Objective The serum albumin level has been reported to be positively associated with blood pressure in some cross-sectional studies. However, the longitudinal relationships between the serum albumin level and incident hypertension have been poorly studied. Methods The incidence of hypertension was calculated for each quartile of serum albumin in 1,385 normotensive men and 855 normotensive women without cardiovascular disease at baseline who revisited for four years thereafter. The hazard ratios (HRs) of incident hypertension were calculated for each one SD increase in the serum albumin level and for the higher quartiles of serum albumin compared with the lowest quartile adjusted for smoking, alcohol drinking, body mass index, proteinuria, estimated glomerular filtration rate, uric acid, fasting glucose, log triglycerides, log high-sensitivity C-reactive protein, white blood cell count, hemoglobin, and systolic blood pressure. Results During four years of follow-up (mean; 3.1 years), 242 men (17.5%) and 89 women (10.4%) developed hypertension. The incidence of hypertension significantly decreased through the quartiles of albumin (p for trend=0.012). The HR (95% confidence interval (CI); p value) of hypertension for each one SD increase in the serum albumin level was 0.779 (0.696-0.872; <0.001). Compared with the lowest quartile of serum albumin, the HRs (95% CIs; p values) of hypertension for the second, third and fourth quartiles were 0.765 (0.574-1.018; 0.066), 0.628 (0.440-0.895; 0.010) and 0.520 (0.364-0.742; <0.001), respectively. Conclusion A decreased serum albumin level was found to be a significant predictor of hypertension in a Japanese health screening population.