Deterrence and Foreign Policy

Author(s): Alexander L. George and Richard Smoke

Source: World Politics , Jan., 1989, Vol. 41, No. 2 (Jan., 1989), pp. 170-182
Published by: Cambridge University Press

Stable URL: https://www.jstor.org/stable/2010406

REFERENCES
Linked references are available on JSTOR for this article:
https://www.jstor.org/stable/2010406?seq=1&cid=pdf-
reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

Cambridge University Press is collaborating with JSTOR to digitize, preserve and extend access
to World Politics

This content downloaded from

200.52.255.133 on Tue, 27 Jun 2023 01:57:21 +00:00
All use subject to https://about.jstor.org/terms
 DETERRENCE AND FOREIGN
POLICY

By ALEXANDER L. GEORGE and RICHARD SMOKE*

THE article by Christopher Achen and Duncan Snidal in this is-

sue, however provocative some of its observations, has the merit of
raising an important question: How should we think about the relation-
ship between deductive theories of deterrence on the one hand and, on
the other, research on deterrence that is at the same time empirical and
oriented toward theory? While forcefully defending the deductive form
of theory, Achen and Snidal also clearly dissociate themselves from the
extreme position that it is a self-contained enterprise that need not take
empirical research seriously. Rather, they recognize that empirical case
studies are capable of contributing, and, indeed, have to some extent al-
ready contributed, to the development of theory, perhaps even to the kind
of deductive theory they favor.'
Their general position on the need for both approaches- although not
the superiority and priority they attach to deductive deterrence theory-
is entirely compatible with our own long-standing position. The question
to be debated is not the superiority of one approach over the other but
rather how to improve both approaches and develop fruitful interaction
between them.2 On this score, the Achen-Snidal article is disappointingly

* For helpful comments on an earlier draft the authors express appreciation to Kenneth A.
Oye, Jack S. Levy, and Robert Jervis who, however, bear no responsibility for the thoughts
expressed in this comment.
I Thus: "in the hands of George and Smoke, the case study approach helps generate theory
in a very direct way" (p. 156); "we emphatically believe that they [case studies] are essential to
the development and testing of social science theory" (p. i67); "indeed, analytic theory cannot
do without case studies" (p. i69; see also p. 159). Achen and Snidal, "Rational J)eterrence
Theory and Comparative Case Studies," World Politics 41 (January i989), 143-69.
2Alexander L. George and Richard Smoke, Deterrence in American Foreign Policy: Theory
and Practice (New York: Columbia University Press, 1974). Sincc publishing this book, wc
have worked steadily to raise standards and explicate procedures for improving the quality of
case studies and their contribution to theory development. We have given serious attention to
most of the concerns about them raised by Achen and Snidal. See A. L. George, "Case Studies
and Theory Development: The Method of Structured, Focused Comparison," in Paul Gor-
don Lauren, ed., Diplomacy: New Approaches in History, Theory and Policy (New York: Free
Press, I979); A. L. George, "Case Studies and Theory Development," paper presented to the
Second Annual Symposium on Information Processing in Organizations, Carnegie-Mellon
University, Pittsburgh, PA, October I5-i6, i982; A. L. George and T. J. McKeown, "Case
Studies and Theories of Organizational Decision Making," in Robert Coulam and Richard
Smith, eds., Advances in Information Processing in Organizations, Vol. 2 (Greenwich, CT: JAI

This content downloaded from

200.52.255.133 on Tue, 27 Jun 2023 01:57:21 +00:00
All use subject to https://about.jstor.org/terms
 RATIONAL DETERRENCE: GEORGE & SMOKE 171

meager in its suggestions, in part because it concentrates on real or alleged

limitations of case-study research and has little to say about how the de-
ductive theory of deterrence, which they admit is deficient, might be im-
proved.
In reporting the results of our empirical research on deterrence fifteen
years ago, we called attention to the limitations of the deductive, abstract
version of the theory then extant; we did not reject the approach itself. In
fact, as Achen and Snidal come close to recognizing (fn. 41, and
even more so in an earlier version of their paper), the reformulation of
deterrence theory we advanced in Chapter 17 of our book presented a
number of more refined propositions that suggested how deductive the-
ory might be improved. This was, to be sure, incidental to our major ob-
jective, which was to demonstrate by means of a relatively new research
strategy ("structured, focused comparisons) how a relatively small num-
ber of historical case studies could contribute to developing a more dif-
ferentiated theory than deductive theory offered, one composed of con-
ditional generalizations (as exemplified in Chapter i8) that had more
specific relevance for policy making.4 In this research strategy, individual
cases are used as building blocks to create a cumulative development of
typological explanatory theory. Since the cases selected for study in this
approach do not constitute a representative sample of the entire universe
(a goal that is probably infeasible in any case), typological theory does not
provide a basis for estimating the frequency distribution of different out-
comes and, contrary to Achen and Snidal (p. i6o), we have explicitly de-
nied any such claim for it.5

