Changes across Periods in the Proportion of Older Americans Reporting Low Happiness

Using five repeated cross-sectional waves of the US General Social Survey (GSS)—2010, 2012, 2014, 2022 and 2024—this study analyses 2,949 respondents aged 65 and over. It examines changes in the proportion reporting that they are “not too happy” between 2010–2014 and 2022–2024, how these changes vary across marital-status and education groups, and their sensitivity to adjustments for survey mode. The analysis combines survey-weighted multinomial logit models and marginally standardised predictions with shared primary sampling unit (PSU) resampling and Holm adjustment for multiple testing. After adjustment for demographic composition, the standardised proportion reporting “not too happy” rose from 12.98% to 20.00%, an increase of 7.02 percentage points, or approximately 54.1% in relative terms; the 95% confidence interval for the percentage-point increase was 3.05 to 11.00. The in-person-only sample also showed an increase, of 6.71 percentage points, indicating that the direction of change does not depend on the inclusion of recent web respondents. This finding does not, however, fully rule out survey-mode or participation-selection effects. Omnibus tests of heterogeneity across marital-status and education groups provided no clear evidence of differences in their increases. The available evidence therefore neither establishes that the increase in low happiness was concentrated in a particular group nor demonstrates that all groups experienced the same degree of change. The estimated increase of 13.20 percentage points among respondents with a bachelor’s degree or higher is exploratory and warrants further investigation. The study’s theoretical and practical value lies in distinguishing current disadvantage from changes across periods and illustrating an analytical approach that combines probability-scale interactions, survey-mode sensitivity analyses and design-based inference for complex survey data. Its findings suggest that ageing policy should broaden the coverage of well-being monitoring while continuing to attend to groups, such as widowed respondents, with higher current levels of low happiness. Further assessment should consider individual health, economic circumstances, social connections and care needs. Author information: Liang Xu holds an MA in Advertising and Marketing from the University of Leeds and a bachelor's degree in Accounting (CPA) from Southwestern University of Finance and Economics. Publication status: Preprint; not peer reviewed. Files: The main manuscript is in English. The accompanying supplementary materials are in Chinese, with an English cover and summary figure, and contain Appendices A1–A4, Tables S1–S26, Figure S1, and a numbered reference list.