1 Introduction
The risk gap refers to the phenomenon that the extent to which people recommend others to engage in risky behaviors differs systematically from the extent to which they are themselves willing to engage in the same risky behaviors (e.g., (Reference Beisswanger, Stone, Hupp and AllgalerBeisswanger et al., 2003, Reference Kray and GonzalezKrayGonzalez, 1999, Reference Ubel, Angott and Zikmund-FisherUbel et al., 2011)). Recent research has provided a novel solution to this puzzle — when deciding whether to engage in the risky behavior themselves, people focus on the extent to which their peers are engaging in the risky behavior (i.e., the descriptive norm), but when recommending the same behavior to others, people are significantly less influenced by their peers’ behaviors (Reference Helfinstein, Mumford and PoldrackHelfinstein et al., 2015). For example, people are more likely to jaywalk if they see most others jaywalking; however, when recommending others whether or not to jaywalk, they are not influenced by whether or not most others are jaywalking. However, although people ignore what their peers are doing when recommending risky behaviors to others, it does not necessarily mean that they are not considering any normative information. In this research, we argue that people do attend to normative information when making recommendations to others, but that different types of normative information — descriptive norms vs. injunctive norms — influence their own decisions as opposed to their recommendations to others, respectively.
People’s perceptions of what their peers are doing capture one form of normative information, namely descriptive norms, which refers to the extent to which others are engaging in a given behavior. Another important form of normative information is injunctive norms, which refers to the extent to which most others would approve or disapprove of a given behavior (Reference Cialdini, Reno and KallgrenCialdini et al., 1990, Reference Kallgren, Reno and CialdiniKallgren et al., 2000). Prior research on individual risky decision-making has focused largely on descriptive norms. For example, peers’ engagement in risky behaviors predicts individuals’ own likelihood of engaging in those behaviors in various domains, such as smoking (Reference Christakis and FowlerChristakisFowler, 2008, Reference KobusKobus, 2003) and alcohol consumption (Reference Borsari and CareyBorsariCarey, 2001, Reference Rosenquist, Murabito, Fowler and ChristakisRosenquist et al., 2010). Receiving information that a friend or a family member engaged in a risky behavior (e.g., bungee jumping) dramatically increases participants’ own likelihood of engaging in the same risky behavior (Reference Bursztyn, Ederer, Ferman and YuchtmanBursztyn et al., 2014).
Although people’s own risky behaviors are influenced by what they believe their peers would do in a similar situation, descriptive norms have been found to have little effect when people’s make recommendations to others (Reference Helfinstein, Mumford and PoldrackHelfinstein et al., 2015). In that study, participants indicated either their own willingness to engage in various risky activities (i.e., invest in risky stocks, drive without a seat belt, bungee jump), or indicated their willingness to recommend engaging in the same risky activities to a loved one. In the same study, participants also rated how many of the people they know engaged in the same risky activities. The study found that participants’ perception of their peers’ risky behavior were positively correlated with their own willingness to engage in the same behavior, but were uncorrelated with their willingness to recommend the same behaviors to their loved one.
Although that study examined the influence of descriptive norms, we hypothesize that, when making recommendations to others, people attend to injunctive norms. Two lines of research are consistent with this hypothesis. First, when making decisions that do not require cognitive elaboration, people are more likely to follow descriptive norms, but when making decisions that require cognitive elaboration, people are more likely to follow injunctive norms (Reference Kredentser, Fabrigar, Smith and FultonKredentser et al., 2012). As people are concerned about defending their decisions when making recommendation to others (Reference Kray and GonzalezKrayGonzalez, 1999, Reference Ubel, Angott and Zikmund-FisherUbel et al., 2011), which is likely to involve cognitive elaboration, injunctive norms should have a stronger impact when people make recommendations to others than when they make decisions for themselves.
Second, people pay more attention to injunctive norms rather than descriptive norms when their goal is to build and maintain social relationships (Reference Cialdini, Trost, Gilbert, Fiske and LindzeyCialdiniTrost, 1998). For example, when primed with words associated with descriptive norms, people were faster at identifying words related to accuracy and efficiency, whereas when primed with injunctive norms, they were faster at identifying words related to social approval (Reference Jacobson, Mortensen and CialdiniJacobson et al., 2010). In the case of the risk gap, people are likely to care more about the relationship when making recommendations to others than when deciding for themselves. Thus, we would expect that injunctive norms, compared to descriptive norms, would have a stronger influence on people’s recommendations to others than when they make decisions for themselves. To test this hypothesis, we conducted a study that closely follows (Reference Helfinstein, Mumford and PoldrackHelfinstein et al., 2015) method.