Prcss, 1985), 21-58. Case-study methodology is discussed also in Richard Smoke, -War: Con-
trolling Escalation (Cambridge: Harvard University Press, I977), chap. 3 and app. B. George
continues efforts to improve case-study methodology and plans additional publications.
A discussion of the uses and limitations of the controlled comparison method of studying
a small number of cases for theory development should not overlook the importance of single-
case studies. Lack of space prevents us from summarizing the arguments on behalf of the
contributions a single case study can make to theory development that have been advanced by
such writers as Harry Eckstein, "Case Study and Theory in Political Science," in F. I. Green-
stein and N. W. Polsby, eds., Handbook of Political Science, Vol. 7 (Reading, MA: Addison-
Wesley, I975), 79-I38, and Lawrence B. Mohr, "The Reliability of the Case Study As a Source
of Information," in Coulam and Smith, op. cit., 65-97.
3 The description of this method in George and Smoke (fn. 2), 95-I03, has since been elab-
orated, and related methodological and theoretical issues more fully discussed, in the two ar-
ticles by George (fn. 2) and in the George-McKeown article (fn. 2).
4 The formulation of contingent generalizations is necessary in order to capture the fact
that deterrence is characterized by the phenomenon of what General Systems Theory refers
to as "equifinality" (and what John Stuart Mill referred to as "plurality of causes" when he
warned that his "method of agreement" and "method of difference" could not be easily em-
ployed for inferring causal relationships). Equifinality refers to the fact that similar outcomes
on a dependent variable (e.g., deterrence failures) occur as a result of different causal pro-
cesses, thus making the search for robust universal causal generalizations infeasible. Abstract
deductive deterrence theory has thus far ignored the phenomenon of equifinality.
5See George (fn. 2, I979), 59-60.

This content downloaded from

200.52.255.133 on Tue, 27 Jun 2023 01:57:21 +00:00
All use subject to https://about.jstor.org/terms
 172 WORLD POLITICS

In an additional departure from existing deterrence theory, we disag-

gregated it into several of its components, which we called "initiation the-
ory," "commitment theory," and "warning and response" theory (Chap-
ters 17, i9, and 20).6 We did so because we believe that an internally more
complex form of deterrence theory will be necessary in order to enhance
its policy usefulness. Thus, for example, we need to know more than ra-
tional deterrence theory tells us before we can better understand under
what conditions, why, and how a state dissatisfied with the status quo will
choose to challenge deterrence. Needed for this purpose is an "initiation
theory" that is much richer and more complex than that provided by the
deductive theory of deterrence.
In another departure from abstract deductive deterrence theory, we
found it necessary to reconceptualize the problem of deterrence for dif-
ferent levels of conflict. These are (i) the deterrent relationship of the two
superpowers' strategic forces; (2) the deterrence of local and limited wars;
and (3) the deterrence of nonmilitary challenges and "sublimited" conflict
at the lower level of the spectrum of violence. We noted that the first of
these three levels had received the greatest and most successful attention
in the formulation of deterrence theory. Largely because deterrence the-
ory at the strategic level, dealing as it does with a relatively simple struc-
tural situation, was so much better developed, theorists were tempted to
employ the logic of strategic deterrence as the paradigm case for thinking
about deterrence in general. This has proven to be quite unsatisfactory,
however, for there are major differences in the problem of applying de-
terrence strategy effectively at the second and third levels of conflict. De-
terrence at those levels is much more dependent upon context than at the
strategic level (although contextual variables also affect strategic deter-
rence witness the concern over conditions and events that can contrib-
ute to crisis instability and inadvertent or accidental war). Thus, situa-
tions of substrategic conflict, in which the policy question of whether and
how to employ deterrence arises, are often very fluid, ambiguous, and
subject to unexpected and substantial changes. The interests and moti-
vations (and hence the objectives) of one or both sides are often much
more complex and unstable than in the simpler, paradigmatic strategic
case.7 And, finally, it is also more difficult to determine the means that

