0% found this document useful (0 votes)
140 views22 pages1992 - Sebenius - Negotiation Analysis - A Characterization and Review
negosiasi
Uploaded by
sharkinCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
140 views22 pages1992 - Sebenius - Negotiation Analysis - A Characterization and Review
negosiasi
Uploaded by
sharkinCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
/ 22
5.26, 1992
60, Florida
spresentihe
‘tribution of
atineligost
‘ganizations
janizationa
2pinclude
sorsity
‘ua
foenia
rsiy
We stongly
>abstactol
damto ipm.
tin Orando,
ddwith your
Volume 38, Number 1 January 1992
‘Ting-peng Liang 1 A Composite Approach to Inducing
Knowledge for Expert Systems
Design
James K. Sebenius 18 Negotiation Analysis: A
Characterization and Review
Thomas R. Hoffmann 39 Eureka: A Hybrid System for
Assembly Line Balancing
Yigal Gerchak & J. David Fuller 48 — Optimal Value Declaration in “Buy-
Sell” Situations
G. Andrew Karolyi 57 Predicting Risk: Some New
Generalizations
S. Chan Choi, Wayne S. DeSarbo & 75 NOTES: A Numerical Approach to
Patrick T. Harker Deriving Long-Run Equilibrium
Solutions in Spatial Positioning
Models
‘Yu-Sheng Zheng 87 On Properties of Stochastic Inventory
Systems
Harry Groenevelt, Liliane Pintelon & 104 — Production Lot Sizing with Machine
Abraham Seidman Breakdowns
Scott Webster 124 New Bounds for the Identical Parallel
Processor Weighted Flow Time
Problem
Yamazaki, 137 On Optimal Arrangement of Stations
Hirotaka Sakasegawa in a Tandem Queueing System with
& J. George Shanthikumar Blocking
About Authors 154
‘Apia ua nse Tenia nan nar 9 ene Se oes Rs din np te Ube pen -19 a, nnn
Sts po al rene nd sofas Sn 0 Set Rn ie Spe TH FOTN Se
‘OE. Lavy Cones Cg Cav No 36:10 Sn Can Panag pa Pc, Re hand ae in oe . Pineda US
NEGOTIATION ANALYSIS: A CHARACTERIZATION
AND REVIEW*
JAMES K. SEBENIUS.
John F. Kennedy School of Government, Harvard University, Cambridge, Massachusetts 02138
“Negotiation analysis" seeks to develop prescriptive theory and useful advice for negotiators
and third partis. It generally emphasizes the parties’ underlying interests (as distinct from the
issues onthe table and the positions taken). alternatives to negotiated agreement, approaches to
productively manage the inerent tension between competitive actions aim" value individually
and cooperative ones to “create” value jointly, as well as efforts to change perceptions of the
‘ame itself. Since advice to one side does not necessarily presume the Full game-theoretc rationality
ofthe other side(s), negotiation analysts often draw on the findings of behavioral decision analysts
and economists. Further, this approach does not generally assume that al the elements of the
"game" are common knowledge. Thus, the negotiation analytic approach tends to de-emphasize
‘the application of ame-theoretc solution concepts o effort to find unique equilibrium outcomes,
Instead, to evaluate possible strategies and tactics, negotiation analysts generally focus on changes
in perceptions of the “zone of possible agreement” and the (subjective) dstibution of possible
regoiated outcomes conditional on various actions. This approach is especially sensitive t0 po-
‘entially unrealized joint gains. 1t has been used to develop prescriptive advice for the simplest
bilateral negotiations between monolithic parties, for negotiations through agents or with inked
“internal” and “external” aspects, for negotiations in hierarchies and networks, as well as for
‘more complex coalitional interactions,
(NEGOTIATION ANALYSIS; GAME THEORY: EQUILIBRIUM CONCEPTS; COMMON
KNOWLEDGE; BEHAVIORAL DECISION ANALYSIS; NEGOTIATION; BARGAINING)
With its primary roots in decision analysis and game theory, “negotiation analysis”
seeks to develop prescriptive theory and useful advice for negotiators and third parties.
In this article, 1 will describe what seem to be distinctive characteristics of this emerging
approach, discuss a number of representative works, sketch the elements of a typical
negotiation analysis, and highlight some promising trends.
‘One might quite reasonably ask about the need for a new approach since the theory
of games already provides a logically consistent framework for analyzing negotiating
situations. Given standard rationality axioms, the interests of the involved parties can
be abstracted into utility functions. The implied expected utility criterion ranks alternative
courses of action both with and without negotiated agreement. Full descriptions of the
courses of action that can be taken by each involved party are encapsulated into “strat-
egies.” Rigorous analysis of the interaction of strategies leads to a search for “equilibria,”
or plans of action such that each party, given the choices of the other parties, has no
incentive to change its plans. And, “{fJor forty years, game theory has searched for the
grand solution,” that would achieve “a prediction regarding the outcome of interaction
among human beings, using only data on the order of events, combined with a description
of the players’ preferences over the feasible outcomes of the situation.”
Game theory has been especially useful for understanding repeated negotiations in
well-structured situations such as various financial markets, It has offered useful guidance
for the design of negotiation and bidding mechanisms, has uncovered some powerful
* Accepted by Irving H. LaValle; received January 24, 1990. This paper has been withthe author 4$ months
for 2 revisions.
