Comparative optimism about infection and recovery from COVID‐19; Implications for adherence with lockdown advice

Abstract Background Comparative optimism, the belief that negative events are more likely to happen to others rather than to oneself, is well established in health risk research. It is unknown, however, whether comparative optimism also permeates people’s health expectations and potentially behaviour during the COVID‐19 pandemic. Objectives Data were collected through an international survey (N = 6485) exploring people’s thoughts and psychosocial behaviours relating to COVID‐19. This paper reports UK data on comparative optimism. In particular, we examine the belief that negative events surrounding risk and recovery from COVID‐19 are perceived as more likely to happen to others rather than to oneself. Methods Using online snowball sampling through social media, anonymous UK survey data were collected from N = 645 adults during weeks 5‐8 of the UK COVID‐19 lockdown. The sample was normally distributed in terms of age and reflected the UK ethnic and disability profile. Findings Respondents demonstrated comparative optimism where they believed that as compared to others of the same age and gender, they were unlikely to experience a range of controllable (eg accidentally infect/ be infected) and uncontrollable (eg need hospitalization/ intensive care treatment if infected) COVID‐19‐related risks in the short term (P < .001). They were comparatively pessimistic (ie thinking they were more at risk than others for developing COVID‐19‐related infection or symptoms) when thinking about the next year. Discussion This is one of the first ever studies to report compelling comparative biases in UK adults’ thinking about COVID‐19.


| BACKG ROU N D
Until a vaccine and/or an effective cure for COVID-19 becomes available, battling the current pandemic will critically depend on how well people follow behavioural advice to adhere to lockdown restrictions, adhere to social distancing rules and engage in effective personal hygiene. It has been reliably established that peoples' perceptions about a situation influence their behaviour. 1 Understanding people's thinking about COVID-19 risk in this pandemic is thus critical in understanding and predicting COVID-19related behaviour in future.
One well-established phenomenon in risk perception is comparative optimism. Comparative optimism is the belief that negative events are less likely to happen to oneself than to others. 2,3 The majority of people of all genders and ages show comparative optimism for a wide variety of risks, including many health hazards. 4,5 For example, people believe that they are less likely than others to be involved in a car accident, to experience a divorce, to fall victim to a crime or to lose their job. 2,3 Although comparative optimism is impressively robust, some systematic variation has been observed.
One well-documented finding is that comparative optimism is more pronounced for risks that people deem controllable, such as lifestyle-related health problems. 6,7 In this paper, we explore the occurrence of comparative optimism in relation to COVID-19-related behavioural risks.
Controllability of COVID-19 risk has been an important factor of the UK Government Public Health advice. At the start of the UK lockdown, the Government communication focused on the idea that staying at home would have direct positive impacts on curbing COVID-19 transmission. The slogans 'Stay at Home, Protect the NHS, Save Lives', recently replaced by 'Stay Alert, Control the Virus, Save Lives', had at their heart the idea that this pandemic was controllable by individuals taking personal action.
At the same time, research has shown that greater perceived controllability of an event enhances the likelihood of greater comparative optimism, 6,7 and so we suggest it is likely that recommendations encouraging protective behaviours may be positively associated with enhanced comparative optimism during the lockdown period.
If this is the case, we would expect to see high rates of comparative optimism concerning aspects of COVID- 19  The question whether people show comparative optimism concerning COVID-19 is important because of its potential psychological and behavioural consequences. Comparative optimism may have elicited the anecdotally observed lack of compliance with lockdown guidelines in the UK. 9 Despite overwhelming public support for continued lockdown prior to a safe, gradual loosening of restrictions, 10 25% of inhabitants of some areas admitted breaking lockdown rules. 11 Such failure to comply with lockdown guidelines may have many causes other than comparative optimism. These include, but are not limited to, misunderstanding of lockdown principles, boredom, loneliness, mistrust in policymakers, belief in herd immunity, and/or the desire to alleviate the assumed loneliness or boredom of others. Surveys carried out by King's College London and Ipsos Mori suggested that by week 5 of the UK lockdown (w/c 20th April 2020), 2 in 5 younger people (18-25 years) were finding lockdown restrictions extremely difficult to cope with, or expected to find it so in the next 4 weeks. 12 This is against a backdrop of misunderstanding of guidelines, such as about how often people are allowed to leave the house and what they are allowed to leave the house for. 12 Alongside these factors, it is likely that people who believe COVID-19 is less likely to happen to them than to others may infer that their actual risk is much smaller than that communicated in the media, and thus that strict adherence to lockdown restrictions is unnecessary in their case.
Previous research has shown that comparative optimism contributes to risk-taking. 13 Greater comparative optimism is known to be associated with more risk-increasing behaviours. 14 Conversely, comparative optimism may also positively contribute to people's mental health 15 whereby in reducing anxiety, it might enable the fostering of positive relationships. 16 13 October 2020 after first online publication: The preceding sentence has been amended to 'one of the first ever studies' instead of 'the first ever study '.] were more at risk than others for developing COVID-19-related infection or symptoms) when thinking about the next year.
Discussion: This is one of the first ever studies to report compelling comparative biases in UK adults' thinking about COVID-19.

