Long-Term Consequences of Parents' Gender-related Stereotype about Mathematics Success: A Propensity Score Matching Approach

• In Event: Learning outcome for marginalized students

Proposal

         The purpose of this study is to understand the long-term causal effect of parents’ gender-related stereotype about mathematics success on high school students’ mathematical success. Using data from a longitudinal study, this study further investigates those consequences on different gender groups through differentiating male and female students.

         This study includes three major parts: 1) a literature review of studies examining the impact of parents’ gender-related stereotype about mathematics and science success on students’ learning outcome and experience; 2) a propensity score matching to account for differences in background characteristics and other characteristics related to mathematics and science learning between parents who have stereotype and their counterparts; and 3) a series of regression models after propensity score matching to show whether parents’ stereotype predicts high school students’ achievement, dropout, degree program level, and major selection.

         This study applies Stuart's (2010) four-step model. First, a propensity score was estimated to better isolate the effect of family background and parental involvement on students’ math motivation and achievement. Second, cases with parents holding stereotype were matched to their counterparts based on these propensity scores. Third, we used multiple imputation to account for missing data as the missing increased with the research went by. Finally, regression models on the matched sample were used to identify the consequences of parents’ math stereotype on math motivation and achievement for females and males separately.

         The data used in this research is the High School Longitudinal Study, 2009-2013 (short for HSLS:09) collected by the United States Department of Education. The survey HSLS:09 first took place in the 2009-2010 school year (base year). Students from over 900 public and private high school currently in 9th grade were stratified-randomly selected. The second wave took place in the spring of 2010 when most of the participants were in their 11th grade. The third wave was launched in 2013, in order to know the cohort’s postsecondary plans and decisions. The sample size is 23,503 in total for students and missing values increase as the data collection processes. In every wave, parents, teachers and school principals were also asked to fill out questionnaires. Therefore, the entire data set includes survey results from four parties: student, parent, teacher, and school.

         This study includes three waves of data to serve propensity score matching and regression models with adjusted data. For the propensity score matching, this study will use data from the first wave of the survey. This data majorly contains the demographic information of students and their families, students' math grades at 9th grade, parents' negative attitude, and parental involvement of math-related activities. The regression models, on the other hand, will use data from the second and the last wave. Variables included as responses are about students’ math achievement (math credits earned and math GPA in 12th grade) and math motivation. The last wave, however, does not contain information about students’ math motivation, so the math motivation variables used in regressions are from the second wave.

         Two obvious difference is found in parents’ helping build or fix things and parents’ stereotype about math success. Parents of male students reported a higher percentage of helping build or fix things than parents of female students. Also, as noted in the previous section, more parents of male students believe that males are better than females in math than parents of female students. Even though for parents who have girls, there were still about 26% of them believed girls are not as good as boys in math.

         According to Austin (2009) and Dietrich and Lichtenberger (2015), 10% or less standardized mean differences are considered good while standardized mean differences greater than 20% are considered large. Based on these thresholds, the matching done in this research demonstrates an excellent balance and demonstrates the success of the full matching used in equating parents who had the stereotype and who did not on observed characteristics.

         We found statistically significant effects in both female and male samples after using propensity score matching. Results show that female students are disadvantaged by their parents’ stereotype about math success. And male students, on the contrary, have significantly higher math motivation if their parents have a stereotype that males are better in math than females.

Author

• Wenjuan Sang, Indiana University