6 This kind of disaggregated theory is an example of what Paul Diesing, drawing upon
Abraham Kaplan, has referred to as "concatenated theory" in his Patterns of Discovery in the
Social Sciences (Chicago: Aldine-Atherton, I971), 22-24.
7 Particular attention must be given not only to the strength of the Initiator's motivation to
challenge the status quo but also to the time he has in which to do so. Moreover, asymmetry
of motivation favoring the Initiator can sometimes compensate for asymmetry of power fa-
voring the Defender. On this point see, for example, A. L. George et al., The Limits of Coercive

This content downloaded from

200.52.255.133 on Tue, 27 Jun 2023 01:57:21 +00:00
All use subject to https://about.jstor.org/terms
 RATIONAL DETERRENCE: GEORGE & SMOKE 173

are likely to provide effective deterrence, i.e., th

tial damage considered necessary and adequate to convey a credible and
sufficiently potent threat.8 As a result, not only are the requirements for
deterrence often more complicated and elusive for those levels of conflict;
they are also more difficult both to identify reliably and to meet.9 The
problems of deterrence at different levels of conflict offer an additional
challenge to deductive deterrence theory that it has not yet addressed and
that are not recognized in the Achen-Snidal article.
We need to comment also on the inadequacy of rational deterrence
theory's current formulation of the utility framework for dealing with
the cost-benefit calculations of a state contemplating whether to challenge
deterrence. The theory's assumption that a would-be initiator decides
whether to "attack" or "not to attack" is oversimplified and inadequate
for prediction. As we noted in our book,Io the Initiator's choice is often
not so limited; rather, the Initiator often has multiple options at his dis-
posal for challenging a deterrence commitment, and his cost-benefit cal-
culations are geared to identifying an option that offers an opportunity
for gain while minimizing the risk of an unwanted response by the De-
fender. A more complex model of strategic interaction than Achen and
Snidal offer is needed to grasp the interplay between a Defender who em-
ploys deterrence strategy and an Initiator who is considering not merely
whether to challenge but how best do so at an acceptable cost-benefit level.
Employing such a model of strategic interaction enabled us to score some
cases as having mixed outcomes, i.e., the deterrence strategy employed
may have succeeded in dissuading the Initiator from choosing some risk-
ier options for challenging the status quo but it failed to dissuade the In-
itiator from employing "limited probes" or "controlled pressure" strate-
gies to bring about change. This conceptualization of the interaction
process also enabled us to identify several different causal patterns of de-
terrence failure (the phenomenon of equifinality we referred to in foot-

Diplomacy (Boston: Little, Brown, I971); Zeev Maoz, "Resolve, Capabilities, and the Out-
comes of Interstate Disputes, I8I6-I976," Journal of Conflict Resolution 27 (June I983), I95-
229; Jack S. Levy, "When Do Deterrent Threats Work?" British Journal of Political Science
i8 (October i988), 433-60.
8 Robert Wilson notes the difficulty that game-theoretic models encounter with contextual
variables, which render most situations indeterminate; "Deterrence in Oligopolistic Compe-
tition," in Paul C. Stern et al., eds., Perspectives on Deterrence (New York: Oxford University
Press, forthcoming).
9 Examples of difficulties encountered in attempts to apply deterrence strategy at substra-
tegic levels of conflict are provided in the analysis of the Eisenhower administration's Middle
East policies in George and Smoke (fn. 2), chap. i I. For a discussion of thirteen ways in which
the characteristics of deterrence at the strategic level differ from those at lower levels of con-
flict, see chap. 2.
Io Ibid., 520-22.

This content downloaded from

200.52.255.133 on Tue, 27 Jun 2023 01:57:21 +00:00
All use subject to https://about.jstor.org/terms
 174 WORLD POLITICS

note 4) which, as Achen and Snidal acknowledge (p. 156), their rational
deterrence theory has not considered and which they regard as of major
importance for theory as well as policy.ii
The preceding remarks have raised the problem relating to levels of
analysis in constructing deterrence theory and, even more, in utilizing it
in policy-making contexts. Achen and Snidal evidently prefer to hold fast
to a deductive theory, one that "Black Boxes" both the decision-making
and strategic-interaction levels of analysis. It is not easy for the reader to
grasp how they reconcile this preference with their recognition of the
value of findings from empirical research that attempt to study actual
events in these two Black Boxes rather than simply make assumptions
about decision-making and strategic-interaction processes. Left unan-
swered in their essay is whether and how they would integrate these em-
pirical findings into a new deductive deterrence theory or how they
would make the connection between the two types of theory. In other
words, how do they propose to deal with the deterrence phenomenon at
different levels of analysis?
This leads us to raise a number of questions regarding the nature of
the deductive rational deterrence theory espoused by Achen and Snidal
and to pose the critical question of how it might be improved and still
remain a deductive theory. Although the major impetus and thrust of
their article is to defend deductive deterrence theory, Achen and Snidal
clearly admit the absence of a satisfactory rational deterrence theory.
They acknowledge that coherent supporting arguments for such a theory
are lacking, and state that what they are really praising are its foreseeable
conclusions when it is properly developed.I2 Nonetheless, they vigorously
praise the contributions it has already made, while at the same time ar-
guing merely that rational deterrence theory should not be abandoned for
what they dismiss as "the patchwork of empirical findings currently
available" (p. 159). And, toward the end of the essay, they further dilute
their defense of this particular theory by concluding: "More importantly,
our goal here is to defend theory, not rational deterrence theory" (p. 159).
While the forthright admission that rational deterrence theory remains
undeveloped is praiseworthy, the authors should at least tell us what they
think an axiomatic-deductive theory must be like in order to be capable

