Associations between Affect Variability, Mean Affect, and Mental Health among College Students during the COVID-19 Pandemic

Abstract

Objective:

The purpose of the study was to test whether associations between affect variability and mental health (i.e., anxiety symptoms, depressive symptoms, flourishing) differ by mean levels of affect during the COVID-19 pandemic.

Methods:

College students (N=1883; M_age=19.81, SD=1.33) completed a survey and 21 daily reports of affect (M=19.41 surveys, SD=4.19). We calculated mean affect and affect variability (i.e., standard deviation) from daily reports. Regression models then tested associations between positive and negative affect variability and mental health.

Results:

Participants with higher positive affect variability had higher anxiety symptoms, depressive symptoms, and surprisingly higher flourishing. Higher negative affect variability was associated with poor mental health for individuals with low mean negative affect, and was associated with better mental health for individuals with high mean negative affect.

Conclusion:

Affect variability may simultaneously tax mental health for certain individuals and enable others to appreciate daily experiences and have better mental health.

Although persistently low positive affect and high negative affect are related to poor mental health,¹ affect variability is another important dimension of affective experiences in daily life with implications for mental health. Affect variability can result from a combination of high emotional sensitivity and deficits in emotion regulation, both of which can be involved in psychopathology broadly.² An individual whose mood is more sensitive to external factors, such as daily experiences or aspects of the environment, or who struggles to manage psychological responses may show larger fluctuations in affect and consequently have poor mental health, irrespective of having overall better mood (i.e., high positive affect, low negative affect). Although higher affect variability is related to elevated risk for poorer mental health and psychological well-being,^3,4 Fragile Desirable Affect Theory proposes that affect variability differentially relates to mental health by mean affect,^5,6 which has been supported in older adults⁷ and adolescents and young adults.⁸ Large shifts in affect may be related to poorer mental health for individuals with generally high affective well-being and be simultaneously beneficial for individuals with poor well-being. These disparate effects have yet to be examined during the COVID-19 pandemic, a time of unpredictability when college students were highly vulnerable to psychopathology.⁹ Further, mental health is not the absence of psychopathology,¹⁰ and studies have not characterized whether affect variability differentially relates to aspects of positive mental health including flourishing. Therefore, the present study examined how and for whom affect variability relates to mental health by testing whether associations between college students’ affect variability and mental health (i.e., anxiety symptoms, depressive symptoms, and flourishing) differed by mean affect during the COVID-19 pandemic.

Affect Variability and Mental Health

Affect variability reflects fluctuations in affect over a timescale (e.g., across days in a week) and is typically measured as the individual standard deviation across repeated measures of affect.^11,12 Greater affect variability could suggest that individuals are having varied experiences that they are responding to throughout the day, which could be both burdensome and fulfilling. However, some degree of variability is needed for individuals to flexibly respond to situational demands, in line with evidence that physiological resources that promote social engagement are related to flexible affective responses to stimuli rather than affective stability.^13,14 Individuals with higher affect variability may also need to modulate cognitive processes to maintain stable affect.^12,15 Coping with more varied affective states can interfere with subsequent daily experiences (e.g., academic performance) or promote further stress proliferation and thereby worsen mental health.¹⁶

Variability can differ by affective valence, or the degree to which affective states are aversive (negative valence) versus pleasurable (positive valence).¹³ Both positive and negative affect variability should be considered because positive and negative affect can serve distinct functional purposes.¹⁴ Higher negative affect variability predicts poor outcomes including future negative affect, anxiety disorder diagnosis, depression risk, and low life satisfaction.^3,19–21 Variability in positive affect has received relatively less attention but may also confer health risk. Individuals with high positive affect variability experience both relatively intense increases in positive affect and corresponding intense declines in positive affect over time. Positive affect can support psychobiological resources for strengthening social relationships,²² such that these intense declines in positive affect may impact daily experiences and contribute to poor mental health. Similar to negative affect variability, higher positive affect variability has been related to lower happiness and well-being across studies, although associations were weaker for positive affect variability than negative affect variability.^3,23

Studies of affect variability have been limited by examining psychopathology rather than positive mental health. Although higher positive and negative affect variability have been related to poorer mental health, another possibility is that variability may also promote positive psychological states such as flourishing (i.e., the presence of mental health, encompassing fulfillment and social well-being) by allowing affective flexibility.^24,25 Higher variability in self-reported positive and negative affect is greater for individuals with high self-esteem,²⁶ although other studies have observed the opposite association.^27,28 It is possible that affect variability may be simultaneously related to both positive mental health and psychopathology, and that affect variability may only be beneficial for certain individuals.

