Gender Differences in Willingness to Guess

Using test conditions that resemble the SAT, the researcher found that women are more likely to skip questions than men, which negatively impacts their overall test score.

• When there is no penalty for a wrong answer, all male and female test-takers answer every question, regardless of whether or not the task is framed as an SAT.
• When there is a penalty for wrong answers, women answer fewer questions than men. On average, when ¼ of a point is deducted for a wrong answer, women skip 3.68 questions while men skip 2 questions. Because women skip more questions, they receive lower scores than their male counterparts who were more willing to guess.
• When the task is framed as a SAT, both men and women skip significantly fewer questions than when the task is unframed. However, women still skip significantly more questions than men (2.035 and 1.063, respectively).
• Knowledge of material, levels of confidence, and risk aversion as measured by the experiment do not fully explain the difference between men and women’s tendencies to skip questions. When men and women are grouped according to risk preference, there are still gender differences in willingness to guess within these groups. The magnitude of the gender gap in skipped questions is similar for both SAT-framed and neutrally-framed versions of the test, suggesting that the salience of a high pressure environment is also not what drives the gender gap.

This study was comprised of four treatments with each subject participating in one treatment. The researchers selected 20 questions from official College Board practice tests for the SAT II Subject Tests in World and U.S. History. In Part 1 of the experiment, subjects answered questions in a setting similar to that of a standardized test, with the option to skip as many questions as they would like. Subjects received 1 point for every correct answer and 0 points for any questions they skipped. Researchers varied the penalty imposed for a wrong answer by either deducting 0 points or ¼ of a point. In Part 2, subjects decided whether or not to accept each of 20 gambles based upon a drawing of random numbers between 1 and 100. Part 3 measured knowledge of the material and level of confidence, because each subject had to answer to the same questions from Part 1 while also estimating the likelihood that his/her answer was correct. Finally, researchers explored how framing the task influenced respondents’ strategies by informing subjects that questions were taken from actual SAT II tests. Across all treatment groups, 406 men and women participated in the study.