The rate of profit as a random variable

This thesis is a systematic attempt to investigate two conjectures about the distribution of company rates of profit: that it should be log-normal (Gibrat 1931), and that it should be gamma distributed (Farjoun and Machover 1983). A large set of company accounts data is analysed, and partial support found for Gibrat and for a generalised version of Farjoun and Machover. The analysis includes a demonstration of different empirical distributions for different profit rate measures, a demonstration of power law tails in all measures of the profit rate, and a demonstration of size effects (differences in tail weights) in financial ratios. Annual variation in the overall skewness and kurtosis of profit rate distributions is shown to be dominated by variation in the power law tails. L-moments, a recent innovation in robust methods to deal with extreme values, are used in conjunction with a size-weighted sampling scheme to identify possible models for distributions of the profit rate at the capital level. Farjoun and Machover derive their hypothesis from a particular conception of the process of capitalist competition. A rival conception, that of Glick (1985), is tested using company accounts data and shown to be vulnerable to criticism concerning the scope of its data set, the test statistic employed, and its choice of profit rate measure. More fundamentally, it is also dependent on doubtful premises about the within-industry distribution of profit rates, as L-moment analysis demonstrates.