist.

Each action taken (or not taken) in this model has a variety of benefits and/or
costs associated with it. These six terms combine to make up the payoffs that are

associated with each outcome of the model. Table 4.] summarizes these costs and

 

 

 

 

 

 

 

 

benefits.
TABLE 4.1
Terms in the Model
Terms Definition
x. Value of A’s demand for i, ie { A, B,

t—¥ C}(zero sum between A and B)
E c. Physical cost of conflict for i, ie {A, B, C}
__ d Domestic political costs

mm Security benefit to B/invested cost
__ undertaken by C

yC/B Autonomy benefit to C/cost to B for
M alliance
__ w, Value for conflict for i, is {A, B, C}

 

 

 

96

 

.. a 3.-

‘»“".~l

\ \
...i\ o- .

. x
7' -
4 J
' ADI-t
. t
1 r
«and
‘ >
‘ *th
‘1'». .
“7“

 

 

 

 

4
‘fi
\1‘ '_‘
y
n
..

 

 

 

 

D
u.
I
v'.
‘.
‘~
'.
\
~ -
't
v

Now that the terms that comprise the potential payoffs have been identified, let us
now turn to a more detailed discussion of what these terms mean.69 The first term to be
considered is x. This corresponds to the value of the demand that A makes of B. It is the
issue at stake for each of the actors that are a party to the criSis. Thus. if A is able to gain
its demand, the status quo is shifted in its favor. B and C, on the other hand, witness the
status quo as being shifted detrimentally against their desired resolution of the status quo.
For A and B, if A makes a demand and B gives in, the result is a zero-sum solution, in
which A’s gain is 8’3 loss. In the case of C, it may not lose all of its benefits from the
alliance if B acquiesces, but it does stand to lose some leverage. How much it values the
demand, of course, will determine to a large extent the actions that C undertakes.

For example, consider the 195 8 Berlin Crisis. The Soviet Union demanded that
Berlin be made a free city, and demanded that the other remaining occupying powers
relinquish their control over the divided city. This would have placed the Soviets at a
tremendous advantage in the center of Europe, with the United States losing valuable
bases and presence in the heart of Germany. Having committed itself to containing
Soviet aggression and ideology worldwide, and having become involved in the Korean
War because of this pledge, the United States could scarcely afford to lose face in front of
the Soviets in Western Europe. American influence, both in Europe and on a global
scale, would have been jeopardized.70

__

69 . . . . . .

Means of operatronalrzrng these terms are detailed in chapter 5, when I discuss the
pr0positions to be tested, and the methods and variables used to conduct the empirical
analysrs of this dissertation.

70 George and Smoke (1974), as well as Slusser (1973) provide excellent accounts
regarding the American commitment to Berlin and its implications for American foreign
policy outside of Europe.

97

.rg‘i

 

1..., t.
a

 

.4. .
u .

n _ _;
n, i
.

. . ‘
"ws

   

 

 

.o
4 .
. v I
>,, A

I. L
_ ,
‘s-
u
.
\
..
r
\“
v
a,
. \ -
4“
e
u
.,‘
s
I»

 

'-..
. .4
‘.,.

~-

I.

n
. v-

u
.,

~

 

If B and C are both joined in an alliance, both states can mutually benefit. As
outlined earlier in this study, great powers obtain autonomy benefits from allying with
minor powers. These smaller alliance members gain security from their defender. I
denote autonomy benefits in this model as y. This is the concession that B makes to C in
order to gain alliance with C. It may require a coordination of some policy issue, or it
may be permission for C to maintain some form of military presence on B's territory.

B also gains from the alliance relationship. I denote the security benefits that B
receives from C as m. These benefits serve as a signal from C to B, and provide a dual
function. First of all, they allow potential adversaries to witness the importance that C
attaches to B, since they are costly investment signals. Secondly, in many instances,
these investments help prepare B for potential conflict, if an adversary decides to press an
issue. This signal can be items such as arms transfers, statements of intention to protect a
protege, or even tightness of the alliance ties. Both of these terms function as signals that
can be witnessed by a hostile aggressor. This trade-off helps shape the perceptions of an
antagonist, because it provides an indication of the value that a defender places on a
prote’gé’s well being.

Again Western Europe in the aftermath of the Second World War provides a
telling example of this phenomenon. The United States was able to use its economic and
military power to help rebuild the war-shattered societies of the West after 1945, and use
its membership in the NATO alliance to work at containing and deterring the Soviet
Union, especially through its presence in West Germany. Western Europe, on the other
hand, was able to gain military and economic good that were needed to rebuild their

economy and strengthen their military might. The United States gained autonomy by

98

investing in Western Europe. This allowed the United States to maintain a presence on
the Continent, and the NATO allies were able to gain security under the American
nuclear umbrella.

We now come to the last three terms that can potentially comprise payoffs to the
model: c, d, and w. The first two terms are specifically cost terms. Costs of conflict
involvement are represented by c. This cost increases as a crisis escalates. Every action
undertaken by a state in times of crisis involves some level of cost that deals with the use
(or potential use) of the armed forces. When A makes a demand, it often accompanies it
with a use (or threat) of force. If B and C choose to respond, it is often with a threat of
military action as well.

In the domestic sphere there exist costs for any military venture as well.

Domestic political costs, denoted as d, often act as constraints on decision-makers. While
I assume, as do many others (e.g. Fearon, 1994a, 1994b; Hart and Eyerrnan, 1996) that
domestic political costs are imposed if a foreign policy venture leads to conflict, or if no
actions are undertaken on behalf of a beleaguered ally, I argue that this is not the only
realm in which domestic considerations come into play. In many countries, notably those
with competitive political regimes, the decision to undertake a foreign policy venture,
such as forming an alliance, is subject to domestic political bargaining as well as
international negotiations. This is particularly evident in democracies such as the United
States, where international agreements are subjected to ratification by the national
legislature. Such costs must be considered as an evaluation of the status quo, as well as

the potential costs that erupt by becoming involved in a militarized crisis.

99

o
. u
', J,
-..e..-
.

-. 4
.A
..... h

.

‘l
a.
4
“'.9
s4,» ,
. .

.I-v

u.....

 

.2.
“PM.
‘ .
d-t
‘ 'c‘
.,.
E“. I.
a it
o
I u .
' .
w.
" .
‘p
r
-...‘_‘

 

 

' 17)

Particularly in democracies, but in authoritarian regimes as well, there is always
the fear that a major foreign policy defeat will cost an office holder his or her position
(Bueno de Mesquita, Siverson, and Woller, I992; Bueno de Mesquita and Siverson,
1995). While the role of domestic politics in international affairs, as discussed in
chapter 3, is one that has not been overly analyzed until recently, it plays a much greater
role than traditional realists or neorealists acknowledge. An example from great power
interaction demonstrates this quite nicely.

In the aftermath of the Treaty of San Stefano and the Congress of Berlin in 1878.
the Balkan Peninsula was reshaped, and Bulgaria, Romania, and Serbia emerged as
independent states. Serbia, desperate for security in the face of potential Bulgarian
irredentism, sought alliance with Austria-Hungary. Austria’s benefit from this alliance
was twofold: economic and military. Economically, Serbia provided Austria with
favorable trading circumstances. Militarily, alliance with Serbia permitted the Austrian
monarchy to actively engage Russia, and permit the eventual annexation of Bosnia-
Herzegovina.

The treaty of alliance with Serbia, although attempts were made to keep its
provisions secret, was quite problematic for the Austrian government. The Hungarian
portion of the Austro-Hungarian empire was overly concerned with the growing Slavic
influence within the empire’s boundaries. There were concessions that had to be made,
and this rift within the empire became known, both to Bulgaria and the Russians, who
were eager to exploit this domestic strife. The annexation of Bosnia did not take place at

this time primarily due to this conflict of interests, and the Serbian-Bulgarian war of 1885

100

 

mt. .,.r

-‘1- .1.

 

 

I~.‘

 

 

erupted in large part due to the Bulgarian impression that Austria was constrained at
home, and would not assist her Serbian ally."

The final term to be discussed is w, the value for conflict. The initiation of a
crisis entails the possibility that the crisis could escalate and lead to armed conflict. Each
state in the model holds some value for possibly engaging in such activities, and this
value is shaped by a variety of factors, as will be discussed below. For now it is
sufficient to state that states hold different values for going to war, based upon what they
hope to accomplish from engaging in fighting. These values are based largely on a
state’s military capacity and its willingness to engage in risky behavior to achieve its
goals.
4.3.3 Discussing Payofls and Types

Having described the possible outcomes of the model in some detail, and
explicated the terms that comprise the payoffs associated with the outcomes of the model,
I now turn to an elaboration of the payoffs associated with each of the outcomes. Payoffs
are defined as a player’s utility for the outcome of a game (Morrow, 1994b, 351). They
are a combination of the benefits and costs associated with each outcome. Table 4.2

outlines the payoffs each state gains from each potential outcome.

 

7' Further discussions of this issue can be found in Langer (1950, 323365).

101

TABLE 4.2

 

 

 

 

 

 

 

 

 

 

 

PAYOFFS FOR THE MODEL
State SQ ACQB BDA BD'I‘ CNAB CNALL
A 0 xA xA-c-d xA-c-d w-c-d w-c-d
B xB +m3- -xB-d x3 +m3- x3 +m3- w-c—d w-c-d
YB-d yB-c-d yB-c
C X), +y3- -xc xA +y3- xA +y3- -d-m w-c-d
mB-d mB-c-d mB-c-d

 

Having outlined the benefits and costs that each state potentially encounters in the

model, and specified the generalized payoffs each player holds for the specified

outcomes, a discussion of players’ types is needed. While states may be similar in their

desires for security and survival (Morgenthau, 1985; Waltz, 1979), or,more generally

viewed as utility maximizing entities (Bueno de Mesquita, 1981; Morrow, 1997), they

possess different attributes in how they pursue these goals. To paraphrase George Orwell,

all states are created equal, but some are more equal than others.

Two factors figure prominently in determining a player’s type: its relative military

capabilities vis-a-vis its adversaries, and its value for changing (or maintaining) the

current resolution of the status quo. In the following sections I address the role of

capabilities in deterrence situations, and the role that risk plays in determining values for

different issues. It is these two concepts, I argue, that determine what types of players

l02

 

exist in the model. Following from this discussion I proceed to outline three distinct
types that player A can assume, and the two types that player C can assume.72
4.3.4 The Role of Capabilities

Much was made of the role of capabilities in chapter 3, and for good reason.
Capabilities, or power, have long been central to the study and conduct of international
affairs. Traditional balance of power theorists (Morgenthau, 1985; Walt, 1987; Waltz,
1979) have long argued that power parity leads to peace. Others, notably adherents of
power transition theory (Organski and Kugler, 1980; Kugler and Lemke, 1996) maintain
that power parity is a path to conflict. Only power preponderance, according to this view,
leads to stability and peace within the international system.

Capabilities are a means of measuring the power that a state possesses. As was
outlined in chapter 3, relative capabilities lie at much of the heart of both theories of
alliances and theories of deterrence.73 In the model developed here, I denote relative
capabilities as kl, where ie I, l = {A, B, C}. If alliances are means to an end, namely, to
deter enemies, then capability aggregation has a justifiably prominent place in the

discussion of extended deterrence via alliance.74 However, capabilities are not the only

 

72 In keeping with most studies of extended deterrence (Fearon, 1994a; Huth, 1988; Huth
and Russett, 1984, 1988, 1990, 1993), it is the capabilities of the aggressor and the
potential defender that are considered most important. A notable exception to this is
Sorokin (1994), who examines the capabilities of the target and the role they play in
determining whether or not to form an alliance in the first place.

73 Singer, Bremer, and Stuckey (1972) define relative capabilities (or power) as a state’s
measure of capabilities in relation to its adversaries. Absolute power, on the other hand,
is concerned with how much total power a state has, without being concerned with its
adversary’s power. See also the recent work on the relative versus absolute gains debate
(Griecol988; Powell 1991; and Snidal 1991).

103

factor that influences deterrence success or failure, as will be discussed in the following
section.
4.3.5 Discussing Risk

If military power were the sole factor in determining deterrence success or failure.
we would never witness asymmetric conflict, or conflict initiation by weaker powers
(Paul, 1994). However, as much of history demonstrates, this is not the case. In a recent
theoretical study, F earon (1994a) demonstrates that it is not mere capability ratios that
determine extended deterrence failure. In fact, he shows that states with fewer
capabilities than the combined attributes of an alliance can actually initiate a dispute and
gain their demands. They merely need to make a demand on an issue that is of low
salience to the defender.

As F earon’s findings demonstrate, capabilities are not the only determinants of an
extended deterrence encounter. Aggressors need to know how much a defender (and
target) value a certain issue as well prior to challenging it. Failure to determine a
defender’s value for an issue can lead to an unwanted conflict. Thus, an attacker’s
decisions in this model must often be made under conditions of risk.

In their relations with one another, states continually encounter situations with
uncertain outcomes. With such uncertainty being present, states must often make choices
that involve some amount of risk. Risk inherently involves making decisions in which
outcomes are not known ultimate clarity.

In initiating a crisis, an aggressor state (A) may not know for certain if a target

state’s ally (C) will intervene. A may feel confident in being able to gain concessions

 

74 . . . . . . . - . .
When usrng relatrve capabilities in this study, I am talking about a combination of
immediate, short-term, and long-term capabilities that a state can call upon in time of

l04

 

. v.1
If... w
1
_ t
.
.h..
In r
, r.
‘51 .»
... ,
-

u 7"
.. . ..
t

.r.
,‘1

.0-

h~1iu

um

 

 

 

 

 

 

 

.
’t.
n
ft
‘x
r»
.1
1
p
‘1'
.
-

 

 

 

from B if B fights alone, but may feel differently if C decides to intervene on B’s behalf.
This decision is often made, therefore, under conditions of uncertainty. Elaboration on
the role that uncertainty plays in deterrence/crisis situations will be detailed in later
sections of this chapter.

Different states assume different attitudes toward risk. These can be termed risk
propensities. “Risk propensities capture the fact that different decision makers may make
different choices when faced with the same set of alternatives solely because of their
attitude towards choosing options with probabilistic outcomes” (Huth, Bennett, and
Gelpi, 1992, 482).

War, as Morrow (1987) discems, is an inherently risky choice. Within the context
of militarized interstate disputes, an initiator may never know with ultimate certainty if a
third party will intervene in the conflict. I view the decision by A in choosing to initiate a
dispute or not as being. one made under conditions of risk. Since war is a distinct possible
outcome of a dispute, but the outcome of the war is probabilistic in nature, A must choose
whether I) it is willing to escalate a dispute to war, and 2) it is willing to fight against the
combined capabilities of an alliance. The term win the payoff structure to the model
captures this risk attitude on the part of state A, and it is also an indication of how willing
C is to fight to save its ally.

The value for conflict, I argue, can be derived from the desire to change the status
quo and the military might k that each state possesses. In regards to power, as I noted
above, I am concerned with immediate, short—term, and long-term military power. Such a
distinction between capabilities in regards to deterrence encounters is not new. Many

leaders, when considering war initiation, have factored in the power of their adversary.

k

crisis.

l05

 

...I- I-)
-.u
1‘. ‘4
".. a.»
.' ,-
- 1 ,
A
c.» .—
~» 9
.A .4'
my...“
> v -_
~>~ . \

 

 

 

 

t
t.

'F
.b"-

 

I“. u
._1."
'0
-
\

 

 

 

\.

 

 

 

The Japanese in the Pacific theater of the Second World War, for example, recognized

that they could not win a long-term war with the United States. Their h0pe had been to
win decisively at the onset, and then be able to come to terms with the United States that
were favorable to their long-term goals. Similar thinking also occurred in the European
theater of the war. Hitler realized that he could quickly overpower Poland, and the rest of
Europe. Although his generals noted that the combined strength of the British and

French could defeat the German army in the event of a protracted war, Hitler was willing
to use blitzkrieg tactics in an attempt to win a short war decisively.

At issue for a potential aggressor, then, is to accurately discern what type of
defender it is facing. Thus, below I describe the different types of A that exist, based
upon the value A holds for w. This value, as mentioned above, is based upon A’s
capability ratios with BC, which is an indirect measure of A’s willingness to run risks to
change the status quo.75 I assume that A’s capabilities are always greater or equal to that
of B. I make this assumption because, as Sorokin (1994) demonstrated, if B is more
powerful that its adversary, then it does not require the additional security benefits of
alliance. I then proceed to discuss the two types of player that C can assume, based upon
the value they hold for w. In regards to C, I maintain that w is based upon the value it
holds for the autonomy benefits it receives from its alliance relationship with B. In the
following section regarding preference orderings, I link types with values for the various

payoffs.

75 Carlson (1998) uses capability ratios as a measure of the costs that states are able to
absorb in a conflict. The reasoning she uses and that put forth here are similar. The
greater the ability of a state to absorb costs, the more likely it is to use force in an attempt
to alter (or preserve) the status quo.

106

 

 

Plgyer A

Player A can be ofthree types. Formally, let tAe TA, TA ={1, 2, 3}. lftA = 1, it
has greater relative capabilities than BC. Thus, kA > k3 > kc. With such an advantage,
then, A prefers to make a demand x than remaining at the status quo. With higher levels
of military power, it has a higher value for the possibility of conflict.

If tA = 2, it has capabilities that are less than the combined capabilities of the
alliance BC. Thus, kA <k3c. By the assumption made above, kA > k;;.76 _ Therefore, if A
knows that C will intervene for certain, it will prefer the status quo to x. The costs are too (a
high. However, if A believes that C has little interest in the issue being challenged, it
will still prefer x to the status quo, even if its capabilities are evenly matched with those
of B. Thus, we see here that capabilities may not be the only indicator of success in
determining crisis outcomes. Success or failure also hinges upon determining each
state’s utility for the specific outcome.

If tA = 3, then it has capabilities that are equal to those of BC.‘ForrnaIly, kA =k3c.
If capabilities are evenly distributed, then decisions are made under conditions of power
parity. This implies that it is more than capabilities that determine the strategies that A
will choose; rather, one must focus on its utility for the status quo versus its utility for x.
m

Player C can assume two types. Types for C are determined by the value they
hold for y with respect to the other costs they must face in maintaining relations with their
minor power ally. The first type is tc = 1. This type of defender is places a strong value

on its relationship with its protege. It has a higher value for its protege, which can be

-—k

107

evidenced by the strength of signal it sends to its ally. Formally, for this type of
defender, y > m. A strongly committed defender will strictly prefer the status quo to any
other outcome.

If tc = 2, then it is a weakly committed defender of B. In such an instance it is
willing to run the risk of letting B fight A alone, if it comes to conflict, rather than
intervening. Formally, y _<_ m. It does not hold the status quo in very high regard, at least
in respect to its relationship with its ally.

4.3.6 Preferences

Now that the types of players have been delineated, our attention is turned to the
preferences that each type of player possesses. Although the payoff structures coupled
with the types of players have the potential to lead to a dizzying amount of different
preference orderings, I place restrictions below on the preferences that each type of state
may hold. While modeling states’ types in such a discrete manner may be cumbersome, I
believe that it captures the essence of actual relations between states.77 Prior to outlining
the preferences of each state, however, a few words about preference orderings are
necessary.

Preference orderings, as Morrow (1994b), and Gates and Humes (1997) point out,

must meet three criteria for them to be valid. Since making choices based upon one’s

7’ The special case of k», = k3 is discussed below in the equilibrium analysis.

77 It had been suggested that I model the types of players as continuous types rather than
as discrete types. While there do exist some arguments for such a modeling strategy, it
should be recalled that the purpose of the model here is to shed light on substantive issues
surrounding extended deterrence via alliance. I fear that by making the mathematics
more complex may hinder an understanding of the substantive results that I derive from
the models.

108

nv,','1

rub.‘
I
rt 1’1
al.-41v
A
~-.
v32
VJ.- \
. . ,
.11
«um.-
\
' D
|
r r .
4
a
30477;,
is“

preferences for outcomes lies at the heart of rational choice modeling, it is imperative that
we understand the situations under which preference orderings can be viewed as logical.

First of all, preferences must be complete. By this is meant that actors prefer one
outcome to another, or they are indifferent (Gates and Humes, 1997, 8). Secondly,
preferences are assumed to be fixed. This means that as the game progresses, actors may
not change their preferences. They can, however, choose to alter how to arrive at their
most preferred outcome. Finally, preferences are assumed to be transitive. This means
that if an actor prefers A to B, and B is preferred to C, then A is preferred to C. This
prevents cycles of preferences from arising, which can give rise to nonsensical behavior.

I now discuss in some detail the preference orderings of each state. I then derive
the equilibria of the game, and discuss its implications. Table 4.3 outlines the

preferences that each state holds.

TABLE 4.3

Preference Orderings of Each State (Ordinal Outcomes)

 

 

 

 

 

 

 

 

State and Type Preference Ordering j
A=1 ACQB>CNAB > CNALL > SQ> BDA >BD*,T]
A=2 ACQB > CNAB >SQ > BD> CNALL > BD*A ]
=3 ACQB > CNAB > CNALL >SQ>BD > 3D“?
‘T B SQ> BDA> BD*A> CNALL >ACQB>CNA97
=1 SQ> BD*A > BDA>CNALL> CNAB>ACQB j
‘“ =2 SQ> BD’tA > CNAB > BDA> CNALL>AC037

 

 

109

. . a
p_ “0‘

.iLAV-

Ir: 7‘ "

C~Ut§d

-.,..- t.

we“ n...

P T'
...A.,

 

‘..
4
ll

1.
A“?
hug“
1 u .
kg \
~

.t.
.\‘i‘

 

 

.
I.

.
”“1
4.

‘5 .
b

Let us examine each state’s preferences in turn. Although the above table outlines
the preferences in ordinal form, that is for ease of presentation. Given that preferences
are at the heart of rationality, it is necessary to determine the course of action that each
state would prefer, given the above situation. We do so by carefully delineating what
cost and benefits terms go into each different preference ordering.

Player A (IA=I):

If A is a type 1 player, it would prefer for Player B to accede to his demands,
given that he has initiated a dispute in hopes of altering the status quo in its favor. If A
were comfortable with the issue at hand, it would have no reason to make a demand of
3.78 If conflict is to be an outcome, he would prefer it to be a bilateral one. It is easier to
defeat a single adversary if conflict is the outcome than it is to defeat two states fighting
alongside one another. The status quo (SQ) is preferred over backing down in both of
its guises. However, A prefers a trilateral conflict to backing down. If A chooses to
back down, it incurs costs that it feels are unacceptable, given that its capabilities are
greater than those combined of BC, kA >k3c.
Player A (IA =2):

If A is a type 2 player, it would once again prefer for Player B to accede to his
demands, given that he has initiated a dispute in hopes of making some gain for himself.
If conflict is to be an outcome, he would prefer it to be a bilateral one. Once again, it is
easier to defeat a single adversary if conflict is the outcome than it is to defeat two states
fighting alongside one another, especially if, as was assumed above, kA>kB. If tA=2, it is

73 . . . . . . .
Recall the above assertion that international disputes occur due to dissatisfaction by
one (or more) parties with the current resolution of a specific issue.

110

1 1.
v
c
\‘,‘i
,—
1 1.
~ 1
.1]- tl
rv;o-‘
I‘il.tt.k
p
v v
.11.
' ‘t
, .

   

l4 1

1.,7
5

 

 

 

willing to make a limited probe against B, and see what action C will undertake. Thus, it
is willing to take this chance and back down if C will intervene. It prefers backing down

(BD) to the possibility of trilateral conflict, the status quo, or the utility of backing down
when it knows that C will not intervene.

Player A (tA=3):

If A is a type 3 player, it would (as in the other two cases) prefer that B give in to
its demand immediately, and save the costs of further crisis involvement. And if A
realizes that B is going to resist, he would rather fight against B alone, especially since
A’s capabilities are greater. However, if A is willing to change the issue and challenge
B, knowing full well that B is allied with C, then it implies that x > 0 (status quo). Thus,
trilateral conflict is preferred to the status quo. Backing down in either of its possibilities
is the least attractive outcome, because costs have been incurred with no appreciable
benefit.

Player B:

Obviously, Player B would prefer that Player A left him alone, and that the status
quo (SQ) prevails. However, if an attack does come, B would prefer to have the dispute
end in BDA. This allows B to realize that its ally was willing to come to its defense. B’s
next preferred outcome is BD*A. Even if its ally C chooses not to intervene, B would
rather witness A’s backing down. However, if conflict arises, Player 2 understandably
would want its ally C to fight on his side. This follows from the assertion above that
alliances oftentimes form for security purposes. Nations join in alliances to increase their
capabilities, and to prevent another nation from attacking them. If this is not to be, then

acquiescing to Player A's demands is preferred to fighting a conflict alone against an

111

aggressor. Although B knows with certainty that by giving into A’s demands it will be
altering the issue at stake, but not to its advantage, it is better than fighting a bilateral
conflict. By fighting, B has the chance that it may lose the conflict, and suffer the new
imposed status quo. In addition, it must absorb the costs that the conflict imposed upon
it.

Player C (If: I):

By functioning as a defender in the alliance relationship between itself and Player
B, Player C is making a commitment to Player B, that it will defend its ally in case it is
attacked. Given that disputes are costly, and have the potential to change the status quo,
a highly committed C would prefer to have the status quo maintained. However, if B
were challenged by A, C would prefer to have A back down without having to threaten to
intervene. A threat to intervene means that C must mobilize its forces, and incurs some
costs. Hence, it would prefer this outcome ( BD*A) to the possibility of conflict.

If conflict is to be the outcome, C would prefer a trilateral conflict. C, being a
strongly committed defender, realizes that the outcome of such hostilities may be in its
favor, and the status quo preserved. (Not to mention that A as an adversary may no longer
pose a threat). A bilateral conflict between A and B is the next preference, because B
may win the conflict on its own, and be able to maintain a status quo which is favorable
to C. Finally, C would least prefer the acquiescence of B to A, because the benefits it
received from the alliance may be severely curtailed, since B’s acquiescence ascertains

that B will lose its value for the issue at stake with a probability of 1.0.

Il2

r ‘.
.mn.

.t‘.“

In.

3"! ,,

u...,_ 11
I, W

J .
...‘_ \-

 

" .«J

“L”,

 

 

..
. I
.a.
'-
‘
\ "u

 

I ‘ I
“"hl
.. ‘
~1 _.{
“\,t
3».
4“.
h

 

 

‘M‘
"i 4
DA ‘
'5
U" "
b

 

Player C (tc = 2):

The case of a less-committed defender is a novel one, for traditional theories of
alliances do not predict such a type. However, as was discussed in some detail above,
empirical evidence indicates that states are, more often than not, unreliable in regards to
their alliance commitments. A player C of this type wishes, as does its highly committed
counterpart, to have the status quo remain. However, if a dispute is initiated by A, and
conflict is the result, it would prefer to have a bilateral conflict between A and B. C may
not intervene due to other factors (high domestic political costs, fear that A’s capabilities
are too great, etc.). However, a victory by B would preserve the status quo, without C not
having to incur the costs of becoming embroiled in conflict. Finally, acquiescence by B
is less preferred than is a trilateral conflict.

4.4 Solving the Model: The Game’s Equilibria

Since the game is one of complete and perfect information, it can be solved
through the use of backwards induction. If there are no information Sets, the equilibrium
concept that is used is that of subgame perfection. A subgame perfect equilibrium (SPE)
is defined as “a set of strategies if for every prOper subgame, the restriction of these
strategies to the subgame forms a Nash equilibrium” (Morrow, 1994b, 129). Backwards
induction calls for us to begin at the end of the game and work our way back up the game
tree, and via this method determine the equilibria of the game Backwards induction
compares each player’s preferences for outcomes at each node of the game, and has them
choose the strategy which will avail them of the greatest expected utility.

As Morrow (1997, 22, note 4) notes, solving games via backwards induction

requires only preferences over outcomes. Fully specified utility functions are not

113

required, as long as chance moves do not figure into the model. Since the model
outlined above does not assume any moves by nature, given the assumptions of
information being common, complete, and perfect, I can proceed by solving the game in
this fashion.”

Having specified the game’s structure and its relevant payoffs, we can now turn to

the solutions of the game. Depending upon types of players, there exist four unique

subgame perfect equilibria in pure strategies in this model.80

Equilibrium 4.1: In a game with complete and perfect information, the status quo (SQ) is
a subgame perfect equilibrium in pure strategies if the following conditions are met: tA
=2 and t3 =1 and to = 1. The equilibrium is {initiate, escalate; ~resist; ~intervene}.

Equilibrium 4. 2: In a game with complete and perfect information, acquiescence by B

(A CQB) is a subgame perfect equilibrium in pure strategies ifll) 1,4:1, t3=1, and tc= 2;
or 2) 1,4:2, 13:1, tC=2 or ”1,433, 13"], lc=2.

Equilibrium 4.3: In a game with complete and perfect information, trilateral conflict
(CNAu) is a subgame perfect equilibrium in pure strategies if 1) tA=1, t3 =1 , and tr =1 or
2) u=3, t3 =1, and tc =1.

Equilibrium 4. 4: If the game is played under conditions of complete and perfect
information, then bilateral conflict (CNAB) is a subgame perfect equilibrium outcome in
pure strategies if we relax I) the assumption of kA>k3 and assume that ’94 =k3 and 2)

assume that under conditions of power parity, B prefers CNAB to AC QB, Hence the game
is played under conditions in which (,4 =2, t3=2, and tc =2.

4. 4.1 Discussion

Under conditions of complete and perfect information, four distinct outcomes are
possible. Deterrence may succeed; B may accede to A’s demand; a trilateral conflict may
erupt; or a bilateral war may be the result. All of these outcomes are intuitively

reasonable, given that information is complete. In the remainder of this section I

 

79 This manner of solving for subgame perfect equilibria in games with complete and
perfect information is quite common in the international relations literature. A good
example can be found in Bueno de Mesquita and Lalman (1992).

114

22""

t... . .-

tutti-4

.9“ 1'.
.. ‘\

 

 

 

‘.

.1‘
V‘ t

“u
.1“

t“\ ,

“1 .

elaborate upon this and discuss some of the shortcomings that arise when informational
uncertainties are not modeled. This discussion leads into a more properly specified
model, in which information is limited.

Three items of interest emerge from this simple model. First, conflict can take
place under conditions of both power parity and power preponderance. This calls into
question the role that capabilities alone play in determining whether deterrence succeeds
or fails. As the model indicates, there is more shaping a state’s behavior than the
distribution of power within the international system, or within a triadic relationship. If
we are to understand the impact that a potential defender has on deterring an adversary,
we need to look more deeply into the issues over which they are fighting, and the values
they place on trying to attain their ideal resolutions of the issue at stake.

Secondly, we notice that conflict can occur under conditions of certainty. This
result is somewhat counterintuitive, although it has been noted by others (Bueno de
Mesquita and Lalman, 1992). This counterintuitive finding is intereSting because so
much of the literature in international politics (e.g. Jervis, 1976, 1988; Kim and Bueno de
Mesquita, 1995) suggests that conflict (with war as the most extreme form of conflict) is
the result of uncertain perceptions about an adversary. However, certain states may place
such a premium on the issues at stake that they cannot be deterred. The quintessential
example of this is Nazi Germany in 1939 (Khong, 1996). Additionally, certain states
may value the issue or status quo to such a degree that they are willing to fight for it,
despite being overwhelmed in regards to military capabilities. North Korea’s continued

advances towards the South fall into this category, especially given the overwhelming

h

80 Appendix A provides the proofs of these equilibria.

115

 

 

 

.9... a.

In). ,

I.\1..b

 

 

I
' >1.
1
», HA- .-
a; '\
4“ ._
-"n,,-
""‘1\.
S‘- w
b
'I.1,L
a.

 

 

 

"In

 

military superiority of the United States, which maintains a large military presence in
South Korea.

Third, we notice that an aggressor does not undertake limited probes, and then
back down in the face of a defender’s intervention. Extended immediate deterrence is
either never encountered, or it always fails. However, the historical record does not
bear this conclusion out. In the nineteenth century, for example, we witness Bulgaria
backing down in the face of Austrian intervention in 1885. In more recent times, the
Soviet Union made a number of demands regarding the status of Berlin, and then backed
dOWn in the face of American resolution to protect Berlin at all costs. How can such
results be explained? Is it merely an artifact of the model’s simplicity, or is there a
different explanation?

I maintain here that the model is correctly specified, and that it depicts the
empirical world quite well under conditions of certainty. This argument follows from an
earlier one made by Bueno de Mesquita (I978). Certainty, he argues can be denoted by
the structure of the international system, and uncertainty emerges when change occurs
within the system. Both of these concepts can be measured, he argues, with alliance ties.

Expanding upon this notion of certainty, I argue that it is not alliance ties per se
that signify certainty in the international system, but the signals sent between allies are
the source of certainty. How an adversary perceives the strength of an alliance
contributes to its views regarding the certainty of intervention by a target’s defender.
Under specific conditions, as was noted in the equilibrium analysis, conflict emerges in
spite of the fact that it is believed that a defender will intervene. In other circumstances

general deterrence succeeds. I maintain that it is the signaling that takes place between

116

 

. i
l... .1
» WN‘
., I
v ~‘c‘

 

 

1|.
"4

’- no

in.

 

 

allies that helps create this aura of certainty. This is not to say that increasing signals of
commitment will always lead to peace. American experiences in Korea demonstrate that
regardless of how much aid and assistance is given to the South, there are continual
challenges on the part of the communist North. Rather, these signals serve as means to
inform an uncertain adversary about what possible factors it may be facing in the case of
a conflict.

To account for the fact that limited probes do take place, and therefore model
international crises more accurately, we need to introduce some elements of uncertainty.
In the expanded model below, I focus on limited information being available to the
aggressor. A does not know what how strong C’s commitment to its ally is, and must
depend upon witnessing C’s actions to determine what type of player it is facing. I will
demonstrate below that conditions of uncertainty result in some very interesting behavior
on the part of a potential aggressor.

4.5 Adding Uncertainty to the Model: Perceptions and Signals

As I build up to a more fully specified model of extended deterrence via alliance,
I discuss the role that uncertainty and signaling commitrnentshave on a state’s
perceptions of its adversaries. I link these concepts to the modeling literature in general,
and to the deterrence literature in particular.

4.5.1 Uncertainty and Perceptions

If we recall the outcomes of the model with complete information for a moment,
it should be quite simple to ascertain why a demand by A is never followed by A backing
down from a crisis it initiated. If A knows everything about its adversaries B and C,

including their capabilities and their values for conflict, then it can predict in advance the

117

 

 

on...
-—
mi.

- .t
1.. .1
...

O

.pu.

‘u
,d"

 

 

 

 

 

 

outcome of the (potential) crisis. As history has demonstrated to the contrary, such
knowledge is not always readily available. We do witness limited probes made by states,
in which they do back down (the Bulgarian challenge to Serbia), and we do witness
conflicts that were not necessarily expected (Germany’s belief that France and Britain
would not come to the aid of Poland in 1939).

Uncertainty in the expanded model is concerned with two issues: the relative
strength of the BC alliance, as measured by military capabilities and the value that C as a
defender has for the alliance. State A begins a crisis witnessing the existence of the BC
alliance, and being cognizant that C has made some signal of commitment to its protege.
These issues of uncertainty and perception are captured in A’s initial beliefs about what
sort of alliance it is facing. Capabilities, as Bueno de Mesquita, Morrow, and Zorick
(1997) demonstrate, often provide uncertainty for potential aggressors. The implications
of signals between allies, however, are not as clearly defined. A state may be willing to
supply its protege with aid and weapons, but be unwilling to commit troops if the protege
is attacked. In other instances the defender may be willing to run the risk of all out armed
conflict in order to protect an ally that it deems particularly valuable.

