Introduction

As the preceding chapters show, different service areas adopt verydifferent approaches towards identifying what works. On the onehand, the health sector has in general adopted a research culture inwhich it is accepted that the services provided should in principlebe exposed to rigorous scientific evaluation (see Chapter Three).Central to this culture is the notion of experimentation, usuallytaken to mean randomised controlled trials. In practice many medicalpractices have not been properly evaluated in this sense, or thestudies that have been undertaken fall some way short of providingincontrovertible guidance. Much activity remains untested throughrandomised trials (Smith, 1991). Strategic policies, such as theintroduction of ‘fundholding’ in general practice (and itssubsequent abandonment), have not hitherto received the same form ofevaluative attention as individual medical procedures (Ham et al,1995; Davies and Nutley, 1999). Yet the principle is enshrined:interventions should be tested before widespread use, andexperimentation (in the form of randomised controlled trials) liesat the apex of a hierarchy of evidence (see Chapter Three, Box3.3).

In contrast, research on effectiveness takes a different form, and isoften less visible, in other parts of the public sector. In areassuch as education, social services and criminal justice, there hasbeen considerable research activity over several decades. However,coverage is patchy, there is less consensus regarding appropriatemethodology, and there is little agreement as to how to use researchevidence to inform policy and practice (see Chapters Four to Six).Although experiments (including randomisation) have at times beenconducted in education and criminal justice, and some too in socialcare, their use is highly contested for a range of ontological,epistemological, methodological, ethical and practical reasons. Thearguments in these areas go beyond disputes over rational/technicalmatters, and the debates about the role of experiments (andrandomisation in particular) expose deep philosophical divides. Inother parts of the public sector the use of experimentation in theform of randomised controlled studies is largely absent in guidingpolicy or practice, for example, in housing, transport, welfare orurban policy (see Chapters Seven to Ten). It remains moot whetherthe absence of randomisation from evaluations in these policy areasarises because of intrinsic limitations of the method, practicalreasons, or lack of fit with the dominant service cultures.