OBJECTIVES: To simulate each of two hypothesised errors in the National Cancer Registry (recording of the date of recurrence of cancer, instead of the date of diagnosis, for registrations initiated from a death certificate; long term survivors who are never notified to the registry), to estimate their possible effect on relative survival, and to establish whether lower survival in the UK might be due to one or both of these errors. DESIGN: Simulation study. SETTING: National Cancer Registry of England and Wales. Population Patients diagnosed as having breast (women), lung, or colorectal cancer during 1995-2007 in England and Wales, with follow-up to 31 December 2007. MAIN OUTCOME MEASURE: Mean absolute percentage change in one year and five year relative survival associated with each simulated error. RESULTS: To explain the differences in one year survival after breast cancer between England and Sweden, under the first hypothesis, date of diagnosis would have to have been incorrectly recorded by an average of more than a year for more than 70% of women known to be dead. Alternatively, under the second hypothesis, failure to register even 40% of long term survivors would explain less than half the difference in one year survival. Results were similar for lung and colorectal cancers. CONCLUSIONS: Even implausibly extreme levels of the hypothesised errors in the cancer registry data could not explain the international differences in survival observed between the UK and other European countries.