Implications of Affect Variability May Depend on Affective Levels

Fragile Desirable Affect Theory posits that associations between affect variability and mental health may differ by mean levels of affect,^5,6,29 another aspect of emotion that consistently predicts mental health.³⁰ For individuals with low negative affect, high variability may suggest days of high negative affect; negative experiences may occur infrequently but result in a spike in negative affect for these individuals. Individuals with high positive affect accompanied with high variability may similarly experience higher sensitivity to environmental or external factors. Therefore, affect variability may be a risk factor for affective disorders for individuals with high affective well-being (i.e., whose emotions are fragile or easily disrupted by daily events). Prior research has found that higher positive affect variability is related to poorer immune function (i.e., higher systemic inflammation, lower antiviral response to vaccination) only among individuals with higher mean positive affect.^31,32 However, studies of affect variability and mental health have rarely accounted for how associations may differ by mean affect.

Research is needed regarding whether higher variability is also related to poorer mental health for individuals with poor affective well-being. Having low variability in affect—also termed rigidity—is a phenotype of depression, especially in conjunction with persistent negative affect.²⁴ Someone with persistent high negative affect or low positive affect may benefit from experiencing a reprieve from these states because poor well-being is a potent risk factor for psychopathology.¹ Indeed, emerging findings indicate that greater negative affect variability is related to worse health outcomes including greater depressive symptoms and depression risk for individuals with low negative affect, and to lower risk only for individuals with very high (over 99^th percentile) mean negative affect.^7,8,33 Results have been less consistent for positive affect variability. One study found that most older adults with greater positive affect variability had greater odds of seeing a mental health professional and of prospectively having depression if they had average to high levels of mean positive affect, but not if they had low levels of positive affect.⁷ In turn, a mega-analysis of adolescents and young adults found no evidence that associations between affect variability and psychopathology differed by mean positive affect.⁸ Those studies only assessed psychopathology (e.g., depression), but positive psychological states (e.g., flourishing, life satisfaction) are not merely the absence of psychopathology and can influence health uniquely from psychopathology.^10,34 Given the inconsistent results and omission of positive mental health as an outcome, further research with large samples is needed to identify how affect variability is related to both psychopathology and positive mental health.

Affect and Affect Variability during the COVID-19 Pandemic

The COVID-19 pandemic provides a unique opportunity to disentangle how mean affect and affect variability relate to mental health. The unpredictable circumstances and hardships incurred by the pandemic resulted in an overall decline in positive affect and an increase in negative affect for most individuals, although affect variability was generally stable between before and during the pandemic.^35–37 Affect variability increased for certain individuals, particularly those who reported greater COVID-19-related stress.³⁸ These findings highlight the uniqueness of mean affect and affect variability during this period of unpredictability, whereas prior studies have been limited by strong relations between variability and mean affect (correlations of .57 and .75 for negative affect, −.36 and −.46 for positive affect).^7,8

The pandemic increased the prevalence of anxiety and depression. Rates were particularly high for college students, who faced academic uncertainty and an unprecedented transition to remote instruction away from peers.^9,39 Identifying whether affect variability is related to poorer versus better mental health for individuals with poor well-being will promote our understanding of the potential correlates and daily experiences of individuals at-risk for psychopathology during this period. Prior studies found that low negative affect variability was related to mental health risk for individuals with very high mean negative affect,⁷ and such associations might be consequential during the pandemic. The pandemic also disproportionately affected the hardships and health of marginalized groups (e.g., racially minoritized, first-generation, LGBTQ+ students).^40,41 Given the prevalence of psychopathology in college students of varied backgrounds, research in large, heterogeneous samples is needed to identify how affect variability, in conjunction with mean affect, may relate to mental health during the pandemic.

Present Study

The present study tested associations between variability in positive affect and negative affect and mental health—with respect to both psychopathology (i.e., anxiety and depressive symptoms) and psychological flourishing—and whether associations differed by mean affect among college students during the COVID-19 pandemic. This study expanded upon prior research by testing flourishing as a positive outcome, and testing associations during a period when college students experienced an unpredictable transition and were at high risk for psychopathology. Affect variability was calculated as the individual standard deviation in affect ratings across 21 days. Positive and negative affect variability were examined separately because positive and negative affect are related but thought to serve distinct functions.¹⁸ Individuals with higher positive and negative affect variability were hypothesized to report greater anxiety and depressive symptoms, as well as lower psychological flourishing, consistent with prior research.^20,23 Building from Fragile Desirable Affect Theory,^5,6,29 associations between higher affect variability and poorer mental health were hypothesized to be strongest for individuals with high mean positive affect and low mean negative affect in line with prior findings of mental and physiological health.^31,32 Because persistently poor affective well-being is also a risk factor for psychopathology,¹ higher positive and negative affect variability were also hypothesized to relate to better mental health for individuals with low mean positive affect and high mean negative affect.

Method

Participants

At a large public university in the Northeastern United States in 2021, 2068 college students completed an electronic psychosocial survey and up to 21 daily surveys. Full-time students at the university between ages 18–24 were eligible for the study. Participants who provided demographic information in the psychosocial survey and completed at least two daily surveys were included in the analytic sample (n = 1883). An identical pattern of results was observed in sensitivity analyses that limited the analytic sample to participants who completed at least one week (n = 1656) and at least two weeks of daily surveys (n = 1454).

