Economist Paul Milgrom is celebrated for his Nobel Prize-winning work on auction theory and design. But he has published a wide range of other innovative, influential research throughout his career – including a book and articles emerging from his 1991-92 CASBS fellowship. Gani Aldashev (CASBS fellow, 2024-25) engages Milgrom on highlights of this often-collaborative or cross-disciplinary work on organizational behavior, the institutional roots of trust and cooperation, social choice for environmental policy, and more.
Economist Paul Milgrom is celebrated for his Nobel Prize-winning work on auction theory and design. But he has published a wide range of other innovative, influential research throughout his career – including a book and articles emerging from his 1991-92 CASBS fellowship. Gani Aldashev (CASBS fellow, 2024-25) engages Milgrom on highlights of this often-collaborative or cross-disciplinary work on organizational behavior, the institutional roots of trust and cooperation, social choice for environmental policy, and more.
PAUL MILGROM: Stanford faculty page | Personal website | Nobel Prize page | Nobel bio | Wikipedia page| CASBS page |
Gani Aldashev: Georgetown faculty page | CASBS page | Google Scholar page |
PAUL MILGROM WORKS REFERENCED IN THIS EPISODE:
Economics, Organization, and Management (Prentice Hall, 1992), coauthored with John Roberts (CASBS fellow, 1991-92)
"Multitask Principal-Agent Analyses: Incentive Contracts, Asset Ownership, and Job Design," The Journal of Law, Economics, and Organization (1991), coauthored with Bengt Holmstrom
"Complementarities and Fit Strategy, Structure, and Organizational Change in Manufacturing," Journal of Accounting and Economics (1995), coauthored with John Roberts
"Complementarities, Momentum, and the Evolution of Modern Manufacturing," The American Economic Review (1991), coauthored with Yingyi Qian, John Roberts
"Complementarities and Systems: Understanding Japanese Economic Organization," Estudios Economicos (1994), coauthored with John Roberts
"The Role of Institutions in the Revival of Trade: The Law Merchant, Private Judges, and the Champagne Fairs," Economics & Politics (1990), coauthored with Douglass North (CASBS fellow, 1987-88) and Barry Weingast (CASBS fellow, 1993-94)
"Coordination, Commitment and Enforcement: The Case of the Merchant Guild," Journal of Political Economy (1994), coauthored with Avner Greif (CASBS fellow, 1993-94), Barry Weingast
"Is Sympathy an Economic Value? Philosophy, Economics, and the Contingent Valuation Method," in Contingent Valuation: A Critical Assessment, J.A. Hausman, ed. (Elsevier, 1993)
"Kenneth Arrow's Last Theorem," Journal of Mechanism and Institution Design (2024)
Other works referenced in this episode:
Oliver Williamson, The Economic Institutions of Capitalism: Firms, Markets, Relational Contracting (Mcmillan, 1985). Much of this book was written at CASBS during Williamson's 1977-78 CASBS fellowship.
Works emerging from Milgrom's CASBS fellowships
Milgrom's collaborations with, intellectual interactions with, or responses to other Nobel Prize winners in this episode:
Narrator: From the Center for Advanced Study in the Behavioral Sciences at Stanford University, this is Human Centered.
In 2020, Paul Milgrom shared the Nobel Prize with his Stanford colleague, Robert Wilson, for their pioneering work on improving auction theory and designing new practical auction formats, which have benefited buyers, sellers, taxpayers, and societies around the world. Anyone interviewing Paul would be remiss not to engage him on this celebrated work, and were no different on this show. But few may know that Milgrom has twice been a CASBS Fellow, first in 1991-92, and again in 1998-99, and that his time spent here on the Hill yielded a trove of collaborative and cross-disciplinary work that remains influential.
