SHARPE Interview at the home of William Sharpe (edited for clarity and readability) This is the second interview in our series with founding contributors to the field of finance. The title of the series is “Masters of Finance”, and I am pleased to announce that our second master of finance is Dr. William Sharpe. Technical support for the taping of this interview is being provided by Dimensional Fund Advisors. We thank them for their assistance. We are taping today in the home of Dr. Sharpe and his wife Kathryn. And from their lovely home in Carmel, California, I bid a welcome to Dr. Sharpe. Sharpe: It is a great pleasure to be here with you. You didn’t run away when we showed up. Buser: Sharpe: True. Buser: Although Dr. Sharpe continues to contribute to a long and distinguished career, his name will always be identified with the Capital Asset Pricing Model, He has won a number of awards for his many contributions. In 1980, he was elected president of the American Finance Association. In that same year he received an award from the American Assembly of Collegiate Business Schools for outstanding contributions in business education. In 1989, the Financial Analysts Federation honored Dr. Sharpe with the Nicholas Molodovsky Award for outstanding contributions to the finance profession. In 1990, Dr. Sharpe shared the Nobel Prize with Harry Markowitz and Merton Miller. Currently, Dr. Sharpe is Professor Emeritus at Stanford. In fact he is a double Professor Emeritus at Stanford Perhaps he will explain what that means in today’s interview. In addition, Dr. Sharpe remains active in a business of his own and in consulting activities. Q. Let me begin with an obvious question. What is it like to win the Nobel Prize? A. Well as I and many of my friends say, if anybody offers you one, you should take it. It is definitely worth having. It is incredible. It’s an unbelievably heady experience, as you can well imagine. You hear from people that you haven’t heard from for a very long time. 1 Q. Any asking for loans? A. Well, you get requests for some really quite remarkable things. You get letters from time to time from people who want you to help them because their brains were implanted with electrodes while they were on airline flights, and they were forced against their will to do various things. Some real kooks come out of the woodwork. But you go to Stockholm, and you are treated like royalty. It is intellectually, emotionally, and in virtually every sense, an experience that is outside the range of comprehension for an ordinary academic, such as myself. Q. Do you recall where you were when you first heard that you’d won the prize? A. Oh yes. Everyone does. We were at an industry and academic conference in Tucson, Arizona. The phone rang about 4:00 in the morning. I had not even been paying attention to the fact that this was about the time for the announcement. The phone rang in the middle of the night There was someone on the other end saying something with a bit of an accent. There had been a fellow trying to get me to give a talk in Belgium, and he seemed to have trouble with time zone differences. He had already called several times at odd hours, and I thought it’s him again. When I realized it wasn’t, I next thought somebody is playing a trick on me. I then very quickly realized it almost certainly wasn’t. So after the call and jumping up and down and waking up Kathy we turned on CNN. In about five minutes the first announcement came. It had about 60 percent of the information right. Then every five minutes they repeated it, and it got a little bit more accurate. We called for room service, sat on the balcony outside our room, and watched the sunrise over Tucson. Then all hell broke loose. Q. Were you aware that you were under consideration? A. No I wasn’t. There are people who try to follow these things and make bets on who might get it each year. I had never been in that particular mode. I believe Harry may have told you this too, but I think Harry and I, and perhaps others in finance, felt that given who had already been awarded the prize, and the work for which they had been given it, the committee probably was not going to include financial economics within the purview of economics for the purposes 2 of the prize. So I , t least felt that my field wasn’t going to be in the running. So I hadn’t really worried about it. Q. What were your thoughts upon learning that you would share the award with Harry Markowitz and Merton Miller? A. Well needless to say, sharing it with Harry was particularly wonderful for me since it was Harry’s work that got me into finance and started me on the path that I’ve taken. So to be able to share it with Harry was a very, very special experience. Merton Miller’s work was in a somewhat different area than my main line of inquiry, but Mert and I knew each other over the years, exchanged papers back and forth, and of course to be associated with Merton Miller was also a very heady experience. Q. You mentioned that Dr. Markowitz helped in your career, could we back up a little bit and tell us how you even got to that point? What was your formal training, starting at the earliest relevant point? A. I started as an undergraduate in a pre-med course because my mother wanted me to be a doctor. Then I switched to business at UCLA. The first two courses were accounting and economics. I hated accounting, at least as it was taught in that course, but I loved economics. I did not know anything about economics until I took the course. So I switched to an economics major. I got a bachelors and a masters in economics. Then I went into the service, and when I came back, I went to work at the Rand Corporation doing research. I also started a PhD at UCLA and worked in the finance area with Fred Weston . He introduced me to Harry’s work, which I fell in love with. I thought this was just wonderful, beautiful stuff. So that’s how I became conversant with the work. I started a dissertation on transfer pricing. Jack Hirshleifer came to UCLA, and since I was starting with his work, it was decided that he should be my chair. He looked at what I thought was a half-finished dissertation, and he said he just didn’t think there was much there. So I went to Harry, who was at Rand, and I asked if maybe there was something I could build on from his work. In some sense, the rest is history. Harry wasn’t on the faculty at UCLA, but I worked out an arrangement with Armen Alchian to work with Harry and have Armen be my dissertation chair. 