Time Spent on Homework and Academic Achievement: A Meta-analysis Study Related to Results of TIMSS

488 independent findings from 74 countries, 429,970 students (225,430 in grade 4 and 204,540 in grade 8) across TIMSS cycles 1999-2015

Fourth and eighth graders in 74 countries

Psicología Educativa 28(1), 13-21; read at abstract depth plus verbatim Results sentences from the publisher's HTML - a clean PDF could not be retrieved, so this is recorded as abstract. Random-effects meta-analysis of the association between reported homework-time bands and achievement on TIMSS, an INDEPENDENT standardized international assessment. The estimates are between-student comparisons of time bands, so causally inert - grade C follows from the design, not the execution.

The largest confirmation of the age gradient on independent standardized tests. Pooled d = 0.186, statistically significant. By grade: fourth grade d = -0.057, eighth grade d = +0.256, with the difference statistically significant. By subject: mathematics d = 0.358 versus science d = -0.009 - a subject gap larger than the age gap, and one that lines up with Eren and Henderson's finding that assigned homework moves maths and nothing else. A cultural moderator: vertical-collectivist countries d = 0.258 against horizontal-individualist d = 0.047. THE MOST UNSETTLING RESULT IS THE TIME TREND: the effect runs -0.270 (1999), 0.036 (2003), 0.251 (2007), 0.439 (2011), 0.525 (2015). A homework effect that swings from strongly negative to strongly positive over sixteen years, measured on the same instrument in the same countries, is not behaving like a stable causal parameter; it is behaving like a marker of something else that changed - curricular intensity, tracking, or who reports homework time.

High. The unit of analysis is a between-student comparison of self- or teacher-reported homework-time bands with no within-family or within-student identification anywhere in the pooled studies. Time spent indexes ability, conscientiousness and family supervision, all substantially heritable and all independently predictive of TIMSS scores.

Replicates the age gradient established by Cooper's two syntheses on a disjoint corpus, a different instrument and 74 countries: elementary near or below zero, secondary around +0.25. That is now three independent syntheses agreeing on the gradient. The temporal instability is unreplicated and, as far as this archive can tell, unexplained.