Take for a moment the relationship between Great Britain and Poland in 1939.
Germany’s invasion of Poland occurred after France and Britain had abandoned
Czechoslovakia to Hitler’s demands. Neville Chamberlain had even remarked that the
British people were unconcerned about the fates of the Czechs, a people they knew little

about. France had behaved in a very similar manner. Based upon these actions, Hitler

118

 

0....

\
.tnu ~<

gun a-

nu...-

I
.v...“

 

Mb 1.I

 

”in
N11

1
no“

'u,

4
.(h

 

 

 

1.
‘ v~
L

 

most likely believed that Poland was not a vital interest to these Western democracies.
and, due to his miscalculation, the Second World War erupted.81
Many discussions regarding perception and uncertainty take the form of
counterfactual arguments, in which one can posit questions speculating on what may have
occurred, given a particular change in some historical event.82 So, in this respect, what if
Britain and France had been firm in their demands at Munich? Or if the United States
had negotiated a stronger treaty with Taiwan in 1954? The result of these differences

could have changed world history. However, it is uncertain if we could ever know for

sure.”

Moving from conceptual thoughts on uncertainty and perception, I now discuss
how to account for such uncertainty in game-theoretic models. Let it be recalled that the
concern here is over the type of player that C is, and how strongly or weakly committed it
is to its minor power protégé. How can different types of players be modeled within a
common framework? The solution to this question can be found in'Harsanyi (1967-
1968). In his seminal article on Bayesian games, Harsanyi allows nature to make an
initial move in a game, and the move stipulates with particular probabilities the type of
player(s) in the model. So, in the model presented below, nature makes a first move and

determines the type of player that C is, with some probability 1:. Thus, nation A is able to

8' See Huth (I 988); Khong (1996); and Thompson (1997) for greater discussions on the
Possibility of deterring Nazi aggression under Hitler’s leadership.

82 Fearon (1990) and the selections in Tetlock and Belkin (1996) all address this issue
from a variety of methodological perspectives.

119

.0.” 1
4..

use”-..

«pd-l

ago-.5.

.1n..,.»

4
Aunt:

sfltju ‘
outnh.

 

.4
. .,
1.
”‘1..'

.

‘37“;

31.1 ,

determine in some probabilistic fashion how committed of a defender C actually is.
Rather than knowing at the onset C’s type, A must use its senses of perception, witness
actions undertaken by C, and make an educated guess as to how sincere C is about
defending its protégé.

Such uncertainty in the model can be potentially overcome by witnessing the
signal sent by C to B in regards to the alliance they share. Signals serve two purposes in
this model of extended deterrence: they demonstrate the value that a defender holds for
the protege; and they provide a measure of C’s willingness to assist its ally in time of
need. Both of these provide A with information regarding C’s type. In this vein, the
discussion now turns to a delineation of signaling within the context of an alliance.

4.6 Extended Deterrence via Alliance with Incomplete Information and
Signaling

By introducing the concept of uncertainty, we are introducing a larger dose of
reality into the model. With uncertainty present in the model, we can allow the signals
sent by C to be indicators of its willingness and to model the types of player that A is
facing. This next section outlines the modifications made to the earlier game in order to
account for uncertainty and signaling.

4. 6.1 Extensions and Modifications to the Previous Model

Having discussed the role that perception plays in international politics, and the
use of signals to help identify types, we can now proceed to outline the expansions made
to the earlier model. To account for these refinements, I must make a few additional

assumptions.

_—k

83 These “what if” scenarios are actually counterfactuals. The recent book edited by
Tetlock and Belkin (1996) provide some interesting insights into the use of such

120

 

 

 

’ ,z
.>\
‘v'
".
1 .
‘ 1""
. ..l .1.

 

 

 

a.

‘W-

h‘ ‘
v “\

 

 

Assumption 1 1: Information in this model is incomplete and asymmetric. C and B know
A ’s type. A is uncertain about C ’s type. In addition, A knows its own type, and B and C
know their own type. A only knows the types of C that are available, and not the specific
type that C is.

Assumption 12: If C is strongly committed to its alliance with B, it will always signal that
it is strongly committed to its prote'gé. Thus, 7t(say stronglstrong) = I and 7: (say
weaklstrong) = 0.84

Assumption I 3: If C is weakly committed to its ally, it may signal strong in an attempt to
deceive its adversary.

These additional assumptions provide additional uncertainty within the model.
To account for this uncertainty, we must introduce the concept of beliefs into our
equilibrium refinement. Players are forced to update their beliefs based upon the actions
that other players take. Thus, we rely on an equilibrium concept known as perfect
Bayesian equilibrium (PBE). A perfect Bayesian equilibrium is defined as “a belief-
strategy pairing such that the strategies are sequentially rational given the beliefs and the
beliefs are calculated from the equilibrium strategies by means of Bayes’s Theorem
whenever possible” (Morrow, 1994b, 176). A PBE requires us, therefore, to utilize the
tools of probability in order to update our beliefs of what is happening in the world
around us. In analyzing the game outlined below, we must account for this uncertainty,
and acknowledge that states update their beliefs as their interactions occur throughout the
game. In the game presented here, the concern is with A’s perceptions of its adversary.

The conceptualization and implications of commitments being viewed as signals

in the context of deterrence and alliances was introduced earlier. In the remainder of this

 

scenarios for generating scientific knowledge.

84 . . . . . - .
. There 18 no incentive Within a deterrence alliance for a defender to Signal weak when it
is actually strong, because such a signal may actually invite an attack upon a prote'ge’.

121

~1—

«.11.‘

.
r
3‘...

:15

 

Q
'I
q
7‘
V!
c
it
4,.
(-
0

section I identify the generic components of a signaling model. and then present my
extended model in a signaling context.
4. 6.2 The Structure of Signaling Games”

Signaling models are a special class of incomplete information, and generally rely
on Bayesian techniques to arrive at solutions. In such a dynamic game a signal serves as
communication prior to the onset of the game itself. Such signals can serve as “cheap
talk” indicators, which are costless to send (Banks, 1990), or they can serve as “costly
signals” that indicate some measure of commitment (F earon, 1994a, 1994b, 1997).86 In
the model developed below, signals are costly, but vary in their intensity depending upon
the type of defender.

In its theoretical form, a signaling model involves two players; a Sender (S) and a
Receiver (R). The progression of the model occurs in the following manner:
1. Nature determines from a set of types T a type t, for the Sender. Formally, T =

I. “4136:; .zifhis type is drawn with a probability distribution of p(ti) >0, and p(ti) +

2. The Sender knows its type, given Nature’s draw. Sender then chooses a signal mi
from a set of signals M, where M = {m ...... ml}-

The Receiver witnesses m (but not Nature’s choice of t-.). The Receiver then chooses
an action at, from a set of possible actions, where A = {a1 ...... aK}.

Lo)

4. Payoffs for the signaling model are given as Us (ti, m, at.) and UR (t1, mj, at).

——_

85 Most of the discussion here comes from Gibbons (l 992). The concept of signaling
stems from Spence (1973). Others that provide good discussions of signaling models in a
purely theoretical manner are F udenburg and Tirole (1993); Kreps (1991); and Osborne
and Rubinstein (1994). Applications within political science include Banks (1991) and
Morrow (1994b). Within international relations the main works are Banks (1990) and

Fearon (l994a, 1994b, 1997).

8" In keeping with the arguments made in chapter 3, I maintain that costless signals do not
exist in the models developed herein.

122

4. 6.3 A Signaling Model of Extended Deterrence via Alliance with Incomplete
Information

Having discussed the additional assumptions, we are now ready to view the
updated model. Figure 4.2 depicts the model of extended deterrence via alliance with
incomplete information and signaling.

[FIGURE 4.2 ABOUT HERE]

Some discussion of the initial moves and the terminology associated with the new
additions to the model are warranted. The model begins with a move by Nature“, which
determines the type of player C is. C can be a strongly committed defender with
probability (‘9, or a weakly committed defender with probability I-®.88 C then makes its
initial move, and signals if it is strongly or weakly committed to its alliance relationship
with B, with m=l being a signal of a strong commitment and m=2 signaling a weak
commitment.

Upon witnessing C’s signal, A must decide whether or not to initiate a demand
against B. Since, by the assumption made above, we know that C will never signal weak
if it is actually strong, A knows that if it witnesses a weak signal it is facing a weak
defender. However, what if A witnesses a strong signal? It must determine whether it

is a true signal of C’s intentions, or if it’s an attempt to bluff A into not initiating any sort

 

87 Nature is defined as a “non-player who takes random actions at specified points in the
game with specified probabilities” (Rasmusen, 1989, 22). As an alternative means of
. envisioning nature, nature can be depicted as chance (or fate) determining certain types.

88 In the complete information variant of the model presented above, we assume that the
value of O is known to be either 0 or 1. In this model there is uncertainty on the part of
the aggressor, and (9 serves as one of the components of calculating A’s prior beliefs.

123

 

. w-x
l -n.,
t .
. >>ht
I
» -’
a

 

 

 

4;y-.

N1. ,
1,‘

. .
, r
\

\ a”

 

 

 

 

of a demand. A believes that C is a specific type of defender with prior beliefs it. Based
upon its prior beliefs about C’s type, A must decide whether or not to initiate.89

As in the above model, B must decide whether or not to resist. If B resists, then C
is faced with deciding whether or not to intervene. This action by C forces A to update
its beliefs about what type of defender it is facing. A updates its beliefs after witnessing
C’s move. Its posterior beliefs about C being an actual strong defender is represented by
11*, where u*= 1 means that A knows that C is strong with absolute certainty.

A few moments should be taken within the context of the theory developed in this
dissertation to discuss the factors that influence a potential aggressor’s updated beliefs.
As was noted above, a defender can assume two distinct types: a strongly committed
defender and a weakly committed defender. As was discussed at some length, the type of
defender is based upon the value it holds for its protégé. How can an'attacker update its
beliefs regarding the type of defender it is facing?

Information and beliefs are intimately connected. States have certain prior beliefs
about their adversaries, and must update them as interactions transpire. Often times,
however, these beliefs are uncertain; information is incomplete or private. Based upon
this limited information, the potential aggressor must make conjectures concerning the
type of defender it is facing.

Beliefs are formed at the extended general deterrence level and updated at the
extended immediate deterrence level in the game outlined above. At the level of
extended general deterrence, a potential aggressor, uncertain about the commitment level

of a defender, can formulate beliefs based upon the strength of signal sent by the defender

——_

89 In the equilibria proofs in Appendix C these posterior beliefs are calculated.

124

 

leafy- 4; ..,
"yttrhg. u.

 

1
-" ‘3" 4

u...,_5l I

1..

.l a :u. .
“Lu as.“

 

‘t
1':"-‘~.l

 

to the protégé, the past actions of the defender in regards to its prote’gé and/or other allies,
and in the benefits that the defender receives from the protégé in return for security
benefits.

Regardless of how these prior beliefs are formulated by a potential aggressor, they
must be updated once a crisis is initiated and (potentially) escalates. Such updating
depends upon the actions a defender takes regarding its threatened protége'. Does the
defender do nothing, offer concessions, match military force with military force? All of
these actions allow the attacker to update its beliefs regarding the type of defender it is
facing. Thus, beliefs and limited information play a major role in determining whether a
crisis will escalate or not.

Having discussed the role that beliefs play in the model developed here, we now
progress to the formal explication of the model. A few modifications to the traditional
signaling format are required. In this model the Sender is C, the defender in the alliance
relationship. The Receiver is A, the potential aggressor. Equilibria will be denoted in the
following manner. UA, B, c (me; at; an; ac; tc), where m is the signal sent by C; a is the
action undertaken by A, B, and C respectively; and t is what type of defender C is. To
account for the actions each player can take, Table 4.4 outlines the meanings behind each

action.

125

TABLE 4.4

Defining Actions in the Signaling Model

 

 

 

 

 

 

 

 

State Action Definition
A 1 Initiate, Escalate
A 2 Initiate, ~Escalate
A 3 ~Initiate, ~Esca1atc
B I Resist
B 2 ~Resist
C 1 Intervene
C 2 ~Intervene

 

 

 

 

 

With these modifications in place, I now move to solving the model with
incomplete information.
4.6.5 The Model ’s Equilibria

In accordance with the above discussion, the solution concept for the model is that
of perfect Bayesian equilibria. I focus on the two different situations that occur. In the
first situation, C signals that it is weak. In the second situation, C signals that it is strong,
but A is uncertain regarding the legitimacy of this signal. I examine the ‘case of a weak
signal first.
Scenario 1: C signals weak.

If A witnesses a weak signal, then it knows with absolute certainty (u = 1.0) that
it is facing a weak defender. Thus, the relevant complete information conditions as
discussed above hold. This is not surprising, since information is complete and perfect in
such a situation. It is in the second scenario that signaling becomes an issue.
Scenario 2: C signals strong

What becomes more interesting is the existence of a strong signal. Then there

exists the possibility that C is bluffing in an attempt to deter A.

126

It

 

 

Ills

31’5", ‘3!
bruins.

7.013 1113

’1 »',.

\l I

*1~...4
»

“:m ..

ml
V',‘ "111.:

 

1“.
.‘,\."\;.
‘4

 

To solve the equilibria of this modified model, I follow the procedure explicated
by Gates and Humes (1997). 1 specify the type of player A is, and draw out all possible
strategy combinations that exist in regards to the two possible types of player C can
assume. I then determine which ones are in equilibrium, and denote what beliefs must
exist for such an equilibrium to occur. In order to do this. I first calculate the prior beliefs
that A has concerning C. G), as was already noted, is the probability that C is a strong
defender, and 1-6) is the probability that C is a weak defender. I now introduce some
more notation regarding beliefs. 1 denote the following probabilities as it. Thus,

7! (signal stronglstrong) = l.
rt(signal weaklstrong) = 0.
1t (signal stronglweak) = or
1r(signa1 weaklweak) = 1-0t

From this discussion we note that if C is actually strong, it will signal with
certainty that it is strong. Conversely, if we witness a weak signal, we can ascertain with
certainty that C is weak. C does have an incentive to signal that it is strong if it is weak,
and naturally has an incentive to signal strong when it truly is strong. We need to specify
A’s belief about C’s strength when it witnesses a strong signal from C. We state this by
using Bayes’ rule.

Bayes’ Rule is a means for an actor to determine its beliefs under uncertainty. We

SPCCify Bayes’ Rule for A in the following manner:

in (tlm) =Mildt)
p(t)1ri(81t) + p(t’)1r1(81t’)

127

 

no a)” 1'
ekuhvl “.1

let; L

.1
,Ntbn-H ll

‘ l
v.1 ‘9‘ .1
in u\'\l

p.11. «1

.C.....‘ .1.

 

I: ‘Ihfll
|_ “>-
.1.l)_,l
'3‘ :tr‘i

"1111.4"
n 'Q‘L

 

'3'.
5

O

f D
{X}

Having this general rule at hand, we can now calculate the beliefs that A holds

concerning the type of defender it is facing.90 These beliefs are denoted as A’s prior

beliefs, having witnessed C’s signal, and are represented by u. The beliefs that A holds

are developed by A given the context within which it interacts with both B and C. Thus.

beliefs are shaped by the signals C sends to B, changes made in the status quo between A

and B, or in the past actions C has undertaken on behalf of its prote’ge’ or other allies.
Noting the origins of these beliefs. I now turn to an explication of how they are

represented in the formal model.

uA (weaklweak) = 1.0

in, (stronglweak) = 0

up, (stronglstrong) = O
@‘HX-CXG)

11A (weaklstrong) = (it-Oct
a-aG) + (9

These denote the prior probabilities that A possesses about C, upon witnessing a
strong signal from C. As noted above, a move by C in the game allows A to know with
certainty what type of defender it is facing. In game-theoretic parlance, C’s actions
reveal its type. Thus, as A makes its final move, its posterior (updated probability),

denoted as u*, is 1.0 in all of the cases here. With these beliefs in hand, we can now

examine the different cases that arise.

¥

90 Derivations of these prior probabilities utilizing Bayes’ Rule can be found in Appendix
B.

128

u
l w"
\ liVl.\
11.11 :13 7‘

1.11L1uy‘i u

I
'1‘ 1 4. '1
.m‘ ~\n
1

.4

d
“1:: li101

\ (571'

-
~v~g|

“CS-£1.11

P'
v

LN“
“4‘41”.
! 1
1T?

1 L
I IV

I
111 .
Mil-1‘.

1,1‘

C036 1. [4:]

In this case, A is an aggressor that is stronger than the combined strength of B and

C. Hence, we know from the definition of tA=1 from above that it will never fail to

initiate a crisis in the first place, and it will never fail to escalate a crisis. Thus, we need

only examine all instances in which A initiates and escalates.

Equilibrium 4.5:

{m=1: a4 =1; ag=I; ac=1: tc=1) with beliefs p: 1.0 is a perfect Bayesian Equilibrium

of the model with the following belief structure: )1) = 9 ; HA" = I . 0.
0+a-a9

The outcome is a trilateral conflict.

Equilibrium 4. 6:

{m=I: aA =1; a3=2,' ac=2s lc=2l is a perfect Bayesian Equilibrium of the model with the

following belief structure: a, = a- 3a; int“ = 1.0.

a-aQ + 0
The outcome is acquiescence by B.

Case 2: (,4 =2
In this case A is an aggressor that is weaker than a combined foe of B and C, but

is stronger than B is alone. Hence, we know from the definition of tA=2 that it will never
escalate against an intervention by C.

Equilibrium 4. 7:
{m=1: a A =2; ‘13:]; ac=1-’ tC=I) is a perfect Bayesian Equilibrium of the model with the

following belief structure: 1.1,. = 9 ; in“ = 1.0.
a+Q + (10

The outcome is back down by A.

Equilibrium 4.8:
{m=1_- a4 =1; a3=2; ac=2: m=2} is a perfect Bayesian Equilibrium of the model with the
following belief structure: ,uA = a—Qa; 1114* = I .0.

a—aQ + 0

The outcome is acquiescence by B.

 

129

._ 1
, .

.1111. I:
-42 ‘5»)

 

v- 1
I If 4
1 1

calla. it

1’. 1‘1'
11.1.1“

Case 3: 1,, =3

In this case, A is equal in power to the combined alliance BC. This then becomes an

analysis of the model under conditions of power parity.

Equilibrium 4. 9:
{m=1 .° a4 =I ; a3=I; ac=15 tc=1} is a perfect Bayesian Equilibrium ofthe model with the
fallowing belief structure: 11.4 = (9 ; int“ = I .0.

a+0 + a@

The outcome is a trilateral conflict.

Equilibrium 4.10:

{m=I: 0,, =1 ,' a3=2; ac=2: tc=2} is a perfect Bayesian Equilibrium of the model with the

following belief structure: in, = a—Qa; 1124* = 1.0.
a-aQ + (9

The outcome is acquiescence by B.

4. 6. 6 Discussion

When compared to the model with complete information, some interesting

differences emerge from the signaling model. For purposes of comparison, Table 4.5
indicates the equilibria that emerge from the model with complete information, and Table

4.6 presents the equilibria that emerge from the model with complete information.

130

 

 

TABLE 4.5

EQUILIBRIA UNDER COMPLETE INFORMATION

 

 

 

 

 

 

 

C’s Type tA=l tA=2 tA=3
tc=1 CNALL SQ CNALL
tc=2 ACQB ACQB, CNABm ACQB

TABLE 4.6

EQUILIBRIA UNDER INCOMPLETE INFORMATION

 

 

 

C’s Type tA=l tA=2 tA=3
tc=1 CNALL BDA CNALL
tc=2 ACQB ACQB, CNAB ACQB

 

 

 

 

 

 

First of all, we notice that it is only under conditions of certainty that extended
general deterrence via alliance is successful, and the status quo prevails. This seemingly
contradicts much of the standard conventional wisdom in international politics that
suggests that bluffing can lead to the status quo being preserved."2 Upon further
reflection, however, it is not that surprising. There is an incentive for challengers to
make limited probes, and see what actions a defender will take. Given the assertion that

alliances have a tendency to be unreliable in regards to their honoring obligations, this is

intuitively plausible.

9' Recall from above that a bilateral conflict is a special case of a target being equal in
military capabilities to the attacker, and the defender is not willing to intervene. This
result holds under both complete and incomplete information scenarios.

131

I

..~<mfll

hi.“ i, .
i

.
Hr. x-x'

‘thn‘l‘I

rinse:
hint: ‘- '
.

1.xm‘41
' 4

--u . l ”.5;-

 

 

 

 

 

If an incentive for states to challenge the status quo exists under conditions of
incomplete information, then how do we explain the preponderance of cases in which
conflict does not break out, and general deterrence holds? As noted above, extended
general deterrence succeeds only under conditions that approximate complete
information and certainty regarding the commitment level of a defender. Based upon
their beliefs about a defender, challengers will determine whether to select themselves
into crises or to forgo any initiation (F earon, 1993, 1994a). This issue of selection
effects, then, hinges largely upon the information and beliefs that a potential aggressor
holds prior to the possible initiation of a crisis. As a aggressor’s beliefs are adjusted
towards 1.0, in believing that a defender will intervene, then conditions of certainty are
approximated.

Secondly, conflict can once again occur under conditions of power parity.93
Once again this suggests that military capabilities are not the only force that drives
disputes. Rather, it is largely a concern the issues that are at stake that drive disputes,
rather than mere calculations of military power. If power were all that mattered, one
would witness few conflicts in the world. The fact that asymmetric initiation occurs

suggests that power is not the end-all to understanding world politics and interactions

among hostile states.

h

- 92 Unfortunately I cannot assume full credit for this insight. Both F earon ( l 997) and
Papayounou (1997) find that attempts at bluffing are not upheld in equilibrium.

93 While I do not relax the assumption here that A and B can be equal in capabilities, the
same results do hold if I do. The result will be a bilateral conflict with no intervention by

C. Interested readers in this proof are advised to consult Appendix A.

I32

tJ'JWJr'
Uh», . w

IJ'QY‘."
“all“.

.3!) 3'
in. ,y‘

‘3'." I

.
"‘1...‘

‘.‘-‘|‘ 4
gt .1!”

T’M‘J

 

 

 

4.7 Conclusion

The above games demonstrate the conventional wisdom that exists regarding
extended deterrence. If a defender is able to make a credible commitment to its ally, then
deterrence should be effective. However, the failure of a defender to credibly commit to
its ally can lead to the ally giving in to an aggressor’s demands.

I have argued that these commitments can be viewed as signals, which are costs that
defending states are willing to incur in their quest for autonomy. Alliances can provide
deterrence for member states if the parties involved all have common knowledge about
the others involved. Only when uncertainty is introduced can deterrence fail. The
models suggest that since signals are costly, the greater the likelihood that a defender will
intervene, so as not to see its sunk costs go to waste. This is a finding that often runs
contrary to the literature on alliance reliability, and warrants further scrutiny.

In the formal models presented here, the theoretical framework outlined in chapter
3 is validated. The actions of a defending state, witnessed by the signals it sends and
receives to and from its protégé, help to create an aura of certainty in the international
system. These signals, envisioned as costs that have been invested, provide a means for
potential aggressor states to formulate beliefs and perceptions about the commitment of a
defender. These perceptions help determine whether an aggressor wishes to exploit a
favorable Opportunity in the international system in order to further its national goals.

Having presented a theory of extended deterrence via alliance in a formal fashion, the
time has come to evaluate the theory against the empirical record. Do costly signals
exchanged between allies provide the desired deterrence benefits? Do they influence an

aggressor’s perceptions and create conditions approximating certainty? Do states make

133

 

nqfly‘Q';

{in-uni.

- ‘ c

z N
.
at» .‘AA

attempts to exploit opportunities in the status quo in Spite of the existence of a deterrence
alliance? Are deterrence alliances effective at ensuring the status quo for their minor
partners? These are some of the questions that emerge from the theory presented above,

and will be subjected to empirical scrutiny in the following chapters.

I34

FIGURE 4.1

EXTENDED DETERRENCE VIA ALLIANCE WITH COMPLETE
INFORMATION

ED

 

 

1m = Initiate
= resrst

in= intervene
e = escalate

 

 

 

I35

EX

 

Strong
Signa

ini ~ini

SQ

I'CS "TCS

int ~int

FIGURE 4.2
EXTENDED DETERRENCE VIA ALLIANCE WITH INCOMPLETE
INFORMATION AND SIGNALING

N
Strong
Signa
~ini ini
B
SQ
~res
ACQ

int

CN BD CN BD“
AB

ALL

136

l—G
C
Weak
i gnal
A
~ini
SQ
res ~res
C
ACQ
~int
~es e ~es e

ini

['85

int ~int

[fiowfii
L \l‘fi}
I

z
",N', 'r
~‘-u\ n
u

‘Jiura
“i“ ll

Elfin"

\

I“ "I
I“
'

I‘ll :.
lib»

CHAPTER 5

PROPOSITIONS, CASES, AND VARIABLES

5.1 Introduction

A common critique of game theoretic models is that they do not lend themselves
to empirical testing, or that they demonstrate the “obvious”. As discussed in chapter 2,
game theoretic models of deterrence are often the targets of those opposed to rational
deterrence theory. As the equilibria from the models in chapter 4 demonstrate,
interesting, even counterintuitive findings can come from formal analysis. Unfortunately,
many opponents of formal theory criticize it for being too elegant, with no empirical
references.

One mistake that I intend on avoiding is not testing my models in an empirical
manner. Modeling solely for the sake of modeling is not something I am interested in.
Models help us to understand a particular question, and identify the important factors
which influence behavior. In this regard, I realize that there may be particulars to certain
wars or conflicts that my models may not address. Aberrations and outliers do occur, and
often provide avenues for further investigations and refinements to theoretical discourse.
However, I feel that the formal models presented in this dissertation adequately address
situations in which extended deterrence via alliance is functioning. It merely becomes a
matter of finding the proper means of testing them against the historical record.

Such concerns are not new. In fact, as I noted in chapter 2 when the rational
deterrence debate was discussed, the most common criticism of formal models is the
difficulty of scrutinizing their results empirically. A large portion of this problem has

been the failure of formalists to draw out testable aspects of their models. Fearon (l994a,

I37

. 31“",

bin—cu

MIA

s
I" A
d

$.L '3‘

 

 

.15.

“in;

in.
my,

 

 

 

237) notes this problem and remarks; “modelers have probably not done enough to draw
out hypotheses that can be tested using simple, plausible, and readily available
measures”. Given that factors such as power, issues, audience costs, and signals are
central to much of international politics, and to the theory developed in this dissertation
in particular, such limitations need to be overcome to the greatest extent possible.

With this in mind, it should be noted that not all of the parameters in the formal
model can be estimated in a statistical/empirical manner. How does one operationalize a
concept such as resolve, or model limited information in a quantitative manner? Often
the best that we can hope for is to test implications from the formal models, and augment
these implications with evidence from the historical realm. 94 Thus, in this chapter I will
be focusing on the Operational definitions of the variables being tested, but when I turn to
actual testing in chapter 6, I will return to the six cases that have been elaborated upon
throughout this body of work. The focus on these cases adds a greater richness and
validity to the theory developed here, and provides some bite to the often-cold numbers
derived from statistical formulations.

Having stated my desire to test my formal models empirically, my attention now
turns to what specifically is being examined by my statistical modeling. My concern
throughout this dissertation has been to model the impact that an existing alliance has on
ensuring effective deterrence. Since deterrence in its extended form is a policy proffered

by one state (a defender) in exchange for some benefit, I feel that is most important to

94 Gates and Quinones (1994) and Gates and Humes (1997) both argue that in many
instances, only implications can be tested from formal models, rather than testing the
models directly. The best attempt within international relations to test a game-theoretic
model directly, with developed measures for utility, is Bueno de Mesquita and Lalman
(1992)

I38

«~H'Jfi'r
H,'..‘\.\.lu

r

)1 ' N‘, r"
ltkfirt'n

WW 1
r! . .nu‘ U
. .

bU.,sii& .

Ill". 1"“

I
Art ‘J'
‘ .¥‘
‘1vl.
.r‘ Jr

" nui-

I
119;.
“ wt

grin
1-|i
Jim

,.
t¥

I
lid“.

'1.
‘4

v
-zr‘

(‘1
re.
,.

concentrate on the factors that shape both a defender’s and a potential aggressor’s
decision calculus. Therefore, I believe that my empirical models are an accurate
portrayal of the factors that A encounters when it considers when making a choice in a
conflict situation. Although much of this focus is on the actions of an aggressor, it does
not imply that the game models are incorrectly specified. The decision made by any of
the actors in the system is conditioned to some extent by what the other two actors have
(or have not) done.

In the remainder of this chapter I develop empirically testable propositions that
are derived from my formal models. I discuss at what level these propositions function
(extended general deterrence or extended immediate deterrence), and the implications
and reasoning behind each. I then discuss the issues of selection bias that occur in
quantitative studies of deterrence. Moving from this discussion on selection effects, I
delineate my set of cases, and explain how I’ve attempted to avoid some of the problems
associated with selection effects. Finally I operationalize the variables, dependent and
independent. that I will use to empirically test the propositions in chapter 6.

5.2 Propositions

As the theoretical framework and formal models suggest, two distinct processes
are at work in the study of extended deterrence via alliance. There exist a number of
situations in which extended general deterrence via alliance is attempted. In most years,
the status quo is not challenged. In the minority of cases in which an actual challenge is
witnessed, escalation to large-scale conflict is not the norm. However, in some of these
cases, extended general deterrence failures escalate, and extended immediate deterrence

is challenged as well. Thus, in order to test in an empirical fashion the influences that

I39

I' 7‘. ‘ "I"
rerdr PM”
Jilfflfl‘rt‘é Z
TONSIIOII
l I

ltrioxlinu 1

Kinetic n
tramps ‘n
Itl filrmd
has been c
l 10W i 3?:
Internist»
Smith I 19'
la: ‘Jtlliz

43% he ll

“limit

various parameters from the formal model have on the success or failure of extended
deterrence at both the general and immediate levels, we need to formulate specific
pr0positions from both levels that can be subjected to empirical scrutiny. In the
following two sections I outline and elaborate upon the testable pr0positions that emerge
from the analysis of the models’ equilibria in chapter 4. I provide some detailed
discussion of the propositions, and how they emerge from the formal model.

It should be recalled from the introduction to this chapter that testing garne-
theoretic models is an often difficult, if not impossible, venture. This is not to say that
attempts have not been made to directly test outcomes of formal models. A technique
developed by McKelvey and Palfrey (I998), logit quantal response equilibrium (LQRE)
has been developed as a means of empirically testing game-theoretic models. Signorino
(I999) applies the LQRE framework to Bueno de Mesquita and Lalman’s (1992)
lntemational Interaction game, and finds limited support for its findings. Altematively,
Smith (1997) develops what he terms a Strategically Censored Discrete Choice model
that utilizes Bayesian techniques. In his reanalysis of the Bueno de Mesquita and Lalman
data he finds overwhelming support for their findings. As evidenced from the brief
evidence presented above, there exists no consensus on how to test formal models
empirically. In this vein, therefore, the pr0positions stated below seek to test implications
from the formal models, rather than test every single parameter that was included in the
models.

5. 2. 1 Propositions regarding Extended General Deterrence
The first set of propositions deals with a larger set of cases; namely, those in

which deterrence has been extended by a great power defender to a minor power protégé.

I40

“afprrc'
Ubhbr b .

301. l:

mfiw:
ri-‘nA-I

ft";
“III. “II

As the equilibrium analysis from chapter 4 demonstrates, at the level of extended general
deterrence, there are five major factors that influence whether deterrence is successful or
not. I argue that extended general deterrence success is a function of these five key
parameters, which are derived from the formal models in chapter 4. F ormally,

Extended general deterrence success = f (m, y, d, k, 11) (1)
where

m = commitment sent from C to B (sunk cost).

y = autonomy benefits C receives from B.

d = domestic political constraints on C.

k = military balance (short term and long term balance).

,u = A ’s prior beliefs about the defender ’3 type.

Proposition 1.0 (Commitment/Security Proposition)

As the level of commitment C sends to its prote'ge' B increases, the greater the likelihood
that extended general deterrence will be successful.

Proposition 1. 1

As the level of arms transferred fiom a defender to its prote'ge’ increases, the greater the
likelihood that extended general deterrence will be successful.

Proposition 1.2

The likelihood that extended general deterrence will succeed increases if the defender
and protégé are united by a defense pact.

Proposition 1.3
The likelihood that extended general deterrence will succeed increases if the defender
establishes military bases on the territory of the protége'.

These first three propositions test the impact that specific commitments made by a
defender on behalf of a prote'gé have on ensuring extended deterrence success. As is
borne out by the equilibrium analysis in the prior chapter, the signal that is sent is

strongly associated with the status quo being preserved. Increasing the signal sent

141

I
»,1.'

Raid.“

h

‘1)!“
S\>Ln..l.
.

‘0‘ .1.
F. El ll

36L '3.le

‘1 ;
Elm.“

regarding security to an ally helps to create a sense of certainty in the international
system, and provides little incentive for a potential challenger to initiate a dispute which
it knows will lead to the intervention by a third party.

If we examine each of these prOpositions in more detail, their implications
become much clearer. In the first instance, arms transfers from a defender to a protége’
help increase the security that a minor power gains from the alliance relationship. As
Sislin (1994) notes, arms tend to be utilized as means of influence. They often provide
political ties, and symbolize the possibility of improved relations as time progresses. As
was argued in chapter 3, if great powers use alliances with minor powers to pursue
interests abroad, then the transfer of arms ties a defender much more closely to the
security and defense of its protégé. If these transfers are witnessed, then the value of the
protégé becomes all the clearer to potential challengers.95

In the second proposition the concern is over the type of alliance that has been
established between the defender and protége'. As discussed in chapter 2, different types
of alliance ties exist. I distinguished between defense pacts, ententes, and non-aggression
pacts. The defense pact is the tightest form of commitment, for it pledges military
assistance in the case a signatory is attacked. Formalizing the alliance commitment in
such a tight manner sends a strong signal from defender to potential challengers, warning
them that the protégé is of considerable value to the defender.

Finally, the establishment of military bases on a prote’gé’s territory is a signal of

commitment that should increase deterrence effectiveness. The establishment of bases

 

95 Additionally, as Bueno de Mesquita (1981) and Ray’s (1990) replication of Bueno de
Mesquita confirms, allies have a tendency to fight one another. Arming an ally
demonstrates a great deal of faith in the intentions of the minor power, for these weapons
could conceivably be used against the defender at some later date.

I42

midis: the
the protege.
great porter
ironwork (
formation. (
internationi

How
thezr soldier
establishme
than autono
irren'entior
Propositior

The greater
(railroad (.

Propositior

lire higher
{railroad I
Proporitlor
Ilia llleflr’rr
sh3 ’

Proposltlor

o v
"6 ‘l'Qiml
(1

ttttraphr

D ,
I roposmo.

he r’ilefih

straddles the line between being a benefit for the defender and being a security benefit for
the protege. As Harkavy (1989) points out, the establishment of bases abroad permits a
great power to maintain a global military presence. According to the theoretical
framework outlined in Chapter 3, this is one of the main goals of asymmetric alliance
formation. Great powers ally with smaller ones in order to gain some leverage in their
international affairs.

