Psychometric problems with the method of correlated vectors applied to item scores (including some nonsensical results)

•The method of correlated vectors (MCV) is often used to study group differences in performance on cognitive test items•It is shown here that this method does not convey much information about the role of g in the ethnic groups being compared•MCV was applied to scenarios in which cognitive tests were replaced by measures of mood or personality in one of the two groups in the analysis•MCV incorrectly suggested group differences in g (a Jensen Effect) even though group differences could not reasonably be on g

Spearman's hypothesis stating that ethnic group differences on cognitive tests are most pronounced on the most highly g loaded tests has been commonly tested with Jensen's method of correlated vectors (MCV). This paper illustrates and explains why MCV applied to item-level data does not provide a test of measurement invariance and fails to provide accurate information about the role of g in group differences in test scores. I focus on studies that applied MCV to study group differences on items of Raven's Standard Progressive Matrices (SPM). In an empirical illustration of the psychometric problems with this method, I show that MCV applied to 60 SPM items incorrectly yields support for Spearman's hypothesis (so-called Jensen Effects suggesting that the group difference is on g) even when the items in the second group are not from the SPM but rather from a test composed of 60 items measuring either anxiety and anger or the big five personality traits. This shows that MCV applied to item level data does not accurately reflect the degree to which item bias or g plays a role in group differences. I conclude that MCV applied to items lacks both sensitivity and specificity.