The claims made in many statistical analyses for magnetic field decay follow from the assumption that radio luminosity declines slower than spin-down luminosity which is suggested by least-squares fits to the data. However, such fits are very sensitive to the behaviors of the fastest and slowest pulsars. If pulsar luminosities are plotted in distance-selected groups, the (radio inefficient) Crab and Vela pulsars are clearly exceptional members and the remaining pulsars are consistent with a mean fixed conversion efficiency of 10–5. Numerical simulations (and theoretical analysis) strongly suggest that the pulsar period distribution peak is caused by luminosity selection and searches limited in distance by dispersion measure, and not by some mechanism that removes old pulsars. We cannot exclude magnetic field decay with some large time constant, ≥107 years say, but only because it would make such a small difference to the expected pulsar statistics, not because it is required by the data.