This thesis pursues the double purpose of measuring, and improving the tools of measurement of, the economic impact of both advertising and pricing decisions by firms in duopolistic industries. In seeking to obtain efficient statistical estimates of the effect of these variables on market demands, we specify various structural economic models from which we derive estimable systems of simultaneous equations. The essential hypothesis is that, at any given period, observations on the variables of these simultaneous-equation econometric models have arisen as the equilibrium outcomes of some specified games of competition between firms.
This work illustrates a new methodology that combines game theoretic considerations and modem econometric and statistical tools. Our empirical findings have, indeed, demonstrated how fruitful and promising such a combination is.
The analysis of data on the U.S. soft drink industry by means of the framework developed in this study produces two types of results. First, we obtain more accurate estimates of the economic impact of advertising, a highly strategic and instrumental variable for firms, than those obtained so far with available techniques. We utilize full information maximum likelihood methods to estimate simultaneous-equation econometric models of the U.S. soft drink industry, each of which incorporates information about a specific form of competition between firms. Second, using recent econometric techniques, we perform some statistical tests which enable us to discriminate among the different models. We are, therefore, in a position of determining which of the various formal representations of the industrial organization of such a sector is most compatible with the available data.