Most participants identified as female (67.66%; 30.75% male; 1.58% another gender identity). A minority of participants identified as a first-generation student (18.32%) and as LGBTQ+ (18.32%). The majority of participants identified as white (76.58%), with smaller percentages identifying as Asian (14.13%), Black (2.97%), Hispanic (2.18%), or multiracial (4.14%). A minority of participants reported a physical disability (1.92%) or a chronic condition (e.g., asthma, diabetes; 13.22%). Despite data being collected during the COVID-19 pandemic, participants rarely reported being in close proximity to someone who tested positive for COVID-19 (N = 455, 1.37% of daily observations) or showing COVID-19 symptoms (N = 625, 1.88%).

Procedure

Participants provided informed consent and completed a 15-minute electronic survey regarding demographic information, mental health, and well-being. They provided their phone number and received daily Qualtrics surveys for 21 days via text message and email at 9 am each day, with a reminder one hour later. Daily surveys are ecologically valid and designed to capture experiences that are representative of daily life beyond the designated period.⁴² Daily surveys initiated before midnight were considered valid, and most participants completed all 21 surveys (80.14%; M = 19.41 surveys, SD = 4.19). Participant information was stored in REDCap. Participants received $15 for completing the psychosocial survey and $2 per daily survey, plus $5 high-completion bonuses for completing at least six daily surveys during the first week and 12 daily surveys during the second and third weeks combined, for a total possible $67. For each bonus, participants also had a 1/100 chance of winning a gift card ($50 for first bonus, $100 for second bonus). Data were collected February-December 2021, during which on-campus housing and dining halls were open and in-person instruction had resumed for all classes. Face masks were required in classrooms and campus meeting spaces. Study procedures were approved by The Pennsylvania State University Institutional Review Board (STUDY000015710).⁴³

Measures

Affect Mean and Variability

Participants rated the extent to which they experienced 10 emotions the previous day; these items were modified from Positive and Negative Affect Scale (PANAS) and have been previously administered in experience sampling.⁴⁴ Participants rated four positive affect items (i.e., cheerful, content, energetic, enthusiastic) and six negative affect items (i.e., anxious, guilty, insecure, lonely, low, suspicious) on a scale from 1 (Never) to 7 (Always). Each subscale showed excellent reliability at both the person (αs = .94, .89) and day level (αs = .81, .70) for positive and negative affect, respectively. Daily positive affect and negative affect were calculated by averaging items across each subscale, with higher values indicating higher levels of each affect. Mean and individual standard deviation were calculated for each participant using all available data for positive affect and for negative affect across the 21 days. Data did not suggest inattention to daily measures; only 21 participants (1.1%) had no variability in either positive or negative affect across days, of whom only 11 reported the scale minimum or maximum daily. Although the individual standard deviation can be biased by missing data, neither positive nor negative affect variability was correlated with the number of completed surveys in this study (rs < |.03|, ps > .2) and the individual standard deviation has been commonly measured in prior studies.^{7,23,32,36,45} Furthermore, we observed an identical pattern of results when limiting the sample to participants who completed at least two weeks of daily surveys (n = 1454) and in sensitivity analyses additionally controlling for number of completed surveys.

Anxiety

As part of the psychosocial survey, participants rated the extent to which they agreed with six items regarding anxiety symptoms (e.g., “My heart races for no good reason.”, “My thoughts are racing.”) from the generalized anxiety subscale of the Counseling Center Assessment of Psychological Symptoms (CCAPS-34)⁴⁶ on a scale from 0 (Not at all like me) to 4 (Extremely like me). Items showed high reliability (α = .87), and a mean was calculated with higher values indicating higher anxiety symptoms. Following Locke et al., cutoff scores of 1.3+ and 2.1+ were used to identify moderate anxiety risk and elevated anxiety risk, respectively.⁴⁶

Depression

Participants rated 10 items regarding how frequently they experienced depressive symptoms (e.g., “I felt lonely”, “I could not get going”) using the Center for Epidemiological Studies Depression (CESD-10).⁴⁷ Each item was rated on a scale from 0 (None of the time) to 3 (Most or all of the time). Two items were reverse-coded. Items showed high reliability (α = .86), and a sum was calculated with higher values indicating higher depressive symptoms. Following Andresen et al., a cutoff score of 11 or higher was used to identify elevated depression risk.⁴⁷

Flourishing

Participants rated the degree to which they agreed with eight items regarding psychological flourishing in the psychosocial survey (e.g., “My social relationships are supportive and rewarding.”, “I am engaged and interested in my daily activities.”) using the flourishing scale.⁴⁸ Items were rated on a scale from 1 (Strongly disagree) to 7 (Strongly agree). Items showed high reliability (α = .90), and a mean was calculated with higher values indicating higher flourishing.

