This article generalizes and extends the kernel block bootstrap (KBB) method of Parente and Smith (2018, 2021) to provide a comprehensive treatment of its use for GMM estimation and inference in time-series models formulated in terms of moment conditions. KBB procedures that employ bootstrap distributions with generalised empirical likelihood implied probabilities as probability mass points are also considered. The first-order asymptotic validity of new KBB estimators and test statistics for over-identifying moments, additional moment constraints and parametric restrictions is established. Their empirical distributions may serve as practical alternative approximations to those of GMM estimators and statistics and to other bootstrap distributions in the extant literature. Simulation experiments reveal that critical values arising from the empirical distributions of some KBB test statistics are more accurate than those from standard first-order asymptotic theory.