Reconsidering gender differences in intelligence: A meta-analysis of gender differences in IQ variability

• In Event: Gender Stereotypes and STEM Motivation: Experimental, Longitudinal, Meta-analytic, and Intervention Research Across the Lifespan

Abstract

The variability hypothesis posits that males tend to be more variable than females in terms of both biological and psychological traits (Shields, 1982; Reinhold & Engqvist, 2013), including general intelligence (Johnson, Carothers, & Deary, 2008). Because of widespread acceptance of this hypothesis, men are thought to be more likely than women to have very high levels of intelligence (Lehre et al., 2008). This hypothesized greater male variability in intelligence has been exploited as one explanation for gender disparities in science, technology, engineering, and math (STEM) fields. More specifically, some have argued that men are overrepresented in the STEM professoriate because they are more likely to possess high levels of natural intelligence, a trait considered requisite for success in male-dominated academic disciplines (Leslie, Cimpian, et al., 2015). Despite the implications of these arguments, efforts to investigate the validity of the variability hypothesis have been limited. In the present research, we examined the empirical evidence for greater male variability in IQ scores, and the related conjecture that gender differences in IQ account for gender disparities in STEM, through meta-analysis.

Following a systematic review of the IQ literature, we meta-analyzed 634 effect sizes to estimate (1) the mean gender difference in IQ and (2) the gender difference in IQ variability. Samples were drawn from more than 125,000 children and adults (mean ages = 2-76 years), born between 1880 and 2010, and residing in 36 countries (55% of samples were North American/European). This sample-level diversity allowed us to consider the effects of multiple cultural/contextual variables on reported effect sizes.

Across all samples, there was a very small gender difference in mean IQ scores (d = -0.05, 95% CI[0.00, -0.09], p < .05), equating to a male advantage of less than 1 IQ point on average (Figure 1). This very small male advantage was largely present across countries (Figure 2) but decreased with historical time, with female and male IQ scores reaching parity by the end of the twentieth century. When controlling for the mean difference, there was no gender difference in IQ variability (CVR = 0.99, 95% CI[0.97, 1.00], p = .118), with the meta-analytic CVR indicating that female IQ scores were 99% as variable as male IQ scores (Figure 1). Moderator analyses revealed regional differences in the magnitude of the CVR (B = -0.04, p = .017; Figure 2). Male IQs were slightly but significantly more variable than female IQs in Western cultures, namely Australia, Europe, and the Americas (CVR = 0.97, 95% CI[0.96, 0.99], p = .013). In contrast, there was no significant gender difference in IQ variability in Eastern cultures, including East Asia and the Middle East (CVR = 1.02, 95% CI[0.99, 1.04], p = .274).

These findings indicate that gender differences in general intelligence are very small in magnitude and dependent on cultural context, contradicting the popular notion that men are more likely than women to have high levels of intelligence. Therefore, the persistent underrepresentation of women in STEM cannot be attributed solely to gender differences in intellectual ability.

Authors

• Jillian Lauer, University of Cambridge
  Presenting Author

• Andrei Cimpian, New York University
  Non-Presenting Author