The study examines four methods to allocate changes in employment to the various employment size categories. For annual data, the reference parameter should be the average size of the enterprise or a dynamic allocation procedure. Due to regression bias, job creation by smaller enterprises is overestimated when figures from the previous year or from start-ups are used. Using the final enterprise size overestimates job creation by large-scale businesses. Quarterly data are best served by a dynamic allocation to size categories. This method of allocation, used as its official method by the American Bureau for Labor Statistics, is the only one that allows allocating new formations and closures to the size categories.