Analytic Strategy

Separate regression models tested associations between affect variability and anxiety symptoms, depressive symptoms, and psychological flourishing, each as continuous variables (see Supplemental Information for conceptual models). Positive affect variability and negative affect variability were tested in separate models, and all models controlled for race (dummy-coded relative to sample majority group, white), LGBTQ+ (0 = non-LGBTQ+, 1 = LGBTQ+), gender (0 = sample majority group, female; 1 = male and different identity), first-generation status (0 = continuing-generation, 1 = first-generation), and age (mean-centered). Supplemental analyses modeled a three-level categorical indicator for anxiety (coded 1 = low, 2 = moderate, and 3 = elevated risk) using ordinal regression and a binary indicator of depressive symptoms (0 = not elevated, 1 = elevated risk) using logistic regression. All ordinal regression models met the assumption of proportional odds.

All models tested same-valence (i.e., positive or negative) mean affect as a moderator of the associations between affect variability and mental health. For statistically significant interaction coefficients, regions of significance were identified using the Johnson-Neyman technique.⁴⁹ This technique identifies the values of the moderator (i.e., mean affect) at which the predictor (i.e., affect variability) relates to the outcome. For models in which the interaction coefficient was not significant (p > .05), the interaction was removed as a predictor and models tested the main effects of affect variability, controlling for mean affect. This modeling resulted in three regression models testing the Variability × Mean interaction for positive affect (one per mental health outcome) and for negative affect.

Because the three mental health outcomes were inter-correlated, structural equation models were specified in post-hoc analyses to test associations between affect variability and anxiety symptoms, depressive symptoms, and psychological flourishing simultaneously, allowing residuals for each outcome to covary. These models initially included all covariates and then omitted nonsignificant covariates from the final models. Data were analyzed in Stata 16.1.

Results

Participants generally reported high anxiety and depressive symptoms, with nearly half reporting moderate to elevated anxiety (24.43% moderate, 24.91% elevated) and elevated depressive symptoms (47.11%). The daily surveys indicated that participants were generally low in negative affect and higher in positive affect (Table 1). Participants reported slightly higher positive affect variability than negative affect variability, as measured by the person-specific standard deviation of ratings across days (Table 1). Histograms of distributions of positive and negative affect variability are presented in Figure 1, and box plots are presented across the sample as well as by gender, LGBTQ+ status, first-generation status, and race in Supplemental Figures S1–S3. Mean negative affect was significantly correlated with negative affect variability, such that individuals with higher mean negative affect tended to have greater daily fluctuations in negative affect (Table 1, Fig. 1). However, mean positive affect was not associated with positive affect variability. Neither positive nor negative affect variability differed by age (Supplemental Figure S4). There were no outliers on negative affect variability and four outliers on positive affect variability (i.e., over four standard deviations above the mean). Excluding outliers resulted in an identical pattern of results.

Table 1.

Descriptive statistics and correlations for study variables.

Variable	M	SD	Min	Max	1.	2.	3.	4.	5.	6.	7.	8.
1. Depression	10.93	6.44	0.00	30.00	—
2. Anxiety	1.39	1.04	0.00	4.00	.71*	—
3. Flourishing	45.35	7.40	8.00	56.00	−.57*	−.35*	—
4. Mean Negative Affect	2.30	0.90	1.00	6.77	.62*	.57*	−.42*	—
5. Negative Affect Variability	0.60	0.31	0.00	2.58	.38*	.32*	−.19*	.54*	—
6. Mean Positive Affect	3.68	1.13	1.04	7.00	−.47*	−.30*	.51*	−.30*	−.18*	—
7. Positive Affect Variability	0.89	0.35	0.00	2.78	.06	.07*	.10	.04	.40*	.04	—
8. Age	19.80	1.33	18.00	24.00	.02	.04	.00	−.06*	−.03	−.06*	.03	—

Note:

^*p < .01.

SD=Standard Deviation.

Figure 1.

Associations between mean emotion and emotion variability (positive affect above, negative affect below). Histograms display distributions along axes.

Positive Affect

Three regression models tested whether associations between positive affect variability across days and anxiety symptoms, depressive symptoms, and flourishing differed by mean positive affect. For each outcome, the Mean × Variability interaction term was nonsignificant (ps > .12). Thus, final models include the main effects of positive affect variability, controlling for mean positive affect. Higher mean positive affect was related to lower anxiety and depressive symptoms and to higher psychological flourishing (Table 2). As hypothesized, higher positive affect variability was associated with higher anxiety and depressive symptoms. Interestingly, higher positive affect variability was also associated with higher psychological flourishing.

Table 2.

Mental health outcomes as a function of positive affect variability and mean positive affect.

