How Institutional Context Affects Degree Production and Student Aspirations in STEM Kevin Eagan, Ph.D. University of California, Los Angeles January 28, 2010
The Problem Higher institutional graduation rates in non-STEM fields relative to STEM fields Push toward accountability standards Relative homogeneity among researchers in science, technology, engineering, and mathematics (STEM) careers Research puts onus on students
Research Questions Institutions’ STEM Degree Production What institutional characteristics affect the production of undergraduate STEM degrees? What factors contribute to institutions’ efficiency at producing undergraduate STEM degrees? Students’ Degree Aspirations What student characteristics predict student degree aspirations at the end of four years of college? What institutional characteristics predict student degree aspirations at the end of four years of college? Do these student and institutional variables have differential effects across specific groups of students?
Theory and Literature: Economic Production Functions
Theory and Literature: Degree Aspirations Status attainment theory (Blau & Duncan, 1967; Sewell, Haller, & Portes, 1969) College student socialization (Weidman, 1989) Primary limitations of degree aspiration studies: operationalization of the dependent variable, under-development of institutional problem, and analytic methods
Methods: Production Function Data: Integrated Postsecondary Educational Data System (IPEDS) Sample: 4-year public and private non- profit bachelor’s degree granting institutions (N=1,428) across 4 years Subsample for additional analyses: 197 public and private, non-profit four-year institutions
Methods: Production Function Dependent Variables DV1: total undergraduate STEM degrees produced each year DV2 (created from first analysis): production efficiency score for each institution-year case Independent variables: Production function: human capital, labor, financial capital Efficiency analysis: selectivity, structural characteristics, climate elements
Methods: Production Function Analyses Stochastic frontier analysis ○ Decomposes error term into two components: randomly distributed error and non-randomly distributed error (inefficiency) ○ More robust than OLS regression ○ Distinct from data envelopment analysis, as SFA accounts for external shocks to the firm Hierarchical Linear Modeling ○ Analyze the relative contributors to production efficiency
Production Function Results Decreasing returns to scale Average efficiency score: 40% Efficiency Negatively affected by: % PT faculty, % URM students Positively affected by: % PT students, % STEM students, selectivity
Methods: Degree Aspirations Data Students ○ 2004 Freshman Survey ○ 2008 College Senior Survey ○ National Student Clearinghouse Institutions ○ IPEDS ○ Student-level aggregates ○ SFA model (efficiency score) Sample: 5,876 students across 197 institutions
Methods: Degree Aspirations Dependent variable: recoded degree aspirations into five categories Independent variables Background characteristics (2004) Pre-College characteristics (2004) Connections to peers and faculty (2008) Campus involvement (2008) Campus climate perceptions (2008) Institutional characteristics ( ) ○ Structural characteristics ○ Aggregated climate elements ○ Production efficiency scores from SFA model
Methods: Degree Aspirations Analyses Response weights Multinomial hierarchical generalized linear modeling ○ Categorical, non-ranked outcome ○ Nested data (students within institutions) ○ Model building
Results: Degree Aspirations – Institutional Predictors Master’s Degree M.D.J.D.Ph.D. Control: Private ++++ HBCU ++ Agg. faculty support +++ Agg. cross-racial interactions +++ Production efficiency NS
Results: Degree Aspirations – Individual Predictors Master’s Degree M.D.J.D.Ph.D. Undergraduate research participation ++ Grad school prep. program +++ Faculty support + College GPA ++++ Find a cure to a health problem + Make a theoretical contribution to science + Be well-off financially +-
Limitations Secondary data analysis Limited controls for institutional (student and faculty) quality in SFA model Timeframe of surveys limits causal inferences Low longitudinal response rate
Discussion Limitation of applying economic theory and efficiency to higher education Balancing democratic mission of higher education with political and economic realities Student preparation Faculty employment Program duplication and coordination Engagement with diversity
Implications for Research Institutional data Utility of efficiency scores in higher education Self-selection bias and causality