Ibid., chap. i8.

See particularly fn. 24 in Achen and Snidal (fn. i): "Analysts continue to struggle pain-
fully for a fully satisfactory version of this game; as Harrison Wagner has remarked to us, 'the
rational theory of deterrence' doesn't exist." However, they go on to offer the confident as-
surance that "the principal conclusions of a legitimate theory of deterrence are foreseeable
even if the supporting arguments are at present incomplete; it is the former that we call 'ra-
tional deterrence theory.'

This content downloaded from

200.52.255.133 on Tue, 27 Jun 2023 01:57:21 +00:00
All use subject to https://about.jstor.org/terms
 RATIONAL DETERRENCE: GEORGE & SMOKE 175

of predicting the success or failure of deterrence in specific cases. That it

is not currently a full-fledged deductive theory is evident, but what is
missing? Have the requirements and standards for such theories been
clearly and fully identified and agreed upon by proponents of deductive
theory? Do Achen and Snidal adhere to the same conception of these re-
quirements as Harrison Wagner or Bruce Bueno de Mesquita; or are they
satisfied with the operationalization of deterrence provided by Huth and
Russett.'3
The task of developing a full-fledged deductive deterrence theory may
be much more demanding than Achen and Snidal anticipate. Even if the
internal logical structure of such a theory is eventually fully developed, it
remains to be seen how well it will perform in making predictions of de-
terrence outcomes in specific cases. Achen and Snidal offer no clear con-
ception of what kind of axiomatic-deductive theory would be capable of
making such predictions. To achieve this capability, rational deterrence
theory requires not merely internal logical consistency; it must also be op-
erationalized so that the presence or absence of the conditions assumed to
be necessary and/or sufficient for deterrence to be effective can be mea-
sured and established in each of the specific cases for which predictions
are attempted. Since the theory is not yet operationalized, it provides no
basis for predicting whether a specific deterrence threat will be suffi-
ciently credible and potent in any particular case. Lack of operationali-
zation leaves rational deterrence theory weakest precisely where it needs
to be strongest if it is to be able to make predictions in specific cases, i.e.,
in assessing the utility calculations of the Initiator in a particular situation.
(It was for this reason that fifteen years ago we emphasized the neglected
importance of "Initiation Theory" in the abstract deductive theory of that
era. '4)
We shall return to problems of operationalizing deductive deterrence
theory but, first, we would like to make several observations about its as-
sumption of rationality. An assumption of some kind of "rationality" in
deductive deterrence theory is necessary and useful as a starting point, but
a general, unrefined assumption of rationality appears to us as grossly in-
sufficient to operationalize the theory adequately. We do not reject the
usefulness of the general assumption of rationality-even "pure ration-
ality"-for the development of theory. But, as Thomas Schelling cau-
tioned many years ago, "whether the resulting theory provides good or
poor insight into actual behavior is ... a matter for subsequent judg-

I3Paul Huth and Bruce Russett, "What Makes Deterrence Work? Cases From i900-i980,"
World Politics 36 (July i984), 496-526.
I George and Smoke (fn. 2), chap. I7.