Today on Human-Centered, a conversation with Nobel Laureate Paul Milgrom, the Shirley and Leonard Elie Professor of Humanities and Sciences in the Department of Economics at Stanford University. He's been at Stanford since 1987, and before that, held appointments at Yale and Northwestern universities. He is an elected member of the National Academy of Sciences, the American Academy of Arts and Sciences, the Econometric Society, and was named a distinguished fellow of the American Economics Association. In addition to the Nobel Prize in Economic Sciences, some of the other honors he's received include the John J. Cardy Award for the Advancement of Science and the BBVA Frontiers of Knowledge Prize. Milgram's conversation partner is Gani Aldashev, a 2024-25 CASBIS Fellow and Professor of International Economics at Georgetown University in Qatar.
The two discussed the innovative work that emerged from Milgram's time at CASBIS. For example, with John Roberts, he improved a classic Oliver Williamson book, written at CASBIS incidentally, through understanding the behavior of organizations in terms of interconnected systems. His joint game theory and historical work with Douglass North, Barry Weingast, and Avner Greif on the reputation-based and institutional roots of trust and cooperation. And an economic philosophy paper born out of an infamous environmental catastrophe, which leads to a poignant discussion of a 2024 Milgram article on social choice for environmental policy that realizes ideas Kenneth Arrow shared with him during Arrow's last week of life.
We'll tally up the long list of nobelis he's engaged with over the years in the episode notes, but for now, let's listen.
Gani Aldashev: Professor Milgrom, welcome. It's a great pleasure and honor to have you with us. You were a CASBS Fellow twice in 1991-92, and then again in 1998-99.
I'd like to start with the amazing work that you did during your first fellowship year that focused mainly on organizational economics. So your book on organizational economics, jointly written with John Roberts, developed to a large extent during your time at CASBS in 1991-92, and it represents a remarkable synthesis of decades of earlier work. Could you please tell us what was your vision for that project?
Paul Milgrom: Well, John and I, John Roberts, that is, and I, had read a book by Oliver Williamson, who Oliver had been my colleague at Yale from 1985-1987. His book, The Economic Institutions of Capitalism, and it was quite an exciting book in terms of the questions that it asked. But it didn't meet the standards of precision that John and I were used to.
And we decided we really wanted to address the same questions, but with greater precision and more deeply. So that was part of it, on my side anyway. And on John's side, he was teaching at the Stanford Graduate School of Business, and teaching cases.
And there was some really puzzling behavior in the cases that he was teaching, that we were trying to understand from an economic perspective. And we wound up deciding that the way to understand a lot of behavior was in terms of systems. That instead of looking at things one at a time, you needed to consider things together as part of a system.
Gani Aldashev: The book is impressive, and it had a lasting impact both in terms of research, it has collected many, many citations, but also as a textbook. So how do you think it changed the way the field of org-econ is taught and understood?
Paul Milgrom: Well, I think it had its biggest impact in law schools rather than in economics departments, judging from where it was adopted, some in business schools as well. We were trying to understand the patterns of behavior we saw, and that we saw that is, and we had a new set of explanations, and they seemed to apply both historically to the way firms had behaved over time. We went back to the history of the Hudson Bay Company as it developed in Canada, all the way up to what was the contemporary puzzles at the time about why Toyota was so successful compared to the large North American auto companies, General Motors and such, which had obvious scale advantages. And we learned something from the way we approached it.
Gani Aldashev: One other key aspect of problems within organizations that you also mentioned in the book, but is also an important strand of your research, was the so-called multitasking problem. And your now classic paper, Multitasking and Incentive Design, jointly written with Bengt Holmstrom, another Nobel Prize laureate, has had lasting influence. Why do you think this idea resonated so widely?
Paul Milgrom: I think it's critical to understanding incentives. When Bengt Holmstrom and I were looking at individual incentives, we looked at, among other things, insurance agents that represented multiple companies, and understood that if you increased the commission rate that an agent got from one company, the sales for another company would fall. They would tend to shift effort from one to the other.
And then we began to wonder, well, why isn't this just like any other producer of anything? Why isn't this just a matter of price theory? That you tend to produce more of the things that have relatively high prices, and less of the things that have relatively low prices.
And it turns out that that's really widely useful in thinking about human behavior.
Gani Aldashev: And do you think that this idea is being applied today in other aspects of organizations, or maybe underapplied in your view?
