Dunn, Wilson, & Gilbert (2003)

pdf

School

Simon Fraser University *

*We aren’t endorsed by this school

Course

SOCIAL PSY

Subject

Psychology

Date

Oct 30, 2023

Type

pdf

Pages

Uploaded by mkd16

10.1177/0146167203256867 ARTICLE PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN Dunn et al. / MISPREDICTING SATISFACTION Location, Location, Location: The Misprediction of Satisfaction in Housing Lotteries Elizabeth W. Dunn Timothy D. Wilson University of Virginia Daniel T. Gilbert Harvard University People tend to overestimate the emotional consequences of future life events, exhibiting an impact bias. The authors replicated the impact bias in a real-life context in which undergraduates were randomly assigned to dormitories (or “houses”). Participants appeared to focus on the wrong factors when imagining their future happiness in the houses. They placed far greater weight on highly variable physical features than on less variable social fea- tures in predicting their future happiness in each house, despite accurately recognizing that social features were more important than physical features when asked explicitly about the determi- nants of happiness. In Experiment 2, we found that this discrep- ancy emerged in part because participants exhibited an isolation effect, focusing too much on factors that distinguished between houses and not enough on factors that varied only slightly, such as social features. Keywords: affective forecasting; impact bias; predictors of happiness; emotion; isolation effect I t is safe to say that chocolate cake, sunny spring days, and true love are more likely to produce happiness than are mud pies, freezing rain, and bad breakups. Yet, even if people accurately recognize what factors will lead to happiness, they may fail to apply this knowledge in imag- ining how they will feel in the future. For example, con- sider the case of a high school senior touring two univer- sities that differ greatly in terms of setting (urban vs. rural), size, and gender ratios but are very similar in terms of dorm life, available extracurriculars, and oppor- tunities for contact with professors. In predicting how happy he or she would be at each university, the rational student should weight the quality of each aspect (e.g., availability of extracurriculars) by its importance to his or her happiness. Yet, we would argue that the student may inadvertently place undue weight on those factors that vary a great deal across universities while placing lit- tle or no weight on important factors that vary less across options in imagining his or her future well-being at each school. This prediction follows from Tversky’s (1972) elimi- nation-by-aspects theory of choice. According to the the- ory, people simplify choices between options by cancel- ing out and disregarding features that are shared across options, a tendency Kahneman and Tversky (1979) term the “isolation effect” (see also Houston & Sherman, 1995; Houston, Sherman, & Baker, 1991). Applying this theory, Hodges (1997) asked participants to choose between three apartments, including two that shared several very positive features but also had unique nega- tive features, as well as one that had a unique set of posi- tive and negative features. Although the two apartments with shared features were more attractive, on balance, than the third apartment, participants exhibited a pref- erence for the latter because they cancelled out the posi- tive, shared features of the first two apartments, paying attention primarily to the unique, negative features. 1421 Authors’ Note: This research was supported in part by a National Sci- ence Foundation Graduate Research Fellowship to the first author and Grant R01-MH56075 from the National Institute of Mental Health to the second and third authors. We thank Jack McArdle for statistical ad- vice and Kevin Carlsmith, Gerald Clore, and Brian Malone for com- ments on an earlier draft. The article is based on the first author’s master’s thesis project. Portions of these results were presented at the 2001 meeting of the Society for Personality and Social Psychology, San Antonio, Texas. Correspondence concerning this article should be ad- dressed to Elizabeth W. Dunn, Department of Psychology, University of Virginia, 102 Gilmer Hall, P.O. Box 400400, Charlottesville, VA 22904- 4400; e-mail: edunn@post.harvard.edu. PSPB, Vol. 29 No. 11, November 2003 1421-1432 DOI: 10.1177/0146167203256867 © 2003 by the Society for Personality and Social Psychology, Inc.

Focusingonunique featureswhile cancelingoutshared features may be a reasonable and efficient strategy for choosing between options. For example, it would be sen- sible for the college-bound student to pay attention pri- marily to features that varied a great deal across Univer- sity A and University B in deciding which school he or she would rather attend. But it would surely be unwise for the student to conclude that he or she would be bliss- fully happy at University A and miserable at University B while neglecting important factors that were similar across these universities. To the extent that people overestimate differences in the emotional consequences of various outcomes due to ignoring shared features, they are likely to enjoy desir- able outcomes less than they expect while reaping unan- ticipated happiness from less desirable outcomes. This tendency may help to explain why individuals are often woefully inaccurate in predicting how various outcomes will affect their happiness (Buehler & McFarland, 2001; Gilbert, Driver-Linn, & Wilson, 2002; Gilbert, Lieberman, & Wilson, in press; Gilbert, Pinel, Wilson, Blumberg, & Wheatley, 1998; Gilbert & Wilson, 2000; Loewenstein & Schkade, 1999; Mitchell, Thompson, Peterson, & Cronk, 1997; Sieff, Dawes, & Loewenstein, 1999; Wilson & Gilbert, in press; Wilson, Wheatley, Meyers, Gilbert, & Axsom, 2000). People frequently overestimate the duration and intensity of their own emotional responses to future events and outcomes, a tendency that Gilbert et al. (2002) term the “impact bias.” In everyday life, people often engage in affective forecasting when envisioning competing possibilities (e.g., receiving tenure vs. not, liv- ing in the suburbs vs. the city, ordering foie gras vs. frog legs). As an important artifact of this process, features that differ between competing options may drive fore- casts, whereas features that are relatively constant take a backseat. In a similar vein, Wilson et al. (2000) showed thatindi- viduals exhibit the impact bias in part because they over- look the fact that the emotional impact of any one event will be mitigated by other events, such as snowstorms, parties, and the other ups and downs of daily life, an oversight Wilson et al. (2000) termed “focalism” (see also Schkade & Kahneman, 1998). We suggest that the isolation effect may produce a different kind of focusing bias, in which individuals ignore the affective conse- quences of important features that are shared across potential outcomes, leading to exaggerated expecta- tions regarding differences in future well-being associ- ated with one outcome versus another. The housing system at a major university provided us with an opportunity to conduct an initial test of this hypothesis. In the spring of their 1st year, students at the university are randomly assigned to spend the subse- quent 3 years of college living in 1 of 12 dormitories (or “houses”), with each serving as a kind of college within the college. Each 1st-year student enters the housing lot- tery along with the roommates he or she has chosen as well as a group of up to 15 other friends (“blockmates”) who are automatically assigned to the same house. Thus, by the time students enter the housing lottery, they know the group of people whom they will live with the follow- ing year. In addition, most 1st-year students are familiar with the location and other physical characteristics of the houses as well as the individual “character” of each house (e.g., one house is known for its spirited intramu- ral teams, another for its parties, another for its historical tolerance of gays and other stigmatized groups). The houses vary a great deal in perceived quality, and most students anticipate that housing assignment will strongly influence their happiness; 1st-year students often stay up all night awaiting their housing assignment and can be seen jumping up and down or accepting consolation from an upper-class student after receiving news that they have been assigned to a desirable or undesirable house. Thus, the housing system provides a natural experiment in that students are randomly assigned to a good or bad outcome that they consider extremely important. Whereas physical features (e.g., location) differ greatly across houses, social aspects of house life (e.g., sense of community) vary less across the houses. Indeed, because one enters the housing lottery with a group of roommates and blockmates, the quality of one’s rela- tionships with these close others should be roughly equivalent regardless of house assignment. Typically, social relationships are a critical determinant of happi- ness (Argyle, 1999; Biswah-Diener & Diener, 2001; Diener, Gohm, Suh, & Oishi, 2000; Diener, Suh, Lucas, Smith, 1999; Larson, 1990; Myers, 1999; Sheldon, Elliot, Kim, & Kasser, 2001); indeed, Diener and Seligman (2002) speculated that perhaps “good social relation- ships are, like food and thermoregulation, universally important to human mood” (p. 83). When asked directly about the determinants of happiness, most people seem well aware of the critical role of close relationships (Campbell, Converse, & Rodgers, 1976; Freedman, 1978; Loewenstein & Schkade, 1999; Pettijohn & Pettijohn, 1996). For example, respondents at Pittsburgh International airport rated family life and friends as most important to happiness while rating income as less important (Loewenstein, 1996, described in Loewenstein & Schkade, 1999). Thus, we predicted that social aspects of house life would be related to actual happiness and thatparticipants would recognize the crit- ical importance of social features when explicitly asked about determinants of happiness. Yet, we anticipated that when predicting their future happiness, students would pay little attention to the 1422 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

