Technical change, the long-run behavior of the United States stock market, and an enquiry into the accuracy of simulations

This thesis has two parts, each with a different subject. Part I studies the connection between technical change and the major movements of the US stock market. Part II adds to the foundations of numerical simulation and computational economics.; The post-war behavior of the US stock market is extremely puzzling. After a decade of relative stability, equity prices (as measured by the market value of US corporations as ratio of the replacement value of their assets) went down by 50% in 1973--1974 and stagnated at that level for the following decade before starting to recover in the mid-1980s. Equity prices increased dramatically during the 1990s and collapsed again in 2000 and 2001. Today they stand at their 1960--72 average level.; The major movements in the US stock market coincide with dramatic changes in the technological possibilities of the US economy. A growing literature suggests changes in technology are the main force behind equity price fluctuations. The first part of this thesis performs a quantitative test of this hypothesis.; To evaluate the impact of technical change on equity prices we use general equilibrium theory. Our analysis indicates technical change is quantitatively relevant for understanding the behavior of the US stock market. The trends of total factor productivity and the energy-saving behavior induced by the energy crisis of the mid-1970s can account for most of the secular movements in equity prices without resorting to animal spirits or irrational behavior.; The second part of this thesis provides a general framework for the simulation of stochastic dynamic models. Our analysis rests upon a continuity property of the set of invariant distributions and a generalized law of large numbers. From these results we establish convergence of the simulated moments to their exact values as the approximation error of the computed solution converges to zero. Furthermore, under an additional regularity condition we show the approximation errors of the simulated moments are of the same order of magnitude as that of the computed solution. The constants bounding these convergence orders can be explicitly computed and shown to depend on primitive parameters.