In a 1990 article in the same journal, Rushton reported significant racial differences in crime rates, based on data from international police records. However, his statistical method, the analysis of variance, did not assess the size of the trends that he considered significant. The re-analysis of his data revealed that the average correlation coefficient was 0.24, suggestion that less than 6 percent of the variance is shared. The trends were also mostly weak and inconsistent, with 70.8 percent of the underlying coefficients classified as low or nonsignificant. Thus, Rushton's own data suggest that relying on race as a predictor of crime in individual cases would result in a 99.9 percent rate of false positives, an absurdly high rate. Tables, note, and 12 references