Download as pdf or txt
Download as pdf or txt
You are on page 1of 21

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/225830407

Mathematical Models of Foreign Policy Decision-


Making

Article  in  Synthese · September 1994


DOI: 10.1007/BF01063911

CITATIONS READS

30 329

3 authors:

Alex Mintz Nehemia Geva


Interdisciplinary Center (IDC) Herzliya Texas A&M University
62 PUBLICATIONS   1,862 CITATIONS    36 PUBLICATIONS   882 CITATIONS   

SEE PROFILE SEE PROFILE

Karl DeRouen
University of Alabama
80 PUBLICATIONS   1,848 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Research assessment View project

peace studies/conflict/international relations View project

All content following this page was uploaded by Alex Mintz on 16 March 2016.

The user has requested enhancement of the downloaded file.


ALEX MINTZ, NEHEMIA GEVA~ AND KARL DEROUEN JR.*

MATHEMATICAL M O D E L S OF F O R E I G N P O L I C Y
DECISION-MAKING: COMPENSATORY
VS. N O N C O M P E N S A T O R Y

ABSTRACT. There are presently two leading foreign policy decision-making paradigms
in vogue. The first is based on the classical or rational model originally posited by yon
Neumann and Morgenstern to explain microeconomic decisions. The second is based on
the cybernetic perspective whose groundwork was laid by Herbert Simon in his early
research on bounded rationality. In this paper we introduce a third perspective - the
poliheuristic theory of decision-making - as an alternative to the rational actor and
cybernetic paradigms in international relations. This theory is drawn in large part from
research on heuristics done in experimental cognitive psychology. According to the
poliheuristic theory, policy makers use poly (many) heuristics while focusing on a very
narrow range of options and dimensions when making decisions. Among them, the
political dimension is noncompensatory. The paper also delineates the mathematical
formulations of the three decision-making models.

There are presently two leading foreign policy decision-making para-


digms in vogue. The first is based on the classical or rational model
originally posited by von Neumann and Morgenstern to explain micro-
economic decisions. The second is based on the cybernetic perspective
whose groundwork was laid by Herbert Simon in his early research on
bounded rationality, and later was refined by John Steinbruner. In this
paper we introduce a third perspective - the poliheuristic theory of
decision-making - as an alternative to the rational actor and cybernetic
paradigms in international relations. This theory is drawn in large part
from research on heuristics done in experimental cognitive psychology.
In the following pages we offer: (a) an overview of these paradigms as
they are applied in the context of decisions on the use of force in
international relations; (b) a description of the mathematical ex-
pressions of these models; and (c) a discussion of these equations in the
context of compensatory vs. non-compensatory principles of decision-
making.
Each of the three paradigms of foreign policy decision-making is
underpinned by different decision rules. A decision rule enables the

Synthese 100: 441-460, 1994.


© 1994 Kluwer Academic Publishers. Printed in the Netherlands.
442 ALEX MINTZ, NEHEMIA GEVA, AND KARL DEROUEN JR.

decision maker to select "the most preferred alternative; generally from


a partial or total ordering, or a prioritization of alternatives" (Sage
1990, p. 232). The rational actor/expected utility model is based upon
a holistic evaluation of alternatives. The cybernetic and poliheuristic
models are each grounded in nonholistic process rules. The poliheuristic
model also posits a noncompensatory process rule (see below). Holistic
decision rules generally rely upon an evaluation of all alternatives.
Only then evaluations are compared to determine the best choice.
Nonholistic models, on the other hand, employ a simplified process
whereby the decision-maker sequentially eliminates or adopts alterna-
tives "by comparing them to each other, or against a standard, either
across dimensions or across alternatives" (Sage 1990, p. 233). Holistic
decision-making is then more formal and demanding, while the non-
holistic heuristic-based models are typically streamlined by rules which
offer cognitive shortcuts.

THREE APPROACHES TO FOREIGN POLICY DECISION MAKING

The Rational Actor/Expected Utility Approach. Since the early 1980s


research on foreign policy decision making appears to be dominated
by studies based on the rational choice paradigm 1 with the proliferation
of game theoretic and expected utility analyses (see e.g., the work of
Steven Brams 1985; Bruce Bueno de Mesqtfita 1981; 1985; and Bueno
de Mesquita and David Lalman 1990, 1992). According to Oneal (1988,
p. 620):

[t]he weight of the evidence indicates that decision makers as a whole do adopt rational
decision making procedures during international crises. The bureaucratic politics, organi-
zational process, and individual stress models are useful reminders of the limits of
rationality, but none captures the essence of decision in time of crisis.

The decision process executed by the rational actor is, according to


Graham Allison (1971, p. 30) " . . . the consistent, value-maximizing
choice within specified constraints". The payoffs or consequences of
each alternative are ranked in terms of the decision-maker's values and
objectives. The decision maker chooses an alternative "whose conse-
quences rank highest in terms of his goals and objectives" (1971, p.
33). If the costs attached to a certain alternative increase, then the
chances of selecting that alternative are reduced. The reverse also holds
FOREIGN POLICY DECISION-MAKING 443

