School improvement has been a considerable part of my work over the past 10 years, reviewing and synthesising evidence on the best pedagogical and other approaches, and conducting trials and other robust evaluations of school improvement interventions. This book has a policy focus, and so most of these trials and reviews are not discussed in any detail here, but they are dealt with in detail in Gorard et al (2017a), and the most recent comprehensive review of evidence on school improvement appears in Gorard et al (2016a). This chapter looks at how evidence is synthesised, what such syntheses can tell us and what school leaders can do to use such evidence successfully.

Until recently, school improvement policies have either been created on little robust evidence at all, or have simply been based on the correlates of the kinds of school effectiveness studies described in Chapter 6. This is a very misleading approach because high- and lowattaining schools might differ in a range of ways that are unrelated to whether their results are better or worse. This includes their type (such as academy or community school), their location and even their décor (potted plants for example). School improvement policies have also been influenced by highly vocal single studies, and developers acting as agents selling their wares via conferences and similar. Teacher action research is widely used but is not really research let alone action research as originally devised (Gorard, 2013a). None of these is appropriate for advising national or local policy on improving schools.

‘Hyper-analyses’

More recently, school decision-makers around the world have been increasingly influenced by ‘hyper-analyses’ of prior evidence. A systematic review of evidence is an attempt to summarise the findings of key studies in response to a specific research question (see Chapter 2 earlier). This is perfectly proper as long as full account is taken of the relative strength and clarity of each study. A meta-analysis is a particular kind of systematic review that combines the ‘effect’ sizes of all of the studies in the review to provide an aggregate ‘effect’ size, or an overall single answer to an effectiveness question. This, again, is perfectly proper as long as full account is taken of the relative strength and clarity of each study, and the studies being aggregated are fully commensurable – same interventions, age range, outcomes and so on.