In this study, we explored the prognostic risk factors of elderly patients (≥65 years old) with lymph node-negative esophageal cancer (EC) and established a nomogram to evaluate the cancer-specific survival of patients. The surveillance, epidemiology, and end results database was used to collect data on patients diagnosed with EC. Univariate and multivariate Cox analyses were used to determine independent prognostic factors, and the nomogram for predicting cancer-specific survival of EC patients was constructed based on the independent prognostic factors obtained from the multivariate Cox analysis. To evaluate the predictive ability of the nomogram, calibration curves, concordance index (C-index), receiver operating characteristic curves, and decision curve analysis were conducted. Kaplan–Meier method was used to analyze the long-term outcomes of EC patients with different risk stratifications. A total of 3050 cases with lymph node-negative EC were randomized into the training cohort (1525) and the validation cohort (1525). Cancer-specific mortality at 1, 3, and 5 years in the entire cohort was 30.7%, 41.8%, and 59.2%, respectively. In multivariate Cox analysis, age (P < .001), marital status (P < .001), tumor size (P < .001), Tumor-node-metastasis stage (P < .001), chemotherapy (P = .011), radiotherapy (P < .001), and surgery (P < .001) were independent prognostic factors. The C-index for the training cohort was 0.740 (95% confidence interval [CI]: 0.722–0.758), and the C-index for the validation cohort was 0.738 (95% CI: 0.722–0.754). The calibration curve demonstrated the great calibration ability of the nomogram. Based on the area under the receiver operating characteristic curve, the nomogram demonstrated a higher sensitivity than the tumor-node-metastasis stage. Decision curve analysis showed the good clinical utility of the nomogram. The risk stratification system was established using the Kaplan–Meier curve and verified by the log-rank test (P < .001). The nomogram and risk stratification system can improve the accuracy of prediction to help clinicians identify high-risk patients and make treatment decisions.