Getting COVID-19: Anticipated negative emotions are worse than experienced negative emotions

Abstract

Objective

When people think about negative events that may occur in the future, they tend to overestimate their emotional reactions, and these “affective forecasts” can influence their present behavior (Wilson and Gilbert, 2003). The present research examined affective forecasting for COVID-19 infection including the associations between emotions and preventive intentions and behavior.

Methods

In two studies, we compared individuals’ anticipated emotions and recalled emotions for COVID-19 infection. Study 1 asked college students (N = 219) and Study 2 asked general adults (N = 401) to either predict their emotions in response to a future COVID-19 infection or to recall their emotions associated with a previous infection.

Results

In both studies, reliable differences in negative emotions emerged. Those who were predicting their feelings associated with a future infection anticipated more negative emotion than those who were recalling their feelings associated with a past infection reported. Greater negative emotion in both studies was significantly associated with being more likely to have been vaccinated as well as higher intentions to get the booster vaccine.

Conclusions

These findings suggest that compared to those who have had a COVID-19 infection, those who have not yet experienced infection anticipate they will experience greater negative emotion, and this may have implications for preventive behaviors. In general, these findings suggest that people may have an impact bias for COVID-19 infection.

1. Introduction

Research has connected individuals' emotions to health behavior decision processes as well as the uptake of health behaviors (for reviews, see Diefenbach et al., 2008; Diefenbach et al., 2008; Ferrer et al., 2016; Williams and Evans, 2014). While many studies have examined the role of current emotions (when relevant to a specific health threat, these are known as “integral emotions”), others have highlighted the importance of anticipated emotions (e.g., Chapman and Coups, 2006). Although less studied, anticipated emotions may be equally important as current emotions to health behavior decisions (Mellers and McGraw, 2001). For example, a growing body of research has shown that emotions like anticipated regret are highly important to enacting future health behaviors like obtaining vaccines (e.g., Brewer et al., 2016; Koch, 2014). In the present research, we examine people's anticipated emotions for contracting a COVID-19 infection, including how they compare to recalled emotions for infection, and how they relate to vaccine behavior.

According to affective forecasting theory, when individuals think about the future, they tend to mispredict their emotions (Wilson and Gilbert, 2003). For some future events, individuals’ affective forecasts may be characterized by an “impact bias,” in which they overestimate the intensity and duration of feelings (Buehler and McFarland, 2001; Gilbert et al., 1998; Kahneman and Thaler, 2006; Wilson and Gilbert, 2003; Wilson et al., 2000). Although most research on affective forecasting has examined non-health outcomes, a growing number of studies have examined health outcomes. Typically, in these studies, healthy people are asked to predict their subjective well-being, including emotions and quality of life, should they experience some health condition. Their estimates are then compared with those who actually have or have had the health condition. In one study, researchers assessed the real-time emotions of healthy patients and patients who had kidney disease (Riis et al., 2005). They found that although the two groups did not differ in their daily emotional experiences, the healthy patients showed an impact bias: When asked what it would be like if they had kidney disease, they predicted fewer positive emotions and more negative emotions than those who actually had the disease experienced. Other studies have highlighted similar findings for different chronic illnesses or health threats like developing pain or gaining weight (e.g., Baron et al., 2003; Loewenstein and Frederick, 1997; Ubel et al., 2005; Walsh and Ayton, 2009; for a review, see Loewenstein and Schkade, 1999). These studies show that compared to people who actually experience these events, those who are asked to predict them overestimate the impact on their well-being.

Although they can be inaccurate, the reason affective forecasts are most important is because they can influence people's present decisions (see Hoerger, 2012 for a review), including health decisions (e.g., Hoerger et al., 2016; Rhodes and Strain, 2008; Ruby et al., 2011). In fact, researchers have argued that affective forecasts may be particularly important in the domain of health where they could influence life and death decisions (Halpern and Arnold, 2008; Rhodes and Strain, 2008). For example, imagine an individual who must choose between two treatments for a medical condition (Rhodes and Strain, 2008). If they anticipate intense regret about potentially experiencing a side effect of one of the treatments, they may choose the alternative treatment that does not have that side effect yet may be the less effective treatment. Anticipating negative emotions may similarly influence serious preventive health decisions like whether to undergo cancer screening. For example, disgust is a particularly important emotion related to colon cancer screening behaviors (Reynolds et al., 2013). If people overestimate their disgust regarding the screening procedure, they may decide to forgo this potentially life-saving behavior (Janz et al., 2007).