true (1971, p. 34; but see Bendor and Hammond 1992). The rational
actor or the classical model requires some rather strict assumptions that
many experimental studies have found to be too demanding to accu-
rately describe the cognitive processes associated with foreign policy
decisions. For instance, the nation as an actor is often assumed to have
one set of goals, one set of options, and one estimate of the conse-
quences of each alternative. In addition, the model typically assumes
that the decision be "faithfully implemented"; and that the outcome
be used as a benchmark for future actions (Onea! 1988, p. 601).
The rational actor perspective in foreign policy decision-making is
best illustrated by the expected utility approach of Bueno de Mesquita
(1981, 1985) and Bueno de Mesquita and Lalman (1990, 1992), which
utilizes axiomatically derived formal equations to replicate the decision
process of leaders. The approach is an elaborate approximation of the
cost/benefit models originally used to study economic decisions.
According to Bueno de Mesquita (1981, 1985), nations are led by
rational, forward-looking, expected-utility-maximizing leaders. The
strategies these leaders select are a function of "the values they attach
to alternative outcomes and the beliefs they hold regarding how their
adversary will respond to their strategic decisions" (Bueno de Mesquita
and Lalman 1990, p. 750). Decision makers evaluate the costs and
benefits associated with each alternative to obtain "the largest net gain
(expected utility) at an acceptable level of risk" (Bueno de Mesquita
1984, p. 228).
Leaders, using the expected utility approach, select the alternative
that maximizes utility in a holistic and comparative fashion (see Bueno
de Mesquita and Lalman 1990, 1992). The usefulness of the expected
utility theory becomes evident when its elegant mathematics generate
results that can be interpreted in terms of standard conflict parameters.
For example, expected utility theory can be used to gain substantive
insight into the role alliances play within both balance of power, and
power transition theories. That is, the expected utility captures these
two theories by explaining what a third party will do when confronted
by an ongoing conflict (see Bueno de Mesquita 1985; Bueno de Mes-
quita and Lalman 1986).
The Bounded Rational/Cybernetic Perspective. The cybernetic de-
cision-making model of Steinbruner (1974) offers a cognitive substitute
to the rational actor model (Steiner 1983, p. 373). Steinbruner notes
that the rational (or analytic) paradigm is powerful, but has limitations.
444 ALEX MINTZ, NEHEMIA GEVA, AND KARL DEROUEN JR.

Cybernetic models refute the classical rational assumptions (see Sylvan


and Chan 1984).
Since Simon's (1957) groundbreaking critique of the rational actor
approach, various scholars have attributed to the expected utility and
other analytical decision-making strategies the requirement for exten-
sive processing time, cognitive effort, concentration and skills that in
many cases are not available, especially under time pressures and rap-
idly changing conditions, and in other cases are not worthy to the
decision (cf. Klein 1989; Beach 1990). Accordingly, these scholars argue
that decision-makers seldom maximize utility. The bounded rational/
cybernetic paradigm (Simon 1959) is designed to address these seem-
ingly complex and counterintuitive decision processes by providing an
alternative decision structure.
The cybernetic decision making process is less exhaustive (non-hotis-
tic) due to cognitive and motivational constraints. The decision process
is less comprehensive, as not all alternatives are considered along all
available dimensions.
The example of the tennis player was given to explain cybernetic
decision making (Steinbruner 1974, pp. 49-50). A seemingly simple
tennis stroke actually requires thousands of decisions about crucial
aspects such as footwork, racquet speed, timing, keeping one's eyes on
the ball, placement and follow-through. But players do not actually
think through each of these aspects every time they make a shot. The
process becomes automatic and routinized and, ostensibly, never varies
very much from the baseline pattern. According to Steinbruner (1974),
the rational paradigm might explain the tennis player's actions by
wrongly assigning him or her the ability to make unconscious calcu-
lations every time the ball is struck. One cannot be sure however, that
the tennis player was acting in accordance with the rational actor model
just because he or she hit a lob over the opponent's head - the ball may
have been mis-hit. Put differently, if there is no way of independently
establishing utility values, the postulate that decision makers will always
choose the alternative that maximizes utility is a tautology (1974, p.
28).
Cybernetic decision making centers around the minimization of un-
certainty through information feedback loops (1974, p. 51). The rational
framework assumes that "the range of possible outcomes is known,
and thereby eliminates the possibility that an outcome might occur
which was not even visualized in advance. It assumes that [the] rules
FOREIGN POLICY DECISION-MAKING 445

[of the game] are specified and stable. For complex problems neither
of these assumptions can be held", however (1974, p. 18).
The cybernetic approach is needed because decision-makers lack the
fundamental cognitive skills needed to carry out the rational schema
during the structural uncertainty-laden policy problems of our time
(Simon 1957; Steiner 1983, p. 423). The bounded rational/cybernetic
paradigm precludes the need to calculate optimal procedures and alter-
natives based on preferred outcomes by eliminating alternatives and
ignoring the environment and the issue of variety (Simon 1957).
The cybernetic approach was applied in the context of decisions on
the use of force by Ostrom and Job (1986), and was later refined by
James and Oneal (1991). By doing so, they captured Simon's satisficing
decision principle which relaxes the cognitively demanding maximiz-
ation assumption of the rational actor perspective. Ostrom and Job
posit that the leader simplifies the decision task by concentrating on
three decision environments (dimensions), and compares the summary
evaluation of the use of force alternative across these dimension to a
satisficing threshold. 2 If this threshold is crossed, the probability of the
use of force increases.
The three decision dimensions (or as stated originally, decision en-
vironments) are the international, domestic, and political. The interna-
tional dimension captures the strategic balance, international tension,
and extent of involvement in ongoing wars. The domestic environment
taps the state of the domestic economy, aversion to war, and the public
attitude towards international tension and towards the strategic balance.
The political dimension captures public approval of the president, politi-
cal success, and position on the electoral calendar. 3
Prior to describing the third and most recent approach to foreign
policy decision-making - the poliheuristic model - it is necessary to
outline a decision principle that distinguishes this model from the pre-
vious approaches. This principle relatesto the noncompensatory versus
the compensatory processing of information that underlies decision
making.

