It is easy to argue that macroeconomics is the most important major component of economics as its main variables, such as aggregate income, consumption, investment, price indices, interest rates, exchange rates, and unemployment, affect the main decision makers in the economy, such as employers, consumers, investors, and government policy makers. Attempts to measure and model the relationships between such variables have involved virtually all of the dominant econometric methodologies and the problems that have arisen and were resolved have been extremely important in the development of econometrics. Amongst the first were the large simultaneous models built with annual or quarterly data constrained by some well-formulated and generally accepted theory. In many cases the model is fully specified from the theory and the only task remaining was estimation of the parameters, but what a difficult task that proved to be in many cases! The traditional search for better estimators and an understanding of their properties returned with the later consideration of rational expectations. Originally the large models were not very dynamic, in contrast with the main alternative approach (called here just 'time series analysis') which concentrated on dynamics,paid little or no attention to economic theory and built models involving only a few variables. Over the years these two approaches have interacted with each other, one side learning from and being influenced by the other. The large models became more dynamic and involved unit roots and cointegration, the time series models considered size, that is. the number of variables used more seriously and payed more attention to the use of economic theory.
展开▼