Background and Goals

Prior research reveals that distinct types of motivation affect the risks that people take in different kinds of situations.²

• People with strong prevention motivation—who care deeply about preserving what they have and avoiding worse states—are risk-averse under most circumstances.³
• By contrast, people with strong promotion motivation—who are most concerned with making progress and avoiding a sense of stagnation—tend to be risk-seeking in most cases.

However, after experiencing a recent loss or gain, these tendencies can reverse. For instance, after incurring a loss, prevention motivation leads to especially risky behavior if such a choice offers the chance of returning to a satisfactory status quo.⁴

Research had not yet established whether prevention motivation would similarly motivate risky behavior after other people in a person’s social context experience a recent gain.

• In this case, even though the person’s actual circumstances have not changed, others’ gains function as a new reference point against which the individual falls short.
• As a result, this situation could feel like a loss because the person has failed to keep up with their peers.

I conducted quantitative experimental research to answer this question.

Business Objective

While this research was conducted in an academic setting, and thus was not instigated by a particular business goal, one can imagine possible objectives that it might help to satisfy. For instance:

• Identify users most likely to make unsustainably risky investments in a digital investing platform
• Increase the average wager amount in a sports betting app⁵

Research Question

• After people learn that their peers have recently experienced a substantial gain—that is, in cases of socially-defined counterfactual loss—does prevention motivation predict the extent of their risky behavior?

Hypothesis

• The strength of people’s prevention motivation will predict greater risky behavior following counterfactual loss, such that people high in prevention will take larger risks when the risky action offers the sole chance to attain the new socially-determined satisfactory state. [Pre-registration available here.]

Methods

I conducted an online experiment comprising two purportedly unrelated studies: one framed as a personality survey (for which they were compensated $1) and a second framed as an opportunity to beta-test a new Bitcoin investing platform (on which they could invest the $1 earned earlier).

Reasoning

• I chose to conduct this experiment online because it offered me the best chance to recruit a diverse sample that approximated the characteristics of the U.S. adult population within the constraints of the research budget.
• I used a “multiple studies” paradigm in order to minimize participants’ reactions to the motivation questionnaires from biasing their responses during the “app-testing” portion of the study that assessed risky behavior.
• I deceptively told participants that the risk assessment task was a real beta-test for a new app—and offered them the opportunity to invest the actual $1 that they had earned by completing the first part of the study—to maximize the study’s ecological validity, as making real (sometimes risky) micro-investments through digital interfaces is commonplace these days and I wanted the task to feel true-to-life.
• I proceeded with a behavioral dependent measure of risk-taking because people’s beliefs about what they would do are not always aligned with what they actually do (i.e., the “attitude-behavior gap“).

Participants

295 U.S. adults recruited on Amazon Mechanical Turk (“M-Turk”)

• Age: ranging from 19 to 71 (M = 36)
• Sex: 50% male; 49% female
• Race/Ethnicity: 81% White; 8% Black; 5% Asian; 2% Hispanic
• Income: 4% $10K–$20K; 12% $20K–$40K; 21% $40K–$70K; 33% $70K–$100K; 17% $100K–$250K; 9% $250K–$500K; 1% $500K+
• Education: 1% high school diploma; 17% some college; 22% associate’s degree; 15% bachelor’s degree; 36% graduate degree; 7% doctorate

Study Materials

“Study 1 – Personality Survey” Portion:

In this portion of the study, participants completed a self-report measure of prevention and promotion motivation.⁶ All participants then received $1 and were invited to participate in “Study 2”, which they were told would involve beta-testing a new Bitcoin micro-investing app.

“Study 2 – Investing App Beta-Test” Portion:

In this portion of the study, participants first learned about the purported DigiVest app, which would allow them to make micro-investments in Bitcoin. Participants were told that they had the opportunity to invest any portion of their $1 reward from “Study 1” in Bitcoin today, from 0% to 100%.

Importantly, these instructions noted that because the app had not yet officially launched, financial performance in the beta-test would be based on historical market data—which we controlled so that Bitcoin’s value increased over the previous three days.

After learning how to use the “app”, participants were placed into two experimental conditions.