Paul Milgrom: Well, yes, certainly. I mean, I think it's everywhere. I think we had written specifically then about incentives for teachers in which if you try to incentivize teachers to improve test scores, they might neglect development of social skills, for example, or other things that were harder to measure.
We were looking at when people had a multi-dimensional task, if you incentivize strongly one dimension of the task, how you might harm performance on other dimensions. And we still see that everywhere. Anytime that there's two kinds of output or output in quality, or especially where one thing is easy to measure and the other is not, you'll get distortions in all the behaviors from incentives that you attempt to apply.
Gani Aldashev: Within the field of organizational economics, you've also written several highly influential papers about complementarities in modern manufacturing. Could you please tell us briefly about this idea, how it came about, and how it's related also to observations that you did while you were during your CASBS year here?
Paul Milgrom: Yeah, well, during the CASBS year, John and I made a couple of visits, one to Toyota factories in Japan, and also to some of the suppliers in Kentucky, and also to the new Toyota plant, which had just been built in Kentucky, to understand why it was so successful, and why its system was so different from the system that General Motors and Ford and Chrysler had been using at the time. The different in many ways, and there was clearly a lot of confusion about that. People were saying, well, you know, personnel management matters.
These guys sing songs together in the morning, as though, you know, somehow or another, that were the secret. And as we looked at it, we realized that we were looking at a different system. The songs weren't an important part of the thing that's a system.
That was Japanese culture. But the use of continuous improvement required workers who could change the jobs they were doing, which required more education and more training for the workers, required machines that were more flexible, as if you were going to be changing the flow of work. And those things all fit together.
If you were not going to be changing your machines, you didn't need flexible workers. If your workers were inflexible, it wasn't going to be a good idea to be changing your processes all the time. That there were certain things that just went together.
And changing processes all the time also fit well with having more frequent new product introductions. So it affected the way you managed workers, the way you managed the factory floor, the kinds of products that you offered. They were all interconnected systemically.
And that was really a useful way to think about the differences among firms and why in some eras one firm and one firm's organization would work better than another. I think this was an era at the time when robotics were emerging, when worker training had improved, and it was time for a new system, a system that worked better in that current environment. And Japan had leapfrogged to the US at that time.
Gani Aldashev: What can this work in your view teach us about organization of work in today's increasingly digital interconnected economy?
Paul Milgrom: Well, that's a great question. I'm not quite sure exactly what it entails. I do think clearly we are going to have more value to workers who can work with AI.
I think AI is, as it takes over tasks or changes the nature of work, having flexible workers becomes even more important. But the specific kinds of tasks that the workers can do, the ones that are likely not to be replaceable, are the ones you're going to want to emphasize in your workforce and train more intensively. I really haven't thought through in detail the full set of implications, but as the environment changes, the kinds of training, the kinds of machines, the kinds of products that a company produces or uses will change, and that we have to think in terms of systems to understand the full effects.
Gani Aldashev: So we're likely to see complementarities at play again.
Paul Milgrom: I do think so, yes.
Gani Aldashev: I would like to go back for a second to your experience of observing firms and organizations directly. So this is a tradition that goes back perhaps to Ronald Coase, yet another CASBS fellow. And could you please share some of the details of that experience, and maybe give an example of a situation where direct observation challenged your assumptions, or led to a breakthrough in your thinking?
Paul Milgrom: Well, when we went to Kentucky to visit Toyota and its suppliers, there were a number of surprises, things that I hadn't expected to see. There were Toyota engineers on the floor of the supplier's workplace, talking to them about how they should be organizing their work so that it fit well with Toyota's procedures. That was a surprise.
The need for the firms to be in close physical proximity, that was associated with the very low inventories, which Toyota had, and which are widely part of manufacturing today. When you get an order, you want it to be delivered quickly. You keep low inventories of inputs.
That, of course, was also complementary to flexibility. It meant that you could change your product and not have wasted inventories sitting around. So, the pieces that fit together were much more numerous than I had expected. It was part of what inspired our thinking about complementarity and its role in organizations.
Gani Aldashev: But this type of direct observation is something that the field of economics doesn't always prioritize in current type of research. Why do you think this is the case, and how is it different from the current type of empirical work that is done in economics?
Paul Milgrom: Yeah, I think the kind of observations you make depends on the kinds of questions you want to ask and where you want to search for answers. I think in economics today, there is a lot of emphasis in empirical economics on policy analysis. We want to know whether a child education program, for example, leads to improved performance by children later in life.
Well, if you just observe the child, then you had to wait years to see what happened later in life. That would be very hard to do. It would take many years of observation to do that.
And the observations you made, chances are, wouldn't be causal. If you want to know who participates in an early childhood education program, it's not a random selection of children. More likely to be children from activist families or wealthier families who do that.
And what's happened, if that's your concern and you want to know what the causal effects are of early childhood education, you need to take a much more sophisticated approach. You need to take into account factors that might lead to random assignment of children. The economists have a variety of methods that they use that allow them to do causal inference.
But mostly that's not just about going out and making observations of the firms who are willing to host you. I was collecting a different kind of information that was suitable to the questions I wanted to ask, which were different from questions about how do we treat income inequality or climate change or early childhood education
Gani Aldashev: So it's question dependent, essentially.
Paul Milgrom: I think observation has to be question dependent, yes.
Gani Aldashev: Going back to your CASBS experience, CASBS is par excellence a place of interdisciplinary interactions and collaborations. I'd like to talk now about another strand of your work, which is also highly interdisciplinary, but in collaboration with other fields, such as history, law, and political science. So your work on multilateral reputation mechanisms with Barry Weingast, also CASBS fellow, Doug North, another CASBS fellow, and Avner Greif, it's an interaction with political scientists and economic historians.
It's a very curious collaboration. My understanding that it touches on the institutional roots of trust and cooperation. Could you please tell us more about how it is to work with such a diverse team of people and a little bit more about this work?
Paul Milgrom: Well, sure. Let me roll that back a little bit. One of the things that happened to me at CASBS is exposure to the research that was going on around me.
We used to have these, I think they were Thursday night seminars where we would listen to specialists in different fields talk about their research. And there were quite a few curious things. So I became curious of the different perspectives that would be offered from the different fields.
And it seems to me that the reputation was one of those things. It was a word that was used in different fields to mean different things in economics and in game theory. There was a reputation theory in the theory of repeated games. Very narrow reputation theory. And in the other social sciences, the word was used more widely.
And so I remember sitting down in a group meeting at some point and talking with Douglass North and Barry Weingast, who were there, about the reputations and the role they played. And Douglass explained to me what he thought was going on, specifically in the Champagne Faires before the rise of the state in France and throughout Europe. And I told him I thought that was a very incomplete theory, because he was just looking at the incentives of one guy and not understanding how the whole system worked. And he said, well, why don't you tell me what you think a complete theory would look like?
I had tried to explain it, but they asked me to go home and write something down. So I did. I went home and I wrote down something that would be a complete theory to explain what was going on in the Champagne Faires. And when I brought it back, they were very excited and said a lot of this theory is exactly what did happen at the Champagne Faires. There were details that needed to be adjusted to be consistent with the historical evidence. But we put together a theory of, a complete theory of how reputation could have functioned in the Champagne Faires.
And it was quite exciting to work with them and to match what was really going on and do it in a way that was complete from the point of view of a game theorist.
Gani Aldashev: It's an impressive interaction between historian on one side and game theorist on the other. How do you think we learn from each other in that interaction? Who learns from whom and in which way?
Paul Milgrom: Well in that first interaction, I must say Barry Weingast played a really central role. He knew a lot of the history. He was a good friend and previous co-author of Douglas North. And he also knows a fair amount of game theory. So he lubricated that interaction a lot. And as for the rest, we were both quite excited.
I was excited to have a new game theoretic based model that was important for understanding an important fact in Douglas' conception of economic history. And he was also, you know, excited by once he understood what I meant by a complete explanation, one that took into account not just the incentives of one party not to cheat because he'd be punished in the future, but the incentives of the punisher, the incentives of those who would judge him, the incentives of the community as a whole. Once he understood that level of conception, he was excited by that. So we had a really good time together.