However, security benefits arise for the protégé as well. No state wishes to see
their soldiers killed and then have their leaders stand by and do nothing. The
establishment of bases, I argue, provide a much greater security benefit to the protege
than autonomy to the defender, because the help increase the likelihood of military
intervention in the case of extended general deterrence being challenged.

Pr0position 2.0 (Autonomy Proposition)

The greater the degree of autonomy benefits that C receives from B, the greater the
likelihood that extended general deterrence will be successful.

Proposition 2.1

The higher the level of exports that goes fiom a defender to its prote'gé, greater the
likelihood that extended general deterrence will be successfitl.

Proposition 2.2

The likelihood that extended general deterrence will succeed increases if the defender is
able to import vital strategic raw materials fiom its prote'ge’.

Proposition 2. 3

The likelihood that extended general deterrence will succeed increases if the prote'ge' is
geographically situated in a strategic locale that is beneficial to the defender.

Proposition 2. 4

The likelihood that extended general deterrence will fail increases if the BC alliance is
democratic in nature. ’

I43

 

mi...

fi"‘..‘.
ind“

Al. (H

\-l cu

"9-
,
pod“

£03).

In these four propositions we are able to see what benefits a great power is able to
obtain from its minor power protégé. The first two propositions deal with economic
issues: trade and supply of vital raw materials. If a great power defender is able to gain
little in terms of security from allying with a minor power, then it must have some reason
for forming the alliance in the first place. Trade and the acquisition of valuable raw
materials would make the relationship more valuable. Being able to export goods to the
protégé helps increase the economic base on the defender’s home front, while the import
of strategic raw materials helps the defender maintain its military might. And, to the
protégé’s benefit, trade often produces what have been termed security extemalities
(Gowa, 1994; Mansfield and Bronson, 1997).

Having an ally in a strategic locale is also beneficial for the defender, as was
discussed in the previous chapter. Territorial contiguity, protection of a trade route, or
the ability to use a protégé’s territory for forward basing are all issues that provide
autonomy benefits to the defender. Take for example, the relationship between Austria-
Hungary and Serbia in the latter part of the nineteenth century. Although the Austrians
did not establish military bases on Serbian territory, they did have permission to move
troops across Serbian soil in the event of a war. Alternatively, the Turks have often been
viewed as good alliance partners due to their control of the straits leading to the Black
Sea.96 Strategic location, therefore, is a primary benefit that a defender can gain from
allying with a minor power.

Finally, a democratic alliance may lead to an increased chance of extended

general deterrence failing. As was noted in chapter 3, the “gains from trade” argument

r44

a ‘ll'
Me:

I t
39‘ Av
M .r

jiw'il‘
In.“

so CI:

espoused by Gartzke and Simon (1996) and Simon and Gartzke (1996), non-democratic
regimes may be more willing to tolerate restrictions and policy concessions demanded by
great power defenders than are democracies. Such states may be more willing to permit
the defender to establish bases on their territory, or align their foreign policy aims with
that of the defender. Much of this can be done due to the fewer constraints placed upon
leaders of non-democratic states. Democracies tend to be more hesitant to give in to the
demands of a defender in regards to policy coordination, or surrendering some issues of
sovereignty, as in base establishments.

Harkavy (I989) elaborates upon the fact that the establishment of bases erodes the
sovereignty of the minor power in such a relationship. On another note, examining
regime type in this manner allows us to get at a measure of policy coordination, albeit in
a rather rough manner. Bueno de Mesquita (1981) has argued that the tau-b measure of
alliance portfolios is a good means of measuring policy similarity, at least in regards to
security issues. In contrast to the tau-b approach, Signorino and Ritter (1999) have
devised a measure, S, to measure policy similarities between states. Unfortunately, both
of these measures use alliance configurations to explain policy coordination. In the study
being conducted in this dissertation, I find it problematic to use alliance configurations to
explain alliance behavior. Hence, I rely upon the regime measure as a means of
examining policy coordination among allies.

Proposition 3.0 (Domestic Political Constraints Proposition)

The likelihood of extended general deterrence succeeding decreases as the domestic
political constraints imposed upon the defender increase.

 

96 Jelavich ( 1973) and Langer (1950) provide excellent descriptions of these two alliances
and the policies that drove their formation.

I45

l ‘b

.I\ “it {I

rl-‘Ir’fliim
PH v'lt) \j
"r

T - '.'..
I'IL' Mt

I I
c] .,
dtsf’tihtd

There s

term at

polio; t

(b
:' 3
’73

truth.

fittest

Proposition 3.1

As the constraints placed upon the executive in a defender state increase, the greater the
likelihood that extended general deterrence will fail.

Proposition 3. 2

The likelihood extended general deterrence success increases if the government in the
defending state is a unified (single party) government.

Foreign policy decisions, as was argued in chapter 3, are not made in a vacuum.
There exist constraints upon leaders regarding the choices they make. The decision to
form an alliance, and protect a friend from aggression, is just merely one of the many
policy choices that state leaders face on a regular basis. In many democratic societies, for
example, foreign treaties are subject to some formal ratification process by the legislative
branch. As these restrictions increase on an executive, he or she is more likely to face
opposition to various policies that are proposed.

Alternatively, greater restrictions placed upon an executive indicate that greater
amounts of bargaining may transpire in order to have a treaty of alliance ratified. As the
United States discovered in 1954, there existed a fair amount of debate within the Senate
regarding the exact nature of the treaty of alliance to be signed with Taiwan. Such
domestic divisions do not exist within countries whose executives are less constrained,
and signal (in some cases wrongly), the lack of resolve on the part of a defending state.

In regards to the second proposition, a unified government may have more leeway
to make foreign policy decisions. This, I argue, is because different factions do not need
to be pleased. Therefore, I maintain that a single party government (democratic or non-
democratic) will have an easier time providing extended general deterrence, for it is

perceived that there are fewer obstacles in the govemment’s path.

146

Proposi

llflk’b

gori'r'l“

Pt I. u

Pr'mflbt
I

.hlhrl

I
7 l.‘ .
(VI-L’Alh
thur

‘ "
.1.‘ fit:
Lt'r
dual/U

m: -_

\L rill
other

ISCUR

Proposition 4.0 (Balance of Forces Proposition)

As the balance of forces favors the BC alliance, the greater the likelihood that extended
general deterrence will be successful.

Proposition 4.1

As the balance of short-term forces favors the BC alliance, the greater the likelihood that
extended general deterrence will be successful.

Proposition 4.2

As the long term balance of forces shifts and favors the BC alliance, the greater the
likelihood that extended general deterrence will be successful.

The balance of military forces between a potential aggressor and the alliance it is
facing is quite important in determining whether an initial challenge is made or not.
Military capabilities, as was argued in chapter 3, help to enhance deterrence credibility.
The ability to inflict punishment on an attacker tends to be a function of the forces that
the defender can marshal once a crisis erupts. Hence, the short-term balance of forces is
important in determining whether or not a crisis is initiated in the first place.

The second proposition, that dealing with the long-term balance of military
forces, addresses the resources that a state can draw upon over a long period of time.
These capabilities can help a state prevail in the case of a drawn out conflict, or a war of
attrition. In such instances, states must delve deep into the resources they have available,
other than manpower. The long-term balance of forces captures these resources, since it
is comprised of the industrial base of a state as well as the manpower it has in the
military.

Proposition 5.0 (Initial Beliefs Proposition)
As a potential aggressor ’s initial beliefs regarding the defender ’s type as being a

strongly committed defender increase, the greater the likelihood that extended general
deterrence will succeed.

I47

Promotion 3.

It defender
untried gent

Proposition 3

lie defender

grater the [1'

These
defender are
3H3 When it
initiate a Cris
to escalate a

henna] am

Uidlblliry 0
ln rh

come ‘0 the

doing the

lead [0 1933

 

Proposition 5.1

If a defender intervened on behalf of any ally in the past, the greater the likelihood that
extended general deterrence will succeed.

Proposition 5. 2

If a defender is actively involved in another dispute that doesn ’t involve its protege. the
greater the likelihood that extended general deterrence will fail.

These two propositions regarding the potential attacker’s perceptions of a
defender are critical for understanding when extended general deterrence will succeed.
and when it will falter. As discussed in chapter 4, many aggressors will be willing to
initiate a crisis against a target, given a favorable opportunity structure, but not be willing
to escalate a crisis if it will lead to a defender’s intervention. Hence, the beliefs that a
potential attacker holds are a good indication of how it views its international
environment, and when it is opportune to launch an attack on a target.

In the first proposition we are able to witness the credibility of a defender
regarding its other allies around the world. This keeps in line with Schelling’s (1960)
statement that the United States was willing to sacrifice troops in South Korea in order to
ensure that its commitment to West Germany was viewed by the Soviets as ironclad.
Willingness on the part of a defender to assist other allies in crisis bodes well for the
credibility of its commitment.

In the second proposition the concern is about the willingness of a defender to
come to the aid of its protége’ when it has already been involved in additional disputes
during the same year. “War weariness” may be the result on the part of the defender, and

lead to less effective deterrence at the general level.

I48

5.2.2 Propositions Regarding Extended Immediate Deterrence

Having outlined the testable propositions that are concerned with the success of
extended general deterrence, and how a deterrence commitment can be communicated as
credible, our attention now turns to a situation that emerges when a crisis erupts. Our
concern turns to ensuring deterrence stability; conditions under which the aggressor does
not fear some form of preemptive strike on the part of the defender. At the level of
general deterrence, more long-term strategic factors shape circumstances. When it comes
to the level of immediate deterrence, different factors come into play.

Extended immediate deterrence = f (k, d, x, ,u*) (2)
Where

k= balance of forces (immediate, short-term, long-term, nuclear capability).

d= domestic political costs of C.

x= issue at stake for A.

[1* = A ’s updated beliefs regarding C.

Once extended general deterrence has broken down, the defender must make a
decision as to whether it will intervene on behalf of its beleaguered protégé or not. As I
noted in the theoretical chapter, Fearon (1997) conceptualizes these actions on the part of
the defender as tying-hands signals. As I maintained above, I differ from Fearon in this
conceptualization in that I maintain that these signals also serve as sunk costs. In
particular, they have an impact upon a state’s reputation, both at home and abroad. These
reputation costs, although not modeled explicitly in this dissertation, impinge upon a
leader’s decision calculus.

These differences aside, Fearon and I are both addressing a similar idea. A

defender must use its actions to signal its intentions to the adversary. These signals can

range from doing nothing to full-scale armed intervention. I argue that as the defender

149

moreases its
possibility th'
the aggressor
Proposition

It the haiunt
immediate (1?

Proposition

Is the her/rm
extended im

Ptoposition

.45 the baian
extended im

Proposition

.43 the long 1
tritirhood rh

PlUposition

Capabilities

 

increases its willingness to fight, it signals these intentions by providing for the
possibility that the crisis may escalate to war. These tying-hands signals reveal its type to
the aggressor, and provide for stability in the immediate deterrence situation.

Proposition 6.0 (Balance of Forces Proposition)

As the balance of forces favors the BC alliance, the greater the likelihood that extended
immediate deterrence will be successful.

Proposition 6.1

As the balance of immediate forces favors the BC alliance, the greater the likelihood that
extended immediate deterrence will be successful.

Proposition 6. 2

As the balance of short-term forces favors the BC alliance, the greater the likelihood that
extended immediate deterrence will be successfitl.

Proposition 6.3

As the long term balance of forces shifts and favors the BC alliance, the greater the
likelihood that extended general deterrence will be successful.

Proposition 6.4

The likelihood of extended immediate deterrence being successful increases if the
defender possesses nuclear weapons.

Military capabilities, as discussed in the previous section, are inherently an issue
when it comes to analyses of deterrence. At the level of general deterrence, hostility
exists, but is not necessarily overtly expressed. As was argued in chapter 3, it is changes
within the aggressor’s environment that determine when it will exploit an opportunity.
In the immediate case, on the other hand, the concern is largely with how can a conflict
be won quickly, and goals attained. In such an instance, therefore, the role of military

capabilities, especially the immediate balance of forces, comes into play.

l50

The immediate balance of forces between the attacker and the combined strength
of the defender and protége' that are in place at the moment a crisis erupts are oftentimes
crucial to determining whether the demand of an aggressor are met, or if the demand is
met with resistance. For example, at the onset of the Second World War, despite their
intentions, France and Great Britain were in no position to intervene military in the
immediate aftermath of the German invasion of Poland. No British or French troops
were deployed to assist the Polish army in case of an attack. Hitler took advantage of this
weakness, and war on the Eastern Front began.

The short-term balance of forces comes into play once an attacker has made a
demand and moved militarily to secure this demand (i.e., seized the territory of the target
state). The short-term balance between the attacked alliance and the aggressor is
important withinthe first few months of conflict, as resources are mobilized and reserve
troops are called up. As the short-term balance shifts towards the embattled alliance. it
becomes increasingly unlikely that the attacker will be able to gain its objectives.

The long-term balance of forces comes into impact if a war becomes one
dominated by attrition strategies. It is dependent upon the resources that a state can call
upon for its war effort over a long period of time. While this balance should have a
positive impact on deterrence success, it should not be that large, given that states do not
tend to plan on fighting wars of attrition.

Finally we turn to a post 1945 phenomenon: the possession of nuclear weapons on
the part of a small number of states. Nuclear weapons are often credited with keeping the
peace between the United States and the Soviet Union dining the Cold War (Sagan and

Waltz, 1995), but they also demonized as being “terrors without deterrence” (Kugler,

ISI

1984). lmai
immediate d
is dirlirult to
nsis may 5

Glen the s

.45 the dome
extended im

Proposition

'H

the existent
that errende
PWposition

l or erirten

“It” likeliho
As
and desires
hill imesn
1‘ Some be
Sacrifice in
defended in

more ac

 

1984). I maintain that nuclear weapons should have an impact on the success of extended
immediate deterrence due to the fear that the crisis may escalate out of control. While it
is difficult to make the nuclear option credible (Powell, 1990), there is the chance that a
crisis may spin out of control and cause such weapons of mass destruction to be utilized.
Given the stress that is often present during times of crisis, and potentially faulty

decision-making procedures, this proposition appears logical.
Proposition 7.0 (Domestic Political Constraints Proposition)

As the domestic political constraints increase for C, the greater the likelihood that
extended immediate deterrence will be successful.

Proposition 7.1

The existence of a democratic regime in the defender ’s state will increase the likelihood
that extended immediate deterrence will be successful.

Proposition 7.2

The existence of a unified (single party) government in the defender ’s state will decrease
the likelihood that extended immediate deterrence will be successful.

As was argued in chapter 3, there exist costs that emanate from the preferences
and desires of domestic political audiences. In the theory deveIOped here, I maintained
that investing in a protége' generates costs through a guns and butter argument. As long
as some benefits is seen in providing “guns” to an ally, the people don’t mind the
sacrifice in “butter”. However, if assistance is provided abroad, and the interests are not
defended in time of crisis, then the domestic audience views the assistance as being

wasted. Hence, I argue, contrary to what many would think, that increased domestic

pressure actually increases the effectiveness of extended immediate deterrence.

152

Hg“:
for 01‘5““r
discerned Cl
(SUPREME 3
the} £0“le
to? 1“ k"
[0 hate 3 P0
Proposition

it .t '5 He‘ll“

dr‘fr'mllt‘t i
Pr. position
ifthe issue c
immediate d

Terr
inremationa
dissertation
of a potenti:
Congress 0
an of the t
present Ind

Giv
1870. 1 mail
hat crises t

tailore of e

 

However, a single party government may be less likely to intervene in a crisis. for
fear of becoming entrapped in an undesired conflict. As Prins and Sprecher (1999)
discerned, coalition governments within parliamentary democracies are more prone to
reciprocate and escalate militarized disputes than are single party governments. This.
they conjectured, was due to the ability of the government to place blame with “the other
guys”. In keeping with this finding, I argue that unified governments will be less likely
to have a positive impact on extended immediate deterrence success.
Proposition 8.0 (Issue at Stake Proposition)

As A ’3 value for x increases, the greater the likelihood that extended immediate
deterrence will fail.

Proposition 8.1
If the issue at stake is territorial in nature for A, the greater the likelihood that extended
immediate deterrence will fail.

Territory, as was discussed in chapter 3, has been an underlying issue of
international conflict for much of human history. Many of the alliances examined in this
dissertation were formed for the purpose of preventing territorial irredentism on the part
of a potential aggressor. Serbia, for example, sought Austrian protection after the
Congress of Berlin in 1880 due to Bulgarian designs on Serbian territory. Or, in the latter
part of the twentieth century, Great Britain’s alliance with Malaysia was an attempt to
prevent Indonesian aggression against the newly formed Malay state.

Given the centrality of territory to so much of conflict that has transpired since
1870, I maintain that territory is a major issue at stake in many crises. Thus, I maintain
that crises that evolve due to territorial issues will be more likely to escalate, and lead to a

failure of extended immediate deterrence.

153

Proposition

entitled rmr:
'l
i

reposi'nm '

or promo:

stormed :nrr.
Pnposrtnrrr :

It (proctor

i’lt .:"L.'..’Illitt/
D i v
i ".ip'ti‘ir'rlm

PC is intuit
dtir‘rrem-e t1

P“position

It , r
h defender

In or
toll aetualls
“5 discugg‘

us.L'e‘ssor h;

Pie'ittt 3X1

 

Proposition 9.0 (Updated Beliefs Pr0position)

As A ’s beliefs about C being a strong defender increase, the greater the likelihood that
extended immediate deterrence will succeed.

Proposition 9.1

If C practices a policy of tit-for-tat military escalation, the greater the likelihood that
extended immediate deterrence will succeed.

Proposition 9.2

If C practices a policy of firm-but-flexible diplomatic response to the crisis, the greater
the likelihood that extended immediate deterrence will succeed.

Proposition 9. 3

If C is involved in an ongoing dispute, the greater the likelihood that extended immediate
deterrence will fail.

Proposition 9.4

If a defender intervened on behalf of its protége' with regards to the potential aggressor
in the past, the greater the likelihood that extended immediate deterrence will succeed

In an extended immediate deterrence crisis, the beliefs of what actions a defender
will actually commit to have a great impact on whether the crisis escalates or not. This
was discussed in chapter 3, and formalized in chapter 4, where the beliefs that an
aggressor has concerning the type of defender were of some consequence for determining
whether extended general deterrence succeeded or failed. In the cases in which general
deterrence faltered, the actions the defender took allowed the attacker to update its beliefs
regarding what type of defender it was actually facing. The pr0positions outlined here
account for the updating on the part of the attacker.

The first two pr0positions are concerned with the military and diplomatic bargaining

strategy that the defender uses when a challenger has made a threat to alter the status quo.

154

ot‘eaotion.
responds I
mohilizati
but it may

loll mobil'

 

The concern on the part of the defender is how to make the aggressor perceive the
alliance commitment as credible, but also keep the commitments stable as well.

In the case of the first proposition, regarding military escalation behavior, the
defender has three alternatives to choose from. It can adopt a policy of strength, a policy
of caution, or a policy of tit—for-tat.97 If a defender adopts a policy of strength, it
responds to provocation on the part of an aggressor with overwhelming military
mobilizations. Such a policy may lend credibility to a defender’s commitment to its ally,
but it may also risk inflaming the aggressor, and leading to a preemptive strike, prior to
full mobilization of the defender.

Alternatively, a defender can adopt a policy of caution, much as the French did
during the Munich crisis. A defender can choose to cautiously intervene in the crisis,
which may prevent harm from befalling the defender, but also leave its credibility open to
question.

Finally, a defender can adopt a policy of tit-for-tat, in which it responds in a
reciprocal fashion to the actions of an aggressor. As Huth (1988) has argued, this sort of
policy combines both deterrence stability and credibility. It allows the defender to signal
its intentions regarding its ally, yet it does not so overwhelm the attacker, and provides
the attacker with a means of withdrawal, that deterrence stability is preserved. Hence, in
the empirical model in the next chapter, I examine the effect that a tit-for-tat policy has

on extended immediate deterrence success.

97 . . . . . . . . . .
The discussrons of military escalation polrcres and bargaining techniques are drawn

from Huth (1988); Gochman and Leng (1983); Leng (1984, 1993); and Leng and

Wheeler (1979). For detailed discussions on a strategy of tit-for—tat and its ability to elicit

cooperation, see Axelrod (1984).

155

0f
processes
made in f

pursue; a b

A b

note 0

difficult lo

attacker.

 

Of course, military strategy tends to be coupled with diplomatic bargaining
processes as well. Borrowing again largely from Huth (1988; see also the references
made in footnote 76), I denote three distinct diplomatic strategies that a defender can
pursue; a bullying strategy, a conciliatory strategy, and afirm-but-flexible strategy.

A bullying strategy adopted by a defender indicates that it is not willing to
negotiate on any issue. This may enhance the credibility of the defender, but makes it
difficult for any compromise to be arrived at, and may invite escalation on the part of the
attacker.

Alternatively, the defender could adopt a conciliatory strategy, in which it makes
no demands of the aggressor, and accedes to the demands made at the onset of a crisis.
Such a policy does nothing to ensure credibility or stability. It actually makes the
defender appear weak and irresolute. and may open the door to more challenges in the
future.

Again, as Huth (1988) discerned, the strategy that offers the greatest hope for
extended immediate deterrence success is a firm-but-flexible response to the actions of
the aggressor. A firm-but-fiexible strategy allows a defender to stand up to the demands
made by an aggressor, but also to introduce some bargaining leverage in other areas of
concern, in an attempt to entice the aggressor to compromise on one issue while gaining
leverage on something else.

The third proposition above is concerned with an attacker’s perception of a
defender’s willingness to fight, given that the defender is currently involved in an

additional dispute. As noted in the analysis of extended general deterrence above, the

156

likelihood of a defender intervening is reduced if it is already involved in an ongoing
dispute elsewhere.

In the fourth and final proposition above, the past behavior of a defender
regarding its protege and attacker should have a positive impact on general deterrence
success. The willingness of a defender to intervene in previous disputes between an
attacker and its protégé send a signal that it values its relationship with its ally quite
highly.

Having described the testable pr0positions that can be derived from the formal
models in chapter 4, I now turn to a discussion of how to operationally measure variables
that can be used to empirically test these propositions. Before proceeding into the details
of measurement, however, some discussion is needed that frames the set of cases that
would comprise a good scientific study of extended deterrence via alliance. In this vein, I
turn to a delineation of the role that selection bias plays in studies of extended deterrence.
Failure to account for this issue can lead to many problems in the statistical estimation of
the propositions outlined above. Since selection effects are so germane to the study of
deterrence, and figure prominently in the criticisms of those that challenge rational
deterrence theory, I feel it necessary to elaborate on the issue in some detail.

5.3 The Problem of Selection Effects in Analyzing Deterrence Situations

As was noted in chapter 3 in the theory specification, and in chapter 4, where the
formal models were deveIOped, the crisis sequence can be viewed as a process. In the
context of the theory developed in this dissertation, crisis escalation is dependent upon

extended general deterrence failing in the first place. Since the theory developed herein

157

examines the impact of a deterrence alliance at both the onset and escalation of a crisis,
great care must be taken in order to prevent bias from entering into the analysis.

Problematic here is the issue of selection bias, which was touched upon in the last
chapter”. In this section and the following I discuss the role that selection bias plays in
the study of extended deterrence relationships.99 In the following chapter, when I present
my statistical results. I discuss how standard econometric techniques fail to account for
these selection effects, and provide an alternate way of examining a model of extended
deterrence via alliance that accounts for both crisis initiation and escalation.

One of the greatest dangers that plagues the study of deterrence is that of selection
bias. Selection bias can best be defined as selecting on the dependentvariable (Geddes,
1991). Selection bias is not only a problem in deterrence situations. but manifests itself
quite clearly in this strand of research. ‘00 The problem is that one finds it difficult to
identify cases of successful deterrence. We witness deterrence failures. Challenges are
made, counter-threats posed, a challenger escalates or back down. Deterrence failures
correspond nicely with crises. What we don’t see are challenges not made. Leaders tend
to be loath to admit that an adversary has deterred them, for it invites more of the same.

For example, it would be difficult to imagine the Soviet leadership publicly announcing

 

98 This issue of selection effects is a current topic of interest in the international relations
literature. See, for example, Achen and Snidal (1989); Gartner and Siverson (1996);
Fearon (1993, 1994a); and Smith (1996, 1997).

99 Throughout this chapter and the next I use the terms selection bias and selection
effects interchangeably. They refer to the same phenomenon.

'00 Geddes (1991) deals with selection effects in the context of comparative politics, and
Dubin and Rivers (1989) develop their bivariate probit model with selection with direct
reference to the study of American voting behavior.

158

that hey

Am neon

l

Poi
pro lemsi
me toned
.tn . in th
eqi libria
pn lems
do nence
tes ;. and
5 1 Se

As
all ough

at isis.

 

that they had ceased making demands regarding Berlin because of the effectiveness of the
American deterrence commitment!

Failing to account for the possibility of selection bias can lead to a variety of
problems in the study of extended deterrence via alliance. Some of these concerns were
mentioned in chapter 2, in the debate surrounding the role of rational deterrence theory.
And, in the previous chapter, issues of selection bias were noted in discussing the
equilibria of the two formal models. In the following two sections I elaborate upon the
problems that selection bias play in both game-theoretic and empirical analyses of
deterrence. The focus in both sections is on how to control for their effects in empirical
tests, and how the two methods of analysis are related.

5.3.1 Selection Eflects and Theory Development

As Fearon (1993, l994a) demonstrates, studies in extended immediate deterrence,
although based on classical rational deterrence theory, do not account for the initiation of
a crisis. Studies in immediate deterrence fail to account for the initial challenge to
general deterrence. As his work has noted, the factors that are operative at the level of
general deterrence (i.e., domestic political costs) may have a reversed impact at the
immediate deterrence level.

Crises, in F earon’s conception, operate as a set of costly signaling encounters.
States can. and will, select themselves in and out of crises depending upon the value they
place on the issues of stake. His reanalysis of Huth and Russet’s (1984, 1988) data
confirms some of his theoretical predictions. Hence, there is a sort of self-selection

occurring here, and is corroborated by the equilibria from the models in chapter 4.

159

Fearon’s conception of crisis is one of crisis bargaining, in which each stage of
the crisis is an exchange of costly signals. At each encounter in the crisis process. states
reveal information regarding their resolve through the actions they take (or don’t take).
His focus on signaling differs somewhat from mine. I argue that when an alliance is
formed, defending states send signals through the mechanisms of formation that allow
potential aggressors to determine the credibility of the commitment. These signals form
the prior beliefs of an aggressor, and subsequent moves on the part of the defender allow
the aggressor to update his belief structures.

Hence, my model is an extension of Fearon’s work in that I allow for a signal to
occur prior to the onset of a crisis, and also allow updating throughout the crisis to take
place. This modeling technique allows for selection bias as Fearon describes it to be
accounted for, and, as I have argued, allows for a more accurate depiction of a deterrence
encounter.

5.3.2 Selection Effects and Empirical Analyses

Turning from the game-theoretic concept of selection bias, we now address the
role that selection bias plays in empirical analyses. As Geddes reminds us, the cases you
choose affect the results you get. In the situation being analyzed here, the problem is
magnified. First of all, how do we know if deterrence is the real purpose of the alliance?
Secondly, what is the relationship between an extended general deterrence failure and an
extended immediate deterrence failure? Third, at an even more basic level, how do we
know that deterrence is the actual reason for the absence of any crisis between
adversaries? This is not a problem for merely deterrence theorists, but all who study

international politics.

160

'3.)
11“»

t“l
Wit

SDI

it:

so

no

“I

St

As Most and Starr (1989, 47) point out, analysts who study interstate conflict
need to account for selection effects in their research strategies. Cases of non-initiation
tend to be coded as missing data. If we identify conflict by determining all cases in
which violence occurred and work backwards, we are not considering cases in which
conflict could have occurred but didn’t. Hence, selecting on the dependent variable
(conflict), or, in the cases being analyzed here, extended general deterrence failure. will
lead to results that simply may be wrong.

From a purely theoretical point of view, this issue of selection bias has plagued
studies of deterrence, and could potentially play a confounding role in the study of
extended deterrence via alliance as well. To counter this potential problem, I follow
some advice given by two analysts concerned with the role that selection plays in
determining cases of attempted deterrence.

Achen and Snidal (1989), in their discussion of deterrence and comparative case
studies, maintain that one needs a sound theoretical reason for determining a population
of cases. In that vein, I examine all cases in which a great power defender has extended a
policy of deterrence to a protege. This allows me to then examine both cases in which a
dispute was initiated, and when it was not. It also provides me with a set of cases in
which targets have alliances with other states, and cases in which they do not. Thus, the
selection bias inherent in many studies of deterrence can be corrected for. By examining
all cases in which alliances were formed, and examining those cases in which extended

general deterrence faltered, as well as those cases in which deterrence was not
V challenged, I am able to eliminate much of the bias that could be introduced into this

study. This allows me to examine the processes by which extended general deterrence

161

sitntions r
examine th
5.4 Ih
POI
deterrence
identified
minor poo
three on rt
hetoeen t
manner in
entire so
made on

defensive

all 35m

(Ott‘ da

r\
Dam
(hill? st
Benita:
Inform}
Dias} Ill

situations move into situations of extended immediate deterrence. I can therefore
examine the decisions made at each point, as the crisis escalates.
5.4 The Set of Alliances

Following the lead of Achen and Snidal (1989) mentioned above. I outline a set of
deterrence alliances using theoretical criteria. I maintain that a deterrence alliance can be
identified in the following manner. First, there must be some identifiable threat to the
minor power that it desires protection from. This can be the result of threats or uses of
force on the part of the potential aggressor, or the product of long-running rivalry
between the twos states. Secondly, the great power defender must make it clear in some
manner that the purpose of its alliance with the minor power is to deter aggression. Thus.
public statements, as the British made on behalf of Poland in 1939, or explicit promises
made within the treaty of alliance, as in the NATO treaty, stipulate the
defensive/deterrent nature of the alliance.

Using data from the Correlates of War (COW) project, I have identified the set of
all asymmetric alliances in the period spanning 1870-1984. The sources for identifying
the alliances come from Singer and Small (1966a, 1966b, 1968), Small and Singer
(1969), with additional data being drawn from Oren (1990). Bennett (1997) updated the
cow data through 1984.‘°'

The question of when to begin analysis is one that continually arises in studies of

international politics. A variety of watershed events exist, and all of them could provide

 

. '0' Data was made available to me by D. Scott Bennett, and is available on his website

(http://www.personal.psu.edu/faculty/d/s/dsb10/), This is the data that was used in
Bennett (1997), and extends the COW alliance data to 1984. Additionally, the data set
incorporates some discrepancies noted by Oren (1992), and codes for Morrow’s concept
of asymmetry as well.

162

a starting point for this study. In using the Correlates of War data, I did have information
on alliances going back to 1816. My reason for starting my analysis at 1870 is largely
due to data availability regarding military bases, arms transfers, and data on international
trade. Data that is available after 1918 is much more accurate than data from the pre
World War I era, but through exhaustive consultation of various sources (outlined in the
coding rules of the independent variables below), I feel confident in the data that I have
assembled that goes back to 1870.

In narrowing down the set of alliances, I adopted the following coding rules. First
of all, due to the fact that the nature of the theory being examined is defensive, I included
only alliances that were coded by the COW project as defense pacts or ententes. These
are the two types of alliances in which some form of military cooperation in case of
conflict is pledged.

Secondly, I examined the context within which the alliance was formed. This
involved reading the treaties that were signed and seeing if they specified a specific
threat. Unfortunately, many of the treaties are quite vague, and secondary sources were
necessary to establish the contextual situation in which the alliances were formed.

Sources for the treaties included Grenville (1974), Grenville and Wasserstein ( l 987) and
Hurst (1972), which are comprehensive compendituns of international treaties from 1815
onwards. Additionally, Langer (1950), Osgood (1964, 1968), and Pribham (1921)
possessed commentary on the circumstances surrounding the formation of many of these
alliances.

Attempting to determine the potential aggressor against a target was an even more

difficult task. For the period of 1870 —1890, Langer (1950) was a primary source, at least

“

I63

within a European context. Other cases were a bit more difficult. I relied to a great deal
on the works of Dupuy and Dupuy (1970), and Langer (1972) as very broad, general
military histories of the world. Other more specific sources on each alliance are
mentioned in Appendix D, where I describe each alliance in some detail.

Third, I had to determine a unit of analysis. While a majority of the alliances
analyzed (17/20) were bilateral in nature, three were multilateral. Alliances that were
multilateral presented a more complicated task. Should they be disaggregated into dyadic
components, or should they be treated differently? These three multilateral alliances
were decomposed into dyads. The reasoning behind this is twofold. First of all, the
theory as presented poses a dyadic (actually, it is triadic in nature if one accounts for the
potential aggressor) specification. Thus, it seems natural to arrange the data in a dyadic
manner. Secondly, and one that is more consistent with historical reality, in the cases in
which there is more than one signatory to the alliance, there are also separate alliance
treaties binding the signatories in a dyadic fashion. For example, in the treaty that joins
the Romania with Germany and Austria-Hungary in alliance, there are separate treaties
between each major power and Romania.

Third, since the focus of this dissertation is on deterrence, I had to ascertain the
time frame in which extended deterrence was one of the goals of the defending state.
This involves, in four of the cases, censoring the data. For example, the Soviet Union
and Mongolia were united by a defense pact from 1936 until 1984. Prior to 1945, when

the Second World War came to a close, the pact was directed against Japanese
’ aggression. However, the Japanese threat after the war was replaced by a threat from the

People’s Republic of China, who laid claim to territory on the two countries’ borders. In

164

1962, Mongolia and China signed a treaty ending their territorial dispute. However, the
Soviet-Mongolian alliance persisted, even though the contentious issue had been
removed. Given the settlement of the issue that threatened the peace between the two
neighbors, I code the deterrent component of the alliance ending in 1962. '02

Having coded the alliances in the above described manner, I arrived at a set of
twenty alliances that were formed with the express notion of a defender providing
extended deterrence benefits to the protege. Table 5.1 denotes the set of alliances, the

duration of the alliance, the potential aggressors, and number of disputes, if any.

 

‘02 Any censoring decisions, and justifications for them, can be found in the descriptions
provided in Appendix D. The other cases are the United States and West Germany
(1954-1972), the United Kingdom and Egypt (1936-1943), and the United Kingdom and
Jordan (1949-1957).

165

ALLIANCES IN WHICH EXTENDED DETERRENCE IS ATTEMPTED

TABLE 5.1

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1870-1984
Defender Protégé Aggressor Duration of Number
Alliance of
Crises

United Kingdom Turkey Russia 1878-1880 0
Austria-Hungary Romania Bulgaria 1883-1916 1
Germany Romania Bulgaria 1883- l 9 1 6 1
Austria-Hungary Serbia Bulgaria 1881-1895 1
Russia China Japan 1 896- 1 900 0
France Belgium Germany 1921-1936 1
France Poland Germany 1921-1939 1
France Czechoslovakia Germany 1924 0
France Czechoslovakia Germany 1925-1939 2
Italy Romania Bulgaria 1926-1930 1
Italy Albania Yugoslavia 1927-1939 1
United Kingdom Egypt Italy 1936-1943 5
Soviet Union Mongolia Japan(until 1936-1962 4

1945)

China (1945-

1962)
United Kingdom Poland Germany 1939 1
United Kingdom Jordan Israel 1949-1956 2
United States South Korea North Korea 1954-1984 10
United States Taiwan China 1954-1979 4
United Kingdom Malaysia Indonesia 1957-1971 5
Soviet Union Vietnam China 1978-1984 3
United States West Germany Soviet Union 1954-1972 3

 

 

 

 

 

 

From this I was able to arrive at 302 years in which eXtended deterrence via
alliance was attempted. This observation, the alliance year, is the underlying unit of
analysis for the empirical studies conducted in the following chapter. A full description
of each alliance, its duration, its reason for existing, and dispute involvement can be
found in Appendix D.