In contrast to these examples, in other situations, anticipating negative emotion could lead to proactive health behavior. For example, in a recent investigation, researchers found that individuals who were thinking about future negative outcomes – from experiencing physical injury to getting food poisoning – predicted they would experience more negative emotion than those who experienced these outcomes reported (Dillard et al., 2021). Importantly, anticipating greater negative emotion was connected to higher preventive behavior intentions. In the present research, we examine if people predicting their feelings for a future COVID-19 infection anticipate more negative emotion than those recalling their negative emotion from a past COVID-19 infection. We further examine associations with negative emotions, such as whether they related to vaccination status and plans to get a COVID-19 booster.

2. Present research

COVID-19 was identified as the third leading cause of death in the U.S. in both 2020 and 2021 (https://www.nih.gov/news-events/news-releases/covid-19-was-third-leading-cause-death-united-states-both-2020-2021) and it has killed over 6 and one-half million people globally (https://www.worldometers.info/coronavirus/coronavirus-death-toll/). Although vaccines have reduced mortality and have been widely available (at least in the U.S.), many people remain unvaccinated. Applying psychological theories can help us understand factors that may underlie the perceived threat of COVID-19 and people's protective behaviors. The present research applies affective forecasting theory (Wilson and Gilbert, 2003) to understand the emotions that may characterize COVID-19 and how these may relate to vaccine uptake. In two studies, we examine people's emotional reactions to a COVID-19 infection. Specifically, we asked individuals who never had the virus to predict how they would be emotionally affected (i.e., anticipate their negative emotions) and individuals who previously had the virus to report how they were emotionally affected (i.e., recall their negative emotions). In line with research described above, we hypothesized that individuals who are predicting their feelings for a future COVID-19 infection may anticipate more negative emotion than those recalling their emotion from a past COVID-19 infection. This hypothesis is also consistent with research finding that people are likely to anticipate intense negative emotion for future events that are distressing and highly important (Hoerger et al., 2010). We further hypothesized that greater negative emotion would be associated with greater likelihood of having obtained the COVID-19 vaccine as well as higher intentions to get the booster. Finally, in line with the predictions about negative emotion, we also explored individuals' perceptions of the severity of COVID-19 infection. Conceivably, individuals who are predicting a future COVID-19 infection may anticipate it to be more severe than those recalling an infection.

2.1. Sample size determination and data availability

The data and questionnaires from both studies are included on the Open Science Framework (https://osf.io/bwu3c/?view_only=55aeabb7db20460eb76282a13772e57c). The studies were not preregistered. In Study 1, which included college students, we collected as many participants as possible over the course of the time available to us (N = 219). In Study 2, which included a more general sample of participants from Prolific.co with a balanced design in terms of COVID-19 infection status, we set a goal to collect 200 participants who previously had or had never had COVID-19 given our financial resources (a goal of 400 total participants). The balanced design of Study 2 gave us 80% power to detect an effect size for a difference in means between two groups of approximately d = 0.28 (Faul et al., 2007). The Method sections below contain information on data exclusions, all manipulations, and all measures in the study.

3. Study 1

3.1. Methods

3.1.1. Participants

Participants included American college students (N = 219) from a mid-sized public university in the Midwest. Three participants did not report demographic characteristics and were not included in primary analyses. Of the remaining 216 participants, 76% were female and the average age was 18.6 (SD = 1.1). Most participants were White (82%; 7% of whom were of Hispanic ethnicity) with some African American (5%) and Asian American (4%) participants (2% other races or not reported). Most participants (94%) reported their health as good, very good, or excellent, and most (85%) reported no existing chronic health condition. The majority of participants (n = 194; 89%) reported that they had received the COVID-19 vaccine, and 60 participants (27%) reported that they had previously been diagnosed with COVID-19.

