Impact Bias

​

What is Impact Bias?

Impact Bias is the tendency to overestimate the intensity, the duration, or both, of one’s emotional reaction to a future event.

More common than underestimating future emotional reactions, however, is the impact bias, whereby people overestimate the intensity and duration of their emotional reactions to future events—even when they know what the future event is likely to entail and they are not in a particularly “hot” or “cold” emotional state at the time of making their forecast.

That sentence is from Timothy D. Wilson and Daniel T. Gilbert, “Affective forecasting: Knowing what to want,” Current Directions in Psychological Science 14 (2005), pages 131–134. The scanned PDF was read. An affective forecast is a prediction of one’s own later feeling. The bias is about that prediction. It is not a claim that the event will be pleasant or unpleasant. The same review says people are generally good at that valence call. A root beer is more pleasant than a root canal. The miss is how strong the feeling will be, and how long it will stay. A useful picture is a calendar with one Saturday written in ink and the other days left blank. The Saturday is real. The blank days are also real, and they will have classes, meals, and other news. The forecast treats the inked day as the mood of the month. Wilson, Wheatley, Meyers, Gilbert, and Axsom (2000) named that narrow focus focalism. It is one source of the bias. It is not the whole of it. The Duration Neglect page is a different measurement. There, people look back on an episode that already happened, and the length of the episode adds little to the remembered evaluation. Here, people look forward, and they give a future feeling more intensity or more staying power than later reports support. The Focusing Illusion page uses focalism as an alias for a related attention error, the one in judgments of a whole life. Wilson and Gilbert (2013) say the two labels describe a similar difficulty. This page does not treat them as one score. The Hot–Cold Empathy Gap is a third cousin, and it often runs the other way: from a calm state, people underestimate a later craving or pain. The 2005 review gives that underestimation as the less common miss, and it gives the Impact Bias as the more common one.

​

Impact Bias in 3 Depths

• Beginner: When one future event feels as if it will set your mood for months, name three ordinary things that will also happen in those months. The blank days on the calendar are the part the forecast drops.
• Practitioner: Before a job decision, a move, or a relationship decision, write two forecasts. One is how you will feel while you are thinking about the outcome. The other is how happy you will be, in general, on an ordinary day. Wilson and Gilbert (2013) treat those as different questions. Do not treat a filled-in week as a cure that has been shown outside the samples below.
• Advanced: The 1998 paper expected the error mainly on bad news, because people neglect the ways they will explain a loss to themselves. The breakup forecasts fit that pattern. The short-term tenure forecasts were too extreme in both directions. A rule that says only losses are misforecast is already wider than that paper.

​

Origin

The duration form has a date and a paper. In 1998, Gilbert, Pinel, Wilson, Blumberg, and Wheatley, then at Harvard, the University of Texas at Austin, and the University of Virginia, published six studies in the Journal of Personality and Social Psychology. They called the error the durability bias: overestimating how long an affective reaction will last.

People are generally unaware of the operation of the system of cognitive mechanisms that ameliorate their experience of negative affect (the psychological immune system), and thus they tend to overestimate the duration of their affective reactions to negative events.

That sentence opens the 1998 abstract. The psychological immune system, in their usage, is the set of cognitive moves that soften a negative feeling, including the explanations people later find. Immune neglect is failing to anticipate those moves. The six topics in the abstract are a romantic breakup, failure to get tenure, an electoral defeat, negative personality feedback, an account of a child’s death, and rejection by a prospective employer. This page uses numbers only from the breakup study and the tenure study. The other four procedures were not extracted. Wilson and Gilbert, in a 2013 comment in the same journal, volume 105, pages 740–748, DOI 10.1037/a0032662, give the later name. The PDF was read. They write that the first studies were about duration, and that the name was the durability bias. They then write that people also mispredict the initial reaction, so they changed the term to the impact bias, for overestimating the initial impact, the duration, or both. They cite Gilbert, Driver-Linn, and Wilson (2002) for that change. That chapter was not opened, so its wording is not quoted here. The 2005 review, which was read, uses the later name and the intensity-and-duration definition above. Focalism is a second source, with its own paper. Wilson, Wheatley, Meyers, Gilbert, and Axsom, “Focalism: A source of durability bias in affective forecasting,” Journal of Personality and Social Psychology 78(5), 2000, pages 821–836, DOI 10.1037/0022-3514.78.5.821. The PDF was read. The abstract’s claim is that people focus on the event in question and not enough on the consequences of other future events. Asking them to think about other future activities reduced the durability bias in their forecasts. The Planning Fallacy is named in that abstract as a literature the authors discuss. This page does not treat that discussion as a measured effect on project dates.

​

Key Points

Four distinctions keep Impact Bias from turning into “nothing matters” or into a single score for every forecast.

​

Applications

These four moves use the questions the studies actually asked. They are not a claim that a diary, a second question, or a conversation with someone who has been through it changes a grade, a hire, or a household’s mood by a measured amount.

