This paper describes two models of an agency that is collecting and reporting observations on a dynamical linear stochastic economy. The first is a "classical" model, with the agency reporting data that are the sum of a vector of "true" variables and a vector of measurement errors that are orthogonal to the true variables. The second is a model of an agency that uses an optimal filtering method to construct least-squares estimates of the true variables. These two models of the reporting agency imply different likelihood functions. A model of the investment accelerator is used as an example to illustrate the differing implications of the models.

This paper explores some of the implications for econometric practice of the principle that people's observed behavior will change when their constraints change. In dynamic contexts, a proper definition of people's constraints includes among them laws of motion that describe the evolution of the taxes they must pay and the prices of the goods that they buy and sell. Changes in agents' perceptions of these laws of motion (or constraints) will in general produce changes in the schedules that describe the choices they make as a function of the information that they possess. Until very recently, received dynamic econometric practice ignored this principle. The practice of dynamic econometrics should be changed so that it is consistent with the principle that people's rules of choice are influenced by their constraints. This is a substantial undertaking and involves major adjustments in the ways that we formulate, estimate, and simulate econometric models.

Estimates of a rational expectations version of Friedman's time-series consumption model are obtained by imposing the pertinent restrictions across the stochastic processes for consumption and income. A likelihood ratio test is used to test the adequacy of three joint hypotheses: namely, Friedman's model, rational expectations, and some arbitrary conditions on the disturbance process in the consumption function. The paper treats both the cases in which income is econometrically exogenous with respect to consumption and those in which it is not. The macroeconomics of this exogeneity condition are briefly discussed.

A dynamic linear demand schedule for labor is estimated and tested. The hypothesis of rational expectations and assumptions about the orders of the Markov processes governing technology impose overidentifying restrictions on a vector autoregression for straight-time employment, overtime employment, and the real wage. The model is estimated by the full-information maximum-likelihood method. The model is used as a vehicle for reexamining some of the paradoxical cyclical behavior of real wages described in the famous Dunlop-Tarshis-Keynes exchange.

A statistical definition of the natural unemployment rate hypothesis is advanced and tested. A particular illustrative structural macroeconomic model satisfying the definition is set forth and estimated. The model has "classical" policy implications, implying a number of neutrality propositions asserting the invariance of the conditional means of real variables with respect to the feedback rule for the money supply. The aim is to test how emphatically the data reject a model incorporating rather severe classical hypotheses.

This paper is a critical review of and a reader's guide to a collection of papers by Robert E. Lucas, Jr. about fruitful ways of using general equilibrium theories to understand measured economic aggregates. These beautifully written and wisely argued papers integrated macroeconomics, microeconomics, finance, and econometrics in ways that restructured big parts of macroeconomic research. (JEL A31, E00, E13, E50)