	Flourishing			Anxiety Symptoms			Depressive Symptoms
Variable	B	SE	β	B	SE	β	B	SE	β
Intercept	44.93***	0.29	—	1.10***	0.04	—	9.96***	0.26	—
Primary Predictors
Positive Affect Variability	1.52***	0.42	.07	0.15*	0.06	.05	1.13**	0.38	.06
Positive Affect Mean	3.22***	0.13	.49	−0.26***	0.02	−.28	−2.53***	0.12	−.44
Covariates
Age	0.14	0.11	.03	0.02	0.02	.02	−0.04	0.1	−.01
Female	1.57***	0.31	.10	0.38***	0.05	.17	0.56*	0.28	.04
Asian	−0.91*	0.43	−.04	−0.28***	0.06	−.09	−0.31	0.38	−.02
Black	−0.68	0.94	−.01	−0.18	0.14	−.03	1.61	0.85	.04
Hispanic	−0.15	0.55	−.01	−0.04	0.08	−.01	−0.07	0.49	.00
Different Identity or Multiracial	−0.57	0.74	−.02	−0.03	0.11	−.01	0.31	0.66	.01
LGBTQ+	−2.06***	0.38	−.11	0.45***	0.06	.17	2.74***	0.34	.16
First-Generation Status	−0.61	0.39	−.03	0.03	0.06	.01	0.71*	0.35	.04

Note: Positive affect variability, positive affect mean, and age were centered at the grand-mean. Female was dummy-coded (Female = 1, all other genders = 0). Each racial category was dummy-coded relative to non-Hispanic white (majority group). LGBTQ+ was dummy-coded (LGBTQ+= 1, not LGBTQ+= 0). First-generation status was dummy-coded (first-generation = 1, continuing-generation = 0).

^*p<.05,

^**p<.01,

^***p<.001.

Supplemental analyses used logistic regression to test whether positive affect variability was similarly associated with clinical anxiety and depression risk. Again, interaction terms were not significant, ps > .07, so models tested main effects of positive affect variability. Higher variability was associated with higher odds of moderate/elevated anxiety (B = 0.30, SE = 0.13, p = .024, 95% CI [0.04, 0.57], OR = 1.11) and elevated depression risk (B = 0.40, SE = 0.15, p = .009, 95% CI [0.10, 0.70], OR = 1.15; Supplemental Table S1).

Negative Affect

Three regression models were then specified to test associations between negative affect variability and the mental health outcomes. In contrast to results for positive affect, Mean Negative Affect × Negative Affect Variability interaction terms were consistently significant (ps < .004), such that the associations between negative affect variability and each outcome differed by mean negative affect (Table 3). The Johnson-Neyman technique was used to probe this interaction and revealed that, among individuals with higher levels of mean negative affect, higher negative affect variability was related to lower anxiety and depressive symptoms and higher psychological flourishing (Fig. 2). Associations were also probed using simple slopes (Fig. 3). Specifically, these associations were statistically significant for individuals with mean negative affect values above 3.42 for anxiety symptoms, 3.02 for depressive symptoms, and 2.66 for flourishing (1.28, 0.83, and 0.43 standard deviations above the average of mean negative affect, respectively). At low levels of mean negative affect, higher negative affect variability was related to greater anxiety and depressive symptoms, but not psychological flourishing. Associations were significant for individuals with mean negative affect values below 1.53 for anxiety symptoms and 2.33 for depressive symptoms (0.84 standard deviations below and 0.06 standard deviations above, respectively, the average of mean negative affect).

Table 3.

Mental health outcomes as a function of negative affect variability and mean negative affect.

	Flourishing			Anxiety Symptoms			Depressive Symptoms
Variable	B	SE	β	B	SE	β	B	SE	β
Intercept	44.62***	0.31	—	1.20***	0.04	—	10.57***	0.23	—
Primary Predictors
Negative Affect Variability	0.47	0.60	.02	0.04	0.07	.01	1.79***	0.45	.09
Negative Affect Mean	−3.56***	0.2	−.43	0.63***	0.03	.54	4.05***	0.15	.56
Negative Affect Variability × Negative Affect Mean	1.80***	0.53	.07	−0.20**	0.07	−.06	−2.11***	0.4	−.10
Covariates
Age	−0.14	0.12	−.02	0.06***	0.01	.07	0.27**	0.09	.06
Female	2.21***	0.33	.14	0.27***	0.04	.12	−0.14	0.25	−.01
Asian	−1.46**	0.45	−.07	−0.28***	0.06	−.09	−0.09	0.34	.00
Black	−1.99*	0.98	−.04	−0.01	0.12	.00	2.89***	0.74	.07
Hispanic	−0.01	0.57	.00	−0.09	0.07	−.02	−0.34	0.43	−.01
Different Identity or Multiracial	−1.29	0.77	−.03	0.06	0.10	.01	1.08	0.58	.03
LGBTQ+	−2.08***	0.40	−.11	0.32***	0.05	.12	2.11***	0.3	.13
First-Generation Status	−1.57***	0.41	−.08	0.07	0.05	.03	1.34***	0.31	.08

Note: Negative affect variability, negative affect mean, and age were centered at the grand-mean. Female was dummy-coded (Female = 1, all other genders = 0). Each racial category was dummy-coded relative to non-Hispanic white (majority group). LGBTQ+ was dummy-coded (LGBTQ+= 1, not LGBTQ+= 0). First-generation status was dummy-coded (first-generation = 1, continuing-generation = 0).

