This dissertation collects three independent essays in the area of Macroeconomics and Macroeconomic forecasting. The first chapter introduces and motivates the three essays. Chapter 2 highlights a serious problem of the Bayesian vector autoregressive (BVAR) models with Litterman???s prior cannot be used to get accurate forecasts of the driftless variables in a mixed drift models. BVAR models with Litterman???s prior, because of the diffuse prior on the constant, do not perform well in the long-run forecasting of I(1) variables either, if they have no drift. This is interesting as in practice most of the macro models include both drift and driftless variables. One solution to this problem is using the Bewley (1979) transformation to impose zero drift to driftless variables in a mixed drift VAR models. A novel feature of this chapter is the use of g-prior in BVAR models to alleviate poor estimation of drift parameters of the Traditional BVAR model. Chapter 3 deals with another possible explanation for the poor performance of the Traditional BVAR models in inflation forecasting. BVAR with Litterman???s prior have the disadvantage of a lack of robustness to deterministic shifts, exacerbated by the ill-determination of the intercept. Several structural break tests show that Australian inflation has breaks in the mean. Chapter 3 uses the Kalman filter to allow parameters to vary over time. The novelty of this chapter is modifying the standard BVAR model, where deterministic components evolve over time. Moreover, this chapter set aside the assumption of diagonality in the prior variance-covariance. Hence, another novelty of this chapter is using a BVAR model with modified non-diagonal variance-covariance matrix similar to the g-prior, where the deterministic components are the only source of variation, to forecast Australian inflation. Chapter 4 moves onto DSGE models and estimates a partially microfunded small-open economy (SOE) New-Keynesian model of the Australian economy. In this chapter, structural parameters of the rest of world (ROW), SOE, and closed economy, are estimated using Australian data as the small economy, and the US as the ROW, with the full information maximum likelihood.
Identifer | oai:union.ndltd.org:ADTP/258798 |
Date | January 2006 |
Creators | Heidari, Hassan, Economics, Australian School of Business, UNSW |
Publisher | Awarded by:University of New South Wales. School of Economics |
Source Sets | Australiasian Digital Theses Program |
Language | English |
Detected Language | English |
Rights | Copyright Hassan Heidari, http://unsworks.unsw.edu.au/copyright |
Page generated in 0.0017 seconds