From a theoretical standpoint, age-related increments in the susceptibility to COVID-19 solicitations are plausible considering that that older adults shift their focus from pursuing long-term, knowledge-focused goals to pursuing short-term, affect-focused goals (Reed and Carstensen, 2012). According to Socioemotional Selectivity Theory (Carstensen et al., 1999; Mather and Carstensen, 2005; Reed and Carstensen, 2012), these goals center around establishing and maintaining positive affect, leading older adults to exhibit a so-called “positivity effect” in their cognitive processing: Unlike younger adults, who are often more affected by negative stimuli (thus exhibiting a “negativity bias,” Kisley et al., 2007), older adults selectively focus on and remember positively valenced information better than they do negative information. Correspondingly, older adults are theorized to place greater weight on benefits than risks (Mikels et al., 2015). This, combined with a reduced ability to learn to avoid risks, may lead older adults to engage in suboptimal risk taking (e.g., Denburg et al., 2001; Mata et al., 2011; Mikels et al., 2013). As a result, older adults might pay undue attention to the benefits associated with COVID-19 solicitations while simultaneously disregarding their potential risks or costs, making older adults more likely to fall for these scams than their younger peers are.

Data concerning vulnerability to fraud, particularly to COVID-19 related scams, are just starting to emerge. The present study was designed to evaluate whether age differences exist in susceptibility to COVID-19 scams, and possible mechanisms that might help explain these differences. Younger, middle-aged, and older adults were asked to indicate their intentions to respond to or interact with five solicitations based on real-life COVID-19 products or scams that may result in financial loss or information/identity theft (or even bodily harm, in the case of products that are supposed to protect against COVID-19). In addition, participants evaluated each solicitation’s perceived genuineness, benefits, and risks.

Grounded in past research, we hypothesized that a higher intention to act on COVID-19 solicitations would be predicted by higher genuineness ratings, higher benefit ratings, and lower risk ratings associated with each of the solicitations (e.g., Hanoch et al., 2006; Schoepfer and Piquero, 2009; Wood et al., 2018). Following from Socioemotional Selectivity Theory, we hypothesized that older adults would perceive higher benefits and lower risks, thus reporting increased intentions to respond to COVID-19 solicitations.

Finally, we examined the role played by individual difference measures that have been shown to be associated with susceptibility to fraud, age, or both: Previous experience with fraud (Titus and Gover, 2001; Schoepfer and Piquero, 2009), receptivity to non-sensical information (“bullshit” information, Pennycook and Rand, 2019), skepticism toward ads (Obermiller et al., 2005; MacAlvanah et al., 2015; Anderson, 2016), stress levels (Anderson, 2013), and impulsivity (AARP, 2003; Anderson, 2016; Jones et al., 2019). Specifically, we evaluated whether covariates that differ between age groups may explain the proposed age differences in scam susceptibility.

United States participants were recruited through Prolific in June 2020, 3 months after the WHO identified the COVID-19 crisis as a pandemic. Prolific is an online panel provider with access to ca. vetted 40,000 participants (Palan and Schitter, 2018). Data collected through Prolific have been found to be comparable to the data collected in laboratory-based settings, with the exception that participants recruited through Prolific are more diverse as well as slightly older (Porter et al., 2018).

Among the potential participants approached, 5 dropped out after and 20 withdrew before beginning the study. Of the N = 231 participants we recruited, 21 were manually excluded due missing data for half or more of the survey, providing no demographic information, and/or failing attention checks; one participant was excluded because they did not provide consent prior to taking the study and should have been screened out automatically but was not. The final sample consisted of N = 210 participants (18–84, M = 48.03, SD = 18.56 years; see Supplementary 2A for a power analysis); three age bins were used to guarantee similar numbers of younger, middle-aged and older adults. In total, we recruited 68 younger (18–40 years, M = 25.67, SD = 5.93), 79 middle-aged (41–64 years, M = 49.86, SD = 7.20), and 63 older adults (65–84 years, M = 69.87, SD = 4.50).

