¡Descarga Question-Behavior Effect: A Meta-Analysis of the Influence of Questions on Behaviors y más Guías, Proyectos, Investigaciones en PDF de Psicología solo en Docsity!
A meta-analytic synthesis of the question-behavior effect Eric R. Spangenberg, Ioannis Kareklas, Berna Devezer, David E. Sprott PII:DOI: S1057-7408(15)00110-2doi: 10.1016/j.jcps.2015.12. Reference: JCPS 521 To appear in: Journal of Consumer Psychology Received date:Revised date: 27 November 20121 December 2015 Accepted date: 11 December 2015
Please cite this article as: Spangenberg, E.R., Kareklas, I., Devezer, B. & Sprott, D.E.,A meta-analytic synthesis of the question-behavior effect, Journal of Consumer Psychology (2015), doi: 10.1016/j.jcps.2015.12.
This is a PDF file of an unedited manuscript that has been accepted for publication.As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proofbefore it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers thatapply to the journal pertain.
ACCEPTED MANUSCRIPT
Question-Behavior Effect Meta-Analysis 1 A Meta-Analytic Synthesis of the Question-Behavior Effect
Eric R. Spangenberg University of California, Irvine Ioannis Kareklas University at Albany, State University of New York Berna Devezer University of Idaho David E. Sprott Washington State University Author Note
Thanks to our many colleagues for their motivation, providing data, advising on coding or analytic procedures, and offering friendly review. A list of those contributing to this Sisyphean effort would be prohibitively long; question-behavior scholars listed in our references serves as a good proxy for identifying those to whom we are grateful. Our thanks also to four anonymous reviewers, the associate editor, former editor Connie Pechmann and editor Amna Kirmani for their patience and helpful direction in the review process. All authors contributed equally to this project.
Corresponding author: Eric R. Spangenberg, Dean, Professor of Marketing and Psychology, University of California, Irving, The Paul Merage School of Business, 4293 Pereira Dr. Suite 5600, Irvine, CA 92697-
- E-mail: ers@uci.edu
ACCEPTED MANUSCRIPT
Question-Behavior Effect Meta-Analysis 3 A Meta-Analytic Synthesis of the Question-Behavior Effect ―There is nothing so theoretical as a good method‖ (Greenwald, 2012) This quote is the title of an article in which Greenwald (2012) documents the near- impossibility of establishing superiority of one among competing theoretical interpretations of a novel, interesting, and empirically well-established phenomenon. One such simple effect was introduced to psychology by Sherman (1980) as ―the self-erasing nature of errors of prediction.‖ The effect is seemingly straightforward: Question a person about his or her behavior regarding a target action and the likelihood of performing the behavior in the future significantly changes. Much early work was conducted under either the label of self-prophecy (Greenwald, Klinger, Vande Kamp, & Kerr, 1988) or mere-measurement (Morwitz, Johnson, & Schmittlein, 1993). Methodological similarities prompted scholars from both streams to join the taxonomically separate areas under an overarching paradigm now referred to as the ―question-behavior effect‖ (Sprott et al., 2006). The question-behavior effect has been demonstrated for a variety of behaviors including: registering to vote and voting in elections (Gerber & Green, 2005a;b; Greenwald, Carnot, Beach, & Young, 1987); influencing consumer purchases (Morwitz et al., 1993); reducing cheating in college (Spangenberg & Obermiller, 1996); impacting risky behaviors among adolescents (Fitzsimons & Moore, 2008); increasing exercise and other health-related behaviors (e.g., Spangenberg, 1997; Sandberg & Conner, 2009); increasing recycling (Sprott, Spangenberg, & Perkins, 1999); reducing implicit gender stereotyping (Spangenberg & Greenwald, 1999); and helping a charity (Liu & Aaker, 2008). Far from exhaustive, this list gives one a sense of the practical import of question-behavior-based interventions. However, a synthesis of conditions
ACCEPTED MANUSCRIPT
Question-Behavior Effect Meta-Analysis 4
under which the effect consistently manifests is lacking. Thus, one aim of our work is to identify effective means and contexts within which the technique can be applied. Although the robustness and practical importance of the effect is evident, its theoretical underpinnings remain a matter of some debate. Much of this debate stems from the sheer number of theories proposed for the effect. For such a relatively simple influence technique, researchers have suggested a remarkable quantity and diversity of explanations (Dholakia, 2010; Sprott et al., 2006), with relatively few having been tested and very few given the attention of more than a single study. Further, most existing support comes from theoretically relevant boundary conditions, rather than direct mediational process