Research

The aim of this research project is to significantly improve our understanding of the effects of non-linearities and rare events in economics and finance, by harnessing the power of Sequential Monte Carlo (SMC) methods. The project will address the limitations of current knowledge by focusing on the following research objectives.

Development of scalable SMC methodology for large dimensional problems

In order to develop algorithms that can deal with the demanding large-scale systems we are interested in, we aim to advance SMC methodology and exploit the parallel nature of compute clusters, i.e. the availability of a large number of processors, to speed up existing algorithms. These algorithmic advances will be utilized in the applications in the other two workstreams.

Application of SMC methods to structural models in economics and finance

Rich structural models in economics and finance naturally give rise to non-linear solutions and latent state spaces, making it a natural but challenging application area for SMC methods. In this workstream, we apply SMC techniques to estimate models used in macro-finance, macro-economics and industrial organization. In addition to the speedups from algorithmic advances, we will use the particular model structure in each problem to design efficient algorithms.

Application of SMC methods to reduced form models in economics and finance

State-of-the-art reduced form models incorporate complex dependencies, structural change and non-linearities to better fit complex economic and financial data, making empirical analysis challenging. Fortunately, SMC is well adapted to deal with the resulting high-dimensional non-linear estimation tasks. In this workstream, we use SMC methods to fit reduced form option pricing models, VAR models with sparse structural change and rich copula panel models.