5.5 The Set of Crises: Measuring Deterrence Failures

The identification of cases of extended general deterrence provides the researcher

with a dilemma. How does one know if deterrence was actually effective? Cases of

I66

deterrence failure are easier to identify. A demand is made, an attack is committed. a
crisis is escalated. However, as the central question in this dissertation states, we are
interested on determining the impact which existing alliances have on the initiation and
escalation of crises. But alliances often have deterrent'effects. Thus, we don’t witness
dispute initiations when deterrence succeeds. If we focus only on cases in which a
dispute is initiated. we are introducing selection bias into our sample. as was discussed at
length above.

The statistical framework adOpted in the next chapter helps control to a great
extent the problems of selection bias that often emerge in analyses of deterrence.
However, we need to also take care and make certain that our population of crises is as
accurate as possible. The above section identified the strategies that were undertaken in
identifying the alliances that were formed with an extended deterrence component. In
this section I describe the manner in which 1 identified the challenges to extended general
deterrence success, and what occurs when extended general deterrence falters.

5.5.1 Identifying Cases of Extended General Deterrence Failure

The first task is to define what a case of extended general deterrence failure looks
like. I adopt similar coding rules to those found in Huth (1988) and l-Iuth and Russett
(1993). A case of extended general deterrence failure can be viewed as occurring as long
as the following criterion is met:

A threat has been made by a potential aggressor against the target. It must be a clearly
defined, explicit threat, with the actual use of or threat to use force and the target must be
' clearly identified. Such actions include one or more of the following:

a. Statement of intent to use force.
b. Buildup of military forces for potential use.

167

c. Preparation of military forces for imminent use.
d. Actual use of force directed against the target.

In order to identify the set of cases of extended general deterrence failure; I
utilized the Militarized Interstate Dispute (MID) data set, as developed by Gochman and
Maoz (1984) and updated by Jones, Bremer, and Jones (1996). According to Gochman
and Maoz, (1984, 587), "a militarized international dispute is a set of interactions
between or among states involving threats to use military force, displays of military
force, or actual use of military force". This corresponds nicely with Snyder and
Diesing’s definition of crisis in chapter 3. This data set spans the years 1816-1992, and is
the most encompassing data set of its sort in the quantitative international relations field.

Unfortunately, the size of the data set also causes a few problems. As the recent
study by Jones. Bremer and Singer (1996) shows, in the years 1816-1992 there have been
at least 2000 MIDs.103 Every single dispute, however, does not threaten a state in a
similar manner. Due to this large number, I chose to focus this study on those disputes in
which force was threatened, displayed, or used at the beginning of a crisis. A threat to
use force is defined as a verbal indication of hostile intent; a display of force involves
military demonstration but no actual combat; and a use of force is one in which active
military operations are used. This includes blockades, border clashes, occupation of
tenitory. seizures of personnel or material, clash, raid, declaration of war, or use of CBR
weapons (Jones, Bremer, and Singer, 1996, 171-173).

In identifying cases of extended general deterrence failure, I first went through the

MID data and identified all cases in which hostility existed between the protége' and the

 

‘03 I say “at least” because data sets are continually updated, and future work may expand
(or decrease) the number of disputes which have occurred since 1816.

168

potential aggressor. I then turned to more detailed sources to ensure that the dispute
actually appeared to be a crisis of some sort, in which basic values were threatened on the
part of the protege. To assist in identifying in more detail these crises, I turned in part to
the ICE data of Brecher and Wilkenfeld (1997) for theiperiod 1918-1984, Tillema (1991)
for events in the post 1945 era. and to Dupuy and Dupuy (1970) and Langer (1950, 1972)
for more general military histories. Additional information was garnered from Small and
Singer (1982), eSpecially in identifying what they term extra-systemic wars'04
Once I determined that the aggressor made a serious threat towards the protege, I
coded extended general deterrence as failing. I coded this on a yearly basis. For crises
that spanned years, I coded extended deterrence as failing in the first year when the crisis
was initiated, and coded it as failing in the second year as well. This is the first dependent
variable (EGDSUCC), with extended general deterrence success being coded as a 1. and
extended general deterrence failure being coded as a 0. I identify 302 cases from 1870-
1984 in which extended deterrence via alliance has been attempted. In these cases, there

exist 257 cases of extended deterrence succeeding and 46 cases in which it fails. Table

5.2 provides some summary statistics regarding this dependent variable.

 

 

 

 

 

TABLE 5.2
FREQUENCY DISTRIBUTIONS OF EXTENDED GENERAL
DETERRENCE SUCCESS
EGDSUCC Frequency Percent Valid Cumulative
Percent Percent

0 (failure) 46 1 1.1 15.2 15.2

1(success) 256 62.0 84.8 100.0
Total 302 73.1 100.0

 

 

 

 

 

 

 

 

‘04 An extra-systemic war is defined by Small and Singer (1982, 52) as a war that exists
between a system member and a political entity that is not construed as a system member.

169

5.5.2 Identz'fi/ing Cases of Extended Immediate Deterrence Failure: The Possibility of
Escalation
Cases of crisis escalation, or extended immediate deterrence failure, provide a
more difficult problem in regards to identification and coding. The MID data codes for
highest level of hostility, but does not code for initial use of force, nor does it code for
escalation dynamics. '05 To construct my variable for extended immediate deterrence
success, I had to turn to other sources than the MID data set. In Appendix D, along with
my descriptions of the twenty alliances that comprise my data set, I also discuss the crises
(if any) in which they become involved. In this appendix I provide bibliographic
information regarding the specific crises. Much of this is taken from the sources
mentioned above, and others are taken from more specific histories of the crises and wars
that surround them.
Adopting the coding rules of Huth (1988, 26-27). I define extended immediate
deterrence success/failure in the following manner. First of all, extended immediate
deterrence succeeds when the attacker refrains from using military force against the

target, or the combat is limited to a small-scale use of force (under 200 casualties). If

 

'05 The MID coding for level of hostility is as follows 1) no militarized action; 2) threat to

use force; 3) display of force; 4) use of force; 5) war (Jones, Bremer, and Singer, 1996).
Carlson (1998) examines crisis escalation within the context of extended deterrence, and
utilizes the MID data, along with data from Huth (1988), in part to code for escalation
processes. I argue that while one can assume that the failure of extended immediate
deterrence as coded by Huth (1988) implies escalation occurred, it is better and more

. accurate to examine the sequence of events that took place surrounding an individual
cnsrs.

170

extended immediate deterrence is successful, the attacker does not attain the goals which
it had when initiating the crisis in the first place.

I code extended immediate deterrence as failing when a large-scale use of force
on the part of the attacker is launched at the protege (totaling more than 200 fatalities in
all parties involved). or the demands of the aggressor are acceded to by the protege and/or

defender.

This leads to my second dependent variable, extended immediate deterrence

success (EIDSUCC). I code the variable as a 1 if extended immediate deterrence
succeeded, and 0 if it failed. Below in Table 5.3 I provide some summary statistics on

the 45 cases of extended immediate deterrence that I have identified for the time period

 

 

 

 

 

 

 

 

 

 

 

 

 

under analysis.
TABLE 5.3
FREQUENCY DISTRIBUTIONS OF EXTENDED IMMEDIATE DETERRENCE
SUCCESS
EIDSUCC Frequency Percent Valid Cumulative
Percent Percent
0 (failure) 19 4.6 41.3 41.3
1 (success) 27 6.5 58.7 100.0
Total 46 11.1 100
5.6 Operationalization of the Independent Variables: Measuring influences on
Willingness

Having described how my dependent variables are measured, 1 now turn to a
discussion of the variables that impinge upon extended deterrence success and failure, at

both the general and immediate levels. In each sectionl discuss the variables that

171

measure the concepts that were introduced as parameters in the formal models in chapter
4.
5.6.] Security Benefits (Invested Costs)

The following paragraphs discuss the variables that measure the concept of
security benefits, or invested costs, that a defender has proffered to a prote’ge’.
Defense Pact

Capabilities may reflect a state’s ability to ward off attack, or defend itself or a
protege, but it needs to signal its value in its ally in some manner, or it will continually be
using its capabilities to fight. As was argued in the theoretical portion of this dissertation.
the signing of an alliance treaty binds two states together for military/security purposes,
and also sends a signal regarding the defender’s interest in its protege. This signal, I
argued above, functions as a security benefit for the protege.

In keeping with this line of reasoning, I limited myself to two classes of alliances,
as discussed above. I then coded whether each alliance was a defense pact or an entente.
I coded the variable DEF PACT as a dummy, with the variable taking on a value of 1 if
the alliance was a defense pact and 0 otherwise. This data was taken from Bennett
(1997)

Arms Transfers

 

“Arms transfers and arms production have appeared as an inevitable concomitant
to war and military preparation throughout human history” (Krause, 1992, 34). The
transfer of armaments between allies serves a threefold purpose. First, it provides

additional security to the smaller alliance partner. Secondly, the transfer of arms

I72

provides a signal of the value defender places on a protege. Third, the arms trade allows
the defender to exert some degree of influence on the protege (Sislin. 1994).106

To construct this variable I examined conventional arms transfers from the
defender to the protégé. I coded arms transfers in the following manner. I took the total
number of conventional arms imported by a protege from the defender. and divided this
total by the total number of arms the protege imported from all countries. The variable
ARMS is measured on an annual basis.

I utilize a measure based upon percentage rather than number of weapons systems
transferred or conversions into constant dollar amounts for two reasons. First of all, I
believe that the percentage measure demonstrates the degree of dependence that the
protégé relies upon the defender for armaments. Secondly, the use of number of arms
transfers, as Kinsella and Tillema (1995) and Schrodt (1983) note, is more of a measure
of political influence. Given that my theoretical question is on the provision of security
rather than the degree of political influence that the defender has With the protege, I feel
that the measure constructed here is appropriate.

Given the difficulties of locating arms transfer data, I have utilized a variety of
sources.’07 In the post 1945 era I have relied heavily on SIPRI (1975), as well as the
SIPRI Yearbooks (1961-1984). In order to ensure validity and reliability across years, I
also used data from the US. Arms Control and Disarmament Agency’s World Military

106 - - -
A good example of the use of arms as a means of political pressure can be seen in the

British embargo of arms to Jordan in the aftermath of the 1948 Arab-Israeli War. Britain

refused to send arms to Jordan until certain conditions were met within the Jordanian

government. On this topic, see Pundik (1994).

'07 Brozska (1982) provides a thoughtful insight into the problems of finding arms
transfer data. Nearly twenty years later, the issues he raises are still germane.

I73

Expenditures and Arms Transfers data (1963-1984). When it came to data from the pre-
World War II era, 1 utilized a diverse set of sources. The Statesman ’s Yearbook (I 886-
1939), Harkavy (1975), Kolodziej (1989), Krause (1992), and Laurance (1992) were all
additional sources that I utilized in constructing this indicator.

gigs

Military bases established on the territory of a protége' combine both elements of
security and autonomy. They permit a great power defender greater leeway for pursuing
international objectives, while providing a greater degree of security for the host country.
Witness, for example, the American naval bases established on the island of Taiwan. The
ability of the American Seventh Fleet to dock and base itself at various ports in Taiwan
allowed for a more rapid American response in the various Taiwan Straits crises, as well
as allowing the United States to maintain a presence in the Far East and pursue a variety
of foreign policy goals in the region.

Due to the fact that there are distinctions between functions that bases serve (see
Harkavy, 1982, ch. 2 on the different roles that bases play), it would be ideal to construct
separate indicators for different base types. However, many bases have a tendency to
serve dual functions (naval Operations and intelligence gathering, for example). Due to
these problems, I do not differentiate between different types of bases. Rather, I
construct an indicator that merely signifies if bases are present or absent year to year
within the protégé.

The variable I construct for bases, BASES, is coded as the number of bases in any
given year that a defender has established on a protégé’s territory. I code this variable on

a yearly basis. For example, Russia and China are allied beginning in 1896. but the

174

Russians do not establish a military base at Port Arthur in China until 1898. Thus, the
first two years of the alliance are coded as 0 for the existence of Russian bases
on Chinese soil.

Data for this variable were drawn primarily from Harkavy (1982, 1989).
Additional information on US. bases in the Cold War era was drawn from Duke (1989)
and Duke and Krieger (1993). For years prior to the First World War these data were
drawn from a variety of diplomatic histories, including Albrecht-Cane (1973), Craig
(1978), Kann (1974), and Jelavich (1973).

5.6.2 Autonomy Benefits

The following paragraphs discuss the benefits that a great power defender is able
to gain from being allied with a smaller power. This is the autonomy portion of the trade-
off between security and autonomy in deterrence alliances.
£229.11

Trade between a defender and a protégé serves as another indicator of the value
the defender has in the protégé. As 1 hypothesized above, the more a defender exports to
a protégé, the greater the likelihood that extended general deterrence will be successful.
Since most minor powers can offer little to their great power patrons in terms of security.
they need to provide other goods that will make them more attractive as alliance partners.
Increased access to foreign markets is a significant autonomy benefit for a defender, and,
as mentioned above, has a tendency to create security extemalities as well.

I constructed my variable EXPORT as a measure of the percentage of exports that
a defender exports to its protege. This is calculated as the ratio of exports of the dyadic

trade between protege and defender over the total and exports of the defender. This

175

variable is derived as a yearly measure. Data was taken from a variety of sources. . I

utilized the Statesman 's Yearbook (1885-1939), the IMF Direction of Trade (1950-1984).
Mitchell (1975), the League of Nations International Trade Statistics ( 1 933-1 93 8), and
the United Nations Yearbook of International Trade Statistics (1950-1984).

Stategic Location

Oftentimes, the protége' in an alliance is strategically located in regards to the
interests the defender is pursuing. 1 code a protege as being strategically located if it
meets one of three criteria. First of all, is it contiguous to the territory of the defender?
Secondly, does the protégé border upon the colonies of a defender? Third, does the
protege control access to major trade routes or choke points on these trade routes? 1 code
the variable STRALOC as 1 if it is strategically located. and 0 if the protege is not
strategically located.
Strategic Raw Materials
Additionally I code whether or not a defender is importing strategic raw materials

from the protege. This variable, STRAW, is coded on an annual basis, and is coded as a
the percentage of strategic raw materials the defender imports from its protége’ divided by
the total number of strategic raw materials the defender imports. Strategic raw materials
in the pre World War I era include coal, iron ore, petroleum, copper, nickel, lead, and
sulfur. In the interwar period cotton rubber, bauxite, manganese, and chromium are
added to the list. In the aftermath of the Second World War titanium and cobalt are
additionally coded as being strategic materials. All of these factors can be used to

increase military production. I coded this variable on an annual basis. Sources include

the Statesman ’s Yearbook (1885-1984), Mitchell (1975), the League of Nations

176

International Trade Statistics (1933-1938), and the UN’s International Trade Statistics
(1950-1984).
Democratic Alliance

I code whether an alliance is a democratic alliance or not by examining if both the
defender and the protége' are democratic or non-democratic. Jointly democratic states are
coded as a 1 and non-democratic or mixed alliance dyads are coded as 0. I code this on a
yearly basis, with data coming from the Polity database.

5. 6. 3 Domestic Constraints

The following paragraphs describe the various constraints and restrictions that are
placed on a government.
Executive Constraints

I code executive constraints for the defender state as a means of measuring
audience costs. The Polity project has developed a scale to identify the amount of
constraints that a state has on the actions its executive branch can undertake. The scale is
a 7-point scale, ranging from 1-7.

A brief description of this scale is as follows. 1) the leader has unlimited
authority. 2) intermediate category. 3) slight to moderate limitations on executive
authority. 4) intermediate category 5) substantial limitations on executive authority. 6)
intermediate category. 7) executive parity or subordination. Further details on the
coding procedures can be found in Jaggers and Gurr (1995), and in the revised Polity II

codebook of Gurr (1997)-

177

mm
In order to determine if a government is a unified (single party) government or
not, I focused on the regime type of the defender. I first noted if the government was a
democracy or a non-democracy, using the criteria described in the following variable
description of democratic defender. I then examined the democratic states that emerged,
and coded on a yearly basis if the defender had a single party government or a divided
(or coalition) government. This variable, UNIFIED, is coded as a 1 if the government is
a single party government and 0 if it is not. Data was taken from Mackie and Rose
(1974), along with the Polity III data.
Democratic Defender
To account for the domestic audience costs that leaders in the defending state

face, I code if the defender is a democracy or not (DEMC). Democracies, as many
scholars have argued (Fearon, 1994a, 1994b; Hart and Eyerman, 1996), face higher
domestic costs in regards to success and failure in foreign affairs. 1 determine on a yearly
basis if the defender is a democracy or not. I utilize the Polity III dataset (Gurr, 1974;
J aggers and Gurr, 1995; McLaughlin et. al., 1998) to determine the state’s yearly
democracy score. I denote any state that receives a 6 or higher on the Polity scale as a
democracy. Additionally, since there do exist regime changes throughout history, I
utilize the Polity 111d data of McLaughlin et. al. (1998) to determine if, at the point of
crisis initiation, the defender was truly a democracy.

5. 6.4 Initial Beliefs of the Attacker

The following paragraphs discuss how I have operationalized the prior beliefs of

the attacker.

178

M!

In a first attempt to measure a potential aggressor’s beliefs about the commitment
of a defender, I examine the defender’s past actions in regards to the other allies it has.
This is an attempt to measure the argument made by Schelling that was mentioned above,
that credibility is often maintained by acting in a variety of venues.

Data for total alliance involvement on the part of a defender was taken from
Bennett (1997). I used the MID and ICB datasets to determine previous encounters in
which a defender had the opportunity to assist another ally. I code my variable
CREDIBLE as a 1 if the defender aided another ally in the previous year against some
threat, and 0 if the defender did not aid another ally or had no opportunity to do so.
Involvement

Secondly, I pr0pose that additional dispute involvement in any given year on the
part of a defender should have an impact on the success of extended general deterrence.
Using the MID and ICB datasets, I code on a yearly basis the variable INVOLVE. I code
it as a 1 if the defender is involved in a dispute with another state and 0 if it is not.

5. 6. 5 Military Balance

The following paragraphs describe how I have measured the military capability

measures that are used in my empirical analyses.

Immediate Balance of Forces

 

The immediate balance of forces is best defined as those as those land forces of
the potential attacker in a position to initiate an attack and those land forces of the
defender and protege to repulse such an attack (Huth, 1988, 58). As in regards to the

long-term balance of forces, I code the immediate balance of forces (IMMF OR) as a ratio

179

of the defender’s capabilities plus the protégé’s capabilities divided by the attacker’s
capabilities.

Data for the immediate balance of forces is not as easily found as is the long-term
balance. Some data was garnered from Huth (1988), Clodfelter (1992), and Dupuy and
Dupuy (1970). Other sources for this data are listed in Appendix D, where sources on the
various alliances and crises are listed.

Short Term Iial_ance of Forces

The short-term balance of forces (STBAL) differs from the immediate balance in
that it is also concerned with what military forces each side can call upon within a short
period of time. It is largely concerned with the fear attacker has in a possible military
intervention on the part of the defender.

Once again the short-term balance is expressed as a ratio between the combined
capabilities of the defender and protégé and the potential attacker. However, in this case,
I adjust for distance if the defender and attacker are not contiguous with the protege. The
distance procedure is discussed below, in the coding for the long-term balance of forces.
Data regarding the short-term balance of forces are drawn from Clodfelter (1992), The
Correlates of War projects dataset on military capabilities, Dupuy and Dupuy (1970)
(1977), Huth (1988), May (1984), Wolfe (1970), The Military Balance (1961-1984), and
the Statesman 's Yearbook (1885-1984). Additional sources that were consulted can be
found in Appendix D, under the descriptions of the various crises.

Long-term Balance of Forces
As elaborated upon earlier, military capabilities play a great role in ensuring

deterrence success or failure. At the level of extended general deterrence, we are

180

preoccupied largely with the long-term balance of forces. This corresponds with an
attacker being concerned about a long war of attrition; namely, what power can a
defender and protege combined raise to fight a war that may drag out. Such a concern is
evident in Japanese and German thinking in the Second World War. Both states believed
that by striking fast, they could neutralize the Allied power enough to prevent the
emergence of a long costly conflict.

I construct two measures of the long-term balance of military capabilities. Both
of them are based upon the Correlates of War’s measure of national capabilities (Singer,
Bremer, and Stuckey, 1972). The measure of national capabilities takes each state’s
military capabilities (percentage of world military personnel and military expenditures)
and multiplies by the sum of that state’s industrial and demographic resources. These
additional resources include percentage of world steel production. industrial fuel
consumption, urban, and total population. The long-term balance of forces, denoted as
LTBAL, is calculated as the ratio of the combined capabilities of B and C divided by the
capabilities of A.

However, there is some discussion concerning the projection of power around the
globe. As Boulding (1962) termed it, there is a “loss of strength” gradient. To put it
simply, power declines with distance. As a defender, the United States for example,
prepares to undertake action when a protégé is threatened, it must make preparations to
potentially use military force away from home. Moving material from the United States
to West Germany, for example, is costly and time-consuming. Travel conditions are
poor, troops become ill, supplies become damaged or lost in transit. To account for these

issues, I utilize an adjustment developed by Bueno de Mesquita (1981, 105), to calculate

181

the adjustments made to take into account the distance the defender must travel in order
to reach the protege.

The variable, termed ALTBAL (adjusted long-term balance) adjusts the
capabilities of both the defender and attacker by the distance they must cover. The
formula (Bueno de Mesquita, 1981. 105) is as follows:

d. . - + 0 . . ' ‘ '
logl Islam" ‘b’m‘ksl “ “6’, where ib is the distance in miles

adj capabilities = capabilities
between A or C to the defender. and miles is the number of miles a state can reasonably
expect to cover in one day. For the time period 1816-1918, Bueno de Mesquita argues
that states could move on average 150 miles per day. For the interwar period, the figure
rose to 375 miles per day. In the post 1945 era, the figure is 500 miles per day.108

Once the adjusted capabilities for A and C are calculated, the balance of forces is
once again derived, with the adjusted capabilities of C being added to the capabilities of
B, and this being divided by the adjusted capabilities of A. '09 This figure is the final
variable used in the analysis to measure the long-term balance of forces.

Nuclear Weapons

Nuclear weapons have long been surmised to have kept the peace between the
superpowers during the Cold War. I maintain that since nuclear weapons have been
viewed as weapons of last resort, they do not play that large of a role at the level of

general deterrence. This keeps in line with Schelling’s (1966) concept of brinkmanship,

which asserts that nuclear weapons may be used in a crisis as it spirals out of control.

 

. ’08 Distances were computed from capital city to capital city for A to B, and C to B.
Distances were calculated using the EUGene program of D. Scott Bennett and Allan
Stain. Data and documentation on the EUGene program can be found at the following
website: (http://wizard.ucr.edu/cps/eugene/eugene.html).

182

The variable (NUCLEARC) is coded as a I if the defender possessed nuclear weapons at
the time of a crisis, and 0 if it did not possess such weapons.
5.6.6 Issue at Stake

The following paragraph discusses howl measured the territorial issue at stake for
each crisis in my data set.
Issue at Stake

Above I hypothesized that as the value that the defender placed on the issue at
stake x increased, the greater the likelihood that extended immediate deterrence would be
successful. As the implications of the formal models suggest. an attacker will, in certain
situations, make a demand of a state, even if it is unevenly matched in terms of military
capabilities. All the attacker must do is make a demand that has little salience to the
minor power’s defender (F earon, 1994a).

As evidenced in so many conflicts throughout the years, territory is an issue that
states have fought over. Since much of the empirical record noted earlier suggests that
states fight primarily over territorial issues, I code whether the issue at stake for the
attacker and target is territorial in nature. I code this variable TERRITORY as a 1 if
territorial concerns are the reason for the crisis, and 0 otherwise. Data for this variable is
taken to some extent from Brecher and Wilkenfeld (1997), as well as from the detailed
histories listed in Appendix D.

5. 6. 7 Updated Beliefs of Attacker
The following paragraphs describe how I measured the updated (posterior) beliefs

of the attacker, once a crisis has been initiated.

'09 All calculations were done using the SPSS statistical package. Data for the
capabilities was derived using the EUGene program.

183

Tit-for-Tat

As was argued above, a military strategy of tit-for-tat should lead to a successful
extended immediate deterrence outcome. A policy of tit-for-tat allows a defender to
signal its resolve, but it is a response to the actions of the attacker, so provocative
measures are not undertaken. I code this variable, TFT. as a 1 if such a policy is adopted
by the defender and 0 if a different policy is ad0pted. In order to determine what policy
of military escalation was adopted by the defender, I consulted various historical
accounts of the 45 extended immediate deterrence encounters that comprise my dataset.
Details on these cases can be found in Appendix D.
Firrn-but-Flexible

In addition to using military means, states pursue diplomatic measures in order to
resolve crises. As was argued above. a firm but flexible bargaining strategy will provide
the greatest leverage in peacefully ending a crisis. This allows a defender to make offers
regarding the ending of the crisis, rather than merely reacting in a negative fashion to the
actions of the attacker. I coded this variable, F BF, as a 1 if the defender adopted a firm-
but-flexible stance in a crisis, and as a 0 if it adopted another form of bargaining
behavior. In order to determine what policy of bargaining behavior was adopted by the
defender, I consulted various historical accounts of the 45 extended immediate deterrence
encounters that comprise my dataset. Details on these cases can be found in Appendix D.

Past Behavior

 

In a first attempt to determine the credibility of a defender, and the influence this
has on the perceptions of a defender, I examine the past behavior of a defender with

respect to its protege. In particular, I examine if the defender has come to the assistance

I84

of its protégé when the protégé has had some sort of adversarial encounter with the
current potential aggressor.

1 code this variable, PAST, on a yearly basis. I code it as a 1 if the defender aided
its protégé in the previous year (or most recent past encounter with the potential
aggressor) and 0 if the defender did not aid its protége’ or had no opportunity to do so. I
used the MID and ICB data to identify past encounters between the involved states.
Ongoing Dispute

1 code for a final variable ONGOING, which measures if the defender is engaged
in any additional ongoing crises. As was argued above, this variable helps determine the
updated beliefs of an aggressor. 1 code the variable as a 1 if yes, and 0 if not. I
hypothesize that ONGOING should have a negative impact on the success of extended
immediate deterrence. The presence of this variable helps control for any cross-sectional
variations that may arise. Data was taken from the MID and ICB datasets.

5.7 Descriptive Statistics: A First Cut at Analysis

Having delineated all of the independent variables to be utilized in the statistical

analysis, I now turn to a first examination of these measures. Table 5.4 provides

descriptive statistics of all of these variables.

185

TABLE 5.4

Descriptive Statistics

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable N Mean Std. Dev. Min. Max.
Past 302 . 1986755 .3996654 0 1
Credible 302 .4801325 .5004343 0 1
Export 302 .0261 122 .03413 79
Strategic 302 .341 05 96 .4748522 0 1
Location
Defense 302 .76821 19 .4226747 0 1
Pact
Arms 302 .6863245 .3521581 .05 1
Bases 302 9.324503 22.10772 0 95
Adjusted 302 .7993212 .2608836 .2462212 1.503868
Long-term
Balance of
Forces
Executive 302 4.675497 2.419603 1 7
Constraints
on C
Democratic 302 .536423 8 .4994992 0 l
C
Involved 302 .5496689 .4983 527 0 1
Democratic 302 .5364 .4995 0 1
Alliance
Nuclear C 302 .3940397 .4894545 0 1
Unified 302 .6721854 .470196 0 1
Short-term 302 1.024783 .4985791 .36 2.88
Balance of
Forces
Tit-for—Tat 46 .6956522 .4652151 0
Firrn-but- 46 .5217391 .505047 0
Flexible
Territory 46 .5217391 .505047 0 1
Ongoing 46 .5217391 .505047 0 1
Dispute
Immediate 46 .561087 .2795888 15 1.1
Balance of
Forces 1 \ \

 

 

 

 

In addition to these descriptive statistics, I perform some initial examinations of

the data to ascertain if any problems exist. As a first cut, I attempt to determine if any

186

110

multicollinearity exists between any of the independent variables. Multicollinearity. as

noted by Hanushek and Jackson ( 1977, 87-89), exists when two or more independent
variables are highly correlated in a sample of the data. High degrees of collinearity
among independent variables calls into question the assumption that the independent
variables are independent of one another.

I examined the correlation matrix of all of the independent variables to determine
whether multicollinearity is a problem with my independent variables. As a good rule of
thumb, multicollinearity is problematic if the absolute value of any variable is correlated
with any other variable at the level of .75 or higher. Upon inspection of this matrix, the
only variables that have any potential problems arising from multicollinearity are XCON
and DEMC. Both of them, as noted above, are a means of measuring domestic audience
costs in democratic states. Hence, it is not surprising that they are correlated at such a
high level (.828). However, since they are not used in the same equation, this high
correlation is not damaging to the empirical analysis conducted in the next chapter.

5.8 Conclusion

Prior to empirically testing the implications of my formal models, I made an
initial examination of the data and its implications. This second examination of the data
is more of a descriptive nature than one assessing the nature of the data’s proper
measurement and Specification. As was noted in chapter 4, there exist six distinct
outcomes of the formal model. In Table 5.5 I break down the 302 cases of attempted

deterrence via alliance, and categorize them with their respective outcomes.

 

no . . . . _
In the next chapter I examine the data for an additional potentially confounding
factors: autocorrelation

187

Table 5.5

Distribution of the Model’s Outcomes

Outcome N um 0 Cases
Status
ACQB

A
DA
CNAB

ALL

 

A number if interesting items emerge from this initial examination of case
distribution. First of all, extended general deterrence has a tendency to be successful. In
256 of the 302 total cases, the status quo is the outcome. In the 46 remaining cases, an
aggressor backs down in 27 of the 46 cases. Thus, in 58.7% of all extended immediate
deterrence cases identified here, the aggressor did not gain its objectives. This is a good
indication that alliances have a decent impact on ensuring that crises do not escalate in
the event that extended general deterrence falters.

In the remaining 19 cases in which extended general deterrence was unsuccessful,
14 of these 19 cases involved the active military intervention of the defender on behalf of
its protégé. This preliminary analysis calls into question the assertion that has been made
regarding the impact that alliances have on the onset of conflict. It also calls into
question the conventional wisdom regarding the reliability of alliances as well.

These preliminary statistics only reveal some rough indicators regarding the
effectiveness of extended deterrence via alliance. Further analysis is required to
determine why alliances serve as such good deterrent. In the following chapter I discuss

the methodological means by which I will test the propositions presented in this chapter.

188

Having outlined the variables that comprise the empirical portion of these tests, I now

turn to the statistical testing of my game-theoretic models.

189

CHAPTER 6

EMPIRICAL TESTS

6.1 Introduction

In this chapter I present a means of testing the propositions that emerged from my
formal models. I have argued throughout this dissertation that extended general
deterrence and extended immediate deterrence are linked as part of a larger process. One
cannot assume that extended immediate deterrence encounters occur without first
acknowledging that extended general deterrence faltered. If one assumes that the two
processes are unrelated, then one is introducing selection bias into the statistical

framework. Such bias can possibly introduce errors into the statistical estimation, and

provide misleading results.

In this chapter I outline a statistical method that can account for this selection bias
in related equations. Specifically, I utilize a bivariate probit model that accounts for
selection (Dubin and Rivers, 1989; Reed, 1999). Since this method is based upon the
maximum likelihood probit model, I first outline the assumptions underlying a probit
specification. I then provide separate probit models for my two equations, extended
general deterrence success and extended immediate deterrence success. These estimates

serve as baselines to which we can compare a more properly specified model that

accounts for the selection process.

Building upon the initial probit estimates, I then delineate why a bivariate probit

model that accounts for selection is the more appropriate means for testing the theory

developed in this dissertation. I present estimates from the bivariate probit model, and

compare them with the standard probit results.

190

Having empirically tested my models, I move to a discussion of what the results
mean in a substantive fashion. I examine the substantive effects of my independent
variables, and interpret their impact within the context of the six cases I have been tracing
throughout this dissertation. I focus on the role that the signals sent by the defender at
both the general and immediate levels have on ensuring deterrence effectiveness, and the
perceptions that a potential aggressor develops given these signals.

6.2 The Probit Framework

Most political scientists that are trained in quantitative methodologies are familiar
with linear regression. Specifically, they tend to be most familiar with ordinary least
squares (OLS), a technique that assumes that the model being estimated is linear in
nature, and that the dependent variable of interest is normally distributed and continuous.
An OLS regression equation assumes the following functional form:

y.==a+,dx.+c,, i=1, ...... n (I)
In the above equation, known as the classic linear regression mOdel, a certain number of

assumptions are made. These are as follows:

1. The disturbance term of the equation has a mean of 0. E [6]] = 0 for all i.

2. Homoskedasticity, which assumes that the variance 02 is a constant for all i.

3. Non-autocorrelation among the disturbance terms, Cov [5,, 5,] = 0, i¢j.
4. The regressor x and the disturbance term a; are not correlated, Cov [x, at] = o for all I
and j.
In the models being estimated in this dissertation, these assumptions are not met.
The classic linear model assumes a continuous measure for the dependent variable y. In

the models being estimated below, the dependent variables are both dichotomous, coded

I91

as 0 and 1. Hence, attempts at estimating the equations of interest with OLS would lead
to results that are misleading.

Instead of linear regression techniques I utilize maximum likelihood estimation
(MLE) procedures (Aldrich and Nelson, 1984; Eliason, 1993; Fisher, 1950; King, 1989).
Maximum likelihood provides a means of estimating equations that are nonlinear in
nature, with dependent variables that are dichotomous (or non-continuous). The logic of
MLE is aptly captured in the following quote. “Find an estimate for G) such that it
maximizes the likelihood of observing those data that were actually observed” (Eliason.
1993, 7-8).

I have the option of using two different model specifications to estimate my

equations. The first functional form would be the logit equation, which has the following

functional form:

1 (2)

Pr (extended general (immediate) deterrence success =
1 + e-AB

where the equation is read the probability of extended deterrence success and XB

are the independent variables in the equation.

Alternatively, the probit specification could be used. The functional form of the

probit equation is:
Pr (extended general (immediate) deterrence success = $(B’x) (3)
where the equation is read the probability of extended deterrence success and B’x

are the independent variables in the equation being estimated.

192

Regardless of which specification is used, the statistical results will be virtually
identical.’ ” In the analysis conducted below I use the probit specification. Although
probit results are a bit more difficult to interpret than are logit results, I feel that probit is

more appropriate, given that the following section will estimate the model with a

bivariate probit specification. ' '2

6. 2. I The Statistical Models to be Tested

Having noted the methodology being used to test my two statistical models. I now
turn to a specification of the equations to be estimated. The first equation concerns itself
with extended general deterrence success. The second equation concerns itself with

extended immediate deterrence success. The first equation, therefore, can be written as

follows:
y=c+blx1+b2x2+b3x3+b4x4+b5x5+u (4)
where y is the dichotomous dependent variable extended general deterrence

success (EGDSUCC), x1....x5 represent the parameters from the formal models that are

being estimated, and the symbol in parentheses afterwards denotes the expected sign of

the statistical coefficient; ’ '3

C = constant term

X] represents the security commitment component of the equation (+)

 

'” Good discussions of the use of binary choice models can be found in Greene (1993,
ch. 21).