evidence. Theoretical uncertainty for such an operationally clear-cut effect may be unsurprising, especially to those who have wrestled with similarly ―simple‖ effects, finding them more complex than initially expected (e.g., halo effect; Thorndike, 1920). While several question-behavior studies contain information critical to our understanding, many fail to inform with regard to its theoretical underpinnings. Furthermore, non-systematic consideration of moderators and boundary conditions has hampered our understanding of the effect. Thus, a critical aim of the current work is to make progress on the challenge identified in Greenwald’s (2012) article quoted at our outset—namely, by advancing theoretical understanding of the question-behavior effect ―method‖ through meta-analytic examination of existing work. Opportunely, there is a sufficient quantity of research and consistency in approach across question-behavior studies to allow for meaningful meta-analysis. This empirical synthesis of over 100 published and unpublished question-behavior effect studies represents the most comprehensive investigation of the phenomenon to date. Contributions of our work are two-fold: First, we document the basic nature of the question-behavior effect and develop an understanding
ACCEPTED MANUSCRIPT
Question-Behavior Effect Meta-Analysis 6
Attitude-based Explanations This family of explanations share the premise that questioning activates attitudes which in turn influence future performance of the questioned behavior. The most researched of these explanations is that of attitude accessibility, holding that questioning makes attitudes associated with the behavior more accessible, with those attitudes guiding future behavioral performance (Morwitz et al., 1993). This account is compelling due to evidence that highly accessible attitudes are strong predictors of behavior. As such, attitudes are likely to be employed as decision criteria when the behavioral opportunity emerges (e.g., Fazio, 1989; Alba, Hutchinson, & Lynch, 1991). Morwitz and Fitzsimons (2004) found support for attitude accessibility, while ruling out other theoretical accounts (i.e., label and intent accessibility). Specifically, they found that asking general intent questions about candy bars increased (decreased) people’s choice of candy bars for which they held positive (negative) attitudes, but had no effect on the choice of other products in the category. The effect of increased attitude accessibility implies an automatic rather than an effortful process—a prediction supported by Fitzsimons and Williams (2000). Attitude accessibility is one of the few theoretical mechanisms that has been tested by researchers in comparison to other proposed theories. For example, Spangenberg et al. (2003, study 2), Janiszewski and Chandon (2007, experiment 1A) and van Kerckhove et al. (2012a, study 3) included contexts addressing the extent to which attitude accessibility can account for the question-behavior effect. While more comparative theory testing is warranted, these studies indicate that attitude accessibility does not always account for observed effects. Also challenging attitude accessibility is work showing that questioning people directly about attitudes does not change behavior to the same degree as does asking an intention question (Chapman, 2001). Two additional attitude-based accounts for the effect have been offered, both receiving
ACCEPTED MANUSCRIPT
Question-Behavior Effect Meta-Analysis 7
limited support. One theorizes that in some situations, automatic activation of cognitions triggered by an intention question may make implicit (rather than potentially conflicting explicit) attitudes more accessible. Support for this account comes from Fitzsimons, Nunes, and Williams (2007) who showed that when participants were asked about intentions regarding vice behaviors (e.g., skipping class, drinking alcohol), they engaged in such behaviors more than a control group. Also suggested is attitude polarization, which holds that the effect of repeated questioning makes accessible attitudes toward behavior more extreme (Morwitz et al., 1993). Later research, however, does not corroborate this theoretical perspective (Morwitz & Fitzsimons, 2004). Consistency-based Explanations Various theories for the question-behavior effect are predicated on some form of consistency between the question and behavior. Primary among these is one of the most enduring consistency-based theoretical perspectives in history—cognitive dissonance. As conceptualized by Festinger (1957), cognitive dissonance creates a motivational state aimed at alleviating psychological discomfort generated by an inconsistency among one’s cognitions. The question- behavior-related explanation has primarily drawn on the discrepancy-from-self-concept view developed by Aronson (1968). In support, Spangenberg et al. (2012) point out that the question- behavior technique is reminiscent of classic induced compliance procedures for eliciting cognitive dissonance. Initially proposed by Spangenberg and Greenwald (1999), answering a behavioral question is proposed to elicit normative beliefs regarding a behavior, as well as prior personal failures to behave normatively; the discrepancy between these sets of cognitions evokes dissonance, motivating behavioral change. Consistent with this explanation, research shows that people make self-predictions in socially normative directions (e.g., Sherman, 1980) and they generally under-perform with regard to engaging in normative behaviors (e.g., Spangenberg,