Gani Aldashev: I think it was Mark Twain who said that history doesn't repeat itself, but it rhymes. So, do you think that work has repercussions for current day, in particular in the period of rising populism and institutional distrust? Do you think, or what do you think is the role for the reputation-based mechanisms? Do they apply, and if so, under which conditions?
Paul Milgrom: Yes, I think reputation-based mechanisms still apply. I think it gives us a way to think about a lot of what's going on in society. Clearly, within a cohesive group, it's easier to maintain a reputation.
Your behavior, when you cheat in a close-knit group, in a family, for example, is observed by many people, and community enforcement is possible. When we have fragmentation, the ability to enforce your behavior across, to have reputation mechanisms, enforce good behavior across communities is poorer. It doesn't work very well, unless the community itself benefits in the kind of work that I did with Avner Greif and Barry Weingast.
Communities of traders who lived in port cities would enforce their rules against other members of their own community in order to create a good reputation for the community. And that would allow somebody from a distant port to trust a trader who was a member of the, you know, of the club of the guild, as it were, for, from a trading community to trust them, even if they didn't expect to see that trader again, because they could demand recompense for any bad behavior from that trader's community. I think, you know, that sort of thing is still possible.
We have to take a look at the specific relationships within and among communities to understand where reputation can be effective.
Gani Aldashev: I see. So, does it imply that the decline in social capital might make these mechanisms less strong? Or, in the place where trust in social capital is weaker, these mechanisms are less biting?
Paul Milgrom: The groups to which they apply will be fewer. I think if, what you mean by less biting is that for any random group of people, it will be less likely that they can rely on a mechanism like that. I think, yes, that's true.
And these days, we have new ways of studying that with network theory, where we look at connections and where the connections are dense and where they're sparse, and are able to assess that from new theoretical perspectives.
Gani Aldashev: This work is built on game theoretic models. And some of your earlier work explores certain key foundational issues in game theory. And here, I think about your work on reputation building in games. And game theory is often seen as a toolkit for economists, but your work seems to argue for its relevance much beyond economics, across probably all social sciences. How do you see game theory in reaching disciplines like political science, sociology or anthropology?
Paul Milgrom: Well, I think the answer is different for the three. I think within political science, there's already a lot of game theory used. People recognize self-interested behavior as being fundamental in thinking about politics, and about strategic behavior taking into account strategic responses as being fundamental.
So, I think game theory has already penetrated many, many political science analyses. In these other analyses, I think there remains a question of whether it's strategic behavior or something else that's motivating behavior, whether, you know, I don't know, emotions or family relationships or culture that is driving behavior. And what game theory does is it provides an alternative lens.
You look at a behavior, and somebody says, well, they're doing that for cultural reasons. And you ask, well, might they be doing it for strategic reasons? And game theory provides you the tools to look at that, say, what might their incentives be? Did they have the information they need to do that in that strategic way? Who are the other players? How might that have influenced their incentives? What was the history of that? Were there histories of deviations from community norms being punished? How are they punished? And how do we understand that in terms of the incentives of the parties?
So it gives you a new set of questions to ask. And the two kinds of analyses can also interact. It can be that culture does govern behavior up to a point, up to the point where the incentives become too strong. And maybe you see that by adding a game theoretic analysis on top of an analysis that looks at other causes.
I remember when I was a student, and my first course of game theory, at the very first lecture, there was a discussion about rationality. And one statement that stayed in my mind is that rationality as understood in game theory is a different kind of rationality as compared to the first course of microeconomics. In particular, it's about strategic rationality, which is taken into account also the response of the other person and thinking about the other person's perspective, some kind of allocentric thinking.
Gani Aldashev: Do you think this kind of thinking is highly relevant also for other disciplines such as sociology and anthropology? You mentioned this already. Do you think this...
Paul Milgrom: It seems to definitely be present in political science, a bit less so in the other social sciences.
So I think that when you and I were learning game theory, we're kind of old guys here, the kinds of things you've said would have been said in classrooms. When I teach game theory these days, I don't describe it in quite the same way that you have just described it. But game theoretic forces don't emerge only from I think about what you did.
