Cigarette smoking continues to be one of the major risk factors for increased morbidity and mortality worldwide. Among many adverse health effects, smoking can induce erythrocytosis, which is commonly believed to result from elevated serum erythropoietin (EPO) levels. Currently, however, this notion is only alleged, without data available to substantiate it. Hence, we analyzed data from the Prevention of Renal and Vascular End-Stage Disease study, a prospective population-based cohort study. Smoking behavior was quantified as number of cigarettes smoked per day and as 24-hour urinary cotinine excretion levels, an objective and quantitative measure of nicotine exposure. In 6808 community-dwelling participants, the prevalence of nonsmokers, former smokers, and current smokers were 29%, 43%, and 28%, respectively. Hematocrit levels were higher in current smokers (41.4%±3.6%) than in nonsmokers (40.3%±3.6%) (P<.001). In contrast, median EPO levels were lower in current smokers (7.5 IU/L; interquartile range [IQR], 5.7-9.6 IU/L) than in nonsmokers (7.9 IU/L; IQR, 6.0-10.7 IU/L) (P<.001). In multivariate linear regression analysis, current smoking, compared with nonsmoking, was independently positively associated with hematocrit levels (β=.12; P<.001) and hemoglobin levels (β=.11; P<.001), but inversely associated with EPO levels (β=-.09; P<.001). In sensitivity analyses, we observed a dose-dependent inverse association of smoking exposure reflected by 24-hour urinary cotinine excretion levels with EPO levels. Contrary to common belief, we identified that in the general population, smoking is inversely associated with EPO levels. Future mechanistic insight is needed to unravel the currently identified association, and if reproduced in other studies, guidelines for diagnosis of secondary erythrocytosis may need to be revisited.