" Rubenstein (1991, p. 923). See the lasic game-theoretic works of Von Neumann and Morgenstern (1944)
and Luce and Rafa (1957); for recent insightful assessments, with special regard to bargaining, sce Roth
(1985), Aumann (1989), Harsanyi (1989), and Rasmusen (1989),
1B
(0025-1909 /92/380 /0018801.25,
Con © 192, The nse of Mangement Sines
comp
intrig
the ec
ofan:
for ui
Siebe
Ye
the st
usefir
cone
choo:
the o
rende
and“
and e
Despi
limits
where
insigh
the si
chara
appre
A
years
appre
taken
Betwi
that 1
“psye
tensic
woul
to gar
ratior
cision
from
this a
At
emery
Gam
devel:
Raiffe
intera
of wh:
Schell
works
had
25ce
(1983) 22138
alysis”
arties.
orging
ypical
heory
iating
can,
native
of the
action
iption
ons in
dance
werful
months
ca9aay
‘= Roth
$01.25
Senet
NEGOTIATION ANALYSIS 19
competitive dynamics, has usefully analyzed many “fairness” principles, has generated
intriguing hypotheses, and now flourishes both on its own and in applications such as
the economics of industrial organization. With nonspecialist audiences in mind, a number
of analysts have recently described some of the most useful contributions of game theory
for understanding negotiating behavior. (See, for example, Weber 1985, Myerson 1991,
Siebe 1991, and Young 1991.)
Yet the dominant game-theoretic quest to predict equilibrium outcomes resulting from
the strategic interactions of fully rational players often suffers from a lack of prescriptive
usefulness. First, on standard assumptions, there are often numerous plausible equilibrium
concepts, each with many associated equilibria—and no a priori compelling way to
choose among them. Second, one’s client, for example, may wish to act rationally, but
the other side may not behave as a strategically sophisticated, utility-maximizer—thus
rendering conventional equilibrium analyses inapplicable. Third, the elements, structures,
and “rules” of many negotiating situations are not completely known to all the players,
and even the character of what is known by one player may not be known by another.
Despite some ingenious theorizing, the frequent lack of such “common knowledge”
limits—from a prescriptive standpoint—much equilibrium-oriented game analysis. Even
where it is possible to bend such a situation into the form of a well-posed game, and gain
insights from it, the result may have lost considerable prescriptive relevance. Rather than
the simplification that is almost always necessary for analytic tractability, the essential
character of the problem may have been changed to fit the game theorist’s exacting
approach.
A spirited and informative interchange in the pages of Management Science a few
years ago debated the explanatory, predictive, and prescriptive merits of the game-theoretic
approach.’ The principal analytic alternative for addressing negotiation situations was
taken to be decision theory; one author sharpened the question in his ttle to “On Choosing.
Between Rev. Bayes and Prof. von Neumann” (Kahan 1983). A few commentators felt
that to abandon the fully rational approach was to be cast into the murky realm of
“psychology.” Largely since the time of that discussion, and reflective of many of the
tensions it illuminated, another analytic approach to negotiation has been emerging. It
would be an unfortunate mistake to characterize this new approach as an “alternative”
to game theory; indeed, it might be called “nonequilibrium game theory with bounded
rationality and without common knowledge” (although other candidates might be “de-
cision analysis with a strong interactive flavor,” or “strategically sophisticated psychology
from a prescriptive point of view”). Adopting the label “negotiation analysis” instead,
this article first seeks to characterize this approach more precisely.
A, Distinctive Characteristics of a Negotiation Analytic Approach
At the cost of some game-theoretic precision and formal rigor, a body of work has
‘emerged that offers prescriptive insights into negotiation. While Luce and Raifla’s (1957)
Games and Decisions was primarily a brilliant synthesis and exposition of game theory's
development since von Neumann and Morgenstern’s (1944) classic work, Luce and
Raiffa began to raise questions about the limits of this approach in analyzing actual
interactive conflict situations. Perhaps the first work that could be said to be in the spirit
‘of what I mean by “negotiation analysis” was The Strategy of Conflict (1960) by Thomas
Schelling, followed by his Arms and Influence (1966). The point of departure of these
works was game-theoretic but they proceeded with less formal argument and their analysis.
had far broader direct scope. Though nominally in the behavioral realm, Walton and
2 Soe, e4, Kadane and Larkey (1982a, b, 1983), Harsanyi (1982a, b), Kahan (1983), Roth and Schoumaker
(1983), Rothkopt (1983), and Shubik (1983), 20 JAMES K, SEBENII
McKersie’s Behavioral Theory of Labor Negotiation (1965) drew on Schelling’s work as
well as rudimentary decision and game theories in highlighting distinctions between so-
called “integrative” and “distributive” bargaining as well as the “intraorganizational”
negotiations that take place in tandem with the bargaining between labor and man-
agement.