3.1.2. Procedure

Data collection occurred over a four-week period from November–December 2021. At this time, the university had COVID-19 restrictions in place including mask requirements in classrooms and other campus spaces as well as vaccine mandates. Students in introductory psychology courses were invited to enroll in a study on “Beliefs and expectations about COVID-19,” which they could link to a website maintained by the university. After self-selecting into the study, participants completed a consent form and then responded to questions including whether they had ever been diagnosed with COVID-19. Participants who answered “Yes” then completed questions about the emotions they experienced and participants who answered “No” completed questions about their anticipated emotions. All participants also answered questions about the perceived severity of COVID-19. They then answered questions about vaccines including whether they had been vaccinated and the degree to which they were interested in getting a booster vaccine when it was available or recommended. Participants also completed other questions not directly relevant to the current project (please see supplementary material for more information). In exchange for participating, students received credit in their psychology course.

3.1.3. Measures

Anticipated and recalled emotions. Four emotions – regret, guilt, anger, and fear – were assessed on 7- point scales from “Not at all” to a “A great deal” with higher numbers representing more intense emotion. Those who reported they had previously been diagnosed with COVID-19 were asked, “When you had COVID-19, to what extent did you feel regretful?” Those who reported they had never been diagnosed with COVID-19 were asked, “If you were to get COVID-19, to what extent would you feel regretful?” Similar questions asked about feeling guilty, angry, and fearful. We analyzed the ratings separately as well as created an overall score by averaging across the four emotions. Reliability (α) for anticipated emotions was 0.85, and for recalled emotions, it was 0.86.

Vaccine behavior. Participants answered one question about their vaccine behavior: Have you been vaccinated for COVID-19? They could respond “Yes” (coded as 1) or “No” (coded as 0).

Booster interest/intention. People who reported being vaccinated also completed two questions about their interest in and plans to obtain a booster vaccine: “If a booster vaccine that protects you from COVID-19 was available for your age group, would you want to get it?” and “If a booster vaccine that protects you from COVID-19 was recommended for your age group, would you want to get it?” Both questions were on 7-point scales from “Not at all” to “Very much”. The two items were combined into a composite (α = 0.97).

Perceived severity. Two questions assessed the perceived severity of COVID-19. Those who reported they had previously been diagnosed with COVID-19 were asked, “How serious would you say it was when you had COVID-19?” Those who reported they had never been diagnosed with COVID-19 were asked, “How serious do you think it would be if you were to get COVID-19?” Both were on a 7-point scales from “No harm at all to “Extremely devastating”. A second question asked about the severity of symptoms of COVID-19. Those who previously had the virus were asked, “When you had COVID-19, were your symptoms generally: mild, moderate, or severe?“, and those who had not had the virus were asked, “If you were to get COVID-19, do you think your symptoms would be generally: mild, moderate, or severe?“. Response options were mild (1), moderate (2), or severe (3).

4. Results

4.1. Descriptives

We first explored differences between those who reported a previous infection (n = 58) and those who did not (n = 158). One-way Analyses of Variance (ANOVAs) revealed that the two groups did not differ with respect to age, race, or gender. The two groups also did not differ in their self-reported general health, presence of existing chronic illness, or other variables that were tangential to hypotheses but were assessed in the survey. (Please see supplemental material for additional measures and analyses.)

4.2. Primary analyses

The means and SDs for the individual emotions are presented in Table 1 . As shown in the Table, anticipated emotions were consistently higher than recalled emotions. To test if these differences were significant, we used one-way Analyses of Covariance (ANCOVAs), controlling for age, race, and gender. With the exception of anger, all of the differences were significant. An overall negative emotion score was also significant, showing that anticipated emotions were higher than experienced emotions, Ms = 3.50 (SD = 1.59) vs. 2.72 (SD = 1.32), respectively, F(1,211) = 9.50, p = .004, effect-size r = 0.26.

Table 1.

Study 1 Means and SDs for anticipated and experienced feelings.

	Anticipated feelings (n = 158)		Recalled feelings (n = 58)		Significance
	M	SD	M	SD	F	p	Effect-size r
Regretful	3.33	1.91	2.47	1.49	8.54	.004	0.24
Guilty	3.25	2.06	2.52	1.88	4.31	.039	0.18
Angry	4.03	1.90	3.48	1.66	3.05	.082	0.15
Fearful	3.39	1.76	2.41	1.40	12.88	<.001	0.29

Note. All feelings were rated on 7-point scales from “Not at all” to “A great deal”. All analyses control for age, race, and gender.

