PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf

This article may be used for research, teaching and private study purposes. Any substantial orsystematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply ordistribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contentswill be complete or accurate or up to date. The accuracy of any instructions, formulae and drug dosesshould be independently verified with primary sources. The publisher shall not be liable for any loss,actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directlyor indirectly in connection with or arising out of the use of this material.

Focalism and the Failure to ForeseeUnintended Consequences

Joyce EhrlingerFlorida State University

Richard P. EibachUniversity of Waterloo

From drug policy to environmental initiatives, well-intentioned actions have often led todramatic unintended consequences. The present research demonstrates that individualsfocus on what is intended in a way that leads them to not sufficiently consider importantunintended consequences. Participants fail to sufficiently weight the unintended conse-quences of attempts to satisfy preferences in a computer simulation (Study 1). Focus onintended consequences, and the resulting error in prediction, is reduced when the inter-relatedness of variables within a system is highlighted (Studies 2 and 3). This researchcan inform means of improving predictions within complex systems.

History is replete with examples of well-intentioned poli-cies producing dramatic unintended consequences.National Prohibition was undertaken in 1920 with thebest of intentions. However, people turned to medicinalalcohol that was legal but also more potent. In the end,deaths by alcohol poisoning increased by 300% duringProhibition (Thornton, 1998). Similarly, proponents ofthe Civil Rights Act of 1991 sought to minimize dis-crimination in the workplace by increasing the rightsof plaintiffs. Instead, fear of legal action made employ-ers more likely to lay off and less likely to hire minorityapplicants (Oyer & Schaefer, 2003).

A common theme in film and literature, both high-brow (e.g., Romeo and Juliet) and less so (e.g., all threeBack to the Future films), is that seemingly insignificantacts can set off a complicated string of unintended con-sequences. Despite a common understanding that unin-tended consequences can arise from seemingly benignactions, their occurrence is often not anticipated. Thegoal of the present research is to better understandwhy it is so difficult to foresee unintended consequencesand, in so doing, to inform ways to foresee unwantedconsequences in the future.

WHY ARE UNINTENDED CONSEQUENCESOFTEN UNANTICIPATED?

The phrases ‘‘unintended consequences’’ and‘‘unanticipated consequences’’ are often treated asthough they are logical equivalents. However, antici-pated consequences are not always intended. Manufac-turers sometimes anticipate, but do not intend, injuriesrelated to the use of their products. Similarly, conse-quences can be intended but not anticipated. Last-ditchefforts are executed with strong intentions to bringabout a perhaps unlikely consequence. People oftenuse unintended and unanticipated consequences nearlyinterchangeably, however, because it is exactly thoseconsequences that are unintended that are often alsounanticipated.

An early analysis by Merton (1936) provides a usefulframe for summarizing past research on why people failto foresee unintended consequences. First, Mertonargued, the failure to anticipate unintended conse-quences sometimes arises from ignorance. Unintendedconsequences frequently appear in systems that areextremely complex and for which not all variables areknown. Merton argued that the failure to anticipatecan also arise from error in our understanding of the rel-evant systems. Predictions and judgments can be heavily

Correspondence should be sent to Joyce Ehrlinger, Department ofPsychology, Florida State University, Tallahassee, FL 32306-4301.E-mail: ehrlinger@psy.fsu.edu

BASIC AND APPLIED SOCIAL PSYCHOLOGY, 33:59–68, 2011Copyright # Taylor & Francis Group, LLCISSN: 0197-3533 print=1532-4834 onlineDOI: 10.1080/01973533.2010.539955

Downloaded By: [Ehrlinger, Joyce] At: 04:33 25 February 2011

biased by inaccurate causal theories (e.g., Nisbett &Wilson, 1977; Ross, 1989). Misunderstanding the causalrelationships in a system can make it difficult to foreseeunintended consequences of one’s actions and choices.

Most of the literature on unanticipated consequencesis devoted to descriptions of specific cases of unintendedconsequences that were not foreseen because under-standing of the relevant systems was either incompleteor erroneous. Stretching across disciplinary lines, thisliterature includes analyses of unintended consequencesin political science (e.g., Holden, 1987; Jervis, 1997;Thornton, 1998), organizations (e.g., Kaminski, 2001;McBriarty, 1988; McKinley & Scherer, 2000), medicineand public health (e.g., Corn, 2000; Fisher & Welch,1999), engineering (e.g., Reason, 1987; Tenner, 1996),ecology (Simberloff & Stiling, 1996), and social systems(Kipnis, 1993; Meddin, 1986; for reviews, see Perrow,1999; Tenner, 1997).

Thus, one means of identifying potential unintendedconsequences is to ensure that one’s understanding ofthe relevant systems is complete and accurate (or, atminimum, that one’s level of confidence and risk consid-erations matches one’s level of understanding of thosesystems). We argue that, in addition, there are psycho-logical features of the way that predictions are madethat can make it difficult to foresee unintended conse-quences. In particular, we argue that there are two waysin which focus on the intended consequence can blindpeople to alternative consequences. First, failures toforesee unintended consequences can stem from anoverly passionate focus on the intended consequence.Merton argued that people might become so consumedby desire for a preferred consequence that they areunable to think beyond that consequence. This articleexplores a second way in which focus can be blinding.We argue that failures to foresee unintended conse-quences can also stem from a dispassionate focus onthe intended consequence. We draw from past researchsuggesting that people’s attention is often focused quitenarrowly (e.g., Schkade & Kahneman, 1998). This dis-passionate narrowing of attention, known as focalism,leads to judgments that are overly influenced by thecurrent object of attention and that place insufficientweight on other relevant variables.

