Cyclical changes of homicide rates: a reanalysis of Brearley's 1932 data

The objective of this analysis is to explore the possibility that Brearley's failure to detect a coherent pattern of seasonal variation in his well-known study of homicide in the United States was due to a lack of statistical refinement. The original data on homicides in the United States between 1923 and 1928 (N = 51,798) were reanalyzed using spectral analysis. The results show a significant seasonal dependency with a peak in the homicide rate in August. The seasonal dependency explains between 23% and 30% of the total homicide variance. The importance of this finding is the degree of explained variance and the location of the peak, which is remarkably consistent with more contemporary studies using similar methods despite substantial sociodemographic changes, suggesting an endogenous etiology.