COMPENSATORY VS. NONCOMPENSATORY STRATEGIES


OF FOREIGN POLICY DECISION-MAKING

Studies of foreign policy decision making have primarily utilized com-


pensatory decision making strategies. Compensatory processes imply
446 ALEX MINTZ~ NEHEMIA GEVA~ AND KARL DEROUEN JR.

that when political leaders choose among alternatives, a high value on


one dimension (e.g., the military balance of force) can compensate for
a low value on another dimension (e.g., political) for the same decision
alternative (see Einhorn and Hogarth 1981).
The compensatory decision strategy may be performed in one of two
ways. According to the linear model each dimension for a decision
alternative is given a value, and the dimensions are combined additively
to produce an overall value for each alternative. The decision-maker
compares the alternatives on the basis of these overall values, and
chooses the alternative with the highest score (see Billings and Scherer
1988; Ford et al. 1989; Payne et al. 1988). The second model is the
additive difference model by which alternatives are compared on each
dimension, and these differences are then summed across dimensions.
This summation results in a preference for one decision alternative
(Olshavsky 1979).
A review of the literature on foreign policy decision-making shows
that the compensatory linear model underlies the decision processes
both for the work generated within the perspective of the rational actor-
expected utility model (see e.g., Bueno de Mesquita and Lalman's work
on the use of force, 1990), and also the research that was triggered
within the framework of the cybernetic/bounded rationality approach
to decision-making (Ostrom and Job 1986; and James and Oneal 1991).
According to these authors, scores are combined to maximize (Bueno
de Mesquita and Lalman 1990) or "satisfy" (Ostrom and Job 1986)
utility.
The compensatory decision strategy is often associated with a holistic
(comprehensive and exhaustive) process where all available alternatives
are considered along all relevant dimensions in an attempt to select the
alternative with the highest net gain. 4
Using a compensatory model, the decision maker may trade-off a
high value on one dimension of an alternative with a low value on
another dimension. The expected utility models of Bueno de Mesquita
and Lalman (1990, 1992) are compensatory in that the elements that
comprise the utility scores are additive. For example, high political
costs associated with the use of force can be compensated for by large
military benefits.
Compensatory models are additive. The essence of this model is that
the various elements of a multidimensional object contribute indepen-
dently to its overall utility. The components are added together for a
FOREIGN POLICY DECISION-MAKING 447

"score" (e.g., "net gain", "critical value", "satisfactory score" etc.).


While compensatory processes can be applied to some situations, non-
compensatory processes can be applied to other incidents.
Einhorn (1971, p. 1) pointed out that the compensatory decision-
making process has typically been conceptualized in terms of a linear
model in which components are added together to form an overall
evaluation of alternatives. Obviously, when linear models are applied
to data which may possess nonlinear characteristics, the model accounts
for less variation than a nonlinear model would have (Brannick and
Brannick 1989; Einhorn 1970). While such models may appear robust,
it is uncertain whether they accurately portray the underlying cognitive
processes. For example, it has been shown that the domestic component
in decisions to use force is important only as far as the leader's ruling
coalition is dissatisfied with the economy or other domestic issue (Mor-
gan and Bickers 1992). That is, the level of total public support is
irrelevant and subsequently a nonlinear relationship obtains between
public support and uses of force.
The poliheuristic, noncompensatory model captures the heuristic nat-
ure of the foreign policy decision making process (see Mintz and Geva
1994). It does not require exhaustive searches or comparisons. Such a
decision process is less complex than the compensatory process.
Compensatory processes are marked by an inter-dimensional infor-
mation search using "a constant amount of information per alternative"
(Billings and Marcus 1983). Noncompensatory selection procedures
usually imply an intradimensional component which is attribute or di-
mension-based, rather than alternative-based, and does not imply a
constant amount of time spent on each alternative (Mintz 1993a, p.
599). Compensatory models are alternative-based as an alternative is
considered along available dimensions before moving on to evaluate
another alternative (Payne, Bettman, and Johnson 1988, p. 536). In
the noncompensatory model, an alternative that does not meet a certain
threshold on the most important dimension will be discarded (Mintz
1993a, p. 599). Leaders "weigh" the relative importance of each dimen-
sion (V~) of the decision alternative (Ai). In the noncompensatory
model, the expected gain on a critical dimension should be higher than
a certain threshold level. If the expected gain along a critical dimension
is below the threshold value, then the alternative is eliminated. In this
sense the model depends upon decision heuristics to deal with the
nonholistic nature of the process. The noncompensatory model facili-
448 ALEX MINTZ~ NEHEMIA GEVA, AND KARL DEROUEN JR.

tates an interactive use of cues or dimensions. Since there is no compen-


sation between dimensions, the search pattern spends differing amounts
of time and uses nonconstant amounts of information across alterna-
tives. This cognitive "shortcut" is what classifies the noncompensatory
approach as heuristic.
It has been suggested that noncompensatory, rather than compensa-
tory linear models, are used in more complicated decision environments
because they are cognitively easier. The definitive feature of noncom-
pensatory models is their multiplicative nature which effectively rules
out compensation between independent variables (Brannick and Bran-
nick 1989; Mintz and Geva 1994).
Since noncompensatory models employ some form of elimination by
dimensions, they can also be classified as satisficing models and not as
optimizing models, as in the case of expected utility models (Mintz
and Geva 1994). Noncompensatory models are concerned with finding
"acceptable", rather than maximizing, alternatives (ibid). They are also
satisficing models because it is possible that not all dimensions will be
considered before a decision is made based on partial information. The
idea behind the noncompensatory model then is to quickly eliminate
alternatives in order to simplify the information search and evaluation
phases of the decision process (Mintz and Geva 1994; Payne, Bettman,
and Johnson 1988, p. 534).