To understand how dispassionate focus can biasjudgment, imagine that you are asked to choose betweentwo vacation spots about which you have several piecesof information. All of Spot A’s qualities are average(e.g., average weather, medium quality hotel, averagebeaches, average nightlife). In contrast, Spot B is char-acterized by extremes (e.g., fantastic weather, terriblebeaches, fabulous hotels, almost no nightlife). If youare like most participants faced with this choice (Shafir,1993), you might find yourself focusing on the extremegood qualities of Spot B and prefer that vacation over

one at Spot A. This seems a reasonable choice until itis compared to choices made by participants asked toimagine that they already hold reservations to bothvacation spots and must choose which vacation toreject. When choosing which vacation spot to reject,participants focused on the extreme negative qualitiesof Spot B and chose to reject that vacation rather thanSpot A. In both cases, participants considered the sameinformation but the particular choice they were making(which to choose vs. which to reject) led them to focusnarrowly on a subset of the relevant information. As aresult, the two groups made inconsistent decisions.

This illustrates just one example from a large body ofresearch in social and cognitive psychology suggestingthat focalism, or a tendency to focus narrowly on oneor a few variables, can lead to neglect of importantinformation and, consequently, inaccurate judgment(e.g., Buehler & McFarland, 2001; Koehler, 1991; Klar& Giladi, 1999; Kruger & Burrus, 2004; Windschitl,Kruger, & Simms, 2003). Further, manipulations thatsubtly change attentional focus can have dramaticeffects on accuracy in judgment (e.g., Shafir, 1993;Storms, 1973; Tversky, 1977).

Applying this research to the problem of unintendedconsequences, we argue that focalism with respect to theintended consequence can result in a neglect of impor-tant information regarding alternative, unintended con-sequences—including information that is knowable andplainly relevant to predictions. Although psychologicalexplanations have received little attention in the litera-ture on unintended consequences, there is some evidencethat those who mention considering possible indirectconsequences of their actions in a simulation game aremore successful in those games (Dorner, 1980, 1996).Although Dorner did not ask participants to make pre-dictions in these studies, one possible reason that thosewho consider the possibility of indirect consequencesperform better is because they are making more accuratepredictions about outcomes in the game.

A second example can help to illustrate this point.People tend to mistakenly predict that people residingin California are happier than are those who reside inthe Midwest (Schkade & Kahneman, 1998). This pre-diction makes sense when focusing on a comparisonbetween the sun and beaches of California and the snowand cornfields of the Midwest. A focus on weather, how-ever, leads people to give too little weight to the manyother variables that might influence a person’s level ofhappiness (e.g., social support, dispositional happiness).Imagine a person who decides to move from Iowa toCalifornia in order to enjoy a happier life. Her focusupon the sun and beaches of California might lead herto not give sufficient weight to potential unintendedconsequences stemming from her move (e.g., a loss ofcloseness with friends and family in Iowa). As a result,

60 EHRLINGER AND EIBACH

Downloaded By: [Ehrlinger, Joyce] At: 04:33 25 February 2011

her prediction that moving to California will lead toincreased happiness might be completely inaccurate.

THE PRESENT RESEARCH

The present set of studies provides empirical demonstra-tions of how a focus on the intended consequence leadsto a failure to anticipate unintended consequencesacross three quite different settings. In Study 1, parti-cipants predicted future states of a simulation modelin which a focus on movements that might satisfyagents’ preferences can make it difficult to recognizepotential unintended consequences of those movements.Studies 2 and 3 explore how manipulations of focus onthe intended consequence affect the likelihood of fore-seeing potential unintended consequences. In Study 2,participants made predictions about the consequencesof potential laws either in a no manipulation controlcondition or after completing a ‘‘defocusing’’ manipu-lation meant to encourage them to consider the largersystem of variables. Study 3 utilized a different meansof ‘‘defocusing’’ participants before asking them tomake predictions about the consequences of a changeto college admissions criteria. We expected that thedefocusing manipulation in Studies 2 and 3 would pro-duce less biased, more accurate predictions.

STUDY 1

Policymakers often act with the intention of increasingthe satisfaction of their constituents. However, policychanges that increase the satisfaction of some constitu-ents can sometimes decrease the satisfaction of others.If policymakers focus narrowly on the effects of theirpolicies on those whom they intend to benefit and ignorepotential negative effects on others, they may overesti-mate the net satisfaction that will be produced by theiractions.

In Study 1 we designed a modified version of aSchelling (1971, 1972) model to test whether the tend-ency to narrowly focus on the preferences of the inten-ded beneficiaries of a change leads people to overlookunintended consequences for others within that systemand thus overestimate the net satisfaction produced bythe change. A Schelling model consists of a simple grid,much like a checkerboard, in which two types of agentsare distributed randomly. Each agent is endowed with apreference for proximity to in-group members. In eachround of the model simulation, agents remain in theiroriginal position if their preferences regarding the pro-portion of neighbors who are in-group members are sat-isfied. Those with a proportion of in-group to out-groupneighbors that fails to meet their preferences move to a

new, randomly selected position. These rounds continueuntil the preferences of all agents are satisfied.

In models of this type, even moderate levels ofin-group preference lead reliably to segregation overtime. Schelling’s original demonstration was powerfulbecause, at the time, few expected this inevitability ofsegregation. We have based the model used in thisexperiment upon Schelling’s initial work but have modi-fied it to our exploration of predicted consequences. Inthe present study, we predicted that participants wouldoverestimate the speed of preference satsifaction byvirtue of focusing too much on the intended consequ-ences of agent movement.