‘The first overall synthesis of the emerging field of negotiation analysis appeared with
the publication of Howard Raiffa's (1982) The Art and Science of Negotiation. An ap-
plication and elaboration of some of these ideas in the context of the mammoth Law of
the Sea negotiations, Negotiating the Law of the Sea: Lessons in the Art and Science of
Reaching Agreement, was published by Sebenius (1984). A set of essays, structured
simulations, and teaching notes commissioned by the National Institute for Dispute
Resolution resulted in publication of The Manager as Negotiator and Dispute Resolver,
by Lax, Samuelson, Sebenius, Weber, and Weeks (1985). This approach was systematized
into an overall method in the first part of Lax and Sebenius’ The Manager as Negotiator
(1986); the second part specialized the method to managerial negotiations within and
among organizations. More recently, Negotiation Analysis (forthcoming, 1991) edited
by H. Peyton Young, furthers this evolving tradition.?
Though negotiation analysis is inspired by the categories, concepts, and techniques
developed in the field of game theory, itis generally distinctive in a number of respects—
that, as I will explain, bring it far closer in spirit to decision analysis (i.c., decomposing
the problem: separating and subjectively assessing probabilities, values, attitudes toward
risk, and time preference; structuring and sequencing the parties’ choices and the chance
events). Though not uniformly present, four features seem to capture much of the spirit
of this emerging approach: an “asymmetrically prescriptive /descriptive” orientation, a
radically subjective perspective, a sensitivity to potential ‘value left on the table,” and a
de-emphasis on game-theoretic solution concepts or equilibria supplanted by a focus on
changes in perceptions of the “zone of possible agreement”—including analysis of moves
to change the perceived “game” itself. I will discuss these before outlining the elements
of a canonical negotiation analysis.
An Asymmetrically Prescriptive / Descriptive Orientation
Unlike the “symmetrically prescriptive” approach of game theory, wherein fully rational
players are analyzed in terms of what each should optimally do given the other’s optimal
choices, a negotiation analyst typically seeks to generate prescriptive advice to one party
given a (probabilistic) description of how others will behave; this is in line with the
decision analytic approach (e.g., Raiffa 1968). In developing prescriptions for one side,
negotiation analysts typically assume intelligent, goal-seeking action by the other parties,
but not full game-theoretic (interactive) rationality. Such descriptive assessments of the
others need not assume tactical naiveté; as contextually appropriate, the assessments can
incorporate none, a few, or many, rounds of “interactive reasoning.” An important
additional perspective is what Raiffa calls “externally prescriptive /descriptive,” a stance
appropriate to advising third parties such as mediators and arbitrators about how best
to act, given assessments of the protagonists.
Concern with the “other side” renders the insightful work of “behavioral decision
analysis” and behavioral economists important to negotiation analysis. Knowledge of
systematic cognitive deviations from strict individual “rationality,” poorly-calibrated and
> A numberof other works—sometimes appearing in The Journal of Confit Resolution and The Negotiation
Journal —focusing on characteristic aspects ofthis emerging approach will be discussed below.
“See, eg, Bazerman (forthcoming), Neale and Bazerman (1991), Einhorn and Hogarth (1988) along with
the other excellent collections of papers in Kahneman, Slovi, and Tversky (1982), the review in Schoemaker
(1983), and Bel, Raiffa, and Tversky (1988).
incor
tactic
thea
are
Ni
up t
spiri
than
perc:
mut:
tions
such
Thir
prec
asses
nege
For
coor
opti
with
class
walk
the!
help
ofth
anal
“zor
that
alter
part
agre
’s
(a98:
descr
coop
negot
Pare
when
dem
inba ork as
ven so-
ional”
man-
d with
\nap-
aw of
nce of
cured
‘ispute
solver,
vatized
stiator
in and
edited
aiques
ects—
dosing
oward
hance
= spirit
ion, a
anda
cus on
moves
ments
tional
otimal
party
th the
eside,
arties,
of the
itscan
ortant
stance
wy best
cision
age of
dand
vation
re with
maker
NEGOTIATION ANALYSIS 21
inconsistent probabilistic assessments, as well as other anomalies not only has direct
tactical implications but also helps build up more structure on the “descriptive” side of
the area’s “asymmetrically prescriptive /descriptive” orientation. Of value in this regard
are the works of Roth and his colleagues that blend game-theoretic and psychological
considerations in rigorous experimental settings.®
2. A Radically Subjective Perspective
[Negotiation analysis is radically subjective in three important senses. First, it is presumed
up to the parties how they assess the probabilities of different events. (This is closer in
spirit to Kadane and Larkey’s 1982a, b views on subjective probability and game theory
than the more classical view as expressed by Harsanyi 1982a, b.) Second, subjective
perceptions of the parties’ underlying interests are taken as sovereign (though not im-
mutable). Less tangible concerns for self-image, fairness, process, precedents, of rela-
tionships can have the same analytic standing as the “harder” or “objective” interests
such as cost, time, and quality that are common to traditional economic approaches.
|, one side is not bound to regard the “other side” as acting in accord with the
precepts of game-theoretic rationality; the other side's likely behavior must be subjectively
assessed in light of available evidence.
3. Sensitivity to “Value Left on the Table”
Much early game-theoretic and economic analysis of bargaining simply assumed that
negotiated agreements would be “efficient,” or would not leave joint gains on the table.