Using a within-subjects design, each participant reviewed five COVID-19 related messages and ads based on real-life scams (see see Supplementary 1A for items concerning each solicitation; the solicitation materials themselves can be obtained from the corresponding author): An email ostensibly issued by the WHO, a text message warning about alleged exposure to COVID-19, a colloidal silver (disinfectant) ad for an “all-purpose remedy” and a vaccine announcement claiming a new vaccine could cure COVID-19 in hours. Finally, to account for the possibility that some might generally distrust COVID-19 related offers, even when they are genuine, participants reviewed a legitimate face mask ad. Although some solicitations entailed potentially interactive elements (such as a button or link), participants simply reviewed pictures of each message or ad without interacting with them.

After providing informed consent, participants completed questions about the five solicitations, the order of which was randomized and counterbalanced across participants. In a next step, participants responded to all individual difference measures. With the exception of the Kessler Stress Scale, which was presented first, the order of these scales was randomized and counterbalanced as well. At the end of the survey, participants provided demographic information and were debriefed about the aim of the study.

All analyses were conducted in RStudio Version 1.3.1093. Age group differences between younger, middle-aged, and older adults were assessed using non-parametric, rank-based between-subject ANOVAs and ANCOVAs (Results were comparable when conducting Kruskal-Wallis tests instead). Bonferroni-corrected post hoc tests comparing two groups at a time were conducted using independent-sample Wilcoxon tests. Regression results were based on generalized linear models, and Pseudo-R² for each regression model was calculated using the Nagelkerke formula in R’s “rcompanion” package (Mangiafico, 2020). Note that unlike R², which indexes explained variance, Pseudo-R² may be better understood as an index of model fit. Because we tested a high number of predictors, we explicitly differentiate between findings that are significant at the conventional alpha value (0.05) and those that remain significant at a stricter alpha value (0.001).

In a first step, we compared responses between the face mask solicitation, which served as a control scenario, and the four fraudulent scenarios. To this end, we averaged participants’ action intentions (Cronbach’s α = 0.66) and ratings of perceived genuineness (Cronbach’s α = 0.52), risks (Cronbach’s α = 0.52), and benefits (Cronbach’s α = 0.73) across the four scams. Compared to the face mask solicitation, participants were less willing to respond to scams (M_mask = 2.44, SD_mask = 1.80, M_scams = 1.73, SD_scams = 0.90, V = 2,404, p < 0.001) and less likely to perceive them as genuine (M_mask = 4.17, SD_mask = 1.54, M_scams = 2.51, SD_scams = 0.98, V = 1,384.50, p < 0.001). In addition, the four scams were perceived to be more risky (M_mask = 3.71, SD_mask = 1.81, M_scams = 5.57, SD_scams = 1.00, V = 16,450, p < 0.001) and less beneficial (M_mask = 3.91, SD_mask = 1.62, M_scams = 2.19, SD_scams = 1.09, V = 1,227.50, p < 0.001). Thus, participants correctly differentiated between genuine and fraudulent solicitations.

To determine whether to include or exclude the face mask scenario in our analyses, we next examined inter-correlations for action intention, genuineness, benefit, and risk ratings across all five scenarios (see Supplementary 2B, Supplementary Table 1). Although the face mask scenario served as a control scenario, responses to this scenario showed considerable correlations with responses to the four scams (most ps<0.05 to <0.001). Furthermore, scale reliability for the four variables of interest (i.e., action intentions, perceived genuineness, perceived risks, perceived benefits) remained about the same or improved when including rather than excluding responses concerning the face mask solicitation. As a result, subsequent analyses were based on averages derived from all five solicitations than rather just the four fraudulent scenarios (Cronbach’s α = 0.75 action intentions, α = 0.52 perceived genuineness, α = 0.62 perceived risks, α = 0.71 perceived benefits). Supplementary Analyses for each individual solicitation are provided in Supplementary 2B, Supplementary Table 2. Supplementary Analyses based on an average of the four scams (rather than all five solicitations) are provided in Supplementary 2C, Supplementary Tables 3–6.

Descriptive statistics and age group comparisons for all demographic, individual difference, and solicitation measures are provided in Table 1.

