Coping with Nasty Surprises: Improving Risk Management in the Public Sector Using Simplified Bayesian Methods

‘Coping with Nasty Surprises: Improving Risk Management in the Public Sector Using Simplified Bayesian Methods’, Asia and the Pacific Policy Studies, 2015, 2(3), 452-466.

Bayesian methods are particularly useful to informing decisions when information is sparse and ambiguous, but decisions involving risks must still be made in a timely manner. Given the utility of these approaches to public policy, this article considers the case for refreshing the general practice of risk management in governance by using a simplified Bayesian approach based on using raw data expressed as ‘natural frequencies’. This simplified Bayesian approach, which benefits from the technical advances made in signal processing and machine learning, is suitable for use by nonspecialists, and focuses attention on the incidence and potential implications of false positives and false negatives in the diagnostic tests used to manage risk. The article concludes by showing how graphical plots of the incidence of true positives relative to false positives in test results can be used to assess diagnostic capabilities in an organisation—and also inform strategies for capability improvement.

Open access PDF of the full original article: Click Here

Managing Risk and Increasing the Robustness of Invasive Species Eradication Programs

‘Managing Risk and Increasing the Robustness of Invasive Species Eradication Programs’,  Asia and the Pacific Policy Studies, 2015, 2(3), 485-493.

Invasive species eradication programs can fail by applying management strategies that are not robust to potentially large but non- quantified risks. A more robust strategy can succeed over a larger range of possible values for non-quantified risk. This form of robust- ness analysis is often not undertaken in eradi- cation program evaluations. The main non- quantified risk initially facing Australia’s fire ant eradication program was that the invasion had spread further than expected. Earlier consideration of this risk could have led to a more robust strategy involving a larger area managed in the program’s early stages. This strategy could potentially have achieved eradication at relatively low cost without sig- nificantly increasing known and quantified risks. Our findings demonstrate that focusing on known and quantifiable risks can increase the vulnerability of eradication programs to known but non-quantified risks. This high- lights the importance of including robustness to potentially large but non-quantified risks as a mandatory criterion in evaluations of inva- sive species eradication programs.

Open access PDF of the full original article: Click Here

 

Solving Large Dimensional CGE Models

‘Solving intertemporal CGE models in parallel using a singly bordered block diagonal ordering technique’, Economic Modelling, 2016, 52, 3-12.

The work introduces a direct ordering method that employs a special feature of an intertemporal Computable General Equilibrium (CGE) model to reorder its first-order partial derivative matrix into a Singly Bordered Block Diagonal (SBBD) form. The matrix can then be decomposed into LU form and solved in parallel. With this method, the numerical results  show a substantial advantage in computational time and memory use for parallel solutions of intertemporal CGE models in comparison to current serial solution methods. A solution for an intertemporal and regional model of the Vietnamese economy is provided as an example and comparator for the different methods.