Research: Supercomputers Help Predict Trade and the Behaviour of Financial Markets
A DCSC case story, provided by Professor Niels Haldrup, Aarhus University
Financial data bases are now being established with observations on prices and volumes of the assets traded. Minute-by-minute, each trade and information in relation to the trade, is registered, giving an enormous stream of data. Since so many assets are traded over time, giant amounts of data are accessible for data analysis. The field is gaining increased research interests in Denmark and there is a growing demand for access to scientific computing.
Also register-data with information on individuals are established and contain an increased complexity of information, both over a cross section, but also over time. In this respect the Danish registration of data on single individuals (via the CPR register) is unique from an international perspective and econometricians and other social scientists worldwide are attracted by the variety of data accessi-ble for Denmark in particular.
In marketing a rapidly developing research area focuses on patterns of trade by individuals. Loyalty cards give access to the trade behaviour of individuals re-garding single commodities traded, which can be combined with other register-based information, to identify important patterns in the way individuals behave to price changes and so on.
In all these cases, the data analysis is becoming increasingly complex. The (econometric) models designed incorporate a number of factors that give rise to strong computing power, a need that is caused both by the huge amounts of data to analyse, but also because of the complexity of the models to be esti-mated.
Another area where much computing power is needed concerns simulation based methods in the social sciences. This includes aspects of game theory, economet-ric estimation methods including monte carlo based methods, and the estimation of complex dynamic systems, for instance dynamic stochastic general equilibrium models.
Read more about the research: