Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Multivariate Modelling of Non-Stationary Economic Time SeriesIntroduction

Multivariate Modelling of Non-Stationary Economic Time Series: Introduction [In this edition the authors draw on further research on multivariate time series methods that are used and might be used in the analysis of economic and financial data. The approach is firmly multivariate and starts with linear models: time series models where the data are taken to be stationary. This implies that the series are generally differenced and that likely relationships relate to growth rates in economics and returns in finance. In this context, the model can be transformed to derive long-run relations in the rate of growth of the economy and this links with other economic factors or to long-run asset price relations. Impulse response functions are then derived and their uniqueness is considered as well as forecasting. The latter is treated in brief as these are the primary concerns of the book by Clements (2005). Forecasting is clearly an important direct and indirect outcome of time series modelling, but for the economics profession such models have meaning related to the identification of economic phenomena and the determination of policy. Forecast performance is also an important component in model selection, for economist, financial analyst and pure statistician. For the economic and financial analyst the nature of the model might be suggestive of the nature of the economy, either in a long-run or a short-run sense. Hence, models that make no economic sense or are not able to be identified might be considered to be inconsistent with theory and as a result may only have a statistical value. There is also an interest in models being well specified and stable over time. There may be reasons for instability by virtue of policy changes, but the interest is then directed to the invariant structures that have implications for exogeneity. In particular the idea that relations invariant to series that are not stable can be viewed as indicative of super exogeneity. This focuses attention on parameters, which are often linked to the long run. Having distinguished between the exogenous and endogenous series, one will then be drawn towards econometric identification. This is the capacity to detect economic structure from the long-run or short-run parameters, as compared with time series identification that relates to the dynamic process driving the short-run behaviour of the data. Causality is related to exogeneity and this considers the extent to which it is possible to detect the forcing variables or to forecast conditional on a subset of the data either in the short run or in the long run. Causality is sensitive to the nature of the system and in the short run relates to the parameters of certain dynamic processes being set to zero. This applies in the long run when some of the variables satisfy the conditions appropriate for cointegrating exogeneity (Hunter, Econ Lett 34:33-35, 1990), but with the exception of systems with only two variables or one cointegrating vector, this is also a requirement for the effect of long-run causality in the short run.] http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png

Multivariate Modelling of Non-Stationary Economic Time SeriesIntroduction

Download PDF
18 pages

Loading next page...
 
/lp/springer-journals/multivariate-modelling-of-non-stationary-economic-time-series-j0Tn9yrAj4
Publisher
Palgrave Macmillan UK
Copyright
© The Editor(s) (if applicable) and The Author(s) 2017. The author(s) has/have asserted their right(s) to be identified as the author(s) of this work in accordance with the Copyright, Designs and Patents Act 1988.
ISBN
978-0-230-24330-9
Pages
1 –19
DOI
10.1057/978-1-137-31303-4_1
Publisher site
See Chapter on Publisher Site

Abstract

[In this edition the authors draw on further research on multivariate time series methods that are used and might be used in the analysis of economic and financial data. The approach is firmly multivariate and starts with linear models: time series models where the data are taken to be stationary. This implies that the series are generally differenced and that likely relationships relate to growth rates in economics and returns in finance. In this context, the model can be transformed to derive long-run relations in the rate of growth of the economy and this links with other economic factors or to long-run asset price relations. Impulse response functions are then derived and their uniqueness is considered as well as forecasting. The latter is treated in brief as these are the primary concerns of the book by Clements (2005). Forecasting is clearly an important direct and indirect outcome of time series modelling, but for the economics profession such models have meaning related to the identification of economic phenomena and the determination of policy. Forecast performance is also an important component in model selection, for economist, financial analyst and pure statistician. For the economic and financial analyst the nature of the model might be suggestive of the nature of the economy, either in a long-run or a short-run sense. Hence, models that make no economic sense or are not able to be identified might be considered to be inconsistent with theory and as a result may only have a statistical value. There is also an interest in models being well specified and stable over time. There may be reasons for instability by virtue of policy changes, but the interest is then directed to the invariant structures that have implications for exogeneity. In particular the idea that relations invariant to series that are not stable can be viewed as indicative of super exogeneity. This focuses attention on parameters, which are often linked to the long run. Having distinguished between the exogenous and endogenous series, one will then be drawn towards econometric identification. This is the capacity to detect economic structure from the long-run or short-run parameters, as compared with time series identification that relates to the dynamic process driving the short-run behaviour of the data. Causality is related to exogeneity and this considers the extent to which it is possible to detect the forcing variables or to forecast conditional on a subset of the data either in the short run or in the long run. Causality is sensitive to the nature of the system and in the short run relates to the parameters of certain dynamic processes being set to zero. This applies in the long run when some of the variables satisfy the conditions appropriate for cointegrating exogeneity (Hunter, Econ Lett 34:33-35, 1990), but with the exception of systems with only two variables or one cointegrating vector, this is also a requirement for the effect of long-run causality in the short run.]

Published: May 10, 2017

Keywords: Unit Root; Time Series Model; Impulse Response Function; Error Correction Model; Future Contract

There are no references for this article.