^*p<.05,

^**p<.01,

^***p<.001.

Figure 2.

Regions of significance for associations between negative affect variability and mental health outcomes at different levels of mean negative affect. Interactions were probed using the Johnson-Neyman technique. Pink regions indicate values of mean negative affect at which the association is non-significant. Blue regions indicate values for mean negative affect at which the association significantly differs from 0. For anxiety and depressive symptoms (above), associations are significant at low and high values of mean negative affect. For flourishing (below), associations are significant only at high of mean negative affect.

Figure 3.

Simple slopes of associations between negative affect variability and mental health outcomes at different levels of mean negative affect (1 standard deviation below the sample mean, the sample mean, 1 standard deviation above the sample mean). Note: SD = Standard Deviation.

Again, supplemental analyses tested whether negative affect variability was related to risk of moderate/elevated anxiety and elevated depression. Associations between negative affect variability and mental health again varied by mean negative affect, as indicated by Mean × Variability interactions; B = −0.61, SE = 0.20, p = .002, 95% CI [−1.00, −0.22]) for anxiety risk; B = −1.31, SE = 0.24, p < .001, 95% CI [−1.78, −0.84] for depression risk (Supplemental Table S2). Consistent with models analyzing continuous anxiety and depressive symptoms, higher negative affect variability was associated with lower risk for elevated anxiety and depressive symptoms at high mean negative affect (values above 3.72 and 3.02, respectively; 1.62 and 0.83 standard deviations above the mean), and associated with higher risk at low mean negative affect (values below 2.72 and 2.33, respectively; 0.49 and 0.06 standard deviations above the mean).

Post-Hoc Analyses: Modeling Mental Health Outcomes Simultaneously

To account for shared variance across outcomes, structural equation models were specified to test associations between affect variability and anxiety symptoms, depressive symptoms, and psychological flourishing simultaneously with residuals correlated between each pair of outcomes. Separate models were tested for positive affect variability and negative affect variability. Models showed appropriate fit (RMSEAs < .01, CFI = 1.00, SRMR < .01, ps = .3), and results mirrored those based on the separate regression models reported above. Positive affect variability was significantly associated with higher anxiety symptoms (B = 0.15, SE = 0.06, 95% CI [0.02, 0.27], β = .05, p = .023), higher depressive symptoms (B = 1.10, SE = 0.37, 95% CI [0.37, 1.83], β = .06, p = .003), and greater flourishing (B = 1.54, SE = 0.42, 95% CI [0.72, 2.36], β = .07, p < .001; Supplemental Fig. S5; Table S3). The association between negative affect variability and each mental health outcomes was significantly moderated by mean negative affect (all ps < .005; Supplemental Fig. S6; Table S4).

Discussion

The present study examined how positive and negative affect variability relate to negative (i.e., anxiety and depressive symptoms) and positive (i.e., flourishing) measures of mental health in college students during the COVID-19 pandemic. Higher positive affect variability was related to higher anxiety and depressive symptoms and, surprisingly, higher flourishing, irrespective of mean affect. In line with the Fragile Desirable Affect Theory,^5,6,29 associations for negative affect variability consistently differed by mean negative affect. High negative affect variability was related to higher anxiety and depressive symptoms for individuals with low mean negative affect. In turn, high negative affect variability was related to higher flourishing and lower anxiety and depressive symptoms for individuals with high mean negative affect, suggesting that affective flexibility may be beneficial for individuals with high negative affect and that persistent, high negative affect may be a stronger risk factor than variable, high negative affect. Taken together, higher affect variability can relate to both better and poorer mental health, and research should consider mean affect when interpreting negative affect variability.

Associations for Positive Affect Variability

Prior research has generally found that higher negative affect variability and, to a lesser degree, positive affect variability are related to poorer health,³ but limited research has assessed how associations between affect variability and mental health differ by mean affect. The present study extended this work by assessing both positive and negative affect variability and examining flourishing as an index of positive mental health, in addition to anxiety and depressive symptoms. Different patterns of results emerged for positive and negative affective variability, as associations with positive affect variability were not moderated by mean positive affect, in alignment with results from studies of adolescents and young adults.⁸ This finding is at odds with prior evidence that higher positive affect variability is related to higher distress and depressive symptoms for individuals with high mean positive affect but not for individuals with low mean positive affect.^7,32,33,50 Results have been similarly mixed for physical health outcomes.^29,31 Fragile Desirable Affect Theory may not have been supported for positive affect variability in the present study because we recruited young adults during the COVID-19 pandemic, when individuals had low positive affect³⁶ and when mean positive affect was less related to affect variability than in prior studies.^7,8,32,36

