I feel deeply honored and privileged to have the opportunity of giving the first Baffi Lecture at the Bank of Italy. Paolo Baffi was not only a distinguished banker and financial expert, he was also a remarkable economist and a visionary social thinker. He had outstanding technical expertise in many different fields, but combined his intellectual eminence with a profound sense of values. As Governor Ciampi put it at the general meeting of the Bank of Italy last May, Paolo Baffi represented “an extraordinary combination of penetrating logic, erudition and moral strength … [he was] not only a gifted student of economics, he had a deep-seated commitment to act for the common good.”1 In remembering Baffi today, we must keep in mind both his intellectual contributions and his general evaluative concerns.

How do economists persuade their readers that one policy is superior to another? A glance at the literature on welfare economics quickly provides the answer to this question: Economists enter policy debates armed with mathematical models, evaluating options on the basis of their consequences. Economists typically classify a policy change as a welfare (or “potential Pareto”) improvement with respect to the status quo if the gain realized by the winners exceeds the harm sustained by the losers. The best policy becomes the one that generates the highest net benefit.

Economists use tastes as a source of information about personal welfare and judge the effects of policies upon preference satisfaction; neoclassical welfare economics is the analytical embodiment of this preference sovereignty norm. For an initial distribution of wealth, the welfare-maximizing outcome is the one that exhausts all possible gains from trade. Gains from trade are defined relative to fixed ordinal preferences. This analytical apparatus consists of both the Pareto principle, which implies that externality-free voluntary trades increase welfare, and applied costbenefit analysis, which attempts to weight costs and benefits when evaluating policies that are not Pareto improvements.