Two data sets representing the two principal experimental techniques used to study the β-decay of short-lived fission fragments are analyzed. The average endpoint and electron energies for nuclei with similar Q-values are the same within statistical accuracy in the two sets. However, the spread of the endpoints is considerably larger in the data set based on integral γ-energies (ANR set) compared with the set based on standard γ-spectroscopy (ENDF set). The averaged β-strength function for the ANR set follows a "universal" trend, decreasing monotonically and rapidly with increasing β-endpoint energy. On the other hand, the β-strength function in the ENDF set is less steep, and at higher excitation energy it is considerably smaller than in the ANR set. These differences imply a systematic bias in one of these two data sets.