Stima di modelli Macro-Finance a tempo continuo

Time series modelling has been widely utilized in the macro-finance area. Usually, models involve data that are generated in smaller time intervals than the sampling interval related to the available data, giving rise to a fundamental problem of econometric modelling: the interval between observations is much larger than the decision intervals of economic agents that the observations reflect. This leads to estimates which are subject to a distortion called temporal aggregation bias. One of the aims of this thesis is to emphasize that the use of such linear time series models, whenever naively specified, is potentially incompatible with a context where the data are generated in finer time intervals than the interval pertaining to the available data. The first part of this work discusses the development of, and issues arising in, the formulation of structural continuous time linear models and the estimation of their parameters using an exact discrete time model; a comparison with an alternative approximation method for stochastic differential equations, namely the Euler-Maruyama scheme, is also considered. The second part of this work concentrates on the estimation of the structural parameters of the Ornstein-Uhlenbeck process, the touchstone for financial variables modelling. Through Monte Carlo experiments we assess the finite sample properties of the maximum-likelihood estimator of the mean reversion parameter under the two different discrete representations - the exact and the approximate one. Simulations are based on Campbell and Viceira’s seminar papers and pertain to both the univariate and the multivariate setting.