Regression analyses were next conducted to examine vaccine behavior as well as interest/intention to get a booster vaccine as a function of negative emotion (as an overall score across the entire sample and then as separate models). Logistic regression was used to examine vaccine behavior (as this was a dichotomous response) and linear regression was used to examine booster interest/intention. Analyses controlled for age, race, and gender in Step 1 of the regressions. Note that the analysis for booster interest/intention only included those participants who had reported getting the vaccine (n = 194). For vaccine behavior, the logistic regression was significant, B = 0.75, SE = 0.22, Wald = 11.86, p = .001. The estimated odds ratio showed an increase of more than double (i.e., twice as likely), [Exp(B) = 2.106, 95% CI (1.38, 3.22)] for likelihood of having obtained the vaccine for every one unit increase of negative emotion. When run as separate models, analyses revealed that anticipated emotions were significantly associated with vaccine behavior, B = 0.92, SE = 0.31, Wald = 9.16, p = .002, but recalled emotions were not, B = 0.25, SE = 0.33, Wald = 0.57, p = .449. For intentions to get the booster, the regression analysis was significant, B = 0.52, SE = 0.09, 95% CI (0.34, 0.70), b = 0.39, t(187) = 5.76, p < .001, and similarly revealed that higher negative emotion was associated with greater interest/intention to get the booster when it was available and recommended. When analyzed separately, significant associations emerged between anticipated emotions and intentions, B = 0.53, SE = 0.10, 95% CI (0.33, 0.74), b = 0.40, t(139) = 5.16, p < .001, as well as recalled emotions and intentions, B = 0.53, SE = 0.22, 95% CI (0.08, 0.99), b = 0.35, t(43) = 2.39, p = .022.

Finally, we also examined perceived severity of COVID-19. One-way ANCOVAs were conducted to examine differences in seriousness and severity of symptoms. Although anticipated seriousness was higher than experienced seriousness, Ms = 3.03 (SD = 1.37) vs. 2.69 (SD = 1.16), respectively, the difference was not significant, F(1,211) = 2.67, p = .104. There was also no significant difference for anticipated or experienced severity of symptoms, Ms = 1.46 (SD = 0.55) vs. 1.50 (SD = 0.50), respectively, F < 1.

5. Discussion

The findings of Study 1 showed that people who think about contracting COVID-19 in the future anticipate higher negative emotion than those who recall their negative emotion associated with an actual infection. Findings revealed significant differences for three negative emotions – regret, guilt, and fear. Negative emotion was also significantly related to behavior as those with greater negative emotion were more likely to have been vaccinated and reported higher intentions to get the booster vaccine. This study may suggest that when people think about getting COVID-19 in the future, they may overestimate their negative emotional reaction, and this anticipated emotion may be equally as important as experienced emotion in relating to behavior. Our findings are in line with the affective forecasting literature which says that anticipated emotions may be higher than experienced or recalled emotions, and anticipated emotions can influence present decision-making (e.g., Dillard et al., 2021; Hoerger et al., 2016; Ruby et al., 2011; Rhodes and Strain, 2008). To our knowledge, few if any studies have applied affective forecasting ideas to the COVID-19 threat.

Although Study 1 provides evidence that anticipated negative emotions for a COVID-19 infection are higher than recalled emotions, we chose to conduct a second study that would address the inherent limitations of a college student sample and the small sample size of those who reported a previous infection. Ideally, for the most effective comparison, we would have equal groups of people who have or have not been infected. Also, collecting data from more general adults would be prudent as a booster vaccine had been recommended to this group.

6. Study 2

The primary goal of Study 2 was to replicate the findings of Study 1 with a more general adult sample. It was also important to address the issue of the smaller and unequal sample size concerning COVID-19 status. Thus, in Study 2, we sought to recruit an equal number of participants who previously had been diagnosed, or not, with COVID-19. We again examined if there were significant differences in negative emotions associated with experiencing COVID-19 infection and if negative emotions were related to vaccine behavior and intentions. The hypotheses for Study 2 were the same as Study 1.

6.1. Participants

Participants included 401 American adults ranging in age from 18 to 74 (M = 32.9, SD = 12.0) of which 69% were female. The majority of participants were White (82%; 5% of whom were of Hispanic ethnicity) with some African American (4%) and Asian American (6%) participants (3% other races or not reported). Most participants (83%) reported their health as good, very good, or excellent, and a majority (72%) reported they had no chronic health conditions. Most participants (n = 332; 83%) reported that they had received the COVID-19 vaccine, and about one-half (n = 160; 48%) of these individuals had also received a booster vaccine.

