School-average achievement negatively effects on academic self-concept (ASC) and career aspirations (Big-Fish-Little-Pond-Effect, BFLPE), even though individual-student achievement has positive effects on these same outcomes. In the largest cross-national test of the BFLPE ever conducted (PISA2006; 398,411 15-year-olds; 14,353 schools, 56 countries), we apply multilevel structural equation models of mediation. We show: (a) Individual achievement had positive effects on ASC and career aspirations, but effects on career aspirations was mediated by ASC in most countries; (b) school-average achievement had negative effects on ASC and career aspirations but in most countries there was total/partial mediation of the contextual effect of school-average achievement on career intentions. The study has important methodological and policy implications for school contextual studies.
Traditionally, BFLPE studies have either applied multilevel models based on single (manifest) measures of each construct that ignore measurement error or single-level SEMs with multiple indicators that ignore the multilevel structure of the data. Recent developments of multilevel structural equation models allow researchers to simultaneously control for measurement error and sampling error and have led to the development of a general framework for multilevel mediation analysis. Here we apply and extend these evolving methodological developments the combine the advantage of SEMs and MLMs into an integrated statistical framework. Analyses were conducted with Mplus 6.1, using multiple imputaton with multiple plausible values of achievement, survey weights provided by PISA, and robust maximum likelihood estimation. Data considered here come from PISA 2006 assessed science competencies of 15-year-old students in nationally representative samples (398,411students from 14,353 schools in 56).
Substantive & Methodolotical Significance
This study is a substantive-methodological synergy, extending previous cross-cultural research on the BFLPE in five important ways: (1) our sample, taken from PISA 2006, included more countries and a more diverse sample of countries than any other previous BFLPE study; (2) ours is the first PISA study (and one of the few BFLPE studies) to focus on ACH and ASC in science, a subject area with high-policy relevance for developed and developing countries; (3) we showed that the BFLPE generalized to career aspirations in science and that these effects were mediated by ASC; (4) we used multilevel structural equation models, cutting-edge quantitative methods, to implement the contextual effects and the multilevel mediation models; and (5) country-level differences in the negative contextual effects of school-average achievement were small and not significantly related country-level cultural and economic development differences.