OR modelling for public sector performance measurement

The overall research theme of this thesis is to analyse, from both a theoretical and an empirical perspective, how Operational Research modelling can be used in support of public sector performance measurement. The thesis includes four papers written by Van der Meer together with a variety of co-authors. Three of these papers discuss the application of a variety of OR models - in particular, Logistic Regression Analysis, Stochastic Frontier Analysis, Data Envelopment Analysis and Bayesian Belief Networks - relating to the performance of the Maritime and Coastguard Agency in coordinating Search And Rescue services around the UK coast. The fourth paper discusses the application of Discrete Event Simulation to support performance measurement in a Musculo-Skeletal Unit operating across two Glasgow hospitals that are both part of NHS Scotland. The thesis investigates five research questions, all of which can be directly traced back to the overall research theme. The evidence base for the investigation consists of the papers referred to above, in conjunction with a wide range of sources from the academic and professional literature. The conclusions may be briefly summarised as follows. First, there are five specific problems and challenges for performance measurement in public service organisations. Second, there are at least two categories of OR modelling approaches that can play an important role in support of public sector performance measurement; namely, models for measuring performance differentials and models for understanding the cause-and-effect relationships driving such performance differentials. Third, while there are advantages to a multimethodology approach to OR modelling, there are also a number of barriers (including paradigm incommensurability). Fourth, the adoption of Critical Realism as a philosophical basis for combining different OR models can resolve problems of paradigm incommensurability. Fifth, OR models that are founded on a realist ontology should be validated according to two specific groups of criteria.