Family factors associated with sixth-grade adolescents' math and science career interests

Career Development Quarterly, Sept, 2004 by Sherri L. Turner, Jason C. Steward, Richard T. Lapan

Math self-efficacy. Math self-efficacy, defined as "beliefs in one's capabilities to organize and execute a course of action necessary to produce a given attainment" (Bandura, 1997, p. 3), was measured using the abbreviated FSMA Math Confidence Scale. This scale consists of eight items. Five items are scored positively, and three are scored negatively. Sample items from this scale are the following: "I am sure that I can learn mathematics" (scored positively) and "I don't think I could do advanced mathematics" (scored negatively). For this scale, M = 3.85, SD = .70, [alpha] = .88, skewness = -.9, and kurtosis = .9

Math outcome expectations. Math outcome expectations, defined as the expectations that math will be important to one's future career, were measured using the abbreviated FSMA Usefulness of Mathematics Scale. The scale consists of four items. Three items are scored positively, and one is scored negatively. Sample items from this scale are the following: "Knowing mathematics will help me earn a living" (scored positively) and "In terms of my adult life, it is not important for me to do well in mathematics in high school" (scored negatively). For this scale, M = 3.92, SD = .88, [alpha] = .86, skewness = -.9, and kurtosis = .5.

Split-half reliability estimates in the original norming sample ranged from .86 to .93 on these five scales (Fennema & Sherman, 1976). In our sample, Cronbach's alpha reliabilities ranged from .72 to .88. Six-month stability coefficients, which were estimated for a subset of the research sample (n = 278), ranged from high to moderate (Math Confidence Scale, r = .69; Father Scale, r = .57; Mother Scale, r = .50; Usefulness of Mathematics Scale, r = .46; Mathematics as a Male Domain Scale, r = .38). We judged each of these sample statistics to be within acceptable limits, and so proceeded to test our research questions.

Results

Structural equation modeling (SEM), using the maximum likelihood method of parameter estimation, was used to explore the hypothetical relationships among the variables. One of the multivariate assumptions underlying SEM is the use of continuous, normally distributed data. However, recent studies have explored the use of categorical data in the SEM procedure. For example, Muthen and Kaplan (1985) examined the performance of a categorical variable model estimator compared with estimators for continuous variables and found that the categorical model yielded a slight underestimation of chi-square but that the parameter estimates and sampling variability were well in line with expected values. Therefore, although one of our model parameters was measured using a categorical variable (family structure), we judged that the SEM procedure would yield adequate parameter estimates of our hypothesized model.

We used a confirmatory analysis approach and therefore imposed an a priori structure on the model that is congruent with the postulates of SCCT. We left all degrees of freedom unbounded. A raw data matrix was used to estimate our observed exogenous indicators of Perceived Mother Support, Perceived Father Support, Family Structure, and Career Gender-Typing and observed endogenous factors of Math Self-Efficacy, Math Outcome Expectations, and Math and Science Career Interests. Three unobserved exogenous indicators, which were associated with error variance, were also included in the model.