This content downloaded from

200.52.255.133 on Tue, 27 Jun 2023 01:57:21 +00:00
All use subject to https://about.jstor.org/terms
 176 WORLD POLITICS

ment. "i Even if one operates with the additional assumption of a "uni-
tary actor," a more specific version of the type of rationality assumed to
be operating is required, if only to indicate how one believes actors in
general, or the specific actor in question who is contemplating a challenge
to deterrence, will deal with the well-known cognitive limits on ration-
ality. (Not all actors deal with these cognitive limits in the same way.)'6
Herbert Simon's concept of "bounded rationality" is of some use but, as
he recognizes, it is often insufficient for fine-tuning rational choice theo-
ries.'7 If the theorist wishes to continue to "Black Box" the decision-mak-
ing process by means of assumptions (instead of getting into the Black
Box to observe and/or estimate what is actually going on, as policy makers
with the assistance of their intelligence specialists attempt to do), then it is
often desirable to develop what Simon calls "auxiliary assumptions" as to
the type of rationality that actors in general, or the particular actors about
whom predictions are to be made, typically exhibit in situations of this
kind.
Such auxiliary assumptions regarding rationality are often added to
their models, as Simon notes, by investigators who formulate and employ
deductive theory. But where do they come from? Not from contemplat-
ing one's navel but rather from reflecting on relevant empirical observa-
tions of actual behavior and converting such observations into auxiliary
assumptions to be added to the theory and incorporated into its internal
logical structure.
Important questions arise in addressing the task of operationalization,
the answers to which are not self-evident. First, which of the key concepts
of rational deterrence theory should be singled out for this treatment?
Achen and Snidal, like other deterrence theorists, emphasize the impor-
tance of (i) the credibility of a deterrence commitment/threat, and (2) the
need for a threat of punishment that is sufficiently strong to overcome the
Initiator's inclination to challenge deterrence by influencing his cost-ben-
efit calculations. Are these the key concepts of the theory to be operation-
alized? Or will deductive theorists avoid dealing with these particular
concepts and settle for operationalizing the Initiator's utility calculations?
This would require the investigator to determine whether, in contem-
plating a possible challenge to the status quo, the would-be Initiator ex-

i5 Schelling, Strategy of Conflict (Cambridge: Harvard University Press, i960), 4. For a

fuller statement of our concern over the limitations of the rationality assumption in deterrence
theory see George and Smoke (fn. 2), 73-77.
i6 See, for example, the discussion of different ways in which individuals attempt to cope
with the cognitive limits on rationality (i.e., inadequate information, inadequate knowledge,
and value complexity) in A. L. George, Presidential Decisionmaking in Foreign Policy (Boulder,
CO: Westview Press, ig80), chap. 2.
7 Simon, "Human Nature in Politics: The Dialogue of Psychology with Political Science,"
American Political Science Review 79 (June i985), 293-304. See especially pp. 295 and 300.

This content downloaded from

200.52.255.133 on Tue, 27 Jun 2023 01:57:21 +00:00
All use subject to https://about.jstor.org/terms
 RATIONAL DETERRENCE: GEORGE & SMOKE 177

periences "positive utility" or "negative utility" in a particular deterrence

situation. If so, how useful would such operationalization of the theory
be for predicting deterrence success and failure?
Second, what kind of predictions would a full-fledged, adequately op-
erationalized deductive deterrence theory be capable of making, and how
useful would such predictions be? Would such a theory modestly claim
(as does Bueno de Mesquita, who regards "positive utility" as a necessary
condition for war initiation') to have identified only a "necessary" con-
dition for deterrence success, leaving unidentified the all-important other
conditions? If, for example, either the credibility or the potency of a de-
terrent threat (or, more generally, "negative utility") is only a necessary
but not sufficient condition for deterrence success, then such a theory does
not predict deterrence success but merely observes that the stipulated nec-
essary condition would be present if a deterrence success occurred. The
predictive power of the theory and its utility would be further diminished
because a so-called "necessary" condition, such as credibility of the deter-
rent threat, could be present (as in the Pearl Harbor case) even in cases of
deterrence failure.
On the other hand, would an adequately operationalized deductive
theory make the much more ambitious claim that it identified one or
more sufficient conditions for deterrence success? If so, what are the can-
didates for a sufficient condition? Credibility of a deterrent commitment,
we already know from historical cases, cannot be considered a sufficient
condition for deterrence success. In the Pearl Harbor case, for instance,
the Japanese accepted the U.S. deterrence commitment as credible but
nonetheless attacked, in part because the United States was also subject-
ing them to strong coercive diplomacy. Again, is a threat of punishment
potent enough to turn the Initiator's cost-benefit calculations against an
"attack" a sufficient condition? Not by itself, quite obviously, since such
a threat must also be credible. Perhaps one can say that a sufficiently po-
tent threat (whatever that is) is at least a "necessary," though certainly not
a sufficient, condition for deterrence success. Such a restricted claim is
consistent with what we have already learned from history: while "mas-
sive retaliation" was an enormously potent threat, it often lacked enough
credibility and relevance to deter some types of challenges to deterrence
commitments made by the United States on behalf of its foreign policy
interests.