quality of social features that they expected to find in the houses because social features varied less than physical features across houses. In other words, we predicted that the tendency to focus on highly variable features (i.e., theisolationeffect)wouldoverwhelmparticipants’explicit recognition that less variable social features are critical to happiness. In two experiments, we examined whether partici- pants overestimated how happy they would be living in a desirable house as well as how unhappy they would be liv- ing in an undesirable house, thereby exhibiting the impact bias. To the best of our knowledge, this repre- sents the first attempt to demonstrate the impact bias using amultiyear longitudinal designwithrandom assign- ment to a major life outcome. In addition, we attempted to demonstrate a mechanism for the impact bias that has not previously been investigated, namely, the isolation effect.As discussed, we predicted thatparticipants would underestimate the importance of features that the houses had in common (social relationships), even though they knew, in the abstract, that these features would be impor- tant to their happiness. We predicted that participants would overestimate the importance of features that var- ied across houses (physical features). EXPERIMENT 1 Method Overview . In the spring of their 1st year (Time 1), shortly before they learned which house they would live in for the subsequent 3 years, college students predicted how happy they would be 1 year later if they were living in each of 12 houses. One and 2 years later (Times 2 and 3), we asked participants to report their actual happiness. We predicted thatpeople would overestimate how happy theywould be if theywere assigned a desirable house and underestimate how happy they would be if they were assigned an undesirable house. Furthermore, people were expected to base their forecasts more on physical features that varied greatly between houses and less on social features that varied relatively little, even though social features would be more related to actual happi- ness than physical features. Participants . At Time 1, 174 1st-years completed our initial questionnaire. Of these, 118 completed the Time 2 survey as sophomores and 84 completed the Time 3 survey as juniors. Time 1 survey . The initial survey asked participants to predict how happy they would be overall at the same time next year if they were living in each of the 12 houses on a scale ranging from 1 ( unhappy ) to 7 ( happy ). Similar one-item measures have been used successfully in other affective forecasting studies (e.g., Gilbert et al., 1998) and possess acceptable psychometric properties (Diener et al., 1999; Fordyce, 1988). In addition, participants esti- mated the extent to which 10 features related to house life (e.g., location, sense of community) would influence their happiness with the house on a scale ranging from 1 ( will have no influence on myhappiness ) to 7 ( will have a large impact on my happiness ). Six of these features were physi- cal qualities of the house, whereas four of the features were related to the quality of social life in the house (see Table 1). All participants completed the survey approxi- mately 1 to 2 weeks before they were randomly assigned to the house where they would spend their sophomore, junior, and senior years. Time 2 survey . One year after completing the initial survey, participants were asked to report their overall happiness on the same 7-point scale used in the first sur- vey. 1 Participants also reported how much each of the physical and social features of house life had influenced their happiness on a scale corresponding to that used at Time 1. In addition, participants were instructed to rate the objective quality of these 10 features of life in their house on a7-point scale (1= poor ,4= average ,7= excellent ). Time 3 survey . Two years after the initial survey, partici- pants again reported their happiness and provided objective ratings of the quality of the 10 features of house life on the same scales used at Time 2. All surveys were conducted via e-mail and were completed between March 19 and March 27 of each year (except for two par- ticipants who completed the Time 2 survey during the first week of April). Results and Discussion Impact bias . Across the sample as a whole, participants’ forecasted happiness was significantly (although not highly) correlated with their overall happiness a year later, r (109) = .26, p < .006. Fifty-three participants were assigned to houses in which they expected to be below their personal mean in predicted happiness (undesir- able house group), whereas 56 participants were assigned to houses where they expected to be at or above their personal mean (desirable house group). 2 Partici- pants who were assigned to houses they considered undesirable were significantly happier overall ( M = 5.38, SD = 1.16) at Time 2 than they had expected to be ( M = 3.43, SD = 1.5), t (52) = 10.71, p < .0001, d = 1.49. In con- trast, participants who were assigned to desirable houses were significantly less happy overall ( M = 5.45, SD = 0.92) than they had expected ( M = 5.96, SD = 0.85), t (55) = –3.04, p < .01, d = 0.41. Consistent with previous research, then, we found that participants overestimated how much an important life event would affect their happi- ness. Indeed, participants assigned to desirable houses Dunn et al. / MISPREDICTING SATISFACTION 1423

Your preview ends here

Eager to read complete document? Join bartleby learn and gain access to the full version