In Schelling’s model, agents are merely nodes in acomputer program that are moved randomly, notattending to the movement of others or to where anideal location might be. Although the intended conse-quence is to move agents to a location where their pre-ferences are satisfied, new locations are chosen atrandom. Thus, the new location an agent is moved tomight no better satisfy his or her preferences than theprevious location and might lead to unintended conse-quences in that it might unsettle out-group memberswhose preferences were satisfied before the arrival oftheir new neighbor. In our experiment, the model wasdescribed as an elementary school classroom in whichboys preferred sitting next to other boys and girls pre-ferred sitting next to other girls. We asked participantsto make predictions about how quickly the classroomwould stabilize given a preference that at least two thirdsof each student’s neighbors shared his or her gender.

It is important to note that we took pains to makesure that all participants understood the model beforethey made predictions. We wanted to rule out the possi-bility that any error in prediction stemmed from misun-derstanding the model and demonstrate, instead, thaterrors in prediction stemmed from focalism. Specifically,we expected that participants would focus on theintended consequence of movement—moving a studentaway from a seat that made him or her unhappy—andthat this would lead them to neglect how movement ofthese students might result in unintended consequences,that is, making previously happy students unhappy.Further, we expected that this failure to anticipate theunintended consequences of movement would lead part-icipants to overestimate how quickly the preferences ofall students in the class could be satisfied.

Methods

Participants. Forty-six undergraduates completedthis experiment for course credit. We excluded 3 parti-cipants who demonstrated a failure to understand themodel either through poor performance on the quizand=or through explanations of their predictions. One

FOCALISM AND UNINTENDED CONSEQUENCES 61

Downloaded By: [Ehrlinger, Joyce] At: 04:33 25 February 2011

additional participant provided predictions in non-numerical terms (e.g., ‘‘a lot’’) that could not be ana-lyzed. After these exclusions, our final sample was 42participants.

Procedure. Upon arriving at the lab, participantswere asked to read a detailed explanation of the model.They read that each grid represented a classroom inwhich girls and boys were seated randomly, each witha preference for sitting next to students who share hisor her gender. ‘‘Sitting next to’’ was defined as sittingin a directly adjacent horizontal or vertical square. Part-icipants were told that students next to each other diag-onally were not included in preference considerations. Ifthe student was the same gender as at least two thirdsof his or her neighbors, the student was considered‘‘happy’’ (the student’s preferences were satisfied) andhe or she remained seated for that round. If, however,the student’s neighbors did not meet the two-thirds cri-terion (i.e., the student was ‘‘unhappy’’), he or she wouldbe moved to a new, randomly selected seat. In eachround, all students whose preferences were not satisfiedvacated their seats simultaneously and were thenassigned new seats at random. The movement createda new seating configuration with, potentially, a newset of happy and unhappy students. As many roundswould be run as necessary to satisfy all students’preferences.

We took pains to ensure that participants understoodthe rules of the model, including that the unhappy stu-dents—those whose preferences were not satisfied—weremoved to a randomly selected new seat. To ensure thatparticipants understood the model, they were given abrief quiz to ensure that participants (a) could correctlyidentify happy and unhappy students in a grid and (b)understood that movement within the model was ran-dom. For example, participants were asked to explainhow the model determines the new seat for an unhappystudent. A second multiple-choice question asked parti-cipants to indicate which of several possible seats wasthe most likely new seat for an unhappy student. Toachieve a perfect score, participants had to indicate thatno seat was more or less likely than any other becausethe new seat would be chosen at random. The exper-imenter scored each participant’s quiz immediately andthen explained the reasoning behind each correctanswer.

Once it was clear that participants understood themodel, they were asked to make predictions about threehypothetical classrooms. Each classroom contained 25seats in a 5! 5 grid and 20 or fewer students seated inthe room. The classrooms differed in the number of stu-dents seated (10, 15, or 20 students) but, in each room,60% of the students were unhappy. For each classroom,participants were asked to predict the number of

students who would be unhappy after two rounds andthe number of rounds required to satisfy the preferencesof all students. They were also asked to describe the fac-tors they considered while making these predictions.

To create accuracy criteria, we created a computersimulation of the model. Given an initial seating con-figuration, the program ran a series of rounds in whichunhappy students were identified and moved to ran-domly determined new seats. The program repeated thistask until the preferences of all students were satisfied.We calculated accuracy criteria for each classroom byaveraging across the results of 100 simulation trials.1

Results and Discussion

Before beginning analyses, we created composite mea-sures of the degree to which participants over- or under-estimated the speed with which students’ preferenceswere satisfied. Specifically, we calculated the degree towhich participants over- or underestimated the numberof unhappy students in Round 3 and the total numberof rounds necessary to satisfy all agents’ preferencesfor each classroom, expressed as a percentage of the rel-evant accuracy criterion. We then averaged across errorpercentages for the three classrooms to create two com-posite measures for each participant—the percentage bywhich they under- or overestimated the number ofunhappy students in the third round and the percentageby which they under- or overestimated the number ofrounds necessary for all students’ preferences to be sat-isfied. Participants’ estimates of the number of roundsnecessary to satisfy student preferences were positivelyskewed. To satisfy the assumptions of normality, we firstcreated the relevant composite measure of predictiveerror. We then added a constant and log-transformedthis percentage error. All reported analyses were con-ducted on this log-transformed data. We then reversedthis transformation to calculate and report mean valuesfrom the analyses.

We predicted that participants would focus on theneeds of unhappy students in the model and, as a result,underweight the possibility of unintended consequencesstemming from their movement. As a result, we expectedthat participants would underestimate both the numberof unhappy students in the third round and the numberof rounds necessary for student preferences to be satis-fied. Indeed, participants underestimated the numberof unhappy students in Round 3, on average, by

1Across 100 trials in our model simulation for each classroom con-figuration, the average number of unhappy students after Round 3 was5.96 in the 10-student classroom, 10.0 in the 15-student classroom, and16.04 in the 20-student classroom. The average number of roundsbefore total segregation was 10.62 in the 10-student classroom, 15.6in the 15-student classroom, and 24.5 in the 20-student classroom.