”2 King (1989. l 12) discusses how to transform probit estimates into logit estimates and

vice versa.

193

xz represents the autonomy component of the equation (+)

X3 represents the domestic costs component of the equation (-, +)

x4 represents the beliefs component of the equation. (+)

u = error term.

The second equation to be estimated is concerned with extended immediate
deterrence, and the determinants of its success or failure. The second equation can also
be written in the following manner:

y = 0+ b/xt +b2x2 + b3x3 + b4x4 + u (5)

where y is the dichotomous dependent variable extended immediate deterrence
success (EIDSUCC); x, .x4 represent the parameters from the formal models that are
being estimated, and the symbol in parentheses afterwards denotes the expected sign of
the statistical coefficient;

c = constant term

X; represents the balance of forces (immediate, short-term, long-term, nuclear

capability of the defender) (+)

x; represents the issue at stake in the crisis (-)

x3 represents the domestic political constraints on the defender (+. -)

x4 represents the updated (posterior) beliefs of the aggressor (-)

u = error term.

Table 6.1 provides an overview of the variables to be examined in the empirical
models being tested at the level of extended general deterrence This table links the level

0f analysis, parameters from the formal models, and their empirical operationalization.

g

”3 In the cases in which two different propositions can be tested, and the different
prOpositions predict different signs on the coefficients, I specrfy the direction for both

194

TABLE 6.1

Parameters and Variables Influencing
Extended General Deterrence Success

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Parameter from Definition , Empirical
Formal Model Operationalization
m Security Arms transfers
Commitment
“ “ Defense pact
established
“ “ Bases established
y Autonomy Benefit Exports
“ “ Strategic Location
“ “ Democratic Alliance
“ “ Import of strategic
raw materials
d Domestic Political Executive
Constraints on Constraints on
Defender Defender
“ “ Unified government
11 Initial (prior) beliefs Credible Defender
of attacker toward other allies
regarding defender’s
type
“ “ Involvement in
other Crises
k Military Balance Short-term Balance
of Forces
“ “ Long-Term Balance
of Forces

 

 

 

 

 

In the same vein, Table 6.2 provides an overview of the variables to be examined
in the empirical models being tested at the level of extended immediate deterrence. This
table also links the level of analysis, parameters from the formal models, and their

empirical operationalization.

estimates.

195

emph

COnce

TABLE 6.2

Parameters and Variables Influencing
Extended Immediate Deterrence Success

 

 

 

 

 

 

 

 

 

 

 

 

Parameter from Definition _ Empirical
Formal Model Operationalization
k Balance of military Immediate balance
forces of forces
“ “ Short-term balance
of forces
“ “ Long-term balance
of forces
“ “ Nuclear capability
of defender
x Issue at stake in Territory
crisis
d Domestic Political Regime type of
Constraints on defender
Defender
“ “ Unified Government
11* Updated (posterior) Tit-for-tat strategy
beliefs of attacker of military
regarding defender escalation
“ “ Firm-but-flexible
diplomatic strategy
“ “ Defender
involvement in
additional ongoing
crisis
“ “ Past behavior of
defender regarding
protege

 

 

 

 

 

6. 2. 2 Empirical estimates

Having described the equations to be tested, we now turn our focus to the
empirical estimates of these equations. Table 6.3 presents the results of the equation

concerning extended general deterrence, and Table 6.4 presents the results of the equation

196

concerning extended immediate deterrence. Both equations were estimated with a probit
specification. ' '4

In these statistical estimates I also make every attempt to control for a problem
that often plagues empirical models of this sort. Of great concern here is the issue of
autocorrelation. Autocorrelation (or serial correlation) is a problem that often plagues
time series models. It generally occurs when “measurement errors lead to the same
difficulties or problems encountered during several successive time periods” (Hanushek
and Jackson, 1977, 143). Autocorrelation, when it is present, causes parameter estimates
to be inefficient, although they remain unbiased. The standard error of the variable is
often increased when autocorrelation is present, and this decreases the t-statistic, which
may cause a researcher to reject it as being insignificant, even if this is not actually the
case.

In data such as those being analyzed here, autocorrelation is often a major issue.
This is because the data under analysis are binary cross sectional time series (BCSTS)
data. ’ ’5 Simply put, the data here is a set of groups, each group being measured

according to some temporal constraint. 1n the analysis being conducted here. the year is

 

”4 All estimates used in this dissertation were derived using the Stata 6.0 statistical

program. Data was entered into SPSS format for ease of manipulation, and then the Stat
Transfer package was used to transform the SPSS data into Stata format. I also ran the
model in SPSS to ascertain that the results did not differ. The statistical estimates were
identical.

”5 Beck and Katz (1995) deal with the issue of autocorrelation in time-series cross
sectional (TSCS) models. However, this assumes a continuous dependent variable.
Their discussion is of importance in that it calls researchers attention to the issue of
autocorrelation in the context of cross-sectional data. Beck, Katz, and Tucker (1998)
extend this discussion to the study of BTSCS data. I utilize their techniques below to
ascertain if autocorrelation is present in my statistical model of extended general
deterrence success.

197

the temporal unit of analysis. The problem that arises is that the dependent variables may
be influenced by temporal problems. In a first attempt to account for any possible
temporal problems with the data, I run the statistical estimates using robust standard

errors. Robust standard errors assist in correcting for temporal autocorrelation among the

data llb

Both models presented below are the result of testing a variety of empirical
specifications in order to obtain the best fit. In both cases at least one variable represents

the parameters from the formal models. as outlined in tables 6.1 and 6.2.

 

”6 For more on the role of robust standard errors, a good introduction is Beck (1996).
Greene (1993, 308-3 09) discusses the role of robust estimates in econometric models.
Beck, Katz, and Tucker (1998) discuss the role that autocorrelation plays in BTSCS
models. For the original discussion of robust standard errors, see Huber (1967) and
White (1980).

198

PROBIT ESTIMATES OF EXTENDED GENERAL DETERREN CE

TABLE 6.3

 

 

 

 

 

 

 

 

SUCCESS
Variable Concept Coefficient P > z
(Robust
Standard Error)
DEFPACT Security .5372159 .013
Benefit (.2166767)
EXPORT Autonomy 8.21895 .05 8
Benefit (4.328146)
DEMBC. Autonomy -.4939739 .034
Benefit (.2328589)
STBAL'” Military 3.256-07 .000
(logged) Balance (4.40e-08)
XCONC Domestic -.0479043 .278
Constraints on (.0442018)
C
INVOLVED Initial Beliefs -.3435184 .081
(1971593)
CONSTANT .7766715 .001
(2298518)
N= 302 Pseudo R‘ = .0784. Cases properly predicted = 90.5%
x2 (6) =89.09 Prob. > x2 =0.000 Log-Likelihood = - 118.76899

Null Model = 84.77%

1 , . .
'7 I take the log of the short-term balance of forces in each equation in order to account

for the overwhelming advantage some countries have in regards to manpower.

199

TABLE 6.4

PROBIT ESTIMATES OF EXTENDED IMMEDIATE

 

 

 

 

 

 

 

 

DETERRENCE SUCCESS
Variable Concept Coefficient _ P > z
(Robust
Standard Error)
IMMBAL Military 1.04627 .205
Balance (.825643)
STBAL Military .0000685 .1 10
(logged) Balance (.0000428)
BASE*ARMS Security .0238293 .003
Benefit (.0080145)
DEMC Domestic .795814 .169
Constraints on (.5786693)
C
TERRITORY Issue at Sake -1 .603661 .004
for A (.5561154)
FIRMBFLE Updated Beliefs .3427852 .497
ofA (.5041317)
Constant 1 .062651 .293
# (1.01146)
N= 46 Pseudo R1 = .3336 Cases properly predicted = 87.0%
it2 (6) =40.49 Prob. > x2 =0.000 Log-Likelihood = - 20.782421

Null Model = 52.2%

Prior to any discussion of these models, I examine the extended general
deterrence model for the presence of autocorrelation. To test for autocorrelation in my
extended general deterrence model, 1 utilized the BTSCS temporal dummy variable
method of Beck, Katz, and Tucker (1998). For my estimated model without any
corrections for temporal correlation, I obtained a chi-square statistic with 6 degrees of

freedom of 89.09. I then created temporal dummy variables to account for any potential

200

autocorrelation within the model.l '8 The model with the temporal dummies yielded a
chi-square statistic with 8 degrees of freedom of 34.3 1. A log-likelihood ratio test of the
two models gives a probability of 0.3564. Hence, duration dependence is not present in
the extended general deterrence model being estimated here. Inclusion of temporal
dummy variables in such a model leads to induced multicollinearity and hence problems
in the estimation, including improper standard errors.

Having examined the model for autocorrelation, I now turn to some initial
discussion of the implications of the estimates. In both models I discuss the overall
power of the model, and the impact that the individual coefficients have on the dependent
variable. I then present some analysis based on the substantive effects that each
independent variable has on the dependent variable. I concern myself first with the
influences on extended general deterrence.
6. 2. 3 Predicting Extended General Deterrence Success

In regards to extended general deterrence, the empirical estimates lend credible
support to the theoretical suppositions made in chapter 3: that potential aggressors
respond to opportunities within their structural environment; constraints on an executive
make it harder for defenders to ensure deterrence credibility; and signals that are
observed from trading security and autonomy play a major role in providing extended
general deterrence success. The overall model is statistically significant at the .0000
level, and predicts 90.5 % of the cases correctly, and all coefficients are in the predicted

direction.

 

”8 These procedures are discussed in some detail in Beck, Katz, and Tucker (1998). For
a Stata program that calculates temporal dummy variables, see Richard Tucker’ 5 website
(www.fas.harvard.edu/~rtucker).

201

Each of the six variables is statistically significant below the .p<. 10 level, and
two (DEF PACT and DEMBC) are significant below significant below the p <. 05 level.
And, as prior theories of deterrence suggest, military capabilities (STBAL) is significant
at the p<. 01 level.

Since MLE coefficients are non-linear in nature, they cannot be interpreted in a
manner similar to OLS coefficients. Therefore, in order to interpret their overall impact
on a dependent variable, we need to resort to a different method. I will use here a
technique called marginal effects, or method of first differences (King, 1999, 106-108).
According to Quinones and Gates (1995, 72, n. 27), “marginal effects are not equal to the
coefficients. ...By calculating the marginal effects of each independent variable, it is
possible to determine the effects of a change in x on predicted probabilities for different
values of the dependent variable”.

In a more technical vein, the following is a discussion of how marginal effects are
calculated. If we hold all independent variables save one constant at their mean, we can
them examine the marginal impact the one remaining variable has on the dependent
variable. This allows us to alter the specification of the model, and see what happens as
values of the independent variables are fluctuated. Table 6.5 demonstrates the

substantive effects the independent variables have on extended general deterrence.

202

TABLE 6.5

SUBSTANTIVE EFFECTS OF INDEPENDENT VARIABLES ON
EXTENDED GENERAL DETERRENCE SUCCESS

 

 

 

Variable Variable Held at VariableHeld at I Percentage Change 1

Minimum Value Maximum Value #
DEF PACT .06 .22 +. l 6
EXPORT .04 .07 +03
DEMBC .06 .01 -.05
STBAL .04 .08 +04
INVOLVED .02 .01 -.01
XCONC .05 .02 -.03

 

 

 

 

 

As can be seen from the substantive effects, the costly signaling theory is lent fair
amount of credence at the level of extended general deterrence. If we move the variable
DEF PACT from its empirical low to its empirical high, the probability of extended
general deterrence being successful increases by 16%. If we examine this in more detail.
it is not surprising.

The signing of a defense pact on the part of a defender signifies a much greater
commitment to defending an ally. Partaking in a defense pact tends to have additional
effects for both parties as well. In the case of the United States and West Germany, the
United States used its power and prestige to protect the nascent Federal Republic from
Soviet aggression. After varied attempts to change the status on the European continent,
the Soviet Union ceased these actions after 1961. The magnitude of the American
commitment to West European security in general, and to German security in particular,

sent messages to the Soviets that the American presence in Europe was permanent. The

203

costly signals that the United States was able to send through its strong alliance
commitment to West Germany helped prevent any challenges after 1961.

Although the costly signal has such a large impact, it should also be noted that the
other variables in the equation also exert influence on whether extended general
deterrence succeeds or fails. As expected, the military balance between the alliance and
potential attacker is quite important. As the short-term military balance is fluctuated
from its empirical low to its empirical high, the probability of extended general
deterrence succeeding increases by 4%.

It should be noted that it is the short-term balance, not the measure of national
capabilities, which has the greater influence on deterrence success. This corroborates the
findings of Huth (19880, who discerned that challengers hope for quick victory, and will
strike when they feel it is opportune. The national capability measure is important for a
long-term war of attrition, but it plays little role in determining whether or not an attack is
imminent. This helps explain why we witness conflicts that are asymmetric in nature;
that is. war initiation by a weaker power. It is the hope of the challenger that if the short-
term balance is in its favor, it can prevail.

The actions taken by Germany during the Czech crisis of 1939 illustrate this
nicely. Germany realized that if it annexed Czechoslovakia, it might possibly incur the
wrath of the British and the French. However, the German short-term balance was
favorable compared to the British and French, who could not mobilize tr00ps quickly
enough to fight the German threat. In this case, the advantage lay with the Germans, and

they exploited it to its full potential.

204

Autonomy benefits play quite a role in determining whether or not extended
general deterrence succeeds or fails. One of the key benefits defenders receive from
allying with minor powers is the ability to gain access to foreign markets. As the
EXPORT variable demonstrates, as the level of export from a defender to a protege
increases, the probability of deterrence being successful increases 3%.

If we go back to the nineteenth century, we see that the role of economics played
quite a part in the decision of Austria-Hungary to ally with Serbia and with Romania. As
Langer (1950) points out, the decision to ally with both of these states was an Austrian
desire and need for greater access to markets (along with an importation of raw materials
needed for the military). Although the importation of strategic materials is not
statistically significant, the ability to export signifies a greater willingness for a defender
to fight on behalf of a protege if a crisis erupts. In the years that Austria maintained its
alliance ties with these two Balkan states, only two crises erupted. Both were quite
minor, and settled due to Austria’s rapid involvement. In her zeal to protect economic
interests, Austria reacted quickly to threats to her small allies.

Two variables that coincide are the democratic alliance variable and the executive
constraints variable. Both of them have a negative impact upon the success of extended
general deterrence. DEMBC decreases the likelihood of deterrence success by 5% as it is
fluctuated between its empirical low and high, and XCONC decreases the likelihood of
deterrence success by 3% as it goes from its low to high values.

As was argued in earlier chapters, both of these variables address regime
characteristics and willingness of states to make concessions to their allies. In the case of

similar regime types, the implication is that democracies will be less willing to coordinate

205

policies with alliance partners, due to the fear that it would infringe upon their
sovereignty. As France found out in the interwar period, the Belgian refusal to
coordinate common defense policies or permit French bases on Belgian soil was largely a
belief on the part of the Belgian government that to do so was a violation of Belgian
independence. However, as the United States has found out in regards to South Korea
and Taiwan, it is much easier to coerce a non—democracy into making concessions over
policy or base establishment.

This leads into a discussion of the impact that executive constraints have upon
ensuring effective deterrence policies. As was suggested in chapter 3, the more
constraints that are placed upon a government, the harder it is for that government to
credibly commit to other states. Hence, democracies may find themselves the targets of
aggression, or find their allies as more common targets due to the conception that the
democracy will not intervene. It appears from the analysis that it is merely the presence
of constraints at the institutional level that matter, rather than the presence or absence of a
single-party unified government. This calls into question some of the beliefs that state
that certain types of government are more or less vulnerable to external pressures.

Rather, it provides an indication that it is an institutional constraint that matters more.

Finally we turn to an analysis of beliefs and their impact on when extended
general deterrence will succeed or fail. Contrary to many other beliefs (Mercer, 1996;
Schelling, 1960), credibility does not appear to be an important factor at the level of
general deterrence. Past behavior also does not appear to play a role in determining
whether extended general deterrence succeeds or fails. These findings, since the faith in

them has driven much of the foreign policy process during the post 1945 era, warrant

206

further investigation. However, it is interesting to note that if a defender has been
involved in a dispute during the previous year, extended general deterrence success
decreases by 1%.

This can, I maintain, be explained by the fact that. attackers recognize
vulnerabilities in defenders, and act upon them. China, for example, tended to initiate
attacks on Taiwan in the 19505 when the United States was either currently involved in
another dispute or had recently been involved in one. “War-weariness” makes it difficult
for defenders to respond to every provocation that arises, at least in the minds of the
challengers. If the issue is that pressing to an attacker, it will often use the defender’s
dispute involvement as a good clue as to when to make a demand.

6. 2. 5 Predicting Extended Immediate Deterrence Success

Having discussed the influences that shape extended general deterrence success,
our attention now turns to the situation that occurs when extended general deterrence
fails. As I argued in chapter 3, while extended general deterrence failure is largely due to
changing opportunities within the international system, when it comes to immediate
deterrence success or failure, it is the more contextual factors that matter, notably the
costly investments that are made and the balance of military capabilities.

When it comes to predicting the success of extended immediate deterrence, the
overall model is quite successful. It is statistically significant at the .0000 level, and all
of the coefficients are in the proper direction. It predicts 87% of the cases correctly, with
a null model predicting 52.2% of the cases properly.

However, when it comes to interpreting the individual variables, the results are

not as successful. Only two variables (TERRITORY and BASE*ARMS) are statistically

207

significant below the p 5. 01 level. Others, notably the military balance variables
(IMMBAL and STBAL) and the domestic constraints variable (DEMC) are close to
being statistically significant, but do not meet the criteria that is conventionally accepted.
Due to this, I suspect that there is some sort of modeling issue that is affecting the results.
It is to account for such problems that I utilize a bivariate probit model with selection.

In the following section I introduce the bivariate probit model that accounts for
selection. As the discussion demonstrates, modeling extended immediate deterrence
encounters without accounting for the failure of extended general deterrence can lead to
erroneous estimates.

6.3 The Bivariate Probit Model with Selection: Accounting for Selection Bias

While the models above are statistically significant in their own right, there is
something lacking. This is especially the case with the equation that estimates extended
immediate deterrence success. As was argued in chapters 3 and 4 extended deterrence
must be viewed as a process, with immediate deterrence encounters‘being a subset of
general deterrence failures. Additionally, as I maintained in chapter 3, and demonstrated
in the formal models in chapters 4, extended immediate deterrence encounters arise
directly out of the failure of extended general deterrence. The two distinct probit models
estimated above, however, do not capture this sequential process.

There exist two distinct alternatives to modeling extended general deterrence and
extended immediate deterrence as an interrelated sequential process. The first, as was
done above. is to model the two equations separately. This assumes that no dependence
exists between the two stages of the deterrence process. However, as our theory

demonstrates, this assumption is not realistic.

208

The second alternative would be to use a modification of the Heckman (1979)
two-step procedure. Huth (1996) utilized such a method in his study of territorial
disputes. This procedure estimates two probits (or logits), and saves their predicted
probabilities. The predicted probabilities from the first equation are transformed into
Mill’s inverse ratio. This new variable would be included in the extended immediate
deterrence equation, and captures the selection effects that may be existent in the models.
Unfortunately, it assumes that one has properly specified the exact population of cases (in
this analysis. the entire population of extended deterrence via alliance cases). It also
produces results that are heteroskedastic and inefficient.

Given that the two options discussed above are problematic, there is a third means
of estimating the two equations to account for the fact that the extended immediate
deterrence equation is a subset of the larger, extended general deterrence, phenomenon.
This is through the use of a bivariate probit model with selection, as developed by Dubin
and Rivers (1989), and utilized by Reed (1999) in a study of conflict initiation and
9

escalation. ' '

A bivariate probit model assumes the following model structure:

yl* =X/fl/ + u, (3)

)0“ = X252 + 112 (4)

——k

”9 Brehm and Gates ( l 997) also utilize a similar framework in their study of bureaucratic

oversight. See Greene (1993, 360-3 63) for a brief, yet detailed description of this
procedure.

209

Let y, * be the dependent variable extended general deterrence success, and let ya”
represent the dependent variable extended immediate deterrence success. We only
observe y;* if y, * is equal to 0; that is, extended general deterrence has failed.
Implicitly assumed in the independent equations-estimated above is the notion
that the value of the error term, u. is uncorrelated between the two models. Hence, it
assumes that the covariance C 0v [u 1. u;] = 0. However, given the nature of the theory
specified above, we cannot assume that the two models are unrelated. Given this belief,
we can assume that C 0v [u ,, u2] = p. This symbol, p, captures the selection effects that
transpire between the two equations being estimated.
With these assumptions, therefore, we can write the likelihood function of the
20

bivariate probit model with selection as follows:l

L08 L = 2108(1'¢(BIX1)
YI=0

+ 2 log 92 ((lels (Bzxz, 'P)
Y]: 09 YZ=0

+ 2108 (1’2 «le19 (BzXz, '9)
VI :19 3’2 :1

Having specified the likelihood function for the model being estimated some
clarifications on the terminology in the function are in order. (I) is the distribution of the
univariate normal function, and 92 is the bivariate normal distribution function. The first
term on the right side of the equation relates to the censored observations that are an issue

of extended general deterrence never fails. The second and third terms on the right side

 

’20 This likelihood function comes out of work conducted by Meng and Schmidt (1986),
and has been used by Brehm and Gates (1997), Dubin and Rivers (1989), and Reed
(1999) in their. studies of similar models.

210

relate to those cases in which extended general deterrence fails but does not escalate
(extended immediate deterrence success), and those cases in which the crisis escalates
(extended immediate deterrence failure).
6.3.1 Empirical estimation with the bivariate probit model with selection

I-Iaving discussed the framework within which we will estimate the two-equation
model, we now turn to an estimate of the fully specified model. The variables in the
bivariate model are identical to those estimated above. The results of the model are

presented in Table 6.7. ’2’

 

121

The model was estimated in Stata 6.0 using the nested command.

211

 

TABLE 6.6

A UNIFIED MODEL OF EXTENDED GENERAL DETERRENCE AND
CRISIS ESCALATION USING BIVARIATE PROBIT ANALYSIS

 

 

 

 

 

 

 

 

 

Variable Unified Model Extended General Extended Immediate
B (std. Error) Deterrence Success Deterrence Success
Only Only
13 (std. Error) 13 (std. error)
DEFPACT .5372 (.2167)** .5372 (.2167)**
EXPORT 8.218 (4.328)* 8.218 (4.328)*
DEMBC. -.4940 (.2329) ** -.4940 (.2329) **
STBAL 3.25e-07 3.25e-07
(logged) (4.40e-08)*** (4.40e-08)***
XCONC -.0479 (.0442) -.0479 (.0442)
INVOLVED -3.435 (1.971) * -3.435 (1.971) *

IMMBAL .9854 (.7628) * 1.046 (.8256)
STBAL .209 (.397) ** .0000 (.0000)
DEMC .637 (1.458) ** .7958 (.5787)

BASE*ARMS .025 (.0074) *** .0238 (.0080) ***
TERRITORY -154 (.5329) *** -1.604 (.5561) ***
FIRMBFLE .3046 (.4871) * .3428 (.5041)
p SELECTION -0.7413 (.0494)***
EFFECT
Log-Likelihood -121.59 -1 18.77 £35678
Sample Size 302 302

*=

p: .10; ** = pf .05; m = p: .01

212

As can be seen with just a cursory glance, the bivariate probit model that accounts
for selection does a much better job of estimating the entire extended deterrence process
than do the two independent probit equations. From this, two issues arise. First of all,
the value for p that captures the selection effects within the model is negative, and it is
statistically significant. If extended general deterrence failure and extended immediate

deterrence success were two unrelated events, then the p would not be significant. In

other words, we could assume that p = 0, and that the two equations are actually
independent. However, this is not the case. By examining the two equations
independently, we are failing to account for the wider range of a larger political process
that is taking place.

Secondly, the bivariate probit model that accounts for selection demonstrates that
the variables that influence extended immediate deterrence success are statistically
significant after the selection effects have been accounted for. Therein lies the major
reason for using this modeling technique: it accounts for problems that tend to go
unrecognized when individual probit estimates are made.

6. 3. 2 General Implications of the Unified Model using Bivariate Probit Modeling

Before turning to the substantive interpretations of the variables that influence
extended immediate deterrence success, I wish to delineate the success that the bivariate
probit with selection has in testing the signaling model outlined in chapter 4. It cannot be
noted enough that extended deterrence is a sequential process. As the empirical evidence
in Table 6.7 demonstrates, costly signaling through investment plays a major role in
determining when extended general deterrence will succeed, and when extended

immediate deterrence will be successful if general deterrence falters.

213

As was seen in the equilibrium analysis of the models in chapter 4. greater
investments on the part of the defender lead to a situation of certainty, and thus lead to
deterrence success. These help ensure deterrence credibility at the level of extended
general deterrence. And, as the unified model demonstrates, these investments,
especially in the guise of arms and bases, help in resolving crises as well. Additionally.
the model lends support to the argument made by F earon (1994a, 1994b) that domestic
political costs play in resolving a crisis. And, as a majority of deterrence theorists
suggest, the role of military might is also quite important at both levels of deterrence.

The results here demonstrate the predictive qualities of the signaling model quite
strongly. In cases where enough investments have been made by a defender on behalf of
a protege, extended general deterrence succeeds. And, when it is challenged, these
investments help to resolve the crisis without escalation. Additionally, the model
predicts, and is borne out by the statistical results, that under certain conditions
deterrence, both extended and immediate, will fail. As the empirical estimates show,
when the issue at stake for the attacker is territorial in nature, extended immediate
deterrence is much more likely to fail. As much of the historical evidence suggests, cases
of deterrence failure tend to emerge from territorial disputes. Thus, the empirical
findings in this chapter, notably with the bivariate probit modeling technique, lend a large
amount of support to the costly signaling theory developed in this dissertation.

The statistical modeling technique helps correct for deficiencies found in

modeling the two deterrence scenarios as distinct processes, and accounts for the

selection bias that often occurs within studies of deterrence. With these thoughts in mind,

214

 

I turn now to a substantive interpretation of the variables that influence extended

immediate deterrence success.

6.3.3 Predicting Extended Immediate Deterrence Success

Having noted that the estimates provided by the bivariate probit model differ for

the extended immediate deterrence equation, I now turn to an analysis of the substantive

effects of these variables on influencing extended immediate deterrence success. Once

again I use the marginal effects technique to test the influence of each independent

variable on the success of extended immediate deterrence success. Table 6.6 presents

these substantive effects.

TABLE 6.7

SUBSTANTIVE EFFECTS OF INDEPENDENT VARIABLES ON

EXTENDED IMMEDIATE DETERRENCE SUCCESS

 

 

 

 

 

 

 

 

 

 

 

 

Variable Variable Held at Variable Held at , Percentage Change
Minimum Value Maximum Value

IMMBAL .09 .16 +07

STBAL .14 .303 +.163
DEMC .132 .232 +.10
BASE*ARMS .10 .25 +.15
TERRITORY .32 -.689 -.36
FIRMBFLE .08 .142 +06

 

 

 

As we examine the factors that influence extended immediate deterrence, we are

immediately struck by the fact that military capabilities play a much greater role in

determining whether deterrence succeeds or fails. Once again, the short-term military

balance is most important, and the immediate balance of forces is also quite influential.

215

This is quite logical and intuitive, given that states do not desire to fight prolonged wars.
As in the case of extended general deterrence, the long-term capabilities of a state do not
appear to matter.

If we turn back to the Serbian Bulgarian War of ’1 885, we notice that the
immediate balance of forces favored the Bulgarian offensive aimed at the Serb forces.
However, when the Bulgarian leadership factored in the Austrian intervention, and the
power that the Austrian army would have been able to provide to the Serbs, they felt that
they had not option but to back away from their demands. The possibility of fighting one
of the strongest armies in EurOpe was not appealing to the Bulgars at the time.

Interestingly enough, the presence of nuclear weapons on the part of the defender
does not serve a deterrent purpose. In the 1978-79 Sino-Vietnamese War, the Soviet
Union had the ability to defend its Vietnamese protége’ with nuclear weapons. The
possibility of a nuclear strike, albeit remote, did not deter the Chinese from launching an
attack against their Vietnamese adversaries. Nuclear weapons do not seem to play a role
in extended deterrence encounters, most likely because they do not pose credible threats.

The establishment of both bases and arms, which are costly investments on the
part of a defender, have a major impact on determining whether extended immediate
deterrence succeeds or fails. If we fluctuate the interactive variable BASE*ARMS from
its empirical low to its empirical high, the probability that extended immediate deterrence
will succeed increases by 15%. This again lends support to the costly signaling argument
advanced in this dissertation. Such actions are costly for defenders to undertake, and to
throw these investments away without any attempt at protecting them is not in their best

interest.

216

The American ability during the Cold War, both within Europe and in regards to
Taiwan, exemplifies this finding. The United States maintained a vast network of
military bases in Western Europe and Taiwan, with many of these bases being
concentrated in West Germany. The United States during the Cold War was the largest
arms supplier to both these countries. In many cases, the United States was the sole
provider of weaponry for these states. These investments demonstrated American
commitment to these states, and also allowed them to protect themselves in the event of a
crisis.

Having noted the fact that leaders do not like to waste costs that have been
invested, we now turn to an analysis of the role that domestic constraints play on
extended immediate deterrence encounters. As F earon (1994a, 1994b) has noted,
defenders who are democratic in nature tend not to back down in crises due to pressures
that they will feel at home. The empirical evidence presented here corroborates this
finding. As we fluctuate the variable DEMC from its empirical low to its empirical high,
the probability of deterrence succeeding increases by 10%. This links in quite well with
the concept that costly signals can be viewed as investments made by defenders. If the
leaders of a democratic state renege on their promises, and let their investments go to
waste, then they may face removal at the hands of the pe0ple they are sworn to serve.

The role of updated beliefs plays a minor role in determining when extended
immediate deterrence will succeed or fail. If a defender adopts a policy that is firm but
flexible (FIRMBFLE), the likelihood that extended immediate deterrence will succeed
increases by 6%. In the case of the United States being a defender of West Germany, we

notice that the United States was never willing to concede on issues regarding Berlin to

217

the Soviets. However, in 1958, and in 1961, the United States was able to make
political overtures to events in the Middle East and in East Asia that mollified the
Soviets, and showed that while the United States would be firm on the issue of Berlin. it
was willing to work with the Soviet Union on c00perating in other spheres.

The final factor that influences extended immediate deterrence success is the issue
of territory. If the issue at stake is territorial in nature for the aggressor, the likelihood
that extended immediate deterrence will succeed decreases by 38%. This is not
surprising, given that a majority of conflicts throughout human history have been
territorial in nature. If we examine again the onset of the Second World War, we see this
in stark detail. Germany, seeking living space in the East, launched an attack on Poland
in September 193 9. Despite French and British intervention, Germany could not be
deterred and forced to back down.

6.4 Conclusion

The more unified structure found within the bivariate probit modeling
specification is a more accurate test of the theory outlined in chapter 3, and formalized in
chapter 4. In many cases, potential aggressors witness events within their strategic
environments. and find it opportune to initiate a crisis and make a demand of another
state. However, as was argued in chapter 3, when extended general deterrence fails, the
success or failure of extended immediate deterrence is largely based upon military might
and the willingness to become involved in a conflict.

The case of the Serbian-Bulgarian War of 1885 provides a perfect illustration of
this concept. Bulgaria, sensing weakness on the part of the Serbs, launched an attack.

Serbia was weakened. and Bulgaria was rapidly gaining on the offensive. However, the

218

threat of Austrian intervention changed the decision calculus of the Bulgarian leaders.
Faced with overwhelming military power, the Bulgarians were forced to back down from
their demands.

Extended general deterrence failure, therefore, can be viewed as a process that
sets the stage for further escalation of conflict. As was noted, in only 46 of the 302 cases
of attempted extended general deterrence did a crisis erupt. The results of the model
above suggest that costly signaling does help enhance a defender’s credibility and create
an aura of certainty within the international system. If we as researchers are interested in
deterrence failure and the escalation of crises, either in the case of alliance commitments
or not, we need to address the issue of interdependence between the two distinct events.
Failure to do so leads to erroneous conclusions at worst and only partially correct
conclusions at best.

In this chapter I have empirically estimated the implications of my costly
signaling model, and have found support for the propositions that were derived from the
formal analysis. Using a dataset that I developed to test cases of extended deterrence via
alliance, I have discerned that costly signaling plays a major role in deterrence
effectiveness. Rather than focus on military capabilities, as so many deterrence theorists
have in the past, I have chosen to examine the actions that a state takes in order to make
its commitments to an ally credible. While not denying the central role of military
prowess and might in the study of deterrence, the findings posed in this chapter suggest
that we need to concentrate more on the interactions between structural environment and
the goals of the individual actors. As references to the various cases suggest, military

might has not always been the best means of providing a peaceful status quo.

This chapter has also highlighted some of the methodological concerns that arise
when we attempt to analyze complex political processes. As the bivariate probit analysis
demonstrates, in order to understand extended immediate deterrence success, we need to
understand the conditions in which extended general deterrence faltered in the first place.
If we are to engage in studies that address such concerns, we need to utilize appropriate
methodologies in order to arrive at reasonable statistical inferences.

I have maintained throughout this dissertation, and demonstrated theoretically and
empirically, that costly signaling can induce deterrence stability. I have argued that we
need to utilize appropriate methodological tools in order to understand both strands of
deterrence as an interdependent process. The empirical estimates lend credence to the
arguments posed in earlier chapter. The empirical results corroborate the findings of the
theoretical model, and are rigorously derived from the formal specification of the theory
developed earlier. Thus. we witness here both additive and integrative cumulation on the
question of alliances and their impact on deterrence success or failure.

In the following chapter I summarize the findings of this dissertation. I discuss
the cumulative findings that have emerged from it, and make suggestions for further

research agendas in the area of alliances and deterrence.

220

CHAPTER 7

CONCLUSION

7.1 Introduction

This dissertation began with the simple question “what impact do existing
alliances have on the success of extended deterrence?” As I have endeavored to answer
this question. I have ventured into game-theoretic models of signaling, cross-sectional
and cross-temporal statistical issues, and in-depth analyses of various alliances. I have
learned that this question, while simple in origin, does not have a simple answer.

In the following pages of this conclusion I summarize the implications of my
analysis, both formal and empirical. In doing so I discuss what has been learned in this
dissertation regarding the additive and integrative cumulative knowledge surrounding the
role that alliances play in deterrence. I conclude with some thoughts on future work and
refinements that can be drawn from this dissertation.

7.2 Implications from the Formal Analysis: Integrative Cumulation

As the theory outlined in chapter 3 suggests, and the formal analysis in chapter 4
demonstrates, investing costs into a prote’gé can help ensure effective extended
deterrence. And, if extended general deterrence fails, the investments made by the
defender help in resolving the crisis by providing a means for the beleaguered protege to
defend itself.

The formalization of the theory sheds light on two aspects of the
alliance/deterrence puzzle. First of all, it introduced a costly signaling approach to

Morrow’s (1991) framework of autonomy and security. Secondly, it demonstrates in

221

some detail the role that limited information and beliefs play in deterrence scenarios. I
examine each in turn.
7. 2. 1 Costly Signaling and the Autonomy/Security T radeoff

As was noted earlier, states enter into alliances for various reasons. Small states
seek security, while larger states seek other benefits. In the cases examined here,
deterrence is the security benefit that is provided by a larger state to its minor power
protégé. In return, the prote'gé provides some sort of benefit that can be broadly
construed as autonomy.

