Fade-Out of Educational Interventions: Statistical and Substantive Sources

What appears to be ineffectiveness of educational interventions in the long run may actually be caused by statistical artefacts in the equating of tests taken at different time points or by the nature of the skill development in the absence of targeted interventions. We use longitudinal data on the full population of public school students in Denmark to estimate central parameters in the equating of reading test scores and in a skill formation model. We compare the model's predictions to observed fade-out in a randomized controlled trial two and four years after the end of the intervention. Predicted and observed estimates consistently show that about half of the initial effect has faded out after four years. However, because of the concave nature of skill development, the treated students maintain more than 80 % of their time lead