62 EHRLINGER AND EIBACH

Downloaded By: [Ehrlinger, Joyce] At: 04:33 25 February 2011

31.23%, t(40)!"12.222, p< .001, and the number ofrounds necessary to satisfy all students preferences by83.92%, t(41)! 28.17, p< .001 (see Table 1).

Next, we explored whether this error in predictionsstemmed from a tendency to focus on unhappy studentsand, as a result, neglect the unintended consequences oftheir movement on other students. Two coders read part-icipants’ open-ended descriptions of the thought pro-cesses behind their predictions and rated whether, acrosstheir descriptions for any of their predictions, participantsspecifically mentioned that movement of the ‘‘unhappy’’students could upset the stability of the ‘‘happy’’ students(a! .81). As expected, those participants who mentionedthat such unintended consequences might occur (only26% of the sample) made more accurate estimates thanthose who made no mention of unintended consequenceswith respect both to the number of rounds required tosatisfy all students’ preferences, t(38)!"2.45, p< .05and, marginally, the number of ‘‘unhappy’’ students inRound 3, t(40)! 1.95, p< .10. These results support ourargument that participants often focused on movementof the unhappy participants in a way that led to a neglectof unintended consequences of that movement and,consequently, inaccurate predictions.

Discussion

Participants underestimated the number of unhappystudents in Round 3 and the total number of roundsneeded to satisfy all students’ preferences. These errorsin prediction are in exactly the direction one wouldexpect if participants gave insufficient weight to thepossibility of unintended consequences of student move-ment. An analysis of participants’ explanations of theirpredictions provides further evidence that a focus onthe intended consequence of improving the situation ofunhappy students contributed to errors in prediction.Only one fourth of our sample mentioned the possibilityof unintended consequences of movement, and thoseindividuals made more accurate predictions regarding

the model than did those who did not mention thispossibility.

Elsewhere we have shown that people also overlookunintended consequences when they work in groups(Ehrlinger, 2004). For instance, in one study, participantsshot basketball free-throws first individually and then inteams of three. Participants anticipated that joining ateam would result in a better score than each individualcould achieve alone. However, they tended to overlookhow joining a team would produce unintended conse-quences in the form of coordination problems. Predic-tions of the team score were overconfident in exactlythe way one would expect if participants placed too muchweight on the presence of additional shooters and too lit-tle on the possibility of coordination errors. Further, ananalysis of participants’ thought listings suggested thatoverconfidence was confined to those who failed to antici-pate the possible unintended consequences of workingtogether. Those teams who anticipated that crowding,balls bouncing off each other, and other difficulties coor-dinating efforts might occur, in contrast, tended to makeaccurate estimates of their team’s score.

These studies provide initial evidence that people failto foresee unintended consequences and that this failurerelates to errors in prediction. In Studies 2 and 3, wesought to balance the experimental control available inthe lab in Study 1 with demonstrations involving judg-ments about current political events. Studies 2 and 3also offer experimental manipulations of focus todirectly investigate the degree to which a failure to fore-see unintended consequences stems from focusing toomuch on the intended consequence and too little uponalternative possibilities.

Studies 2 and 3 employ manipulations that draw atten-tion away from the point of focus with the goal of increas-ing consideration of alternative outcomes. Indeed, pastresearchers have successfully reduced the tendency toplace too much weight on the focal point when makingjudgments (e.g., Wilson, Wheatley, Meyers, Gilbert, &Axsom, 2000). The tendency to overestimate the lon-gevity of emotional reactions to a favorite sports team’sperformance, for example, is attenuated after peoplecomplete a diary outlining the events they were likely toexperience on a day soon after the game (Wilson et al.,2000). Completing this diary, Wilson and colleaguesargued, reminded participants to focus on additional lifeevents, thus reducing focus upon the game and producingmore accurate estimates of future happiness.

In Studies 2 and 3, participants completed tasks inwhich they assessed the importance of variables eitherindividually or in a way that highlights their interrela-tions. We expected that highlighting the degree to whicheach variable is related to all others in the system wouldlead participants to think more about how a change inone variable would have unintended consequences on

2Degrees of freedom vary across analyses because participants leftsome questions blank.

TABLE 1Underestimation and Consideration of Unintended Consequences in

a Modified Schelling Task

% Underestimation of:MentionedPossibility

Did Not MentionPossibility Difference

Unhappy students inRound 3

21.02% 34.63% 13.64%

Rounds untilpreferences satisfied

71.67% 91.61% 19.94%

FOCALISM AND UNINTENDED CONSEQUENCES 63

Downloaded By: [Ehrlinger, Joyce] At: 04:33 25 February 2011

other variables. As a result, we expected participants inthis ‘‘defocused’’ condition to make predictions of theoverall consequence of change in ways less consistent withthe intended consequence, relative to control participants.

STUDY 2

In 2003, California held a gubernatorial recall electionsparked, in part, by a $38 billion budget deficit.Although Californians were quite troubled by the exist-ence of this deficit, opinion polls suggested that theyrejected every cut in service proposed to reduce the defi-cit as well as nearly every proposal to increase taxes(Baldassare, 2003). When these attitudes are consideredin isolation, they make perfect sense. It is the rare personwho, if asked in isolation, would prefer to pay highertaxes. However, people might be willing to pay highertaxes to reduce a dangerous budget deficit if, by sodoing, they will protect a favored government programfrom budget cuts. We argue that people tend to focusnarrowly on the proposal at hand (e.g., a proposal toraise or lower taxes) without consideration of how thatproposal might affect other aspects of the budget.