K E Y W O R D S
comparative optimism, COVID-19, lockdown, risk perceptions, unrealistic optimism may thus have both desirable and undesirable consequences for the behavioural response to the pandemic.
There is ample reason, therefore, to examine the extent to which comparative optimism occurs in risk perceptions concerning COVID- 19. This paper reports UK data from an online international survey testing the following hypotheses: 1. People are comparatively optimistic about COVID-19; people will report being less likely than others to experience negative COVID-19-related outcomes (eg getting infected in the first place) and more likely to experience a positive outcome (ie full recovery if infected).
2. Stronger comparative optimism will occur for those aspects of COVID-19 that people have been encouraged to view as personally controllable (ie the likelihood of getting infected or infecting others) than for those aspects that they view as being less personally controllable (ie how ill one gets once infected).

| Design
We conducted an online cross-sectional international survey across 10 countries at a time, whilst many countries were controlling people's movement through various COVID-19 lockdown restrictions.
On 24th May 2020, there were N = 15 084 recorded responses, from people residing in a variety of countries in Europe (eg UK, Denmark, Norway, Sweden, Greece, France, Poland, Belgium, The Netherlands), the United States and Australia. This paper presents data from UK-based participants only. We report here only on the UK data because the particular way in which lockdown was eased in the UK, focusing on personal responsibility, makes comparative risk perceptions particularly pertinent to the UK context. The comparative optimism questions asked participants to consider 'the average person of your age and gender' and to rate how likely it was that a series of COVID-19-related events would happen to themselves as compared to that average person. Some events could arguably be considered as being within the person's control (eg How likely is that over the past month you have accidentally infected others with COVID-19/ that you will get infected with COVID-19), whilst some were likely to be seen as uncontrollable (eg How likely is that if infected you will need hospitalization/ that you will find yourself in an Intensive Care Unit). Participants answered on 5-point Likert scales ranging from 'Extremely likely'

| Materials
(1) to 'Extremely unlikely' (5). Finally, participants reported on a variety of demographic information, including the existence of underlying health conditions identified by the NHS as making them more vulnerable to COVID-19.

| Participants
This paper reports data from a UK sample of adults (N = 645) who completed the survey during the period 24 April-10 May 2020 8.00 GMT (ie weeks 5-8 of the COVID-19 lockdown).

| Statistical methods and Power
Cronbach's alpha analysis was used to assess the scale for internal consistency. Exploratory principal component analysis (PCA) was used to examine whether the comparative optimism scale consisted of identifiable subscales. For this, we used a Varimax rotation, thus not allowing the factors to correlate, and an Eigenvalue of 1.
No further analyses relating to the PCA (eg Goodness of Fit) were performed.
Using guidance for analysing comparative optimism data 18 we converted participants' responses into a comparative optimism continuous scale from −2 to +2, such that positive scores indicated comparative optimism, zero indicated a perceived likelihood equal to that of another average person of one's age and gender, and negative scores indicated comparative pessimism. We tested the occurrence of comparative optimism/ pessimism on each subscale and then on each individual item through single-sample t tests against a hypothesized population mean of 0.
A repeated measures ANOVA was used to test for differences in comparative optimism across the extracted PCA factors. Gender differences in responses to the three subscales were explored through independent samples t tests. All analyses were repeated with and without participants who had self-identified as being at higher risk for COVID-19 to ensure that their responses did not bias our findings.
Finally, statistical power was assessed post hoc. The required sample size for a single-sample t test to detect a small effect (d = 0.25) was N = 210. The survey was thus sufficiently powered to detect a small effect, using a single-sample t test, with a hypothesized population mean of 0 and a 95% confidence level.

