Objectives: Despite media and public dialog portraying loneliness as a worsening problem, little is known about how the prevalence of loneliness has changed over time. Our study aims to identify (a) temporal trends in episodic and sustained loneliness (lonely in 1 wave vs consistently lonely in 3 consecutive waves); (b) trends across sociodemographic subgroups by sex, race/ethnicity, birth cohort, education, employment status, marital status, and living alone; and © longitudinal predictors of loneliness in middle-Aged and older Americans (≥50 years). Methods: Based on Waves 3 (1996) to 14 (2018) of the Health and Retirement Study (n = 18,841-23,227), we conducted a series of lagged mixed-effects Poisson regression models to assess trends of episodic and sustained loneliness in the overall and sociodemographic subgroup samples (by sex, race/ethnicity, birth cohort, education, employment, relationship, and living alone status). To examine the predictors of episodic and sustained loneliness, we used a multivariate mixed-effects Poisson regression model with all sociodemographic variables entered into the same model. Results: Episodic loneliness prevalence decreased from 20.1% to 15.5% and sustained loneliness from 4.6% to 3.6%. Trends were similar across most subgroups. Males, Caucasians, those born in 1928-1945, with university education, working, married/partnered, and those not living alone reported lower episodic and sustained loneliness, although associations with sustained loneliness were stronger. Discussion: Contrary to common perceptions, loneliness has decreased over 20 years of follow-up in middle-Aged and older Americans. Several sociodemographic subgroups have been identified as having a higher risk of loneliness, prompting targeted public health attention. © 2023 The Author(s). Published by Oxford University Press on behalf of The Gerontological Society of America.