Access to all documents
Unlimited textbook solutions
24/7 expert homework help

were not significantly happier than participants assigned to undesirable houses, t < 1. Features related to actual happiness . To examine whether forecasters go astray partly because they focus on the wrong factors in imagining their emotional futures, we first sought to demonstrate that the quality of social fea- tures in the houses predicted participants’ actual happi- ness. We summed participants’ Time 2 ratings of the objective quality of the four social features in their house and their ratings of the objective quality of the six physi- cal features in their house to create social and physical quality composites, which we centered at 0 by subtract- ing the composite means. We entered the composite quality of social features ( α = .50) and the composite quality of physical features ( α = .55) into a regression predicting happiness at Time 2; in this analysis and sub- sequent analyses, we first examined the main effects of social and physical quality in a simultaneous regression, and then the interaction term was added to the model in a second step. There was a marginally significant rela- tionship between overall happiness and the quality of social features, t (114) = 1.66, β = .16, p = .10, whereas overall happiness was unrelated to the quality of physical features, t (114) = 0.66, β = .06, p = .51, or to the interac- tion of social and physical feature quality, t (113) = –1.24, β = –.12, p = .22. We examined these relationships again the following year using Time 3 ratings of physical and social features to predict happiness at Time 3. The quality of social fea- tures was again positively related to overall happiness, t (77) = 2.75, β = .35, p < .01. Unlike at Time 2, the quality of physical features was negatively related to overall hap- piness, t (77) = –2.27, β = –.29, p < .03. 3 The two-way inter- action of physical and social features was nonsignificant, t (76) = –0.42, β = –.05, p = .67. Thus, across their sopho- more and junior years, participants’ happiness was con- sistently related to their ratings of social life in the house. In contrast, the relationship between physical features and happiness was nonsignificant or in the opposite to the expected direction. Features related to forecasted happiness . Do forecasters recognize the relationship between social features and their future happiness? At Time 1, when asked directly about the determinants of their happiness, forecasters predicted that the four social features would have a slightly greater impact on their happiness ( M = 4.91, SD = 1.10) on average than would the six physical features ( M = 4.67, SD = 0.93), t (104) = 2.33, p < .02. Yet, in imagining how happy they would be in specific houses, forecasters might overlook the role of social fea- tures. To explore this possibility, we examined the rela- tionship between participants’ affective forecasts at Time 1 and their ratings of the physical and social fea- tures of their house at Time 2 (recall that participants did not rate these features atTime 1).We entered Time 2 ratings of physical quality and social quality, followed by the two-way interaction, into a regression predicting forecasted happiness at Time 1 for the house to which the participant was actually assigned. Ratings of the phys- ical quality of the house at Time 2 were strongly related to forecasted happiness, t (106) = 4.32, β = .39, p < .0005, suggesting that participants were aware of the physical characteristics of the houses and relied heavily on this information in predicting their future happiness.Incon- trast, ratings of social features at Time 2 were unrelated to forecasted happiness, t (106) = 0.26, β = .02, p = .80, suggesting that participants were unable to predict the quality of social features they would experience in the house or placed little weight on this information in fore- casting their happiness. The two-way interaction was nonsignificant, t (105) = –1.49, β = –.14, p = .14. The overemphasis placed on physical features helps to account for the discrepancy between participants’ affective forecasts and experiences. We entered partici- pants’ Time 2 ratings of social and physical features into a regression predicting the difference between actual and forecasted happiness and then added the two-way interaction to the model. Participants who reported high satisfaction with the physical features of their house exhibited relatively low overall happiness compared to their forecasts, t (106) = –3.96, β = –.37, p < .0005, suggest- ing that physical features did not deliver the degree of happiness participantshadanticipated;neither the qual- ity of social features, t (106) = 0.81, β = .08, p = .42, nor the two-way interaction, t (105) = 0.34, β = .03, p = .74, signifi- cantly predicted the difference between overall happi- ness and predicted happiness. Summary of results . As expected, we found strong sup- port for the impact bias; participants overestimated how happy they would be in desirable houses and how miser- able they would be in undesirable houses. Our results suggest that forecasters may have erred by focusing on physical features such as location while virtually ignoring the quality of social life in the houses. When our 1424 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN TABLE 1: Physical and Social Features Relevant to House Life Physical Features Social Features Location Relationships with roommates Attractiveness Relationships with blockmates (friends who enter housing lottery together) Room size Relationships with tutors (graduate/ faculty advisers who live or eat in the house) House size Sense of community in the house Facilities (e.g., for athletics, arts, etc.) Dining hall

participants were sophomores and juniors, their ratings of house social life were reliably related to their actual happiness, whereas their ratings of house physical fea- tures did not consistently predict happiness. Yet, the affective forecasts they made as 1st-years were strongly related to the quality of physical features and unrelated to the quality of social life they later found in the houses. Of interest, participants accurately reported that social features would strongly influence their happiness when they were explicitly asked about the determinants of happiness. Thus, although quality of social life was related to happiness and participants recognized that this would be the case, participants’ forecasts were unre- lated to the quality of social life they found in the houses. Instead, forecasts were driven by physical qualities of the houses, which failed to deliver the lasting pleasure par- ticipants anticipated. Because the Time 1 survey did not ask participants to predict the quality of social and physical features in all 12 houses, we could only examine the relationship between Time 2 ratings of social and physical quality and Time 1 affective forecasts. It is possible, then, that participants estimated the quality of social life they would find the fol- lowing year and used these estimates in their affective forecasts, but thattheir estimates were highly inaccurate, thereby eliminating any relationship between forecasts and the quality of social life they actually experienced. To tackle this potential problem, we asked a new sample of 1st-years to predict the quality of each of the social and physical features they expected to find in each of the 12 houses. In addition, we tested the isolation effect more directly by asking some participants to think about fea- tures that would be the same across houses immediately before forecasting. We predicted that this manipulation would reduce the isolation effect, thereby increasing the weight participants placed on social features. Finally, Study 2 addressed a potential alternative explanation of the results of Study 1, namely, that there was insufficient variance in people’s ratings of social features to allow for correlations with predicted or actual happiness. EXPERIMENT 2 Method Time 1 survey . We gave a longer version of the Time 1 survey used in the first experiment to a new sample of 144 1st-year participants at the same university, who com- pleted it approximately 1 to 2 weeks before receiving their housing assignment. In addition to predicting how happy they would be 1 year later if they were living in each of the 12 houses, participants predicted how much the 10 features of house life examined in Study 1 would influence their happiness. Next, participants rated the objective quality of each of the physical and social fea- tures they expected to find in each of the 12 houses. For example, in considering a given house, they rated how good the location would be and how good the sense of community would be. They were asked to write “dk” or “don’t know” if they felt they could not predict how good a certain feature would be in a given house; otherwise, all of the scales were equivalent to those used in the previ- ous study. Finally, participants rated how much each of the features would vary across different houses on a 7- point scale (1 = would not vary at all , 7 = would vary a great deal ) as well as how easily they could predict the quality of each feature (1 = very difficult to predict , 7 = very easy to pre- dict ) and how easily they could visualize each feature (1 = very difficult to visualize , 7 = very easy to visualize ). Manipulation of isolation effect . Immediately prior to completing the survey, a subset of participants was ran- domly assigned to answer questions designed to manipu- late their focus on features that varied relatively little across houses; these participants were asked to write about aspects of the houses and house life that would be “pretty much the same” across houses immediately before forecasting. To reduce potential demand charac- teristics, we included a separate question that asked par- ticipants to write about aspects that would “vary a great deal” across houses. One group of participants ( N = 28) first wrote about what would vary and then about what would be the same across houses (same-last condition). We expected these participants to place increased weight on social features because they thought about aspects of house life that would be similar across houses immediately before forecasting. Another group of participants ( N = 40) answered the two questions in the reverse order (vary-last condition), such that they thought about highly variable aspects of house life immediately before forecasting. Control par- ticipants ( N = 76) were not asked any questions before completing the main part of the survey. If participants focused on highly variable features by default, then par- ticipants in both the control and the vary-last conditions should place greater weight on physical features than social features in making affective forecasts, whereas par- ticipants in the same-last condition should show a reduc- tion in this bias. Time 2 survey . One year after completing the initial survey, participants were asked to complete the same questions used at Time 2 in Experiment 1; participants rated their actual happiness as well as the affective impact and objective quality of the various physical and social features of their house. A total of 90 participants returned the Time 2 survey (63% of original sample). Dunn et al. / MISPREDICTING SATISFACTION 1425

Results and Discussion Replication of Study 1 . As in Study 1, we found strong support for the impact bias. Participants who were assigned to houses they considered undesirable were sig- nificantly happier overall ( M = 5.0, SD = 1.11) than they had expected to be ( M = 3.82, SD = 0.86), t (35) = 5.26, p < .001, d = 0.89, whereas participants who were assigned to desirable houses were less happy ( M = 5.41, SD = 1.35) than they had anticipated ( M = 6.02, SD = 0.94), t (51) = 2.81, p < .007, d = 0.39. Although participants in desirable houses reported somewhat greater overall happiness than participants in undesirable houses, this difference did not reach significance, t (86) = 1.52, p = .13, d = 0.33. Consistent with Study 1, there was a modest correlation between forecasted happiness and actual happiness, r (88) = .19, p < .08. Replicating Study 1, participants predicted that the quality of social features in the house would influence their happiness ( M = 5.35, SD = 0.82) more than the qual- ity of physical features would ( M = 4.92, SD = 0.85) when they were asked directly about determinants of happi- ness, t (141) = 5.71, p < .0001. However, we predicted that participants’ affective forecasts would actually be driven by the quality of physical features in the house rather than by the quality of social life participants anticipated. For each house, we entered ratings of physical quality (mean α = .81) and social quality (mean alpha = .84) into a simultaneous regression predicting forecasted happi- ness and then added the two-way interaction to the model. In this way, we conducted 12 separate between- subjects regressions (one for each house). Across the 12 houses, there were no significant positive relationships between participants’ ratings of social features and their forecasted happiness; for one house, there was a margin- ally significant negative relationship between antici- pated social quality and forecasted happiness (see Table 2). 4 In contrast, significant or marginally significant posi- tive relationships emerged between ratings of physical features and affective forecasts for 10 of the 12 houses. The two-way interaction was nonsignificant for 11 of the 12 houses. In line with our hypotheses, then, partici- pants placed a great deal of weight on physical features of the houses and neglected social features entirely in imagining their future happiness—despite recognizing on an explicit level that social features mattered more than physical features. Manipulation of isolation effect . We expected that asking participants to consider features that would be similar across houses would lead them to think about social fea- tures, whereas asking participants to think about fea- tures that would vary across houses would lead them to think about physical features. As a check on this manipu- lation, three research assistantswho were unaware of our hypotheses coded participants’ responses to the ques- tions regarding what would vary and what would be the same across houses. For each question (vary and same), the coders noted whether participants mentioned each of the social and physical features of the houses. The coders showed considerable agreement, with effective reliabilities of .90, .88, .89, and .88 for ratings of the num- ber of times participants mentioned physical features under the vary question, social features under the vary question, physical features under the same question, and social features under the same question, respec- tively. Averaging coders’ ratings, we found that partici- pants mentioned physical features ( M = 1.88, SD = 1) sig- nificantly more than social features ( M = 0.48, SD = 0.67) when asked what would vary across houses, t (67) = 9.22, p < .001, and mentioned social features ( M = 0.69, SD = 0.74) significantly more than physical features ( M = 0.42, SD = 0.60) when asked what would be similar across houses, t (67) = –2.36, p < .02, as expected. To examine whether our manipulation influenced the weight participants placed on social features, we used a multilevel linear model, an approach thatallowed us to include all of our data in a single powerful analysis while accounting for the lack of independence between a single individual’s ratings of the 12 houses (Bryk & Raudenbush, 1992; Kashy & Kenny, 2000; Kenny, Kashy, & Bolger, 1998). Although our model used continuous predictors, this analysis is analogous to a between-within ANOVA with repeated measures on one factor and a between-subjects manipulation on another factor. Multi- level models are increasingly used in repeated-measures designs in which each participant provides ratings on the dependent and independent variable at a series of time points; in such models, time is the lower-level unit nested 1426 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN TABLE 2: Summary of Regression Analyses Predicting Forecasted Happiness From Ratings of Social and Physical Qualities for Each House Social Physical Social × Quality Quality Physical House N β t β t β t House 1 40 –.13 –.80 .64 3.96*** –.44 –2.79** House 2 31 –.07 –.28 .53 2.02** .09 0.49 House 3 30 –.24 –.73 .66 2.00* .02 0.07 House 4 41 .10 .56 .66 3.66*** .19 1.37 House 5 35 –.07 –.22 .45 1.44 .40 1.92 House 6 32 –.26 –1.30 .66 3.30** –.23 –1.39 House 7 30 –.58 –1.74* 1.13 3.36** .31 1.81 House 8 36 –.11 –.41 .60 2.30** .31 1.48 House 9 38 –.09 –.59 .82 5.09*** .18 1.30 House 10 28 –.29 –.90 .93 2.90** .10 0.51 House 11 33 –.08 –.29 .70 2.54** .03 0.19 House 12 30 .00 .00 .58 1.58 .34 1.70 * p ≤ .10. ** p ≤ .05. *** p ≤ .001.

Your preview ends here

Eager to read complete document? Join bartleby learn and gain access to the full version

Access to all documents
Unlimited textbook solutions
24/7 expert homework help

within person, the upper-level unit. Our data can be thought of as similar to a repeated-measures design, with each house providing a measurement point. In our model, house is a lower-level unit nested within partici- pant, the upper-level unit. Condition represents an upper-level variable because the manipulation was at the level of the participant. Affective forecasts (the depend- ent variable) and ratings of physical and social quality are lower-level variables measured at the house level. We entered the upper-level variable of condition and the lower-level variables of anticipated social quality and anticipated physical quality, along with the interactions of these three variables, into a multilevel linear model predicting forecasted happiness (measured at the lower level of house). In essence, this multilevel model (a) cre- ated a regression equation predicting forecasted happi- ness from physical and social quality ratings for each par- ticipant across houses, (b) combined these equations to create an aggregate, and (c) compared whether the regression coefficients differed by condition. Inline with our expectations, preliminary analyses revealed that the vary-last condition did not differ from the control condi- tion; thus, we collapsed across these two conditions, which we contrasted with the same-last condition. The multilevel model revealed a very strong positive relationship between anticipated quality of physical features and forecasted happiness, t (336) = 12.02, B = 0.23, p < .001, whereas the anticipated quality of social features was unrelated to forecasted happiness, t (336) = –0.70, B = –0.03, p = .49 (see Table 3). 5 The difference between the weight placed on physical versus social fea- tures may have been magnified in this analysis because there was much greater variability in any one partici- pant’s ratings of physical features (mean SD = 4.4) than social features (mean SD = 1.76) across houses. Because of this statistical artifact, the relative size of these main effects should not be overinterpreted, but this finding is at least consistent with the 12 between-subjects regres- sions showing that participants’ affective forecasts were driven by the quality of the house’s physical features rather than by social features. The only other significant effect was the expected two-way Social Quality × Condition interaction, t (336) = 2.59, B = 0.08, p < .01. We plotted the mean forecasted happiness predicted by the model at 1 SD above and below the mean social quality rating for participants in the same-last condition and the control/vary-last condi- tion based on standard guidelines (Aiken & West, 1991). As seen in Figure 1, this interaction demonstrated that participants placed greater weight on the quality of social features if they were asked to think about features of house life that would be similar across houses immedi- ately before forecasting. This finding, we should note, is inconsistent with the interpretation that there was too little variance in ratings of social factors to allow for predict- ability. People in the same-last condition, as expected, did appear to place greater weight on social factors than other participants when making their forecasts. Although thismodel provides support for our hypoth- eses, it is based on only 62 out of our 144 participants (43%). Many participants had missing data for the social and physical composite variables because they failed to predict the quality of at least one of the house features, most likely due to the fact that some of the features (e.g., tutor quality) were relatively difficult to predict (recall that participants were given the option to say “don’t know”). To compensate for this shortcoming, we con- ducted an additional analysis in which we substituted anticipated roommate quality for the social quality com- posite and anticipated quality of location for the physical quality composite. Roommate quality was rated as both Dunn et al. / MISPREDICTING SATISFACTION 1427 TABLE 3: Multilevel Model Predicting Forecasted Happiness From Condition and Social and Physical Composites B df t Fixed effects Physical quality .23 336 12.02*** Social quality –.03 336 –0.70 Condition .14 60 1.38 Social Quality × Condition .08 336 2.59** Physical Quality × Condition –.02 336 –0.94 Social Quality × Physical Quality –.01 336 –1.30 Physical Quality × Social Quality × Condition –.01 336 –1.60 Random effects Intercept (individual differences in forecasts) .27 2.4** Physical quality .00 0.53 Social quality .01 1.49 Residual 1.06 11.89*** NOTE: Total observations = 1,728; observations used = 404; R 2 = .54. * p ≤ .10. ** p ≤ .05. *** p ≤ .001. 3 4 5 6 7 Low High Quality of Social Features Forecasted Happiness Same-last Vary-last & Control Figure 1 Relationship between social features composite and fore- casted happiness by experimental condition.

the easiest to predict and the most important of the social features, and location was rated as both the easiest to predict and the most important of the physical fea- tures. By using these flagship variables as proxies for social and physical qualities, we were able to include 121 out of 144 participants (84%) in our analysis. This analysis revealed a very strong relationship between anticipated location quality and forecasted hap- piness, t (1,163) = 21.37, B = 0.52, p < .001, as well as a smaller relationship between anticipated roommate quality and expected happiness, t (1,163) = 5.29, B = 0.32, p < .001 (see Table 4). Consistent with the analysis using composite variables, we found a significant Roommate Quality × Condition interaction, t (1,163) = 2.27, B = 0.11, p < .05; participants in the same-last condition placed greater weight on anticipated roommate quality than did participants in the control and vary-last conditions (see Figure 2). We also found a significant three-way Location Quality × Roommate Quality × Condition interaction, t (1,163)= 2.45, B = 0.04, p < .05. 6 Plotting this interaction revealed that participants in the same-last condition consistentlyused roommate quality in predict- ing their future happiness, whereas the other partici- pants placed weight on roommate quality only when location was relatively good. Although our first model combined all the relevant variables and included 43% of the sample and our sec- ond model used only the flagship variables and included 84% of the sample, results across the two analyses were consistent. In both models, we found that participants placed greater weight on physical features than social features in forecasting their future happiness. More important, we found that participants increased their use of social features if they were asked to consider aspects that would be similar across houses immediately before forecasting. This provides support for our argu- ment that participants gave low weight to social features in making affective forecasts because of the isolation effect. Our analyses, of course, rest on the assumption that participants viewed physical features as more vari- able than social features. This assumption was strongly supported; participants reported thatthe quality of phys- ical features would vary much more across houses ( M = 5.33, SD = 0.74) than would the quality of social features ( M = 3.05, SD = 0.97), t (116) = 22.61, p < .0001. By manipulating people’s focus on variables that would vary little across houses, we found support for the isolation effect in affective forecasting. It is possible, of course, that other factors also contributed to people’s tendency to underweight social features. Because of the drive for certainty, forecasters may place excessive weight on features that can be predicted with confidence. Par- ticipants reported that the quality of physical features was easier to predict ( M = 5.28, SD = 0.87) than the qual- ity of social features ( M = 3.47, SD = 1.35), t (138) = 13.30, p < .0001, a difference that may have increased forecast- ers’ tendency to attend to physical features over social features. Features related to actual happiness . If social features were completely unpredictable, participants would be entirely justified in ignoring these features when making forecasts. Although plausible, this interpretation is not supported by the data; participants’ ratings of the actual social quality of their assigned house at Time 2 corre- lated strongly with their predicted ratings of social qual- ity for that house at Time 1, r (21) = .60, p < .004, and this correlation was similar in magnitude to the correlation between predicted and actual physical quality ratings, r (32) = .52, p < .003. 1428 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN TABLE 4: Multilevel Model Predicting Forecasted Happiness From Condition, Anticipated Roommate Quality, and Antici- pated Location Quality B df t Fixed effects Location quality .52 1163 21.37*** Roommate quality .32 1163 5.29*** Condition –.05 119 –0.93 Roommate Quality × Condition .11 1163 2.27* Location Quality × Condition –.01 1163 –0.49 Roommate Quality × Location Quality .05 1163 2.46* Location Quality × Roommate Quality × Condition .04 1163 2.45* Random effects Intercept (individual differences in forecasts) .28 4.02*** Location .03 3.32*** Roommate .10 2.24* Residual 1.04 22.54*** NOTE: Total observations = 1,728; observations used= 1,290; R 2 = .55. * p ≤ .10. ** p ≤ .05. *** p ≤ .001. 3 4 5 6 7 Low High Quality of Roommate Relationship Forecasted Happiness Same-last Vary-last & Control Figure 2 Relationship between roommate quality and forecasted happiness by experimental condition.

