Impact bias or underestimation? Outcome specifications determine the direction of affective forecasting errors

People rely on affective forecasts to anticipate the emotional impact of future events. Consequently, affective forecasts have an integral role in decisions, from everyday choices about what to eat for lunch to those with consequential personal, medical and legal implications (e.g., Blumenthal 2004; Ditto, Hawkins, and Pizarro 2005; Eastwick et al. 2008; Halpern and Arnold 2008; Maroquin, Nolen-Hoeksema, and Miranda 2013; Mellers and McGraw 2001; Miloyan and Suddendorf 2015; Riis et al. 2005; Woodzicka and LaFrance 2001). Affective forecasters are generally accurate when predicting whether an event will be pleasant or unpleasant, but often err in their predictions of how pleasant or unpleasant the event will be (Artz 1996; Mathieu and Gosling 2012; Rachman and Artz 1991). Early affective forecasting research demonstrated a widespread impact bias, a systematical tendency for forecasters to overestimate the hedonic impact of future events, ranging from HIV test results over tenure decisions, to their favorite sports team winning or losing a game (for reviews, see Gilbert and Wilson 2007; Wilson and Gilbert 2003, 2005, 2013). Recently, several countervailing cases have been documented in which forecasters systematically underestimate the hedonic impact of future events, such as the emotional pain induced by milder forms of sexism or sad fictional stories, and the pleasure gleaned from highly probable or distant future prizes (Andrade and Van Boven 2010; Bosson, Pinel, and Vandello 2010, Ebert and Meyvis 2014; Dunn et al. 2007; Gilbert et al. 2004a; Nordgren, Banas, and McDonald 2011). We propose a process account that can explain these seemingly incongruous findings and predict the direction of affective forecasting errors. The hedonic impact of an event is influenced by general outcome specifications of the event such as its magnitude, duration, psychological distance, and probability of occurrence (Ebert and Meyvis 2014; Mellers et al. 1997). Some outcome specifications such as magnitude and duration are positively correlated with the hedonic impact of the event (e.g., presumably, winning more money feels better than winning less money, and a longer commute feels worse than a shorter commute). Other outcome specifications such as probability and psychological distance are negatively correlated with the hedonic impact of the event (e.g., presumably, winning $100 feels better if one had a 1% chance of winning than a 99% chance, and receiving $100 today is more exciting than receiving $100 a year from today; Brandstaetter, Kuehberger, and Schneider 2002; Ebert and Meyvis 2014; Mellers et al. 1997). Critically, affective forecasters are typically more sensitive to the effects of outcome specifications on the hedonic impact of an event than are experiencers (Aknin, Norton, and Dunn 2009; Buechel et al. 2014; Dunn, Wilson, and Gilbert 2003; Ebert and Meyvis 2014; Gilbert et al. 2004b; Hsee and Zhang 2004). Because experiences are more affect-rich than the simulations upon which affective forecasts rely, experiences usurp more attentional resources necessary to attend to and be impacted by outcome specifications (Buechel et al. 2014; Ebert and Meyvis 2014; Morewedge et al. 2010). We suggest that consequently, the relationship between outcome specifications and hedonic impact is amplified in affective forecasts compared to experiences, resulting in both overestimation (i.e., impact bias) and underestimation (i.e., its inverse) of hedonic reactions. The direction in which an outcome specification modulates the hedonic impact of an event allows us to predict which affective forecasting error is more likely to occur. In general, we predict that when outcome specifications are positively correlated with the hedonic impact of an event (e.g., magnitude or duration), forecasters should overestimate the extent to which high specification values will intensify its impact and low specification values will diminish its impact. Conversely, when outcome specifications are negatively correlated with the hedonic impact of an event (e.g., probability or psychological distance), forecasters should overestimate the extent to which low specification values will intensify and high specification values will diminish its impact. To test our theory, we simultaneously manipulated both kinds of outcome specifications in two experiments. We expected that when both specifications were aligned in their modulation of hedonic impact-high magnitude and low probability or low duration and high psychological distance-they would have an additive effect, leading forecasters to systematically overestimate (i.e., impact bias) or underestimate (i.e., its reverse) the hedonic impact of experiences. When the specifications were misaligned, we predicted that their effects would be countervailing, leading to less extreme predictions and forecasting errors.