Esophageal cancer is one of the most commonly diagnosed malignant tumors in China. This study aimed to examine the temporal trend of esophageal cancer mortality rates during the period 1987-2009 in both rural and urban settings and to detect the effects of year of death and year of birth on the trends using joinpoint regression analysis and an age-period-cohort model. Age-standardized mortality rates were calculated by the direct method using the world population of 1960, and joinpoint regression was performed to obtain the annual percentage change (APC) in mortality rate. Poisson regression models were fitted to evaluate the period and cohort effects after adjusting for age. During the period 1987-2009, age-standardized mortality rates showed an overall significant decrease for rural females (APC=-2.3, 95% confidence interval [CI]: -3.3%, -1.2%), urban males (APC=-1.8, 95% CI: -2.6%, -1.0%) and urban females (APC=-3.7, 95% CI: -4.9%, -2.4%), but the decrease was not statistically significant for rural males (APC=-0.9, 95% CI: -2.0%, 0.3%). After adjusting for age and with the birth cohort of 1900-1904 or period 1987-1991 as reference, the relative risk of successive cohorts decreased steadily and that of more recent periods kept relatively stable. The decreasing birth cohort effect in the recent generations could correspond to increased adoption of healthy dietary habits and life-styles in the population.