National opinion polls suggest that question framingcan be used to highlight the interrelations betweenbudget categories. Gallup and Investor’s Business Daily,for example, each asked respondents about approachesto stimulating the economy during the winter of 2001–2002. The Gallup poll (2002) asked a question that high-lighted the interrelations among possible strategies.They asked, in a single question, whether participantsthought the economy should be stimulated primarilythrough tax cuts or primarily through increased spend-ing on, for example, ‘‘benefits for recently unemployedworkers and construction projects.’’ Roughly even num-bers of participants favored tax cuts versus increasedspending, but fewer than 8% of respondents said thatCongress should pursue both strategies. When Investor’sBusiness Daily and the Christian Science Monitor (2001)asked participants about the same strategies in separatequestions, 66% favored cutting taxes to stimulate theeconomy and 72% favored increased spending. Withoutspecifically highlighting the degree to which Congressmight have to choose primarily one or the other strat-egy, we argue, respondents focused on each strategy inisolation rather than as part of a larger, limited budget.

Study 2 offers an experimental analog of how peoplemight think about categories of national spending in iso-lation. Participants were asked to first indicate howmuch money each national spending category should,ideally, receive. Half of the participants were assignedto allocate national spending in terms of percentagesof the total budget. This task forces attention to theinterrelations among spending categories in that spend-ing more in one category leaves less for other categories.

After completing this budget allocation task, weasked participants to rate the consequences of twoproposals—a large tax cut and a large increase in socialwelfare funding. Each proposal represents somethingvery valuable if considered in isolation, but each wouldrequire a substantial amount of money. Thus, a strongcommitment to one should suggest a weaker commit-ment to the other. We expected participants in the defo-cused condition, relative to control participants, to moreoften think about how a tax cut might limit the moneyavailable for desired government services. We alsoexpected defocused participants to more often thinkabout the cost of an increase in welfare spending and,as such, to wonder about where that money would comefrom. Generally, we expected defocused participants tomore often mention possible unintended consequencesof the proposals and, as a result, to make predictionsof the overall effect that were less in the direction ofthe intended effect.

Method

Participants. Fifty-six undergraduates participatedin exchange for extra credit in psychology courses.

Procedure. Participants were asked to first completea federal budget task. They were seated at a computerand asked to read a Web page on which there was a listof the spending categories that make up the federal bud-get (as cited at http://www.nathannewman.org/nbs/ aspart of a similar national budget allocation task). Thelist included information regarding how the budgetwas allocated in 2001 both in terms of the dollar amountallotted to each category and what those dollar amountstranslated to in terms of a percentage of the total bud-get. Participants could learn more about each categoryby clicking on its name to find a paragraph describingthe types of spending making up that category.

Participants were asked to indicate how they thoughtthe national budget should be allocated in one of twoways. In the control condition, there was a pull-downmenu next to each category that participants used toindicate whether they thought funding should remainthe same, be increased, or be decreased. Each pull-downmenu constituted a scale ranging from increasing fund-ing by 100% to decreasing funding by 100%.

In the defocused condition, pull-down menus werereplaced with textboxes in which participants wrote apercentage of the total budget that they wished to allo-cate to that category. A calculator at the bottom of thepage displayed what percentage of the total budget hadbeen spent thus far. Participants were not allowed tocontinue to the next task until they had allotted exactly100% of the budget, no more and no less. Because thistask requires some decrease in funding in some category

64 EHRLINGER AND EIBACH

Downloaded By: [Ehrlinger, Joyce] At: 04:33 25 February 2011

for every increase in funding elsewhere, it was meant tohighlight the degree to which the categories were notindependent and, instead, were part of a system.

After completing the budget allocation task, parti-cipants were asked to imagine two measures that mightbe introduced to Congress. The first described a largetax cut, intended to ease the tax burden paid byAmericans. Participants predicted how positive theoverall effect of this tax cut would be on a scale from1 (extremely negative) to 9 (extremely positive). Theywere also asked to explain the reasons for this predic-tion. Finally, participants answered the same two ques-tions with respect to a proposal for a large increase insocial welfare funding.

Results and Discussion

We expected participants who were defocused, in thatthey allocated the national budget as percentages, to holdless positive expectations of both proposals than controls.Indeed, those encouraged to think about the budget as asystem made predictions of the overall effect of both pro-posals that were less positive. Defocused participantsmade less positive predictions of both the effect of thetax cut (4.82 vs. 5.44), t(53)! 1.75, p< .10, and of theincrease in social welfare spending (5.52 vs. 6.33),t(54)! 2.02, p< .05, than did controls. The intendedeffects of both proposals, explicitly stated within thequestion, were quite positive. Thus, reducing focus onthe intended effect led participants to make predictionsthat were less in the direction of that intended effect.

To determine whether the observed difference wasdue to differences in consideration of possible unin-tended consequences, open-ended responses were readby two coders, blind to condition. Coders first made ajudgment about whether each response made referenceto any type of unintended consequence (a! .75) andthen, separately, coded for the mention of two specifictypes of consequences. They coded for whether parti-cipants mentioned (a) how the tax cut might indirectlyaffect the availability of government services (a! .88)and (b) how an increase in social welfare spending mightindirectly affect future taxes or the availability of othergovernment services (a! .85).

