Daily positive and negative affect during the COVID-19 pandemic

Abstract

The COVID-19 pandemic influenced emotional experiences globally. We examined daily positive and negative affect between May/June 2020 and February 2021 (N = 151,049; 3,509,982 observations) using a convenience sample from a national mobile application-based survey that asked for daily affect reports. Four questions were examined: (1) How did people in the United States feel from May/June 2020 to February 2021?; (2) What demographic variables are related to positive and negative affect?; (3) What is the relationship between experienced stressors and daily affect?; and (4) What is the relationship between daily affect and preventive behavior? Positive affect increased, and negative decreased over time. Demographic differences mirrored those from before the pandemic (e.g., younger participants reported more negative and less positive affect). Stressors such as feeling unwell, experiencing COVID-19 symptoms, exposure to COVID-19, and lack of sleep were associated with less positive and more negative affect. Exercising protective behaviors predicted future affect, and affect also predicted future protective behaviors (e.g., less protective behavior when happy but more when grateful and thoughtful). The implications for public health communication were discussed.

Keywords: COVID-19 pandemic; demographic differences; negative affect; positive affect; protective behavior; stressors.

Figure 1

The HWF application and sample note. (A) The HWF app: opening screen, COVID-19-related symptoms, affect ratings. (B) Demographic, affect, and protective behaviors data collected by the HWF app. (C) Distribution of users by geographic region, gender, age, living alone, race/ethnicity, county-level income, population density, number of children in household, and occupational category.

Figure 2

Positive and negative affect during the COVID-19 pandemic compared to May/June 2020. IPW GEE regression of (A) positive affect and (B) negative affect on pandemic month dummy variables. Covariates: demographics, sleep duration, whether feeling unwell, exposure to COVID-19, testing for COVID and test results, county-level COVID case rates and death rates, and county-level R_t values. Shown are 95% confidence intervals (n = 3,509,982 responses from 151,049 users). For coefficients, confidence interval values, and p-values, see Supplementary Table S3.

Figure 3

Associations between demographic variables and (A) positive and (B) negative affect. IPW GEE linear regression of (A) positive and (B) negative affect on demographics variables. Covariates: month dummy variables, sleep durations, whether feeling unwell, self-reported exposure to COVID-19, testing for COVID and test results, county-level case rates and death rates, and R_t values. Shown are estimated coefficients with 95% confidence intervals (n = 3,509,982 responses from 151,049 users). For coefficients, confidence interval values, and p-values, see Supplementary Table S4.

Figure 4

Associations between stressors and positive (A) and negative (B) affect. IPW GEE linear regression of (A) positive affect and (B) negative affect on different categories of stressors. Covariates: month dummy, county-level COVID-19 case rates and death rates, and testing for COVID and test results. Shown are estimated coefficients with 95% Cis (n = 3,509,982 responses from 151,049 users). For coefficients, confidence interval values, and p-values, see Supplementary Table S5.

Figure 5

Protective behavior predicting future (A) positive and (B) negative affect. IPW GEE linear regression models were performed separately for those who responded whether they stayed at home (n = 3,133,829 responses from 115,241 users) and those who indicated having left home in the previous 7 days (n = 1,701,325 responses from 98,919 users), but the results were combined into one figure for simplicity of presentation. Covariates: month dummy variables, demographics, sleep duration, whether feeling unwell, self-reported exposure to COVID-19, testing for COVID and test results, and county-level R_t values. Shown are estimated coefficients with 95% Cis. For coefficients, confidence interval values, and p-values, see Supplementary Table S6.

Figure 6

Positive affect (A) and negative affect (B) predicting future protective behavior. Three GEE logistic regression models were fit to predict staying home (n = 3,133,829 responses from 115,241 users), wearing facial covering, and social distancing (both n = 1,821,042 responses from 105,447 users) using average scores of affect variables in the previous 7 days. Covariates: demographics, month dummy variables, sleep duration, whether feeling unwell, and self-reported testing for COVID and test results. Shown are log ORs and the corresponding 95% CIs. For OR values, confidence interval values, and p-values, see Supplementary Table S7.