The South African business cycle and the application of dynamic stochastic general equilibrium models

[摘要] ENGLISH ABSTRACT: This dissertation considers the use of Dynamic Stochastic General Equilibrium(DSGE) models for the analysis of South African macroeconomic business cyclephenomena. It includes four separate, but interrelated parts, which follow alogical sequence.Therst part motivates the use of these models before establishing thetheoretical foundations for these models. The theoretical foundations are accompaniedby detailed derivations that are used to construct a model for asmall open economy.The second part considers the properties of South African macroeconomicdata that may be used to estimate the parameters in these models. It includesa discussion of the variables that may be included in such a model, as well asvarious methods that may be used to extract the business cycle. Thereafter,the sample size for the dataset is established, after investigating for possiblestructural breaks in therst two moments of the data, using various univariateand multivariate techniques. Thenal chapter of this part contains an investigationinto the measures of core in ation, whereby a comparison of trimmed means, dynamic factor models and various wavelet decompositions are appliedto data for South Africa.The third part considers the application of the dataset that was identi edin part two, in a DSGE model that incorporates features that are typical ofsmall open economies. It includes a discussion that relates to the role of theexchange rate in these models, which is found to contain key information. Inaddition, this part also includes a optimal policy investigation, which considersthe reaction function of central bank.Thenal part of this thesis considers more recent advances that have beenapplied to DSGE models for the South African economy. It includes an exampleof a nonlinear model that is estimated with the aid of a particlelter,which is then used for forecasting purposes. The forecasting results of bothlinear and nonlinear versions of the model are then compared with the resultsfrom various Vector Autoregression (VAR) and Bayesian VAR models.

[发布机构] Stellenbosch University