Gestational age is an important birth characteristic examined in epidemiological studies. Though there are well-documented problems with the reporting of gestational age on birth certificates, schemes for addressing this issue have not been systematically evaluated. With singleton births from the 1995–97 US linked birth/infant death files, we compared a handful of perinatal outcome estimates derived from the resulting analytical files using two published methods often used to manage inconsistent gestational age data. The first method (Alexander et al., 1996), provides cut-points for implausible birthweight–gestational age combinations and excludes infants with birthweights outside a plausible range. The second (Zhang and Bowes, 1995), provides different cut-points for implausible birthweight–gestational age combinations and then substitutes the clinical gestational age estimate for the original value, if available, reducing the number of births at the affected gestational age, but excluding fewer births from the resulting analytical files. The Alexander method excluded 0.4% of our study population and the Zhang method reassigned and excluded 1.0% and 0.2%, respectively; however, over 20% of birth records with gestational age 28–30 weeks were modified by either method. Using either method, more high-risk than low-risk and more black than white births were excluded. These differential exclusions affected corresponding perinatal outcome estimates and relative risks between maternal risk groups for preterm delivery, gestation-specific infant mortality and birthweight; overall infant mortality rates were not affected. Systematic comparisons between results of different studies will need to consider the data modifications used, the populations affected, and the outcomes assessed when drawing conclusions.