THE POLIHEURISTIC MODEL

The poliheuristic model of decision making has been developed recently


as an alternative to the expected utility and cybernetic models. The
model recognizes the use of poly (many) heuristics by decision makers
to simplify complex foreign policy decisions (see Mintz and Geva 1994).
Heuristics compensate for incomplete information as they provide cog-
nitive shortcuts to intricate foreign policy matters by organizing the
information in such a manner as to facilitate the decision process (Snid-
erman, Brody, and Tetlock 1991, p. 19).
Heuristic-based models, are particularly well-suited to decisions when
one goal or outcome is understood to be paramount (Mintz 1993a).
For example, when domestic politics are most highly regarded, all
other dimensions are diminished and the decision process is streamlined
(1993a). Domestic politics become the sine qua non, and subsequently
FOREIGN POLICY DECISION-MAKING 449

multiple outcomes need not be assessed alongside a multitude of trade-


offs (see Axelrod 1976).
The poliheuristic model of foreign policy decision making also as-
sumes that actual decision-making behavior is not even boundedly
rational. Instead, it employs simple heuristics. It differs from the other
process models of foreign policy decision-making as it is typically associ-
ated with a non-exhaustive process in which "a decision is being made
prior to the completion of the consideration of all alternatives along
all dimensions" (Mintz 1993a; Mintz and Geva 1994, forthcoming).
According to the poliheuristic model, comparisons are made within a
(very) restrictive alternative set and attribute set. The selection of an
alternative is not based upon the detailed conscious consideration of
all aspects of the alternatives. Thus, rather than choosing an alternative
that maximizes utility on the basis of an exhaustive comparison process,
or selecting an alternative that "satisfices" a certain criterion, the theory
suggests that foreign policy decisions are often based on the rejection
or adoption of alternatives on the basis of one or at most a few criteria.
According to Mintz (1993a, pp. 598-599), rather than relying on "holis-
tic decision rules that require the evaluation and comparison of all
alternatives across different dimensions, the decision maker adopts
heuristic decision rules that do not require detailed and complicated
comparisons of alternatives and rejects or adopts alternatives on the
basis of one or a few criteria". Moreover, the poliheuristic model
represents a decision making process in which alternatives are not
rejected or adopted on the basis of comparison to a maximizing criterion
(1993a).
According to the poliheuristic model of decision, political leaders
almost by definition take into account (explicitly or implicitly) political
factors and consequences while making policy decisions. The model
suggests, then, that leaders "will eliminate an option that is below the
'cutoff' level on the political dimension. In other words, one set of
calculations that enters into their decision making calculus is political,
the other is substantive (e.g., economic, strategic, diplomatic, legal,
social, and so on). The context of the substantive dimension(s) that
enters into the calculus of decision making varies as a function of the
substance of the decision (i.e., it can be economic, military, etc.). The
noncompensatory principle suggests that regardless of the score on this
dimension, however, the use of force debate is moot if this alternative
is likely to politically damage the leaders" (Mintz 1993a, p. 60i).
450 ALEX MINTZ, NEHEMIA GEVA, AND KARL DEROUEN JR.

Leaders do not make tradeoffs across dimensions to compensate for


a low score on the political dimension. An alternative that is likely to
damage the political prospects of the leader is rejected without consider-
ing the "score" on the other dimensions. Remaining options are then
evaluated (sequentially) based on how they score on other dimensions.
The noncompensatory process does not always continue, however, until
only one alternative is left. There are situations where several remaining
alternatives all pass a threshold value and are then considered based
on a more analytic processing (or by selecting a lexicographic decision
rule) (Mintz 1993a). Even when one alternative is left (by default), a
final refinement of the default choice is typically performed "by trying
to minimize costs and maximize benefits" (1993a, p. 600).
The basic assumption of the poliheuristic model is that a necessary
though not sufficient condition for the use of force is that it will not
undermine the political fortunes of the leader (i.e. politicians will not
shoot themselves in the foot by pursuing policies that will damage
them politically). Available alternatives are screened along the most
important dimension, and the 'surviving' alternatives (chosen or unre-
jected) are scrutinized along the second and subsequent dimensions,
until a choice is made. It should be pointed out that only relevant
and nontrivial dimensions (e.g., political, military, economic, etc.) are
considered (1993a, p. 601).
While the expected utility theory of the use of force (Bueno de
Mesquita and Lalman 1992) and the cybernetic model of the use of
force (Ostrom and Job, 1986; James and Oneal, 1991) also capture the
salience of domestic politics to foreign policy decisions (see also Russett
1990; Mintz and Russett 1992), they fall short of recognizing the non-
compensatory nature of the decision process. In fact, none of these
studies views domestic politics as a prerequisite for the use of force in
a noncompensatory decision-making environment. In the next section
we describe the formal representations of the expected utility and cyber-
netic decision-making models discussed above, and specify the elimin-
ation by aspect version of the poliheuristic model.