3 Q. When I meet Dr. Hirshleifer, I will have to thank him for pointing you in the direction of finance. A. Well I can tell you I’ve thanked him many, many times. Q. Did it strike you as odd then, or now, that you would ultimately use similar tools from transfer pricing? I assume you had been doing optimization problems with Lagrangian multipliers. Or perhaps I should just ask, what was the mechanism that you had been analyzing in transfer pricing? A. using It’s going to take me a little while to reconstruct. But as I recall, I was a linear programming Input/Output formulation. I was working with Lagrange multipliers, but not in the traditional sense. It was more shadow prices from linear programming. Q. And shadow prices were what I was thinking about. Were you working with duality problems or things like that? A. Exactly. Q. You then worked with Dr. Markowitz, who was also working with linear programming and constraints and— A. Well Harry’s work was all quadratic, and the shadow prices were much closer to Lagrangean’s, or at least Kuhn Tucker conditions. Q. So, there wasn’t, it wasn’t as big of a switch as it might have seemed? A. No, I was in a group that Rand called the Logistics Department. We were doing big linear programming problems and using a lot of the then-new operations research tools. So I was bringing an OR mindset and technology to economic issues. Q. Your first article, which appeared in Management Science, includes an acknowledgment. And over the years I’ve read a lot of acknowledgments. I have even written a few. But I cannot recall one as, as sweet or as kind as your acknowledgment. Do, do you remember it? A. Actually it wasn’t my first article. My first article was in the Naval Research Logistics Quarterly, on a totally different subject. Q. Harry had an earlier article there as well. A. Yes. Q. I stand corrected. In your second article, I wonder if you would read that part of the acknowledgment to Dr. Markowitz. 4 A. I am often accused of putting all my important results in footnotes. Maybe this is one of them. I’ll skip the first sentence. The remaining text, referring initially to “the author” reads “His greatest debt, however, is to Dr. Harry M. Markowitz at the Rand Corporation, with whom he was privileged to have a number of stimulating conversations during the past year, It is no longer possible to segregate the ideas in this paper into those which were his, those which were the author’s, and those which were developed jointly. Suffice it to say that the only accomplishments, which are unquestionably the author’s are first the computer program, and then the writing of the article.” I just noticed that on the other page there is an article by two authors, one of whom is Marty Leibowitz, now a very close friend. Go figure. I was certainly hedging my bets with the footnote. One of the things that has happened, as I’m sure you’re aware, is that financial economics is an old enough field now that there’s a, there’s a cadre of, of younger researchers — and some older researchers — who are doing histories of the field. Among other things they are trying to decide whether or not a particular idea was attributable to Markowitz, or to Roy, or to some monk from the 13th century. So there is a lot of dispute, and I am glad I wrote that because at least there, I don’t claim anything more than I can establish. I definitely wrote the paper. Q. Well it’s a remarkable acknowledgement. Dr. Markowitz thinks that your contribution was a lot more substantial than the footnote indicates. A. Yes he does I think one of the problems that we all have, especially when we work with PhD students, is taking into account the interchange of ideas. And this is true with colleagues as well, whether they are formal coauthors or not, Ideas are ideas. We all know they result from conversations and batting things back and forth. It always amazes me when people say “ I know that was my idea, and it was only my idea”. Unless you’re in a garret somewhere, I don’t know how you can say that. In particular Harry indicated that he thought a simple model of security returns would offer a very good way to solve the mechanical problem of trying to estimate lots of covariances. But he did not realize the great simplification in the problem that would result, beyond the mechanics of to get estimates for all of these co-variances. As for the single index model, or the one factor model itself, it came straight out of Harry’s work. 5 Q. Do you recall your original phrase? You have mentioned single index and single factor, but that wasn’t the phrase you used. A. No it wasn’t. I called it the diagonal model because, if you take advantage of the structure, you can write the covariance matrix as a diagonal matrix with zeroes on the off diagonal and an additional equation At that time, when I was programming algorithms on the IBM 709 every little bit helped computationally. Now it seems almost ludicrous that we worried about these algorithmic simplifications. Now we just solve problems with brute force on our PCs. Q. The new computers are amazing, but Harry indicated that one of his efforts was to try to program [the mean-variance problem] for 25 securities, and he failed. He had to give up. It was too ambitious (Laugh) for the computer technology