Recent policy applications of control theory methods in dynamic economic models raise the issue of imposing the transversality condition for determining a unique optimal control policy. In a stochastic framework this issue involves alternative methods of estimation, which are discussed here both theoretically and empirically. The economic implications of the alternative methods are analyzed here in some detail through several recent dynamic models in economic growth and exchange rate instability.

This paper discusses three key elements of stochastic growth in the Schumpeterian dynamics. These elements comprise the new entry of firms in an industry, the displacement of the old technology by the new and the nonlinear impact of learning by doing on the growth of innovating firms. Each of these elements has important implications for the new theory of endogenous growth.

Production and cost frontiers of a firm in an industry are directly affected by the uncertainty of market demand and the uncertainty of input availability. The nonparametric approach of data envelopment analysis is generalized here in both static and dynamic directions by incorporating these uncertainties.

Two types of statistical models are empirically applied to test the pattern of volatility in the exchange rate markets. One considers the autoregressive models and tests the random walk hypothesis. The other considers the conditional variance process and tests the hypothesis of chaotic dynamics. Empirical results mostly support the random walk hypothesis and also the existence of Lorenz-type chaos.

Which of the two forces, past trends or future expectations plays a more dominant role in exchange market volatility? This hypothesis is econometrically tested here for four advanced industrial countries, France, UK, Japan and Germany over the period 1985 to 1995.