MATHEMATICAL MODELS OF FOREIGN POLICY


DECISION-MAKING
Zinnes (1976) has comprehensively reviewed mathematical models of
international relations. Since the early 1960s, such models have also
FOREIGN POLICY DECISION-MAKING 451

been applied to international crisis behavior (Oneal 1988, p. 602).


Researchers turned their attention toward postdicting behavior of lead-
ers in times of foreign policy crisis. The Cuban Missile Crisis was a
major stimulus of this new research endeavor. By the 1980s, the genre
had turned towards trying to predict the war and peace decisions of
leaders (see Zinnes 1980). Predicting the use of force in times of crisis
became a particularly appealing agenda (e.g. Bueno de Mesquita 1981,
1985; Ostrom and Job 1986). What follows is an overview of the mathe-
matical application of the three leading decision-making models of the
use of force.
Compensatory models. In general terms the additive compensatory
model is specified simply as (see Brannick and Brannick 1989; Dawes
1964):
k

(1) Y = a + ~ biXi,
i--1

where the dependent variable is force/no force, and the independent


variables are the decision dimensions or cues. When there is more than
one independent variable, this model specification allows for compen-
sation to occur. Since the variables are added, a low score on one
variable will not necessarily preclude the use of force. This is consonant
with the rational actor paradigm in which decision-makers make de-
tailed assessments of costs and benefits. Subsequently, decision-makers
in the rational actor paradigm are fully prepared to think in terms of
trade-offs by comparing costs and benefits (Kinder and Weiss 1978, p.
708; Lebow 1981, p. 101, Oneal 1988, p. 575).
Bueno de Mesquita and Lalman (1992), in their expected utility
approach, provide a much more sophisticated formal conception of
the compensatory model of decisions to use force. The international
interactions game that the authors axiomatically derive is used to assess
the costs/benefit determinations decision-makers are said to practice.
The expected utility calculation used when leaders must decide whether
to use force, for example, is specified as:
(2) pi(ui[A i _ ~ p i _ ai((1 - pi)]) +
+ (1 - Pi)(ui[Aj - ~biPi - ai(1 - pi)]).

In this formulation, demands are represented by A. Thus [ui(At)] is


the value gained if one's demands are won, while [ui(Aj)] becomes the
452 ALEX MINTZ, NEHEMIA GEVA, AND KARL DEROUEN JR.

value lost by acceding to an opponent's demands, pi is the subjective


probability of winning, while (1 - pi) is the subjective probability of
losing. The term ~biP' denotes the domestic political costs associated
with uses of force rather than negotiation. This term captures costs
such as electoral defeat, violent demonstrations, and coups d'etat-
events which constrain decisions to use force (Bueno de Mesquita and
Latman 1992). The final term, ai(1 - p i ) , captures the costs involved
in fighting away from home measured in terms of tangibles such as life
and property. By calculating their expected utilities in this way, leaders
"evaluate the risky components of outcomes as lotteries" (Bueno de
Mesquita and Lalman 1992, p. 42).
The expected utility given by the above equation is then compared
to that of the other consequences of a binary choice set of using versus
not using force: e.g., capitulation, negotiation and acquiescencel The
compensatory nature of the above model is evident by the fact that
political costs (as well as other costs) are subtracted in the equation
from the gains accruing from demands. Hence, high political costs can
be compensated for by large gains from demands. Furthermore, in
order to make the choice between 'use of force' and 'no use of force'
the decision maker has to compute utilities for each alternative separ-
ately before making the choice. In other word, the decision-maker
performs an alternative-based decision process.
Other treatments of the compensatory decision process were pro-
vided by Ostrom and Job (1986) and James and Oneal (1991) in their
modeling of decisions to use force. These models differ from the ex-
pected utility approach in that satisficing behavior is subsumed within
a cybernetic framework consisting of three specific dimensions. The
Ostrom and Job (1986; see also James and Oneal 1991; Oneal and Lian
1992a, 1992b) model is specified as:

(3) Pr(Yt = 1) = P r ( Y * > h)


= Pr[(Z aiIit + ~ bjtDjt -1- ~ CkPkt + et) > hi,

where the dependent variable Yt takes the dichotomous form: force (1)
or no force (0); and a, b, and c represent the weights or relevance of
each of the dimensions: the international (I), domestic (D), and the
political (P). The probability that force will be used is then determined
by whether or not a threshold is crossed after the summation of the
three dimensions. The model is then estimated using probit analysis.
FOREIGN POLICY DECISION-MAKING 453