of that era. He had to settle for a ten security application. A. Yes Q. Well to use a baseball analogy, if your first finance paper was a solid hit, your next effort was clearly a homerun. I am speaking about your 1964 Journal of Finance paper, Capital Asset Prices: a Theory of Market Equilibrium Under Conditions of Risk. How did you come up with that bold idea? A. Well it was almost perfectly natural. It started with Harry’s work. Harry said that if you have estimates or risk, returns and correlations, here is how you should build your portfolio. It was very definitely a normative model : here is how to do portfolio choice, as we call it today. My dissertation had three parts. The first piece was the algorithm you have referenced for solving the portfolio problem under the single index model assumptions in an efficient way. The second piece of was again very much normative in tone, but it had a little bit of a behavioral aspect. Through Fred Weston I made contact with an investment advisor in Los Angeles. He and I talked, and I asked him to think probabilistically. He asked “How?” He knew about bell shaped distributions, so I made some grid paper and got him to draw distributions for 30 or 40 stocks, and to make some other subjective estimates. I then punched all that information on cards and processed it to show him the implications a la Markowitz of his estimates. The portfolios that came out typically had something like 60 percent in a stock called Haloid Xerox. I asked him if he thought they were plausible portfolios, or if he felt there was something wrong somewhere, either with his 6 estimates of the prospects or with the assumptions I and Markowitz had made to get the portfolios. For the third part of the dissertation I did what any well trained micro-economist would do, I asked if everybody did what Markowitz said they should do, what would be the implications for equilibrium in capital markets? This part was positive economics. It looked at equilibrium using the single index model as the maintained hypothesis about the way in which securities related one to the other. So I got the PhD degree and went to University of Washington. I thought there were some really nifty results in the third part of the dissertation I had the capital market line, the security market line, betas, the whole nine yards. But it all seemed hostage to this very severe assumption about the interrelations of security returns. So I set about seeing if I could generalize the model It turned out that it didn’t take much effort to get the same results without the restrictive assumption. That is how the published version of the CAPM came to be. I had the greneralization within a few months after finishing the dissertation, but given the vagaries of publishing and refereeing, it took a while to publish. Q. As a former editor of the Journal of Finance, you are being kind. (Laugh) Authors sometimes use words other than vagaries to describe the review process. But I understand there was an unfortunate time lag in what turned out to be a race with other researchers. You said that any good micro economist would be led to this result. At least three other very good micro economists were indeed in hot pursuit. At the time, were you aware that you were in a race? A. No, not at all. I was rather well cut off from the mainstream. I was at Rand full time while I did my dissertation, and then I went to the University of Washington. I just wasn’t in the mainstream of what was going on in the East Coast. Jan Mossin was working on some of these issues, somewhat later while he was at UCLA. But, the others, John Lintner and Jack Treynor were on the East Coast. I just wasn’t in that loop. Q. They had that three hour time advantage over you. (Laugh) A. That’s true. Versus all of the advantages we had, including climate. Q. Or distractions. (laugh) Dr. Markowitz was surprised, and relieved, when I pointed out to him that he had actually beaten Roy, who he thought of as a major 7 competitor, by a full three months. I would like to go on record by saying that you beat Lintner by five months. So if Harry’s relieved, you should be even more so. With the benefit of what is now almost precisely 40 years since the delayed publication (laugh) of your paper in September of 1964, what is your current view of what your 1964 paper established? A. I like to call it the original CAPM because it was the first of many equilibrium theories dealing with risk and return, broadly construed, in capital markets, or more generally the determination of asset prices and how asset prices relate to future probabilistic forecasts. If you stand back and ask, what were the two big practical messages that came out of that undertaking, one of them certainly, if overly trivialized, is to diversify, diversify, diversify. T formal result is to hold the market proportions of risky securities, and then lever up or down to suit your risk tolerance. But, the idea that holding a broad portfolio in market value proportions was an important and sensible investment strategy was the first big message. The second message was, yes Virginia, there is a reward in the form of higher expected return for bearing risk. But not just any risk. It is much more specific. It is the risk, to use the current broader context, of doing badly in bad times. Beta was a measure, in the particular mean-variance context, of doing badly in bad times. High beta securities are likely to tank in bad markets. The key idea was that it is not all risk, but the central societal risk that an efficient capital market is going to reward with higher expected returns. I think those were the two big messages. The fascinating thing is that when we go to much more sophisticated, much cleverer and much more detailed models of equilibrium, those same two messages keep coming through. Sure there are variations on the theme, but by and large, at least 