The accuracy of suicide statistics was assessed by comparing published Health Department suicide rates for an area of New York City with Medical Examiner records. For the period 1968--1979, records from the Medical Examiner's Office were searched to determine all deaths classified as definite suicides. Another group of deaths was considered suicide by the Medical Examiner but never classified as such. These deaths we labeled "assigned suicides." When definite suicides were compared with all deaths considered suicide by the Medical Examiner (definite and assigned suicides), black suicide was underestimated by 80 per cent and white suicide by 42 per cent. Underestimation was the same for males and females but varied by age group. In 1968, when the seventh revision of the International Classification of Deaths (ICD) was used, Health Department suicide rates for blacks were almost identical to Medical Examiner rates, while white rates were underestimated by 25 per cent. In 1969--1970, when the eighth revision was used, Health Department statistics underestimated black suicides by 82 per cent and white suicides by 66 per cent. Reasons for the underestimations were related to the methods used in committing suicide by the two ethnic groups and to the ways that suicide classification has changed from the seventh to eighth revision. Implications for research using official death certificate reports are discussed.