ACCEPTED MANUSCRIPT
Question-Behavior Effect Meta-Analysis 9
because response fluency serves as evidence for behavioral disposition. Janiszewski and Chandon (2007) provide evidence that differential response fluencies underlie the effect. In their first experiment, they asked participants a sole-intent question about a novel brand of ice cream, dual-intent questions about two novel brands, a general-intent question about purchasing ice cream, or did not ask an intent question. Consistent with their theory, only the sole-intent condition influenced behavior where response fluency for only one brand was established. The fluency-based explanation suggests that the question-behavior effect is driven not only by the question, but also by processing at the time of behavioral opportunity. This notion is akin to Sherman’s (1980) original (yet still untested) proposal of script evocation; he argued that questioning creates a ―good person‖ script that is activated when a behavioral opportunity arises and in turn guides behavioral performance. Proposed by Spangenberg, Greenwald, and Sprott (2008), ideomotor theory provides yet another theoretical view consistent with fluency, suggesting that a question activates a perceptual image or idea of an action being questioned and this image guides future behavioral performance. This interpretation is somewhat supported by Levav and Fitzsimons (2006) who demonstrated that ―easy to represent‖ behaviors were more likely to exhibit question-behavior effects. This automatic ―everyday actions‖ interpretation is consistent with findings that question–behavior effects are more likely to occur without conscious cognitive mediation (Fitzsimons & Shiv, 2001; Fitzsimons & Williams, 2000). Motivation-based Explanations A motivation-based process proposed by van Kerckhove et al. (2012a) holds that questioning activates an intention that uniquely guides future behavioral performance by increasing accessibility of intentional concepts in memory and enhances commitment to perform a certain action. These researchers compared the effects of questioning intentions directly with
ACCEPTED MANUSCRIPT
Question-Behavior Effect Meta-Analysis 10
attitudes, finding the question-behavior effect to be stronger when participants answered an intention versus an attitude question. Moreover, by manipulating the timing of response latency measurement (before or after making a choice), the authors showed that once the intention is fulfilled by making a choice, intention-related information was inhibited, providing support for a motivational mechanism. Another of their studies showed the question-behavior effect to persist as the time interval between questioning and choice increased—a finding not replicated with an attitude measure. Another motivation-based approach implicated in the literature is that of implementation intentions (e.g., Dholakia, 2010)—a self-regulatory method using if-then planning to achieve a particular goal (Gollwitzer, 1999; Gollwitzer & Oettingen, 2008). The underlying idea is that questioning facilitates an if-then plan in one’s mind, guiding future behavioral performance. While conceptually similar, an important operational distinction exists between these paradigms; namely, the question-behavior effect relies upon a single question, while implementation intentions are often generated more elaborately by asking people to describe in detail ―where, when, and how‖ they will perform a focal behavior (Bamberg, 2000). Perkins et al. (2007) suggested a motivation-related mechanism, whereby questioning may activate social identities (i.e., self-definitions of group identification) that motivate behavior consistent with activated identities. In essence, social identity serves as a superordinate goal inspiring future behavior. As a point of divergence from van Kerckhove et al. (2012a), social identity activation was originally aimed at accounting for the observed role of social norms in the question-behavior effect and suggests a process driven by self-esteem. Within a socially normative context, Perkins et al. (2007) found that asking respondents if they would recycle in the future activated self-knowledge related to recycling, thereby boosting self-esteem.
