外研社阅读大赛ORGANIZATIONAL BEHAVIOR AND HUMAN DECISION PROCESSES
Vol.66,No.2,May,pp.131–152,1996
ARTICLE NO.0044
When Time Is Money:Decision Behavior under
Opportunity-Cost Time Pressure
J OHN W.P AYNE AND J AMES R.B ETTMAN
Fuqua School of Business,Duke University
AND
M ARY F RANCES L UCE
Wharton School,University of Pennsylvania
regulatory rules(Einhardt,1989)and often involve Decison-making dilemmas can ari becau error
s stress due to the need to make choices under time pres-may result either from deciding too soon or from de-sure.However,such time stress is often due not so laying decisions too long.Delay can result in lost op-much to strict deadlines as it is to the potential oppor-portunities or reductions in payoffs from the most ac-tunity cost of delaying decisions.As Einhardt(1993, curate decision.This paper investigates decision pro-
p.121)notes,‘‘the decision-making dilemma in such cess in environments where there is time stress due
environments comes from the fact that it is easy to to the opportunity cost of delaying decisions.First,
make mistakes by deciding too soon and equally inef-using computer simulation,the relative accuracy of
fective to delay choices or to imitate others.’’In particu-alternative decision strategies is examined in environ-
lar,delay in deciding can result‘‘in win-ments that differ in terms of the levels of opportunity
cost of delay.The lexicographic choice rule is shown dows of opportunity clo’’(Einhardt,1993,p.1
21). to be a very attractive decision process in situations In some cas,the longer the delay in making the deci-where there is opportunity-cost time pressure.Two ex-sion,the lower the expected return(value)from even periments test the adaptivity of actual decision behav-the most accurate of decisions.For example,if a com-ior to the prence or abnce of opportunity-cost time pany delays a new product introduction to get test mar-pressure along with variations in goals(accuracy em-ket results,competitors may obrve the test,develop
phasized vs.effort savings emphasized),dispersion in
their own versions of the product,and introduce the probabilities or weights across the outcomes of the
versions before thefirst companyfinishes test market-choice options,and the degree of correlation among
ing(Ono,1995).More generally,the company’s pros-the outcomes.Subjects were generally adaptive to op-
香港留学费用pects may continuously degrade with delay;the longer portunity-cost time pressure.However,failures in
the delay,the more effective competitors’respons adaptivity were identified when choice environment
may be.
properties with conflicting implications for adapta-
tion were prent simultaneously.In particular,under We investigate decision making under situations opportunity-cost time pressure,subjects received a where there is opportunity-cost time pressure.We are lower expected payoff when the goal was to emphasize interested in how people adapt their decision-making choice accuracy than when the goal was to emphasize strategies(Payne,Bettman,&Johnson,1993)when savings in effort.The question of when adaptivity in faced with opportunity-cost environments and in the decision making might fail is discusd.᭧1996Academic possible limits or even failures in adaptivity that may Press,Inc.ari in such environments.In particular,we focus on
how people decide when there are conflicts between the INTRODUCTION processing implications of opportunity costs and the
implications of other variables such as goals and the ‘‘High-velocity’’environments are characterized by
degree of conflict(intercorrelation)between outcomes. rapid changes in technology,demand,competitors,or
Next,we review prior rearch on decision making
under time pressure.Then we summarize an accuracy–This rearch was supported by a grant from the Decision,Risk,
effort framework for adaptive decision behavior and and Management Science Program of the National Science Founda-
report a simulation study of decision strategies bad tion.Address reprint requests to John W.Payne at the Fuqua School
of Business,Duke University,Box90120, that framework.Using the simulation results and
1310749-5978/96$18.00
Copyright᭧1996by Academic Press,Inc.
All rights of reproduction in any form rerved.
132PAYNE,BETTMAN,AND LUCE
prior rearch,we develop a t of hypothes regard-strategies,e Payne et al.,1993,or Svenson,1979). ing adaptivity(and possible failures in adaptivity)to
The strategy a decision maker us is highly contingent opportunity-cost time pressure and report two experi-upon a host of task and context factors,such as the ments that test the hypothes.Finally,we discuss
number of alternatives available,information format, the implications of the results for adaptive decision respon mode,and the correlations among attributes making.