​

Case Study

The counted case is Study 2 of Gilbert, Pinel, Wilson, Blumberg, and Wheatley (1998): assistant professors forecasting a tenure decision, set against former assistant professors who had already received one. The pool was the College of Liberal Arts at the University of Texas at Austin, every department except psychology. The college’s records listed people considered for promotion from 1984 through 1994. Current assistant professors in the same departments were added from the faculty directory. Department chairs were asked to correct the lists. Anyone up for tenure in the year of the study was removed, so that a person would not change category mid-study. That left 97 current assistant professors, the forecasters, and 123 former assistant professors. Of the 123, 92 had been promoted to associate professor with tenure, and 31 had been denied it. The paper calls this a 75 percent tenure rate and notes that people who left before a formal decision were not in either experiencer group, because they could not be classified. Each person got a questionnaire from a professor at the University of Virginia about “the lives of people who are or have been faculty members.” Happiness in general was a single item from 1, not happy, to 7, very happy. The same people also completed items from Diener, Emmons, Larsen, and Griffin’s Satisfaction With Life Scale and from Kammann and Flett’s Affectometer 2. Across participants the single item correlated r(98) = .86 with the life-satisfaction items and r(97) = .83 with the Affectometer items. The degrees of freedom differ because a few people skipped items. Only the single item was analyzed further. Forecasters then estimated general happiness at a series of lags after a positive decision and after a negative one, from the time they learned the result out to 10 years. Complete replies came from 33 of 97 forecasters (the paper prints 34.02 percent), 47 of 92 promoted professors (51.10 percent), and 20 of 31 denied professors (64.50 percent). There were not enough people at each single year, so the paper splits experiencers into a recent class, a decision within the past five years, and an ancient class, a decision six to ten years earlier. Recent promoted professors (n = 25) and recent denied professors (n = 7) were on average 2.6 years past the decision. Ancient promoted professors (n = 22) and ancient denied professors (n = 13) were on average 8.5 years past it. Forecasts were averaged the same way: years 1 through 5, and years 6 through 10. On the 1-to-7 scale, forecasters put the first five years after a grant of tenure at 5.90 (SD = 1.09) and after a denial at 3.42 (SD = 1.37), both n = 33. Recent promoted professors reported 5.24 (SD = 1.39). Recent denied professors reported 4.71 (SD = 1.98). For the later window, forecasters put a grant at 5.65 (SD = 1.35) and a denial at 4.97 (SD = 1.81). Ancient promoted professors reported 5.82 (SD = 0.91). Ancient denied professors reported 5.23 (SD = 1.74). Forecasters expected a large early gap that would close. The outcome-by-time interaction on their predictions was F(1, 32) = 67.14, p < .001. Among experiencers, outcome, time, and their interaction were not significant: F(1, 63) = 2.07, p = .155; F(1, 63) = 2.00, p = .16; and F(1, 63) = 0.01, p = .94. Promoted and denied professors were not reliably different in the first five years, F(1, 63) = 0.77, p = .38, or in the next five, F(1, 63) = 1.44, p = .23. The paper says the small number of denied professors means those nulls need caution, and that the decision at least did not have a dramatic, robust effect on general happiness in this sample. The accuracy contrasts compare a forecast with the matching experience. Recent promoted professors were less happy than forecasters expected, F(1, 56) = 4.14, p = .047. Ancient promoted professors matched the forecast, F(1, 53) = 0.27, p = .61. Recent denied professors were happier than forecasters expected, F(1, 38) = 4.36, p = .04. For the ancient denied group the paper prints F(1, 44) = 0.20, p < .66, and it calls that long-term forecast accurate. The printed p is unusually high for a “less than” statement. It is reported here as printed, not corrected. The lesson is the shape of those four contrasts. The short window was misforecast in both directions. The long window was not. These figures belong to that college, those years, and those response rates. They are not a happiness rate for another university, another decade, a hiring meeting, or a household. The authors themselves treat the design as cross-sectional. A person who expects to be miserable after a denial may also be the person who works hardest not to be denied, so the denied group can be a selected group. They ran a later longitudinal election study for that reason. This page leaves that study’s counts unused.

​

Boundaries and Failure Modes

Impact Bias fails as a slogan in three places. First, it is not every mistake about the future, and it is not a retrospective weight. Duration Neglect is about how little a finished episode’s length counts in memory. Projection Bias is about carrying today’s preference into a later preference. The 2005 review’s cold-to-hot underestimates are the other direction from the Impact Bias. Using this name for a forgotten duration, a present-tense craving, or a life-satisfaction judgment about climate goes past the forecasts these papers measured. Focalism is one mechanism inside the Impact Bias. The Focusing Illusion page already tells that attention story for a different judgment. Wilson and Gilbert (2013) say the difficulty is similar. The samples and the questions are not the same. Second, the tenure contrasts will not travel as a rate. The denied cells are 7 and 13 people. People who left before a formal review were excluded. Forecasters and experiencers are different people, answering in different years. The response rates are 34.02, 51.10, and 64.50 percent of the pools the paper defined. The single item is general happiness, not happiness about the decision. A department that treats 4.71 as the happiness of denied professors, or 3.42 as the cost of a denial, has left the table. The breakup means are from 532 students in one introductory course at the same university, after 39 incomplete questionnaires were removed from 571. They are not a rate for other couples. Third, the usual misuse is to hear the short-window error as permission. The paper does not say a denial feels like a grant on the day the letter arrives. The analyzed indexes average years, and the first-day ratings were not the contrasts above. It also does not say the long window is a law. The ancient denied contrast is the one whose printed p is p < .66. Nor does a diary, in the study that invented the demonstration, come with a later mood. Study 1 of Wilson and colleagues (2000) says so. Study 2 of that paper found fans, on later Mondays and Tuesdays, not reliably happier after wins, and one open-date peak fell on Thanksgiving week. The authors say a comparison across the 1995 forecasts and the 1997 reports has to stay tentative, because the games and the samples were not the same.

​

Common Misconceptions

Three shortcuts turn a specific forecast error into a rule about feelings.

​

Related Concepts

These six pages sit next to Impact Bias. Each one is a different mistake.

​

One-Line Takeaway