Close-Up of “Bitcoin Performance Update” Manipulation:

Participants in both conditions decided how much of their $1 from completing the first portion of the study would be “invested” in Bitcoin today, with the remainder kept in savings. On-screen messaging reminded all participants that investing in Bitcoin was a risky behavior:

Close-Up of Primary Dependent Measure:

After making their investment decision, participants answered several brief questions about their “app” experience and noted whether they had previously invested in Bitcoin.

Additionally, participants who indicated that they hadn’t previously invested in Bitcoin answered a few questions probing the emotions that they imagined they would feel during the study experience if they had invested in Bitcoin earlier. These items included the extent to which they would feel relieved (an emotion associated with prevention motivation) and happier (an emotion associated with promotion motivation).

Analytic Strategy

• Conducted a linear regression analysis to test for the hypothesized interaction between participants’ prevention motivation and their assigned experimental condition (Counterfactual Loss vs. Control) in predicting risky investment behavior⁷
• Performed exploratory moderated mediation analyses to investigate whether hypothetical relief and/or happiness served as the mediating mechanism between the focal interaction and risky investment behavior⁸
• To access the raw data and R code used to conduct these analyses, please visit the Open Science Framework page for this project.

Results

Primary Analysis:

• As hypothesized, prevention motivation predicted risky investing behavior when participants experienced a socially-defined counterfactual loss.⁹
• However, contrary to my hypothesis about the direction of this effect, stronger prevention motivation predicted decreased risk-taking in this condition.

To explore these results further, please visit the Shiny app associated with this publication, which dynamically generates the model-predicted Bitcoin allocation based upon user-defined levels of prevention and promotion motivation.

Exploratory Moderated Mediation Analysis:

This analysis revealed that hypothetical relief fully mediated the relationship between prevention motivation and risk-taking in the counterfactual loss condition.¹⁰

While exploratory, this result provided additional evidence for the relevance of loss-framed situations—even counterfactual ones—to the prevention motivational system (as prior research indicates that relief is a prevention-focused emotion).

Recommendations and Potential Impact

While this research was conducted in an academic context and thus did not result in any recommendations for a specific business, my findings have important implications for companies whose products and services involve the user or customer taking on some level of risk.

To consider the two possible business objectives proposed above:

• If the business objective is to identify users especially likely to make unsustainably risky investments on a digital investing platform:
  • This company could assess users’ prevention and promotion motives by adding a few simple multiple-choice questions from the validated Regulatory Focus Questionnaire to the registration process.
• If the business objective is to increase the average wager amount in a sports betting app:
  • This company could highlight fellow users’ recent winnings to users with especially weak prevention motivation.

Limitations and Learnings

While this research provided some clarity into our research question, it had some important limitations:

• It is unclear if participants believed that the risky Bitcoin investment would allow them to catch up with their peers who had recently experienced gains.
  • Prior research indicates that such a belief is necessary to motivate risky behavior among people with strong prevention motivation after an actual loss. (Because I prioritized ecological validity in this study, and thus proceeded with the Bitcoin micro-investing paradigm, such clarity about the likelihood of recouping the counterfactual “loss” was not possible.)
  • Future research should investigate how prevention-motivated people respond to counterfactual loss when the probability of catching up with their peers is made explicit.
• While the manipulation of counterfactual loss was experimental, prevention motivation was measured, so we must be careful about any causal inferences that we draw.
  • Subsequent research should seek to replicate these findings with an experimental manipulation of prevention strength (vs. weakness).

The process also revealed a few unexpected learnings.

• In retrospect, it would have been useful to ask participants explicitly whether they believed that investing in Bitcoin offered the possibility of catching up with their peers—that is, “keeping up with the Joneses”.
  • This was one of the first research projects I conducted as a PhD student. It helped me to learn that in subsequent research, I should take extra care during the study design phase to consider possible mediators and moderators that might be worth assessing based on different possible patterns of results.
• From a more theoretical perspective, although actual and counterfactual loss both involve the failure to meet a reference point of some kind, they seem to produce distinct consequences.
  • Thankfully, I had a clear alternative hypothesis (see the full published article for more detail!) that helped me to understand why the direction of the focal effect was the opposite of what I had expected. It was a useful reminder to thoughtfully consider these kinds of alternatives while designing future studies.