90 percent of the time those messages come through in theories that are much richer than the original CAPM and other models built on mean-variance assumptions. Q. I have some CAPM trivia questions for you. First, we are accustomed to thinking about all four letters of the CAPM. I am curious. When did the M get attached? The title of your paper, for instance, refers to Capital Asset Prices. I do not recall seeing the M in CAPM or a reference to a Capital Asset Pricing Model. Do you remember when that particular term was added? A. I’ve always thought it was Gene Fama that first came up with that. 8 Q. Okay. A. And as for the Sharpe ratio, I think it was Treynor and Black who came up with that. Q. You didn’t even come up with the label for your own Sharpe ratio? A. No I called it (and I still think it’s a better term), the reward to variability ratio. But I think it was Gene that first used the term CAPM. And of course, in that era once somebody at Chicago gave something a name that was it. For example, the risk free rate was forever denoted Rf. Q. That is also on my list of trivia questions. Do you recall your label for the risk free rate? A. I can’t remember it. … Q. You liked rho. A. Oh, but that doesn’t work now because we use rho for correlations. So, in retrospect, that was a bad decision. In fact many of my decisions about symbols at that time were determined by my typewriter. I had one of the old typewriters with keys that came up from a near-horizontal position. I finally managed to get somebody to alter some of the keys. So I had a lower case sigma, a lower case mu, and I might have had a rho. But that was as far as I could go. And then there came a time when somebody stole it from my office at the University of Washington. I was bereft. It’s hard to imagine now in this era where we all have scientific word processing and other ways of using any symbols we want. It was very awkward then. Q. I remember that the Journal of Financial & Quantitative Analysis, just photocopied whatever you produced. A. Yes So if you had to write in the symbols by hand, which some of us did because we didn’t have modified typewriters, they just photocopied that. I have also gone back to look at my dissertation a couple of times, and that whole dissertation was typed by a typist on velum. You had a translucent piece of paper and produced a a reverse carbon image. If the typist made a mistake, he or she had to physically cut it out and then tape in a replacement The graphs in the dissertation I drew with a ruler and free hand. They are really ludicrous. It’s embarrassing. Q. That was state-of-the-art back then. 9 A. It was. Q. Continuing with our CAPM trivia quiz, over the years some of my students have wondered if the C and A are redundant. I told them it wouldn’t have made nearly as nice of a title if we called it the AP-M. We need the C in there. (Laugh) But what inspired you to call it “Capital Asset Pricing”, and did you mean to communicate anything other than just asset pricing? A. That’s an interesting question. I don’t know the answer. I wasn’t thinking about acronyms. I think I was trying to connote the fact that we are talking about assets in capital markets. Whereas with assets, perhaps I was thinking, you would think of cars and trucks. Q. So traded assets? A. Yes Securities, perhaps. Q. Well, I am glad that you [settled on capital assets]… because the buzz word is, CAP-M. With securities, it would be SAP-M, or something. A. Yes, or SPAM, I’m not sure what it would have been. But think about my 1970 book, it was called Portfolio Theory and Capital Markets. So I must have liked the word capital at that time of my life. Q. Yes. It was a capital idea. A. I think that’s been taken by Peter Bernstein. Q. Oh rats. Well that’s never stopped me before so— A. True. Q. Good things, I steal. Out of personal curiosity, and I might be the only one in the universe to have ever worried about this particular turn of events, but you mentioned your 1970s book. I happen to have it here. It was the first book I used to teach a course in portfolio analysis. In this book you reverse the axes from your paper. In the paper you have risk on the vertical axis and expected return on the horizontal axis. I have a picture of that graph, which I will show you. But if I were to put that on the board today, without labels, nobody would recognize it as having anything to do with the Capital Asset Pricing Model. This is the Capital Market Line. [Pointing to a page from the article.] The intercept is way down here on the horizontal axis, and the graph looks very strange. But this is also the way Dr. Markowitz oriented the mean and variance portfolio problem. In 10 contrast, by the time of your 1970 book, only six years later, you have reversed the graph. Under the now standard orientation, expected return is measured on vertical axis, and risk, whether it is portfolio risk or ultimately beta risk, is measured on the horizontal axis. I much preferred that switch. Do you recall when and how you made the conversion? A. I don’t recall in detail. I remember some time after I did the original piece, I read Sir John Hick’s work. I can’t remember exactly how he had the axes, but they were different from Markowitz. As I started to try to teach the material, I found that the newer way seemed much more natural. People tend to think that a higher Y value is better. You want to get up in the world. So it seemed to me to be a good idea to put expected return on the vertical axis. Then expected return more clearly became a good – more is better. As for risk, I flirted with reversing the axis so as to read it from right to left. But there was no natural way to do that. So I figured I was stuck with having the horizontal axis show that more is bad for risk. It seemed to me at the time, from experimenting with students, that this was a better way to present the material. And since