Objective There is little long-term data on the association between the serum albumin levels and mortality in community-based populations. We aimed to determine whether the serum albumin level is an independent risk factor for all-cause and cause-specific death in a community-based cohort study in Japan. Methods In 1999, we performed a periodic epidemiological survey over a 15-year period in a population of 1,905 healthy subjects (783 males, 1,122 females) who were older than 40 years of age and who resided in Tanushimaru, a rural community, in Japan. Over the course of the study, we periodically examined the blood chemistry of the study subjects, including their serum albumin levels. Their baseline serum albumin levels were categorized into quartiles. Results The baseline albumin levels were significantly associated with age (inversely), body mass index (BMI), diastolic blood pressure, lipid profiles [high density lipoprotein-cholesterol (HDL-c), low-density lipoprotein-cholesterol (LDL-c) and triglycerides] and estimated glomerular filtration rate (eGFR). After adjusting for confounders, a Cox proportional hazards regression analysis demonstrated that a low serum albumin level was an independent predictor of all-cause death [hazard ratio (HR): 0.39, 95% confidence interval (CI): 0.24-0.65], cancer death (HR: 0.43, 95% CI: 0.18-0.99), death from infection (HR: 0.21, 95% CI: 0.06-0.73) and cerebro-cardiovascular death (HR: 0.19, 95% CI: 0.06-0.63). The HRs for all-cause and cerebro-cardiovascular death in the highest quartile vs. the lowest quartile of albumin after adjusting for confounders were 0.59 (95%CI:0.39-0.88) and 0.15 (95%CI: 0.03-0.66), respectively. Conclusion The serum albumin level was thus found to be a predictor of all-cause and cerebro-cardiovascular death in a general population.