Many previous studies have focused on the ability of a defender to provide
security to a protege, and this study is no exception. As the empirical evidence
demonstrates, the ability of a defender to invest costly measures in a protege distinctively
helps extended general deterrence succeed. And, in the cases in which extended general
deterrence breaks down, these invested costs make it more likely that the defender will
actually intervene on its protégé’s behalf.

In most of the cases analyzed, the defender was able to gain some meaningful
autonomy benefit from its protégé. Such a benefit, either in terms of increased access to
foreign markets or change in a protégé’s policy, was quite indicative of when extended
general deterrence would succeed, and when it would fail. In a tangential manner, the
exchange of autonomy benefits could be viewed as a costly signal sent from the protege
to the defender, as a means of indicating how serious and tight the alliance relationship is.

Of course, there are some limitations on what the formal model explains and
predicts. Costly signaling does not ensure perfect general deterrence. Challenges do

occur. and at times they do escalate to large-scale conflict. The beginnings of the Second

222

World War have their roots in such deterrence failure, in spite of the costly signals that
France and Great Britain sent regarding their guarantees to Polish sovereignty. Certain
types of states can be deterred due to fear of third party intervention. After 1961, the
Soviet Union ceases to challenge the United States over the status of divided Germany.
Alternatively, with the exception of a very minor challenge in the Second Balkan War,
the presence of an Austrian-Romanian alliance deterred Bulgarian designs on Romanian
territory. However, Germany under Hitler, or North Korea for much of the Cold War,
continually made challenges against states that had powerful allies. The possibility that
alliances may not deter aggression, in spite of costly signaling, highlights the role that
information and beliefs play in extended deterrence situations. It is to a discussion of
these topics that I now turn my attention.
7.2.2 Information and Beliefs

Costly signaling and the transfer of security and autonomy benefits are public
events. What is unknown is the willingness of a defender to come to the aid of an
embattled ally. Potential aggressors must base their initial beliefs upon something, and
costly signaling is a good mechanism for demonstrating the willingness a defender
possesses in regards to fighting on behalf of a protégé. However, why does costly
signaling not ensure perfect deterrence? This question continually arises. The answer, I
maintain. and as was supported by the formal analysis in chapter 4, is that different states
have different willingness functions when it comes to making challenges and attempting
to achieve their goals. Alternatively, different states also have different willingness
1 functions when it comes to deciding whether or not to intervene in a crisis on behalf of an

ally.

223

These different types of players, as explicated in the formal analysis, place
different values on the status quo. In the case of a defender, it is largely based upon the
benefits received less the costs invested in a protégé. In regards to a challenger, it is
largely based upon capability ratios and the perceptions it has of the sincerity of the
defender. These perceptions, I have argued, are often as (if not more) important as
military balances.

Once again I take an example from the onset of the Second World War to
illustrate the role that beliefs and information play in an attacker’s decision calculus.
Germany realized that Poland was allied with Great Britain and France. Both defenders,
particularly the British, had made promises in regards to Polish security. The military
might of Germany was less than the overall military might of the French, British, and
Poles. If one adheres to a strict interpretation of deterrence, these capabilities should
have prevented the Second World War from ever taking place. However, Germany was
not deterred, and Poland was attacked. The question remains. Why?

The answer to this question, as is suggested by the formal analysis, is that certain
states value changing the status quo to such an extent that they are difficult, if not
impossible, to deter. Much of this. I argue, can be drawn from the actions a challenger
witnesses. So, in the case of Germany prior to the 1939, Hitler was able to witness the
British and French failure to act during the Rhineland Crisis of 1936, and the debacle at
Munich in 193 8. Given this, then, Germany’s initial beliefs were that Britain and France
would do nothing to protect Poland.

Of course, upon witnessing actions taken by a defender once a crisis is initiated,

challengers have the ability to update their beliefs. Some challengers will back down, as

224

the Bulgarians did in 1885, upon witnessing Austrian intervention on behalf of their
Serbian allies. On the other hand, Germany continued to press its attack in 193 9, and
extended immediate deterrence also failed. These illustrations suggest two items of
interest. First of all, certain states are difficult, if not impossible, to deter once they have
initiated a crisis. They hold the issue at stake so dear to their heart that they are willing to
run the risk of fighting (and potentially losing) a war in order to attain their goals.
Secondly, it suggests that bluffing is not a strategy for defenders to follow. If a
commitment is credible, then this will be revealed at the time of an initial challenge. If
the commitment is not credible, then why make it in the first place? Such counterfactual
events are difficult to test, it is true, but the evidence from this analysis suggests that the
more a commitment can be made credible through costly signaling, the greater the impact
it will have on ensuring extended deterrence effectiveness.
7.3 Implications from the Empirical Analysis: Additive Cumulation

Moving from the formalization of the theory to the empirical testing of the theory
against the historical record, my focus turns more to the additive cumulative knowledge
that has been gained from the statistical analysis conducted in. this dissertation. This is
not to suggest, however, that it is only through formal modeling that we can uncover
integrative knowledge. The findings of the signaling model are merely conjectures if
they are not tested against evidence obtained from the empirical realm. In that vein,
therefore, I discuss the empirical findings of my models, and show how they belie some
of the conventional wisdom regarding alliances and extended deterrence. I then turn to a

discussion of the methodologies used and the data that were developed, and elaborate

225

upon how they have helped us move forward in the study of alliances and their deterrent
impacts.
7.3.1 Contradictory Findings

Three items of importance emerge from the empirical examination of data
surrounding extended deterrence via alliance. First of all. alliances are quite good
mechanisms for providing extended general deterrence. Of the 302 possible
opportunities for extended general deterrence via alliance to fail, it only did so on 46 of
those occasions. And, in those 46 cases, the defender failed to intervene in only 5 of
these events.

This finding casts suspicion on the results of Sabrosky (1980), who suggested that
alliances tend to be quite unreliable. While it should be noted that my results are
applicable to the set of deterrence alliances that I analyzed, and may not be generalizable
to all cases of alliances, the results are still intriguing and call for further scrutiny on the
alliances and war question. Alliances in this case do appear to be quite good at deterring
aggression, and defenders do come to the aid of embattled proteges. The rationale
behind. this. I maintain, is the impact of costly signaling.

The impact of costly signaling is the second issue of importance to be discussed
here. More than beliefs, more than military capabilities, the trade-off between security
and autonomy benefits are paramount in determining the success or failure of extended
general deterrence. As was noted in the formal analysis, the goods or services traded
between allies serves as a public indication of the value that each state places in the
’ alliance relationship. This differs from many of the earlier studies of extended

deterrence, which have tended to focus on the role of capabilities and their ability to

226

make a deterrence commitment credible. The substantive effects that were presented in
the previous chapter demonstrate that costly signals, particularly the security benefits that
are transferred from a defender to a protégé, are quite important in maintaining the status
quo.

The third item of interest that emerges from the empirical analysis of the costly
signaling theory developed here is the fact that different factors influence extended
deterrence at its different levels. Much of the previous work that has been conducted on
extended deterrence has focused on the immediate level, due to the easier task of
identifying success and failure. However, as the bivariate probit with selection results
indicate, the two processes of extended general deterrence failure and the success or
failure of extended immediate deterrence cannot be separated. As the results
demonstrate, failure to account for the larger process can lead to erroneous findings.

Costly signaling arguments have rarely been subjected to empirical scrutiny, so
the findings in this dissertation are both novel and unique. They demonstrate that it is
possible to operationalize concepts from formal models and submit them to rigorous
empirical testing. Having noted that, I turn to a discussion of the genuine additive
findings this dissertation has contributed to our knowledge of alliances and deterrence.
7.3.2 New Data and Methodological Techniques

As mentioned above, the dataset that was created in this dissertation was designed
to test models that were the result of both extended general deterrence success and
failure, and extended immediate deterrence success and failure. These data represent a
novel development in the study of extended deterrence, although it should be recalled that

they are specific to the alliances within the dataset. As noted in chapter 2, there exist no

227

scientific studies that analyze extended deterrence as a complete process. The data
devised here sought to remedy this oversight, and used the presence of alliances as agents
of extended deterrence as a means of determining a set of cases. This is truly additive
cumulation in the sense that Zinnes (1976, 1981) mentions in her work.

Additionally, two new methodologies were utilized in the empirical work
conducted here. I used new techniques deveIOped to account for cross-temporal issues.
As the model above found, autocorrelation was not present. However. the procedure is
simple enough, and can correct temporal issues so readily, that it should become a tool
used by all analysts who encounter binary cross-sectional time series data.

Finally, in an attempt to test the entire deterrence process, from crisis initiation to
potential escalation to conflict, I utilized a methodology, the bivariate probit model with
selection, to account for the selection effects that often occur within the study of
deterrence. If we are to model conflict initiation and escalation properly, we need to
account for the fact that they are part of a larger process. At each level within the process
different factors may be at work, and this methodology allows us to capture these
disparate conditions.

7.4 Concluding Remarks and Directions for Future Research

This dissertation has provided a good foundation for understanding the role that
costly signaling plays in providing successful deterrence to minor power allies.
However, there is still work to be done on the topic of extended deterrence via alliance.

In the remainder of this chapter I discuss three directions for future research.

228

7. 4.1 An Extension of the Data

As was noted earlier, the dataset developed here is limited to the time period
1870-1984. One of the first tasks that should be undertaken is an effort to extend the
dataset, both back in time and forwards. A preliminary examination was made of the
alliances that existed between 1816-1870, and while some of them can be viewed as
deterrence alliances, there do exist data limitations. Data on international trade, for
example, is often difficult to find for periods prior to 1870. Such data that does exist is
often suspect. Similar concerns exist regarding the establishment of military bases
abroad, and the transfer of weaponry. Such data is needed by the international relations
research community, and a project that seeks to locate such data and make it available
would be a boon to researchers worldwide.

Secondly, the dataset needs to be extended forward in time. The data as it exists
now ends while the Cold War is still in full force. Now that the superpower rivalry has
faded into the annals of history, we need to know if alliances still provide means of
security to minor powers, or has the functioning of international relations also changed
along with the international climate. Given the heated debate surrounding NATO
expansion, and events taking place within the former Yugoslavia, it is apparent that
alliances still play a role in international affairs. Time will tell if they continue to play a
deterrence role, and if this is a policy that statesmen should be eager to pursue.

7.4.2 The Role of Domestic Politics in Alliance Affairs

A second topic that is ripe for further examination is the role that domestic

political considerations play in alliance politics. The decision to join an alliance, as any

foreign policy decision, is often a source of controversy. In the analysis above, both

229

formal and statistical, domestic constraints were hypothesized to play a role in the
success or failure of extended deterrence in both of its incarnations.

In this vein, therefore, two issues need to be addressed. First of all, what
constraints operate on a domestic level besides the ones analyzed in this study. In
democracies, do electoral cycles play a role? Are there additional constraints upon
authoritarian regimes that may have a pacifying effect on their leaders? What about
domestic economic conditions? All of these factors may play a role in determining the
willingness of state leaders to become involved in a crisis.

Secondly, the focus should also encompass the leadership in the government of
the attacker. Much has been made in the literature on domestic politics and international
conflict on the domestic constraints that are faced by a defender. Indeed, the entire
audience costs argument focuses on this scenario. However, what restrictions do the
leadership within an aggressor state face? Do similar constraints Operate for the initiation
Of a crisis, or are other factors at work? In order to provide a more nuanced rigorous
theory of domestic politics and their influence on alliance behavior, such considerations
must be accounted for.

7. 4. 3 Analyzing Crisis Bargaining

A third area Of interest is a further analysis, both formally and empirically, is a
detailed examination Of the events that occur as a crisis progresses from extended general
deterrence failure to resolution either through extended immediate deterrence success or
escalation to conflict. Formally, we need to account for the actions each side takes as the
crisis escalates. Through the use of discount rates, we can arrive at conclusions Of what

level of hostility a defender (or aggressor, or even a protégé) is willing to absorb before it

230

decides to capitulate to an opponent. Through the use of bargaining models we can begin
to understand some more Of the crisis dynamics that occur in the process Of extended
deterrence failure and crisis escalation
7. 4.4 Some Concluding Thoughts

This dissertation has called into question some of the conventional wisdom on the
role that alliances play on ensuring effective deterrence. Alliances actually do have a fair
amount of success in deterring aggression, and at assisting an embattled protége' if a crisis
does escalate. However, this does not mean that all alliances will always serve as
deterrent structures, or does it imply that all states can be deterred. Hitler’s Germany and
North Korea are but two examples of states that repeatedly challenged the status quo.

The implications from this dissertation suggest that costly signaling can be quite
beneficial in ensuring credibility in international politics. It should be noted that the
costly signaling framework applies to situations outside of deterrence as well, and that
this framework is not bound to studies of security and conflict. The theory does suggest
that we must move beyond “mere” structural theorizing when it comes to strategic issues,
and look at the incentive structures, domestically and internationally, that states face.
Doing so allows us to focus on the Opportunities that states encounter, and also to focus
on the calculations that drive their willingness to act.

lntemational politics, both in the empirical world and in the research community,
is moving forward at a dizzying pace. This dissertation has attempted to examine one of
the recurring entities in international politics, alliances, and determine when they will
provide deterrence benefits. In that, it has succeeded to a large extent. With the end of

the Cold War and the restructuring of the international system, many believed that the

231

age of alliances and concerns over deterrence were past. But as long as states threaten
others, and as long as major powers are willing to defend smaller states, deterrence will
be a cornerstone of security policy for many states. In many instances, deterrence will be
a benefit extended through a formal alliance. Only by. understanding how it has
functioned in the past can we hope to understand and predict the impact that military

alliances will have on ensuring deterrence success in the future.

232

APPENDICES

233

APPENDIX A

PROOFS OF THE EQUILIBRIA FROM THE MODEL WITH COMPLETE
INFORMATION

Proof of Equilibrium 1

Equilibrium I : In a game with complete and perfect information, status quo (SQ) is a
subgame perfect equilibrium in pure strategies if the following conditions are met: tA =2
and t3 =1 and tc = 1.

Proof: We can solve this equilibria through straightforward backwards induction. If we
start at the last two nodes, A has a choice to make between CNALL and BD. At the other
node A must choose between CNAB and BD*. At the one node A prefers BD to CNALL,
since when I), =2, the potential costs outweigh its utility for the new resolution Of the
issue; x-c> 0, with 0 being the value to A for the status quo. At the other node, A prefers
CNAB to BD“. Again, by definition of tA =2, A would prefer to fight against a lone B,
since kA >kB.

We now turn to the choice made by C. C must choose whether to intervene (int)
or not to intervene (~int). Since tc =1 , it is a strongly committed defender, and will
always intervene on B’s behalf. its utility for the current status quO is greater than any
costs that may be incurred; m-y > d+c. C prefers a trilateral conflict to having B fight
alone.

B must now make a choice between resist (r) and not resist (~r). B’s choice lies
in resisting, where it knows the outcome will be BD, or in not resisting, and having the

outcome be acquiescence by B (ACQB). Since t3 =1, and it knows that its defender is

willing to intervene, B is willing to resist aggression on the part of A. This keeps with

234

the logic of alliances as capability aggregation units interested in preserving security. If
aggression is to take place, B would rather have an ally fight alongside him.

The final choice belongs to A. A must choose between fighting a trilateral
conflict (BD) and accepting the current resolution Of the issue (SQ). By the preference
orderings above, A would rather not incur the costs of initiating a crisis and then being
forced to back down. Hence, a crisis is never begun, and extended general deterrence
prevails. The subgame perfect equilibrium can be written as {~initiate, ~escalate; resist;
intervene}, where the first choice refers to A’s chosen strategies, the second choice refers
to 8’3 chosen strategy, and the third choice refers to C’s chosen strategy.

Proof of Equilibrium 2

Equilibrium 2: In a game with complete and perfect information, acquiescence by B

(A C QB) is a subgame perfect equilibrium in pure strategies if I ) tA=1, t3=1 , and tC= 2;
or 2) m=2. tB=I, tc=2 or 3) tA=3, t3=1, tc=2.

Proof: 1 will address each case separately.

Case I : In the first case, tA=l , t3=l, and tc= 2. We start again at the final two nodes, and
work backwards to the top of the model. If tA=1, then it has greater military capabilities
than the combined attributes of the alliance it faces: kA >k3c. At the left node A must
choose between a trilateral conflict (CNALL) and backing down (BD). At the right node
A must choose between a bilateral conflict (CNAB) and backing down without
intervention by C (BD*). On both sides A chooses conflict, since x > c + d.

The choice now falls to C, who must decide whether or not to intervene. Since
tc= 2, it is a weakly committed defender, unwilling to intervene militarily. Its utility for

the status quo is less than the costs of becoming involved in a potential militarized

235

conflict; m-y < c + d. Hence, in choosing between the possibility of trilateral conflict
(CNALL) over bilateral conflict (CNAB), C will choose to not intervene.

The choice now falls to B. Since t3=1, we know that it is weaker in terms of
military capabilities vis a vis A. And, since B knows that its ally will not intervene
militarily, it realizes that its choice lies between fighting A alone and acquiescing to A’s
demand. Thus, B chooses tO acquiesce (ACQB).

Once again the final move belongs to A. It knows that B will give in to its
demand, because its ally has refused to intervene. Therefore, A will initiate a crisis
against B, and B will accede to A’s demand. The subgame perfect equilibrium can be
written as {initiate, escalate; ~resist; ~ intervene}.

Case 2: In the second case, tA=2, tB=1, tc=2. . If we start at the last two nodes, A has a
choice tO make between CNALL and BD. At the other node A must choose between CNAB
and BD“. At the one node A prefers BD to CNALL, since when tA =2, the potential costs
outweigh its utility for the new resolution of the issue; c+d > x. At the other node, A
prefers CNAB to BD*. Again, by definition of tA =2, A would prefer to fight against a
lone B, since kA >k3.

The choice now falls to C, who must decide whether or not to intervene. Since
tc= 2, it is a weakly committed defender, unwilling to intervene militarily. Its utility for
the status quo is less than the costs of becoming involved in a potential militarized
conflict; m-y < c + (1. Hence, in choosing between the possibility Of trilateral conflict
(CNALL) over bilateral conflict (CNAB), C will choose to not intervene.

The choice now falls to B. Since tB=1, we know that it is weaker in terms Of

military capabilities vis a vis A. And, since B knows that its ally will not intervene

236

militarily, it realizes that its choice lies between fighting A alone and acquiescing to A’s
demand. Thus, B chooses to acquiesce (ACQB).

The final move again belongs to A. Although A is weaker than the alliance as a
whole, it is still stronger than B. Knowing that C willnot intervene, A can initiate a crisis
against B without fearing third party intervention. Hence, A makes its demand. and B
once again gives in. The subgame perfect equilibrium is {initiate, escalate; ~resist;
~intervene}.

Case 3: We start again at the end of the game, and work backwards. A moves first, and
must decide whether it prefers conflict (either bilateral or trilateral) to either forms Of
backing down. On either side Of the game tree A chooses conflict over backing down. A
knows that its capabilities are equivalent to those of B and C combined, so it must be
willing to fight in order to change the issue at stake. x > c+d.

The choice now falls to C, who must decide whether or not to intervene. Since
tc= 2, it is a weakly committed defender, unwilling to intervene militarily. Its utility for b
is less than the costs of becoming involved in a potential militarized conflict; m- y < c +
d. Hence, in choosing between the possibility Of trilateral conflict (CNALL) over bilateral
conflict (CNAB), C will choose to not intervene.

The choice now falls to B. Since t3=1, we know that it is weaker in terms of
military capabilities vis a vis A. And, since B knows that its ally will not intervene
militarily, it realizes that its choice lies between fighting A alone and acquiescing to A’s
demand. Thus, B chooses tO acquiesce (ACQB).

The final move again belongs to A. Although A is weaker than the alliance as a

whole. it is still stronger than B. Knowing that C will not intervene, A can initiate a crisis

237

against B without fearing third party intervention. Hence, A makes its demand, and B
once again gives in. The subgame perfect equilibrium is {initiate, escalate; ~resist;
~intervene}.

Proof of Equilibrium 3

Equilibrium 3: In a game with complete and perfiect information, trilateral conflict
(CNAu) is a subgame perfect equilibrium in pure strategies if 1) tA=I, t3 =1, and tr =1 or
2) tA=3, (3 =1, and tc =1.

Proof: 1 will address each case separately.

Case I : We start again at the final two nodes, and work backwards to the top of the
model. If tA=l , then it has greater military capabilities than the combined attributes Of
the alliance it faces: kA >k3c. At the left node A must choose between a trilateral conflict
(CNALL) and backing down (BD). At the right node A must choose between a bilateral
conflict (CNAB) and backing down without intervention by C (BD*). On both sides A
chooses conflict, since x > c + d.

We now turn to the choice made by C. C must choose whether to intervene (int.)
or not to intervene (~int). Since tc =1, it is a strongly committed defender, and will
always intervene on B’s behalf. Its utility for the current status quo is greater than any
costs that may be incurred; m- y > c + d. C prefers a trilateral conflict to having B fight
alone.

B must now make a choice between resist (r) and not resist (~r). B’s choice lies
in resisting, where it knows the outcome will be BD, or in not resisting, and having the

outcome be acquiescence by B (ACQB). Since t3 =1, and it knows that its defender is

willing to intervene, B is willing to resist aggression on the part of A.

238

Finally, the choice once again belongs to A. A must choose between not initiating
a crisis and accepting the status quO (SQ), or initiating the crisis and ending up with a
trilateral conflict (CNALL). Since A is militarily superior to the BC alliance, it initiates
the crisis, and conflict is the result. The subgame perfect equilibrium is {initiate,
escalate; resist; intervene}.

Case 2: We start again at the end of the game, and work backwards. A moves first, and
must decide whether it prefers conflict (either bilateral or trilateral) to either forms of
backing down. On either side of the game tree A chooses conflict over backing down. A
knows that its capabilities are equivalent to those of B and C combined, so it must be
willing to fight in order to change the issue at stake. U(x) > 0.

We now turn to the choice made by C. C must choose whether to intervene (int)
or not to intervene (~int). Since tc =1, it is a strongly committed defender, and will
always intervene on B’s behalf. Its utility for the current status quo is greater than any
costs that may be incurred; m- y > c + d. C prefers a trilateral conflict to having B fight
alone.

B must now make a choice between resist (r) and not resist (~r). B’s choice lies
in resisting, where it knows the outcome will be BD, or in not resisting, and having the
outcome be acquiescence by B (ACQB). Since t3 =1, and it knows that its defender is
willing to intervene, B is willing to resist aggression on the part Of A.

Finally, the choice once again belongs to A. A must choose between not initiating
a crisis and accepting the status quo (SQ), or initiating the crisis and ending up with a
trilateral conflict (CNALL). Since A, while not militarily superior to the BC alliance,

places a high value on changing the status quo. Thus, it initiates the crisis, and trilateral

239

conflict is the result. The subgame perfect equilibrium is {initiate, escalate; resist;
intervene}.
Proof of Equilibrium 4
Equilibrium 4: If the game is played under conditions of complete and perfect
information, then bilateral conflict (CNAB) is a subgame perfect equilibrium outcome in
pure strategies if we relax I) the assumption of kA>k3 and assume that kA=k3 and 2)
assume that under conditions of power parity, B prefers C N A 3 to AC Q3, Hence the game
is played under conditions in which m=2, t3=2, and tc =2.
Proof: This is a special case in which tB=2. In all the other instances, kA>kBC or kA<
km. This is the only case in which the capabilities Of B are a factor. We start again with
A’s decision at the final two nodes.

A has a choice to make between CNALL and BD. At the other node A must
choose between CNAB and BD*. At the one node A prefers BD to CNALL, since when
tA =2, the potential costs outweigh its utility for the new resolution of the issue; c+d > x.
At the other node, A prefers CNAB to BD*.

The choice now falls to C, who must decide whether or not to intervene. Since
tc= 2, it is a weakly committed defender, unwilling to intervene militarily. Its utility for y
is less than the costs Of becoming involved in a potential militarized conflict;
(U) y < c + d. Hence, in choosing between the possibility of trilateral conflict (CNALL)
over bilateral conflict (CNAB), C will choose to not intervene.

B now must make its choice, and choose between bilateral conflict (CNAB) and
acquiescence (ACQB). In this instance, B is not inferior to A in terms Of military

capabilities. Therefore, B will resist A’s demand, and choose the possibility of fighting

over giving in.

240

A once again makes the final decision. Although A is weaker than the alliance as
a whole, it is as strong B. Knowing that C will not intervene, A can initiate a crisis
against B without fearing third party intervention. Hence, A makes its demand, and B
fights back. The outcome in equilibrium is bilateral Conflict. The subgame perfect

equilibrium is {initiate, escalate; resist; ~intervene}.

241

APPENDIX B

DERIVATIONS OF BELIEFS VIA BAYES’ RULE

In this appendix I provide the derivations via Bayes’ Rule to determine the prior
beliefs that A possesses upon witnessing a strong signal from C.
Recall from above that
1: (signal stronglweak) = or
7t(signal weaklweak) = l-Ot

Additionally, recall that A is actually a strongly committed defender with
probability O,
and is actually a weakly committed defender with probability l-O.

Given this, we can then calculate the prior beliefs that A possesses upon

witnessing a strong signal from C.

11A(C =strong l C signals strong) = ptstrongj" pt signals stronglstrong)

p (strong)*p (signals stronglstrong) + p(weak)* p Signals stronglweak)

pA(C =strong] C signals strong) = (O) 1 = O
(O) 1 + (l-O)(0t) O + Ot+OOL

11A(C =weak | C signals strong) = = DIweal_<I* pl siggals stronglwea_k_)
p (weak)*p (signals stronglweak) + p(strong)* p Signals stronglstrong)

 

11A(C =weak | C Signals strong) = i (l-O) or = OL-OLO
(l-O)(OL)+O(1) a-aO+O

242

APPENDIX C
PROOFS OF THE EQUILIBRIA OF THE SIGNALING MODEL WITH
INCOMPLETE INFORMATION

In this appendix I provide proofs for the six equilibria that emerge from the
signaling model in chapter 4. It should be recalled that a weak signal on the part of a
defender provides identical results as the model with complete information did. Thus,
this appendix focuses on the scenarios in which the defender (C) Signals strong, and the
challenger (A) must determine what type of defender it is facing. In each scenario I
outline the strategies that each state follows, and how the updated beliefs of A are
calculated for the equilibria that arise.
Scenario 1: tA =1 and tc =1
Strategies

In this first scenario, we can rule out by definition a number of strategy choices
for player A. If A is a type 1 player, it always has an incentive to initiate and escalate a
dispute. Its dominant strategy is to play {initiate, escalate}. And, if C is a type 1 player,
it will always come to the assistance of its protege. And, given C’s type, B will always
resist.

Given this, there exists only one case in which equilibrium occurs. This is
{initiate, escalate; resist; intervene}. The outcome is a trilateral conflict.

The logic is as follows. First of all, as a type 1 player, A always has an incentive
to initiate a crisis. And, due to its military capabilities, it always has an incentive to

escalate the crisis. B, knowing that its ally will intervene, due to C’s type, will always

243

resist. And C, being a type I defender. will always intervene on behalf of an embattled
protege.

In this case, the updating Of beliefs on the part of the attacker does not matter. It
will always play the same strategy, regardless of what type Of defender it is facing.
Beliefs

Thus. A’s initial beliefs that C is a strongly committed defender are

11,. = O , as derived in Appendix B.
O + Ot+O0t

And, A’s initial beliefs that C is a weakly committed defender are

M = Ot-OLO , as derived in Appendix B.
Ot- OLO + O

And its updated beliefs are uA*= 1.0. These updated beliefs are calculated by

using Baye’s Rule, and update the first set of beliefs outlined above.

HA*(C =strong! C signals strong) = (O) 1
(9) 1 + (l-O)(a)

TO update, this becomes
1(1) =1.0
1(1) + (1-9)(0)

The fully specified equilibrium is {initiate, escalate; resist; intervene};
“A = L; “K”: 1.0
O + OL+OOL

Scenario 2: ti, = 1 and tc =2
Strategies

In this second scenario, we can once again rule out a number of strategy
combinations. If A is a type 1 player, it always has an incentive to initiate and escalate a

dispute. Its dominant strategy is to play {initiate, escalate}. And, if C is a type 2 player,

244

it will not come to the assistance of its protege. And, given C’s type, B will always
acquiesce.

Given this, there exists only one case in which equilibrium occurs. This is
{initiate, escalate; ~resist; ~intervene}. The outcome is acquiesence on the part of B

The logic is as follows. Again, as in the first scenario A is a type 1 player. Thus.
A always has an incentive to initiate a crisis. And, due to its military capabilities, it
always has an incentive to escalate the crisis. B, knowing that its ally will not intervene,
due to C’s type, will acquiesce, rather than resist. And C, being a type 2 defender, will
not intervene on behalf Of an embattled protege. Noting that C will not intervene, A
would escalate the crisis if B chose to resist. However, since B chose to concede the
issue at hand, the escalation does not occur.
Beliefs

Once again the updating of beliefs on the part of the attacker does not matter. It
will always play the same strategy, regardless of what type of defender it is facing.

Thus, A’s initial beliefs that C is a strongly committed defender are

”A = O , as derived in Appendix B.
O + Ot+O0t

And, A’s initial beliefs that C is a weakly committed defender are

H»: = Ot-OtO , as derived in Appendix B.
Ot- OLO + O

And its updated beliefs are uA*= 1.0. These updated beliefs are calculated by
using Baye’s Rule, and update the second set of beliefs outlined above.

11A*(C =weak l C Signals strong) = f ( l-O) a
(1'@)(a) + 69(1)

245

This becomes HI) = l

1(1) + (0)1
The fully Specified equilibrium is {initiate, escalate; ~resist; ~intervene};
11A =m: nA*= 1.0.
01- OLO + O

Scenario 3: tA = 2 and tc =1
Strategies

This scenario is more difficult to analyze, due to the fact that as a type 2 player, A
does not have a dominant strategy set of {initiate, escalate}. This is especially so if A
believes that it is facing a type 1 defender. Thus, there are five distinct cases that emerge.
Case 1:

A (a) plays initiate.
And (b) plays escalate if C plays intervene.

B plays resist if A plays initiate.

C plays intervene if A plays initiate.

This case is not an equilibrium. If A witnesses an intervention by C, it knows that
C is a type 1 defender. In such a scenario, A will choose to change its strategy and not
escalate against B.
Case 2:

A plays ~initiate.

B plays ~ resist.

C plays ~ intervene.

This case is not an equilibrium. Under conditions of uncertainty, A always has an

incentive to challenge the status quo. Thus, it will always initiate a crisis, in an attempt to

determine if it can make gains.

246

Case 3:

A (a) plays initiate
And (b) plays ~ escalate if C plays ~ intervene.

B plays resist.

C plays ~intervene.

This case is not an equilibrium for two reasons. First Of all. if C is a type 2 player
and does not intervene, then A has no incentive to back down. Instead, A will escalate
the crisis against B. Secondly, if C is a type 2 player, B will not resist in the first place,
SO there will be no opportunity for A to escalate the crisis anyway.

Case 4:

A (a) plays initiate
And (b) plays escalate if C plays ~intervene.

B plays ~ resist.

C plays ~intervene.

This case is not an equilibrium. By defmition of C’s type, it will always intervene
on behalf of its protégé if a crisis erupts.
Case 5:

A (a) plays initiate
And (b) plays ~ escalate if C plays intervene.

B plays resist.

C plays intervene.

This case is an equilibrium. A has an incentive to make a probe, given that it is

uncertain about C’s type. And, upon witnessing C’s intervention, it will back down.

247

Beliefs

In this case the updating of beliefs on the part of the attacker does matter. A will
play a different strategy, depending upon what type of defender it is facing.

Thus, A’s initial beliefs that C is a strongly cOmmitted defender are

pi, = O , as derived in Appendix B.
O + Ot+OOt

And, A’s initial beliefs that C is a weakly committed defender are

m = Ot-OtO , as derived in Appendix B.
Ot- OLO + O

And its updated beliefs are pA*= 1.0. These updated beliefs are calculated by
using Baye’s Rule, and update the first set of beliefs outlined above.

uA*(C =strong] C signals strong) = (O) 1
(O) 1 + (1-O)(0t)

To update, this becomes
1( 1 ) =1 .0
1(1) + (1-9)(0)

The fully specified equilibrium is { initiate, ~escalate; resist; intervene};

('9 ; “A: 1.0*
O + Ot+OOt

11A

Scenario 4: tA = 2 and tc =2

Once again this scenario is more difficult to analyze, due to the fact that as a type

2 player, A does not have a dominant strategy set of {initiate, escalate}. This is

especially SO if A believes that it is facing a type I defender. Thus, there are four distinct

cases that emerge.

248

Case I:

A (a) plays initiate.
And (b) plays escalate if C plays intervene.

B plays resist if A plays initiate.

C plays intervene if A plays initiate.

This case is not an equilibrium. If A witnesses an intervention by C, it knows that
C is a type 1 defender. In such a scenario, A will choose to change its strategy and not
escalate against B.

Case 2:

A plays ~initiate.

B plays ~ resist.

C plays ~ intervene.

This case is not an equilibrium. Under conditions of uncertainty, A always has an
incentive to challenge the status quo. Thus, it will always initiate a crisis. in an attempt to
determine if it can make gains.

Case 3:

A (a) plays initiate
And (b) plays ~ escalate if C plays ~ intervene.

B plays resist.

C plays ~intervene.

This case is not an equilibrium for two reasons. First Of all, if C is a type 2 player
and does not intervene, then A has no incentive to back down. Instead, A will escalate
the crisis against B. Secondly, if C is a type 2 player, B will not resist in the first place,

so there will be no opportunity for A to escalate the crisis anyway.

249

It should be noted that if the assumption that kA>kB is relaxed, and we assume that
kA= k3, then this case can be an equilibrium. The logic behind this can be found in
Appendix A.

Case 4:

A (a) plays initiate
And (b) plays escalate if C plays ~intervene.

B plays ~ resist.

C plays ~intervene.
This case is an equilibrium. Since C refuses to intervene, B will acquiesce to A’s
demands. C’s inaction reveals its type. The equilibrium is {initiate, escalate; ~resist;
’intervene}. The outcome is acquiescence by B.
Beliefs

In this case the updating of beliefs on the part of the attacker does matter. A will
play a different strategy, depending upon what type Of defender it is facing.

Thus, A’s initial beliefs that C is a strongly committed defender are

MA = O as derived in Appendix B.
O + or+OOt

 

And, A’s initial beliefs that C is a weakly committed defender are

11A = Ol-OLO, as derived in Appendix B.
01- OO + O

And its updated beliefs are pA*= 1.0. These updated beliefs are calculated by

using Baye’s Rule, and update the second set of beliefs outlined above.

 

 

HA*(C =weak l C signals strong) = i ( I-Ol or
(1'@)(01)+ @(1)
This becomes 1(1) = 1
1(1) + (0)1

250

Thus, the fully specified equilibrium is { initiate, escalate; ~resist; ‘intervene};

MA = 01-09 ' llA*= 1.0.
or- OtO + O

 

Scenario 5: tA = 3 and tc =1
Strategies

This scenario is easier to analyze than is scenario 2. If A is a type 3 player, it
always has an incentive to initiate and escalate a dispute. We can thus rule out any
strategy combinations that have A not initiating or escalating a crisis. Its dominant
strategy is to play { initiate, escalate}. And, if C is a type 1 player, it will always come to
the assistance of its protege. And, given C’s type, B will always resist.

