Expectations : some econometric aspects

Abstract

To express an economic model Including expectations in a manageable form econometricians frequently replace expectationa with a functional form resulting from an hypotheala as to their formation. The most popular expectationa hypothesea are overviewed in Chapter 2. Survey data, if available and suitable, are an alternative to expectations hypothesea to represent expectations. How to ascertain the suitability of a survey aeries, and expectations hypothesis, Is discussed throughout the thesis. Identification of models involving expectations, conditional on the assumed expectations hypothesis, is examined by introducing e third class of economic variable, expectations, which enables the contribution to (or sub tr ac t i onfrom) ldentlflability, of expectations, to be assessed. The naive, adaptive and extrapolative hypotheses in general do not effect the ldentlflability of a model identified when expectations are assumed to be exogenous. The distributed lag and weakly rational models of expects* tions require additional assumptions to enable a model including them to be Identified. These findings are elaborated in Chapter 3. It is demonstrated that whether or not current fully rational expectations models are identified may depend upon the assumptions made regarding the information set utilized in forming the rational expectation. Conditions for ident iflability given Wickens' assumptions are developed and compared with those derived by Wallis, given his assumptions, lhe conditions are found to differ. Efficient conditional estimation of expectations models is discussed. FIML methods are appropriate in all instances; in general non linear, cross equation restrictions are involved when any hypothesis (except of course the naive) is assumed true. Consistent estimation methods for current and future rational expectations models which have been suggested in the literature are reviewed. A generalized time series approach producing consistent estimates of future rational expectations models is developed. As identification and estimation of expectations models are conditional upon the assumed hypothesis being true it is desirable to test the validity of this assumption. A model selection procedure is advanced as an improvement on methods which isolate and test a single hypothesis. Whether or not survey data are available a model selection procedure is appropriate. The model selection procedure is also proposed to select an appropriate series when a number are available. All procedures extend to include a number of structural models. An example of the model selection procedure to choose a suitable series and structural model simultaneously is presented. Many and varied survey series are required but these series should be quantitative not qualitative. Qualitative series need to be transformed to quantitative series for use in econometric models. The behavioural models underlying transformation methods may not be representative of the survey group; a number of plausible behavioural models not consistent with the transformation methods are discussed in Chapter 5. The often used Carlson-Parkin transformation method assumes the distribution of the variable concerned, across respondents, is normal. This assumption is tested in Chapter 6, with respect to expected inflation, and found to be invalid. Data obtained by such a transformation method should be carefully considered before being employed. Show more