Evidently some combination of these two conditions- credibility and

potency of deterrence threat- is relevant for deterrence. Leaving aside
the important and not easily answered question of their required mag-

i8 See Bruce Bueno de Mesquita, The War Trap (New Haven: Yale University Press, i98i).

This content downloaded from

200.52.255.133 on Tue, 27 Jun 2023 01:57:21 +00:00
All use subject to https://about.jstor.org/terms
178 WORLD POLITICS

nitude in any particular situation, difficult questions for deductive theory

remain. First, if both these conditions are held to be necessary and/or suf-
ficient, how can they be operationalized? This task is most formidable.
The alternative to avoid dealing directly with them and to focus, in-
stead, on the would-be Initiator's "positive" or "negative" utility-would
leave us with a disappointing, almost trivial, theory that could claim to
identify only a necessary condition for challenging deterrence and, hence,
could not really predict deterrence outcomes.
A second issue would arise if, in applying such a theory, one encoun-
tered historical cases in which both conditions required by the theory
were measured as being present but, nonetheless, deterrence failed. How
would the deductive theory explain such anomalous results? Would the
proponents of the deductive theory "save" it by arguing that the incorrect
predictions must be due to measurement errors that could in principle be
reduced or eliminated by improvements in its operationalization? Or
would they concede the possibility that the deductive theory leaves out
variables (e.g., domestic political constraints, psychological impediments
to optimal information processing, misperception and miscalculation of
the opponent's intentions) that affect deterrence outcomes, or that it
makes defective assumptions about rationality and strategic interaction?
How would deductive theorists decide which of these two possible expla-
nations-measurement error or failure to consider other relevant varia-
bles-is correct?
Still another issue-the familiar and vexing one already encountered
in empirical research on deterrence-concerns the possibility that the ab-
sence of a challenge to deterrence might be spuriously scored as a deter-
rence success. Thus, the conditions postulated by the theory as necessary
and sufficient for deterrence success might be established as present in a
particular case in which, to be sure, deterrence was not challenged, but
nevertheless a claim that the theory has made a successful prediction
might not be justified.
Alternative explanations for the absence of a challenge to the status
quo, which are not encompassed by deductive deterrence theory, would
also have to be considered. Thus, although dissatisfied with the status
quo, the Initiator might not have contemplated challenging deterrence
and, hence, need not have been deterred. Or, while considering a possible
challenge, the Initiator may have decided not to attack for other reasons
altogether (e.g., domestic or allied constraints, internal policy disagree-
ments, the expectation of eventually securing a favorable resolution to his
grievance by diplomacy), and not because of the presumed credibility and
potency of the deterrence threat.
Finally, we feel obliged to say that the failure fully to specify and to

This content downloaded from

200.52.255.133 on Tue, 27 Jun 2023 01:57:21 +00:00
All use subject to https://about.jstor.org/terms
 RATIONAL DETERRENCE: GEORGE & SMOKE 179

operationalize rational deterrence theory encourages its defenders-un-

wittingly to be sure-to employ it in ways that risk giving the theory a
tautological and untestable character. In its present non-operationalized
form the theory is difficult, if not impossible, to refute; as a result, it is
being endowed with a validity that is highly questionable, insufficiently
qualified, and certainly premature. This is so because the theory, lacking
operationalization, is not capable of generating predictions for specific
cases. Instead, the outcomes of specific cases-once they are known- can
all too easily be claimed to be consistent with the theory. After the fact,
one can attribute historical cases of deterrence failure to lack of sufficient
credibility and/or to an insufficiently potent threat of punishment. Simi-
larly, ostensible cases of deterrence success can all too readily be chalked
up to the use of a deterrence strategy that met the postulated require-
ments of the theory.
One way of adjusting to some of the difficulties we have noted would
be to abandon the effort to state necessary and/or sufficient conditions
and settle for the goal of constructing and validating a theory of deter-
rence that would at least be capable of making predictions on a probabi-
listic basis. A theory couched in probabilistic terms is not without value-
if its expected performance can be specified and validated; but it should
also invite efforts to identify additional conditions under which deter-
rence efforts are likely to succeed or fail. The kinds of "conditional gen-
eralizations" that can be abstracted from historical case studies of the
kind we presented in our earlier work might be helpful in this respect.'9
Certainly efforts to develop deductive deterrence theory should con-
tinue and be pressed forward. We have argued that the standards and
requirements for such a theory and the type of predictions it would be
capable of making should be explicated rather than obscured by general
arguments for the superiority of this type of theory. If this is done, it will
become obvious that the achievement of a robust deductive deterrence
theory is far removed from, and in any case of quite uncertain value for,
providing strong guidance for policy-making purposes. Neorealist struc-
tural theory in international relations, with which Achen and Snidal cor-
rectly connect rational deterrence theory, addresses only one level of anal-
ysis. The foremost expositors and proponents of neorealist structural
theory explicitly concede that it is not a theory of foreign policy. Admit-
tedly, a theory of foreign policy, strictly speaking, does not exist and,
given the complexity and difficulty of the task, is not likely to emerge in
robust form for many years, if ever. Neorealist structural theory, in our
view, is necessary, but quite insufficient by itself either for adequately un-