ANCOVA results suggested that older age is associated with various protective factors. To examine the predictive power of each assessed covariate, we regressed action intentions on all demographic, individual difference, and solicitation measures (corresponding regression models predicting perceived genuineness, benefit rating, and risk rating are provided in Supplementary 2D, Supplementary Tables 7–9). When examining each predictor separately (Table 2, middle columns), males, married participants, and those relaying a more conservative worldview were more inclined to respond to the solicitation (all ps < 0.05). In addition, stronger intentions were positively associated with bullshit receptivity, sensation seeking, positive urgency, negative urgency, perceived genuineness, and benefit ratings, and negatively associated with ad skepticism and risk ratings (all ps < 0.001).

Due to the possibility that the relationship between action intentions and age may be curvilinear (i.e., peaking in middle age), we examined both age and age². Neither proved predictive of action intentions, although age² approximated significance. Thus, we entered age² instead of age into a joint model in which all covariates were considered simultaneously (Table 2, right columns). At a conventional alpha value (0.05), this model suggested that higher action intentions were associated with higher educational attainment, being married, having previously fallen victim to financial fraud, and perceiving the solicitations as genuine. At a stricter alpha value (0.001), action intentions were linked to higher levels of positive urgency, higher benefit ratings, and lower risk ratings. In this model, age² did not approximate significance.

We found no consistent evidence for age differences in the willingness to respond to COVID-19 scams. At odds with Socioemotional Selectivity Theory, older adults perceived COVID-19 solicitations to offer significantly fewer benefits than both younger and middle-aged adults did. This difference was no longer significant when accounting for age differences in the susceptibility to non-sense (“bullshit”) information, which older adults appear more wary of (Erlandsson et al., 2018).

Our findings thus contribute to an emerging literature suggesting that older age might be associated with certain protective factors that might lower older adults’ vulnerability to fraud. To illustrate, a novel study finds that older adults score higher on measures of emotional understanding than those under the age of 65, and that higher levels of emotional intelligence, including higher levels of emotional understanding, shield consumers against falling for scams (Mueller et al., 2020). Other recent studies report that older adults are less amenable to persuasion and more sensitive to scam-related and financial risks than younger generations are (Rolison et al., 2014; Rolison et al., 2019; Mueller et al., 2020).

Our findings add to the understanding of decreased scam susceptibility in old age (see Modic and Lea, 2013; Mueller et al., 2020) but the mechanisms underlying the potentially increased scam susceptibility of younger age groups remain poorly understood. In an effort to identify potential predictors of COVID-19 scam compliance, we studied the role played by a range of individual difference measures. Various covariates predicted COVID-19 scam susceptibility when examined by themselves but only a limited number of predictors remained significant when other covariates were accounted for.

With regard to demographic variables, we found that sensu past research, educational attainment (Titus and Gover, 2001; DeLiema et al., 2017) and marital status (DeLiema et al., 2017) were associated with increased vulnerability to fraud. It has been theorized that this is because better-educated people hold “wider interests, engage in a broader range of activities, and have more consumer participation in the marketplace than other demographic groups, thereby increasing their exposure to fraudulent solicitations and transactions” (Titus and Gover, 2001, p. 133). It is possible that since marriage allows spouses to pool their finances, married consumers, too, are more likely to make certain investments or purchases that expose them to fraud attempts.

With regard to individual difference measures, we found that only two individual difference measures were associated with participants’ action intentions when also accounting for other predictors. Specifically, we found that positive urgency predicted higher action intentions across both by itself and in conjunction with other factors. Furthermore, a previous history of fraud susceptibility (Titus and Gover, 2001; Schoepfer and Piquero, 2009) emerged as a new predictor when controlling for other covariates.

Finally, with regard to solicitation-based measures, stronger action intentions were associated with stronger perceptions that a solicitation was genuine in nature, high in benefits, and low in risks (e.g., Hanoch et al., 2006; Schoepfer and Piquero, 2009; Wood et al., 2018). Thus, the observed combination of risk factors in the joint regression model suggests that scam susceptibility is the result of an impulse control issue that is not easily inhibited, not even by past experiences of scam victimization.