Optimistic Bias in Personal Predictions

EVIDENCE OF OPTIMISTIC BIAS IN PERSONAL PREDICTIONS

Henrik Ibsen maintained “that man is in the right who is most clearly in league with the future” (in Sprinchorn, 1964). By similar logic, one might argue that the clearest demonstrations of optimistic biases are those that have revealed systematic discrepancies between people’s predictions and the outcomes they ultimately attain. Empirical studies of the planning fallacy, for example, have shown that people expect to complete personal projects in less time than it actually takes to complete them (e.g., Buehler, Griffin, & Ross, 1994). Evidence that predictions often exceed outcomes has not been limited to estimates of task completion times, however. Documenting a tendency familiar to most students, if not their instructors, Shepperd, Ouellette, and Fernandez (1996) found that students expect to receive higher scores on exams, at least when those exams are still some time away, than they actually receive when they take those exams (see also Doleys & Renzaglia, 1963; Radhakrishnan, Arrow, & Sniezek, 1996). Second-year MBA students were found to overestimate the number of job offers they would receive, the magnitude of their starting salary, and how early they would receive their first offer (Hoch, 1985). Even the forecasts of professional financial analysts have been found to be optimistically biased: Although the studied analysts were reasonably able to anticipate periods of growth and decline in corporate earnings, their forecasts consistently overestimated earnings realized (Calderon, 1993; see also Barefield & Comiskey, 1975). Also, in a set of studies by Mitchell, Thompson, Peterson, and Cronk (1997), incipient vacationers were found to anticipate greater enjoyment during upcoming trips than they actually expressed during their trips, suggesting that people may be overly optimistic even when asked to anticipate their own evaluations of future experiences.

Despite the conceptual clarity of these results and the lack of ambiguity of the criterion against which predictions are regarded optimistic, researchers have largely resorted to other methods in their investigations of optimistic bias. These methods, although less direct, have yielded a number of important insights about the nature of optimistic predictions.

One alternate strategy has been to compare people’s predictions to outcomes attained by similar others. For example, Sherman (1980) asked college students to predict how they would behave in one of a variety of specific situations, and then compared their predictions to the observed behavior of students who had actually been placed in one of the targeted situations. The results of Sherman’s studies revealed that students’ predictions were optimistically biased: Participants consistently reported that they would behave in more socially desirable ways – for example, that they would be more resistant to unwanted social influence, or more likely to donate time to a worthy charity – than did people who had not first been asked to make predictions about their behavior. By the same logic, many newlyweds – who almost uniformly expect that their marriages will endure a lifetime (Baker & Emery, 1993) – can be considered unrealistically optimistic when their expectations are viewed against the high rate of divorce in contemporary society.

Research comparing personal expectations to population base rates has not uniformly revealed evidence of optimistic bias, however. Several studies have shown that, for important and personally relevant events such as unwanted pregnancy, people’s expectations can be quite accurate with respect to the base rates of relevant populations (Whitley & Hern, 1991; Gerrard & Luus, 1995). People also appear to consistently overestimate the prevalence of, and their own likelihood of experiencing, low-frequency events (including extremely negative, life-threatening events such as AIDS) and to underestimate high-frequency events (e.g., Lichtenstein, Slovic, Fischoff, Layman, & Combs, 1978; Rothman, Klein, & Weinstein, 1996). One might hesitate to label the overestimation of negative but statistically infrequent outcomes as evidence of genuine, psychologically meaningful pessimism, however, as these estimates may simply reflect difficulties interpreting and reporting extreme probabilities. Indeed, these apparent pessimistic responses may make the unwanted consequence seem likely enough to warrant preventative action, but still sufficiently unlikely to not be the cause for feelings of hopelessness, panic, or self-defeating behavior. Consistent with this interpretation, Rothman et al. (1996) found that even people who appeared pessimistic when their predictions were compared to population base rates appeared to be optimistic when assessed on an alternate measure of optimistic bias that relies on people’s assessments of their relative likelihood of experiencing future events.

By far the most commonly used method of assessing excessive optimism relies on the comparison of what people expect for themselves and what they expect for others (e.g., Weinstein, 1980; Perloff & Fetzer, 1986). Results from hundreds of empirical investigations have shown that, on average, people tend to view themselves as more likely to experience positive outcomes, and less likely to experience negative ones, than the average members of the group from which they have been drawn (Weinstein, 1998b). For example, most people expect that they have a better-than-average chance of living long, healthy lives; being successfully employed and happily married; and avoiding a variety of unwanted experiences such as being robbed or assaulted, injured in an automobile accident, or experiencing health problems (Weinstein, 1980). Although such comparisons cannot be used to identify optimistic biases within any one individual – any one person may be (and up to half of any group will be) more likely to attain favorable outcomes relative to the average member of their group – the proportion of individuals expecting to be above the median in terms of future outcomes is commonly found to be impossibly large. Representative studies from our own lab, for example, have found that between 85% and 90% of respondents claim that their future will be better – more pleasant and less painful – than the future of an average peer (Armor, 2000, unpublished raw data). This suggests that at least some (and possibly many) are assessing their relative prospects too favorably.