Summary: | The identification of best practices is a crucial aspect of benchmarking (Camp, 1989) but there is little published evidence to show that a framework exists which identifies such practices in the public sector (Kouzmin el al., 1999). This thesis examines how the benchmarking of local government services might be improved through using Data Envelopment Analysis (DEA). Seminal work in 1978 by Charnes, Cooper and Rhodes (CCR) provided the original ratio model of DEA for Constant Returns-to-Scale (CRS) and this was extended in 1984 by Banker, Charnes and Cooper (BCC) to include Variable Returns-to-Scale (VRS). Whilst it is recognised that DEA has sincc been developed, this research applies the primal CCR and BCC efficiency models to estimate productive efficiency over an II-year period (from 1993/94 to 2003/04) across 4 significant service areas provided by local government: Housing Benefit and Council Tax Benefit (HB and CTB); planning control; houschold waste collection and recycling; and highway maintenance. Input and output variables arc selected from published Audit Commission Performance Indicators (PIs), Best Value Performance Indicator (BVPI) data, and statistics produced by the Chartered Institute of Public Financial Accountants (CIPFA). Performance data is pooled into between 3 and 5 windows where the same input and output variables exist. DEA scores, run under CRS provide a ranked performance position for each council averaged across the data windows. Spearman's rank correlation confirms the statistical significance between each of the ranks across the data windows. A sample of 51 high and low-ranked councils across London borough, metropolitan borough, unitary, district and county councils (as applicable) is comparcd and contrasted with performance reviews published by the Audit Commission. A total improvement summary for each type of local authority is estimated against the frontier of best practice while scale effects are calculated under VRS constraints. The Malmquist index has been used to estimate productivity changes over time. Using data from one year only, hierarchical cluster analysis is calculated to identity benchmarking partners. The research identifies the issues of using published performance data in terms of breadth, availability, clarity, annual consistency, the lack of non-discretionary factors and the reporting of data on a per-unit basis. The study also highlights the limitation of applying DEA consistently, resulting from the absence of a framework for the selection of input and output variables for efficiency estimation in local government. Findings from this study demonstrate that DEA can be used to discriminate between high and low-efficient service providers. This allows improved benchmarking against best performing councils to be undertaken and total improvement estimates of £543m p.a. are projected across the 4 services, should best practices be adopted. Important similarities have been found between the high and low-ranked efficient councils and published Audit Commission performance reviews. Through the use of dendrograms, benchmarking clusters for each service area by council type are visualised on a stepwise basis. The contribution made by this research emanates from the discussion of how DEA has been applied in practice to provide an efficiency measure for local government services. Nevertheless, the research acknowledges the risks inherent in basing findings on the analysis of PIslBVPls and secondary data, which may be restricted in terms of availability and quality.
|