A1375
Title: On the optimality of score-driven models
Authors: Alessandra Luati - Imperial College London (United Kingdom) [presenting]
Paolo Gorgi - Vrije Universiteit Amsterdam (Netherlands)
Christopher Sacha Aristide Lauria - University of Bologna (Italy)
Abstract: Score-driven models are shown to be optimal with respect to a novel, intuitive, high dimensional and global optimality criterion, defined as Conditional Expected Variation optimality. The property formalises the use of the score as the driving force in updating models for time-varying parameters. To prove the aforementioned property, a point of contact between the econometric literature and the time-varying optimisation literature is established. As a matter of fact, Conditional Expected Variation optimality can be naturally viewed as a generalisation of the monotonicity property of the gradient descent scheme. Differently from information-theoretic optimality criteria based on the Kullback-Leibler divergence between the model density and the true density, the Conditional Expected Variation regards Euclidean distances in the parameter space, holds on the whole parameter space and is trivially extended to the case in which the time-varying parameter is multidimensional.