ACCEPTED MANUSCRIPT
Question-Behavior Effect Meta-Analysis 12 Theoretical Predictions for Moderating Variables A range of variables consistently occurs across empirical demonstrations of the question- behavior effect allowing application of meta-analytic techniques. While some of these variables do little to enlighten us regarding the nature of the effect, others are informative as to theoretical mechanisms underlying the effect. We now turn our attention to this latter category of variables, including: (1) characteristics of the question and (2) characteristics of the target behavior. For each potential moderator under these two categories, we explore whether attitude-, consistency-, fluency- and/or motivation-based explanations make specific predictions regarding magnitude or direction of the effect. Theoretical predictions are summarized in Table 2. It is important to note that not every explanation makes a theoretical prediction for each of the meta-analytic variables. ––––––––––––––––––––– –––––––––––––––––––––^ Insert Table 2 about here Characteristics of the Question Question response modality. A variety of approaches have been used to ask the question initiating a question-behavior effect. Response modalities for question administration have included computer surveys, paper and pencil instruments, snail mail inserts, telephone calls, posters, and face-to-face interviews. Modes of questioning have been extensively evaluated by researchers (e.g., Holbrook, Green, & Krosnick, 2003) with response modality affecting participants’ performance on self-report questionnaires by triggering different mechanisms (Richman, Kiesler, Weisband, & Drasgow, 1999). In terms of the question-behavior effect, an important dimension underlying response modality is the level of anonymity provided during question administration. When the question is answered under non-anonymous conditions (e.g., an individualized mailer), a consistency-based mechanism predicts larger question-behavior effects due to increased salience of norms and normative dimensions of the questioned behavior
ACCEPTED MANUSCRIPT
Question-Behavior Effect Meta-Analysis 13
(e.g., Bowling, 2005; Cialdini & Goldstein, 2004). The remaining three explanations for the question-behavior effect would not reasonably predict anything regarding response modality. Type of question. Various types of questions have served as independent variables in question-behavior research including measures of intentions, expectations, and predictions. Interestingly, few studies have employed attitude measures, even though one of the major explanations for the effect is built around attitudes becoming more accessible after questioning. The use of different question types is based largely on area-specific research traditions. For example, self-prophecy researchers have most often used self-predictions (e.g., ―Do you predict that you will recycle?‖), while mere-measurement scholars have typically employed measures of intent (e.g., ―Do you intend to increase your number of steps to 12,500 steps per day over the next week?‖). Measurement research finds that relatively subtle changes in survey questions can result in outcome variation—a point raised by prior question-behavior scholars (e.g., Dholakia, 2010; Sprott et al., 2006). For example, a meta-analysis of the theory of reasoned action shows that expectations and intentions differentially relate to other constructs within that framework (Sheppard, Hartwick, & Warshaw, 1988). The attitude-based account of the question-behavior effect, however, does not predict a moderating effect of question type, as predictions, intentions, and expectations should make attitudes equally accessible (denoted as a null effect in Table 2). In terms of consistency, a prediction should likely challenge a person’s prior failings to perform a behavior as norms would dictate (in contrast to a less constrained intention or expectation question), thereby eliciting greater dissonance and larger effects. Similarly, for motivation-based views, the prediction should yield larger effects since the ―I will perform behavior X‖ wording implies a more concrete implementation plan to perform the behavior in the future (Gollwitzer,
- than an intention or expectation. Similar to attitudes, the type of question is not expected
ACCEPTED MANUSCRIPT
Question-Behavior Effect Meta-Analysis 15
turn giving rise to behaviors favoring that particular feature (e.g., choosing high quality items). Attitude accessibility does not predict an effect of the question’s response scale, as the mere act of questioning should activate preexisting attitudes, regardless of the response scale (assuming the behavior is familiar and one toward which people hold well-defined attitudes). In terms of consistency, questions with dichotomous scales would be more likely to increase the accessibility of social norms associated with the questioned behavior (e.g., ―Will you donate a few hours of your time to assist the American Cancer Society in your community?‖ with response options: ―Yes‖ or ―No‖), as compared to questions with continuous or multinomial response scales. The fluency account similarly predicts stronger question-behavior effects for questions with non-continuous scales, since questioning to choose among choice options should be more redundant (and therefore more fluent) with actual behavioral choice. The motivation account does not make any predictions regarding the influence of a question’s response scale. Question-behavior overlap. Prior question-behavior research has only peripherally touched upon the issue of overlap between the question and the questioned behavior. The closest work is by Sprott, Smith, Spangenberg, and Freson (2004) who explore how changes to the question can influence the effect’s magnitude. In that work (study 2), specific predictions resulted in larger effects compared to general predictions or a control condition. While these manipulations were based on a specific question, the findings are more readily interpreted as reflecting the degree of overlap between the prediction and the behavior. All four theoretical explanations predict a positive relationship between overlap and the size of the question-behavior effect. Attitude research has shown that consistency in terms of time, action, context, and target between an attitude measure and behavior yields stronger prediction of behavior (Ajzen & Fishbein, 2005); as such, greater attitude-behavior overlap should lead to stronger question-
ACCEPTED MANUSCRIPT
Question-Behavior Effect Meta-Analysis 16