(Einhorn&Hogarth,1981;Hammond,1990;Payne et
al.,1993).1get over
TIME PRESSURE AND DECISIONS Given a t of available strategies and facing various
task demands,how do people decide how to decide?One Rearch suggests three major ways in which people framework for explaining contingent decision behavior respond to decision problems under time pressure.focus on the accuracy and cognitive effort character-First,people accelerate their ,spend less izing the available strategies(Payne et al.,1993;for time processing each item of information;e Ben related ideas,e Beach&Mitchell,1978,and Bocken-Zur&Breznitz,1981).Second,processing tends to be holt,Albert,Aschenbrenner,&Schmalhofer,1991). more lective under time stress,focusing on the more Cognitive effort has been measured by the number of important and/or negative information about alterna-elementary information processing ,ac-tives(Ben Zur&Breznitz,1981;Wallsten,1993).quiring an item of information,comparing two pieces Third,decision strategies may shift as a function of of information,or multiplying one value by another) incread time pressure(Payne,Bettman,&Johnson,needed to make a decision using a specified decision 1988;Svenson,Edland,&Slovic,1990).rule(Bettman,Johnson,&Payne,1990).Accuracy has There may be a hierarchy of respons to time pres-been measured in a variety of ways,including a com-sure.Under moderate time pressure,Payne et al.parison of the expected payoff when using a choice heu-(1988)report that their subjects accelerated processing ristic with the expected payoff from using a more nor-and procesd somewhat lectively.Under more -mative rule like weighted additive value or expected vere time pressure,people accelerated their processing,value maximization.
were even more lective,and changed strategies from The basic hypothesis of the effort/accuracy frame-a more depth-first(alternative-bad)to a more work is that the strategy ud to decide reprents an breadth-first(attribute-bad)pattern of processing as adaptive balancing of the goals of being as accurate as time pressure incread.As discusd in the next c-possible while conrving limited cognitive resources. tion,this shift in processing pattern is consistent with Although people are often adaptive decision makers, an accuracy/effort framework for strategy lection failures in adaptivity exist.For example,Kleinmuntz (Payne et al.,1993).and Schkade(1993)note that the choice of a strategy The studies cited above all ud deadlines to impo for decision making will be bad on subjective ex ante time ,a time limit of15s per choice in Payne estimates of relative effort and accuracy and the esti-et al.,1988).With very few ,Kerstholt,mates will not always be veridical.Adaptivity may also 1994),the terms‘‘time pressure’’and‘‘time constraint’’fail in situations characterized by multiple task or con-are ud interchangably(Svenson&Maule,1993).text variables with differing implications for adaptive Deadlines can make the u of some normative strate-processing.
gies impossible due to limits on how much information A number of studies have validated the effort and can be procesd in a given time,thus limiting the accuracy model of strategy lection.Creyer,B
ettman, range of feasible approaches for adapting to time pres-and Payne(1990),for example,found that subjects ac-sure(Keinan,1987).Becau of our interest in studying quired more information,took more time,were less adaptivity,we focus on situations with opportunity-lective in the processing of information,were more cost time pressure,where there is time stress but a alternative-bad,and were more accurate when the broad range of strategies is al was to maximize accuracy rather than to minimize
THE ADAPTIVE DECISION MAKER:AN EFFORT/
ACCURACY FRAMEWORK1Task effects describe factors associated with the general structure
of the decision problem,including time pressure.Context effects refer
to factors associated with the particular values of the options in the People often have a variety of information processing
specific decision t under consideration,such as the similarity strategies for dealing with judgment and choice tasks,
among options.In general,task factors are more likely to be noticed ranging from careful and reasone
d examination of al-
a priori or early in the decision.Context factors,on the other hand, ternatives to simple and fast rules of thumb(heuristics)are often noticeable only after some information has already been (for an overview of the properties of various decision
procesd.
133
DECISIONS UNDER OPPORTUNITY-COST TIME PRESSURE
effort(e also Stone&Schkade,1994).Thus,there
was a shift in strategies in the direction that one would
predict using the effort/accuracy framework.