The Ostrom and Job (1986) model assumes that the decision-maker
creates a composite index with the three dimensions, which greatly
simplifies the decision environment. The model is compensatory be-
cause the equation allows for trade-offs among the dimensions since
the three dimensions are combined additively. It is an alternative-based
compensatory model since the calculus of decision is performed for one
alternative (use of force), and if the overall score exceeds a threshold
level, the alternative has a high probability to be chosen.
Noneompensatorymodels. The poliheuristic model of foreign policy
decision making presented above (and in greater detail in Mintz 1993a,
1993b; and Mintz and Geva 1994, forthcoming) suggests that foreign
policy decisions are typically based on a noncompensatory strategy. In
a choice situation, if a certain alternative is unacceptable on a given
dimension (e.g., it is unacceptable politically), then a high score on
another dimension (e.g., the military balance of forces) c a n n o t compen-
sate for it (or counteract it), and hence the alternative is eliminated.
The model is nonadditive: alternatives are eliminated if a score on a
critical dimension(s) is below a cutoff value (see Mintz 1993a, p. 599).
As was pointed out above, this model enables the decision maker to
reject an alternative on the basis of one (or a few) dimensions rather
than to evaluate an alternative along all other dimensions. It is a
dimensional based model, where alternatives are processed (one by
one) along a few critical dimensions. Furthermore, it is a nonexhaustive
process as not all alternatives must be reviewed along all dimensions
before a decision is made. The value of an alternative on a single
dimension is processed first and alternatives that score below the cutoff
value on this dimension are eliminated (Mintz 1993a; Mintz and Geva
1994). According to these authors (p. 8), "political leaders review
alternatives in light of a political dimension and reject all alternatives
that may damage them politically, then move to a substantive dimension
and reject alternatives that are low on this dimension".
To date, empirical support for the poliheuristic model primarily rests
on experimental data which utilized process-tracing methods rather
than structural equations. Yet, it should be noted that a major process
component of the poliheuristic model of foreign policy decision making
is based on the elimination-by aspect decision rule originally developed
by Tversky (1972).
The E B A model (Tversky 1972) views the choice as a covert sequen-
tial process where each alternative (Ai) is described by a set of aspects
454 A L E X MINTZ~ N E H E M I A G E V A , AND K A R L D E R O U E N J R .

(characteristics). At each stage of the process, an aspect (Vj) is selected.


The selection of an aspect eliminates all alternatives that do not contain
the selected aspect (Ranyard 1976, p. 299). The selection continues
until a single alternative remains. The selection of a particular aspect
is proportional to its weight. Aspects that are common to all the alterna-
tives under consideration do not affect the choice probabilities.
In the language of set theory, for a binary choice of the two alterna-
tive A1 ("use of force") and A2 ("international sanctions"), the aspects
common to both alternatives (e.g., political, economic, etc.) are
A1 A A2 = o9. The aspects unique to A1 (i.e., the military costs of
the operation, the 'rally' round the flag effect, etc.) are A1 A A2 =
a, while the aspects unique to A2 (e.g., diplomatic efforts) are
A2 A fi~ = /3. Thus, the probability of choosing alternative A1 over
A2 is:

(4) Pr(A~, 12) = U(~)


+

where U is a measure function on the set of aspects, often interpreted


as a utility function (Ranyard, 1976). The probability of selecting alter-
native A2 ("sanctions") over A~ ("use of force") is in contrast:

(5) Pr(A2, A~) = U(/3)


+

Ranyard (1976) pointed out that these equations are inconsistent with
behavioral observations. We therefore present an alternative formula-
tion based on Maddala (1983). Consider a case with three alternatives,
A~ ("use of force"), A2 ("sanctions"), and A3 ("do nothing"), charac-
terized by four aspects (or attributes) Vt, V2, V3 and V4 (the V's are
the utility values for aspect i), where:
A1 = (V1, V4)
A2 = (V2, V4)
A 3 = (V3)

The order by which the aspects are screened affects the choice. For
instance if V4 is selected as the first aspect for consideration, then, A3
is eliminated from further consideration (as it does not contain the V4
aspect). Hence the choice proceeds with the selection of alternative A1
or A2. Since aspect V4 is common to both alternatives it is not used in
the decision. Thus, the choice between As and Aa depends on the
F O R E I G N POLICY D E C I S I O N - M A K I N G 455

probability of selecting VI as the first aspect compared to selecting 7 2 .


Stated formally:

(6) Pr(llA1, A2, a3) = V1 +


V~ + V2 + V3 + V4

+ V4 Vl
V 1 -[- 7 2 -t- V 3 "[- V 4 7 1 -}- 7 2

= (probability that V1 will be selected as the first aspect) +


(probability that V4 will be selected as the first aspect) x
(probability that V~ will be selected as the second aspect),
as the probability that an aspect will be selected is pro-
portional to its weight.
Using the U.S. decision to attack Iraq as an example, Mintz (1993a)
has specified a "decision board" consisting of three alternatives ("use
of force", "containment", and "do nothing") and a set of five feasible
aspects (or dimensions) associated with each alternative (presidential
popularity, the state of the economy, projected military costs, etc.).
According to the poliheuristic model of decision-making, then,
5
~] V i = S
i=I

and the probability of selecting the use of force alternative, is:

Pr(I[A1, Az, A3) = V--2 + Va V1


S S V1 + V3
The noncompensatory nature of an aspect in this model implies that
alternatives that "own" that aspect (above a pre-determined threshold
level) "survive" the screening and the elimination by aspect process,
while all other alternatives are rejected. Remaining alternatives are
then screened based on a new ("substantive") aspect.
It is acknowledged that this version of the poliheuristic model be-
comes computational infeasible for large choice sets. Recognizing this
problem, Tversky and Sattath (1979) specialized the model to situations
in which the alternatives may be represented by a tree graph. Such a
case is analyzed in Mintz and Geva (1994), see also Figure 1.
This version of the poliheuristic model has obtained significant sup-
port in process tracing designs, in the analysis of several case studies
456 ALEX MINTZ, NEHEMIA GEVA, AND KARL DEROUEN JR.

Iraq's invasion of Kuwait [

Action Do nothing

/5--...
Use of
force
Economic
sanctions
Political
denunciation

Military Blockade
assault
Fig. 1. Tree structure of the model.