Footnotes

1. Model predictions shown here are for participants with low prevention motivation (–1 SD below the mean) versus high prevention motivation (+1 SD above the mean), both with the average level promotion motivation in the counterfactual loss condition. A general linear hypothesis test reveals that this difference in the model-predicted investment at these two levels of prevention motivation (approximately $0.15) is significant (p = .035). ↩︎
2. Zou, X., Scholer, A. A., & Higgins, E. T. (2020). Risk preference: How decision maker’s goal, current value state, and choice set work together. Psychological Review, 127(1), 74–94. https://doi.org/10.1037/rev0000162 ↩︎
3. Crowe, E., & Higgins, E. T. (1997). Regulatory focus and strategic inclinations: Promotion and prevention in decision-making. Organizational Behavior and Human Decision Processes, 69(2), 117–132. https://doi.org/10.1006/obhd.1996.2675 ↩︎
4. Scholer, A. A., Zou, X., Fujita, K., Stroessner, S. J., & Higgins, E. T. (2010). When risk seeking becomes a motivational necessity. Journal of Personality and Social Psychology, 99(2), 215–231. https://doi.org/10.1037/a0019715 ↩︎
5. I’ll refrain from commenting on the ethics of such an objective here! ↩︎
6. Regulatory Focus Questionnaire (Higgins et al., 2001). For the scale validation paper, which includes all 11 items, see: Higgins, E. T., Friedman, R., Harlow, R. E., Idson, L. C., Ayduk, O. N., & Taylor, A. (2001). Achievement orientations from subjective histories of success: Promotion pride versus prevention pride. European Journal of Social Psychology, 31(1), 3–23. https://doi.org/10.1002/ejsp.27 ↩︎
7. In this linear regression analysis (conducted via a straightforward “lm()” call in R), the percentage of participants’ $1 that they chose to “invest” in Bitcoin was regressed on participants’ experimental condition, prevention motivation, promotion motivation, and the interactions between: (a) experimental condition and prevention motivation and (b) experimental condition and promotion motivation. ↩︎
8. To conduct these moderated mediation analyses (as path analyses using the lavaan package in R), I first standardized all variables to facilitate effect size comparisons across models. Then, within each analysis, I specified two regression models and formulas for the conditional indirect, direct, and total effects of interest in accordance with recommended practice for moderated mediation (Preacher, Rucker & Hayes, 2007) and Model 8 within Hayes’ PROCESS macro (Hayes, 2018). In the first model, I regressed the potential mediator on participants’ experimental condition, prevention motivation, and the interaction between experimental condition and prevention motivation, while also controlling for promotion motivation and the interaction between experimental condition and promotion motivation. In the second model, I regressed the dependent variable (the percentage of the $1 “invested” in Bitcoin) on the same predictor variables plus the potential mediator of interest. Within these models, standard errors and confidence intervals were estimated using bootstrap methods. ↩︎
9. When examining the results of the regression analysis, the only significant predictor of Bitcoin allocation was the focal interaction between experimental condition and prevention motivation (b = –13.41, p = .011). Probing this interaction revealed that the effect of prevention motivation on Bitcoin allocation was significant in the counterfactual loss condition (b = −8.31, SE = 3.92, t = −2.12, p = .035; 95% CI [−16.03, −0.59]) but non-significant in the control condition (p = .140). These results indicate that, when facing socially-defined counterfactual loss, as prevention motivation increased, participants allocated less to a risky Bitcoin investment. ↩︎
10. The interaction between prevention motivation and counterfactual loss was associated with hypothetical relief (β = −0.13, p = .022), and the effect of hypothetical relief was associated with Bitcoin allocation (β = 0.42, p < .001). Importantly, the indirect effect of prevention motivation on Bitcoin allocation through hypothetical relief was significant and negative for individuals in the counterfactual loss condition (β = −0.09, SE = 0.04, p = .021, 95% CI [−0.16, −0.01]), but not the control condition (p = .440). These results indicate that the effect of prevention motivation on Bitcoin allocation is fully mediated by hypothetical relief in the counterfactual loss condition. ↩︎