Stronger tests of the effort/accuracy framework have
been conducted by using computer simulations of the
effort and accuracy characterizing typical decision
strategies to generate hypothes about processing pat-
terns and then carrying out experimental studies of FIG.1.Sample decision problem.
actual decision behavior to test the hypothes.Of
particular relevance to the prent studies,the simula-strategies:when the correlations between outcomes tion model of Payne et al.(1988)showed that certain were negative,implying greater conflict,people in-heuristics like the lexicographic rule and elimination-cread their processing and that processing was more by-aspects are actually more accurate than more nor-alternative-bad.Im
portantly,tho subjects who mative procedures like additive utility under vere adapted more to different correlation levels performed time deadlines(becau there is not enough time to better in the n of lecting options with higher fully complete the more complex adding rule).The pre-values.
ferred strategies under time constraints generally in-We wish to examine how a decision maker might volved processing at least some information about all change his or her strategy for processing information the alternatives as soon as possible(breadth-first or when faced with an opportunity-cost decision environ-attribute-bad strategies).The simulation model also ment.Although the above rearch provides some guid-showed that heuristics involving attribute-bad pro-ance about expected strategy changes,we u computer cessing were relatively more accurate when the disper-simulation to investigate the effects of opportunity-cost sion in probabilities was high rather than low.2In ex-time pressure in more detail.In particular,we examine periments with human subjects,people adapted their the relative accuracy of various decision strategies and decision process in ways consistent with the simula-how opportunity-cost time pressure may interact with tion results(Payne et al.,1988),shifting decision strat-different context variables that have been previously egies in the direction of more lective and more attri-shown to impact the acc
phpinarrayuracy of decision strategies bute-bad processing under vere time pressure.(i.e.,dispersion in weights and interattribute correla-Interestingly,Einhardt(1989)found that manage-tion).The simulation reprents an evaluation of differ-ment teams that simultaneously procesd two or more ent strategies in various task environments by an ide-alternatives in a breadth-first fashion made quicker alized adaptive decision maker.We then u the results and better decisions in‘‘high-velocity’’environments of the simulation and previous rearch to generate than tho teams that procesd alternatives using hypothes for our experiments.
more quential,depth-first(alternative-bad)consid-
A SIMULATION INVESTIGATION OF eration of one alternative at a time(e also Judge&
OPPORTUNITY-COST TIME PRESSURE Miller,1991).The results suggest that adaptivity in
environments characterized by opportunity-cost time The Decision Task
pressure may be also involve a shift toward more lec-The decision task in the simulation was a risky tive and more attribute-bad processing as time pres-choice.Each decision problem had four alternative sure is incread.gambles with four outcomes(e Fig.1for an example). In addition to ad
aptivity to salient task factors like Outcome values ranged from$.01to$9.99.The proba-time constraints and goals,people also have been bility that a given outcome ,Outcome1)is shown to be adaptive to more subtle context factors the same for all alternative gambles.However,each like the dispersion in weights characterizing a decision gamble may have a different payoff for that outcome. problem(Payne et al.,1988)and whether the correla-Such choice problems allow us to easily relate con-tion between outcomes is positive or negative(Bett-quences to choice among gambles in our experimental man,Johnson,Luce,&Payne,1993).For example,,people can play lected gambles for real Bettman et al.(1993)showed that people respond to money).3
interrattribute correlation by shifting their processing
3Our risky-choice context with multiple outcomes is structurally 2To illustrate dispersion of probabilities,a four-outcome alterna-similar to riskless multiattribute choice problems(Keeney&Raiffa, tive(gamble)with a low degree of dispersion might have probabilities
1976).Given that a good deal of rearch on choice processing has of.26,.29,.18,and.27.Alternatively,a gamble with a high degree ud riskless stimuli,much of the terminology describing processing of dispersion might have probabilities of.56,.06,.14,and.24.
