Computationally intensive econometrics using a distributed matrix-programming language

Publisher Website
Google Scholar
Abstract
This paper reviews the need for powerful computing facilities in econometrics, focus- ing on concrete problems which arise in nancial economics and in macroeconomics. We argue that the profession is being held back by the lack of easy to use generic software which is able to exploit the availability of cheap clusters of distributed computers. Our response is to extend, in a number of directions, the well known matrix-programming interpreted language Ox developed by the rst author. We note three possible levels of extensions: (i) Ox with parallelization explicit in the Ox code; (ii) Ox with a parallelized run-time library; (iii) Ox with a parallelized interpreter. This paper studies and implements the rst case, emphasizing the need for deterministic computing in science. We give examples in the context of nancial economics and time-series modelling.