For example, the so-called Nash solution (1950), along with other like approaches to
cooperative games (i.e., in which binding commitments are possible), posited Pareto
optimality as a reasonable feature of a negotiated outcome. Despite a world peppered
with needless deadlocks, poor agreements, soured relationships, strikes, and wars, many
Classically trained economists (still) react with incredulity that “rational” bargainers might
walk away from unrealized joint gains. By contrast, most negotiation analysts anticipate
the likelihood of ex post Pareto-inefficient agreements, and devote considerable effort at
helping the parties to “expand the pie."*
4. A Focus Away from Equilibrium Analysis and Toward Perceptions
of the Zone of Possible Agreement
In contrast to a game-theoretic preoccupation with solution concepts and equilibrium
analysis, negotiation analysts typically focus on subjective perceptions of the so-called
“zone of possible agreement” —as well as how those perceptions change. They presume
that each party can at least roughly assess and reassess the attractiveness of its no-agreement
alternatives. The set of possible agreements that, from the standpoint of each involved
party, is better in value or utility terms than no agreement, comprises the zone of possible
agreement, Since each party would rather accept any settlement in the zone of possible
agreement rather than no agreement (assuming the process does not generate spite, conflict,
+ See, e8. Roth and Malouf (1979, 1981), Roth, Malouf, and Murnighan (1981), Roth and Murnighan
(1982), and Roth and Schoumaker (1983)
‘From the standpoint of economic logic, Schelling (1960) discusses common inefficiencies in quasiconstant
‘sum bargains with complete information. Lax and Sebenius (1986) discuss the “Negotitor’s Dilemma” and
describe the frequent ineficiencies resulting from the tension between negotiators’ desires to create value by
and to claim value by behaving competitively. Further, a number of authors have shown thatthe
‘negotiators’ rational selinteested behavior when bargaining with incomplete information can lead to ex post
Pareto-inffcient equilibria, For example, Chatterjee (1982, 1985) discusses the inefficiencies that can result
when parties possess private information, and Myerson (1979) along with Myerson and Satterthwaite (1983)
‘demonstrate the tradeoff between honest revelation of privately held information and ex post Paretofciency
in bargains with incomplete information. 22 JAMES K, SEBENIUS
escalation, or its equivalent), Schelling (1960) made the potent observation that the
outcome of such a situation could only be unraveled by a “logic of indeterminate situ-
ations.” In trying to develop such a logic, negotiation analysts tend to focus on actions
that can change perceptions of the zone, typically, in ways subjectively expected to yield
more favorable distributions of negotiated outcomes.
Let me offer four, somewhat related, reasons that many negotiation analysts may focus
less attention on the solution concepts and equilibria that characterize the bulk of relevant
game theory and mathematical economics—without abandoning or denying the validity
and likely value of such work
(a) Many Solution Concepts; Many, Many Possible Equilibria. Though there exist
many solution concepts and criteria, with various appealing properties and different
predictions, it is often neither a priori nor empirically clear which is the “best” or most.
applicable. For example, as central as the minimax criterion is to classical game theory
in purely competitive situations, if Player One believes that Player Two is not employing,
‘a minimax strategy, pethaps on the basis of experience or empirical studies, there are
often strategies superior to minimax for Player One (e.g., Kadane and Larkey 1982a).
More generally, Shubik (1982) and Mas-Colell (1989), for example, describe an im-
pressive number of candidate solution concepts for cooperative games.
Even with as powerful a concept as that of the Nash equilibrium in noncooperative
‘games (see, ¢.g., Kreps 1989), itis often impossible, even with the imposition of increas-
ingly stringent requirements or refinements, to limit a game's equilibrium outcomes to
unique or even small number of points.” Often there is an infinitude of such outcomes.
As Tirole (1988, p. 446) noted when explaining why “we are now endowed with nearly
a dozen refinements of perfect Bayesian equilibrium,” the “leeway in specifying off-the-
equilibrium-path beliefS usually creates some leeway in the choices of equilibrium actions;
by ruling out some potential equilibrium actions, one transforms other actions into equi-
librium actions. Hence it is not surprising that one often ends up with a continuum of
perfect Bayesian equilibria.”
Despite recent insights into how rational players might select from among multiple
Nash equilibria (Harsanyi and Selten 1988), the rationale for a particular choice may
ultimately seem arbitrary. As Kadane and Larkey (1982a, pp. 115-116) remark, “we do
not understand the search for solution concepts that do not depend on the beliefs of each
player about the others’ likely actions, and yet are so compelling that they become the
obvious standard of play for all those who encounter them.” This seems especially apt
in light of their observation that “solution concepts are a basis for particular prior dis-
tributions” and hence “the difficulty in non-zero sum, N-person game theory of finding,
aan adequate solution concept: no single prior distribution is likely to be adequate to all
players and all situations in such games.”
These considerations might suggest that, rather than discovering multiple equilibria
and imposing restrictions to winnow them down to single predicted outcome, negotiation
analysts should pethaps pause before winnowing and assess a (subjective) distribution
over the plausible equilibria. If this issue of “multiple concepts, multiple equilibria” were
the only question, however, equilibrium methods might still be the mainstay of those
interested in prescriptive approaches to negotiation. Yet, as I will next argue, important
premises for meaningful equilibrium analysis itself may not hold for prescriptive purposes.