us the term attribute instead of outcome.We u this standard
化妆使眼睛变大134PAYNE,BETTMAN,AND LUCE
The Decision Strategies bers of alternatives and attributes difficult.Con-
quently,we manipulated opportunity-cost time pres-In solving choice problems of the kind just described,
sure by using the number of elementary information the decision maker must arch among the probabili-
processing operations(EIPs)needed to complete the ties for each outcome and the values(payoffs)associ-
weighted additive rule in the particular decision envi-ated with the outcomes for each alternative.Different
ronment being studied as a standard of comparison. decision strategies can be thought of as different rules
动名词作定语To illustrate our approach,refer to the decision prob-for conducting that arch.In the simulation,we exam-
冠词练习题lem given in Fig.1.To implement the weighted adding ined six choice rules(equal weight additive value
strategy,a decision maker wouldfirst read the proba-(EQW),lexicographic(LEX),elimination-by-aspects
bility for Outcome1,then the value for Gamble A on (EBA),satisficing(SAT),weighted additive value
Outcome1,and then multiply the two values.Then (WADD),and random choice(RAND);e Payne et al.,
he or she would read the probability for Outcome2, 1993,for definitions).Thefirst four choice rules repre-
the value for Gamble A on Outcome2,multiply the nt heuristics in that potential information is ignored
two values,and add that product to the product for (effort is saved)at the cost of a potential loss in choice
Outcome1.This process would continue for all four quality(accuracy is lower).The four heuristics vary
outcomes,followed by the same process for Gamble B. substantially in how much information is ud and in
Then the weighted values for Gamble A and Gamble how the information is ud.For instance,the LEX
B would be compared.After similar processing for and EBA rules are associated with the u of relatively
Gambles C and D,the decision maker would choo limited amounts of information and processing of infor-
the gamble with the highest weighted value.Therefore, mation within attributes(attribute-bad),whereas
using the method for calculating EIPs described in the EQW and SAT rules are associated with processing
Payne et al.(1993),there would be32read EIPs(acqui-across attributes(alternative-bad).Also,rules like
sitions of information),16product EIPs(weighting op-LEX,EBA,and SAT tend to be associated with more
erations),12addition EIPs(compensatory combina-lective ,different amounts of informa-
tions of information),and3comparison EIPs(de-tion are procesd across alternatives or attributes).In
termining which of two values is larger)needed to solve general,greater variability or lectivity in arch is
the problem in Fig.1using the WADD rule,for a total associated with the u of noncompensatory decision
of63EIPs.If a person ud WADD or expected value strategies where a good value on one attribute cannot
maximization to choo among the alternatives,option compensate for a bad value on another attribute(e
D would be lected,with an expected value(weighted Payne,1976).The weighted additive rule and the sim-
sum)of$5.44.
ple random choice rule reprent the complete u of
Now consider implementing a lexicographic rule on information and the complete lack of information
the same problem.The decision maker wouldfirst read arch,respectively.The WADD rule also is associated
the probability for Outcome1,then the probability for with consistent processing(the same amount of infor-
Outcome2,and then compare the two,retaining the mation is examined for each alternative and attribute),
larger.Similarly,the remaining probabilities would be consideration of tradeoffs,and alternative-bad pro-
read and compared to ascertain the most probable out-cessing of information.Such rules are often viewed as
come.Then the values on that outcome for each gamble normative models of how decisions should be made
would be read and compared tofind the largest value (Keeney&Raiffa,1976;Frisch&Clemen,1994).
on the most probable outcome.The gamble with that
largest value would be lected.To solve the problem The Task Variable of Opportunity-Cost Time Pressure
in Fig.1using the LEX rule would require a total of One could manipulate opportunity-cost time pres-
only17EIPs(8reads,6comparisons,and3elimination sure for a decision problem by reducing the payoff re-
operations).The lexicographic choice rule would lect ceived as a direct function of the amount of time or
option B,with a lower expected value of$5.24.
effort spent in making the decision.However,that
Dividing the number of EIPs needed to solve the makes comparison across problems with different num-
problem using the LEX rule by the number of EIPs
needed to solve the same problem using the WADD terminology below.For example,the standard term attribute-bad rule indicates that the LEX rule should be about27% processing describes processing where information on one attribute(17/63)as effortful as the WADD rule.Now assume is examined for veral alternatives before another attribute is con-that the decision environment involves an opportunity-sidered.In our risky-choice context,that term will refer to examining
,there will be value loss due to delay.Further, values for one outcome across veral alternatives before considering
another outcome.assume that the value loss to delay means that any
135
DECISIONS UNDER OPPORTUNITY-COST TIME PRESSURE
outcome received after the time(effort)needed to u a in the simulation.Thefirst context factor varied was complete WADD strategy would lo25%of its original
the dispersion in probabilities(low vs.high).,be worth75%of its original value).For in-ties of.26,.29,.18,and.27exemplify low dispersion stance,if the WADD rule were ud in such an environ-
for a four-outcome problem,with probabilities of.56, ment to lect option D and Outcome1occurred,then.06,.14,and.24an example of high dispersion.Disper-the payoff would only be.75times$7.84,or$5.88.The
sion in probabilities is a context factor that is relatively expected payoff in that ca,given the lection of op-easy to notice.
