Chapter 114 Experimetrics: The Use of Market Experiments to Evaluate the Performance of Econometric Estimators

Experiments with human subjects have been widely used to test hypotheses derived from economic theory and to provide stylized facts about empirical regularities in economics. One of the clear implications of experimental economics is that the efficiency and distributional properties of market allocations are dependent on the market institutions through which exchange takes place. This makes clear the limitations of economic theories of markets that are institution-free. Econometric theory, as conventionally taught in graduate schools and applied by many economists, is largely institution-free. Properties of econometric estimators are derived from assumptions about the data generating process (DGP) and the error distribution. A clear example is provided by the most central simultaneous equations in economics, supply and demand equations. The basic results in econometric textbooks, that ordinary least squares is an inconsistent estimator and two stage least squares is a consistent estimator, are derived from the assumption that the DGP is competitive equilibrium with zero-mean random error.