We also found that higher positive affect variability was simultaneously related to greater anxiety and depressive symptoms and to greater psychological flourishing, suggesting that positive affect variability may be independently related to positive and negative measures of mental health. This finding aligns with evidence that mental health is not merely the absence of psychopathology.^10,34 It has been posited that there may be a healthy level of affective flexibility—as opposed to stable high positive affect—such that individuals are invested in their daily experiences but avoid clinical psychopathology.²⁵ High positive affect variability may suggest that individuals are sensitive to daily experiences and consequently have large day-to-day fluctuations in affect. Recent evidence suggests that variability in use of emotion regulation strategies across circumstances can be adaptive in daily life.⁵¹ Variability in affect across days may be similarly beneficial for flourishing. Previous research has found that a moderate level of positive affect variability is related to favorable diurnal cortisol profiles, and cortisol profiles have in turn been prospectively related to mental health.^45,52 This association also may have emerged because the sample comprised college students, who tend to be particularly reactive to daily academic and social experiences.^53,54 College students may feel both fulfilled and distressed from social experiences, and associations should be replicated among older adults.

Differences in Associations for Negative Affect Variability by Mean Negative Affect

Consistent with prior results in college students,³³ young adults,⁸ and older adults,⁷ associations for negative affect variability differed by mean negative affect and supported Fragile Desirable Affect Theory.^5,6,29 Our results extend this theory to the context of college students during the COVID-19 pandemic and to flourishing as an aspect of positive well-being. Higher affect variability was related to poorer mental health for individuals low in negative affect and to better mental health for those high in negative affect. For individuals with low negative affect, greater fluctuations may suggest heightened sensitivity to external factors such as negative experiences, and individuals with higher negative affect variability may consequently be at risk for poorer mental health. In turn, for those with higher negative affect, fluctuations may suggest that they experience days of relatively low negative affect, as opposed to persistently high negative affect. These “breaks” in negative affect may be protective against psychopathology and promote flourishing. Whereas past findings observed these protective effects only for individuals with very high mean negative affect (over the 99^th percentile),^7,8 associations emerged at less extreme values in the present study, potentially because the pandemic resulted in overall higher levels of day-to-day negative affect.³¹

People can have high affect variability due to heightened sensitivity to the environment, difficulties with regulating responses, environmental instability, or a combination of these factors. People naturally differ in their proclivity for mounting psychobiological response to stimuli and for modulating these responses.^2,3 Interventions that target affective sensitivity and regulation—or emphasize how enhanced sensitivity can confer flourishing—may effectively reduce risk of psychopathology for individuals with high affect variability. Furthermore, the frequency and severity of stressors varies across individuals. Past research has found that associations between positive affect variability and cortisol, one indicator of physical health, differ by both mean levels of positive affect and perceived stress.⁵ Those from marginalized backgrounds (e.g., racially minoritized, less educated) experience unique stressors that shape daily affect,^55,56and the COVID-19 pandemic disproportionately impacted marginalized groups.⁵⁷ Community-based interventions may focus on scaffolding less stressful environments in addition to emotion-focused approaches. Future studies with daily measures of specific stressors or that allow participants to describe their daily experiences may be needed to identify whether individuals’ sensitivity to the environment or differences in the environment (e.g., stressor severity or frequency) explain associations between affect variability and mental health.

Importantly, associations were observed during the COVID-19 pandemic, a period of unpredictability that greatly impacted college students’ emotions.⁵⁸ These findings suggest that affect variability could relate to risk and potential resilience during such periods. We may have been well-positioned to identify moderating effects of mean affect in the association between affect variability and mental health because individuals tended to have poorer affective well-being (i.e., lower mean positive emotion, higher mean negative emotion) and elevated psychopathology during the COVID-19 pandemic.³⁷ Results were likely affected by consequences of social isolation but not illness related to the pandemic given that participants rarely reported COVID-19 symptoms during the study period. We did not find that associations between positive affect variability and mental health differed by mean positive affect during the pandemic, in line with findings from a recent mega-analysis of adolescents and young adults.⁸ Still, it is possible that moderated associations may emerge during less stressful time periods in line with prior studies.^7,31–33,50

Demographic Differences in Mental Health

College students’ demographics were also related to mental health, irrespective of affect levels and variability. LGBTQ+ students had poorer mental health for all outcomes than non-LGBTQ+ students, possibly because of minoritized stress and low institutional support.⁵⁹ Risk may have also been exacerbated by the COVID-19 pandemic because many individuals returned to communities that may not affirm their identities.⁶⁰ Female and Asian participants consistently reported higher anxiety and lower flourishing than non-female and white participants. Gender differences in depression risk have been observed among college students both before and during the COVID-19 pandemic.^61,62 Racial differences may be due to anti-Asian stigma and discrimination related to the pandemic.⁶³ Associations may not have emerged for depressive symptoms because of the extent to which depressive symptoms increased for college students generally during the COVID-19 pandemic. It is important to note that studies suggest that women and LGBTQ+ college students reported lower anxiety during relative to before the pandemic at certain institutions, potentially because unsupportive institutions could contribute to greater stress for these groups during in-person instruction.⁶⁰ It is possible that the magnitude of differences may have shifted between before and during the pandemic, although the present study lacked measures of mental health prior to the COVID-19 pandemic.

