Options
Leveraging Projection to Increase the Predictive Validity of Self-Reports: Four Experiments in the Domain of Consumption Motives
Type
conference paper
Date Issued
2015-02-27
Author(s)
Raghubir, Priya
Abstract
It is well known that respondents view themselves in a more positive light than others. Prominent domains include personal traits (“better-than-average effect,” Alicke et al. 1995), abilities (“above-average effect,” Dunning, Meyerowitz, and Holzberg 1989), and likelihood judgments (“unrealistic optimism,” Weinstein 1980; “self-positivity bias,” Raghubir and Menon 1998). However, as everyone cannot be above average, these comparative self-reports are biased. An important theoretical and managerial question is: “How best can self-reports be de-biased?” A common technique to elicit reliable responses to sensitive questions is projection: ask about other people and use this information as a proxy for self-judgments. For socially sensitive questions, these (de-biased) proxy-reports have been shown to have higher predictive validity for behavioral intentions than the (biased) self-reports (Fisher 1993). This paper presents an alternate method that leverages projection to “de-bias” (vs. “replace”) self-reports by utilizing contextual manipulations: When other-judgments precede self-judgments (i.e., other-self order) projection should be greater as the other is judged to be more similar and, hence, self-other differences should be attenuated (vs. when self-judgments are elicited first, i.e., self-other order, Raghubir and Menon 1998). The key question is whether other-judgments (projection techniques) or self-judgments elicited after other-judgments (our method) are better predictors of future behavior. To assess this question we first ran three field experiments (Ns=957, 789, and 1095) among participants at an international marathon. Prior to the race, the runners were asked for their motivations to run marathon that were grouped into signaling-related (more sensitive) and fitness-related (less sensitive) reasons. Respondents rated their reasons either before or after rating reasons for the average runner (Studies 1 and 2). In Study 3, we additionally manipulated the referent other (average, similar, inferior, or superior runner). We further assessed the perceived self-other similarity and the extent to which judgments of the other person were anchored on the self (Studies 2 and 3). After the race, we contacted the participants again and in Study 2 and 3 (Ns=552 and 764 matched observations) assessed the stated importance of signaling-related products (e.g., finisher medal, finisher t-shirt). “I run for fitness, not for fame”. In all three studies runners believed that they ran more for fitness rather than signaling motives (ΔFitnessSelf–SignalingSelf, ΔMSelf =.78, .94, and .80 all >0, Motive main effect: Fs>245, ps<.001). This finding replicates the socially desirable bias in consumption motives, which is at an absolute level. The “Nobler-than-thou” effect. In each study we found that the runners believed that they ran for fitness rather than signaling to a greater extent (see above) than the other runner they judged (ΔFitnessOther–SignalingOther, ΔMOther=.09, .05, and -.24 all < their corresponding ΔMSelf, TargetxMotive interaction: Fs>229, ps<.001). We refer to this as the “Nobler-than-thou” effect, which is relative to a referent other. 139 The attenuating effect of order-of-elicitation. In every study we demonstrated that the “Nobler-than-thou” effect (= ΔFitnessSelf–SignalingSelf > ΔFitnessOther–SignalingOther) was lower in the other-self condition (ΔMSelf=.57, .86, and .66 vs. ΔMOther=.19, .17, and -.14) than in the self-other condition (ΔMSelf=.99, 1.02, and .93 vs. ΔMOther=-.01, -.06, and -.34, TargetxMotivexOrder interaction: Fs>16, ps<.001). Process evidence for similarity (established by mediation and moderation). As predicted, Studies 2 and 3 showed that in the other-self order the average runner was perceived as more similar (Mother-self =.21 and .34) than in the self-other order (Mself-other=-.10 and -.12, Fs>5, ps<.05, with similarity as mediator at 95% CIs). The order effect was moderated by the type of other in Study 3 (4-way interaction: F=2.5, p=.06), that is, it was eliminated for similar peers (self-other order: ΔMSelf=.85 and ΔMOther = .17. vs. other-self order: ΔMSelf =.76 and ΔMOther =.29, TargetxMotivexOrder: F=2.0, p>.10) compared to superior or inferior peers (Fs>5, ps≤.02), which corroborates that perceived similarity drives the order effects. Role of projection. Studies 2 and 3 also revealed that the respondents anchored less on themselves when judging the average other in the self-other order (Mself-other=2.94 and 2.66 vs. Mother-self =3.40 and 3.47, Fs>13, ps<.001). Study 3 shows that this holds for average, inferior, and superior others (Fs>2, ps<.10), but not for similar others (Mself-other=3.27 vs. Mother-self=3.42, F<1). Taken together, the findings suggest that eliciting responses about others first makes them appear more similar, and leads to greater projection of own motives onto judgments of others, unless the other person is similar to begin with. Predictive validity. Pooling the data across studies, we regressed the importance of signaling-related products onto the fitness and signaling motives projected for others (projection technique: Only Other), the fitness and signaling motives given for oneself (Only Self), and all four predictors together (Both Self and Other), separately for each order condition. First, we find that the motives people give for themselves are better at predicting the importance they assign to signaling-related products as compared to the motives that they give to others and do so parsimoniously, particularly in the other-self order (R2 Only_Other =.03, R2 Both_Self_and_Other =.09, and R2 Only_Self=.09) versus the self-other order (R2 Only_Other=.02, R2 Both_Self_and_Other=.07, and R2 Only_Self=.06). Second, in the other-self order, it is the (de-biased) signaling motives people give for themselves that drive how important signaling-related products are perceived rather than the projections for others (which are not significant once self-values are entered to the regression), suggesting that eliciting self-reports after other-reports is a superior alternative to only using other reports (projection techniques). Generalizability. Study 4 (N=230 mTurk) replicated the findings for luxury purchases. Consumers judged that they purchase luxury brands more for quality than status reasons compared to others, particularly if self-reports preceded other-reports. Moreover, we found that self-deceptive needs for social approval (vs. self-enhancement or impression management, Paulhus 2002) drive the effects. To summarize, we show that when self-judgments are elicited after other-judgments, they are better predictors compared to not only when they are elicited first, but also as compared to only eliciting other-judgments (viz., projection techniques). As such, encouraging projection via contextual manipulations that increase the similarity between the self and the other (order-of-elicitation, type of referent other), attenuates self-other biases in socially sensitive consumption motives and increases the predictive validity of self-reports.
Language
English
HSG Classification
contribution to scientific community
HSG Profile Area
SoM - Business Innovation
Refereed
Yes
Event Title
Society for Consumer Psychology (SCP) Winter Conference 2015
Event Location
Phoenix, US
Event Date
26.-28.02.2015
Subject(s)
Division(s)
Eprints ID
239621