(b) Significant Departures from Full Game-Theoretic “Rationality.” As discussed
(in the second reason) above, while people in mixed-motive situations normally exhibit
intelligent, purposive behavior, there are important departures from the “imaginary,
idealized, super-rational people without psyches” (to quote Bell, Raiffa, and Tversky
7 See, eg., sequential equilibrium (Kreps and Wilson 1982), perfection and subgame perfection (Selten
1975), stability (Kohlberg and Mertens 1986),
1988
alone
in 4(
analy
of ur
all,
Yet
ama
ratio
piri:
prog
and”
ysis
ic
edge
theor
funet
isme
devel
most
are"
infor
so or
In
price
to sh
migh
whic
with
perfe
and
com
in pr
of off
in of,
M
frequ
sibly
have
migh
withi
ploys
that |
these
accep
Furth
taint
equil
price
“In
pp. 3 hat the
te situ-
actions
to yield
wy focus
relevant
validity
re exist
lifferent
or most
theory
ploying
vere are
1982a).
verative
increas-
omes to
tcomes,
v nearly
off-the-
actions;
aultiple
ce may
“we do
ofeach
me the
ally apt
ior dis-
finding
te to all
vilibria
otiation
ibution
a" were
f those
portant
poses.
seussed
exhibit
inary,
(versky
1 (Selten
NEGOTIATION ANALYSIS 23
1988, p. 9) needed by von Neumann and Morgenstern expected utility maximizers, let
alone as required by sophisticated concepts of sequential strategic interaction alluded to
in 4(a). Of course, this need not detract from the utility of a fully rational “baseline”
analysis—to understand the possible responses of a rational other side—nor the value
of urging consistent, if not fully “rational,” behavior on the subject of one’s advice. After
all, well-structured, repeated negotiations may penalize departures from rational behavior.
Yet many negotiating situations are neither well-structured, repeated, nor embedded in
a market context. And much behavior more than trivially departs from the canons of
rationality. Recognizing this, researchers might hope that, after developing a better em-
pirical understanding of negotiators’ behavior, they might “redo” the game-theoretic
Program using a more behaviorally faithful theory (e.g., “prospect theory” of Kahneman
and Tversky 1979). Unfortunately, an even more fundamental problem with game anal-
ysis would carry over to an impede such a new quest.
(c) Often, the Game's Structure, Rules, and Possible Moves Are Not Common Knowl-
edge. Luce and Raiffa (1957, p. 49) noted that a fundamental requirement of game
theory is that “each player . . . is fully aware of the rules of the game and the utility
functions of the players. . .” and that “this isa serious idealization which only rarely
is met in actual situations.” While a great deal of theory, some of it ingenious, has been
developed for such games of “incomplete information” (¢.g., Harsanyi 1967-1968),
most of it rests on the assumption that the rules of the game and the utility functions
are “common knowledge” in the sense of Aumann (1976)—that each player knows this
information, knows that the others know it, that the others know that he knows it, and
so on.
Imagine that an analyst sought to give advice to one party in an extremely simple
price negotiation—in which a potential buyer and seller negotiate over time about how
to share the “surplus” over their reservation prices that would result from a sale. One
might naturally model this situation as Rubinstein’s (1982) simple bargaining game in
which two players, who must share a pie of size 1, make sequential alternating offers,
with a discounting factor, d, in operation. The result that there is a unique subgame
perfect equilibrium to this game (in which the offering player offers d/(1 + d))is elegant
and insightful, yet it depends absolutely on the full structure of this encounter being
common knowledge. If the players did not have common priors on any number of what
in practice would be tactically variable elements of the game as specified (eg., the sequence
of offers, on the length of time between offers, the mutually expected “rationality” involved
in offer and acceptance decisions), Rubinstein’s analysis would be inconclusive.
Moteover, while not a criticism of this spare model, one might expect a range of other
frequently encountered factors to be potentially decisive—and the players may not plau-
sibly have common knowledge of them. Take the role of expectations. One player might
have just read Schelling’s (1960) “Essay on Bargaining” and become convinced that she
might shape the other player's expectations by making an unchanging sequence of offers
within the zone of possible agreement. She might also use other contextually dependent
ploys to attempt a credible commitment such that the other player would come to believe
that he faced a “take it or leave it offer” that was better than no-agreement. Adding in
these expectational and/or commitment considerations, the result might be impasse or
acceptance of her fixed offer—even though these outcomes were not in equilibrium.®
Further, the essence of many simple distributive bargains involves each player's uncer
tainty about the other’s no-agreement alternatives, and hence, reservation price. Most
equilibrium results require the probability distributions from which these reservation
prices are drawn to be common knowledge of the players, In practice, that condition is,
In this connection, see the “thought experiment” and empirical results of Roth and Schouraker (1983,
pp. 1338-1339). 24 JAMES K, SEBENIUS
<-—— Player One's Minimum
Pareto
1080- Frontier
$
=
5 850.
s
a
650-| Player Two's
Minimum, |
450.
No-Agreement Score
250.
500 1000 1500 2000
Player One
FlGURE 1. Multiple Issue Negotiation,
2500 3000
often unmet, with each party potentially engaging in strategic manipulation of the other
side’s perceptions of no-agreement alternatives (as well as the subjective valuation of
those alternatives),
More generally, the full set of actual and potential players, interests, beliefs, issues,
alternatives to agreement, rules, and agreements, are often only imperfectly known, and
even the character of what is known by one party is not known by others. Indeed, purposive
action by involved or excluded parties can often change the set of involved players, bring
in or exclude issues, raise or lower the salience of different interests, alter the “rules” of
the interaction, or take other actions to change the collective perception of the “game's”
configuration.
