Modelling patterns of improvement over time: value added trends in English secondary school performance across ten cohorts

Abstract

This paper looks at underlying patterns of school effectiveness through analysing a GCSE examination data‐set over a period of ten cohorts (1993–2002) in one very large English school district. Both value added and raw score approaches were explored by employing different statistical multilevel models to examine time trends of school and pupil performance from two perspectives: consistent (linear) and inconsistent (non‐linear) school improvement. Overall, levels of measured attainment for the vast majority of the schools increased over the decade and the results indicate that one in four schools had significantly higher value added improvement trajectories (linear) than would be expected over the decade—in comparison to the average school. Those schools with a lower value added starting point in 1993 were more likely to make significant improvement. However, underlying these linear improvement trajectories it appears that only one in 16 schools managed to improve continuously for more than four years at some point over the decade in terms of value added.

Acknowledgements

This research has been made possible through the sustained support offered by Lancashire Local Education Authority—and especially the heads and the LEA officers of the assessment support team. We should like to put on record our thanks for their continuing invaluable assistance whilst stressing that the responsibility for the analyses and interpretation reported here is entirely our own. We would also like to give special thanks to colleagues who have provided very important support to the project at one time or another since the beginning, including Neil Scanlon, Rebecca Smees, Peter Mortimore and the late Desmond Nuttall.

Notes

1. The GCSE scores were calculated using the GCSE grades recoded as follows: A* = 8, A = 7, B = 6, C = 5, D = 4, E = 3, F = 2, G = 1 or other = 0. Note that prior to 1994 A* grades were not available (i.e. the GCSE score range for 1993 was 0 to 7, for 1994–2002 it was 0 to 8). Also, prior to 1998, half GCSEs were not available/included in the statistics.

2. Note that in 2004 a new GCSE scoring system was introduced, which increased the scale by a factor of approximately six points representing an improvement of one grade.

3. Pupil mobility between schools is the main reason for drop‐out from the study. The levels reported in this study compare favourably with levels achieved in other English LEAs (see Dobson & Henthorne 1999). Centre No. 212, whose SEN records were all missing for 2001, was retained in the dataset. Also the percentage of Lancashire LEA schools in the study was never less than 85% although due to local government re‐organisation in the 1990s some schools left the LEA.

4. Interestingly, the squared ‘year’ term was found to be statistically significant in Model B indicating a slight drop‐off in the raw improvement trend over the ten‐year period.