Hicks had done it one way, and Harry had done it another, I figured that there wasn’t, at that point, a fully standard way. So that is what I chose to do. Q. Thank you. That cleared up that, that problem for me, and maybe no one but myself, but I am glad you did [make the change] because I also found it much easier and a much more natural way to present [the material]. Students take to it very naturally. A. I think so. Q. The next question is also departure from Dr. Markowitz in that you focus on risk free borrowing and lending, and with some caveats that there may be restrictions, your presentation of the model is based on an assumption that individuals can freely borrow and lend at a common risk free rate. How did that particular inspiration, to take a step beyond your mentor in this case, how did you come up with that idea? A. Well that’s easy. It came from Jim Tobin’s separation theorem, which he developed looking at a different issue. He was interested in monetary aspects. But he had shown that if you have a set of risky securities, you can simplify the 11 capital market world to a two-asset case and focus on the problem he was addressing. I can’t give you the date now, I’d have to look it up. Q. 1958? A. Good for you. He had shown that you could separate the problem into finding the optimal combination of risky securities, which in today’s problems would be the maximum Sharpe ratio portfolio, and then allocating your money between borrowing and lending and that combination. So that was pure Tobin. And it was very helpful algorithmically. I was able to build an algorithm which could find that portfolio rather than having to find all of the efficient portfolios. And that led to the equilibrium results. Q. In the equilibrium result, we now think of [the symbol] M as the common designator for the market portfolio. That particular jargon was not in your 1964 article, nor even in the 1970 book. Do you recall when the notation M for the tangency portfolio began to emerge and when you started using that? A. I can’t tell you that. It may have been in my dissertation. Curiously enough in my dissertation, though it was more specifically about the market portfolio, that was rather an artifact of the single index model. I was trying to be general, in the 1964 article. I overreached a bit by trying to have a flat section on the efficient frontier. Generally it can be very close to flat, and it probably is very close to flat, but it is not quite flat. But when the portfolio came to be called M, I can’t tell you. Q. For the next bit of trivia, what about the term, beta. We now think of the Sharpe Capital Asset Pricing Model as a relationship involving beta. Yet [the term,] beta, was not in the 1964 article, or even in the 1970 book. Do you remember at what point the world coalesced around the term, beta? A. I think I was trying to avoid beta, because in statistics, that term designates a scaled covariance. Of course, that particular scaled co-variance is what you would call beta in a regression equation if you have chosen to label the intercept alpha and the slope beta. But other folks began calling it beta at some point, and so I caved. Q. You caved? A. That’s, that’s as close as I can give you specifics. 12 Q. You had the characteristic line early on. So you were you were using beta. A. Yes, the covariance between return and market return divided by the variance of market return was the slope of a regression line. I was trying, especially in the dissertation and I think in the 1964 article, to keep people focused on the future. And so yes it is a regression line. But I wanted to emphasize that it is a regression line that is fitted to points, each of which corresponds to a possible future outcome. I was trying to avoid the trap that almost everybody, including sometimes myself, fell into, saying \ that this is a theory that says if you look at historic measures of beta, through statistical analyses of whatever sort, that it will tell you something about, not only historic average returns, but also future expected returns. The theory doesn’t say anything about the relationship between history and future prospects. The theory says something about expected future returns and future measures of covariance with the market. Q. Okay. How about the, the jargon security market line? [The term] capital market line is all over your 1964 paper, [but the term] security market line is not. The elements are in the 1964 paper, but that particular jargon again didn’t show up in that paper, do you remember when that began? A. That is interesting. I think that jargon was in my 1961 dissertation. I think both of them were in the 1961 dissertation. Q. We will blame the editors of the Journal for that. A. Perhaps. Q. Well, if you do let me know because these trivia issues, maybe other people don’t care, but as the amateur historian, I am just curious to know how things [got started], that are now commonplace and everyone uses them, and it seems to me that they are brilliant. Somebody deserves credit, and probably it is you, for these brilliant breakthroughs. A. Let me beg to differ with you. I don’t see either of those terms very much anymore – the Capital market line or security market line. You see them in textbooks from time to time. But the thing that does frustrate me is that in practice it is common for a consultant to justify estimates of expected returns by showing you a graph with expected return on the vertical axis and standard deviation on the horizontal axis, where each dot is an asset class. They draw a 13 line and show that the points are pretty close to a straight line. Of course, in the original CAPM they are not supposed to be close to a line in that space, they are supposed to be close to a linein E-beta space. Consultants almost never show that, partly because they don’t really think about a world market portfolio, which is what