Even where information is “almost common knowledge,” Rubenstein (1989) shows
that equilibrium results may be sharply different from the fully common knowledge
"On a personal note, in my own negotiation experience—which, apart from academics and consulting,
includes ime inthe U.S. Commerce and Stat Departments, membership for three years on the U.S, Delegation
to the United Nations Law of the Sea Negotiations, and four years of full-time merger and acquisition work on
‘Wall Steet—relatively few negotiating situations have conformed even approximately tothe exacting requirements
‘of common knowledge (including the common prior distributions of negotiator “types"—arrayed by reservation
Dries, utility functions, ete-—required by Harsanyi’s analysis). Some discussion of Harsanyis doctrine of
“common priors" can be found in Sebenis (1984, ep. Chapter 4) and Scbenius and Geanakoplos (1983)
situatio
commo,
Whater
consiste
itself mi
and the
nongan
of com:
(a)
conside
even wi
the neg
corpora
progran
situatio
tions. (
and mo
issues.)
results
reporte:
knowle:
fully in
suggest:
equilibr
questioy
Gointly
Thes.
along w
tional”
empiric
situatio
validate
static at
example
the gam
of negot
that see
In pa:
below,
for mov
such an
on the :
analogo
“increas
In the
uninfort
his subje
subjecti>
of factor
A sim 2other
tion of
issues,
‘mand
‘posive
. bring
les" of
ume’s’
shows
wledge
sulting,
legation
ork on
rements
wine of
983).
NEGOTIATION ANALYSIS 25
situation. More generally, as Aumann (1989, p. 31) unequivocally concluded, “The
common knowledge assumption underlies all of game theory and much of economic theory.
Whatever be the model under discussion, whether complete or incomplete information,
consistent or inconsistent, repeated or one-shot, cooperative or non-cooperative, the model
itself must be assumed common knowledge; otherwise the model is insufficiently specified,
and the analysis incoherent.” (emphasis supplied ) By contrast, as will be discussed below,
rnongame-theoretic analyses can yield useful advice in situations that lack some degree
of common knowledge.
(d) Widely Scattered Negotiated Outcomes in Practice. In line with the theoretical
considerations described above, negotiated outcomes in practice are extremely varied—
even where the underlying negotiating situation is identical. For example, Figure | plots
the negotiated results of pairs of private managers (including the heads of some large
corporations) and public officials (including a U.S. Senator) at two advanced executive
programs held at Harvard. Each side was given a confidential description of a realistic
situation with a private multiattribute value function explicitly embedded in the instruc-
tions. (The carefully structured exercise was originally constructed by John Hammond
and modified by Richard Zeckhauser to include an oil-for-aircraft swap and several side
issues.) After the participants enthusiastically carried out the simulation, the plotted
results were widely scattered.'° Despite the highly structured exercise, and the realism as
reported by the participants, this encounter had many elements that were not common
knowledge; and debriefings of the experienced participants suggest they were not acting
fully in accord with (strategic) rationality assumptions. This typical experimental outcome
suggests that, rather than further refine existing concepts in search of unique negotiated
equilibria in games of common knowledge, useful insights may also come from other
questions aimed at understanding the processes that lead to better and worse outcomes
Gointly and individually),
‘These four considerations—the number of plausible solution and equilibrium concepts
along with the multiplicity of equilibria in many games, the deviations from fully “ra-
tional” behavior, the frequent lack of common knowledge, and the widely scattered
empirical results—cast doubt on the ease with which the structure and rules of a given
situation can be reliably mapped onto a unique negotiated outcome. They do not in-
validate the game-theoretic program; afterall, the body of existing theory is anything but,
static and will certainly make significant progress on these and other challenges. (For
‘example, Nau 1990 has recently reported some interesting results on how the “rules of,
the game” become common knowledge.) Yet these four considerations have led a number
of negotiation analysts to de-emphasize equilibrium methods in favor of other approaches
that seem to yield useful prescriptive theory and advice.
In particular, the negotiation analytic approach generally highlights elements, described
below, that affect the parties’ perceptions of the zone of possible agreement, and searches
for moves that improve the subjective distribution of possible negotiated outcomes. By
such an “improvement,” I mean an increase in subjectively expected utility conditional
on the action and subsequent reactions. (Such “improvement” has a subjective basis,
analogous to the Rothschild-Stiglitz 1970 characterization of a subjectively perceived
“increase” in risk.)
In the skeptical view of Harsanyi (1982a), this approach might boil down to “the
uninformative statement that every player should maximize expected utility in terms of
his subjective probabilities without giving him the slightest hint of how to choose these
subjective probabilities in a rational manner.” Yet, as described below, distinct classes
of factors have been isolated that appear to improve distributions of negotiated outcomes.
°© 4 similar plot for another such exercise is reproduced in Raiffa (1982, p. 138) 26 JAMES K, SEBENIUS
Psychological considerations can help as can cultural observations, organizational con-
straints and patterns, historical similarity, knowledge of systematic decision-making biases,
and contextual features. Less than full-blown game-theoretic reasoning can offer insight
into strategic dynamics as can blends of psychological and game-theoretic analysis (e.e.,
models of “semirational” behavior; see Rothkopf 1983). When one relaxes the assumption
of strict strategic sophistication, Raiffa’s (1982, p. 359) conclusion is appealing: that
some “analysis—mostly simple analysis—can help.”
