The effectiveness of schools is sometimes distinguished from the performance of their students at the end-point of an educational cycle. The idea is that the achievement of students, and hence of schools, is to be measured in terms of the extent to which students have been educationally transformed. A school may achieve high scores in exit examinations but (the students having achieved high entry scores) may still not have transformed them to any great extent. On the other hand, a school with low entry scores may have transformed students considerably but may still end up with low exit scores compared with national norms. It has created more ‘added value’ than the first school. Furthermore, factors for which a school is not directly responsible,

like the social class, poverty or ethnic grouping of the students, may have a decisive effect on the ability of a school to transform them. Any attempt to assess the effectiveness of schools, it is maintained, needs to take account of these factors as well as the value added. Some also maintain that one needs to assess the performance of a school against its potential for achievement (Jesson and Mayston 1988). Agreement about the need to measure effectiveness has not led either to agreement as to how it should be done or even whether it can be done. The most popular approach seems to be multi-level modelling (Goldstein 1987) which assumes that the data can be fitted to a linear model. However, the approach assumes that there is a certain amount of statistical error in the data which can only be interpreted within certain bands of probability (confidence intervals). The practical upshot is that effectiveness measures for most schools show an overlap for the great majority, with a small distinguishable number of high and low achieving schools (e.g. Gray and Wilcox 1995). The possibility of measurement is further compromised when students change schools during the interval between the two measurement points. All these considerations suggest that the statistical measurement of school effectiveness is an inexact, controversial and inaccessible science of little direct use to the public. When researchers have tried to identify the factors underlying

effectiveness (e.g. Mortimore et al. 1988) they have often been accused of pointing to the obvious or commonsensical (White 1997a). School effectiveness needs to be distinguished from school improvement, which is an attempt to increase the effectiveness of schools. School improvement based on effectiveness research is also thought to be of limited value (Gray and Wilcox 1995). Nevertheless,

the desire for accountability is likely to ensure that the search for means of assessing effectiveness will continue. One possibility is through the use of inspection, which is widely used in some countries, such as the UK, or through close analysis of educational practices. The focus then moves from outcomes to processes. One problem that all approaches face is that it makes no sense to assess the effectiveness of schools unless one is clear about what they are effective for. This means that they must be assessed against their effectiveness in achieving educational aims which must first be agreed upon. Effectiveness methodology has had a great influence on govern-

ment policy in England, to the extent that schools are currently provided with a software package for gauging whether, for any given pupil, they are educating them effectively, given the background characteristics of that pupil. One may accept that some schools get their pupils to progress better than other schools with similar characteristics, but still acknowledge that our present understanding of why this is so is very limited. For example, some recent work suggests that one needs fine-grained descriptions of neighbourhoods in order to obtain an accurate picture of the constraints under which schools operate, descriptions which are far more detailed than anything currently provided by government data (Butler and Webber 2007). If this is true, then making schools accountable through the present measurement of added value may not be justifiable.