tion D,would be.75times the expected value of option
The cond context variable was the correlation D,or$4.08(.75times$5.44).More vere opportunity-among the outcome values.This variable was t at
sherry怎么读
negative(high conflict among outcomes,or an average cost environments would be associated with higher
loss due to delay and hence receiving lower percent-correlation between outcomes of0.33),4zero,and posi-ages of the original values after the decision was made.
tive(low conflict,or an average correlation between To continue our example,if taking63EIPs to solve outcomes of.6).The correlation among outcomes is a decision in such an opportunity-cost environment
more difficult to notice than the dispersion in probabili-means an expected value loss of25%(you keep75%),ties becau it requires comparing values for veral
alternatives on veral attributes.A major question what would happen if you were to u a LEX rule to
make the decision more quickly?First,u of the LEX examined in our experimental work is how the pres-rule would lead to a different ,option B,and
ence of salient goals and opportunity-cost time pres-that option has a lower expected value,$5.24.However,sure may interact with more subtle context factors to one would expect that the value loss would not be as
determine how,and how well,people adapt their deci-great as with the WADD rule since the decision was sion behavior.
made more rapidly.We assume that the amount lost
Procedure
due to delay is proportional to the number of EIPs ud
proxyby the LEX rule relative to WADD.That is,instead of The six decision rules were individually applied to
a loss of25%,the loss would be27%of25%(i.e.,(17/1000decision problems in each of the24conditions
63)times25%),or roughly6.7%.Then the expected defined by a4(levels of time pressure—0,25,50,or payoff for the LEX rule would be the EV of Gamble B75%discount)by2(degree of dispersion—low or high) times the proportion of the value retained,or$5.24by3(correlation structure—negative,zero,or positive) times(10.067),or$4.89.More generally,the formula factorial.Payoffs for each problem were randomly gen-is Expected PAYOFFÅ(10((EIPs Rule/EIPs WADD)*erated from a distribution bounded ,$0.00) Value Loss))*Expected Value Rule Choice.,$10.00)using correlation structures Thus,there is some value loss due to using the LEX specified by using the appropriate correlation values rule to decide in the opportunity-cost time pressure between ,0.33,0,or.6)in the columns of the environment,but it is less than the value loss from Microsoft Excel5for Windows spreadsheet reprent-using the WADD rule.This is true even though the ing each decision problem.The dispersion levels for LEX rule does not lect the‘‘best’’choice in terms of probabilities were generated by using one of two meth-expected value.Nevertheless,in this ca a quicker ods designed to produce low levels of dispersion or high decision using a generally less accurate heuristic(LEX)levels of dispersion,as appropriate for the condition in yields a higher expected payoff($4.89)than a slower the simulation(e Johnson&Payne,1985,p.402for decision usi
ng the rule which is most accurate when details).The decision rules and simulation were pro-there is no time pressure(WADD,$4.08).grammed using Excel and Decisioneering’s Crystal In the simulation,opportunity-cost time pressure or Ball,software for running Monte Carlo simulations us-value loss(discount)due to delay was varied across ing spreadsheets.5
four levels:no discount,25%discount,50%discount,
and75%discount.The levels reprent the loss in4
It can be shown that for problems with n outcomes,the maximum value if the decision maker us as many EIPs as the average negative intercorrelation among all pairs of outcomes is[01/ WADD rule.The various levels of discount or value loss
(n01)](Green&Krieger,1986).
5Note that the simulation includes some simplifying assumptions. reprent the degree to which an environment pun-
For example,WADD weights values by probabilities,whereas other ishes delays in decision making,with higher values
,subjective expected utility,prospect theory)u subjec-leading to larger loss.汽车防盗器安装
tive values and/or decision weights.In addition,we weight each EIP
equally in our effort measure.Although different EIPs do require Context Variables
different amounts of time(Bettman et al.,1990),earlier simulation In addition to the task variable of opportunity-cost work has shown that the simpler equal weight assumption leads to time pressure,two context variables were manipulated
nearly identical conclusions(Payne et al.,1988).