What Predicts Achievement in Higher Education? A Review of Meta-Analyses

• In Event: Meta-Analytic Methods and Concepts at Crossroads: Relations of Student and Context Variables With Meaningful Outcomes

Abstract

Effective higher education brings competence, financial, career, health, and other benefits for individuals and countries (e.g., Groot & Maassen van den Brink, 2007). A key question in the design of higher education concerns the sources of students’ academic achievement: Which characteristics of students, teachers and instruction are associated with higher learning outcomes than others? Meta-analyses provide answers to this question by combining effect sizes obtained in multiple studies.

We present a systematic review of meta-analyses on correlates of achievement in higher education. Several such reviews have been published for K-12 education or for education in general but there are no reviews of meta-analyses on achievement in higher education. Thus, it is unclear how different the results are for higher education as compared to other educational levels and little is known about possible moderator variables.

Our thorough literature search [database: PsycINFO in April 2015 with search string (achievement or grades or competence or performance or learning or gpa) and ("higher education" or college or university or tertiary); plus an additional scan of relevant journals, books etc.] resulted in 110 articles. 79 of them were excluded due to various exclusion criteria (no meta-analysis, no inclusion of achievement or students in higher education, focus in a single subject, a subgroup of students, a specific country, or a single test). When the same correlate of achievement was examined in several meta-analyses, we only included the effect size based on the highest number of original studies, which was usually also the most-recent meta-analysis. The remaining 45 meta-analyses investigated 42 instructional variables and 63 student variables (overall, 105 variables) in more than 3,397 effect sizes from more than 2,433,189 participants.
The average effect sizes d reported by the meta-analyses ranged from d=-0.52 to 1.91. For 12% of the variables there were no effects (|d|<.11), for 35% small effects (0.11≤|d|<0.35), for 37% medium effects (0.35≤|d|<0.65), and for 15% large effects (|d|≥0.65).

On average, effect sizes for instructional variables were largest for stimulating social interaction (e.g., asking open-ended questions), followed by stimulating meaningful learning (e.g., clear course objectives and requirements), assessment (e.g., student peer- or self-assessment), presentation of content by teacher (e.g., clarity and understandableness), technology (e.g., online learning), and finally, extra-curricular trainings (e.g., academic skills trainings).

On average, effect sizes for student variables were largest for intelligence and prior achievement (e.g., college admission test results), followed by self-regulated learning strategies (e.g., effort regulation), motivation (e.g., self-efficacy), personality (e.g., conscientiousness), and finally, context (e.g., SES).

Findings revealed several moderators (e.g., course level “introductory vs. advanced”). For 19 variables, we compared results over different educational levels (primary, secondary, tertiary). In part, effects sizes varied greatly by educational level but there were also variables with comparable effect sizes indicating that some learning mechanisms are valid in all educational levels (e.g., advanced organizers).

Our findings can inform instructors in tertiary education about the most efficient instructional methods (e.g., course organization or encouragement of questions and discussion) and can also inform researchers about fruitful areas of further study.

Authors

• Bianca A Simonsmeier, University of Trier

• Franzis Preckel, University of Trier

• Michael Schneider, University of Trier