This study aimed to assess survival by ethnicity for all New Zealand children (0-14 years) diagnosed with cancer during 1990-1993 and second, to determine the accuracy of the children's routinely collected ethnicity information. Four hundred and nine children were followed up using two largely independent sources. We compared survival using the Cox model. Maori and Pacific Island children had the same survival as non-Maori/non-Pacific children for 'all cancers combined' (hazard ratios (HR): 0.98; 95% confidence interval (CI): 0.64-1.50 and 1.01; 95% CI: 0.53-1.89 respectively) and acute lymphoblastic leukaemia (HR: 1.09; 95% CI: 0.45-2.62 and 0.99; 95% CI: 0.24-4.16, respectively). The ethnicity data stored within the National Health Index and the New Zealand Cancer Registry showed reasonably close agreement with ethnicity data provided by the children's mothers at interview (Kappa statistics: 0.82 and 0.81 respectively) while the Mortality Collection showed only moderate agreement (Kappa statistic: 0.63). The point estimates in this study provide no evidence of ethnic disparities in survival from childhood cancer in New Zealand. However, even in this national study there were small numbers of Maori and Pacific children leading to wide confidence intervals. We therefore recommend cautious interpretation.