6.2. Procedure

Data collection occurred on December 21st, 2021. Adults were recruited through Prolific.co, an international survey company that specializes in recruiting participants for behavioral research. Prolific.co states that they have a pool of 150,000 participants worldwide who have completed demographic and screening questions. Any participant who matches study criteria sees the study on their dashboard and can opt-in. The eligibility criteria for the current study included having U.S. nationality, speaking English as a first language, and currently residing in the U.S. We collected two different samples with the goal of collecting 200 participants in each sample: one that reported being diagnosed with COVID-19 and one that reported not being diagnosed with COVID-19 (according to Prolific.co screening criteria). The average compensation rate was $12.30 per hour with the study taking most people between 5 and 7 min. After they read a consent form, participants answered questions about COVID-19 including whether they had previously been diagnosed with the virus, their anticipated or recalled emotions, vaccine behavior, future booster intentions, and other questions.

6.3. Measures

With the exception of booster vaccine intentions (see below), the measures in Study 2 were identical to Study 1. Reliability (α) for anticipated emotions was 0.87, and for recalled emotions, it was 0.82.

Booster vaccine intentions. Given that Study 2 included a more general adult sample and boosters had now been recommended for this age group, the booster vaccine intentions measure differed from Study 1. Participants were asked two questions: “How likely are you to get the booster vaccine in the next three months?” and “Are you planning to get the booster in the next few months?” Questions were on 7-point scales from “Not at all likely” to “Extremely likely” and “Definitely not” to “Definitely yes”, respectively. The two items were combined into a composite (α = 0.98).

7. Results

7.1. Descriptives

We first tested differences between those who reported a previous infection (n = 198) and those who did not (n = 203). One-way Analyses of Variance (ANOVAs) revealed that the two groups did not differ with respect to age or gender. However, there was a racial difference such that there were more individuals who self-reported White (88% vs. 75%) and less who self-reported Asian (11% vs. 1%) in the previously infected group compared to the never infected group. The two groups did not differ in self-report their perceived general health, presence of an existing chronic illness, or other variables assessed in the survey (for more information, please see supplementary material).

7.2. Primary analyses

The means and SDs for individual emotions are presented in Table 2 . Anticipated emotions were again higher than recalled experienced emotions, and ANCOVAs revealed that all emotion differences were significant. The overall negative emotion score was similarly significant, Ms = 3.92 (SD = 1.65) vs. 3.10 (SD = 1.58), F(1,395) = 25.59, p < .001, effect-size r = 0.25.

Table 2.

Study 2 Means and SDs for anticipated and experienced feelings.

	Anticipated feelings (n = 203)		Recalled feelings (n = 198)		Significance
	M	SD	M	SD	F	p	Effect-size r
Regretful	3.87	1.93	2.77	1.91	30.91	<.001	0.28
Guilty	3.39	2.07	2.64	1.86	14.75	<.001	0.19
Angry	4.22	1.91	3.46	2.11	14.59	<.001	0.19
Fearful	4.22	1.88	3.58	2.02	10.87	.001	0.16

Note. All feelings were rated on 7-point scales from “Not at all” to “A great deal”. All analyses control for age, race, and gender. One participant from the recalled feelings group did not report guilt and was not included in that analysis or in the overall negative emotion analyses.

Regression analyses were next conducted to examine vaccine behavior as well as intentions to get a booster vaccine as a function of overall negative emotion. Logistic regression was used to examine vaccine behavior (as this was a dichotomous response) and linear regression was used to examine booster intentions (note that the analysis for booster intentions included only those who had received the vaccine but not the booster; n = 172). All analyses controlled for demographic characteristics, and any missing values (<1% of cases) were excluded listwise. For vaccine behavior, the logistic regression was significant, B = 0.37, SE = 0.09, Wald = 15.55, p < .001. The estimated odds ratio showed an increase of 45%, [Exp(B) = 1.449, 95% CI (1.21, 1.74)] for likelihood of having obtained the vaccine for every one unit increase of negative emotion. When analyzed separately, anticipated emotions were significantly related to having obtained the vaccine, B = 0.49, SE = 0.15, Wald = 11.00, p = .001, but recalled emotions were not significant, B = 0.24, SE = 0.13, Wald = 3.32, p = .068. For intentions to get the booster vaccine, the overall regression analysis was significant, B = 0.28, SE = 0.08, 95% CI (0.11, 0.44), b = 0.25, t(166) = 3.38, p = .001, and similarly revealed that higher negative emotion was associated with higher intentions to get the booster in the next few months. When run as separate models, anticipated emotions were significantly associated with intentions, B = 0.33, SE = 0.12, 95% CI (0.10, 0.56), b = 0.32, t(69) = 2.82, p = .006, but recalled emotions were not, B = 0.22, SE = 0.13, 95% CI (0.03, 0.48), b = 0.18, t(92) = 1.74, p = .085.

