Greater variance in male intelligence test scores can be explained by sex-biased test design

Abstract

The “Greater Male Variability Hypothesis” (GMVH) of mental ability is a century-old hypothesis that males have been overrepresented in elite academic, political, and economic roles because males are more variable in intellectual ability than females. Proponents of the GMVH argue that males and females have similar average intelligence, but males are over-represented at the extreme high end of the distribution because they are more variable in intelligence. The evidence for this hypothesis is mixed and consists of greater male variance in test scores on some intelligence tests - especially in test scores from the early 20th century. However, I show that greater male test score variability results when tests are designed primarily to test intelligence in males. For example, to evaluate exclusively male Army draftees in the First World War or mostly male secondary school applicants in the 1930s United Kingdom. I show that typical question selection procedures, when used to design a test for males, tend to increase the variance of male scores relative to female scores. It is often assumed that intelligence measures are independent of test design procedures, but I show that designing a test for one population leads to statistical irregularities when applied to another. This implies that males are not overrepresented in elite positions because they are more variable in baseline intelligence, but that this may be due to other factors.