Inspire – issue 12

Page 26




Decision makers within the financial world, be they policy makers, wealth managers, private investors or central bankers, all have to make decisions under uncertainty because decisions are made in relation to a future that we cannot predict with complete accuracy. For example, central banks have to decide if and how the base rate should change, and investors have to decide how to allocate their portfolios, based on current and projected economic and financial conditions. Those involved with modelling and forecasting are continuously looking for ways to improve forecast accuracy, the objective being to minimise the errors associated with their projections. This is of huge importance to decision makers because at the heart of decision making are the forecasts used to inform these decisions. Hence the way the forecasts are prepared, the information they convey and how this information is utilised to inform decisions will directly impact the decisions made and the resulting benefits and costs. A key question here is, what tools can we develop and use in order to make the best possible decisions we can, given we are operating in an uncertain environment? My research is concerned with macroeconomic and financial decision making under uncertainty. This uncertainty could stem from how economic relationships should be modelled, the accuracy with which these models are estimated, the uncertainty relating to unpredictable future events that may come to pass and the uncertainty due to changes over time in the relationships we are modelling. 26

I am a financial economist and I am interested in how modelling and forecasting techniques can be used to deal with these types of uncertainty, and whether accounting for these uncertainties leads to better decision making within the context of asset allocation. These statistical and mathematical modelling tools are not just restricted to use in finance and economics, but are of importance to anyone interested in forecasting – be it forecasting stock returns, exchange rates, inflation, rainfall or the demand for electricity. Further, it is essential for those using forecasts to know the best way to assess forecast accuracy – how can we determine whose forecasts are the best? Specifically, decision makers may have several different forecasts generated by different models for the same variable, so how can they assess these forecasts and determine which one(s) to use? Conventionally, statistical metrics are used. However, given that forecasts inform decisions, it is fundamental to use evaluation criteria that reflect the

objectives of the user and assesses forecasts in the context in which they will be used. So for instance, an investor wants to know how she can maximise her wealth and would be interested in using these forecasts to assist her in allocating her portfolio. In this case the investor wants to know the value, in terms of potential gains in wealth, of a set of asset price forecasts, and further, the expected gains in wealth if she bases her decisions on a particular set of forecasts. Broadly, my work considers: (1) the interactions between the macroeconomy and the financial markets (using modelling techniques that accommodate the time variation in the economic relationships we seek to model); (2) the uncertainty about the forecasts we make; and (3) evaluating these forecasts in the decision-making context for which they are intended, and how these will (a) affect the decisions we make and (b) enable us to make better decisions. In particular, I use statistical and mathematical techniques to model