More important, overall happiness at Time 2 was better predicted by anticipated quality of social features than by anticipated quality of physical features at Time 1. We entered Time 1 ratings of anticipated social quality and physical quality into a regression predicting overall happiness at Time 2 and then entered the two-way inter- action into a second model. Anticipated social quality at Time 1 was marginally related to overall happiness at Time 2, t (15) = 1.78, β = .65, p < . 10, whereas predicted physicalqualitywasunrelatedtooverallhappiness, t (15)= –0.17, β = –.06, p = .87. The two-way interaction was nonsignificant, t < 1. Unfortunately, these analyses included only 18 of the 90 participants who completed both surveys, due pri- marily to missing data for the Time 1 social and physical composite variables. Thus, we conducted an additional analysis in which we substituted anticipated roommate quality for the social quality composite and anticipated quality of location for the physical quality composite, which allowed us to include 65 out of 90 participants (72%) in our analyses. We found that anticipated room- mate quality significantly predicted actual happiness, t (62) = 3.12, β = .37, p < .003, whereas anticipated loca- tion quality did not significantly predict actual happi- ness, t (62) = 1.2, β = .14, p = .24. The two-way interaction was not significant, t < 1. Thus, anticipated social quality was at least as reliable a predictor of future overall happiness as anticipated physical quality. Another indication that participants in Experiment 2 made valid predictions of social feature quality is that these ratings were correlated with Experi- ment 1 participants’ ratings of the actual quality of tutors and sense of community in the houses, r (11) = .50, p < .12; in other words, 1st-years’ perceptions of the social quality of the houses corresponded with sophomores’ experiences in the houses. These findings cast doubt on the interpretation that forecasters simply ignored social features because their quality wasimpossible to predict. Furthermore, it is important to note that our analyses examining the relationship between forecasted happi- ness and forecasted feature quality relied on those par- ticipants who felt that they could predict the quality of house features. The large number of “don’t know” responses we received suggests that our survey enabled participants to admit ignorance when appropriate. We find it compelling that even participants who felt that they could make reasonable predictions about the qual- ity of social features ignored this information in making affective forecasts. Still, the ease of predicting physical feature quality relative to social feature quality may have contributed to participants’ tendency to overweight the latter and underweight the former. Future research may confirm the role of subjective predictability in moderat- ing the extent to which various features influence fore- casts, but the present research provides the strongest evi- dence for the role of perceived variability in moderating the relative influence of various features; participants placed excessive weight on highly variable physical fea- tures relative to less variable social features, and this ten- dency was attenuated when participants were led to think about features that would be the same across houses immediately before forecasting. One might argue that ratings of social features were unrelated to forecasted happiness because of a statistical artifact: A relationship between social ratings and fore- casted happiness could not emerge if there was insuffi- cient variability in ratings of social features. Indeed, there was greater variance in physical features (mean SD = 5.56) than social features (mean SD = 4.29) across indi- viduals, within houses. Similarly, although relevant only to the multilevel analyses, there was greater variance in physical features (mean SD = 4.4) than social features (mean SD = 1.76) across houses, within individuals. The critical point, however, is that Time 1 ratings of social fea- tures predicted overall happiness at Time 2 while failing to predict Time 1 affective forecasts. In addition, the fact that our manipulation of the isolation effect successfully altered participants’ use of social features is inconsistent with the possibility that ratings of social features were too invariant to allow correlations with forecasts. Therefore, it seems unlikely that Time 1 ratings of social features failed to predict Time 1 affective forecasts primarily because of insufficient variability. GENERAL DISCUSSION In two experiments, we found strong support for the impact bias; participants predicted that housing assign- ment would have far more influence on their emotional well-being a year later than it actually did. Because we used a longitudinal design and participants were ran- domly assigned to a major life outcome that continued to shape their lives at college, this study provides the most powerful evidence to date for the impact bias. As sophomores and juniors, participants’ actual happiness was more consistently associated with their ratings of social features of house life than with their ratings of physical features of their house. Although these latter results were correlational, our findings dovetail with pre- vious research demonstrating that quality of social life is moreimportanttohappinessthanarematerialcomforts. When asked explicitly about the determinants of their future happiness, participants recognized that social aspects of house life would be more important thanphys- ical qualities. Yet, Experiment 1 provided initial evi- dence that participants expected to be much happier in houses with good physical qualities, whereas we found no evidence that participants based their forecasts on the quality of social life in the houses. Indeed, partici- Dunn et al. / MISPREDICTING SATISFACTION 1429

Your preview ends here

Eager to read complete document? Join bartleby learn and gain access to the full version

