A Spatial Econometrics Approach of Financial Complexity
Coordinator: Sophie Béreau
Date: 1/10/2013 -
The research project aims to develop econometric models that are relevant to explore financial complex phenomena in general and properly account for the implications of individual interconnections in the understanding of risk dynamics in the financial markets in particular. While a vast literature has intended to shed light on contagion and spillovers in financial market, the recent financial crisis has revealed that globalization and the resulting intensification of financial markets interconnectedness contributed to a drastic increase in risk. This evolution calls for a rethinking of both the models and the empirical methods used for the assessment of risk dynamics. To this end, we propose first to develop a synthetic and consistent framework to proper model individual interdependencies and their impact on aggregated dynamics, by combining elements from two different literatures. On the one hand, graph theory and network statistics help to assess and model how interdependencies are structured and characterized among a set of individuals. On the other hand, spatial econometrics provides solutions to identify and estimate regression models assuming some functional form for individual interdependences. Combining the two approaches should thus provide an innovative way to model financial interdependencies as well as contagion phenomena. One key contribution will be to assess in what respect relaxing the assumption of exogenous fixed predetermined distance matrix in standard spatial specifications impacts the identification of the parameters of the model as well as the statistical properties of the estimators. Second based on those theoretical explorations, we will propose relevant estimation methods of the model and compare their performances. Finally, a further step will be to extend our previous model to a nonlinear one, which calls again for both proper specification band estimation methods.
Application of High Performance Computing in Short-term Scheduling of Electric Power Systems under Uncertainty
Coordinator: Anthony Papavasiliou
Date: 1/10/2013 - 30/09/2015
The purpose of this project is to develop parallel algorithms for the short-term scheduling of electric power systems under uncertainty caused by the large-scale integration of renewable energy sources and demand response. This project responds to an increasing need for the improvement of day-ahead and real-time power systems scheduling and market clearing by leveraging parallel computation.