In the very early days of the COVID-19 pandemic outbreak, fear and anxiety rapidly spread across the population in Germany (1–3). In turn, people started to hoard toilet paper and canned foods or even stole disinfectants from hospitals. To curtail infection rates, governmental authorities announced contact prohibitions, lockdowns, and most prominently mandatory mask-wearing. Some safety behaviors like hand-washing and physical distancing are adherent and actively prescribed by government authorities. An adherence to such safety behaviors is socially desirable and requires individuals to incur an immediate cost for the sake of society. However, panic buying and hoarding are rather egoistic behaviors as resources are limited and deprive others of them. Hence, while fear is spreading across the population (2), motives to wash hands seem to differ from motives to panic buy and hoard. Higher vulnerabilities to the virus caused by, e.g., cancer or chronic somatic diseases, but also mental illnesses, can increase fear and anxiety, which eventually leads to some safety behavior (4–6). The current strongly increasing numbers of new infections during the “second wave” on one hand, and political demonstrations against COVID-19 protection-regulations on the other hand, indicate the strong necessity to understand underlying motives for distinct safety behaviors.

Recent research on the behavioral-psychological impact of COVID-19 reveals the prevalence of safety behaviors like panic buying strongly increased since the onset of political measures during the COVID-19 crisis (3, 7–9). Especially panic buying has recently been a matter of strong interest for the public, the government, and research. Arafat et al. (10) showed that ~80% of the media reports on COVID-19 focused on the impact and causes of panic buying. Over half of these reports showed images of empty shelves to illustrate the topic. Furthermore, a quarter highlighted the rumor about panic buying and the remedial measures. Investigating the causes for panic buying, Arafat et al. (7) showed that a sense of scarcity seems to be the most prominent factor alongside an increased demand and importance of a product and the anticipation of rising prices facilitated this effect. Furthermore, the authors reported rumors, safety-seeking behavior, uncertainty, anxiety reduction, and taking control, but also social learning and lacking trust in governmental actions to face the pandemic as reasons for panic buying [see also (11)]. Müller and Rau (12) recently reported a link between present-bias and panic buying, while patience was associated to staying at home and avoiding crowds. On the other hand, the same study showed that all safety behaviors appear to be linked to fear, or at least COVID-19-related concern. In a systematic review, Yuen et al. (9) defined four factors that caused panic buying during a health crisis: a perception of threat and scarcity of products, fear of the unknown, panic buying as a coping behavior to deal with anxiety and to regain control, and social psychological factors.

However, Oosterhoff (13) showed that the belief that COVID-19 is a severe disease was associated with physical distancing. Self-interest was negatively associated with physical distancing, yet positively with hoarding behavior. Also, disinfection behavior was linked to greater social responsibility, while hoarding was negatively related to social responsibility and social trust. Yet, panic buying and hoarding seem to be rather related to impulsive and egoistic motives, which are particularly amplified by uncertainty [see also Chen et al. (14)]. Also, Nivette et al. (15) found that people scoring high on antisocial behaviors exhibit less compliance with public health measures such as frequent disinfection. Campos-Mercade et al. (16) showed clear associations between pro-sociality and norm-compliance. Sanitizer and face-mask use and self-isolation are self-imposed measures to stay safe oneself, but also to protect members of the community. Hence, it appears that the underlying motive may include adherence, trust toward authorities, as well as altruism.

Taken together, panic buying seems to differ from other safety behaviors. Dysfunctional safety behaviors like panic buying and hoarding maintain individual safety while neglecting possible cost for others. However, adherent safety behaviors are those that slow the spread of COVID-19, while reducing individual life-quality. They highly conform with governmentally recommended public health measures.

