Optimization and its discontents
Already outlined towards the end of the nineteenth century, the constrained maximization model was refined and extended to the study of both consumers and firms, with, respectively, utility and profit as objective functions.
Its enhancement opened the way to a conception of economics based on two pillars: optimizing behaviour of agents, interpreted as individual rationality, and equilibrium of markets, interpreted as consistency of individual choices. Paul Samuelson (1947) gave a major boost to these developments by rewriting many problems of economics as optimization problems, with extensive use of mathematics. The discipline gained unity and coherence as apparently diverse subjects appeared to share the same underlying structure and to allow use of the same mathematical techniques.Maximization did not remain confined to the normative character that it had with Gossen, and over time came to be seen as an interpretation of actual behaviours too. The revealed preference approach developed by Samuelson (1938) and others aimed precisely to reformulate the model in a way that allowed checking it against empirical data; among others, Hal Varian (1982) performed tests using US consumption data.
The optimization model has not been beyond dispute, though: at least since the 1950s, Herbert Simon and others contended that actual decision makers lack the cognitive capacities to solve maximization problems and, rather, content themselves with “satisficing” behaviour, choosing options that are acceptable though not optimal. Along these lines, they developed a “bounded rationality” approach as an alternative to the seemingly strong requirements of the individual maximization model. Yet it was difficult to identify universal principles enabling economists to devise general models of nonmaximizing behaviour, so that bounded rationality hypotheses maintained an ad hoc character, in that they represented specific types of behaviours but were hardly robust to changes, however small, in their characterization. Partly for this reason, the optimization model retained its prominence until very recently.