The objective of the present study was to test the hypothesis that volumetric formulae for fetal weight estimation are more tolerant to measurement error than exponential models. A mathematical model of normal fetal growth was constructed, using published British reference standards for biometric variables. Observed measurements were computer-generated by contaminating reference measurements with error terms according to their published coefficients of variation. The error in weight estimation was computed as the percentage difference between weight estimates derived from the observed biometric variables and the true measurements. A total of nine weight estimation formulae were tested. Campbell's formula appeared to be most affected by observational errors, especially before 38 weeks. In this range, they varied up to 9%. The most tolerant was Shepard's formula, with errors of only approximately 2.8%. Other formulae showed errors of approximately 5−6%. With the exception of Campbell's formula, the effect of gestational age was minimal. There was no correlation between percentage error and fetal size. Combining ultrasound biometric variables into a fetal weight estimate does not always exaggerate the errors of the original measurements. There were no significant differences between volumetric and exponential formulae.