Can project-based learning improve students’ mathematics learning outcomes: a three-level meta-analysis

Against the backdrop of the worldwide shift toward competency-driven basic education reform, project-based learning (PBL) is widely recognized for its student-centered design and authentic problem-solving approach. Empirical evidence on the effectiveness of Mathematics PBL is highly varied; traditional meta-analysis often fails to properly address nested data structures. This study conducted a three-level meta-analysis spanning the past 20 years, a total of 35 eligible studies were included. In the primary meta-analytic model, two statistically outlying studies were identified and excluded based on Cook’s distance. Ultimately, 33 studies with 84 effect sizes were incorporated into the main effect analysis.to comprehensively quantify PBL’s impact on students’ mathematical achievement. The results indicated a significant overall effect size for PBL (Cohen’s d  = 0.529, 95% CI [0.38, 0.68]), but publication bias analyses suggested potential overestimation (trim-and-fill corrected estimate = 0.275); therefore, findings should be interpreted cautiously. Moderator analyses revealed that cultural background and intervention duration significantly influenced outcomes. Notably, PBL demonstrated strong effectiveness within a 1-to-8-week timeframe. Non-cognitive abilities showed a numerically larger effect than cognitive abilities, but the difference was not statistically significant ( p  = 0.083). No significant moderators were found for educational level, class size, group size, mathematics knowledge domain, publication type, or the number of technology types used, suggesting that PBL is broadly generalizable across various mathematics teaching contexts. Given the presence of publication bias, these results provide moderate rather than robust support for the localized application of mathematics PBL. Systematic review registration https://www.crd.york.ac.uk/PROSPERO/view/CRD420251105265 , identifier CRD420251105265.