Milgrom on Auctions, Theorems, and the practice of Market Design, in the AER

 The latest issue of the AER publishes a version of Paul Milgrom's Nobel lecture:

Auction Research Evolving: Theorems and Market Designs  By Paul Milgrom

American Economic Review 2021, 111(5): 1383–1405   https://doi.org/10.1257/aer.111.5.1383

Here are a few of the introductory paragraphs:

"Game-theoretic modeling of auctions began in the 1960s with a pair of seminal papers by William Vickrey (1961, 1962) and the brilliant but unpublished doctoral dissertation of Armando Ortega-Reichert (1968). Robert Wilson (1977, 1979) became the next important contributor to auction theory research and, as Wilson’s student, I was inspired to make auctions and bidding the subject of my doctoral dissertation.

...

"Most of my work published in academic journals is theoretical, proving theorems about the properties of abstract models, but developing and participating in real-world mechanisms requires more than that. Two important lessons that I learned from working on high-stakes auctions are that they operate in an almost infinite variety of contexts, and that this variety is the reason for the paradoxical importance of including unrealistic assumptions in models built to understand and illuminate reality. No single set of assumptions is adequate to describe all the various settings in which auctions are used, and too much specificity in models can blind the analyst to important general insights.

...

"Why do economists rely on such unrealistic assumptions? It is because a well-chosen simplification can remove the dust and smoke that obscures our view of the workings of economic forces. Although we celebrate the resulting theorems for the insights they deliver, we can apply them successfully only by being vigilant, working hard to understand not just the insights that simplified analyses provide but also how the designs and rule choices they inspire must be adapted to withstand the dust and smoke and also the much larger disturbances of the particular worlds in which the mechanisms will operate."

And here are the concluding paragraphs:

"Auction theory has changed substantially since I made my first studies in what were still its early days. Although the “unrealistic” models of those times have proved their worth in guiding practical auction designs, some of that guidance was off point. In my own work, this showed up in the traditional analysis of the exposure problem. Despite the theoretical worst-case conclusion that exposure problems are intractable, we found that they could sometimes be quite manageable in practice.

"For the future, simulations and computational methods are likely to be increasingly important. Yet, it still takes theory to understand problems and the scope of proposed solutions. The time has come for old methods and new to work hand in hand."