This article reports data from children aged about 5 years in their 1st year of schooling. Scores on tests of both literacy and numeracy at the start and end of the year were used to derive value-added measures of progress using both residual gain analysis and multilevel (ML) modelling. Results indicated that the school was more effective in explaining pupil progress than the class; that the schools and/or classes which add the greatest value in literacy tended also to be those which added the greatest value in numeracy; and that the numbers of classes and schools which were significantly different from the sample mean depended greatly on which method was used. It is concluded that while ML modelling is preferable to residual gain analysis for research, it may not be as suitable for feedback purposes. (Contains 3 tables and figures.)