Access to all documents
Unlimited textbook solutions
24/7 expert homework help

pants who found themselves in houses with good physi- cal features reported lower happiness than they had anticipated. Consistent with Study 1, participants in Study 2 placed relatively little weight on social features in predicting their future happiness while placing a great deal of weight on physical features of the houses—despite the fact that actual happiness a year later was related to the perceived quality of social features at Time 1. Again, par- ticipants recognized the important role social life would play, reporting that social features mattered more on average than did physical features when questioned directly about determinants of happiness. Asking partici- pants to consider what would be similar across houses immediately before forecasting led them to place greater weight on social features, which provides evidence that the isolation effect plays a role in affective forecasting; social features were generally disregarded because par- ticipants paid little attention to important features that were relatively constant across outcomes, unless explic- itly instructed to do so. Our results are interesting in light of the conflict between research showing that people recognize that social relationships are more important than material comforts in producing happiness and other research demonstrating that people’s behavior often flies in the face of this stated belief (e.g., Putnam, 2000; Schor, 1991). The present research suggests that people may truly believe that social relationships are more impor- tant than material comforts, but they may lose sight of this belief when imagining their happiness with alterna- tive outcomes that vary more on material than social dimensions. Again, we do not wish to argue that people always underweight social features and overweight physi- cal features. Exactly the opposite pattern might emerge if participants were asked to predict their happiness given a set of options that varied greatly on social dimen- sions relative to physical dimensions (e.g., if they were asked to imagine how happy they would be in the same house with different sets of roommates). The key finding of our research is that individuals may easily lose sight of factors that they realize are important to their happiness if these factors do not vary greatly across the options they are imagining. Of course, one might argue that the isolation effect reflects a reasonable strategy rather than a dysfunctional bias; perhaps it is wise to focus primarily on features that differ widely between competing options while ignoring features that are similar across options. We would argue that this tendency might sometimes be functional when choosing between options but makes far less sense when imagining one’s future happiness given a specific out- come. For example, it would be quite reasonable to focus on location when choosing between houses that varied a great deal on this dimension but not on others. When imagining life in a specific house with a terrible location, however, it would be unwise to expect lasting misery while neglecting the happiness derived from liv- ing with one’s wonderful roommates. Limitations . Because several of our findings depended on correlational analyses, caution must be exercised in interpreting our results. For example, the positive rela- tionship between happiness and quality of social fea- tures at Time 2 and Time 3 could indicate that both of these variables depended on a third variable, that happi- ness leads to greater satisfaction with social relation- ships, or that strong social relationships promote happi- ness. Research by Oishi and Diener (2001) suggests that general happiness is more likely to influence global rat- ings of social relationships than to influence more spe- cific ratings used in our study (e.g., relationships with roommates). In any case, based on previous research, it seems reasonable to assume that the relationship between social life in the house and happiness is proba- bly bidirectional, meaning that high-quality social rela- tionships have at least some positive influence on happiness. It is important to remember that not all of our analy- ses were correlational; participants were randomly assigned to houses and we randomly assigned people to condition in Study 2. If this were not the case, one might argue that good physical features failed to promote hap- piness in a consistent fashion because people who seek out material comforts are more difficult to please than their less-materialistic peers. Yet, in our research, partici- pants were assigned at random to houses that would be expensive on the open market (because of their central locations, beautiful architecture, and large rooms) or to houses that would be relatively inexpensive. Thus, although the form of our analyses precludes strong causal inferences, the internal validity of the present research is stronger than most studies on the correlates of well-being. Future directions . Because our research tested these ideas in the single domain of a university housing system, it would be worthwhile to expand this line of research through laboratory studies or field studies in a different domain. Although researchers have already examined the spurious factors that influence reports of subjective well-being (e.g., Schwarz & Clore, 1983), we believe that special biases come into play at the forecasting stage. As well as focusing exclusively on features that vary between outcomes, affective forecasters may place disproportion- ate weight on features that are easy to predict or visual- ize. Future research should focus on examining these variables directly to test whether forecasters show a 1430 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