Next, we examined perceived severity of COVID-19. One-way ANCOVAs were conducted to examine differences in anticipated vs. experienced seriousness and severity of symptoms. For seriousness of contracting the virus, anticipated seriousness was significantly higher than experienced seriousness, Ms = 3.70 (SD = 1.47) vs. 3.05 (SD = 1.42), respectively, F(1,396) = 20.79, p < .001, r = 0.22. However, there was no significant difference for anticipated vs. experienced severity of symptoms, Ms = 1.51 (SD = 0.60) vs. 1.48 (SD = 0.59), respectively, F < 1.

8. Discussion

In a general adult sample, we again found evidence of higher anticipated than recalled negative emotions associated with COVID-19 infection. Compared to those who were reporting their emotions associated with a previous infection, those who were anticipating emotions for possible future infection predicted significantly higher regret, guilt, anger, and fear. Greater negative emotion was advantageous in that it was positively associated with having obtained the vaccine and intending to get a booster dose. In this study with general adults, findings also revealed that those anticipating a future infection rated COVID-19 as more serious than those who were recalling a past infection. However, the two groups did not differ in their ratings of symptoms as mild, moderate, or severe.

In replicating the findings of Study 1 with a general adult sample and with balanced groups (i.e., equal numbers of those who had COVID and those who had not), Study 2 provides additional support for the idea that anticipated negative emotions are more intense than recalled actual emotions. Study 2 also showed that negative emotion was again associated with vaccine likelihood and booster intentions but notably, when analyzed as separate models, anticipated but not recalled emotions were significantly related to these outcomes. It is also important to note that these associations between negative emotions and intentions were among U.S. adults of which booster doses, at the time of data collection, were currently available and recommended. The results of both studies may suggest that compared to reality, the prospect of getting COVID-19 may be characterized by more intense negative affect, which is related to preventive behaviors.

9. General discussion

In two studies with college students and general adults, we compared affective forecasts to affective experiences of a COVID-19 infection. In both studies, when individuals thought about the prospect of contracting COVID-19, they anticipated more regret, guilt, anger, and fear than individuals who had the virus recalled experiencing. Higher negative emotion was meaningful in that it was related to greater likelihood of having been vaccinated as well as higher intentions to get the booster.

Although similar differences in anticipated versus recalled negative emotions were observed in both the college students and general adults, the negative emotions were overall higher in the latter group. In the sample of general adults, perceived severity of COVID-19 also significantly differed among those anticipating versus recalling infection, a finding which was not observed in the college students. Together, these findings may suggest that relative to college students, the general adults felt more threatened by COVID-19. On one hand, this notion of greater perceived threat among an older sample is reasonable given that age is a risk factor for more severe disease. On the other hand, the anticipation of greater negative emotion among the older sample does not fit with recent studies finding that older individuals, compared to younger, are faring better emotionally during the pandemic (including some of the same emotions we tested; Carstensen et al., 2020; Knepple Carney, Graf, Hudson and Wilson, 2021) or that older adults are more optimistic about COVID-19 (Bruine de Bruin, 2021). However, this distinction may relate to emotions about how one would fare with COVID-19 infection (as measured in our research) versus how one is coping emotionally with the pandemic. In fact, although several studies have examined people's emotions during the pandemic, none that we know of have examined people's anticipated or recalled emotional reactions to contracting COVID-19.