2 Method
We hypothesized an interaction between a within-participant factor (perceived norms) and two between-participant factors, norm type (injunctive vs. descriptive) and decision type (deciding for self vs. making recommendations to other). We maximized statistical power by using multiple trials, using mixed within- and between-participant design, and treating trials as a random factor (Reference Judd, Westfall and KennyJudd et al., 2012).
2.1 Participants
A survey seeking 400 US residents was posted on Amazon Mechanical Turk. In response, 415 participants (244 women, 156 men, 4 other, 11 unreported; mean age 37.25 years) completed the study.Footnote 1 Participants were randomly assigned to one cell of a 2 (descriptive vs. injunctive norms) X 2 (self vs. other) design.
2.2 Measures
We presented the participants with 30 risky behaviors from the Domain Specific Risk-taking Scale (Reference Blais and WeberBlaisWeber, 2006). This scale captures risk-taking in five domains, including social, recreation, health/safety, financial, and ethics. First, all participants were asked to rate their perceived norms for each behavior. However, the type of perceived norms that participants rated varied by condition. Those in the descriptive norms condition were asked: “Please indicate to what extent your peers would engage in the described activities or behaviors below.” They were then presented with the 30 risky behaviors and asked to rate each on a 7-point scale ranging from “−3 Definitely would not” to “3 Definitely would.” Those in the injunctive norms condition were asked: “Please indicate to what extent your peers would think people should (or should not) engage in the described activities or behaviors below.” They were presented with the same 30 behaviors and a rating scale ranging from “−3 Definitely should not” to “3 Definitely should.” The order of the 30 items was independently randomized for each participant.
Next, the participants in the self condition were asked to rate the question: “Please indicate the likelihood that you would engage in the described activity or behavior if you were to find yourself in that situation,” on a scale ranging from “1 extremely unlikely” to “7 extremely likely.” Those in the other condition were asked to rate the question, “Please indicate the likelihood that you would encourage a loved one to engage in the described activity or behavior if you were to find your loved one in that situation,” on the scale ranging from “1 extremely unlikely” to “7 extremely likely.”
Finally, following (Reference Helfinstein, Mumford and PoldrackHelfinstein et al., 2015), we collected data on two control variables. All participants were asked to rate the perceived riskiness of each behavior (i.e., “How risky do you perceive this activity to be?”) on the scale ranging from “1 Not at all risky” to “7 Extremely risky.” Further, all participants were asked to rate the perceived benefits of each behavior (i.e., “Please indicate the benefits you would obtain from each situation”) on a scale ranging from “1 no benefits at all” to “7 great benefit.” The survey ended with a list of demographic background questions.
3 Results
As an initial examination of our hypothesis, for each participant, we computed the correlation between the independent variable (i.e., participants’ perceptions of the norm) and the outcome variable (i.e., their likelihood of engaging in / recommending the risky behavior), across the 30 items. Of the 415 participants, six participants gave uniform responses to all the items used to measure the independent variable and/or outcome variable, and were thus dropped from the analysis. The final sample size was 409 (238 women, 156 men, 15 unidentified; 309 White/Caucasian, 35 African American, 16 Latin American, 13 Native American, 30 South/East/Southeast Asian, 6 unidentified; mean age 37.25 years, SD = 0.61).
Consistent with our hypothesis, in the self condition, participants’ decisions to engage in the risky action were more strongly predicted by the perceived descriptive norm, than by the perceived injunctive norm. By contrast, in the other condition, participants’ decisions to recommend the risky action were more strongly predicted by the perceived injunctive norm, than by the perceived descriptive norm. The mean correlations with responses were:
Self, descriptive: r = .67, SE = .03, 95% CI [.62, .73]
Self, injunctive: r = .63, SE = .02, 95% CI [.58, .69]
Other, descriptive: r = .59, SE = .03, 95% CI [.54, .64]
Other, injunctive: r = .72, SE = .02, 95% CI [.67, .78]
Figure 1 summarizes this pattern of results for a parallel analysis using regressions instead of correlations.Footnote 2
Figure 1: Predicted beta (with standard error) of perceived norm effect.