As expected, defocused participants were more likelythan control participants to mention possible unin-tended consequences of both the tax cut (75.0% vs.44.0%), v2(2)! 5.77, p< .05, and of the increase in socialwelfare spending (46.4% vs. 16.0%), v2(2)! 5.79, p< .05.In particular, defocused participants were more likely tospecifically mention one category of spending having anindirect, unintended consequence on another categorythan controls. Relative to control participants, defo-cused participants more often spoke of possible unin-tended consequences of a tax cut on government

services (64.3% vs. 24.0%), v2(2)! 9.11, p! .01, and ofan increase in social welfare spending on taxes or otherareas of spending (17.9% vs. 0%), v2(2)! 4.93, p< .10.One defocused participant provides a particularly clearexample of how the manipulation led to decreased focus.While explaining the logic behind her prediction, shesaid, ‘‘Doing the budget balancing exercise earlier justshowed how many things our federal budget has topay for. If taxes were decreased, where would moneyfor all those programs come from? Lots of things wouldsuffer as a result.’’

We conducted an analysis to determine whetherconsideration of unintended consequences mediated thecondition effect on expectations for the overall outcome.First, we converted participants’ predictions regardingthe outcome of the proposed tax cut and of the pro-posed increase in welfare spending to z scores and aver-aged them to create an index of predicted impact. Wecreated an index of participants’ consideration of unin-tended consequences in the same way, computing zscores for consideration of each proposal and then aver-aging the individual z scores together. The defocusingmanipulation predicted expectations regarding the over-all consequences of the proposals, b!".34, p< .05.Defocusing also led to greater consideration of unin-tended consequences, b! .41, p< .005. However, con-sideration of unintended consequences failed tosignificantly predict expectations of the overall effectin a multiple regression controlling for the effect ofcondition, b!".16, p! .28. Thus, consideration ofunintended effects did not appear to mediate the rela-tionship between condition and expected consequencesalthough the data were in the predicted direction.

Summary

Spending money in one area has the indirect effect ofensuring that less money will be spent on some other area.Most voters know on some level that the governmentdoes not have an endless supply of money and, as such,know that policy decisions require some degree of weigh-ing and balancing which areas of spending are moreimportant. Still, Study 2 demonstrates that they maynot think about how spending in one area indirectlyaffects others. By asking participants to allocate thenational budget as a percentage, the degree to which vari-ables were interrelated was highlighted such that, whenthey were asked to consider specific issues, participantsmore often thought about how each proposal mightindirectly affect other important responsibilities of thegovernment. The manipulation successfully reducedfocus, leading to greater consideration of possible unin-tended consequences as well as less positive expectationsof the proposal. Study 3 serves as a conceptual repli-cation, using a different means of manipulating focus.

FOCALISM AND UNINTENDED CONSEQUENCES 65

Downloaded By: [Ehrlinger, Joyce] At: 04:33 25 February 2011

STUDY 3

College admissions officers are faced with a very difficulttask in that they must identify the students, out of manyapplicants, who most deserve to attend their university.Much like balancing a budget, admissions officers needto do more than just identify qualities that they thinkimportant. Instead, they must also determine the relativeweight to be placed on each characteristic. If the interre-lations between factors that go into the decision processare highlighted, will laypeople adopt an approach moresimilar to that of admissions officers? Will this under-standing of how the factors are interrelated promoteconsideration of possible unintended consequences?We explore these questions in Study 3 by asking parti-cipants to indicate the degree to which they thought aseries of factors were important to consider whenmaking admissions decisions.

As in the previous study, participants indicated theirbelief about the importance of these attributes in one oftwo ways. Control participants were asked to rate thedegree to which each characteristic was important toconsider. Participants in a ‘‘defocused’’ condition werealso asked to indicate how important they thought eachcharacteristic but did so by making a pie chart ratherthan responding to individual questions. We predictedthat constructing a pie chart would highlight the sys-temic nature of the problem because allocating a largewedge to one characteristic leaves much less room toallocate to all other characteristics.

We expected that people would, generally, considerthe importance of each potential factor in admissionsdecisions in isolation. As in the previous study, however,we expected that a relatively simple manipulation wouldhighlight the degree to which different characteristics areinterrelated. Further, we expected that highlighting thesystemic nature of the problem would lead people tomore often consider possible unintended consequencesof changes to the admissions decisions and, as a result,make less positive predictions about the overall impactof such changes.

Methods

Participants. Participants were 49 undergraduateswho received extra credit for their participation.

Procedure. Participants were asked to imaginethemselves as members of an advisory board for a localcollege who wished to reevaluate their criteria for collegeadmissions. In that hypothetical role, they indicated therelative importance of 10 factors relevant to admissionsdecisions (e.g., high school grade point average, standar-dized test scores, participation in extracurriculars). Con-trol participants rated the importance of each factor on

a series of scales ranging from 1 (unimportant) to 9(extremely important). Participants in the ‘‘defocused’’condition, instead, indicated the importance of each fac-tor through a pie chart. They were provided a circle(16.5 cm diameter) divided by lightly dotted lines mark-ing wedges for each 5% of the chart. Participants wereasked to label wedges in whatever way they thoughtappropriate to indicate the importance of each factorfor the final admissions decision.

After indicating the importance of each factor, parti-cipants were asked to imagine that the board was con-sidering placing greater emphasis on participation inextracurriculars. They were asked to predict how posi-tive the consequences of this change in emphasis wouldbe on a scale from 1 (very negative) to 9 (very positive).As in previous studies, they were also asked to explainthe logic behind their prediction.

Results

We predicted that ‘‘defocused’’ participants would con-sider unintended consequences of this change moreoften than control participants and, as a result, be lessoptimistic about the overall effects of this change. Asexpected, defocused participants made predictions ofthe overall effect of the proposed change that were lesspositive (M! 4.96) than predictions made by controls(M! 6.31), t(47)! 3.05, p< .005. To examine whetherthis difference stemmed from a difference in consider-ation of unintended consequences, two coders, blind tocondition, read through open-ended responses andnoted whether each participant mentioned possibleunintended consequences (a! .75). As predicted, defo-cused participants were far more likely to talk of unin-tended consequences than were controls, v2(2)! 7.89,p< .05. Indeed 47.8% of defocused participants madesome mention of possible unintended consequences,compared to only 19.2% of controls.