19 George and Smoke (fn. 2), chap. i8.

This content downloaded from

200.52.255.133 on Tue, 27 Jun 2023 01:57:21 +00:00
All use subject to https://about.jstor.org/terms
 180 WORLD POLITICS

derstanding and explaining much of importance in international rela-

tions or for managing foreign policy. Empirical studies of decision mak-
ing and strategic interaction are necessary to fill, to some extent at least,
the important vacuum it leaves.
Deductive deterrence theory in its present primitive, undeveloped
form (which, unlike Achen and Snidal, we do not equate with the virtue
of "parsimony") contributes very little to an understanding of the uses
and limitations of deterrence strategy as an instrument of foreign policy.
The possibility of unwarranted and dangerous uses of abstract deterrence
theory as a basis for prescriptive advice to policy makers worried many in
the early 1970s.20 Ideas associated with the theory stimulated or accom-
panied the development of strategic theory, which, in turn, has provided
a framework within which policy makers (not deductive theorists) have
tried to address the operational requirements of strategic deterrence.
Contrary to Achen and Snidal, however, we argue that whatever contri-
bution abstract deterrence concepts made to the development of strategic
theory, it was not matched by a similar contribution to an understanding
of the uses and limitations of deterrence strategy at lower levels of con-
flict.
Whether deterrence theory is derived from deductive exercises or em-
pirical research, we continue to argue-as we did fifteen years ago-for
a sharper distinction between deterrence theory and deterrence strategy,
and for a better understanding of the relationship between the two. De-
terrence theory (limited and imperfect as it is) should be regarded as, at
best, an aid to devising deterrence strategies. Its most relevant and re-
sponsible use for policy-making purposes is not its prescriptive power but
rather its diagnostic function, i.e., its assistance to policy makers in as-
sessing the configuration of a situation in which some kind of challenge
to deterrence may occur. Since, as we argued and attempted to demon-
strate in our book, deterrence at the substrategic levels of conflict is highly
context-dependent, there is a critical need in policy making for good sit-
uational analysis.2I This would assess the nature and strength of the Ini-
tiator's motivation, how urgently he feels the need to challenge deter-
rence, the options available to him for doing so, the kind of utility
calculations and assessment of his options he is likely to be making, and
which of them he is likely to choose, if any. Situational analysis of this

20 See, e.g., ibid., chap. 3-

21On situational analysis and the diagnostic function of theory, see George and Smoke (fn.
2), chap. i6; also George (fn. i6), chap. I4. For a more detailed discussion of policy-relevant
theory see George and Smoke (fn. 2), app. A longer version of this appendix was published
by Smoke and George, "Theory for Policy in International Relations," Policy Sciences 4 (De-
cember I973), 387-4I3-

This content downloaded from

200.52.255.133 on Tue, 27 Jun 2023 01:57:21 +00:00
All use subject to https://about.jstor.org/terms
 RATIONAL DETERRENCE: GEORGE & SMOKE 181

