Certain recently reported statistical regularities relating to the dispersion of firms' growth rates have begun to attract attention among IO economists. These relationships take the form of power law or scaling relationships and this has led to suggestions that the underlying mechanisms which drive these relationships may have some interesting analogies with the mechanisms which drive scaling relationships in other fields. In particular, it has led to suggestions that there may be some subtle correlations among the growth rates of the different constituent businesses that comprise the firm.In this paper, I report some new empirical evidence in this area and I put forward a new candidate explanation for the relationships we observe. This candidate explanation does not rely on any correlation mechanisms; rather, it is consistent with the view that the typical firm consists of a number of (approximately) independent businesses. The size distribution of the constituent businesses within firms is modelled by reference to an analogy with the partitions of an integer.