systematic bias toward placing excessive weight on easily predictable and visualizable features. Policy implications . Our research suggests that individ- uals may sometimes err by placing excessive weight on highly variable features and inadequate weight on less variable features in imagining the emotional conse- quences of competing options. A similar bias may emerge when policy makers envision options meant to increase the well-being of their constituents. For exam- ple, in the “slum clearance” projects of the 1950s and 1960s, urban planners chose to tear down small, dilapi- dated tenement buildings, constructing massive towers in their place. Policy makers apparently focused on the drastic improvement in physical features that this change would provide and overlooked the ramifications for social life of ripping up small, tightly knit tenement communities to create giant, anonymous collections of tower dwellers (Ross & Nisbett, 1991). Thus, individuals and policy makers alike may go astray if, as in our studies, they focus on alternatives that vary a great deal on fea- tures (e.g., physical aspects of tenements vs. high-rise apartments) that will be relatively inconsequential for people’s well-being. NOTES 1. After asking participants to report their overall happiness, we asked them to report their happiness with their house, specifically. When we substituted house-specific happiness for overall happiness in the analyses reported in this article, highly consistent results emerged. Therefore, only results for overall happiness are reported, but results of the house-specific analyses are available from the authors. 2. We used this idiographic approach to assess whether a given house assignment represented a positive or negative outcome for each individual. This approach was useful because evaluations of the houses varied from individual to individual, although there was substantial agreementabout the relativedesirabilityof the houses; the averagecor- relation between any one participant’s forecasted happiness and the overall sample’s forecasted happiness for the 12 houses was r = .69, p < .0001. 3. Although the negative relationship between quality of physical features and happiness was unexpected, other studies have found a negative relationship between increases in material comfort (e.g., money) and well-being (e.g., Thoits & Hannan, 1979). Although this paradoxical finding is intriguing, it was inconsistent across time points and should be interpreted with caution. 4. Many participants failed to provide complete data for all 12 houses, resulting in a large amount of missing data in each of the 12 regressions. This potential problem is dealt with in detail in the subse- quent multilevel modeling section. 5. Although not of central interest in this article, the model also provides estimates of the random effects. The intercept estimate indi- cates that thereweresignificant individualdifferencesin meanlevelsof forecasted happiness, collapsing across houses; in other words, some people expected to be happier than others. The small random effects of physical and social quality indicate that the weight placed on these variables varied little across participants. 6. The only other significant effect was an unpredicted two-way Location Quality × Roommate Quality interaction. Because this inter- action was not consistent across models and did not involve condition, it will not be discussed further. REFERENCES Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and inter- preting interactions . Newbury Park, CA: Sage. Argyle, M. (1999). Causes and correlates of happiness. In D. Kahneman, E. Diener, & N. Schwarz (Eds.), Well-being: The foundations of hedonic psychology (pp. 353-373). New York: Russell Sage. Biswah-Diener, R., & Diener, E. (2001). Making the best of a bad situa- tion: Satisfaction in the slums of Calcutta. Social Indicators Research , 55 , 329-352. Bryk, A. S., & Raudenbush, S. W. (1992). Hierarchical linear models: Applications and data analysis methods . Newbury Park, CA: Sage. Buehler, R., & McFarland, C. (2001). Intensity bias in affective fore- casting: The role of temporal focus. Personality and Social Psychology Bulletin , 27 , 1480-1493. Campbell, A., Converse, P. E., & Rodgers, W. L. (1976). The quality of American life: Perceptions, evaluations, and satisfactions . New York: Russell Sage. Diener, E., Gohm, C. L., Suh, E., & Oishi, S. (2000). Similarity of the relations between marital status and subjective well-being across cultures. Journal of Cross-Cultural Psychology , 31 , 419-436. Diener, E., & Seligman, M. E. P. (2002). Very happy people. Psychologi- cal Science , 13 , 81-84. Diener, E., Suh, E. M., Lucas, R. E., & Smith, H. L. (1999). Subjective well-being: Three decades of progress. Psychological Bulletin , 125 , 276-302. Fordyce, M. W. (1988). A review of research on the happiness mea- sures: A sixty second index of happiness and mental health. Social Indicators Research , 20 , 355-381. Freedman, J. (1978). Happy people . New York: Harcourt Brace Jovanovich. Gilbert, D. T., Driver-Linn, E., & Wilson, T. D. (2002). The trouble with Vronsky: Impact bias in the forecasting of future affective states. In P. Salovey & L. Feldman-Barrett (Eds.), The wisdom of feel- ing (pp. 144-166). New York: Guilford. Gilbert, D. T., Lieberman, M. D., & Wilson, T. D. (in press). The pecu- liar longevity of things not so bad . Unpublished manuscript, Harvard University. Gilbert, D. T., Pinel, E. C., Wilson, T. D., Blumberg, S. J., & Wheatley, T. P. (1998). Immune neglect: A source of durability bias in affec- tive forecasting. Journal of Personality & Social Psychology , 75 , 617- 638. Gilbert, D. T., & Wilson, T. D. (2000). Miswanting: Some problems in the forecasting of future affective states. In J. P. Forgas (Ed.), Feel- ing and thinking: The role of affect in social cognition. Studies in emotion and social interaction, second series (pp. 178-197). New York: Cam- bridge University Press. Hodges, S. D. (1997). When matching up features messes up deci- sions: The role of feature matching in successive choices. Journal of Personality & Social Psychology , 72 , 1310-1321. Houston, D. A., & Sherman, S. J. (1995). Cancellation and focus: The role of shared and unique features in the choice process. Journal of Experimental Social Psychology , 31 , 357-378. Houston, D. A., Sherman, S. J., & Baker, S. M. (1991). Feature match- ing, unique features, and the dynamics of the choice process: Predecision conflict and postdecision satisfaction. Journal of Exper- imental Social Psychology , 27 , 411-430. Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica , 47 , 263-291. Kashy, D. A., & Kenny, D. A. (2000). The analysis of data from dyads and groups. In H. T. Reis & C. M. Judd (Eds.), Handbook of research methods in social and personality psychology (pp. 451-477). New York: Cambridge University Press. Kenny, D. A., Kashy, D. A., & Bolger, N. (1998). Data analysis in social psychology. In D. T. Gilbert, S. T. Fiske, & G. Lindzey (Eds.), Hand- book of social psychology (Vol. 1, pp. 233-265). New York: Oxford Uni- versity Press. Larson, R. W. (1990). The solitary side of life: An examination of the time people spend alone from childhood to old age. Developmental Review , 10 , 155-183. Dunn et al. / MISPREDICTING SATISFACTION 1431

Loewenstein, G. (1996, May 3). Explaining public indifference to income inequality . Paper presented to the MacArthur Foundation Meeting on Inequality, Cambridge, MA. Loewenstein, G., & Schkade, D. (1999). Wouldn’t it be nice? Pre- dicting future feelings. In D. Kahneman, E. Diener, & N. Schwarz (Eds.), Well-being: The foundations of hedonic psychology (pp. 85-105). New York: Russell Sage. Mitchell, T. R., Thompson, L., Peterson, E., & Cronk, R. (1997). Tem- poral adjustments in the evaluation of events: The “rosy view.” Journal of Experimental Social Psychology , 33 , 421-448. Myers, D. G. (1999). Close relationships and quality of life. In D. Kahneman, E. Diener, & N. Schwarz (Eds.), Well-being: The founda- tions of hedonic psychology (pp. 374-391). New York: Russell Sage. Oishi, S., & Diener, E. (2001). Re-examining the general positivity model of subjective well-being: The discrepancy between specific and global domain satisfaction. Journal of Personality , 69 , 641-666. Pettijohn, T. F., II., & Pettijohn, T. F. (1996). Perceived happiness of college students measured by Maslow’s hierarchy of needs. Psycho- logical Reports , 79 , 759-762. Putnam, R. D. (2000). Bowling alone: The collapse and revival of American community . New York: Simon & Schuster. Ross, L., & Nisbett, R. E. (1991). The person and the situation: Perspectives of social psychology . New York: McGraw-Hill. Schkade, D. A., & Kahneman, D. (1998). Does living in California make people happy? A focusing illusion in judgments of life satis- faction. Psychological Science , 9 , 340-346. Schor, J. (1991). The overworked American: The unexpected decline of lei- sure . New York: Basic Books. Schwarz, N., & Clore, G. L. (1983). Mood, misattribution, and judg- ments of well-being: Informative and directive functions of affec- tive states. Journal of Personality & Social Psychology , 45 , 513-523. Sheldon, K. M., Elliot, A. J., Kim, Y., & Kasser, T. (2001). What is satisfy- ing about satisfying events? Testing 10 candidate psychological needs. Journal of Personality & Social Psychology , 80 , 325-339. Sieff, E. M., Dawes, R. M., & Loewenstein, G. (1999). Anticipated ver- sus actual reaction to HIV test results. American Journal of Psychol- ogy , 112 , 297-311. Thoits, P., & Hannan, M. (1979). Income and psychological distress: The impact of an income-maintenance experiment. Journal of Health & Social Behavior , 20 , 120-138. Tversky, A. (1972). Elimination by aspects: A theory of choice. Psycho- logical Review , 79 , 281-299. Wilson, T. D., & Gilbert, D. T. (in press). Affective forecasting. In M. P. Zanna (Ed.), Advances in experimental social psychology . San Diego, CA: Academic Press. Wilson, T. D., Wheatley, T., Meyers, J. M., Gilbert, D. T., & Axsom, D. (2000). Focalism: A source of durability bias in affective forecast- ing. Journal of Personality & Social Psychology , 78 , 821-836. Received June 11, 2002 Revision accepted December 26, 2002 1432 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

Your preview ends here

Eager to read complete document? Join bartleby learn and gain access to the full version