Class size and student outcomes in Europe : executive summary in English

Class size reductions are popular with parents, teachers and politicians. In recent years various contributors to the public debate have increasingly expressed the opinion that class size does not matter. For example, the popular science author Gladwell (2013) uses class size reduction as an example of a "thing we are convinced is such a big advantage [but] might not be such an advantage at all" (as cited in Schanzenbach, 2014). Likewise, in a recent report the OECD concludes that "overall, evidence of the effect of differences in class size on student performance is weak" (OECD, 2016, p.349). These strong statements are made without (OECD) or with very selective (Gladwell) reference to evidence, let alone a careful assessment of the evidence. The same can often be said of advocates of class size reductions. Our systematic review of the empirical studies on the effects of class size on student outcomes in Europe shows mixed evidence. Some credible studies report substantial beneficial effects of smaller class sizes, whereas other equally credible studies find effects that are rather small or not significantly different from zero. While some studies show that smaller classes did indeed offer sustained benefits, using this evidence for prescriptive policy purposes typically requires an extrapolation to a different population. Such extrapolations require a solid understanding of how population characteristics, incentives and constraints enter the production function and mediate class-size effects. While there are some studies that investigate such mediating factors, the evidence falls short in providing definitive answers. Based on the current evidence it is therefore hard to give an unqualified recommendation about how and when to use class size policies to improve student outcomes. Additional research is therefore necessary to provide better answers to the questions whether, why, when and for whom class size matters. These questions must be answered through research design, and the evidence must be causal. Field experiments are typically considered the "gold standard". While the single large-scale randomised experiment regarding class size (Project STAR) has produced many insights, they have their own threats to internal and external validity (Heckman and Smith, 1995). We therefore believe it is important to collect evidence from a variety of causal designs. These can be nonexperiment designs derived from maximum class-size rules, or alternative experimental designs (Rockoff, 2009). Finally it is worth pointing out that the notion of a class-size effect depends on traditional modes of teaching and learning where we can think of class size as a well-defined and policy-relevant proxy for inputs in schools' production function. While today's schools are remarkably similar to the schools of our grandparents, technology and pedagogical innovations may change the nature of classrooms and thereby the relationship between class size and inputs. Estimates of class-size effects are likely to have expiration dates when the production function is changing.