Sometimes they emerge because I learn what you will do if I behave in a certain way. And sometimes they apply not just to humans. I'm working, for example, on pricing on the internet or the way bidders bid in auctions on the internet.
And those bidders happen to be machines that are actually learning. So we see that sometimes learners who show no awareness at all of who their competitors are just become aware of what happens when they've acted in a certain way. We'll find that game theoretic analyses work well predicting their behavior, too.
And sometimes not, by the way. It turns out that some of these systems are unstable under learning and that you can't learn the rational behavior because the environment is changing too rapidly. Others are changing right while you're changing.
But I'm heterodox in the way I teach the applications of game theory these days. I think that it can depend on the game, and I try to explain how to my students. And it can depend on the setting and who the players are.
And I think in the future we're likely to have games that are played by machines against humans. You know, we'll also have where you're bidding in something and somebody else is using an AI. And you know, they will be wondering how to apply game theory in those settings as well.
Gani Aldashev: Amazing. So it's a field that's still in the making.
Paul Milgrom: Oh, it is. Yes.
Gani Aldashev: Some of your most impressive technical work in economics has to do with generalizing existing theoretical models. And here I think about your papers with Chris Shannon and other colleagues on Monotone Comparative Statics and Supermodularity. You've greatly contributed to the generalization of these models and extending their applicability beyond narrow functional form assumptions. Why do you think this type of theoretical work is so important?
Paul Milgrom: Yeah, I think that when we are looking at theories, almost always, when I'm applying theories, I do apply theories, as you know. I have a consulting business as well. I want to know what the scope of these theories are, which assumptions are important, how general are the conclusions, and knowing the conclusions in their most general form, and knowing that certain assumptions that were used in the derivation of the first result are not important.
They can be quite helpful to me in figuring out whether the theory is applicable in any new situation. So I think, you know, maximum generality for a theoretical conclusion is quite important in understanding its scope and understanding when it applies.
Gani Aldashev: This work of deepening and generalizing models is painstaking work that takes often many years. And do you see there is a tension between the pressure for novelty, especially empirical, in academic publishing, and the more, the quieter value of generalizing foundational models?
Paul Milgrom: So I think that that distinction is exaggerated. I think that two things about the, you've mentioned the empirical side. On the empirical side, when you test a theory, knowing its scope allows you to figure out how you can test it, what data might apply, so that you can test whether the theory is effective.
And knowing its limits might also help you do what you're calling theoretical novelty. You might say, in the classical theories of microeconomics, we need convexity assumptions for market clearing prices to exist. And then the theoretical novelty might be, how does an economy work when there are no market clearing prices?
But then you'd want to know when there are no market clearing prices. The generality helps raise the question and tell you where the examples might lie and help you explore novel questions that arise outside the scope of current theory.
Gani Aldashev: So during your CASBS year, you seem to have also written a paper about philosophy, economics, and contingent valuation method, which came out as an interaction with a philosopher, right? Cicilla Boca, if I remember well?
Paul Milgrom: So that was related to the Exxon Valdez oil spill and the question of who was harmed. And are you harmed because you feel bad about the death of fish in the Gulf, near Alaska? And how do we think about that? And so that was the issue that came up.
It was not that long after the Exxon Valdez oil spill. And I was just very curious about what the right way to think about that was. You know, what if I buy you a nice new tie, and it makes him feel good, does that make the tie more valuable?
Should society think that the tie is more valuable? What if we like it better than you like it? How do we decide how much that tie is worth when it has to do with my concern about what happens to somebody else or to something else?
And I don't think that that issue has been resolved yet, but I noted how, you know, the what about if he cares about you and you like the fact that I gave him a tie? Have I benefited him by gifting the tie? Those were the kinds of issues that I was addressing.
Gani Aldashev: And it's actually quite far from the other type of work that you've been doing, right? So it's almost applied philosophy.
Paul Milgrom: It came up because of the oil spill, actually, and where I read some things that I thought made no sense, and I just explored what their logical consequences were, and then talked about what you could include without running into paradox.