As in the previous study, we conducted a mediationalanalysis to determine whether the effect of the defocus-ing manipulation on predictions of the overall outcomewas mediated by the tendency to consider possible unin-tended consequences. As predicted, the defocusingmanipulation strongly predicted expectations regardingthe overall consequences of the change. Further, thedefocusing condition predicted greater consideration ofunintended consequences. Finally, a multiple regressionformula including both condition and consideration ofunintended consequences revealed that considerationof unintended consequences does lead to less positivepredictions of the overall effect even after controllingfor the effect of condition (see Figure 1). Inclusion ofconsideration of unintended consequences in the modelresulted in a weaker relationship between the conditionand predictions of the overall effect, Sobel z! 2.00,

66 EHRLINGER AND EIBACH

Downloaded By: [Ehrlinger, Joyce] At: 04:33 25 February 2011

p< .05. Thus, it seems that a difference so small as con-structing a pie chart rather than rating features on sep-arate scales highlights the degree to which possiblefactors are interrelated such that greater emphasis onone requires less emphasis placed elsewhere. The defo-cusing condition led to less optimistic predictions byleading people more often to consider possible unin-tended consequences of the proposed change.

GENERAL DISCUSSION

This article provides evidence that a failure to foreseeunintended consequences stems, in part, from a tend-ency to focus on the intended consequence in isolation,rather than as part of a system of interconnected vari-ables. In Study 1, focus on the intended consequenceled to predictions that did not sufficiently account forthe possibility of unintended consequences. Participantswho did consider potential alternative consequencesmade more accurate predictions. Studies 2 and 3 pro-vided more direct evidence that focus on the intendedconsequence leads to predictions reflecting confidencethat it will come to pass.

In Studies 2 and 3, simple manipulations were used tohighlight the degree to which the variables under con-sideration are related and increased weight given toone requires that less weight be given to the others. Indi-viduals who were defocused in Study 2 were less opti-mistic about potential consequences of a hypotheticaltax cut and a hypothetical increase in social welfarefunding. After completing a task that highlighted thedegree to which different areas of government spendingwere interconnected, individuals more often consideredhow advocating one category of spending might impactother categories. This simple change led to more fre-quent consideration of unintended consequences andfewer positive judgments. Similarly, participants inStudy 3 more often recognized the possibility of unin-tended consequences stemming from a change in admis-sions decisions when defocused through a task thathighlighted the interrelations between factors in anadmissions decision, relative to control participants.

Encouraging Greater Accuracy in Predictions

The final two studies are important both because theydemonstrate how focus on intended consequence leadspredictions astray and because they suggest a tool forimproving predictions. Studies 2 and 3 suggest thatdecision makers who seek to make the best choices poss-ible should seek to highlight the relationships betweenvariables in the system of interest. This might be a diffi-cult task in many cases. It is considerably easier to ident-ify all relevant variables in a budget or in an admissionsdecision than it is to identify all relevant variables in anecosystem. Thus, it might be impossible to highlightinterrelations in a meaningful way as the systembecomes highly complex. Still, defocusing manipulationssuch as those used in Studies 2 and 3 might improve pre-dictive accuracy so long as the system in questionremains relatively simple and easily defined. It may bethat we can adopt means of highlighting the systemicnature of a problem to more often anticipate and, thus,take steps to avoid unintended consequences.

To identify additional means of anticipating unin-tended consequences, we should review how nonpsycho-logical factors influence the likelihood of anticipation.Merton argued that a failure to foresee unintended con-sequences stems, in part, from ignorance and error inunderstanding the system under consideration. We canlearn from past experiences and educate others aboutthe ways in which variables within a system interactand about the potential for unintended consequenceswithin that system. Indeed, memorable unintended con-sequences from the past might also serve much the samefunction as the defocusing manipulations used in Studies2 and 3. Highlighting previous strings of consequencesshould remind participants of the degree to which vari-ables are interrelated and, as a result, improve antici-pation of unintended consequences.

CONCLUSION

In this article, we have shown that the failure to foreseeunintended consequences stems, in part, from the psy-chology of the predictor. In particular, we have shownthat the failure to anticipate unintended consequencesstems from a tendency to focus on what one intends ina way that blinds one to alternative possibilities. As aresult, individuals tend to be overconfident that theintended consequence will come to pass and that therewill be no additional unwanted consequences. This workrepresents one of the first attempts to understandpsychological sources of the failure to foresee unintendedconsequences. Further, it suggests intriguing ways inwhich one might encourage more accurate predictionsand, consequently, better decisions within complex sys-tems. Given the prevalence of unintended consequences

FIGURE 1 A model of how reduced focus affects predictions regard-ing admissions decisions.

FOCALISM AND UNINTENDED CONSEQUENCES 67

Downloaded By: [Ehrlinger, Joyce] At: 04:33 25 February 2011

stemming from acts big and small, the means of improv-ing judgment suggested by the present work can provideimportant contributions both at the level of basicpsychological research and in terms of applications.

ACKNOWLEDGMENTS

Portions of this research were presented at the 2004meeting of the Society for Personality and SocialPsychology, Austin, Texas.

We thank Richard Darlington, David Dunning,Carol Dweck, Thomas Gilovich, and Dennis Reganfor feedback on this research and earlier drafts of thisarticle. We also thank Tasha Campbell, Alex Columbus,Tracy Decker, Whitney Hewitt, Chris Minunni, LauraReadinger, and Rachel Smallman for assisting with datacollection as well as Rick Dale and Ronald Kimball fortechnical assistance.

