Mathematics Task-Specific Self-Efficacy and Actual Performance among Secondary School Students: Examining the Relationship between Confidence and Accuracy

This study examined the relationship between task-specific mathematics self-efficacy and actual performance among 248 senior secondary school students (111 males, 125 females, 12 missing gender data) in Kaduna state. Unlike studies correlating a global self-efficacy score with a separately administered achievement test, this design elicited students' confidence immediately before each of fifteen multiple-choice mathematics items, using validated self-efficacy (α = .848) and performance (Guttman split-half = .715) instruments, analysed with descriptive and inferential statistics. At the student level, mean confidence correlated weakly with percentage correct (Pearson r = .12, 95% CI [-.01, .25], p = .059; Spearman ρ = .18, p = .007, n = 237). A crossed-random-effects mixed logistic regression linking item-level confidence to correctness showed that higher confidence significantly predicted greater odds of a correct response (OR = 1.06, 95% CI [1.03, 1.09], p < .001), confirmed by a cluster-robust GEE model (OR = 1.06, p = .002). Calibration analysis, however, revealed pronounced positive bias: mean confidence exceeded mean accuracy by .34 (t(236) = 21.46, p < .001), with 93.7% of students overconfident and 62.9% of high-confidence responses (≥7) incorrect. Findings indicate that confidence weakly but reliably discriminates correct from incorrect responses, yet students are substantially overconfident relative to actual performance. Implications for instruction, formative assessment, and self-monitoring are discussed.