kind is likely to provide a better understanding of the deterrence problem

and of what variant of deterrence strategy (alone or together with other
policy options) is likely to be most appropriate and effective in that par-
ticular situation. Since the context of even a single deterrence situation
can and does change over time, situational analysis is a continuing task.
Middle-range theory, derived through empirical analysis and expressed
in the form of conditional generalizations, together with general models
of the opponent's behavioral characteristics, are of some help in situa-
tional analysis, as are the findings of the more recent empirical studies of
deterrence subjected to critical analysis by Achen and Snidal.
As these remarks indicate, the policy maker (unlike the deductive the-
orist) makes the attempt to get into the Black Boxes of decision making
and strategic interaction. Would deductive theorists say the policy maker
should not do so but, instead, merely rely on abstract rational deterrence
theory? In our view deductive deterrence theory can give the policy
maker very limited help in assessing the decision-making process of the
opponent or in anticipating the outcome of the interaction between his
own deterrence strategy and the adversary's choice of action. For this, the
policy maker has to rely largely on the information and analysis provided
by his intelligence and advisory system.
We have saved for the end of our comment three other major limita-
tions of deterrence theory whether deductive or empirical. The first is
that the theory cannot define its own scope and relevance as a means of
achieving foreign policy goals.22 That is to say, as implied by the title
given to our comment, deterrence theory per se provides no criteria to
indicate when a deterrence policy should be applied in foreign policy.
Often, but certainly not always (and particularly not at the substrategic
levels of conflict), the criteria are self-evident or easily identified on an ad
hoc basis. But, viewed in the broader context of foreign policy, as it should
be, deterrence theory is best understood as a contingent policy theory: "if
a state believes it necessary or desirable to employ deterrence in a partic-
ular situation, then the general requirements for devising an appropriate,
hopefully viable deterrence strategy are the following. . ." Only a foreign
policy theory, or a theory of statecraft, can provide guidance as to when a
state should employ deterrence to protect specified types of interests as
against (or in addition to) other strategies for protecting its interests and/
or resolving points of conflict with other states. Deterrence strategy, in
other words, must be viewed as only one of a number of different instru-
ments of foreign policy. Failure to recognize this-either by theorists or

22 See George and Smoke (fn. 2), 77-78.

This content downloaded from

200.52.255.133 on Tue, 27 Jun 2023 01:57:21 +00:00
All use subject to https://about.jstor.org/terms
 182 WORLD POLITICS

policy makers- can, and does, easily lead, as we argued in our book and
continue to argue, to a narrow over-reliance on deterrence strategy in for-
eign policy.
Another, related limitation of deterrence theory is its exclusive preoc-
cupation with threats of punishment as a means of persuading an adver-
sary not to challenge a situation in which there is disagreement. Critics of
deterrence, extending back to the early years of the Cold War, have em-
phasized (sometimes overemphasized) that threats can be counterproduc-
tive. We have argued that early deterrence theorists erred in their evident
assumption that a viable theory of deterrence can be developed indepen-
dently of a broader theory of processes by which nations influence each
other, one that encompasses the utility of positive inducements as well as,
or in lieu of, threats of negative sanctions.23 Thus, for example, the cred-
ibility and potency of a deterrence threat might not be so critical if instead
(or in addition) sufficient positive incentives for remaining at the status
quo are offered.
Finally, deterrence theory ignores the frequent need to regard deter-
rence, even when it is necessary and effective, as a strategy to buy time.
Deterrence can severely frustrate an adversary who is strongly motivated
to change a status quo that he regards as invidious, especially when he
feels it legitimate to do so. The consequences of continued frustration are
not easily predictable and are not always favorable to the deterring
power. Deterrence success in the short run is not always beneficial in the
longer run; the adversary may become more desperate to mount a chal-
lenge and may proceed to acquire greater resources for doing so. Under
such circumstances the most reliable benefit of successful deterrence may
be more time time which is best used not in a possibly futile effort to
maintain deterrence indefinitely but to work out, if possible, an accom-
modation of conflicting interests so as to reduce reliance on deterrence
and avoid overt conflict.24

23Ibid., chap. 2I.

24 For additional discussion see George and Smoke (fn. 2), 5-6; also George (fn. i6), 252-54.
The importance of "promises" (as against "threats") was briefly noted by Schelling (fn. I5),
43-46, I3I-37, 175-77. Important leads for the development of a broader influence theory were
given by David A. Baldwin, "The Power of Positive Sanctions," World Politics 24 (October
I971), I9-38, and "Inter-Nation Influence Revisited," Journal of Conflict Resolution I5 (De-
cember I971), 47I-86. The importance of conceptualizing the strategy of coercive diplomacy
to include possible use of "carrots" as well as "sticks" was emphasized and illustrated in
George et al. (fn. 7). The concept of "crisis bargaining" developed by Glenn Snyder and Paul
Diesing includes accommodative as well as coercive actions; Conflict Among Nations (Prince-
ton: Princeton University Press, 1977). Other writers, too numerous to mention here, have
also emphasized the need for a broader, multifaceted theory of inducement or influence. More
recently, Janice Gross Stein and Richard Ned Lebow have emphasized the need for more at-
tention to various forms of "reassurance" to supplement or replace deterrence in conflict sit-
uations.

This content downloaded from

200.52.255.133 on Tue, 27 Jun 2023 01:57:21 +00:00
All use subject to https://about.jstor.org/terms