(e.g., Iraq, Grenada, Dien Bien Phu), and has also been subjected to
limited empirical testing.

CONCLUSION

In this paper we extended the Mintz (1993a), and Mintz and Geva
(1994) research on the poliheuristic model of foreign policy decision
making by specifying a mathematical model of the decision to use force.
The most important aspects of this model are (1) the nonexhaustive
and nonholistic characteristics of the process (i.e. its focus on a very
restrictive set of alternatives and dimensions), (2) its noncompensatory
nature (i.e. the fact that a low "score" on some variables, e.g. political,
cannot be compensated for by a high score on other dimensions), and
(3) the sequential nature of the decision. We offer this model as an
alternative to the expected utility and to the cybernetic theories of the
use of force (see Bueno de Mesquita and Lalman 1990, 1992; and
Ostrom and Job 1986). As was pointed out above, the core elements
and tenets of this theory were tested and received considerable support
using experimental and process tracing methodologies (such as a simul-
FOREIGN POLICY DECISION-MAKING 457

ated "decision board" or other AI designs). The results obtained in


these studies are nontrivial (see Mintz and Geva 1994, Chap. 8), as
they reveal a close correspondence between the theoretical model and
cognitive processes underlying foreign policy decisions.

NOTES

* The authors thank Raymond Dacey for his helpful comments.


1 This represents a shift from previous (1960s and 70s) preoccupation with the effects of
bureaucratic and organizational aspects of foreign policy decision-making.
2 The maximizing premises underlying the expected utility model call for a holistic process
in the search of the best alternative. The satisficing principle underlying the cybernetic
process implies that the inter-dimensional process can stop after a review of an alternative
if that alternative surpasses a satisficing threshold,
3 It should be pointed out that the Ostrom and Job (1986) cybernetic equation reflects
the 'current situation' in which the decision is being made, while the Bueno de Mesquita's
expected utility formulation looks 'forward' at expected consequences of the decision.
4 The above generalization should be considered with some caution. The cybernetic
model of the use of force of Ostrom and Job (1986) is a compensatory model, despite
the fact that the number of alternatives and dimensions is rather limited (e.g., two
alternatives that are considered along three dimensions). However, within this restricted
content domain the decisi0n-maker is perceived to operate via a compensatory process.
Outlining the probability of the use of force as a linear combination of three dimensions
(or decision environments), implies that a high score on one dimension (e.g., interna-
tional) can compensate for a low score on another dimensions (e.g., domestic or political),
since the decision-maker selects the alternative that surpasses a certain "satisfactory"
threshold level.

REFERENCES

Allison, G.: 1971, Essence of Decision: Explaining the Cuban Missile Crisis, Little,
Brown, Boston.
Axelrod, R.: 1976, 'The Cognitive Mapping Approach to Decision Making', in R. Axel-
rod (ed.), Structure of Decision, Princeton University Press, Princeton.
Beach, L. R.: 1990, Image Theory: Decision Making in Personal and Organizational
Contexts, John Wiley, London.
Bendor, J. and T. H. Hammond: 1992, 'Rethinking Allison's Models', American Political
Science Review 86, 301-22.
Billings, R. and S. Marcus: 1983, 'Measures of Compensatory and Noncompensatory
Models of Decision Behavior: Process Tracing Versus Policy Capturing', Organiza-
tional Behavior and Human Performance 31, 331-52.
Billings, R. and L. Scherer: 1988, 'The Effects of Response Mode and Importance on
Decision-Making Strategies: Judgment Versus Choice", Organizational Behavior and
Human Decision Processes 41, 1-19.
458 A L E X MINTZ~ N E H E M I A G E V A , AND K A R L D E R O U E N J R .

Brains, S. J.: 1.985, Superpower Games: Applying Game Theory to Superpower Conflict,
Yale University Press, New Haven.
Brannick, M. T. and J. P. Brannick: 1989, 'Nonlinear and Noncompensatory Processes
in Performance Evaluation", Organizational Behavior and Human Decision Processes
44, 97-122.
Braybrooke, D. and C. Lindblom: 1963, A Strategy of Decision, The Free Press of
Glencoe, New York.
Bueno de Mesquita, B.: 1981, The War Trap, Yale University Press, New Haven.
Bueno de Mesquita, B.: 1984, 'Forecasting Policy Decisions: An Expected Utility Ap-
proach to Post-Khomeini Iran', PS 17, 226-36.
Bueno de Mesquita, B. 1985, 'The War Trap Revisited', American Political Science
Review 79, 156-73.
Bueno de Mesquita, B. and D. Lalman: 1986, 'Reason and War', American Political
Science Review 80, 1113-28.
Bueno de Mesquita, B. and D. Lalman: 1990, 'Domestic Opposition and Foreign War',
American Political Science Review 84, 747-65.
Bueno de Mesquita, B. and D. Lalman: 1992, War and Reason, Yale University Press,
New Haven.
Dawes, R.: 1964, 'Social Selection Based on Multidimensional Criteria', Journal of
Abnormal and Social Psychology 68, 104-9.
Einhorn, H.: 1970, 'The Use of Nonlinear, Noncompensatory Models in Decision Mak-
ing', Psychological Bulletin 73, 221-30.
Einhorn, H.: 1971, 'The Use of Nonlinear, Noncompensatory Models as a Function of
Task and Amount of Information', Organizational Behavior and Human Performance
6, 1-27.
Einhorn, H. J. and R. M. Hogarth: 1981, 'Behavioral Decision Theory: Processes of
Judgment and Choice', Annual Review of Psychology 32, 53-88.
Ford, K. J., N. Sehmitt, S. L. Scheehtman, B. M. Halts and M. L. Doherty: 1989,
'Process Tracing Methods: Contributions, Problems and Neglected Research Ques-
tions', Organizational Behavior and Human Decision Behavior 43, 75-117.
James, P. and J. R. Oneal: 1991, 'Influences on the President's Use of Force', Journal
of Conflict Resolution 35, 307-32.
Kinder, D. R. and J. A. Weiss: 1978, 'In Lieu of Rationality: Psychological Perspectives
on Foreign Policy Decision Making', Journal of Conflict Resolution 22, 707-36.
Klein, G.: 1989, 'Recognition-Primed Decisions', in W. B. Rouse (ed.), Advances in
Man-Machine System Research, Vol. 5, JAI Press, Greenwich, CT, pp. 47-92.
Lebow, R.: 1981, Between Peace and War: The Nature of International Crisis, Johns
Hopkins University Press, Baltimore.
Maddala, G. S.: 1983, Limited-Dependent and Qualitative Variables in Econometrics,
Cambridge University Press, Cambridge.
Mintz, A.: 1993a, 'The Decision to Attack Iraq: A Noncompensatory Theory of Decision
Making', Journal of Conflict Resolution 37, 595-618.
Mintz, A.: 1993b, 'The Decision to Stop the War in the Gulf: A Study of Risk Aversion
in International Politics', Presented at the Annual Meeting of the American Political
Science Association, Washington, D.C.
F O R E I G N POLICY D E C I S I O N - M A K I N G 459