Given this, there exists only one case in which equilibrium occurs. This is
{initiate, escalate; resist; intervene}. The outcome is a trilateral conflict.

The logic is as follows. First Of all, as a type 3 player, A always has an incentive
to initiate a crisis. And, due to its military capabilities, it always has an incentive to
escalate the crisis. B. knowing that its ally will intervene, due to C’s type, will always
resist. And C, being a type 1 defender, will always intervene on behalf of an embattled
protége'.

In this case, the updating of beliefs on the part of the attacker does not matter. It
will always play the same strategy, regardless of what type Of defender it is facing.
Beliefs

Thus. A’S initial beliefs that C is a strongly committed defender are

MA = O , as derived in Appendix B.
O + a+O0t

And, A’s initial beliefs that C is a weakly committed defender are

251

1,1,, = Ot-OLO , as derived in Appendix B.
Ot- OtO + O

And its updated beliefs are uA*= 1.0. These updated beliefs are calculated by
using Baye’s Rule, and update the first set of beliefs outlined above.

uA*(C =strong! C signals strong) = (O) l
((9) 1 + (1-9)(0t)

This becomes
1(1) =1.0
1(1) + (1-9)(0)

The fully specified equilibrium is {initiate, escalate; resist; intervene};

MA = (’9 ; “A1“: 1.0.
O + Ot+OOt

Scenario 6: tA = 3 and tc =2
Strategies
In this final scenario, we can once again rule out a number of strategy
combinations. If A is a type 3 player, it always has an incentive, to initiate and escalate a
dispute. Its dominant strategy is to play {initiate, escalate}. And, if C is a type 2 player,
it will not come to the assistance of its protege. And, given C’s type, B will always
acquiesce.
Given this, there exists only one case in which equilibrium occurs. This is
{ initiate, escalate: ~resist; ~intervene}. The outcome is acquiesence on the part of B
The logic is as follows. Again, as in the fifth scenario A is a type 3 player. Thus,
A always has an incentive to initiate a crisis. And, due to its military capabilities, it
I always has an incentive to escalate the crisis. B, knowing that its ally will not intervene,
due to C’s type, will acquiesce, rather than resist. And C, being a type 2 defender, will

not intervene on behalf of an embattled protege. Noting that C will not intervene. A

252

would escalate the crisis if B chose to resist. However, since B chose to concede the
issue at hand, the escalation does not occur.
Beliefs
Once again the updating of beliefs on the part of the attacker does not matter. It
will always play the same strategy, regardless of what type of defender it is facing.
Thus, A’s initial beliefs that C is a strongly committed defender are

m = O , as derived in Appendix B.
O + Ot+OOl

And, A’s initial beliefs that C is a weakly committed defender are

up, = Ot-OtO , as derived in Appendix B.
Ot- OLO + O

And its updated beliefs are uA*= 1.0. These updated beliefs are calculated by

using Baye’s Rule, and update the second set Of beliefs outlined above.

 

 

MA*(C =weak | C signals strong) = _ (l-O) or
(1-@)(0L) + 9(1)
This becomes 1(1) = l
1(1) + (0)1
The fully specified equilibrium is {initiate, escalate; ~resist; ~intervene};
11A 2%; 1111*: 1-0-
a- OtO + O

253

APPENDIX D

Deterrence Alliances, 1870-1984

This appendix discusses the various alliances. that were used in the empirical
portions of chapters 5 and 6. In each description I denote the members, potential
aggressors, the duration of the alliance, the duration of the deterrence commitment. I
then provide a description of the alliance and its rationale for existence. discuss any crises
that emerged between the potential aggressor(s) and the protégé, and delineate the source
material used in determining the information outlined above.

1. Alliance Members: United Kingdom, Ottoman Empire

Potential Aggress0r(s): Russia

Duration of Alliance: 6/4/1878- 12/31/1880
Duration of Deterrence Commitment: 6/4/1878- 12/31/1880
Reasons for Alliance 's Existence: In 1878 Turkey and Russia concluded yet another war.
which was ended with the Treaty Of San Stefano and subsequent revisions at the
Congress Of Berlin. The implications of the treaty shifted the balance of power in the
Balkans, with the creation of Bulgaria and Serbia, and fear on the part of the Turks and
English of increased Russian involvement in the region. The treaty Of alliance signed
between Great Britain and Turkey permitted the British access to the Black Sea, while
specifically pledging British support for the Ottoman Empire against Russia. In

exchange for military assistance from Britain, Turkey relinquished control of Cyprus to

the British.
Crisis Involvement: None.

Sources: Hurst (1972); Jelavich (1973, 1991), Langer (1950).

254

2. Alliance Members :Austria-Hungary, Romania

Potential Aggressor(s): Bulgaria

Duration of Alliance: October 1883-December 1915
Duration of Deterrence Commitment: October 1883-December 1915
Reasons for Alliance’s Existence: The aftermath of the Russo-Turkish War. formalized at
Congress Of Berlin, established a number of new states on the Balkan Peninsula.
Romania, being one of these new states, was ever fearful of territorial expansion by its
Bulgarian neighbors. On Austria’s part, alliance with Romania provided access to
Romania’s natural resources, as well as provided a means to confront Russian expansion.
Hence, while the Romanians were wary of Bulgarian designs on Romanian territory,
Austria used this alliance as a means of containing Russian expansion in the Balkans.
Access to Romanian territory was the benefit for Austria, while deterrence benefits
against the Bulgarian threat was the benefit accrued by the Romanian state.
Crisis Involvement: In the history of the AustrO-Romanian alliance, only one minor crisis
erupted. In 1913, in the aftermath of the First Balkan War, which effectively ended
Ottoman influence on the European continent, Bulgaria turned on its former allies. In an
attempt to gain more territory, Bulgaria attacked Greece, Romania, and Serbia. Austria
made a diplomatic threat to Bulgaria regarding its commitment to Romania, and the
combined forces of Greece, Romania. and Serbia were able to rapidly defeat Bulgaria and
end her aspirations for a greater Bulgarian state in the Balkans.
Sources: Choucri and North (1975); Hurst (1972); Jelavich (197 3, 1991), Langer (1950),

Taylor ( I 954).

255

3. Alliance Members: Germany, Romania

Potential Aggressor(s): Bulgaria

Duration of Alliance: October 1883-December 1915
Duration of Deterrence Commitment: October 1883-December 1915
Reasons for Alliance’s Existence: The aftermath of the Russo-Turkish War, formalized at
Congress Of Berlin, established a number of new states on the Balkan Peninsula.
Romania, being one of these new states, was ever fearful Of territorial expansion by its
Bulgarian neighbors. On Germany’s part, alliance with Romania provided access to
Romania’s natural resources. as well as provided a means to confront Russian expansion.
Bismarck in particular was wary Of Russian encroachment in this region, and even more
than the Austrians were concerned about containing the Russian threat. Hence, while the
Romanians were wary Of Bulgarian designs on Romanian territory, Germany used this
alliance as a means of containing Russian expansion in the Balkans. Access to Romanian
‘ territory was the benefit for Germany, while deterrence benefits against the Bulgarian
threat was the benefit accrued by the Romanian state.
Crisis Involvement: In the history of the German-Romanian alliance, only one minor
crisis erupted. In 1913, in the aftermath of the First Balkan War, which effectively ended
Ottoman influence on the European continent, Bulgaria turned on its former allies. In an
attempt to gain more territory, Bulgaria attacked Greece, Romania, and Serbia. Germany
also made a diplomatic threat to Bulgaria regarding its commitment to Romania, and the
combined forces Of Greece, Romania, and Serbia were able to rapidly defeat Bulgaria and

end her aspirations for a greater Bulgarian state in the Balkans.

256

Sources: Choucri and North, 1975; Craig (1966), Hurst (1972); Jelavich (1973, 1991).
Langer (1950), Taylor (1954).
4. Alliance Members: Austria-Hungary, Serbia

Potential Aggressor(s): Bulgaria

Duration of Alliance: January 1881-December 1895
Duration of Deterrence Commitment: January l881-December I895
Reasons for Alliance’s Existence: In the aftermath Of the Congress of Berlin, Austria
made a deliberate overture to maintain a presence in the Balkans. In large part this was
due to the possibility of annexing Bosnia-Herzegovina. Serbia for its part was concerned
over Bulgarian attempts to expand at the expense of its neighbors. In an attempt to keep
Serbia under Austrian suzerainty, if not outright domination, Serbia was compelled in
1881 to Sign a treaty of alliance with the Austrian empire. In return for protection, Serbia
was to make no foreign policy decisions without Austrian consent, allow Austria access
to her markets at favorable terms, and permit the passage of Austrian troops through
Serbia in time Of war.
Crisis Involvement: The only war to take place on the European continent in the decade
of the 18808 involved Serbia and Bulgaria. Serbia, witnessing political turmoil within
Bulgaria, struck first. Unfortunately for the Serbs, the Bulgarians were able to rally and
make threatening overtures towards Serbian territory. Faced with imminent defeat.
Serbia appealed to its Austrian ally, who threatened the Bulgarians with Austrian
intervention if they did not cease their advancement towards Belgrade. The Serbian-
’ Bulgarian War lasted scarcely a fortnight in the autumn of 1885 (November 14-28), and

established Bulgaria as the dominant state on the Balkan Peninsula.

257

Sources: Blainey (1988, 60-63), Hurst (1972), Jelavich (1973, 1991), Langer ( 1950).
Taylor (1954).
5. Alliance Members: Russia. China

Potential Aggressor(s): Japan

Duration of A lliance: May 1896 to December 1900
Duration of Deterrence Commitment: May 1896 to December 1900
Reasonsfor Alliance’s Existence: In the aftermath Of a disastrous war with Japan, China
looked elsewhere to guarantee her security. Russia, hoping to expand in the Far East.
provided the perfect opportunity. In return for a base at Port Arthur (established in
1898), Russia provided China with military aid and support against possible renewed
aggression by the Japanese.
Crisis Involvement: None
Sources: Hurst (1972), Huth (1988), Taylor (1954).
6. Alliance Members: France, Belgium

Potential Aggressor(s): Germany

Duration of A ll iance: September 1921 to December 1936
Duration of Deterrence Commitment: September 1921 to December 1936
Reasons for Alliance’s Existence: Given Belgium’s experiences in the First World War, it
is not surprising that she turned to a great power to ensure her future. France, on the
other hand, was interested in containing Germany, and ensuring that she was confronted
in the event of German rearrnament.
Crisis Involvement: In 1936, three years after Hitler’s ascension to power, Germany

sought tO reoccupy the Rhineland. This region Of western Germany was ordered

258

demilitarized by the treaty of Versailles, and had been a constant reminder to the German
people of their defeat in the First World War. In March 1936 German troops entered the
demilitarized Rhineland, which was threatening to the Belgians. Regardless of the fact
that this was a major violation Of the Versailles Treaty, the French did little but protest to
the League Of Nations. While the major powers denounced the German actions, no
military moves were made to prevent Hitler’s actions. In large part this failure of France
to stand up to German aggression led to the cessation of the Franco-Belgian alliance at
the end of 1936, with the proclamation Of Belgian neutrality.
Sources: Brecher and Wilkenfeld, (1997): Grenville (1974); Hurst (1974), Lebow (1981),
Reiter ( 1996).
7. Alliance Members: France, Poland

Potential A ggressor(s) : Germany

Duration of Alliance: February 1921 to September 1939
Duration of Deterrence Commitment: February 1921 to September 1939
Reasonsfor Alliance’s Existence: The First World War transferred much former German
territory to the newly independent state of Poland. After numerous clashes in the early
19208, Polish-German relations settled into an uneasy peaceful coexistence. Poland,
fearful of German desires for her territory, particularly the Danzig area, sought alliance
ties with France. The arrangements called for France to guarantee Polish security, and
established greater trading ties between the two states.
Crisis Involvement: The German invasion Of Poland on September 1, 1939 sparked the

Second World War. In the aftermath Of the Munich debacle, France took a stronger

259

approach to German aggression. When Poland’s sovereignty was violated in the autumn
of 1939, France was quick to declare war on Germany.
Sources: Brecher and Wilkenfeld (1997), Khong, (1996), Seton-Watson ( I967).
Schweller (I 998), Thompson (1997).
8. Alliance Members: F rance, Czechoslovakia

Potential Aggressor(s): Germany

Duration of A ll lance: I 924
Duration of Deterrence Commitment: 1924
Reasons for Alliance ’5' Existence: In 1924, due to Czech fears of German resurgence, the
French Signed a treaty of alliance with Czechoslovakia. Limited in nature, it was
replaced in 1925.
Crisis Involvement: None.
Sources: Grenville (1974), Hurst (1972), Seton-Watson (1967).
9. Alliance Members: France, Czechoslovakia

Potential A ggressor(s) : Germany

Duration of Alliance: 1925-1939
Duration of Deterrence Commitment: 1925-1939
Reasons for Alliance’s Existence: In 1925, France signed a tighter treaty of alliance with
Czechoslovakia, including provisions for military intervention in case Czech sovereignty
were threatened.
Crisis Involvement: The crises surrounding Czechoslovakia and Germany in 1938 and
1939 underscore the failure Of British and French foreign policy in the 19305. In 1938,

demanding that German people in the Sudetenland portion Of Czechoslovakia be

260

admitted to the German Reich, Hitler threatened war. The French, although bound by a
treaty Of alliance with the Czechs, signed away the Sudetenland at the Munich conference
without a Struggle. Scarcely was the ink dry on that document than German troops
annexed the remainder of Czechoslovakia in the early spring of 1939. In spite of their
treaty, the French did nothing but protest diplomatically.
Sources: Brecher and Wilkenfeld (1997), Craig (1966), Seton-Watson (1967).
I 0. Alliance Members: Italy, Romania

Potential Aggressor(s): Bulgaria

Duration of Alliance: 1926-1930
Duration of Deterrence Commitment: 1926-1930
Reasons for Alliance ’s Existence: Romanian-Bulgarian relations had been tense ever
since the formation of the Balkan states after 1880. In the aftermath of the First World
War, the interest in a greater Bulgaria emerged again. Italy, seeking a role as a great
power in European affairs, allied with Romania against the Bulgarian threat in exchange
for trade and influence within Romanian affairs. The short-lived alliance accomplished
its task, and kept the Bulgarians at bay.
Crisis Involvement: In the one clash that occurs between Bulgaria and Romania, in 1936.

Italy intervened with a show of force. Bulgaria, overwhelmed, was unable to gain its

territorial Objectives and backed down.
Sources: Burgwyn (1997), Seton-Watson (1967).
I 1. Alliance Members: Italy, Albania

Potential Aggressor(s): Yugoslavia

Duration of A Iliance: 1927-1939

261

Duration of Deterrence Commitment: 1927-1939
Reasons for Alliance’s Existence: Italy had long sought control over Albania, largely due
to its ports and locale. Albania, on its part, feared its newly formed Yugoslav neighbor.
In 1937, Italy formalized its relationship with Albania in the guise of a security alliance.
It maintained this until 1939. when it invaded Albania and occupied it for the duration of
the Second World War.
Crisis Involvement: In 1927, Yugoslavia was embroiled in conflict with Albania over
territorial issues. Italy intervened on behalf of its minor ally, and was able to prevent
Yugoslavia from gaining hegemony over Albania.
Sources: Brecher and Wilkenfeld (1997); Burgwyn (1997); Seton-Watson (1967).
12. Alliance Members: Great Britain, Egypt

Potential A ggressor(s) : Italy

Duration of A lliance: 1936-1956
Duration of Deterrence Commitment: 1936-I943
Reasons for Alliance’s Existence: The British granted Egypt sovereignty in 1936, and
signed an alliance Of 20 years duration with the Egyptian government. Fearful Of Italian
expansion in the Mediterranean and in Africa, Egypt agreed to permit the British access
to the Suez Canal, and permitted the maintenance of British bases on Egyptian soil.
Crisis Involvement: During the Second World War, Britain and Italy fought battles in the
desert from 1940-1943. After the defeat of Italy. Egypt resumed its attempt to remove

the British from their territory, finally succeeding in 1956.

Sources: Brecher and Wilkenfeld (1997); Clodfelter (I992); Marlowe (1965).

262

I 3. Alliance Members: Soviet Union, Outer Mongolia

Potential Aggressor(s): Japan (until 1945); China (after 1945)

Duration of Alliance: 1936-I984
Duration of Deterrence Commitment: 1936-1962
Reasons for Alliance 's Existence: Mongolia, after gaining independence, was continually
fearful Of Japanese aggression. In 1936 the Soviet Union signed a treaty of alliance with
Mongolia, and relegated the Mongolian state to a virtual satellite of the Soviet Union.
The Soviet Union was able to use Mongolia as a base for its increasing conflicts with
China and Japan, and in return Mongolia gained security and all of its military supplies
from its Soviet patron. Prior to 1945, Japan was the major concern of the Mongolians,
and, from 1945-1962, when territorial differences were settled, the Mongolians were
wary Of communist China.
Crisis Involvement: Throughout the 19305, as Japan extended its empire into China and
Manchunko, Mongolia continually sparred with Japanese troops. In 1936 the Soviet
Union involved itself in the conflict between Japan and Mongolia, and found itself
involved again in 1938 and 1939, in the Changkugeng and Nomonhan conflicts. After
the Second World War, Mongolia had a clash over territorial issues with China in 1947-
1948, which the Soviets became involved in militarily as well.
Sources: Brecher and Wilkenfeld (1997); Coox (1977); Friters (1949); Tang (1959);
Tillema (1991).
14. Alliance Members: United Kingdom, Poland

Potential Aggressor(s): Germany

Duration of Alliance: 1939

263

Duration of Deterrence Commitment: 1939
Reasons for Alliance ’s Existence: In the aftermath of Munich, and German demands
made on Poland regarding Danzig, the British realized that German aggression must be
halted on the continent. In the spring of 1939 Prime. Minister Chamberlain made a
statement to the British Parliament stating that Britain would go to war to protect Polish
sovereignty.
Crisis Involvement: Britain’s commitment was soon tested. On September I, 193 9,
Germany invaded Poland. Britain declared war on Germany, and the Second World War
began.
Sources: Brecher and Wilkenfeld (1997); Craig (1966); Khong, (1996), Seton-Watson
(1967), Schweller (1998); Thompson (1997).
15. Alliance Members: United Kingdom, Jordan

Potential Aggressor(s): Israel

Duration of Alliance: 1936-I957
Duration of Deterrence Commitment: 1949-1957
Reasons for Alliance ’s Existence: Great Britain and Jordan signed an alliance agreement
in 1936, when the Jordanian state was granted autonomy. Although tensions between the
states flared during the Second World War, they cooled with the formation Of the Israeli
state in 1948 and the subsequent war. In 1949 Great Britain issued a public pledge,
saying it would protect Jordan from Israeli aggression. In return, the British were able to
maintain their rights tO airfields and military bases within Jordan.
Crisis Involvement: Although numerous border skirmishes occurred after 1948, only two

major crises erupted before the end of the British-Jordanian alliance in 1957. In 1953. as

264

 

a response to Jordanian infiltration into Israel, Israeli troops fired upon the Jordanian
village Of Qibya. After Jordan appealed to the United Nations and the British, Israel
promised to cease its border incursions against Jordan. In 1956, Israeli forces retaliated
for Jordanian raids against the city Of Qalqilya. Jordan responded by asking for Iraqi
assistance militarily, and inviting Iraqi troops into Jordan. Israel threatened retaliation if
this occurred. The crisis ended with a British Show of planes. and an Israeli promise not
to invade Jordan unless she were physically attacked.
Sources: Blechman (1972); Brecher and Wilkenfeld (1997); Shimshoni (1988); Tillema
(1991).
I 6. Alliance Members: United States, South Korea

Potential A ggressor(s) : North Korea

Duration of Alliance: 1954-1984
Duration of Deterrence Commitment: I954-1984
Reasons for Alliance ’s Existence: Inthe aftermath Of the Korean War, the United States
determined that an increased American presence in Southeast Asia was necessary to
prevent the further spread Of communism in that region. In 1954. the United States
formalized its security arrangements with South Korea in the form Of a military alliance.
The United States gained access to markets for trade. and military bases in South Korea
in exchange for providing weapons and security to the South Korean regime.
Crisis Involvement: Almost immediately after the cease fire was signed in 1953 ending
the Korean War, violations on the part of North Korea began. At various points during
the 1950s and 19605, the North Koreans continually infiltrated the South, in an attempt to

destabilize the South Korean regime. The United States troops stationed in South Korea

265

continually mobilized and repelled incursions from the North. In 1968 the United States
was embroiled in a crisis with North Korea regarding the USS Pueblo, and in 1976 the
United States was once again enmeshed in a crisis regarding North Korean attacks on
American and South Korean soldiers. Since then the KOrean peninsula has been
relatively quiet.
Sources: Bandow and Carpenter (1992); Brecher and Wilkenfeld (1997); Curtis and Ham
(1983); George and Smoke (1974); Tillema (1991).
17. Alliance Members: United States, Taiwan

Potential A ggressor(s) : China

Duration of Alliance: 1955-1979
Duration of Deterrence Commitment: 1955-1979
Reasons for Alliance ’s Existence: In its attempts to contain Soviet and communist
influence worldwide, the United States increasingly supported the Nationalist Chinese
regime on Taiwan. Communist China maintained that Taiwan was merely a rebellious
province, not an independent entity. Ratified in 1955, the United States treaty of alliance
with Taiwan guaranteed protection of the home islands, with vagaries mentioned
regarding the offshore islands of Quemoy and Matsu. In return, the United States was
able to gain export markets on Taiwan, as well as maintain various naval bases. The
alliance continued until 1979, when the United States formally recognized the People’s
Republic of China and ended its formal diplomatic relations with Taiwan.
Crisis Involvement: Four crises erupted for the United States regarding Taiwan. In 1954-
1955, the United States sent the Seventh Fleet to halt Chinese shelling of the offshore

islands, which, as was mentioned above, were not wholly within the protection purview

266

 

of the alliance treaty. Similar Situations occurred in 1958-59, and in 1962. In each
instance China made threatening overtures towards Taiwan, and the United States
continually intervened with a Show of naval force in order to force the Chinese to back
down.
Sources: Blechman and Kaplan (1979); Brecher and Wilkenfeld (1997); Christensen
(1996); George and Smoke (I974); Whiting (1975).
18. Alliance Members: United Kingdom, Malaysia

Potential Aggressor(s): Indonesia

Duration of Alliance: 1957-1 971
Duration of Deterrence Commitment: I957-197l
Reasons for Alliance ’s Existence: In 1957, Malaysia became an independent member of
the British Commonwealth. The British, concerned with Indonesian hostility directed
towards their former colony, and seeking to maintain good economic ties and military
bases in Malaysia, formed an alliance in 195 7 with the Malay state. In return for
protection being provided by British bases and forces, Malaysia maintained its good
economic relations with the British and permitted British garrisons to remain on the
peninsula.
Crisis Involvement: Indonesian Opposition to the Malay Federation came to a head in
1962. and Indonesian incursions began in early 1963. Until 1966, when a treaty ended
hostilities between Malaysia and Indonesia, guerrilla warfare raged along the borders Of
Malaysia, instigated by the Indonesian government. British soldiers stationed on the
Malay Peninsula took an active role in suppressing these raids.

Sources: Brecher and Wilkenfeld (1997); Tillema (1991).

267

I 9. Alliance Members: Soviet Union, Vietnam

Potential Aggressor(s): China

Duration of Alliance: 1978-1984
Duration of Deterrence Commitment: 1978-1984
Reasons for Alliance ’s Existence: China and Vietnam, communist neighbors, finally split
due to ideological differences in the aftermath of American involvement in the Vietnam
War. Vietnam allied itself closely with the Soviet Union, desiring the weaponry and
trade it would receive from the international leader of the communist bloc. In return, the
Soviet Union gained access to former American military bases that were remnants of the
American involvement in Southeast Asia, and had a strong ally on China’s border.
Crisis Involvement: China and Vietnam became embroiled in two conflicts between 1978
and 1984. On Christmas Day, 1978, Chinese troops invaded Vietnam in a Show of
displeasure over Vietnamese actions in Cambodia. The Soviet Union failed to provide
any Significant military support. Again, in 1984, China sent troops across the Vietnamese
border in an attempt to once again restrain Vietnamese actions in Cambodia. The Soviet
navy staged a landing of 400 troops off the shore of Vietnam as a show of solidarity, but
made no direct military intervention.
Sources: Brecher and Wilkenfeld (1997); Horn (1987); McGregor (1988); Organski and
Kugler (1980).
20. Alliance Members: United States, West Germany

Potential Aggressor(s): Soviet Union

Duration of Alliance: 1954-1984

Duration of Deterrence Commitment: l954-1972

268

Reasonsfor Alliance’s Existence: As tensions between the United States and the Soviet
Union began to increase after the Second World War, occupied Germany became the
symbol Of a European continent divided by ideology. With the establishment of two
German states in 1949, a Soviet blockade of Berlin in 1948, and American concern in
communist expansion in the aftermath of the Korean War, the United States decided to
bring West Germany into the NATO alliance. West German entry into NATO
guaranteed an American presence in Europe, as the United States sought to contain
communist influence worldwide. In return, the West German state was guaranteed
security from its eastern neighbors.
Crisis Involvement: Two major crises erupted between West Germany and the Soviet
Union from 1954-1972, and both were concerned with the status of divided Berlin. In
1958, and lasting until late in 1959, the Soviet Union declared that it desired that Berlin
be turned into a “Free City”, and Western occupation Of West Berlin should come to an
end. This would have effectively brought West Berlin under East German control, and
deeply damaged America’s reputation regarding its commitment to a free Europe. After
both superpowers mobilized military forces, an agreement was reached excluding nuclear
weapons from Berlin, but it remained a divided city.

In 1961, after two years of political wrangling over the status Of Berlin, the Soviet
Union erected the Berlin Wall. This effectively ended East German defections to the
West for Berlin, and stabilized East German society. The fear on the part of the United
States was of a Soviet military attempt to unify Berlin by force. The United States

mobilized its military, and Khrushchev ended his demands for a decision to be made on

269

the Berlin question. In 1972 the occupying powers finally signed a treaty that solidified
the status of Berlin as a divided city.

Sources: Brecher and Wilkenfeld (1997): Hanrieder (I989); Slusser (1973).

270

BIBLIOGRAPHY

271

Bibliography

Achen, Christopher and Duncan Snidal. 1989. “Rational Deterrence Theory and
Comparative Case Studies.” World Politics 41 (2): 153-169.

Albrecht-Carlie, Renee. 1973. A Diplomatic History of Europe: Since the Congress of
Vienna, revised edition. New York: Harper & Row

Aldrich, John H. and Forrest D. Nelson. 1984. Linear Probability, Logit, and Probit
Models. Newbury Park: Sage Publications.

Aldrich, John, John Sullivan and Eugene Brogatt. 1989. "Foreign Affairs and Issue
Voting: DO Presidential Candidates 'Waltz' before a Blind Audience?“ American Political
Science Review 83 (1): 123-41.

Allison, Graham. 1971. Essence of Decision. Boston: Little and Brown.

Altfeld, Michael. 1984. “The Decision to Ally: A Theory and Test." Western Political
Quarterly 37 (4): 523-44.

Altfeld, Michael and Bruce Bueno de Mesquita. 1979. “Choosing Sides in Wars.”
International Studies Quarterly 23 (1): 87-112.

Axelrod, Robert. 1984. The Evolution of Cooperation. New York: Basic Books.

Bandow, Doug and Ted Galen Carpenter, Eds. 1992. The (IS-South Korean Alliance:
Time for a Change. New Brunswick: Transaction Publishers.

Banks, Jeffrey. 1990. “Equilibrium Behavior in Crisis Bargaining Games. " American
Journal of Political Science 34 (3): 599-614.

---. 1991. Signaling Games in Political Science. New York: Harwood.

Banks, Jeffrey and Joel Sobel. 1987. “Equilibrium Selection in Signaling Games.”
Econometrica, 55 (3): 647-62.

Beck. Nathaniel. 1996. “Reporting Heteroskedasticity Consistent Standard Errors." The
Political Methodologist 7: 3-6.

Beck, Nathaniel and Jonathan Katz. 1995. "What to DO (and Not to Do) with Time-
SerieS--Cross-Section Data." American Political Science Review 89 (3): 634-47.

272

Beck. Nathaniel, Jonathan Katz and Richard Tucker. 1998. “Beyond Ordinary Logit:
Taking Time Seriously in Binary Time-Series Cross-Section Models.” American Journal
of Political Science 42 (4): 1260-88.

Bennett, D. Scott. 1997. Testing Alternative Models of Alliance Duration,
1816-1984." American Journal of Political Science 41(3): 846-78.

Bennett, D. Scott and Alan Stam. 1998. " Is Instrumental Rationality a
Universal Phenomenon?" Paper presented at the 56th Annual Meeting of the
Midwest Political Science Association. April, Chicago, IL

Blainey, Geoffrey. 1988. The Causes of War. 3rd edition. New York: The Free Press.

Blechman, Barry. 1972. “The Impact of Israel’s Reprisals on Behavior Of the Bordering
Arab Nations directed at Israel.” Journal of Conflict Resolution 16 (2): 155-81.

Brecher, Michael and Jonathan Wilkenfeld. 1997. A Study of Crisis. Ann Arbor:
University Of Michigan Press.

Brehm. John and Scott Gates. 1997. Working, Shirking, and Sabotage. Ann Arbor:
University Of Michigan Press.

Boulding, Kenneth. 1962. Conflict and Defense. New York: Harper and Row.

Brozska. Michael. 1982. “Arms Transfer Data Sources.” Journal of Conflict Resolution
26(1): 77-108.

Bueno de Mesquita, Bruce. 1978. “Systemic Polarization and the Occurrence and
Duration of War.” Journal of Conflict Resolution 22 (3): 241-68.

---. 1981. The War Trap. New Haven: Yale University Press.

---. 1985. “Toward a Scientific Understanding of lntemational Conflict: A Personal
View.” International Studies Quarterly 29 (2): 121-36.

Bueno de Mesquita, Bruce, Gary Woller, and Randolph Siverson. 1992. “War and the
Fate of Regimes: A Comparative Analysis.” American Political Science Review 86(3):
638-46.

Bueno de Mesquita, Bruce and David Lalman. 1992. War and Reason: Domestic and
International Imperatives. New Haven: Yale University Press.

Bueno de Mesquita, Bruce, and Randolph Siverson. 1995. “War and the Survival of
Political Leaders.” American Political Science Review 84 (3): 841-55.

273

Bueno de Mesquita. Bruce, James Morrow, and Ethan Zorick. 1997. “Capabilities.
Perception, and Escalation,” American Political Science Review 91 (1): 15-27.

Burgwyn, H. James. 1997 Italian Foreign Policy in the Interwar Period 1918-1940.
Westport: Praeger Publishers.

Bull, Hedley. I977. The Anarchical Society. New York: Columbia University Press.

Carlsnaes, Walter. 1992. “The Agent-Structure Problem in Foreign Policy Analysis.”
International Studies Quarterly 36 (3): 245-70.

Carlson, Lisa. 1998. “Crisis Escalation: An Empirical Test in the Context Of Extended
Deterrence.” International Interactions 24 (3): 225-54.

Carr, EH. 1949. The Twenty Years’ Crisis. London: Macmillan.

Chemoff, Fred. 1995. After Bipolarity: NA T0 and the Vanishing Threat. Ann Arbor:
University of Michigan Press.

Choucri, Nazli and Robert C. North. 1974. Nations in Conflict: National Growth and
International Violence. San Francisco: W. H. Freeman and Company.

Christensen, Thomas. 1996. Useful Adversaries: Grand Strategy, Domestic
Mobilization, and Sino-American Conflict, 1947-1958. Princeton: Princeton University
Press.

Christensen, Thomas and Jack Snyder. 1990. “Chain Gangs and Passed Bucks:

Predicting Alliance Patterns in Multipolarity.” International Organization 44 (2): 137-
68.

---. 1997. “Progressive Research on Degenerative Alliances.” American Political
Science Review 91 (4): 919-22.

Clodfelter, Micheal. 1992. Warfare and Armed Conflicts. London: McFarland.
Coneybeare, John and Todd Sandler. I991. “The Triple Entente and Triple Alliance
1880-1914: A Collective Goods Approach.” American Political Science Review 84(4):
1197-1206.

Coox, Alvin D. 1977. The Anatomy of a Small War: The Soviet-Japanese Struggle for
Changkufeng/Khasan, 1938. Westport: Greenwood Press.

Craig, Gordon. 1966. Germany, 1866-1945. New York: Oxford University Press.

Curtis, Gerald L. and Sung-joo Han. 1983. The (IS-South Korean Alliance. Lexington:
Lexington Books.

274

Dessler, David. 1989. “What’s at Stake in the Agency-Structure Debate? " International
Organization 43 (3): 441-73.

Deutsch, Karl and J. David Singer. 1964. “Multipolar Power Systems and lntemational
Stability.” World Politics 16 (4): 390-406.

Diehl, Paul. 1992. “What are they Fighting For? The Importance of Issues in
lntemational Conflict Research.” Journal of Peace Research 29 (3): 333-44.

Dion, Douglas. 1998. “Evidence and Inference in the Comparative Case Study.”
Comparative Political Studies 30 (2): 127-45.

Dubin, Jeffrey and Douglas Rivers. 1989. “Selection Bias in Linear Regression, Logit.
and Probit Models.” Sociological Methods and Research, 18 (3): 360-90.

Duke, Simon. 1989. United States Military Forces and Installations in EurOpe. Oxford:
Oxford University Press.

Duke, Simon and Wolfgang Krieger, Eds. 1993. US. Military Forces in Europe: The
Early Years, 1945-1970. Oxford: Oxford University Press.

Dupuy, R. Ernest and Trevor N. Dupuy. 1970. The Encyclopedia of Military History
flom 3500 B. C. to the Present. Revised edition. New York: Harper & Row.

Elman, Colin and Miriam Fendius Elman. 1997. “Lakatos and Neorealism: A Reply to
Vasquez.” American Political Science Review 91 (4): 923-26.

Eyerman, Joseph and Robert Hart. 1996. “An Empirical Test of the Audience Costs
Proposition: Democracy Speaks Louder than Words.” Journal of Conflict Resolution 40
(3): 597-617.

F earon, James. 1990. “Counterfactuals and Hypothesis Testing in Political Science.”
World Politics 43 (2): 169-95.

---. 1993. “Selection Effects and Deterrence.” Typescript. University Of Chicago.

---. 1994a. “Signaling versus the Balance Of Power and Interests.” Journal of Conflict
Resolution 38 (2): 236-69.

---. 1994b. “Domestic Political Audiences and the Escalation of lntemational Disputes.”
American Political Science Review 88 (4): 577-99.

---. 1995. “Rationalist Expectations for War.” International Organization 49(3): 379-
414.

275

---. 1997. “Signaling Foreign Policy Interests.” Journal ofConflict Resolution 41 (1):
68-90.

Friedman, Gil and Harvey Starr. 1997. Agency, Structure, and International Politics:
From Ontology to Empirical Inquiry. London: Routledge.

Friedman, Jeffrey, ed. 1996. The Rational Choice Controversy. New Haven: Yale
University Press.