‘Measured against an “equilibrium” standard, a subjective assessment of the distribution
of possible negotiated agreements that only implicitly embodies the interactive elements
‘may seem analytically thin. Yet complementary standards—logical consistency, system-
atic insight into the negotiation process, or practical utility—may be more appropriate
for evaluating the work of negotiation analysts and suggesting new directions for inves-
tigation.
In short, distinctive features of a negotiation analytic approach include an asymmet-
rically prescriptive / descriptive analytical orientation; a radically subjective presumption;
the expectation of purposive action but not full, strategic rationality; sensitivity to Pareto-
inferior outcomes; a focus on changes in perceptions of the bargaining set rather than
on equilibrium and solution concepts; and a willingness to analyze situations that are
not fully specified in advance. Depending on one’s starting point, this might be “decision
analysis plus” or “game theory minus.” But to flesh out the characterization, it is useful
to review the elements of a typical negotiation analysis.
B. Elements of a Negotiation Analytic Approach
Full negotiation analytic accounts generally consider the following basic elements with
respect to the actual and potential parties: their perceived interests; alternatives to ne~
gotiated agreement; the linked processes of “creating” and “claiming” value; and efforts
to “change the game” itself. These basic elements can be found and analyzed in the
simplest bilateral negotiation between monolithic parties as well as in the most complex
coalitional interactions.
Interests, Issues and Positions
It is often important to distinguish parties’ underlying interests from the issues under
negotiation, on which positions or stands are taken, The connection among positions
on issues and interests is rarely a simple one. Sometimes a focus on deeper interests can
unblock a stubborn impasse over incompatible positions that relate only partially to the
parties’ real concerns; in other cases, emphasizing interests will only generate hopeless
conflict when mutually beneficial agreement on certain overt issues could be reached.
And focusing on issues and positions (rather than interests) can be tactically advantageous.
In virtually all cases, an important first analytic step is to probe deeply for interests,
distinguish them from issues and positions, and to carefully assess tradeoffs. Raiffa (1982)
offers an extended discussion of assessing tradeoffs in negotiation, building on extensive
work by Keeney and Raiffa (1976). (See also Keeney 1988 and Keeney and Raiffa
forthcoming, 1991.) Barclay and Peterson (1976) demonstrate a methodology for assessing
negotiating tradeoffs in the context of a base negotiation. Lax and Sebenius (1986) offer
a simplified discussion of the principles behind such tradeoffs, while Wierzbicki (1983)
critically surveys the methodologies of multiobjective analysis. (See also Saaty's 1980,
1990 analytic hierarchy method.) A book by Nagel and Mills (forthcoming), Multicriteria
Methods in Alternative Dispute Resolution, develops technological and computer-based
means for making such assessments in a variety of negotiating settings.
When individuals or groups with different concerns constitute a negotiating “side,” it
is no longer in general possible to specify overall tradeoffs; however, carefully tracing
which
faction
formal
analysi
shared
such ¢
“value
decisio
2. alt
Peo}
jointly
propo:
course
party's
gotiate
Alte
price g
to cou
anxieti
nature
interp:
ante n
nation
throug
may b
Outrig
comm
in arm
benefit
contin
Evic
often 1
for ex:
discov
ciently
altern:
the dy
analys
Raiffa
Wh
“reser
play t
course
risky,
fewer
the ne
reach,
strong
cultive
prepar
influer nal con-
biases,
r insight
sis (ee.
amption
ng: that
Fibution
ements
system-
‘ropriate
or inves-
ymmet-
mption;
»Pareto-
ver than
that are
decision
is useful
nts with
sto ne-
d efforts
din the
somplex
sunder
vositions
ests can
ly to the
hopeless
reached.
tageous.
aterests,
1 (1982)
xtensive
4 Raiffa
assessing
46) offer
(1983)
's 1980,
‘ieriteria
or-based
side,
tracing
NEGOTIATION ANALYSIS a
which set of interests is ascendant according to the internal bargaining process of given
factions may continue to provide insights, (Wilson’s 1968 work on “syndicates” suggests
formal conditions under which a “group utility function” exists.) One result of such
analysis of interests may be the disaggregation of a side into factions whose interests are
shared enough to justify treating the faction as another distinct party. For cases in which
such disaggregation is not sensible, Keeney, Renn, and von Winterfeldt (1983) discuss
“value tree” analysis, whereby effective preferences of larger groups can be assessed for
decision-making purposes, including policy negotiations.
2. Altematives to Negotiated Agreement
People negotiate in order to satisfy the totality of their interests better through some
jointly decided action than they could otherwise. Thus, for each side the basic test of a
proposed joint agreement is whether it offers higher subjective worth than that side’s best
course of action absent agreement. In examining a negotiation, one should analyze each
party’s perceptions of its own—and the others’ evaluations of their—alternatives to ne-
gotiated agreement.
Alternatives to agreement may be certain, with a single attribute: an ironclad competing
price quote for an identical new car. They may be contingent and multi-attributed: going
to court rather than accepting a negotiated settlement can involve uncertainties, trial
anxieties, costs, time, and precedents that contrast with the certain, solely monetary
nature of a pretrial accord. Alternatives may change over time with new information,
interpretations, competitive moves, or opportunities. Without agreement, the status quo
ante may be superseded by something much worse for one side: a now-neutral island
nation may intend to lease its naval base to one superpower if current negotiations fall
through with the other. In multi-party negotiations, one side’s alternatives to agreement
may be the set of agreements that could be reached by potential opposing coalitions.
