Knowledge matters: Anchoring effects are moderated by knowledge level

Previous research into the relationship between knowledge level and anchoring effects has led to mixed conclusions. This paper presents four studies that used a diverse set of stimuli and paradigms to further investigate this relationship. In Study 1, greater knowledge was associated with smaller anchoring effects—both when knowledge was measured using subjective self-assessments and when using an objective knowledge measure. In Study 2, participants from the USA and India tended to exhibit smaller anchoring effects when answering questions about their own country as compared with questions about the other country. In Study 3, higher knowledge was associated with smaller anchoring effects when examined at an idiographic level. Finally, in Study 4, providing participants with information designed to increase their knowledge led to a decrease in anchoring effects. The consistency of the results across our four studies provides evidence that anchoring effects are moderated by knowledge level in many situations. Copyright © 2012 John Wiley & Sons, Ltd. When people are estimating a numeric value, the introduction of an irrelevant anchor value will often cause their estimate to assimilate towards the anchor (for reviews, see Chapman & Johnson, 2002; Epley, 2004; Furnham&Boo, 2011). Anchoring effects are quite robust and have been observed in estimates across a wide variety of domains including answers to general knowledge questions (Tversky & Kahneman, 1974) and math problems (Smith & Windschitl, 2011), estimates of real-estate prices (Northcraft & Neale, 1987), decisions about criminal sentences (Englich, Mussweiler, & Strack, 2006), and in the outcomes of negotiations (Galinsky&Mussweiler, 2001). Because anchoring effects have been observed in so many different situations, understanding how people might overcome the biasing influence of anchors is important. It seems plausible that one factor that would moderate anchoring effects is a person’s level of knowledge. That is, high-knowledge people would be less influenced by anchors than their lowknowledge counterparts. However, the relationship between knowledge level and anchoring effects is unclear. Some researchers have speculated that high knowledge people “should” be less influenced by anchors (e.g., Mussweiler & Strack, 1999). However, recent research suggests that highknowledge and low-knowledge people are equally biased by irrelevant anchors (e.g., Englich, 2008; Englich et al., 2006; Englich & Soder, 2009; see also Northcraft & Neale, 1987). Given this discrepancy, the studies described in this manuscript were designed to investigate the relationship between knowledge and anchoring across a variety of domains. *Correspondence to: Andrew R. Smith, Appalachian State University, Departmen E-mail: [email protected] Copyright © 2012 John Wiley & Sons, Ltd. Background of Anchoring Effects and Theory In their classic study on anchoring effects, Tversky and Kahneman (1974) first asked participants whether the percentage of African countries in the United Nations was higher or lower than an ostensibly random number (predetermined to be 10% or 65%). This anchor value influenced participants’ subsequent estimates of the actual number of African countries in the UN (higher estimates when the anchor was 65% rather than 10%). Numerous accounts have been offered to explain why people anchor on such irrelevant values in paradigms such as these. These accounts can be grouped into three categories: (i) enhanced accessibility of select knowledge, (ii) anchoring and insufficient adjustment, and (iii) priming. The first account suggests that anchors cause people to recruit biased pools of information (Mussweiler & Strack, 1999; Strack&Mussweiler, 1997; see also Chapman& Johnson, 1999). Mussweiler and Strack’s “selective accessibility” model explains anchoring by assuming that when participants compare the target estimate to an anchor, they first test whether the target is equal to the anchor value. Because people tend to engage in hypothesis-consistent testing (Klayman & Ha, 1987), they will likely think about information consistent with the anchor value. When participants provide their absolute estimate of the target value, they rely on the biased set of information that has been recruited. Therefore, estimates following a comparison with an anchor tend to assimilate toward the anchor value. t of Psychology, ASU Box 32109, Boone, NC 28608-2109, USA. Received 23 January 2012, Accepted 7 October 2012 98 Andrew R. Smith et al. The second account, “anchoring and insufficient adjustment”, assumes that the anchor provides a starting point that people use when making their judgment (Epley & Gilovich, 2001; Tversky & Kahneman, 1974; see also Simmons, LeBoeuf, & Nelson, 2010). As information is recruited about the target, people adjust their estimate away from the anchor. These adjustments, however, tend to be insufficient (Epley & Gilovich, 2004). That is, people tend to stop adjusting once they reach a plausible estimate (Quattrone, Lawrence, Finkel, & Andrus, 1981). Because there is generally a large range of plausible estimates, adjustments that start from a low anchor stop at the lower end of this range, whereas adjustments from a high anchor terminate at the upper end of the range. The third category of accounts, numeric and magnitude priming, posits that anchors prime numbers or magnitudes similar to the anchor value. For example, in one study, participants’ arbitrary ID numbers influenced their estimates of the number of physicians in the phonebook (Wilson, Houston, Etling, & Brekke, 1996). Presumably, viewing the ID number increased the accessibility of similar numbers. When participants generated their estimates, these primed numbers were more likely to come to mind, thereby influencing their estimates (see also Critcher & Gilovich, 2007; Wong & Kwong, 2000). The magnitude priming account is similar, but rather than priming numbers, it assumes that anchors prime magnitude concepts (e.g., “large” and “small”), and these concepts influence the estimates that people give (Oppenheimer, LeBoeuf, & Brewer, 2008). Empirical Findings Regarding Knowledge Level as a Moderator As previously mentioned, a number of studies have tested for expertise or knowledge level as a moderating factor of anchoring effects, and the findings are mixed. In the next two sections, we review the findings suggesting the knowledge does and does not moderate anchoring effects. Findings Suggesting Knowledge Moderates Anchoring Wilson et al. (1996) found that participants who reported they were more knowledgeable about the number of physicians in the phonebook were less influenced by anchors when estimating this value. However, participants were asked about their knowledge level immediately after making their estimate. Therefore, their knowledge judgments could have reflected their confidence in their estimates more so than their knowledge about the topic (Englich, 2008). It is also quite likely that people who felt they were adversely influenced by the anchor might also be less confident in their estimate. Therefore, it is unclear whether the results truly indicate that greater knowledge leads to smaller anchoring effects. Mussweiler and Englich (2003) tested German participants on anchoring tasks before and after the introduction of the Euro in 2002. Before the introduction of the Euro, German participants showed larger anchoring effects when making price estimates in Euros as compared with German Marks. However, after the introduction of the Euro, the pattern was reversed. Presumably, greater experience with the Euro reduced Copyright © 2012 John Wiley & Sons, Ltd. the participants’ uncertainty when making price estimates, and their decreased uncertainty led to small anchoring effects. In addition to the “direct” tests of knowledge and anchoring described earlier, a study by Blankenship, Wegener, Petty, Detweiler-Bedell, and Macy (2008) provided indirect evidence that knowledge can mitigate anchoring effects. Participants answered anchoring questions while either under cognitive load or not. Participants also learned either anchor-consistent or anchor-inconsistent information. Participants exposed to anchor-inconsistent information exhibited smaller anchoring effect, but only when not under cognitive load. In other words, limiting participants’ ability to use their background knowledge influenced the magnitude of anchoring effects. Finally, Mussweiler and Strack (2000) conducted a study where participants estimated the age of an ambiguous target (“Xiang Long”) after exposure to high or low anchor. Importantly, some participants were led to believe the target was a person, whereas others were given no such information (i.e., the target could be a person, a national monument, or a mountain). Participants exhibited smaller anchoring effects when they could specify the category the target belonged to as compared with when they lacked this information. Findings Suggesting Knowledge Does Not Moderate