No research has yet directly investigated a possible distinction of these constructs with predictive or associative dynamics, although this separation would have an immense practical value. If safety behavior was one-dimensional, good policies should find an optimal tipping point at which people comply with public health measures. Yet, a conceptual distinction would propose more specific interventions to reduce panic buying and other dysfunctional behaviors on one hand, and increase adherent safety behaviors on the other hand. For instance, tailor made information or behavioral intervention campaigns could attempt to target specific risk groups which have shown increased dysfunctional safety behavior, but refuse to comply with hygiene measures. It is thus necessary to understand the psychological and environmental influences and underpinnings of both, dysfunctional safety behaviors like panic buying and adherent safety behaviors like mask wearing.

The present study assesses whether adherent and dysfunctional safety behaviors actually share the same correlates, or whether they are embedded into very different behavioral, attitudinal, intra-, and inter-individual contexts. Based on the literature, it is hypothesized that adherent and dysfunctional safety behaviors are two different sub-constructs, which are both correlated and particularly linked fear and anxiety, but show different demographic, psychological, behavioral and contextual correlates. This would suggest that political interventions by governmental authorities should very specifically aim at certain groups of individuals and adapt interventions in accordance to their motives in order to promote adherent while preventing dysfunctional safety behavior at the same time.

Table 1 shows the aggregated characteristics of the sample.

To define which features were predictive of the two safety behavior dimensions, all variables, including demographic, behavioral, trait- and attitudinal variables were regressed on adherent and dysfunctional safety behavior. In both models, the assumption of homoscedasticity did not apply (Breusch-Pagan test: p < 0.001). Results of a heteroscedasticity-robust regression, however, yield almost identical results to the ordinary least squares regression (see Supplementary Material). None of the predictors showed a critical multi-collinearity with variance inflation factors above 5. Treatment-coded regression parameters are displayed in Table 2.

The regression estimates and the marginal effects revealed similar, as well as divergent correlates of adherent and dysfunctional safety behavior. Figure 1 illustrates the marginal effects for the most pronounced differences in the regression models. COVID-19-related fear is positively associated with both, adherent and dysfunctional safety behavior (F-test in the adherent safety behavior model: F_{(1, 15262)} = 2673.15, p < 0.001, F-test in the dysfunctional safety behavior model: F_{(1, 15262)} = 430.53, p < 0.001, see Figure 1A). Yet, this association was more pronounced in the model predicting adherent safety behavior with non-overlapping confidence intervals. On the other hand, while generalized anxiety showed a positive relationship with both safety behaviors [F_{(1, 15262)} = 47.70, p < 0.001, for adherent and F_{(1, 15262)} = 99.82, p < 0.001, for dysfunctional safety behavior], the association with dysfunctional safety behavior appeared to be stronger (see Figure 1B) even though confidence intervals slightly overlapped (see Table 2).

Trust in governmental actions to face COVID-19 showed a strong positive relationship with adherent safety behavior [F_{(1, 15262)} = 442.43, p < 0.001], but a negative relationship with dysfunctional safety behavior [F_{(1, 15262)} = 75.33, p < 0.001, see Figure 1C]. Similarly, the subjective level of information was positively related to adherent safety behavior [F_{(1, 15262)} = 136.28, p < 0.001], but showed a negative association with dysfunctional safety behavior [F_{(1, 15262)} = 26.51, p < 0.001, see Figure 1D].

Age was differently associated with both safety behaviors. Although there was an upward trend in dysfunctional safety behavior across age [F_{(6, 15262)} = 15.59, p < 0.001], only small differences were found when predicting adherent safety behavior [F_{(6, 15262)} = 3.87, p = 0.001]. More specific, dysfunctional safety behavior increased for aged higher than 34 (see Figure 1E and Supplementary Material).

Similarly, respondents from smaller-sized communities showed more adherent safety behavior [F_{(3, 15262)} = 38.37, p < 0.001]. There was a gradient from metropoles to rural communities. Only the difference between small towns and rural communities was not significant (see Supplementary Material). Such a gradient did not emerge for dysfunctional safety behavior [F_{(3, 15262)} = 2.58, p = 0.052, see Figure 1F and Supplementary Material for marginal effects].

