The relative and absolute risks of perinatal death that are estimated from observational studies are used frequently in counseling about obstetric intervention. The statistical basis for these estimates therefore is crucial, but many studies are seriously flawed. In this review, a number of aspects of the approach to the estimation of the risk of perinatal death are addressed. Key factors in the analysis include (1) the definition of the cause of the death, (2) differentiation between antepartum and intrapartum events, (3) the use of the appropriate denominator for the given cause of death, (4) the assessment of the cumulative risk where appropriate, (5) the use of appropriate statistical tests, (6) the stratification of analysis of delivery-related deaths by gestational age, and (7) the specific features of multiple pregnancy, which include the correct determination of the timing of antepartum stillbirth and the use of paired statistical tests when outcomes are compared in relation to the birth order of twin pairs.