Gani Aldashev: I'd like now to turn to the work for which you received the Nobel Prize in 2020, namely, you work on auctions. And it's impressive that you work on auctions has not only reshaped theory, but also had transformative real-world impact, in particular, in spectrum allocation auctions. What has it been like to see your ideas implemented at scale, and what have been the biggest surprises or lessons in working with policymakers?
Paul Milgrom: Well, I must say, it's been very exciting and scary. I started off as someone who was working on theory and not expecting that my ideas would just be taken and put to work in a multi-billion-dollar application the first time they were used. And that was a little bit frightening, but also quite exciting.
And I think that was pretty surprising. It was surprising to me that somebody with as little experience as I had would have their ideas just adopted and put to use. But it wasn't the only time it happened.
It turns out, as I discovered, that I was mostly approached only for very hard problems, problems that nobody else had a clue how to do. And I learned that if I had the only idea in town, then they were going to do that idea because they didn't have anything else to do. And I think the thing I found so surprising is that in, you know, in transactions where there were tens of billions of dollars at stake, people would just take my ideas and use them almost without change. And they did that, I came to understand, because they had no clue what else to do. These were new kinds of applications. So it was very exciting and very scary.
And I, well, I'll remember, I have one story to tell you that may go with this. It was bidding in a radio spectrum auction in 2007 or 2006. And I had recommended to a bidder that I was advising to make a jump bid, the largest jump bid in the history of the spectrum auctions, a $750 million jump bid, which was a very large amount of money even today for a company.
It was even more for my client in 2006 or 2007. And the head of the bidding team said, well, I'm going to need to talk to the chairman of the Board of Finance. I'm going to need to talk to the CEO.
And I had recommended that we plan to make this bid on a Monday morning. And I didn't hear, and I didn't hear. And on Sunday night, I got a call from the executive vice president in charge of the bidding team. He says, we're going to do it. You better be right. Which gives you a pretty good idea how I felt, right?
Gani Aldashev: Oh my God.
Paul Milgrom: And we were right, by the way. It worked beautifully, exactly as we hoped it would.
Gani Aldashev: That's amazing.
Paul Milgrom: Yeah.
Gani Aldashev: But working with messy and imperfect institutional contexts in real world auctions also has its challenges, right? What were the biggest challenges that you had?
Paul Milgrom: Well, there were many. The theories that we wrote down on paper were never exactly right. They were always simplified.
They always omitted some factors that were important, which created doubt in our own minds, as well as in my mind and my team's mind, as well as in the minds of the people who were listening. There was also the conservatism in some cases of the people who were taking advice. They didn't want to be the first to try something new, which is why, as I said to you earlier, the big impacts I had is where there was no alternative.
Nobody else had any idea, and they decided to go with it. But occasionally, as I mentioned in connection with this jump bid, there was the alternative of doing something cautious. And I guess perhaps it's a surprise that such a daring strategy, as we had proposed, was adopted.
It was the largest investment the company had ever made, and it was huge. And they did it, and it worked, and everybody applauded afterwards. I got a t-shirt that said that they gave me, I don't remember exactly, but they made up a t-shirt about this event that they gave me back then.
That was my reward.
Gani Aldashev: You mentioned earlier that your consulting work also creates you some interesting problems to study. So there is a two-way interaction between the applied work and theoretical work. How does it work in your case?
Paul Milgrom: It depends how much time we have. In the US. Incentive Auction, the Broadcast Incentive Auction, I was hired in 2011, I think, and the auction happened five years later, in 2016, and we had the opportunity to do a lot of research to run simulations, to do theoretical research, to do computational tests, to see what was possible and what could be done.
And it was motivated by creating what was $30 billion in transactions. It was a lot of money that was transacted in that auction. And yeah, the research was motivated by that, and it was exciting research.
I'm looking forward to seeing if the same research has other applications. That was a pretty big one, but it may not be the last one, there may be more. After that five years, the team continued to work on simulations on whether there were better ways of doing what we had done.