Pronounced differences also occurred across professional groups in predicting adherent safety behavior [F_{(1, 15262)} = 10.86, p < 0.001]. Predominantly, physicians, nursing staff, and people working as paramedics, firefighters and policemen showed less dysfunctional safety behavior than people having other occupations. Likewise, people indicating current unemployment showed more dysfunctional safety behavior than the other groups, except the group with other occupations (see Table 2 and Supplementary Material).

Respondents who indicated to watch TV and visit websites of public institutions to stay informed on COVID-19 showed more adherent safety behavior (all F-values > 98, p < 0.001), while respondents receiving information from acquaintances showed more dysfunctional safety behavior [F_{(1, 15262)} = 14.32, p < 0.001]. Finally, the presence of a mental disease predicted adherent safety behavior positively [F_{(1, 15262)} = 13.89, p < 0.001], but dysfunctional safety behavior negatively [F_{(1, 15262)} = 53.03, p < 0.001].

The COVID-19 pandemic challenged individual lives and still do to this date. While individuals were obliged to isolate themselves and showing correct hygiene behavior, others hoarded toilet paper and groceries. These adherent and dysfunctional safety behaviors may be manifestations of different motive structures and contexts. The necessary question arises, how governmental authorities can encouraged adherent safety behavior while attenuating dysfunctional safety behavior.

To investigate this highly relevant question, we hypothesized that both types of safety behaviors are associated with different set of features in terms of socio-demographics, psychopathology, personality, and COVID-19-specific attitudes. We were able to collect the largest data sample in Germany and one of the largest worldwide on the impact of COVID-19 on mental health. We specifically asked for adherent and dysfunctional safety behaviors during the COVID-19 pandemic. Despite the exploding amount of literature on the current pandemic, no other study has yet been able to report a comparable dataset profoundly investigating distinct safety behaviors in times of COVID-19 in such a detailed manner. In line with the hypothesis, the current results show that adherent and dysfunctional safety behaviors mostly differed in some of their correlates and even revealed opposite associative directionalities with others. However, some correlates were similar.

Certainly, fear and anxiety are to some degree causal for safety-oriented and preventive behaviors. Here, COVID-19-related fear, as well as generalized anxiety, separately showed positive associations with both safety behaviors. It is important to keep in mind that regression parameters are already conditioned on each other: fear and anxiety-related estimates represent the isolated contribution of each of these dimensions of safety behavior. Also, recent literature suggests a distinction between the fear of COVID-19 and anxiety (2, 5, 6, 33, 34). Importantly, the link between COVID-19-related fear and adherent safety behavior appears to be more pronounced than the link between COVID-19-related fear and dysfunctional safety behavior. A reverse pattern is observed for generalized anxiety – although the confidence intervals slightly overlap: the regression coefficient for generalized anxiety is steeper in dysfunctional safety behavior compared to the adherent safety behavior. These findings tentatively suggest that dysfunctional safety behavior originates from a more omnipresent feeling of threat, while adherent safety behavior results from a direct concern about COVID-19. Even more broadly, dysfunctional safety behavior could be a rather egoistic response to the feeling of overall threat. Indeed, the link between the feeling of threat, stress, and selfish behavior is being discussed for over a century now. Since Cannon (35) defined the concept of fight-or-flight, it is still found up to this point that people under acute and chronic stress incline toward less altruism, less moral decision making, and more egoistic choices (36–39). Furthermore, during the initial course of the COVID-19 pandemic, many communities suffered from shortages of goods due to previous panic-buying [see, e.g., (40)]. The perception of the risk of material deprivation may have been amplified by the subjective feeling of stress (2, 41). Thus, dysfunctional safety behavior may neutralize the subjective feeling of threat by ensuring long-term material security.