Our findings are in line with affective forecasting theory, and the specific error known as the impact bias. The impact bias occurs when people overestimate the intensity and duration of their future emotions (Gilbert and Wilson, 2007; for a review, see Wilson and Gilbert, 2003). Early research on the impact bias showed it for outcomes such as breaking up with a romantic partner or failing to get a job promotion, but it has since been found for many diverse events and outcomes (Dunn et al., 2003; Finkenauer et al., 2007; Gilbert et al., 1998; Hoerger, 2012; Hoerger et al., 2009; Kermer et al., 2006; Sieff et al., 1999; Van Dijk, 2009). Researchers have argued that the impact bias likely underpins many health decisions, but relatively few studies have tested the bias and its behavioral implications (Halpern and Arnold, 2008; Rhodes and Strain, 2008). Given our findings that anticipated emotions were more intense than recalled experienced emotions, our data are suggestive of an impact bias for COVID-19 infection. These data are among the first to apply affective forecasting ideas to this unusually novel and severe virus.

Although our research is an important first step in highlighting the potential of an impact bias for COVID-19, our studies do not provide definitive evidence. This is because we assessed recalled emotions which may differ from actual experienced emotions. For example, it could be that participants who were recalling their emotions from a past infection experienced just as much negative emotion as those who were anticipating an infection, but they remember the emotions as less intense. This idea would be supported by research suggesting that recalled emotions are susceptible to various cognitive biases and processes (for a review see Levine and Safer, 2002). For example, one's expectations about how they should have felt, one's coping or adaptation since the event, and even personality factors may influence recalled emotions (Hoerger et al., 2009; Ottenstein and Lischetzke, 2020; Wilson et al., 2003). Arguably, some of these factors could influence one's anticipated emotions too. However, a future study that uses a within-subjects, longitudinal design, assessing the same individuals before, during and after they experience COVID-19, can provide definitive evidence of an impact bias (see more discussion of this idea in the Limitations section).

One question raised by our findings is, would it benefit people to learn that individuals who contract a virus like COVID-19 may experience less negative emotion than others predict? On one hand, reducing negative emotion in those who have never experienced infection could have the undesired effect of discouraging preventive behavior like getting vaccinated. Indeed, our data would support this notion. On the other hand, many people have experienced high distress due to the pandemic (Belen, 2022; Shafran et al., 2013). While emotions associated with infection may play only a small role in this distress, learning that these emotions may be overestimated (and that people may do better than they anticipate) could be helpful information. Related to this, one strategy to reduce negative emotions surrounding the COVID-19 pandemic is to encourage mindfulness (Dillard and Meier, 2021; Emanuel et al., 2010). Mindfulness is about focusing one's attention on the ongoing, present moment (Brown and Ryan, 2003). People who practice mindfulness may be less inclined to think about future outcomes, or anticipate strong negative emotions associated with these outcomes.

The question above relates to a broad dilemma, faced by researchers in psychology, medicine and other fields, about using emotions to promote health behaviors. That is, to what extent is it acceptable to use, or to increase, people's existing negative emotions to motivate health behaviors? For example, to encourage women to get mammograms, is it appropriate to use interventions to increase their fear (or other negative emotions), or to not correct their existing strong negative emotions about breast cancer? Although some women may hold stronger negative emotions than warranted (e.g., they may be of lower-than-average risk), correcting them could have the unfortunate consequence of reducing their likelihood of getting screened. The answer to this dilemma may well depend on factors such as context (e.g., whether there is a ‘right’ preventive action that is appropriate for most people) or emotion threshold (e.g., when is a negative emotion too much, leading to additional distress, and when is it just enough to motivate behavior). In general, more research should be devoted to determining the conditions relating to this dilemma and affective forecasting is a ripe context for investigating them.

In both studies, we found that individuals who anticipated or recalled greater negative emotion associated with COVID-19 infection were more likely to have been vaccinated and they also reported higher intentions to get the booster. Although our data were correlational, they fit with the broad literature that show emotions, including anticipated ones, can be a powerful influence on heath behaviors (e.g., see Williams and Evans, 2014 for a review), including vaccine behavior (Brewer et al., 2016; Chapman and Coups, 2006; Wilson et al., 2003). Our findings also fit with recent research finding that emotions like fear and regret are positively associated with COVID-19 vaccination and other preventive behaviors (Coifman et al., 2021; Reuken et al., 2020; Wolff, 2021). More research is needed on associations between different types of emotions and health behaviors. For example, are experienced emotions as important as recalled or anticipated emotions in motivating health behavior? And does accuracy of recalled or anticipated emotions matter in this context? Testing associations between these emotions and health behavior may be difficult as the emotions likely share overlap especially for health threats people are familiar with and have prior experience.