you ought to be using as the base for your betas. Q. I didn’t mean to accuse [practitioners] of using the academic world… A. Fair enough. Q. I had mentioned that the terms “security market line”, “capital market line” and “market portfolio” were in your 1970 book but were not in your 1964 paper. Now that you have checked your dissertation, are those terms in your dissertation? A. I can find the security market line. I haven’t looked yet for the capital market line. But, but there is the security market line as we said earlier, with the axes reversed, but the text says this is the equilibrium in the security market. This was in 1961. Q. The same picture is in the 1964 article. So we’re just talking about jargon as opposed to substance. A. Yes. And, and again I haven’t looked for the words “capital market line”, but you’ve convinced me now that those came about somehow or other between the 1964 article, which actually was written in 1962, and the book in 1970. Q. The next thing [on my list is that] we are accustomed to seeing the security market line [as an equation]. How about the equation? If I were to ask my students, “What is the Capital Asset Pricing Model”, most of them would probably respond with a equation, expected return as linear [function of] the beta for the security. Do you remember the first time you used the equation or saw the equation written out rather than a verbal description of the picture you showed me? A. Well, some version is in the dissertation in 1961. Q. It’s in the dissertation as well. Okay, then moving on to empirical tests. It is bold, I think, for any micro economist to presume that some theory they have discovered might actually describe the real world. Yet very quickly you published, in 1966, a study of mutual fund performance where you looked at average return and [historical] beta. Did you expect to find confirmation of the security market line and your Capital Asset Pricing Model? 14 A. I don’t know if I did or not. I guess as I look back, not only on my own work, but on other work that was done in that period, I am struck with how incredibly naïve I and some otheres were about empirical work. The idea that somehow or other you could look at a swatch of history and look at average realized returns and think that those were really good estimates of ex ante expected returns, was as , if anything, extremely naïve. Historical measures of risks and correlations are not too bad as estimates of future risks and future correlations. But average returns, as we now know and as we should have known then, can be pretty awful indicators of expected returns. I think we expected we could do a better job testing ex ante theories using ex post data than we had any right to believe. I think most of the people who did those early tests, including myself, although I did only a very small number of them and did them very crudely, thought we were getting a lot of corroboration for our theories, not only the original Capital Asset Pricing Model, but for extensions of it and other theories. Later we went through a period in finance where we began to use more powerful techniques and concluded that, while under the maintained assumption that historical means are good proxies for expected returns, the historic data are not consistent with this theory or that theory. But hopefully, now most people understand that you have to take historical data with a lot of grains of salt, when you use them to test an ex ante theory of any kind. Q. Are you referring to the later studies by Black, for example, where they broke the periods down into essentially before you discovered the CAPM and subsequent to discovery? A. Well, as a classic case in point, if you take individual securities, or even groupings of individual securities and look at relationships between historic average returns and historic measures such as beta and expect that you are going to get a really crisp indication of what, at any given time during that historic period, the relationship was between ex ante expected return and ex ante beta, you are kidding yourself. You don’t even have enough data to measure whether or not there was an overall market risk premium with any high degree of confidence, let alone measure things down to those miniscule levels. I think we expected, and some still expect, far more out of empirical work using historic 15 results than historic results can provide for us. There just isn’t enough information in the historic recotd to sort out some of the subtleties. Q. When you wrote your 1970s text on portfolio theory and capital markets, I assume you used it in classes that you were teaching at the time? A. I did. Q. I used it as well. It is the first text I used for this. A. That’s at least two of us that bought it. And our students bought it. Q. I was under the impression that [there were] no competitors at the time. So that was it for the subject matter. There were papers available, and you could put packets together. Dr. Markowitz had a book, but it really wasn’t geared as an instructional vehicle. Did you have any idea that you would be able to write a successful textbook, which is a very different skill from fundamental research? A. Well, I hoped I would. I was very fortunate to have been at UCLA and Rand. At UCLA, I took a course in economic history. Warren Scoville, the instructor, required us to write a two page paper once a week on a specific subject. He edited our papers as much for style as for substance, and he insisted that we learn how to write so that somebody could understand what we were saying. At Rand, all of the papers we produced had to go through an editor who took out every passive verb and changed it to an active verb, cut every long sentence into two or three, struck out redundancies, etc., meticulously. What we produced was readable and not just “academic”. So I was very fortunate to learn something about how to write in a way that could communicate ideas to people. I guess I thought I had written enough at various levels that I might be able to write a successful textbook. And certainly given the interest in a book covering those subjects at that time, the book was quite successful. Q. I particularly liked your personification of people with certain types of utility functions. I am going to show you a couple that I have reconstructed and ask if you recall the names of the two individuals that had these particular shaped utility functions. A. The first was probably something like chicken. Q. The first was Mr. Chicken. A. Good. Q. And the second was Mr. Fearless. 