behavior effects. Cognitive dissonance (as detailed by Sprott et al., 2004) triggered by questions overlapping with a particular behavior are more easily addressed by changes in behavior than dissonance generated by a general question (which could be alleviated via a variety of approaches). Similarly, response fluency should guide behavior more effectively when question and behavior overlap because a simulated response would be more readily applied to the current behavioral decision. The motivation-based account makes a similar prediction. Specifically, from an implementation intentions perspective, the action plan laid out at the time of questioning should lead to the performance of consistent behaviors. If the intent-question asks more general questions, respondents may not specify the target behavior in their implementation plans. Specificity of the question. Research has long established that the level of specificity associated with a question can influence how people react to a survey and whether those measures predict actions. For example, measuring attitudes toward specific behaviors is more predictive of subsequent behaviors than are more general attitude measures (Ajzen & Fishbein, 2005). In contrast, attitude accessibility makes a null effect prediction regarding question specificity. We expect that similarly accessible attitudes will be generated from questions with varying levels of specificity. As noted earlier, only one paper in the question-behavior arena has explored the nature of the question (Sprott et al., 2004), finding specific questions resulting in larger effects. We contend that these findings are more supportive of the question-behavior overlap than question specificity; as such, a consistency-based explanation would not predict anything regarding question specificity. In contrast, as outlined by Janiszewski and Chandon (2007), fluency suggests larger effects for specific questions as general questions should make different behavioral responses equally fluent. As expected in their studies, the question-behavior effect manifested following single intention questions regarding a specific brand, but not when
ACCEPTED MANUSCRIPT
Question-Behavior Effect Meta-Analysis 18
questioning on behavior predict a stronger effect for self-reported (vs. observed) behavior as there should be greater salience of consistency between attitudes and behaviors when one reflects on the questioned behavior (as with self-reports). In support of this expectation, a meta-analysis of the theory of planned behavior found that intentions correlated more highly with self-reported than observed behaviors (Armitage & Conner, 2001). Similarly, the consistency-based explanation of dissonance predicts stronger effects for self-reported behaviors, as reflection upon the behavior at time of performance would likely increase salience of prior failures and/or social norms associated with the behavior. Motivation- or fluency-based views predict a null effect as one would expect similar outcomes from an activated intention (per the motivation of van Kerckhove et al., 2012a) and likewise for levels of fluency (per Janiszewski & Chandon, 2007). Timing of behavior. Procedures used to elicit question-behavior effects vary in terms of elapsed time between questioning and focal behavior measurement. Delay between the question and the behavior ranges from no delay (e.g., exp. 1, Fitzsimons & Shiv, 2001) up to a one-year delay (e.g., Dholakia & Morwitz, 2002). Cognitions generated by a question should become less salient after longer time delays, as the cognitive drivers of the effect would likely decay over time. Therefore, attitude-, consistency-, and fluency-based explanations predict a diminishing influence of questioning over time. For motivation-based processes, however, we expect that the activation of intentions should persist until the moment of intention fulfillment/completion. In support of such a null effect, van Kerckhove et al. (2012a) showed the effect of questioning to persist as the interval between the intention-question and choice task increased. Novelty of behavior. Novelty of the behavior should moderate the size of the question- behavior effect due to differential cognitive structures developed around less versus more familiar behaviors. (Novelty in our analyses was coded as a continuous variable with higher
ACCEPTED MANUSCRIPT
Question-Behavior Effect Meta-Analysis 19
values indicating more novel [i.e., less familiar] behaviors.) Attitude-based explanations predict a negative effect of novelty. As Janiszewski and Chandon (2007, p. 310) aptly observed, ―It is unlikely that consumers hold attitudes toward novel products, so it is unlikely that responding to an intent question could increase attitude accessibility.‖ In addition, prior research has found that attitudes based on direct experience are more accessible (Fazio et al., 1983) and have a greater influence on behavior (Fazio & Zanna, 1981) than those based on indirect experience. Given that novel behaviors are less likely to be experienced, we expect weaker effects of questioning for such behaviors. Some work, however, suggests a positive effect of novelty on the question- behavior effect. For example, Morwitz et al. (1993) found stronger effects of measuring intent for those without prior experience with personal computers (vs. those more experienced); while not explained by attitude accessibility, these findings are consistent with a fluency mechanism (Dholakia, 2010). In particular, questions regarding novel (vs. familiar) behaviors are more likely to create response fluency due to a lack of competing cognitions. When a novel behavioral opportunity is present, a single behavioral response may be evoked with relative ease. Similar to attitude accessibility, dissonance predicts a negative relationship between novelty and effect size as social norms and/or prior behavioral failures are less likely to exist for relatively novel behaviors. The motivation-based perspective, on the other hand, predicts a null effect of novelty. Regardless of how novel a behavior is, once an intention is activated, people should feel equally motivated to satisfy that intention with their subsequent actions. Psychological and social risk of nonperformance. Varying types of risk are associated with performing or not performing behaviors. Regarding risk and prior research in the area, most prominent is work examining adolescent behaviors considered socially risky. Fitzsimons and Moore (2008) show that questioning can increase participation in short-term risky behaviors by
