|
Abstract:
|
The task of this work is to discuss issues conceming the specification, estimation, inference and forecasting in multivariate dynamic heterogeneous panel data models from a Bayesian perspective. Three essays linked by a few conraion ideas compose the work.Multivariate dynamic models (mainly VARs) based on micro or macro panel data sets havebecome increasingly popular in macroeconomics, especially to study the transmissionof real and monetary shocks across economies. This great use of the panel VAR approachis largely justified by the fact that it allows the docimientation of the dynamic impactof shocks on key macroeconomic variables in a framework that simultaneously considers shocks emanating from the global enviromnent (world interest rate, terms of trade, commonmonetary shock) and those of domestic origin (supply shocks, fiscal and monetary policy,etc.).Despite this empirical interest, the theory for panel VAR is somewhat underdeveloped.The aim of the thesis is to shed more light on the possible applications of the Bayesian framework in discussing estimation, inference, and forecasting using multivariate dynamic models where, beside the time series dimensión we can also use the information contained inthe cross sectional dimensión. The Bayesian point of view provides a natural environment for the models dlscussed in this work, due to its flexibility in combining diíferent sources of information. Moreover, it has been recently shown that Bayes estimates of hierachical dynamic panel data models have a reduced small sample bias, and help in improving the forecasting performance of these models.
|