16 A. They were both male. Isn’t that embarrassing? Q. I was going to mention that. By giving names to these individuals, you gave life to a topic that was very dry and very difficult to teach. Students got a chuckle out of Mr. Chicken, but they always remembered this shape. They could also easily understand why Mr. Chicken would choose the minimum-variance portfolio. (Laugh) It allowed me, and others, to teach the algebra of the minimum variance portfolio, which we couldn’t have gotten away with without it. I did make one addition. I always wished that you had also identified a person who would have chosen to put everything in the market portfolio, the tangency portfolio. I elected to name that person, Ms. Egalitarian. She was a women, in deference to the women’s movement that was gaining momentum at that particular point in time, and she was egalitarian because she was willing to trade off expected return and variance. In your entire book, that was the only change I would have recommended making. (Laugh) A. Let me just suggest that you need to modify that a little bit because there is a school of thought that argues that cap weighted portfolios are inefficient, and you ought to weight equally. So I would not want to call her Ms Egalitarian now out of concern people might think I am advocating equal weighting. Q. Okay. (Laugh) Let’s change her last name. But she would be Ms something, for sure. A. We could do it in French. Actually, I still have that characteristic in the equilibrium simulators that I am working with now. I insist on naming all my actors. I either use names from the Social Security registry of the most popular names, or I come up with names of my own. I have a behaviorist with a kinked utility function, and I named him Daniel in honor of Danny Kahneman. So if you look at my current work, you will see that all of the people have names. But they are not all names designed quite as explicitly to reflect their utility functions. Q. Well, it was a very clever strategy. Things just seemed to fall in place [for teaching] this theory using this approach. A. Thank you. Q. You have talents beyond your normal academic talents, as a lot of students that couldn’t have learned the material otherwise can attest. A. You are very kind. You are going to make me blush. 17 Q. Well we’ll just think it’s a suntan here… A. Okay. Q. You did a lot of early work on portfolio performance, and then you revisited it, and had another block buster idea with your performance attribution work. I guess I come back to the same question, how did this epiphany overtake you? [What made you realize] there was something to be done in this area? A. You are talking about returns-based style analysis? Q. Yes. A. Well, this was born out of necessity. In 1986 I was interested in the problems of traditional pension funds. For many years I had given seminars and worked with pension fund people. It seemed to me that there were some really interesting issues in that area. Some of the problems could be addressed by bringing to bear existing theory from finance. Others needed some new theory, and new empirical work. I wanted to get my arms around those issues a bit more efficiently than I thought I could as an academic. So my wife and I set up a research and consulting firm. We worked with people on the staffs that ran pension funds for AT&T, General Motors, Hewlett Packard, and others, , learning what their key problems were, and then seeing what could be done to help them. It became fairly clear early on that what we needed was a single factor model that could deal with all the pieces of the fund, the bond portfolios, the stock portfolios, the Lord knows what portfolios. Q. By single you mean integrated as opposed to the one factor? A. Yes. I’m sorry. A multiple factor model, but just one model. There were at that time stock factor models, bond factor models, this factor model, that factor model. But for high level decisions, which involved asset allocation, manager selection, manager performance, etc., what you needed was one model with a reasonably parsimonious number of factors, 10, 15, 20 maybe. If you could map every single piece of that portfolio onto one or more of those factors, everything would add up, and you could get the desired addition, subtraction and aggregation properties. The question was how we were going to do this. There was no way I was going to be able to look at all of the securities. A couple of the funds I was working with at that time had a hundred managers who collectively might have 18 held 3,000 to 10,000 securities. There was no way I was going to look at each security and figure out how it mapped onto up to 15 different factors. That just wasn’t going to be feasible. The question was how to create and implement this grand edifice in a nice neat consistent way, without doing empirical work which was impossible for me. I thought initially, ah ha, regression analysis. Let’s regress the historic return for a particular investment manager’s portfolio on 15 factors, large value stocks, long bonds, whatever. I did that with some test data and got mostly garbage. So that was the end of that. Then I realized that one of the reasons it was garbage was that I was getting estimates