Though structures designed according to codes do not generally fail, it is interesting to know their true probability of failure in view of the economic consequences.A simply supported uniformly loaded box beam is considered. Based on real statistical data derived from tests or adequate and realistic assumptions, the factored momentMfand the resisting momentMrare calculated using Monte Carlo simulation. In order to take into account some important phenomena for prefabricated beams, such as residual stresses, initial crookedness, and end conditions, the random variable α, which multipliesMr, and its variation coefficientVαare introduced. The α coefficient is defined as the tested capacity divided by the code predicted capacity. After interpolation, the variableZ = Mr − Mfis assigned a normal (Gaussian) frequency distribution. Then, the probability of failure is evaluated using a well tested mathematical technique.