Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.247362
Title: Non-linearities in macroeconomics : evaluation of non-linear time series models
Author: Galvão, Ana Beatriz Camatari
Awarding Body: University of Warwick
Current Institution: University of Warwick
Date of Award: 2001
Availability of Full Text:
Access through EThOS:
Access through Institution:
Abstract:
This thesis evaluates different specifications of non-linear time series models applied to macroeconomic problems. The evaluations investigate whether linear models are a good representation of the data, and which non-linear specifications are comparatively better in three different applications. In addition, the implications of the evaluation to the understanding of macroeconomic problems and to economic predictions are analysed. The first evaluation concerns univariate non-linear time series models aimed at reproducing the asymmetries of the business cycles. Using business cycle stylised facts and conditional mean functions and surfaces, the results support the use of non-linear models that can generate a three-phase cycle as the specification that can reproduce all the business cycle features, including the asymmetries in the shape of the cycle. The second assessment is of models that characterise the non-linearities of the US term structure of interest rates. The forecast evaluation of different specifications of threshold vector equilibrium correction models, which are estimated for long- and short-term interest rates and their spread, shows that the inclusion of non-linearity improves short-horizon forecasts. However, when compared with AR models, the gains from nonlinearity only occur when the predictions for the spread are evaluated at long horizons. The third assessment concerns non-linear bivariate systems that account for the effect of non-linearities and/or structural breaks when the spread is employed as leading indicator. Different specifications are evaluated using their prediction of the probability of two definitions of recessions. Models with non-linearities and structural breaks perform better at predicting the probability of recession than linear models and models with only non-linearity or structural break. The results of the evaluation of univariate time series models improve the understanding of the connection between these models and business cycle asymmetries. The winner of the forecast competition of bivariate systems of interest rates and their spread indicates that the expectation theory of the term structure of interest only holds for the period in which the spread is negative, even though the spread can predict changes in the long-term rate in a specific state. In addition, the result that structural breaks and non-linearities are important to predict US recessions when the spread is the leading indicator changes the timing of a predicted recession for 2001.
Supervisor: Not available Sponsor: CAPES (Brazil)
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.247362  DOI: Not available
Keywords: HB Economic Theory Economics
Share: