This paper draws on a study of the introduction of market testing in the UK
civil service to explore ways in which managers involved in the implementa
tion of the new policy constructed evaluations of its impact. It is structu
red around three arguments. The first concerns the problems of evaluating '
what works' in the multi-stakeholder, multi-goal context of public manageme
nt The research highlights a range of overlapping and sometimes conflicting
evaluation criteria across different organizational and occupational group
ings. The second argument explores the difficulties of evaluation in the co
ntext of shifting policy objectives and the dynamic nature of institutional
change. The research shows how the practitioners involved shaped and resha
ped their construction of events over time as unanticipated benefits and di
sbenefits became evident. It also suggests ways in which they responded to
the changing policy context, constructing new rules and norms of action ove
r time. The third argument concerns the different levels of analysis underp
inning managers' constructions of 'efficiency' and 'effectiveness', and how
these constructions were deployed as strategies of legitimation in shaping
the process of institutional change.