The financial crisis of 2007 has had an effect on most of us. The near collapse of large financial institutions, the bailout of banks by national governments, the contraction of the real estate market and downturns in stock markets around the world had dramatic effects on employment, housing, investment returns and on pension funds. Economic models did a poor job of predicting this crisis and the resulting global recession.
Could better macroeconomic models, coupled with high-performance computing, lead to an improved understanding of such financial trends? Could they help us increase confidence in decision making, give us better returns on investments and enable businesses to better forecast future trends and make informed strategic decisions?
Modern economic modelling is reliant on complex mathematical models. These models are powerful tools that can provide useful insights into the behaviour on the economy, but they are limited by the rules and data which underpin the model. If a model is applied outside its intended scope then it may give misleading information and be a poor tool for policy-makers.
The ability to build and test economic models against the appropriate data is of vital importance. In partnership with OSTC Wales Limited, Dr David Meenagh of the Cardiff University Business School has embarked on an exciting project to improve macroeconomic models using High Performance Computing (HPC) Wales’ supercomputing technology. Here, David discusses the exciting research he is involved in:
‘The project we have embarked on with OSTC will reawaken the statistical testing of macroeconomic models. We aim to build up an agreed methodology of testing that can be widely applied to such models in a scientific way.’
‘We propose to extend the methods to unfiltered (raw) data that may well be non-stationary (continually changing). To date, attention has been focused on filtered stationary data, where the filtering process may well remove important information from the raw data, with unknown consequences. Stationary data continuously returns to a fixed equilibrium path, whereas with non-stationary data the equilibrium path to which it continuously returns is constantly shifting. This extension also makes it possible to test models of growth, where previously convincing tests have been hard to formulate.’
‘Simulating a model on unfiltered data is highly computer intensive, as estimating the model parameters requires the running of millions of simulations of the model.’
‘HPC Wales are a vital partner in our project due to the compute-intensive nature of the tests. Their support is enabling us to reduce dramatically the time taken to run the simulations. Without their technology, we would not be able to estimate large macroeconomic models in this way; it would take years to carry out the work.’
David has been working with the Business School at Cardiff University for the past 14 years, driven on by the thought provoking research being undertaken, a desire to have an impact on the decisions of policy-makers, and the opportunity to bring new investment opportunities to Wales.
‘The impact of this project will be wide-ranging in the applied macroeconomics domain as we believe academic practice will be quick to adopt the methods. Our ultimate aim is to see applied macroeconomists in policy-making adopt the practice as routine. The gains to policy-making from scientifically selecting models that are compatible with the data will be substantial.’
The use of scientific methods in policy-making is an area which has been gaining momentum in recent times, for example the “Test, Learn, Adapt” paper from the Behavioural Insights Team at the Cabinet Office promotes the use Randomised Controlled Trials in developing public policy.
‘The support staff at HPC Wales have been very helpful in answering all the questions we’ve had, and outlining how the technology can be efficiently applied within our project.’
To find out more about how high performance computing could benefit your research, please contact us.