| Procedure
As the data were collected during lockdown, the need to engage in novel yet reliable methods of recruitment defined this process.
Potential respondents were invited to participate through researcher networks on various social media including Facebook, Twitter and WhatsApp groups and through direct e-mail. Contacts within these networks were asked to pass the link to the survey onto their networks, in a snowballing recruitment method deemed appropriate for online research especially with people who would normally experience barriers to participating in research. 19 When potential participants

| Scale reliability
The Cronbach's alpha analysis of our data showed that the scale was reliable overall (α = 0.705).

| Demographic characteristics analysis
The demographic characteristics of the sample appear in Table 1.
In line with data on health information preferences 20 the sample was predominantly female and White. Although we sought data on years of post-primary school formal education, 70% of the sample did not respond with the vast majority reporting in ways to suggest they had found the wording confusing. As a result we have not re-

| Dimensions of comparative optimism
The results of the PCA revealed three clear factors explaining a total of 71.39% of the variance. The underlying rotated factors and the items' respective factor loadings appear in Table 2.
We interpreted these factors as follows.

| Comparative optimism per Subscale
As shown in Figure 1, one-third to almost half of the participants felt that as compared to the average other person of their age and gender, they were somewhat or extremely unlikely to need hospitalization, to find themselves in ICU, or to need a ventilator if they were infected by the virus, whereas considerably fewer felt that they were somewhat or extremely likely to experience those events. In addition, more than half believed they were somewhat or extremely likely to make a full recovery as compared to the average other person of their age and gender. Figure 2 shows the vast majority of participants reporting they were somewhat or extremely unlikely to have accidentally infected others last month, to accidentally infect others next month or to get infected themselves next month, as compared to the average other person of their age and gender. In sharp contrast, and as shown in Figure 3, about half of the participants thought that they were somewhat to extremely likely to get infected in the next year and, when infected, to develop COVIDrelated symptoms as compared to the average other person of their age and gender.

| Extent of comparative optimism
Having converted scores on each item into a continuous scale from −2 to + 2, we tested whether these risk perception scores were statistically different from a hypothesized population mean of zero.
Negative scores indicated that the participant's risk estimate was comparatively pessimistic, positive scores suggested that the estimate was comparatively optimistic, whilst scores not significantly different from 0 suggested that the reported risk perceptions were no different from those of the average other person of the same age and gender. The results appear in Table 3. Table 3 shows significant comparative optimism for Subscale 1 and Subscale 2, as well as for each individual item of these subscales.
Participants thought they were unlikely compared to the average other person to be hospitalized, to be admitted to ICU, to need a ventilator, to infect others and to get infected by others, either recently or in the next month. We found comparative pessimism for Subscale 3 and for two of its three individual items. Participants generally reported being more likely to get infected and develop symptoms in the next year compared to the average other person but felt they were no more likely than anyone else to infect others.
We compared the subscales that yielded comparative optimism through a repeated measures ANOVA with subscale (subscale 1 measuring uncontrollable aspects vs subscale 2 measuring controllable aspects) as a within-subjects variable. Comparative optimism was significantly higher for subscale 2 than for subscale 1 (F (1,642) = 81.37, P < .001), revealing that participants showed stronger comparative optimism for those aspects of COVID-19 that may be deemed as controllable than for those considered controllable.

| Gender effects
Because women were so greatly over-represented in our sample, we wished to establish the generality of our findings across genders.

| Risk factor effects
We repeated all the analyses above excluding those participants who reported underlying conditions that might predispose them to worse outcomes if infected with COVID-19. These analyses replicated the above findings. As a result, we have included the data from people reporting to be at higher risk for COVID-19 complications.

| D ISCUSS I ON
On the basis of these data, we suggest that UK adults who meet the demographic characteristics of our sample display comparative optimism concerning many aspects of COVID-19. Where participants showed comparative optimism its pattern was consistent with earlier findings showing that comparative optimism is stronger for controllable than for uncontrollable events. 6 On the basis of the above, we conclude that UK adults may be comparatively optimistic about the chances of coming to harm due to COVID-19 at the moment or having caused harm themselves previously. Future research is needed on the implications of comparatively optimistic thinking for future compliance with government guidelines on managing COVID-19.

ACK N OWLED G EM ENTS
The author teams wish to thank all our participants across Europe, Australia, Canada and the United States for taking the time to assist with this study whilst in COVID-19 lockdown.