And we discovered ways that were better. We could have saved the government over a billion dollars. Billion and a half dollars, I think, was the number we came up with, had we handled some detail differently.
So, you know, I've continued to look afterwards. It's sometimes very hard to test counterfactuals to say what if. If I had done it differently, what would have happened?
But in this case, we had a simulation model we built, and we continued to work on it for years afterwards, and discovered, yep, we could have done it differently. And it would have been billions, or at least a billion and a half, better for the government than what we did. And what can you say? You know, we did what we could in five years.
Gani Aldashev: But sometimes the horizons that you have are also much shorter, right? In some of the consulting work, the deadlines are yesterday.
Paul Milgrom: Yeah, that's right. Sometimes they're much shorter. The very first one, the very first auction, it was a matter of just a few months.
And I remember, again, I had never done any, you know, practical work like this before. And I remember it was in the era when I was hired by Pacific Bell, which was the local telephone company here in California and Oregon at the time. And Nevada, they also covered.
And they just had a few months, and the FCC just had a few months to decide what auction form it was going to use. And yeah, and I remember telling them, that's not enough time for me to do anything. And they said, well, look at what the government has proposed. And I looked at what the government had proposed, and I said, well, I don't know what the best thing to do is, but I can do better than that. And that was it. So I made a proposal.
Bob Wilson and I made this proposal. And it was adopted, because it was way better than what the government had had in mind, and much better for my client, by the way, than what the government had in mind. And the auction ran. And that was what got copied all over the world. I mean, it got adopted in basically all the English-speaking countries. It got adopted in Canada and Australia and the UK and New Zealand. All of them adopted this. And some non-English-speaking countries as well in Germany and elsewhere.
Gani Aldashev: So sometimes an improvement on the existing is already good enough.
Paul Milgrom: Yeah, sometimes it's already good enough. Yes.
Gani Aldashev: Is there a big remaining question or questions that you still like to see answered in the coming several years?
Paul Milgrom: Oh my, there are so many. And they range from purely technical ones at the boundaries between economics and computer science, for example, to the big social science questions for economists. We care about polarization and climate change and income inequality and lots of questions to ask in those domains.
I mostly have been going back to my roots in theory working on somewhat smaller questions, but ones that I think are still pretty important, where I can have an impact in my remaining years. Working on, as I said, the intersection between computer science and economics, which I think raises lots of interesting smaller questions. The most recent paper that I published was called Kenneth Arrow's Last Theorem.
You guys, I don't know whether Ken was, was he here? Kenneth was up here, one of the great thinkers of the 20th century. And I had visited Ken multiple times while he was dying, at the last days of his life.
He was 95 years old when he passed. And he was thinking about, well, the two things he wanted to talk about, the only two things he wanted to talk about when I visited him, were income inequality and climate change. And he sketched to me ideas that he had about social choice over climate policy.
And he told me that he had begun his research career. The first thing he was famous for was Arrow's Impossibility Theorem about the impossibility, essentially, of rational decision making about public policies. And he said he would end his career if he had time with a new impossibility theorem about thinking about environmental policy.
And he sketched the ideas to me. And then he died. And I did nothing for several years.
And then I decided, you know, those deserve to be written down. So I published a paper just some months ago entitled Kenneth Arrow's Last Theorem, in which I described an impossibility theorem for environmental choice. And Ken's first impossibility theorem says, well, the only way to make decisions rationally that weights the preferences of people in society is to have a dictator.
And this theorem has a very similar conclusion, that the only way to do things is to have a dictator, but the dictator will wind up being whoever is the most patient person. That in order to have an efficient policy, the only way to do that was to respect only the environmental preferences of the most patient person. So, that's something I've just finished working on.
Gani Aldashev: Thank you so much and all the very best for your future exciting projects.
Paul Milgrom: Well, thank you. It's been a pleasure talking to you.
Narrator: That was Paul Milgrom in conversation with Gani Aldashev. As always, you can follow us online or in your podcast app of choice. And if you're interested in learning more about the center's people, projects and rich history, you can visit our website at casbs.stanford.edu.
Until next time, from everyone at CASBS and the Human Centered team, thanks for listening.