What's in a Trend? A Comment on Gray, Goldstein and Thomas (2001), ‘Predicting the Future: The role of past performance in determining trends in institutional effectiveness at A level’

Gray et al. (British Educational Research Journal, 27, 2001, p. 391) present evidence that ‘predicting future “value added” performance from past trends is unreliable’. This note argues that the authors treat the notion of trend in an unproblematic manner; namely, as deterministic time trends. Yet the movements in the data identified by the authors as school improvement trends may not be generated by the deterministic process they assume. Instead, these upward and downward movements may reflect a purely stochastic (i.e. random) process or, at least, a process that allows for both stochastic and deterministic elements. In either case, once we treat the notion of trend as problematic, we have to allow for the possibility of a purely random element in school improvement data that is greater than even the authors suggest. Accordingly, the implications of this discussion enhance rather than detract from the authors' findings on the weakness of improvement trajectories and the unreliability of using them to predict future results.