Esophageal and gastric cancers share some risk factors. This study aimed to compare the long-term trends in mortality rates of esophageal and gastric cancers in China to provide evidence for cancer prevention and control. Mortality data were derived from 103 continuous points of the Disease Surveillance Points system during 1991-2009, stratified by gender and urban-rural locations. Age-period-cohort models were used to disentangle the time trends of esophageal and gastric cancer mortality. The downward slope of the period effect for esophageal cancer was steeper than that for gastric cancer in rural areas. Cohort effect patterns were similar between esophageal and gastric cancers, with an inverse U-shape peaking around the late 1920s and early 1930s. A second peak, appearing around the 1950s, was weaker than the first but apparent in males, especially for esophageal cancer. The more marked changes in period effect for esophageal cancer in rural areas suggest esophageal cancer screening practices are effective in reducing mortality, and similar programs targeting gastric cancer should be implemented. The similarities of the cohort effects in these two cancers support the implication of nutrition deficiency in early childhood in the development of upper gastrointestinal cancer.