Models testing negative affect indicated that Black and first-generation college students were higher in depressive symptoms and lower in psychological flourishing than white participants and continuing-generation students, respectively. Black students at primarily white institutions may have poorer mental health related to marginalization than white students and Black students attending historically Black institutions, and first-generation students may have poorer mental health than continuing-generation students because of lower socioeconomic resources, greater academic pressure, and lower institutional support.^64,65 These differences were not significant in models testing positive affect as a predictor, suggesting that positive affect may account for differences in mental health by race and first-generation status.

Implications for Higher Education

These results highlight that positive and negative affect are uniquely related to health outcomes during the COVID-19 pandemic and that maintaining consistently high positive affect during such times of unpredictability can be beneficial for individuals. Educational institutions can offer resources for promoting emotion regulation (e.g., cognitive reappraisal training, challenge versus stress mindset) and create opportunities to disrupt current emotional patterns and enhance variability during times of unpredictability. For instance, encouraging students to practice mindfulness, savor positive experiences, or gratitude journaling may promote positive affect despite potentially turbulent negative affect. In light of our results that affect variability can benefit certain individuals, resources could be tailored for students and acknowledge that flexibly responding can benefit individuals with high negative emotion. Emerging evidence also suggests that cultivating a sense of purpose dampen affect reactivity,⁶⁶ and college students may feel overwhelmed and aimless as they discover what areas they may want to pursue for a career or higher education. College students can be encouraged to be invested in their academic, social, and extracurricular activities but also maintain stable routines and identify activities that are stimulating but consistent in order to maintain high levels of flourishing while reducing anxiety and depressive symptoms.

Given students’ investment in their academics, future studies are needed to identify how affect variability relates to students’ learning and academic experiences. Students can have affective responses to learning,⁵⁴ and negative affect can disrupt learning.⁶⁷ Our results highlight that students’ backgrounds can impact their proclivity for affect variability, which may in turn impact their learning experiences. Restructuring classes to be less emotionally evocative (e.g., fewer high-stakes exams) may benefit college students’ mental health and facilitate a more equitable environment.

Limitations

Generalization of findings to other populations is limited by study design and measures. The sample included primarily white college students. Affect variability may differentially relate to mental health among more racially diverse samples or young adults who did not enroll in college, as people of racially minoritized and lower socioeconomic status often experience additional hardships that impact affect and mental health.^55,56 Although participants completed up to 21 daily checklists, they reported overall affect only once per day. Future studies can use ecological momentary assessment to assess affect variability within a day and environmental factors (e.g., stressful live events, daily hassles) that contribute to affect variability. The daily protocol was administered immediately after the psychosocial survey due to logistical feasibility, despite affect variability theoretically contributing to mental health. Daily measures are generally indicative of individual’s daily life beyond the measured period,⁴² but prospective studies are needed to verify directional associations. This study also included only one measure of positive mental health, and future studies should incorporate additional measures (e.g., life satisfaction, meaning). Finally, abbreviated versions of the PANAS were used, which could impact the results. Discrete negative emotions such as anger and sadness were omitted from the negative affect scale, which could have underestimated negative affect estimates during the COVID-19 pandemic. The circumplex model of affect posits that both valence and arousal are distinct dimensions of affect,¹⁷ and future studies should incorporate items with varied arousal.

Conclusions

Results suggested that affect variability is not necessarily indicative of poorer mental health and that associations with negative affect variability differ by mean negative affect during the COVID-19 pandemic. Higher positive affect variability was simultaneously associated with higher anxiety and depressive symptoms and higher flourishing, potentially because heightened positive affect variability can suggest greater fluctuations in positive affect in response to daily positive and negative experiences. Implications of negative affect variability differed by mean affect. At high mean negative affect, negative affect variability was related to lower anxiety and depressive symptoms, suggesting that individuals with persistent high negative affect had poorer mental health than individuals with higher variability. In contrast, at low mean negative affect, higher negative affect variability was related to greater anxiety and depressive symptoms. Associations may be specific to college students during the COVID-19 pandemic, when individuals had poorer well-being and greater unpredictability. Results highlight the importance of assessing affective valence and accounting for both an individual’s mean affect and fluctuations in affect when considering associations between affect variability and mental health.