Friters, Gerald M. 1949. Outer Mongolia and Its International Position. Baltimore: The
John Hopkins University Press.

Fudenburg, Drew and Jean Tirole. 1993. Game Theory. Cambridge: MIT Press.

Gartner. Scott Sigmund and Randolph Siverson. 1996. “War Expansion and War
Outcomes.” Journal of Conflict Resolution 40 ( 1): 4-15.

Gartzke, Erik and Michael Simon. 1996. “Gains from Trade versus Social Contract
Models: Testing Alternative Theories Of Alliance Formation.” Paper presented at the
annual meeting of the lntemational Studies Association Meeting, Midwest, October, St.
Louis.

Gates. Scott and Sherry Bennett Quinones. 1994. “Game Theoretic and Empirical
Methodologies: Ever the Two Shall Meet?” Political Methodologist 6 (2): 30-6.

Gates, Scott and Brian Humes. 1997. Games, Information, and Politics: Applying Game
Theoretic Models to Political Science. Ann Arbor: University of Michigan Press.

Gates, William and Katsuaki Terawasa. 1992. "Commitment, Threat Perceptions, and
Expenditures in a Defense Alliance." International Studies Quarterly 36 ( 1 ): 101-18.

Geddes, Barbara. 1991. “How the Cases You Choose Affect the Answers You Get:
Selection Bias in Comparative Politics.” Political Analysis 2: 131-50.

George, Alexander. 1979. “Case Studies and Theory Development: the Method Of
Structured. Focused Comparison.” In Paul Gordon, ed Diplomacy: New Approaches in
History, Theory, and Policy. New York: Free Press.

George, Alexander and Richard Smoke. 1974. Deterrence in American Foreign Policy:
Theory and Practice. New York: Columbia University Press.

---. 1989. “Deterrence in Foreign Policy.” World Politics 41 (2): 170-82.

Geva, Nehemia and Alex Mintz, Eds. 1997. Decision-making on War and Peace: The
Cognitive-Rational Debate. Boulder: Lynne Rienner.

276

Gibbons, Robert. 1992. Game Theory for Applied Economists. Princeton: Princeton
University Press.

Gibler, Douglas. 1996. “Alliances that never balance: the Territorial Settlement Treaty.”
Conflict Management and Peace Science 15 ( 1 ): 75-98.

Gibler, Douglas and John Vasquez. 1998. “Uncovering” the Dangerous Alliances. 1495-
1980.” International Studies Quarterly 42 (4): 785-807.

Gochman, Charles and Russell Leng. 1983. “Realpolitik and the road to war: an analysis
Of attributes and behavior.” International Studies Quarterly 27 (1 )2 97-120.

Gochman, Charles and Zeev Maoz. 1984. “Militarized Interstate Disputes: Patterns,
Processes and Procedures, 1816-1976.” Journal of Conflict Resolution, 28 (4): 585—616.

Goertz, Gary. 1994. Contexts of International Politics. Cambridge: Cambridge
University Press.

Gowa, Joanne. 1994. Allies, Adversaries. and International Trade. Princeton: Princeton
University Press.

Green, Donald and Ian Shapiro. 1994. Pathologies of Rational Choice Theory. New
Haven: Yale University Press.

Greene, William. 1993. Econometric Analysis. 3rd ed. New York: Macmillan.

Grenville, J AS. 1974. The Major International Treaties 1914-1973. London:
Methuen.

Grenville. J AS and Bernard Wasserstein. 1987. The Major International Treaties
since 1945. New York: Methuen.

Grieco. Joseph. 1988. "Anarchy and the Limits of Cooperation: a Realist Critique of the
newest Liberal Institutionalism." International Organization 42 (3). 485-507.

Gurr, Ted Robert. 1974. “Persistence and Change in Political Systems, 1800-1971.”
American Political Science Review 68 (4): 1482-1504.

Gurr. Ted Robert, Keith Jaggers, and Will Moore. 1997. Polity II Codebook. Boulder:
Center for Comparative Politics, University of Colorado.

Hagan, Joe. 1993. Political Opposition and Foreign Policy in Comparative
Perspective. Boulder, CO: Lynne Rienner.

Hanrieder, Wolfgang. 1989. Germany, America, EurOpe. New Haven: Yale University
Press.

277

Hanushek, Eric and John Jackson. 1977. Statistics for Social Sciences. New York:
Academic Press.

Harkavy, Robert E. 1975. The Arms Trade and International Systems. Cambridge:
Ballinger.

---. 1982. Great Power Competition for Overseas Bases: The Geopolitics of A ccess
Diplomacy. New York: Pergamon Press.

--- 1989. Bases Abroad: The Global Foreign Military Presence. New York: Oxford
University Press.

Harsanyi. John. 1967-8. “Games with Complete Information Played by Bayesian
Players.” Parts 1-3. Management Science 14: 159-82, 320-334, 486-502.

Harvey, Frank. 1998. Rigor Mortis or Rigor, More Tests: Necessity, Sufficiency, and
Deterrence Logic.” International Studies Quarterly 42 (4): 675-707.

Heckman, James. 1979. “Sample Selection Bias as a Specification Error.” Econometrica
47(2): 153-62.

Holsti, Kalevi. 1992. Peace and War: Armed Conflicts and International Order, 1648-
1989. Cambridge: Cambridge University Press.

Holsti, Ole R., P. Terrance Hopmann, and John D. Sullivan. 1973. Unity and
Disintegration in International Alliances. New York: John Wiley & Sons.

Hopmann, P. Terrance. 1998. “Complexity and Uncertainty in lntemational Systems
(book review of System Eflects, by Robert Jervis). ” Mershon International Studies
Review 42 (2): 313-6.

Hopmann, P. Terrance, Dina Zinnes, and J. David Singer, Eds. 1981. C umulation in
International Relations Research. Denver: University of Denver.

Horn, Robert C. 1987. Alliance Politics Between Comrades: The Dynamics of Soviet-
Vietnamese Relations. Los Angeles: Rand/UCLAC SSIB.

Hurst, Michael. 1972. Key Treaties for the Great Powers, 1814-1914. New York:
Newton Abbott, David and Charles.

Huth, Paul. 1988. Extended Deterrence and the Prevention of War. New Haven: Yale
University Press.

---. 1990. “The Extended Deterrence Value of Nuclear Weapons.” Journal of Conflict
Resolution 34 (2): 270-90.

278

---. 1994. “When do States Take on Extended Deterrence Commitments?” in Francis
Wayman and Paul Diehl, eds. Reconstructing Realpolitik. Ann Arbor: University of
Michigan Press.

---. 1996. Standing Your Ground. Ann Arbor: University of Michigan Press.

---. 1997. “Reputations and Deterrence.” Security Studies 7 (1): 72-99.

---. 1998. “Major Power Intervention in lntemational Crises, 1918-1988.” Journal of
Conflict Resolution 42 (6): 744-70.

Huth, Paul and Bruce Russett. 1984. “What Makes Deterrence Work?” World Politics,
36 (3): 496-526.

---. 1988. "Deterrence Failure and Crisis Escalation", International Studies Quarterly 32
(1): 29-45.

---. 1990. “Testing Deterrence Theory: Rigor Makes a Difference.” World Politics, 42
(3): 466-501.

---. 1993. “General Deterrence between Enduring Rivals: Testing three Competitive
Models.” American Political Science Review 87 (1): 61-73.

Hyland, William. 1987. Mortal Rivals. New York: Random House.

Iida, Keisuke. 1993. "When and How do Domestic Constraints Matter?" Journal of
Conflict Resolution 37 (3): 403-26.

Institute for lntemational Strategic Studies. 1961-1984. The Military Balance. London:
Institute for lntemational Strategic Studies.

lntemational Monetary Fund. 1950-1984. Direction of Trade. New York: lntemational
Monetary Fund.

Jaggers, Keith and Ted Gurr. 1995. “Tracking Democracy’s Third Wave with the Polity
III data.” Journal of Peace Research 32 (4): 469-83.

Jelavich, Barbara. 1973. The Ottoman Empire, the Great Powers, and the Straits
Question 1870-1887. Bloomington: Indiana University Press.

---. 1991. Russia ’s Balkan Entanglements. New York: Cambridge University Press.

Jervis, Robert. 1970. The Logic of Images in International Relations. Princeton:
Princeton University Press.

279

---. 1976. Perception and Misperception in International Politics. Princeton: Princeton
University Press.

---. 1988. “War and Misperception.” In Robert Rotberg and Theodore Rabb, eds. The
Origin and Prevention of Major Wars. New York: Cambridge University Press.

---. 1997. System Effects. Princeton: Princeton University Press.

Jervis. Robert, Richard Ned Lebow, and Janice Gross Stein. eds. 1985. Psychology and
Deterrence. Baltimore: Johns Hopkins University Press.

1011, James. 1992. The Origins of the First World War. Second edition. Oxford: Oxford
University Press.

Jones, Daniel, Stuart Bremer and J. David Singer. 1996. “Militarized Interstate
Disputes, 1816—1992: Rationale, Coding Rules, and Empirical Patterns.” Conflict
Management and Peace Science 15 (2): 163-213.

Kahn, Hermann. 1965. On Escalation: Metaphors and Scenarios. Baltimore: Penguin
Books.

Kahneman, David and Amos Tversky. 1979. Prospect Theory: An Analysis of Decision
under Risk.” Econometrica 47 (2): 263-91.

Kann, Robert A. 1974. A History of A ustria-Hungary. Berkeley: University of
California Press.

Kegley, Charles and Gregory Raymond. 1990. When Trust Breaks Down. Columbia:
University of South Carolina Press.

Keohane, Robert. 1971. “The Big Influence of Small Allies.” Foreign Policy 1 (2):
161-82.

Khong, Yuen Foong. 1996. “Confronting Hitler and Its Consequences”, in Tetlock,
Philip and Aaron Belkin, Eds. Counterfactual Thought Experiments in World Politics.
Princeton: Princeton University Press.

Kilgour, D. Marc and Frank Zagare. 1991. “Credibility, Uncertainty, and Deterrence.”
American Journal of Political Science, 35 (3): 303-34.

---. 1993. “Asymmetric Deterrence.” International Studies Quarterly 37(1): 1-27.

---. 1994. “Uncertainty and the Role of the Pawn in Extended Deterrence.” Synthese,
100(3): 379-412.

280

Kim. Woosang and Bruce Bueno de Mesquita. 1995. “How Perceptions Influence the
Risk of War.” International Studies Quarterly 39(1): 51-66.

King, Gary. 1989. Unifying Political Methodology: The Likelihood Theory ofStatistical
Inference. Cambridge: Cambridge University Press.

Kinsella, David and Herbert Tillema. 1995. “Arms and Aggression in the Middle East:
Overt Military Interventions, 1948-1991.” Journal of Conflict Resolution 39 (2): 306-29.

Kolodziej, Edward. 1989. Making and Marketing Arms: the French Experience and its
Implications for the International System. Princeton: Princeton University Press.

Krause. Keith. 1992. Arms and the State: Patterns of Military Production and Trade.
Cambridge: Cambridge University Press.

Kreps, David. 1991. A Course in Microeconomic Theory. Princeton: Princeton
University Press.

Kugler, Jacek. 1984. “Terror without Deterrence.” Journal of C onflict Resolution 28
(4): 470-506.

Kugler, Jacek and Douglas Lemke, eds. 1996. Parity and War: Critical Reexaminations
of the War Ledger. Ann Arbor: University of Michigan Press.

Kuhn, Thomas. 1962. The Structure of Scientific Revolutions. Chicago: University of
Chicago Press.

Lakatos, Imre. 1970. "The Methodology of Scientific Research Programmes", in
Lakatos and Alan Musgrave, eds. Criticism and the Growth of Knowledge. Cambridge:

Cambridge University Press.

Lalman, David and David Newman. 1991. “Alliance Formation and National Security.”
International Interactions 16 (2): 229-253.

Lalman, David, Joe Oppenheimer and Piotr Swistak. 1993. “Formal Rational Choice
Theory: A Cumulative Science of Politics. In Ada F inifter, ed Political Science: The
State of the Discipline 11. Washington, DC: American Political Science Association.

Langer, William L. 1950. European Alliances and Alignments 1871-1890. New York:
Alfred A. Knopf.

---. 1972. An Encyclopedia of World History, 5'” ed. Boston: Houghton Mifflin
Company

Laurance, Edward J. 1992. The International Arms Trade. New York: Lexington Books.

281

Lebow, Richard Ned. 1981. Between Peace and War: The Nature of International Crisis.
Baltimore: Johns Hopkins University Press.

Lebow, Richard Ned and Janice Gross Stein. 1987. “Beyond Deterrence. " Journal of
Social Issues 43 (1): 5-71.

---. 1989. “Rational Deterrence Theory: I Think, Therefore I Deter.” World Politics 41
(2): 208-24.

---. 1990. “Deterrence: The Elusive Dependent Variable.” World Politics 42 (4): 336-
69.

Leng, Russell. 1984. “Reagan and the Russians: Crisis Bargaining Beliefs and the
Historical Record.” American Political Science Review 78 (2): 338-55.

---. 1993. Interstate Crisis Behavior, 1816-1980. Realism versus Reciprocity.
Cambridge: Cambridge University Press.

Leng, Russell and Hugh Wheeler. 1979. “Influence Strategies, Success, and War.”
Journal of Conflict Resolution 23 (4): 657-63.

Levy, Jack. 1988. “When do Deterrent Threats Work?” British Journal of Political
Science 18 (4): 485-512.

~--. 1989. “Quantitative Studies of Deterrence.” In Stern, Paul C., Robert Axelrod.
Robert Jervis, and Roy Radner, eds. 1989. Perspectives on Deterrence. 1989. New
York: Oxford University Press. ,

---. 1997. “Prospect Theory and The Rational-Cognitive Debate.” In Geva, Nehemia
and Alex Mintz, eds. 1997. Decision-making on War and Peace: The C ognitive-
Rational Debate. Boulder: Lynne Rienner.

Lijphart, Arend. 1971. “Comparative Politics and Comparative Method. " American
Political Science Review 65 (3): 682-93.

Liska, George, Nations in Alliance.“ The Limits of Interdependence, 1962. Baltimore:
Johns Hopkins University Press.

Lowi, Theodore. 1992. “The State in Political Science: How We become what We
Study.” American Political Science Review 86 (l ): 1-7.

Mackie, Thomas and Richard Rose. 1974. The International Almanac of Electoral
History. New York: Macmillan.

282

Mansfield, Edward and Rachel Bronson. 1997. “Alliances, Preferential Trading
Arrangements, and International Trade.” American Political Science Review 91(1): 94-
107.

Marlowe, John. 1965. Anglo-Egyptian Relations 1800-1956. London: Frank Cass.

May, Ernest R. ed. 1984. Knowing One 's Enemies: Intelligence Assessment Before The
Two World Wars. Princeton: Princeton University Press.

McGregor, Charles. 1988. The Sino- Vietnamese Relationship and the Soviet Union.
London: lntemational Institute for Strategic Studies.

McKelvey, Richard and Thomas Palfrey. 1998. “Quantal Response Equilibria for
Extensive Form Games.” Experimental Economics 1(1): 9-41.

McLaughlin, Sara, Scott Gates, Havard Hegre, Ranveig Gissinger, and Nils Petter

Gleditsch. 1998. “Timing the Changes in Political Structures: A New Polity Database.”
Journal of Conflict Resolution 42 (2): 231-42.

Mearsheimer, John J. 1983. Conventional Deterrence. Ithaca: Cornell University Press.

Meng, Chun-Lo and Peter Schmidt. 1986. “On the Costs of Partial Observability in the
Bivariate Probit Model.” International Economic Review 26 (1): 71-86.

Mercer, Jonathan. 1996. Reputation and International Politics. Ithaca: Cornell University
Press.

---. 1997. “Reputation and Rational Deterrence Theory.” Security Studies, 7(1): 100-
13.

Midlarsky, Manus. 1988. The Onset of World War. Boston: Unwin Hyman.
---. I989. “Hierarchical Equilibria and the Long-run Stability of Multipolar Systems.”
In. Manus Midlarsky, ed. Handbook of War Studies. Ann Arbor: University of Michigan

Press.

Mitchell, Brian. 1975. European Historical Statistics. 1750-1970. New York: Columbia
University Press.

Mo, Jongryn. 1994. "Two level games with Endogenous Domestic Coalitions." Journal
of Conflict Resolution 38 (3): 402-23.

Moe, Terry. 1979. “On the Scientific Status of Rational Models.” American Journal of
Political Science 23 (1): 215-43.

283

Monroe, Kristin Renwick. 1996. The Heart of Altruism: Perceptions ofa Common
Humanity. Princeton: Princeton University Press.

Morgan, Patrick M. 1983. Deterrence: A Conceptual Analysis, 2"“1 ed. Beverly Hills:
Sage Publications.

Morgan, T. Clifton. 1994. Untying the Knot of War. Ann Arbor: University of
Michigan Press.

Morgan, T. Clifton and Glenn Palmer. 1997. “Room to Move: Security, Proaction and
Institutions in Foreign Policy Decision Making.” In Randolph Siverson, ed. Strategic
Politicians, Institutions, and Foreign Policy. Ann Arbor: University of Michigan Press.

Morgenthau, Hans J. 1985. Politics Among Nations: the Struggle for Power and Peace.
6th ed. New York: Knopf.

Morrow, James. 1987. “On the Theoretical Basis of a Measure of National Risk
Attitudes.” International Studies Quarterly 31 (4): 423-38.

---. 1988. “Social Choice and System Structure in World Politics.” World Politics 41
(1): 75-97.

---. 1989. “Capabilities, Uncertainty, and Resolve: A Limited Information Model of
Crisis Bargaining.” American Journal of Political Science 74 (3): 941-72.

--—. 1991. "Alliances and Asymmetry: An Alternative to the Capability Aggregation
Model of Alliances." American Journal of Political Science 35 (4): 904-33.

---. 1992. “Signaling Difficulties with Linkage in Crisis Bargaining.” International
Studies Quarterly 36 (2): 153-72.

---. 1993. "Arms vs. Allies: Trade-Offs in the Search for Security. " International
Organization 47 (2): 207-33.

---. I994a. “Alliances, Credibility, and Peacetime Costs.” Journal of Conflict Resolution
38 (2): 270-97.

---. 1994b. Game Theory for Political Scientists. Princeton: Princeton University Press.
---. 1997. “A Rational Choice Approach to lntemational Conflict.” In Geva, Nehemia
and Alex Mintz, eds. 1997. Decision-making on War and Peace: The Cognitive-

Rational Debate. Boulder: Lynne Rienner.

Most, Benjamin and Harvey Starr. 1989. Inquiry, Logic and International Politics.
Columbia: University of South Carolina Press.

284

Mueller, John. 1973. War, Presidents, and Public Opinion. New York: Wiley.

Nalebuff, Barry. 1987. “Credible Pretrial Negotiation.” Rand Journal of Economics. 18
(2): 198-209.

Nicholson, Michael. 1992. Rationality and the Analysis of International Conflict. New
York: Cambridge University Press. '

Nincic, Miroslav. 1992. Democracy and Foreign Policy. The Fallacy of Political
Realism. New York: Columbia University Press.

Norpoth, Helmut. 1987. “Guns, Butter and Government Popularity in Britain.”
American Political Science Review 81 (3): 949-59.

Olson, Mancur and Richard Zeckhauser. 1966. “An Economic Theory of Alliances.”
Review of Economics and Statistics 48 (2): 266-79.

O’Neill, Barry. 1995. “Weak Models, Nil Hypotheses, and Decorative Statistics.”
Journal of Conflict Resolution 39 (4): 731-47.

Oren, Ido. 1990. “The War Proneness of Alliances.” Journal of Conflict Resolution 34
(2): 208-33.

Organski, A.F.K. and Jacek Kugler. 1980. The War Ledger. Chicago; University of
Chicago Press.

Osborne, Martin and Ariel Rubinstein. 1994. A Course in Game Theory. Cambridge:
MIT Press.

Osgood, Robert E. 1962. NA T O: The Entangling Alliance. Chicago: University of
Chicago Press.

---. 1968. Alliances and the American Foreign Policy. Baltimore: Johns Hopkins
University Press.

Ostrom, Charles and Francis Hoole. 1978. “Alliances and War Revisited. "
International Studies Quarterly 22 (2): 215-36.

Ostrom, Charles and Brian Job. 1986. "The President and the Political Use of Force."
American Political Science Review 80 (2): 541-66.

Papayounou, Paul. 1997. “Intra-alliance Bargaining and US. Bosnia Policy.” Journal
of Conflict Resolution 41 (1): 91-116.

Paul. T.V. 1994. Asymmetric Conflicts: War Initiation by Weaker Powers. New York:
Cambridge University Press.

285

Posen, Barry. 1984. The Sources of Military Doctrine. Ithaca: Cornell University Press.

---. 1984-85. “Measuring the EurOpean Conventional Balance: Coping with Complexity
in Threat Assessment.” International Security 9 (3): 47-88.

Powell, Robert. 1985. “The Strategic Foundations of Nuclear Deterrence Theory.”
Political Science Quarterly 100 (I): 75-96.

---. 1987. “Crisis Bargaining, Escalation. and MAD.” American Political Science
Review 81 (3): 717-35.

---. 1990. Nuclear Deterrence Theory: The Search for Credibility. Cambridge,
Cambridge University Press.

---. 1991. "Absolute and Relative Gains in International Relations Theory", American
Political Science Review, 85 (4):, 1303-20.

---. 1993. “Guns, Butter and Anarchy.” American Political Science Review 87 (1): 115-
32.

Pribham, E.F. I921. The Secret Treaties of A ustria-Hungary, 1879-1914. New York:
Howard F ertig.

Prins. Brandon and Christopher Sprecher. 1999. ‘Institutional Constraints, Political
Opposition, and Interstate Dispute Escalation: Evidence from Parliamentary Systems,
1946-89.” Journal of Peace Research 36 (3): 271-87.

Pundik, Ron. I994. The Struggle for Sovereignty: Relations Between Great Britain and
Jordan 1946-1951. Oxford: Blackwell.

Putnam, Robert. 1993. "Diplomacy and Domestic Politics: The Logic of Two Level
Games", in Peter B. Evans, Harold K. Jacobson. and Robert D. Putnam, eds. Double-
Edged Diplomacy. Berkeley: University of California Press.

Quinones, Sherry and Scott Gates. 1995. “Economic Risk and the Politics of
Protectionism.” International Interactions 21 (1): 63-83.

Raknerud, Arvid and Havard Hegre. 1997. “The Hazard of War: Reassessing the
Evidence for the Democratic Peace.” Journal of Peace Research 34 (4): 385-404.

Rasmusen, Eric. 1989. Games and Information. Oxford: Blackwell.
Ray, James Lee. 1990. “Friends as Foes: lntemational Conflict and Wars between

Formal Allies.” In Charles Gochman and Alan Sabrosky eds. Prisoners of War? Nation
States in the Modern Era. Lexington: D.C. Heath.

286

 

---. 1995. Democracy and International Conflict. Columbia: University of South
Carolina Press.

Reed, William. 1999. “A Unified Model of Conflict Onset and Escalation: Capabilities.
Democracy, and Status Quo Evaluations”, American Journal of Political Science,
(forthcoming). '

Reiter, Dan. 1994. “Learning, Realism, and Alliances: The Weight of the Shadow of the
Past.” World Politics 46 (3): 490-526.

---. 1996. Crucible of Beliefs: Learning, Alliances, and World Wars. Ithaca: Cornell
University Press.

---. 1998. “The Emerging Significance of Military Strategy in the Study of lntemational
Conflict.” Paper presented at the annual meeting of the Peace Science Society
(lntemational), East Brunswick, New Jersey, October, 1998.

Rosenau, James N, ed. 1976. In Search of Global Patterns. New York: The Free Press.

Rothstein, Robert L. 1968. Alliances and Small Powers. New York: Columbia
University Press.

Russett, Bruce. 1963. “The Calculus of Deterrence.” Journal of Conflict Resolution 7
(1): 97-109.

Sabrosky, Alan. 1980. "Interstate Alliances. Their Reliability and the Expansion of
War". In J. David Singer, ed. The Correlates of War 11: Testing some Realpolitik Models.
New York: Free Press.

Sagan, Scott and Kenneth Waltz. 1995. The Spread of Nuclear Weapons: A Debate.
New York: Norton.

Sandler, Todd and Keith Hartley. 1995. The Economics of Defense. New York:
Cambridge University Press.

Schelling, Thomas. 1960. The Strategy of Conflict. Cambridge: Harvard University Press.
---. 1966. Arms and Influence. 1966. New Haven: Yale University Press.

Schrodt, Philip. I982. Preserving Arms Distributions in a Multi-polar World: a
Mathematical Study. Denver: University of Denver Press.

---. 1983. “Arms Transfers and lntemational Behavior in the Arabian Sea Area.”
International Interactions 10 (2): 101-27.

287

Schroeder, Paul. 1976. “Alliances, 1815-1945; Weapons of Power and Tools of
Management”. In Klaus Knorr, ed. Historical Dimensions of National Security
Problems. Lawrence: University Press of Kansas.

Schultz, Kenneth. 1998. “Domestic Opposition and Signaling in lntemational Crises.”
American Political Science Review 92 (4). 829-44.

Schweller, Randall. 1997. “New Realist Research on Alliances: Refining, not Refuting.
Waltz’s Balancing Proposition.” American Political Science Review, 91 (4): 927-30.

---. 1998. Deadly Imbalances: Hitler’s Strateg of World Conquest. New York:
Columbia University Press.

Sen, Amatrya. 1970. Collective Choice and Social Welfare. New York: North Holland.

Seton-Watson, Hugh. 1967. Eastern Europe Between The Wars [918-1941. 3'ded.
New York: Harper Torchbooks.

Shimshoni. Jonathan. 1988. Israel and Conventional Deterrence: Border Warfarefi'om
1953-1970. Ithaca: Cornell University Press.

Signorino, Curtis. 1999. “Strategic Interaction and the Statistical Analysis of
lntemational Conflict.” American Political Science Review 93 (2): 279-98.

Signorino, Curtis and Jeffrey Ritter. 1999. “Tau-b or not Tau-b: Measuring the Similarity
of Foreign Policy Positions.” International Studies Quarterly 43 (1): 115-44.

Simon, Michael and Erik Gartzke. 1996. “Do Democracies Flock Together, or do
Opposites Attract?” Journal of Conflict Resolution 40 (4): 617-35.

Singer. J. David. 1961. “The Level of Analysis Problem in lntemational Relations.”
World Politics 14 (1): 77-92.

Singer, J. David and Melvin Small. 1966a. "National Alliance Commitments and War
Involvement, 1815-1945", Peace Research Society (International) Papers 5: 109-140.

---.. 1966b. "Formal Alliances, 1815 —1939.” Journal of Peace Research 1 (1): 1-32.

---. 1968. “Alliance Aggregation and the Onset of War, 1815-1945.” In J. David Singer,
ed. Quantitative International Politics. New York: Free Press.

Singer, J. David, Stuart Bremer, and John Stuckey. 1972. “Capability Distribution,

Uncertainty and Major Power War, 1820-1965.” In Peace, War and Numbers, Bruce
Russett. ed. Beverly Hills: Sage.

288

Sislin, John. 1994. “Arms as Influence: The Determinants of Successful Influence.”
Journal of Conflict Resolution 38 (4): 665-89.

Siverson, Randolph, ed. '1997. Strategic Politicians, Institutions, and Foreign Policy.
Ann Arbor: University of Michigan Press.

Siverson, Randolph and Joel King. 1979. “Alliances and the Expansion of War.” In J.
David Singer, ed. T o Augur Well: Early Warning Indicators in World Politics. Beverly
Hills: Sage. ‘

---. 1980. “Attributes of National Alliance Membership and War Participation, 1815-
1965.” American Journal of Political Science 24 (1): 1-15.

Siverson, Randolph and Julie Emmons. 1991. “Birds of a Feather.” Journal of Conflict
Resolution 35 (3): 285-306.

Siverson, Randolph and Harvey Starr. 1991. The Diflusion of War. Ann Arbor:
University of Michigan Press.

Slusser, Robert. 1973. The Berlin Crisis of 1961 . Baltimore: Johns Hopkins University
Press.

Small, Melvin and J. David Singer. 1969. “Formal Alliances: An Extension of the Basic
Data.” Journal of Peace Research 6 (3): 257-82.

---. 1982. Resort to Arms: International and Civil Wars, [816-1980. Beverly Hills:
Sage. .

Smith, Alastair. 1995a. A Theory of Alliances. Ph.D. Dissertation, University of
Rochester.

---. 1995b. “Alliance Formation and War.” International Studies Quarterly 39 (4): 405-
25.

---. 1996. “To Intervene or not to Intervene; A Biased Decision.” Journal of Conflict
Resolution 40 (1): 16-40.

---. 1997. “The Effect of Foreign Policy Statements on Foreign Nations and Domestic
Electorates.” In Randolph Siverson, ed. Strategic Politicians, Institutions, and Foreign

Policy. Ann Arbor: University of Michigan Press.

---. 1998. “International Crises and Domestic Politics. ” American Political Science
Review 92 (3): 623-38.

Snidal, Duncan. 1986. “The Game Theory of World Politics.” World Politics 38 (1): 25-
57.

289

---. 1991. "Relative Gains and the Pattern of lntemational Cooperation." American Political
Science Review 85 (3): 701-26.

Spence, Michael. 1973. “Job Market Signaling.” Quarterly Journal of Economics 87
(3): 355-74.

Stam, Alan. 1996. Win, Lose, or Draw. Ann Arbor: University of Michigan Press.

Statesman’s Yearbook. 1886-193 9. Statistical and Historical Annual of the State of the
World. London: Macmillan.

Stockholm lntemational Peace Research Institute (SIPRI). 1975. Arms Trade Registers:
The Arms Trade with the Third World. Stockholm: Almqvist and Wiksell.

---. 1961-1984. World Armaments and Disarmaments. Oxford: Oxford University Press.

Snyder, Glen H. 1961. Deterrence and Defense: Toward A Theory of National Security.
Princeton: Princeton University Press.

---. 1984. "The Security Dilemma in Alliance Politics." World Politics 36 (4): 461-95.
---. 1991. “Alliances, Balance, and Stability.” lntemational Organization 45 (1): 121-42.
---. 1997. Alliance Politics. Ithaca: Cornell University Press.

Snyder, Glenn and Paul Diesing. 1977. Conflict Among Nations. Princeton: Princeton
University Press.

Sorokin, Gerald. 1994. “Alliance Formation and General Deterrence: A Game-Theoretic
Model and the Case of Israel.” Journal of Conflict Resolution 38 (2): 298-325

Sprout. Harold and Margaret Sprout. 1956. Man-Milieu Relationships in the Context of
Politics. Princeton: Center of lntemational Affair, Princeton University.

---. 1965. The Ecological Perspective on Human Affairs. Princeton: Princeton
University Press.

---. 1968. An Ecological Paradigm for the Study of Politics. Princeton: Center of
International Affairs, Princeton University.

Starr, Harvey. 1978. "Opportunity and Willingness as Ordering Concepts in the Study of
War." International Interactions 4 (4): 363-87.

Stern, Paul C., Robert Axelrod, Robert Jervis, and Roy Radner, eds.. 1989. Perspectives
on Deterrence. 1989. New York: Oxford University Press.

290

Tang, Peter S. H. 1951 Russian and Soviet Policy in Manchuria and Outer Mongolia
1911-1931. Durham: Duke University Press.

Taylor, A.J.P. 1954. The Struggle for Mastery in Europe 1848-1918. Oxford: Oxford
University Press.

Tetlock, Philip and Aaron Belkin. Eds. 1996. Counterfactual Thought Experiments in
World Politics. Princeton: Princeton University Press.

Thompson, William R. 1997. “Anglo-German Rivalry and the 1939 Failure of
Deterrence.” Security Studies, 7 (2): 58-89.

Tillema, Herbert. 1991. International Armed Conflict since 1945: A Bibliographic
Handbook of Wars and Interventions. Boulder: Westview.

United Nations. 1950-1984. Yearbook of International Trade Statistics. New York:
United Nations.

United States Arms Control and Disarmament Agency. 1975-1984. World Military
Expenditures and Arms Transfers 1963-1990. Washington, DC: Government Printing
Office.

Vasquez, John. 1987. "The Steps to War: Toward a Scientific Explanation of Correlates
of War Findings." World Politics 40 (I): 108-145.

---. 1993. The War Puzzle. Cambridge: Cambridge University Press.

---. 1997. “The Realist Paradigm and Degenerative versus Progressive Research
Programs: An Appraisal of Neotraditional Research on Waltz’s Balancing Propositions.”
American Political Science Review 91 (4): 899-912.

---. 1998. The Power of Power Politics. Cambridge: Cambridge University Press.

Walt, Stephen. 1987. The Origins of Alliances. Ithaca: Cornell University Press.

---. 1991. “The Renaissance of Security Studies.” International Studies Quarterly 35

(2): 211-39.

---. 1997. “The Progressive Power of Realism.” American Political Science Review 91
(4): 931-6.

Waltz, Kenneth N. 1959. Man, the State and War: A Theoretical Analysis. New York:
Columbia University Press.

---. 1964. “The Stability of a Bipolar World.” Daedalus 93 (4): 881-909.

---. 1979. Theory of International Politics. New York: Random House.

291

---. 1997. “Evaluating Theories.” American Political Science Review 91 (4): 913-18.

Ward, Michael Don. 1982. Research Gaps in Alliance Dynamics. Denver: University of
Denver Press.

Wayman, Frank. 1990. "Alliances and War: A Time Series Analysis". in Charles
Gochman and Alan Sabrosky eds. Prisoners of War? Nation States in the Modern Era.
Lexington: D.C. Heath.

Wendt, Alexander. 1987. “The Agent-Structure Problem in lntemational Relations.”
International Organization 17 (3): 335-70.

Whiting, Allen. 1975. The Chinese Calculus of Deterrence. Ann Arbor: University of
Michigan Press.

Wolfe, Thomas W. 1970. Soviet Power and Europe 1 945-1 9 70. Baltimore: Johns
Hopkins University Press.

Wu, Samuel. 1990. “To Attack or not to Attack: a Theory and Empirical Assessment
of Extended Immediate Deterrence.” Journal of Conflict Resolution 34(4): 531-52.

Zagare, Frank C. 1987. The Dynamics of Deterrence. Chicago: University of Chicago
Press.

---. 1990. “Rationality and Deterrence.” World Politics 42 (3): 23 8-60.

Zinnes, Dina. 1976. The Problem of Cumulation”. In James Rosenau, ed. In Search of
Global Patterns. New York: Free Press.

---. 1980. "Three Puzzles in Search of a Researcher", International Studies Quarterly 24
(3): 315-42.

---. 1991. “Musings on Mathematical Modeling.” Conflict Management and Peace
Science 11 (2): 1-17.

Zinnes, Dina, Harvey Starr, Barbara Hill and Stephen Portnoy. 1985. "From War to
War: Stochastic Models of Wartime Alignment Patterns." In Michael D. Ward ed.
Theories, Models, and Simulations in International Relations. Boulder: Westview Press.

Zinnes, Dina and Robert Muncaster. 1997. “Prospect Theory versus Expected Utility
Theory: A Dispute Sequence Appraisal.” In Geva, Nehemia and Alex Mintz, eds. 1997.
Decision-making on War and Peace: The Cognitive-Rational Debate. Boulder: Lynne
Rienner.

292

 

NICHIGQN STQTE UNIV LIBRQRI

II III IIII II II II III II