Temporal specification issues in applied econometric modelling

Abstract

Temporal specification in applied econometrics is concerned with the identification, formulation and analysis of the linkage between aspects of a model and a data set which relate to time. One of the most visible manifestations of the temporal features of a model or data is the basic time interval. In fact, such an interval is intrinsic to every data generation process operating in the real world and, hence, is present in every economic model. Similarly, the data observation period characterizes all empirical data of a time series nature. This thesis deals with some of the theoretical and empirical issues arising from the study of the relationship between these two time intervals. We commence by considering a general framework which deals with temporal specification in applied econometric research. This is followed by a discussion of a number of assumptions essential in all empirical work that relies on combining models with data. The econometric methodology embodying the analysis of temporal characteristics of a model and a data set is then compared with the current practice of neglecting the temporal dimension of applied work. Next, we present a taxonomy of possible outcomes with respect to the temporal characteristics of a model and a data set. It is used to survey some studies which incorporate various methods of adjusting temporal characteristics of a model and/or a data set. The theoretical analysis in the thesis concentrates on temporal aggregation. We describe a model-based approach to studying the consequences of such a transformation of a model and a data set. Some theoretical results are obtained for the effects of aggregating data and models in order to demonstrate the importance of temporal specification in econometric modelling. Special attention is paid to the examination of dynamic models under temporal aggregation since they are indispensable to applied time-series econometrics. The efnpirical analysis contains two applications from macroeconometrics as practical illustrations of the temporal specification analysis. A short-run money demand function for Australia is used to assess the consistency of the postulated basic model interval and the da ta observation interval. The result of this assessment indicates temporal consistency of the model developed by accounting for temporal characteristics of the theoretical model and the data. This outcome contrasts with the rejection of temporal consistency for a commonly used partial adjustment model which neglects the temporal specification issue. The impact of neglecting this aspect of econometric modelling is also investigated in the context of testing the short-run neutrality of money in Australia. We also furnish some results based on Monte Carlo experiments which provide small sample evidence on the consequences of temporal aggregation in dynamic econometric modelling. We conclude the thesis with an overview of the significance of the temporal specification issue in applied econometrics and suggest some extensions for future research.