STEREOTYPE THREAT, GENDER, AND MATH PERFORMANCE: EVIDENCE FROM THE NATIONAL ASSESSMENT OF EDUCATIONAL PROGRESS Current Draft: September 15, 2009 1 First Draft: January 11, 2008 Thomas E. Wei Harvard University Abstract Stereotype threat posits that when individuals are primed about a fixed characteristic of theirs that is negatively stereotyped in relation to a task (i.e., girls cannot do math), subsequent performance on the task is adversely affected. A limitation of most stereotype threat studies is that they have been based upon small and non-representative samples in lab settings, raising concerns about their external validity. I use micro data from the 1978 to 1999 National Assessment of Educational Progress (NAEP) – a large and representative field assessment of 9, 13, and 17 year-old U.S. children – where through a design quirk, students are randomly assigned to test blocks, some of which include gender prime questions while others do not. I exploit this natural experiment by comparing the gender gap in math test scores of students receiving gender primes to those with placebo questions. I find little evidence of negative stereotype threat and strong evidence of stereotype reactance with girls performing better relative to boys for some gender primes. The impact of gender primes on math test performance appears to be sensitive to the exact phrasing used in the primes. 1 Comments welcome and can be directed to [email protected]. I thank Joan Stoeckel from the Educational Testing Service for valuable assistance regarding the NAEP data, as well as Susan Dynarski, Richard Freeman, Brian Jacob, Lawrence Katz, Todd Pittinsky, and seminar participants at Harvard for their support and invaluable comments on earlier drafts. All remaining errors are my own.- 1 -Research in psychology and behavioral economics emphasizes the importance of salience on actions and judgments (cf. Kahneman and Tversky, 1984). For instance, recency bias posits that because of mental availability, individuals overestimate the probability of a plane crash after they have seen one reported in the news. Increasing awareness of gender stereotypes in math may similarly prime students’ perception and performance. More concretely, societal norms may implicitly or explicitly suggest that math is more suitable for boys. This norm, reinforced through peer interactions, parental attitudes, teacher expectations, and other priming channels, is hypothesized to negatively affect girls’ subsequent performance and ultimately reinforce their own stereotypes and behaviors, as well as society’s.2 The result is a self-fulfilling prophecy. This phenomenon is known as “stereotype threat,” or “being at risk of confirming, as self-characteristic, a negative stereotype about one’s group” (Steele and Aronson, 1995). Economists have explored many hypotheses for the observed gender gap in labor market outcomes (Goldin, 1994; Altonji and Blank, 1998; Blau and Kahn, 2000), the most common being occupational segregation (MacPherson and Hirsch, 1995). Divergent incentives to invest in human capital as a result of discrimination (Royalty, 1996) and differences in the division of labor (Becker, 1985) are often used to explain why women and men select into different jobs. However, stereotypes about what women can and should do in the labor market can influence whether employers choose to statistically discriminate and whether women choose to train in certain areas. This is particularly relevant in the male-dominated fields of math and science, which generally require more education and pay higher wages. Some psychology experiments 2 For example, recent work by economists suggests that teachers’ subjective assessments of students are influenced by their own biases (cf. Lavy, 2004; Burgess and Greaves, 2009; Hanna and Linden, 2009; Mechtenberg, 2009). Psychologists have also argued that stereotypes can influence women’s desire to study math and science, which subsequently affects their career decisions (cf. Spencer, Steele, and Quinn, 1999). However, further empirical research on stereotype threat and the other theories to account for the gender gap in math is warranted, especially given new evidence that the gap arises at a much earlier age than previously thought (Fryer and Levitt, 2008).- 2 -have shown that stereotype threat can reduce women’s desire to study quantitative fields (Davies et al, 2002; Gupta and Bhawe, 2007). Economists recently began experimentally documenting gender differences in preferences (Croson and Gneezy, 2009), such as for competition (Gneezy, Niederle, and Rustichini, 2003; Niederle and Vesterlund, 2007), but have done less research on stereotype threat and its possible impact on preferences.3 Most of the work on stereotype threat has been conducted in lab settings with small and highly selected samples, potentially raising concerns about external validity. On the other hand, the small number of observational studies with large and representative samples typically lack exogenous variation in exposure to primes, which leads to concerns about drawing causal inferences since other confounding factors are likely to be correlated with exposure to primes. I address the limitations of these approaches by analyzing the micro data from the long-term trend National Assessment of Educational Progress (NAEP), a large and representative test of 9, 13, and 17 year-old U.S. students spanning 22 years. Students are randomly assigned to test blocks, some of which include gender primes while others do not. Comparing the gender gap in math test scores of students receiving gender primes to those receiving placebo questions reveals any possible priming effects on gender gaps in math. Thus, these data provide an unusual natural field experiment, which sheds light on the relevance of stereotype threat outside of the lab. I find little evidence of traditional stereotype threat, but instead find null effects with some primes and reactance effects with other primes, whereby girls perform significantly better relative to boys. In the latter case, the hypothesized mechanism is that explicitly increasing the salience of gender stereotypes induces additional effort from girls to “prove everyone wrong.” This accords with evidence that girls more intensely reject gender-math stereotypes than boys in the NAEP as well as with prior studies documenting stereotype