This article provides an illustration of how multilevel modelling, which is the standard method for assessing the "school effect" in school effectiveness research, can be combined with the regression-discontinuity approach. While multilevel modelling yields only estimates of relative school effects (differences between schools), the regression-discontinuity approach allows the assessment of the absolute effect of schooling (schooling versus no schooling). A combination of both approaches yields an estimate of this absolute effect and its variation between schools. This study applied the method in a secondary analysis of data derived from the Third International Mathematics and Science Study (TIMSS). (Contains 9 tables, 8 figures, and 9 notes.)