Saturday, April 05, 2008

Randomised Trials in Education - Towards A Practical Approach

Dr Andrew Leigh of the Australian National University, would like to see public policy become more evidence based, particularly in the field of education. He has suggested that randomised trials of the effects of factors such as teachers' salaries and class sizes on educational outcomes would be desirable, to give policy makers, and the rest of us, the information needed to make sound decisions on education policy. In this post I will outline a practical scheme for conducting such a randomised trial.

The trial will require at least four schools. Due to the rigorous controls that are outlined below, these schools will need to be newly constructed - I propose that a National Institute of Experimental Education (NIEE) be established, to manage these experimental schools. Specific details of school organisation are detailed below.

For a trial on the effects of class sizes and teachers salaries four distinct school regimes are needed:

School One: a school with low teacher pay and large class sizes;
School Two: a school with low teacher pay and small class sizes;
School Three: a school with high teacher pay and large class sizes;
School Four: a school with high teacher pay and small class sizes.

This requirement is dictated by the difficulty of constructing an appropriate control group for such a randomised trial; a control group of children who are not schooled at all might create problems for Australian society at large and it is difficult to identify anything that might serve as an educational placebo.

Blinding - of the kind applied in clinical trials of medical treatments - will be difficult but not impossible.

First, students in each of the schools must be kept in ignorance of the existence of the other schools in the trial, and indeed, the existence of any schools unlike the one that they are attending. This can be achieved by drafting representative random samples of children of school entry entry age from all states. The Federal Parliament will need to pass legislation conferring this power on the NIEE.

Similarly, teachers must be selected at random - without regard for experience or competence - from across the states, and the National Institute of Experimental Education will need the power to draft teachers in this way. Teachers drafted to the NIEE will be placed on strict employment contracts stipulating, in particular, that in no way are they to alert the students to the fact that an NIEE school is not a normal school. Heavy sanctions, possibly including criminal penalties, will be imposed on any NIEE teacher breaching this term of their employment contract.

To rigorously control for possible exogenous factors (such as differences in home environment), all students at NIEE schools will be boarders. A rigorous security regime to ensure their separation from the outside world will be maintained. To prevent students inadvertently learning of the existence of other schools which might differ from their own, students will have no access to radio, television, print media or the internet.

A viable first trial would consist of six years of primary education under such conditions, followed by release of half of the experimental students into the general population, where their academic performance in secondary school would be closely monitored.

The other half of the experimental population would receive a further six years of schooling under the above conditions, and their performance on standard university entrance examinations analysed. After their release into the general population, NIEE researchers would monitor their performance in tertiary education.

Longitudinal monitoring of NIEE schooled students is also indicated to establish the effects of these four distinct educational regimes in preparing students for stable employment and social relationships.