This study examines whether middle leadership (ML) and digital instructional leadership (DIL) are associated with teachers' innovative behaviour (TIB) and whether teacher self-efficacy (TSE) has an indirect association with TIB through these leadership variables in Indonesian vocational schools.
A cross-sectional survey of 350 teachers was conducted using SmartPLS 4 with the path-weighting algorithm and 5,000 bias-corrected and accelerated (BCa) bootstrap resamples. This study evaluated measurement quality, structural paths, indirect effects, explanatory power, partial least squares prediction (PLSpredict), and cross-validated predictive ability test (CVPAT).
ML and DIL were positively associated with TSE (ß = 0.513 and 0.412) and TIB (ß = 0.335 and 0.245); TSE was associated with TIB (ß = 0.445; all p < 0.001). Indirect associations through TSE were significant for ML (ß = 0.228, 95% BCa CI [0.173,0.288]) and DIL (ß = 0.183, [0.144,0.233]). R2 was 0.712 for TSE and 0.874 for TIB. All endogenous indicators had positive Q2predict, lower partial least squares structural equation modelling (PLS-SEM) errors than the linear benchmark, and significant CVPAT advantages. Extremely high reliability, elevated indicator variance inflation factors (VIFs), and heterotrait–monotrait (HTMT) = 0.899 indicate substantial measurement overlap. A domain-preserving reduced-item sensitivity analysis lowered the TSE–TIB HTMT to.873 and attenuated, but did not eliminate, the TSE→TIB and indirect associations.
The single-wave, single-source design does not support causal mediation. The recruitment flow, school cluster, and response rate information were unavailable. The study also relied on adapted response scales and a global reflective measurement specification.
Higher education providers and school systems may pilot integrated middle-leadership, digital-instructional, and efficacy-building professional development, with independent outcome evaluation.
The study evaluates two leadership domains and combines explanatory and out-of-sample predictive assessments in an under-researched vocational school context.
