Abstract
Affective forecasting—the process in which individuals predict how they will feel at some point in the future—has become a major topic of research interest in social psychology (Kushlev & Dunn, 2011). The broad message emanating from the literature is that, on balance, people are generally poor at predicting how they will feel after future events (Gilbert, Pinel, Wilson, Blumberg, & Wheatley, 1998; for a review, see Ayton, Pott, & Elwakili, 2007). For example, Hoerger, Quirk, Lucas, and Carr (2009) showed that football fans were not as upset after their team’s loss as they had predicted they would feel. Similar effects have been reported for events of all types, ranging from such relatively inconsequential sports events (Hoerger et al., 2009; Wilson, Wheatley, Meyers, Gilbert, & Axsom, 2000) to major life-changing events, such as being diagnosed with HIV (Sieff, Dawes, & Loewenstein, 1999).
Two Conceptions of Accuracy
Past research on affective forecasting has been based on a conceptualization of accuracy as the mathematical difference between the predicted feeling and the actual feeling. However, there are other legitimate ways to conceptualize accuracy. For example, imagine the job of a director of an agency that provides mental-health support to individuals diagnosed with HIV. In budgeting for the coming year, it would be useful for the director to know that individuals tend to be poor at predicting how they will feel in an absolute sense—that they tend to think they will be more upset after learning about their positive HIV status than they turn out to be (cf. Sieff et al., 1999). But now imagine that the agency’s resources are limited, so the director has the job of allocating them to where they can be used most effectively. In this case, it would be useful for the director to know in a relative sense whether those people who predict they will be more upset actually end up being more upset than the people who predict they will be less affected by the news. In other words, it could be the case that people are generally inaccurate in an absolute sense (all think they will be more upset than they turn out to be) but accurate in a relative sense (those who think they will be the most upset end up being the most upset).
To date, virtually no research has focused on accuracy of affective forecasting in the relative sense, as opposed to inaccuracy in the absolute sense. In conducting the analyses reported here, our goal was to examine the magnitude of people’s relative accuracy in affective forecasting and compare that with the magnitude of their inaccuracy as assessed by the more conventional absolute index. Specifically, we investigated whether individuals are as accurate in the relative sense as they are inaccurate in the absolute sense.
To obtain robust estimates of the effects, we took a meta-analytic approach, summarizing the effects in all published studies for which both relative accuracy and absolute inaccuracy could be computed in the same sample.
Method
Most studies of affective forecasting assess forecasts of affect after an event (e.g., people’s predictions of how they will feel after failing a driving test) and measures of actual affect after an event (e.g., assessments of people’s affect after they have failed a driving test), but typically, the two measures are not obtained for the same individuals, or data are aggregated so that a given individual’s forecast is not linked to his or her actual affect. To evaluate relative accuracy and absolute inaccuracy in the same study, one must use a within-subjects design, collecting affective forecasts and subsequent assessments of affect in the same individuals. Therefore, in identifying studies to be included in our meta-analysis, our first selection criterion was a within-subjects design. Our second criterion was that the forecast concerned the same event that was actually experienced by the participant.
Keyword searches for “affective forecasting” identified 46 potentially relevant articles. Fourteen met our selection criteria. The 12 corresponding authors of these 14 articles were sent an e-mail requesting their raw data. Of these 12 authors, 11 provided data. Thus, our analysis included data from a total of 11 articles, and 16 studies, with a total participant count of 1,074. These data yielded 129 effect sizes. (See the Supplemental Material available online for a list of the articles and a more detailed presentation of analyses and results than space allows here.)
Effect-size estimates were converted to a common metric, d, and a random-effects model was used. Effect sizes were weighted by their inverse variance, and 95% confidence intervals (CIs) were computed. Tests of homogeneity revealed the effect sizes to be heterogeneous, so we conducted moderator analyses (reported in the Supplemental Material).
Results and Discussion
The overall estimated effect size for relative accuracy was 0.56 (95% CI = −.08–1.21), and the overall estimated effect size for absolute inaccuracy was 0.49 (95% CI = −0.07–1.06). Thus, in general, it is as appropriate to conclude from the published literature that individuals are accurate at predicting how they will feel (in the relative sense) as to conclude that individuals are inaccurate at predicting how they will feel (in the absolute sense).
These findings provide an important qualification to the widely understood take-home message from the affective-forecasting literature—that people are generally poor at predicting how they will feel. The meta-analytic findings suggest that this conclusion is only partially true. In fact, the mean correlation between predicted and actual affect was .28, which shows that when accuracy is computed in a relative sense, people are reasonably good at predicting how they will feel.
Both conceptions of accuracy are reasonable, and each one can be appropriate, depending on the circumstances. The findings suggest that a more nuanced understanding of the affective-forecasting literature is warranted. Such an understanding would account for the fact that people are both accurate (in the relative sense) and inaccurate (in the absolute sense) at predicting how they will feel.
Footnotes
Acknowledgements
We are grateful to Robert Wilson for assisting in coding the studies and to Peter Ayton, Roger Buehler, Elizabeth Dunn, Catrin Finkenauer, Daniel Gilbert, Michael Hoerger, Deborah Kermer, Gillian Ku, Kent Lam, Lisbeth Nielsen, and Nick Sevdalis for providing data from their research. We are particularly grateful to Erika Patall for her advice regarding the meta-analyses.
The authors declared that they had no conflicts of interest with respect to their authorship or the publication of this article.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