The finding that adherent safety behavior, but not dysfunctional safety behavior is related to the subjective feeling of information and trust in governmental intervention fits well to the fact that adherent safety behavior is also rather related to fear and concern about COVID-19. Adherent safety behavior could arise as a product of the person's engagement with the pandemic, which would lead to overall higher levels of information, higher trust in governmental actions (perhaps even due to a higher level of information), and overall higher levels of concern. Positive associations between adherent safety behavior and media consumption for the sake of staying informed on COVID-19 further support the argument that more personal engagement with the COVID-19 pandemic results in more adherent safety behavior. Media consumption may spark fear itself, but could also function as a reassuring safety behavior itself. More engaged individuals would more likely attempt to remain up to date on recent developments to evaluate risks and regain a feeling of control. This idea is in line with the negative correlation between generalized anxiety and subjective levels of information regarding COVID-19 [(42), but see also (19)]. Again, there is some evidence that hints toward a difference in quality between COVID-19-related fear and generalized anxiety during the COVID-19 pandemic (5, 33, 34). On the other side, we only find meaningful associations between acquiring information on COVID-19 from acquaintances and dysfunctional safety behavior.

In the data, further differences in adherent and dysfunctional safety behavior occur between age groups. While respondents' age hardly shows any association with adherent safety behavior, older respondents (i.e., older than 34) indicate to engage in more dysfunctional safety behavior. Such increase in dysfunctional safety behavior might reflect an age-related feeling of threat by the virus [see (2) for a detailed insight of the distribution of fear across age groups]: from early on, it has been evident that people of higher age have an increased likelihood to suffer from an unfavorable course of COVID-19, which could eventually result in death (43). Thus, in anticipation of even longer self-isolation in case of infection than in younger people, preparation seems legitimate. Also, elderly are oftentimes less mobile than younger people and lockdowns make longer trips to grocery shops even more difficult. Finally, the pandemic could have cemented already pre-existing consumption styles with elderly individuals tending to buy more of some hygiene products (44).

A rather unexpected finding is that inhabitants of metropoles exhibit less adherent safety behavior than inhabitants of small towns and rural dwellers, a pattern which is not present in dysfunctional safety behavior. Here, pragmatic reasons might contribute the most: rural dwellers could simply have less difficulties to avoid crowds. Perhaps their decision threshold to even use public transit or travel around is also higher due to more autarky. Otherwise put, people living in large cities are more reliant on public transit, and partly, avoidance of crowds is hardly possible. Furthermore, Peters (45) describes the countryside as more vulnerable to potential COVID-19 outbreaks due to a lack of health services. Causes for an increased dysfunctional safety behavior could be manifold. Yet again, it is important to keep in mind that such differences cannot be explained by direct confounding via e.g., occupation or age. These variables have been conditioned upon in the regression model.

Summarizing, the presented results hint toward a differential associative, and thus contextual, embedding of adherent and dysfunctional safety behavior. Both behaviors appear to be independent of each other, which makes a differentiation theoretically and practically reasonable. People with high levels of adherent safety behavior show higher levels of authoritative trust and subjective information levels. They indicate more specific fear of COVID-19 and seem to gather more information via public news channels. Thus, adherent safety behavior could be promoted by increasing governmental responsibility, medial education, and by inducing realistic highly specialized and justified respect of a possible infection with COVID-19. Contrary to that, generalized anxiety is rather associated with panic buying and other dysfunctional safety behaviors. Accordingly, the present data suggest that governmental actions and COVID-19-specific elucidation campaigns should not target the people's general fears triggering dysfunctional safety behavior. They should rather very carefully provide profound information about virus-specific risks and possible protective countermeasures aiming for adherent safety behavior.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by Ethics Committee of the Essen Medical Faculty (20-9307-BO). Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

BW and AS: planning, study design, data collection, data analysis and interpretation, and manuscript writing. ND, VM, MF, and HK: data interpretation and editing. E-MS, MT, and AB: planning, study design, and supervision. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.