Equations for predicting the meat, fat and bone proportions in beef carcasses using the European Union carcass classification scores for conformation and fatness, and hindquarter composition were developed and their accuracy was tested using data from 662 cattle. The animals included bulls, steers and heifers, and comprised of Holstein–Friesian, early- and late-maturing breeds × Holstein–Friesian, early-maturing × early-maturing, late-maturing × early-maturing and genotypes with 0.75 or greater late-maturing ancestry. Bulls, heifers and steers were slaughtered at 15, 20 and 24 months of age, respectively. The diet offered before slaughter includes grass silage only, grass or maize silage plus supplementary concentrates, or concentrates offered ad libitum plus 1 kg of roughage dry matter per head daily. Following the slaughter, carcasses were classified mechanically for conformation and fatness (scale 1 to 15), and the right side of each carcass was dissected into meat, fat and bone. Carcass conformation score ranged from 4.7 to 14.4, 5.4 to 10.9 and 2.0 to 12.0 for bulls, heifers and steers, respectively; the corresponding ranges for fat score were 2.7 to 11.5, 3.2 to 11.3 and 2.8 to 13.3. Prediction equations for carcass meat, fat and bone proportions were developed using multiple regression, with carcass conformation and fat score both included as continuous independent variables. In a separate series of analyses, the independent variable in the model was the proportion of the trait under investigation (meat, fat or bone) in the hindquarter. In both analyses, interactions between the independent variables and gender were tested. The predictive ability of the developed equations was assed using cross-validation on all 662 animals. Carcass classification scores accounted for 0.73, 0.67 and 0.71 of the total variation in carcass meat, fat and bone proportions, respectively, across all 662 animals. The corresponding values using hindquarter meat, fat and bone in the model were 0.93, 0.87 and 0.89, respectively. The bias of the prediction equations when applied across all animals was not different from zero, but bias did exist among some of the genotypes of animals present. In conclusion, carcass classification scores and hindquarter composition are accurate and efficient predictors of carcass meat, fat and bone proportions.