There is, then, a lot at stake, especially given the huge financial implications of reducing
class sizes and deploying more staff. But it was our conclusion (Blatchford, Goldstein
and Mortimore, 1998) that there was little research on class size in the UK and little
reliable information about the effects of class size differences on pupils’ educational
attainments, still less on whether the number of children in the classroom affected
processes such as teaching, learning and peer relations. There was also little information
available in the UK on whether extra support staff in classes are beneficial, and
whether it is more important in larger classes. Our aim was therefore to contribute in a
substantial way to the class size debate by conducting a sustained enquiry into the
educational consequences of class size and pupil adult ratio differences.
Our interest in the subject began in the early 1990s when Peter Mortimore, who was then
the Director of the Institute of Education, and Peter Blatchford, were asked by the
National Commission on Education to write a briefing document on the educational
effects of class size differences (Mortimore and Blatchford, 1993). A main conclusion
was that there was a wide gap between professional experience - i.e., that other things
being equal smaller classes are better educationally - and research evidence, which was
not clear. A main impetus for the research reported in this paper was an attempt to bridge
this gap.
There were two aims. The first was to study the effect of class size differences on pupils’
educational attainments. Our study was conceived in the context of our reviews of
previous research (Blatchford, Goldstein and Mortimore, 1998; Goldstein and
Blatchford, 1998) and our identification of limitations in research designs. In the UK,
correlational studies that used cross sectional designs, that is, examined relationships
between class size and children’s achievements at one point in time, are difficult to
interpret because of uncertainties over whether other factors (e.g., non-random allocation
of pupils to classes) might confound the results. In contrast, the strength of the research
design used in the STAR project is that by randomly allocating teachers and pupils to
classes, it is in theory easier to draw unambiguous conclusions concerning the causal
role of class size. The STAR project is an impressive and bold study, but we have
identified several theoretical and practical difficulties with experimental designs like the
STAR project (Goldstein and Blatchford, 1998, see also Grissmer, 1999, Hanushek,
1999, Mitchell et al, 1991, Prais, 1996), that suggested to us the value in using an
alternative non- experimental design. It was our belief that it would be more productive
to research class size effects by seeking to capture the complex world of education rather
than control one feature of it. It would, in other words, be more valid to seek better
understanding of the effects of class size differences by measuring and examining
relationships between class size and other factors as they occurred naturally in schools,
and to make adjustments for possibly relevant factors such as family income. Most
importantly, in order to overcome problems of cross-sectional studies, the design would
need to be longitudinal, in order to control for pupils’ prior attainments. Previous
research has shown that the strongest effect of class size differences is on the youngest
children in school and so the design would need to include school entry measures. (The
STAR project did not obtain school entry attainments, and so cannot verify the supposed
random assignment of pupils to classes.) We felt this type of design would provide a
valuable source of evidence for policy because it would be more authentic. We could,
for example, examine class size effects across the full range of class sizes, not just a few
selected sizes. This could be important for policy recommendations, for example, if there
are certain class sizes, or class sizes below or above a certain number, which have