To test whether the difference in the correlation score differs across conditions, we regressed each participant’s correlation score on the decision type condition (making recommendations to other = −.5, deciding for self = .5), the type of norm condition (−.5 = injunctive norms, .5 = descriptive norms), and the interaction of these two variables. We found a significant interaction, b = 0.17, 95% CI [.07, .27], SE = .052, t = 3.33, p < .001. To unpack the interaction, we regressed the correlation score on the decision type in each of the norm type condition respectively. When making recommendations to others, injunctive norms show a significantly stronger influence than descriptive norms (b = −0.13, 95% CI [−.21, −.06], SE = .04, t = −3.55, p < .001). When deciding for self, the two types of norms showed no difference in the influencing power (b = 0.04, 95% CI [−.03, .11], SE = .04, t = 1.11, p = .267). Next, we regressed the correlation score on the norm type in each of the decision type condition respectively. Descriptive norms showed as a significantly stronger impact when participants were making decision for themselves than making recommendations to others (b = 0.08, 95% CI [.004, .16], SE = .04, t = 2.09, p = .038). By contrast, injunctive norms showed a significantly stronger impact when participants were making recommendations to others than making decision for themselves (b = −0.09, 95% CI [−.16, −.02], SE = .033, t = −2.69, p = .0078).Footnote 3
4 Discussion
The current findings demonstrate that different types of normative information influence individuals’ decisions to engage in a risky behavior themselves and to recommend this behavior to others. Consistent with prior research (Reference Helfinstein, Mumford and PoldrackHelfinstein et al., 2015), we found that the extent to which people think that their peers’ are engaging in various risky behaviors has a stronger influence on people’s own decisions about whether to engage in the risky behaviors than on whether they would recommend others to engage in the same behaviors. However, expanding the work of (Reference Helfinstein, Mumford and PoldrackHelfinstein et al., 2015), we found that different types of normative information have different decision weights when people make recommendation to others than when making their own decisions — the extent to which they think their peers would approve a given risky behavior. Thus, the conclusion that normative information is irrelevant when people recommend risky behaviors to others was premature; instead, people focus on different types of normative information when making decisions for themselves and when making recommendations to others. Further, our findings suggest that, when people seek independent advice or recommendations from others, instead of just an individual’s personal assessment, they are also exposed to second hand normative biases.
Our findings contribute to the broader field of decision-making for self versus others, which seems to have concluded that people make more rational decisions when making recommendations to others than deciding for themselves (e.g. (Reference Beisswanger, Stone, Hupp and AllgalerBeisswanger et al., 2003, Reference Helfinstein, Mumford and PoldrackHelfinstein et al., 2015, Reference Zikmund-Fisher, Sarr, Fagerlin and UbelZikmund-Fisher et al., 2006)). For example, people display fewer cognitive biases (Reference Ubel, Angott and Zikmund-FisherUbel et al., 2011) and place greater weight on more important factors (Reference Kray and GonzalezKrayGonzalez, 1999) when advising others than when deciding for themselves. However, in the case of risky decisions, our findings indicate that it is not necessarily the case that recommendations to others are better or worse than decisions for the self; instead, these two types of decisions are influenced by different normative information. An important question for future research is whether existing asymmetries in self-other decision-making might also be due to differential effect of descriptive vs. injunctive norms rather than more vs. less biased decision-making.
The current findings highlight the importance of considering social influence in the research of risky decision-making. For example, studies using data from the Health and Retirement Study administered by the Institute for Social Research at the University of Michigan showed that investors find the market more attractive and are more likely to buy stocks (compared to relatively less risky options such as investing in bonds) when more of their peers participate (Reference Hong, Kubik and SteinHong et al., 2004, Reference Ivkovic and WeisbennerIvkovicWeisbenner, 2007). In these studies, the perceived peer behavior is akin to the descriptive norm.
Future studies should also examine both the descriptive and injunctive norms, as well as involving behaviors such as making recommendations to others. Although we argued that people’s perceived norms influenced their own risky decisions and their recommendations to others, future studies should take a closer look at the causal mechanisms. Results from the current study are correlational. Future studies should manipulate the perceived norms and ask whether it directly causes shifts in risk preference. Further, future studies should also examine the sources of people’s perceived norms. Are these perceptions relatively accurate, or are they systematically biased in certain directions?
To conclude, the pervasive effect of normative information is relevant not only to our own decisions to engage in a risky behavior, but also to our recommendations to others. The current study identifies an important influence of injunctive norms when people make recommendations to others. Perceptions of whether a risky behavior is approved or disapproved by others influence the likelihood that individuals will recommend others to engage in the behavior. Further, the effect of injunctive norm on risky behavior recommendation is independent of perceived risk and perceived benefits of the behavior. Overall, our finding demonstrates the importance of differentiating descriptive and injunctive norms. We call for future research to compare and contrast the influence of these two different types of norms on making decisions other than just risky decisions, considering a wider range of situations and a wider range of tasks.
Open access
