Whether small businesses create more jobs than large enterprises is a question controversial in empirical economic research. Methodological literature shows that the method used to assign enterprises to size classes is of crucial importance and can lead to biased results. In a comparison of four alternative approaches we find that allocating enterprises to size classes according to their initial size leads to a systematic overestimation of the job-creating performance of small enterprises, while an allocation using the end size of enterprises systematically overestimates the job-creating performance of large enterprises. Allocating job gains to size classes by using average firm size or a dynamic allocation method both lead to unbiased results. For quarterly data the dynamic method of allocating job changes to size classes has clear advantages compared to the other methods, as it allows to allocate job gains and job losses symmetrically to size classes, meets the criterion of additivity over time and allows to take into account job creation and job losses by start-ups and exits in a quite natural way.