Mintz, A. and N. Geva: 1994, The Polyheuristic Theory of Decision: A Noncompensatory


Approach to Foreign Policy Decision Making, forthcoming.
Mintz, A. and B. Russett: 1992, 'The Dual Economy and Arab-Israeli Use of Force: A
Transnational System?', in S. Chan and A. Mintz (ed.), Defense, Welfare and Growth,
Routledge, London.
Morgan, C. and K. Bickers: 1992, 'Domestic Discontent and the External Use of Force',
Journal of Conflict Resolution 36, 25-52.
Olshavsky, R. W.: 1979, 'Task Complexity and Contingent Processing in Decision Mak-
ing', Organizational Behavior and Human Performance 24, 300-16.
Oneal, J. R. 1988, 'The Rationality of Decision Making during International Crises',
Polity 20, 598-622.
Oneal, J. R. and B. Lian: t992a, 'A Reexamination of the Domestic and International
Influences on the President's Use of Force', Paper presented at the Annual Meeting
of the American Political Science Association, Chicago, IL.
Oneal, J. R. and B. Lian: 1992b, 'Presidents, the Use of Military Force, and Public
Opinion', typescript, University of Alabama.
Ostrom, C. and B. Job: I986, 'The President and the Political Use of Force', American
Political Science Review 80, 541-66.
Payne, J., J. Bettman, and E. Johnson: 1988, 'Adaptive Strategy Selection in Decision
Making', Journal of Experimental Psychology 14, 534-52.
Ranyard, R. H.: 1976, 'Elimination by Aspects as a Decision Rule for Risky Choice',
Acta Psychologica 40, 299-310.
Russett, B. (1990), Controlling the Sword, Harvard University Press, Cambridge, MA.
Sage, A. P.: 1990, 'Human Judgment and Decision Rules', in A. P. Sage (ed.), Concise
Encyclopedia of Information Processing in Systems and Organizations, Pergamon Press,
New York.
Simon, H.: 1957, 'A Behavioral Model of Rational Choice', in H. Simon (ed.), Models
of Man: Social and Rational, John Wiley, New York.
Simon, H.: 1959, 'Theories of Decision-Making in Economics and Behavioral Science,
American Economic Review 49, 253-283.
Sniderman, P., R. Brody and P. Tetlock: 1991, Reasoning and Choice, Cambridge
University Press, Cambridge.
Snyder, G. and P. Diesing: 1977, Conflict Among Nations, Princeton University Press,
Princeton.
Steinbruner, J.: 1974, The Cybernetic Theory of Decision, Princeton University Press,
Princeton.
Stciner, M.: 1983, 'The Search for Order in a Disorderly World: Worldviews and Prescrip-
tive Decision Paradigms', International Organization 37, 373-413.
Sylvan, D. and S, Chart: (1984), 'Foreign Policy Decision Making: An Overview', in D.
Sylvan and S. Chan (eds.), Foreign Policy Decision Making, Praeger, New York.
Tversky, A.: 1972, 'Choice by Elimination', Journal of Mathematical Psychology 9, 341-
67.
Tversky, A. and S. Sattath: 1979, 'Preference Trees', Psychological Review 86, 542-73.
yon Neumann, J. and O. Morgenstern: 1947, Theory of Games and Economic Behavior,
2nd ed., Princeton University Press, Princeton.
460 A L E X MINTZ~ N E H E M I A G E V A , AND K A R L D E R O U E N J R .

Zinnes, D. A.: 1980, 'Why War? Evidence on the Outbreak of International Conflict',
in T. R. Gurr (ed.), Handbook of Political Conflict, Free Press, New York.
Zinnes, D. A.: 1976, Contemporary Research in International Relations, Free Press, New
York.

Program in Foreign Policy Decision Making


102 Bolton Hall
Texas A&M University
College Station
Texas 77843-4348
USA

View publication stats

You might also like