Outright threats by one party to change the no-agreement alternatives of another are
‘common. Or, the best alternative to negotiated agreement may be to keep negotiating:
in arms control, for example, failure to agree may involve worse relations, foregone
benefit and altered settlement possibilities, but in any case necessity may remain for
continued dealings among the same partes.
Evidently, decision analysis (including multi-attribute value and utility theory) can
often help assess alternatives to agreement. When there are many possible alternatives—
for example, many potential purchasers, cach with associated uncertainties and costs of
discovery for the seller—optimal search theory can provide strategies for searching eff-
ciently and valuing the expected findings from the search (Lax 1985). When the parties’
alternatives to agreement are interdependent, concepts from game theory—including
the dynamics of threats and counterthreats as well as the many variants of coalitional
analysis—can help bargainers understand their alternatives (Luce and Raiffa 1957;
Raiffa 1982).
While this evaluation provides a strict lower bound for the minimum worth (the
“reservation price”) required of any acceptable settlement, alternatives to agreement also
play tactical roles. The more favorably that negotiators portray their best alternative
course of action—whether this means a course that is less costly, more efficient, less
risky, with earlier benefits, with more desirable linked attributes (such as reputation), or
fewer undesirable ones (such as bad precedents) —the smaller is the ostensible need for
the negotiation and the higher the standard of value that any proposed accord must
reach. Moves “away from the table” that shape the parties’ alternatives to agreement can
strongly affect negotiated outcomes. Searching for a better price or another supplier,
cultivating a friendly merger partner in response to hostile takeover negotiations, or
preparing an invasion should talks fail to yield a preferable outcome may have greater
influence on the negotiated outcome than sophisticated tactics employed “at the table” |
|
|
28 JAMES K, SEBENIUS
such as clever opening offers or patterns of concession." This poses an interesting prob-
lem of allocating scarce effort at the table versus away from the table (Lax and
Sebenius 1985)
From Structure to Outcome: Favorable Changes in the Perceived Zone of Possible
Agreement. With the elements of the negotiation in place, but without an explicit model
or formal theory (equilibrium-based or other) adequate to map structure and tactics
‘onto bargaining outcomes, how can an individual negotiator or interested third party
decide what to do? In the (often implicit) view of many negotiation analysts, the nego-
tiator’s subjective distribution of belief about the negotiated outcome conditional on
using the proposed tactics must be compared with his subjective distribution of beliefs
about the outcome conditional on not using them. The tactic is attractive if the former
distribution gives him higher expected utility than the latter.'? Specifying these distri-
butions may require an internalized and subjective model of the bargaining process since
no such general model exists; where there is a well-developed and applicable game-theoretic
‘model, of course, it can be used. Of course, the “better” the empirical and theoretical
basis for the assessment, the “better” the subjective distributions of outcome:
‘Thus far, the discussion has proceeded in a decision analytic spirit. Yet, negotiation
problems have a special structure and dynamics that derive from the joint decision-
‘making that is inherently involved. A clear understanding of this structure and its char-
acteristic processes can improve the basis for assessing outcome distributions. While
game theory emphasizes the fully rational, strategic character of the interaction, nego-
tiation analysis systematically blends a sustained focus on what might be called the “tech-
nology of cooperation” with a more boundedly rational view of the strategic, adversarial
elements, as described in the next sections.
3. Creating and Claiming Value
‘The lure of joint action lies in the prospect of each party’s doing better than its alter-
natives to agreement. It is therefore crucial to understand the bases for joint gains and
to envision possible agreements. In most negotiations, the potential value of joint action.
is not fully obvious at the outset.
Creating Value. “Creating value”—that is, reaching mutually beneficial agreements,
improving them jointly, and preventing conflict escalation—requires an approach often
associated with “win-win,” “integrative,” or “variable sum” encounters. To generate
gainful options, itis normally helpful for information to be shared openly, communication
enhanced, creativity spurred, joint problem-solving emphasized, and hostilities produc-
tively channeled. Many analysts offer insights into “creating value” by cooperative be-
havior." Regardless of whether one adopts a cooperative style or not, it is useful to have
an analytic guide as to the underlying bases for joint gains. Three distinct classes of factors
In many settings, visibly improving one's no-agreement alternatives will improve one’s outcomes; varied
‘examples and laboratory experiments provide empirical support for this intuition. Yet, there are instances in
which this will nt be the case. For example, in a marriage, finding a potential replacement mate during a
dispute might seriously damage how one's spouse values the marriage. Or, a disgruntled subordinate whose boss
is known to value loyalty might be better off if he did not generate other offers before going to his boss 10
improve his ob; the breach of loyalty could reduce the boss sense of obligation to keep his employee happy.
Because no general model exist of how such tactics affect the distribution of outcomes, assessment depends on
‘experience, limited empirical evidence, equilibrium calculations, and logic.
That the changed distribution of outcomes gives higher expected utility than the original still need not
mean that the negotiator will get a better outcome than he would have by eschewing the tactics, A random