REFERENCES

Baldassare, M., with The James Irvine Foundation. (2003, June).PPIC statewide survey: Special survey on the California State Bud-get. San Francisco: Public Policy Institute of California. Availablefrom http://www.ppic.org/content/pubs/survey/S_603MBS.pdf

Buehler, R., & McFarland, C. (2001). Intensity bias in affectiveforecasting: The role of temporal focus. Personality and SocialPsychology Bulletin, 27, 1480–1493.

Corn, J. K. (2000). Environmental public health policy for asbestos inschools: Unintended consequences. Boca Raton, FL: CRC Press.

Dorner, D. (1980). On the difficulties people have in dealing withcomplexity. Simulation & Games, 11, 87–106.

Dorner, D. (1996). The logic of failure: Recognizing and avoiding errorin complex situations. Cambridge, MA: Perseus.

Ehrlinger, J. (2004). Complexity neglect: Psychological sources ofthe failure to anticipate unintended consequences. (DoctoralDissertation, Cornell University, 2004). Dissertational AbstractsInternational, 65(7), 3551B.

Fisher, E. S., & Welch, H. G. (1999). Avoiding the unintended conse-quences of growth in medical care: How might more be worse? Jour-nal of the American Medical Association, 281, 446–453.

Gallup. (2002, January 7–9). Gallup poll. Retrieved January 10, 2004,from http://www.pollingreport.com/budget.htm

Holden, C. (1987). U.S. anti-abortion policy may increase abortions.Science, 238, 238.

Investor’s Business Daily & Christian Science Monitor. (2001,November 7–11). Investor’s Business Daily and Christian ScienceMonitor poll. Retrieved January 10, 2004, from http://www.pollingreport.com/budget.htm

Jervis, R. (1997). System effects: Complexity in political and social life.Princeton, NJ: Princeton University Press.

Kaminski, M. (2001). Unintended consequences: Organizational prac-tices and their impact on workplace safety and productivity. Journalof Occupational Health Psychology, 6, 127–138.

Kipnis, D. (1993). Unanticipated consequences of using behaviortechnology. Leadership Quality, 4, 149–171.

Klar, Y., & Giladi, E. E. (1999). Are most people happier than theirpeers, or are they just happy? Personality and Social PsychologyBulletin, 25, 585–594.

Koehler, D. J. (1991). Explanation, imagination, and confidence injudgment. Psychological Bulletin, 110, 499–519.

Kruger, J., & Burrus, J. (2004). Egocentrism and focalism in unrealisticoptimism (and pessimism). Journal of Experimental Social Psy-chology, 40, 332–340.

McBriarty, M. A. (1988). Performance appraisal: Some unintendedconsequences. Public Personnel Management, 17, 421–434.

McKinley, W., & Scherer, A. G. (2000). Some unanticipated conse-quences of organizational restructuring. Academy of ManagementReview, 25, 735–752.

Meddin, B. (1986). Unanticipated and unintended consequences of thegroup process: Enhancing successful group functioning. SocialWork with Groups, 9, 83–92.

Merton, R. (1936). The unanticipated consequences of purposive socialaction. American Sociological Review, 1, 894–904.

Nisbett, R. E., & Wilson, T. D. (1977). Telling more than we canknow: Verbal reports on mental processes. Psychological Review,84, 231–259.

Oyer, P., & Schaefer, S. (2003, Summer). The unintended consequencesof the ‘91 civil rights act. Regulation, 42–47.

Perrow, C. (1999). Normal accidents: Living with high-risk technologies.New York: Basic Books.

Reason, J. T. (1987). The Chernobyl errors. Bulletin of the BritishPsychological Society, 40, 201–206.

Ross, M. (1989). Relation of implicit theories to the construction ofpersonal histories. Psychological Review, 96, 341–357.

Schelling, T. C. (1971). Dynamic models of segregation. Journal ofMathematical Sociology, 1, 143–186.

Schelling, T. C. (1972). A process of residential segregation: Neighbor-hood tipping. In A. H. Pascal (Ed.), Racial discrimination ineconomic life (pp. 157–184). Lexington, MA: D.C. Heath.

Schkade, D. A., & Kahneman, D. (1998). Does living in Californiamake people happy? A focusing illusion in judgments of life satisfac-tion. Psychological Science, 9, 340–346.

Shafir, E. (1993). Choosing versus rejecting: Why some options areboth better and worse than others. Memory & Cognition, 21,546–556.

Simberloff, D., & Stiling, P. (1996). Risks of species introduced forbiological control. Biological Conservation, 78, 185–192.

Storms, M. D. (1973). Videotape and the attribution process: Revers-ing actors’ and observers’ points of view. Journal of Personality andSocial Psychology, 27, 165–175.

Tenner, E. (1997).Why things bite back: Technology and the Revenge ofunintended consequences. New York: Vintage.

Thornton, M. (1998). The potency of illegal drugs. Journal of DrugIssues, 28, 725–740.

Tversky, A. (1977). Features of similarity. Psychological Review, 84,327–352.

Wilson, T. D.,Wheatley, T., Meyers, J. M., Gilbert, D. T., &Axsom, D.(2000). Focalism: A source of durability bias in affective forecasting.Journal of Personality and Social Psychology, 78, 821–836.

Windschitl, P. D., Kruger, J., & Simms, E. N. (2003). The influence ofegocentrism and focalism on people’s optimism in competitions:When what affects us equally affects me more. Journal of Person-ality and Social Psychology, 85, 389–408.

68 EHRLINGER AND EIBACH

Downloaded By: [Ehrlinger, Joyce] At: 04:33 25 February 2011