T HAS traditionally been argued that, if firms create external economies and diseconomies, the proper role of a welfare-maximizing government is to constrain the behavior of firms by arranging rates of taxes and subsidies in order to equate private with social benefit. We attempt to establish both the conditions under which this classical policy prescription might work and is needed, and those under which it cannot be expected to work. First, we argue that motivation exists for firms themselves to try to eliminate externalities in production through merger. Second, we attempt to show that technological externalities can be divided neatly into two cases, which we label "separable" and "non-separable," respectively. Third, if merger has not eliminated the externalities, we argue that the classical scheme of per unit taxes and subsidies can be clearly successful in equating private with social benefit only in the separable cases. Fourth, if the externality is non-separable, we argue that it is not clear that the classical prescription can work even at the conceptual level, since problems of uncertainty and the non-existence of equilibrium arise. Finally, we note that this latter possibility poses some difficult problems for policy-makers, and we attempt to outline and explore briefly alternative policy approaches. The analytic approach which we shall employ involves the consideration of two firms in a competitive industry. The traditional or classical approach, on the other hand, often involves an analysis of externality between competitive industries. We choose to depart from this traditional approach for several reasons. First, the firm is an entity which fits more easily into the framework of our analysis. Second, and more fundamental, it is individual decision units-firmswhich react to externalities so that it seems more "natural" to conduct the analysis at that level.2 Furthermore, concentration upon the industry (as opposed to the firm) requires a certain amount of aggregation which tends to mask some of the more important and interesting points at issue. This aggregation is especially misleading with respect to public policy regulation, where the problem is made to appear much more simple than it actually is. Finally, utilization of the firm as the basic analytic unit gives a level of generality 1 This paper was written as part of the project "The Planning and Control of Industrial Operations" under a grant from the Office of Naval Research and the Bureau of Ships at the Graduate School of Industrial Administration, Carnegie Institute of Technology. The authors would like to express their appreciation to Professors W. W. Cooper and J. F. Muth, both of Carnegie Institute of Technology, Dr. R. R. Nelson, Council of Economic Advisers, and Professor James M. Buchanan, University of Virginia, for their very helpful comments and criticisms. 2 Interestingly enough, J. de V. Graaff also considers that externalities are a phenomenon which relates to the firm rather than the industry, and, furthermore, he seems to think this point quite important (see his Theoretical Welfare Economics [Cambridge: Cambridge University Press, 1957], p. 19).