It is important to consider the timing of this research which occurred during Fall 2021. In a recent large-scale longitudinal investigation, researchers examined both American and Chinese adults’ emotions and behavior over the course of the pandemic (Li et al., 2021). They found that negative emotions like fear, anxiety, and worry were heightened in the beginning of the pandemic, but later, during phases of ongoing risk, returned to baseline levels. Their research also showed that while emotions were predictive of preventive behaviors like wearing a mask early in the pandemic, they were not predictive later. In the present research, we observed meaningful differences between anticipated and recalled emotions associated with COVID-19 infection, and both were associated with vaccine behavior. Thus, although emotional reactions have apparently lessened, our findings may speak to the power of affective forecasting and its implications for present behavior.

10. Limitations

This research is not without limitations. Most importantly, both studies used a between-subjects design in which participants were not randomly assigned yet were asked different questions depending on their experience with COVID-19 infection. Although we believe their negative emotion differences related to affective forecasting errors, the differences may have been due to other factors. For example, people who have contracted COVID-19 and people who have not may differ in various ways. Notably, we did not find differences for demographics like age or gender, or various psychosocial variables that were measured in the surveys (see supplementary material for details). Given that an experimental design would be impossible as one cannot randomly assign people to have a COVID-19 infection or not, future studies might incorporate additional baseline measures (e.g., COVID exposure, self-protective behaviors) when assessing these groups. A second related limitation is that although our method of comparing anticipated to recalled emotions is an approach that has been used to test affective forecasting errors (e.g., Dillard et al., 2021; Gilbert et al., 1998; Sieff et al., 1999), the preferred method is to use a within-subjects, longitudinal design (e.g., Smith et al., 2008; Wilson et al., 2003; Wilson et al., 2000). For example, people would be measured before and after a COVID-19 infection occurs, and their anticipated and experienced emotions can be directly compared. Of course, this design presents logistical challenges such as the difficulty in assessing people as they are experiencing an infection or having to follow people until an infection occurs (not knowing if it will occur). Following people over time may also allow researchers to examine prospective, actual behavior as opposed to the present studies’ approach which examined retroactive vaccine behavior and booster intentions. Although intentions may be a reliable predictor of behavior (Webb and Sheeran, 2006), finding associations between negative emotion and actual behavior would provide more direct support for the notion that the impact bias has behavioral implications. This may be particularly relevant if COVID vaccines become a yearly recommendation.

Finally, another limitation relates to the biases inherent in recalled emotions. First, individuals who were recalling their infection could have experienced it days, weeks, or even months before being in the study. Length of time since an outcome occurred can bias one's memory for the emotions they experienced during the outcome – in the direction of over or underestimating emotions (Wilson et al., 2003). However, others have found that people are relatively accurate in recalling past emotional experiences, especially in the short-term (Hoerger, 2012). At the time of our study, COVID-19 diagnosis was a new, recent phenomenon, having been around for a little over one year, and all participants' infections would have fallen in that same time frame. Nonetheless to resolve this issue, future studies might assess another group of individuals – those who are currently experiencing COVID-19 infection. However, as mentioned above, this assessment presents logistical challenges.

11. Conclusions

People frequently try to predict their future emotions, and this affective forecasting has consequences for behavior. In two studies, we found that anticipated negative emotions for people thinking about contracting COVID-19 were worse than recalled emotions for people who actually experienced COVID-19. The present findings support affective forecasting theory and are suggestive of an impact bias for COVID-19 infection. Furthermore, the emotions surrounding infection appear important for preventive behavior.

Compliance with ethical standards

The authors have no potential conflicts of interest to disclose. The research complies with APA ethical standards and was approved by the Institutional Review Board. All participants completed an informed consent prior to participating.

The data and questionnaires from both studies are included at an Open Science Framework website (https://osf.io/bwu3c/?view_only=55aeabb7db20460eb76282a13772e57c).

Credit author statement

Both authors (Dillard and Meier) contributed to all aspects of this manuscript including Conceptualization, Methodology, Resources, Project administration, Data curation, Formal analysis, and Writing – original draft, review, editing, and revision.

Handling Editor: Blair T. Johnson

Appendix A. Supplementary data

The following is the Supplementary data to this article.

Data availability

The data are available online (address is in Methods section).