The traditional method for conducting sensitivity analysis is to repeatedly solve a model while varying the parameters. The solution is then obtained as some average of these optimal solutions under those different conditions or states of the world. The present work presents results of conducting sensitivity analysis using a method more firmly ground in mathematical programming theory. The present analysis models the investment decisions in a case with large uncertainty in demand: the steel industry in Argentina. Special emphasis is devoted to the recent history, where a recent shift in economic policy (1976-1981) towards allowing free competition with imported products resulted in a severe crisis for the steel industry and its trading partners. An increase in exports was observed during this period which is not likely to continue if there is a recovery process. In the first sections, the relation of steel production and economic growth is analyzed in the context of the world situation of the industry, setting the background for the analysis of the Argentinian industry as a case study. The results of the present model adequately describe the existence of unutilized capacity observed in the industry, as well as the recent increase in exports. The most important conclusion of the model is that the traditional method of conducting sensitivity analysis results in significant inefficiency of the reached decisions, involving large losses for a case such as the steel industry considered here. / Master of Arts / incomplete_metadata