such as that a given manager is minus 300 percent in treasury bills, minus 400 percent in small stocks, plus 900 percent in large stocks and so on. And I knew that wasn’t true. So what do we have as prior information to use when we do the analysis? Well, we knew that most of the managers at that time had no net short positions in any asset class. Well that is easy. Put lower bonds at zero. In some cases, we also knew they had certainly no more than 50 percent in a particular asset class. If we knew that, we could put an upper bound of 50 percent. That was also easy. The result was a quadratic programming problem, and I had quadratic programming algorithms lined up all around the office for other purposes. So I started trying that with the data, and voila, I got sensible results. When I say sensible, I mean this. I would go to the pension fund folks and say my analysis indicates manager X-Y-Z’s style is like 20 percent of this and 30 percent of that and 50 percent of this other thing. What do you think? And then they would say yes, I think that’s right on. Perhaps it was the power of suggestion but probably not Then maybe we would go to the manager. Or we would get a case where it did not seem quite plausible, and then we would go to somebody, the manager or the sponsor, and we would figure out that from an economic viewpoint it was plausible. We got all this corroboration as we started applying the technique, so we applied it to more managers with more sponsors. It turned out to be very useful. Was it perfect? No. Could you improve on it with more detailed information in many cases? Yes. But it turned out to be a very inexpensive, very easily implemented procedure, which could do a very good job. It is good enough, to give you an indication. For r some of the large pension 19 sponsors with whom I still work, the technique is giving -- out of sample so no cheating -- R squares in the high 90s. Q. Wow. A. This is at the fund level. I am not saying I get every single manager right. But at the fund level, I am explaining a huge amount of the month to month variation in performance. So it is really a very useful technique. It is a technique that combines data with priors, or with information from other sources than the data at hand. For example at Financial Engines, where we use the procedure routinely across a large group of managers and mutual funds, we use other information, and not just zero-one bounds. Q. You had previously worked with industry, including some early work with Merrill Lynch on beta methodologies. How did that relationship develop? A. Actually, there were some things even earlier than that. At Rand, I did a lot of practical research as well as theoretical research and application of theory. So I had been very much in the mode of theory and practice. Theory is good for practice, and practice is good for helping you figure out what theory to work on and whether a theory is useful. So I had been very much in the sort of ambidextrous state of having a foot in each camp. Sometimes I had one more weight on one foot, sometimes more on the other. When I first started teaching full time, I continued to do work for RAND on the economics of computers, pricing of aircraft landing spots for commercial airliners, etc. I also did a little half day consulting assignment for IBM implementing my diagonal algorithm on one of their very early machines. The Merrill Lynch liaison came about when I was at the University of Washington. I was approached either by Gil Hammer or maybe Jack Treynor who asked “Would you like to work with us”. I of course said yes, and I went to New York once a month for a couple of days, over a numbers of years. Q. And you helped them estimate betas? A. Well, I helped them with the structure of their beta service, and their performance measurement and attribution service. I was helping apply, and to some extent develop, procedures. I am not an econometrician. They had some very good statisticians and econometricians there, so I didn’t get down and dirty 20 with the details of things like windsorizing the data. I dealt with economics and the finance theories. Q. Early on, they estimated betas for up markets and betas for down markets. Did you have anything to do with that? A. We were all involved with that, speculating about it, playing with it, and trying to understand what it told us. It was a very cooperative effort. It was a very good group. Q. You also worked early on with Wells Fargo in the development of products that we now take for granted, index funds among others. How did you get started with that? A. I am not entirely sure. Bill Foss and I probably knew each other from other venues, but when I first went to Stanford, I remember inviting Bill down to talk to my class. We talked, either then or later, and felt that we could do interesting things together. So I started working with Bill and with the folks at Wells Fargo. Q. Did you help them develop products like index funds? A. Well they did develop them, and I had a little role in that. I had a call from a fellow from Samsonite. I think he was the son of the president. He had just finished an MBA at Chicago. He called me to say he had read, heard about, and had been taught some of my work. He was thinking they should invest in the market portfolio, or something of that type. I put him in touch with Bill Fouse, but the fund wasn’t actually implemented the right way. It was closer to equal weighst than market weights. In any event I believe the first Wells Fargo index fund was the Samsonite fund. So I had a little there as a marriage counselor. Q. In the early days, we did not have passively managed index funds, but we were still trying to teach MBA students to assume that prices were determined by investors who held the risk free asset and the market portfolio. There were no good proxies for the market portfolio early on, and at least my MBAs thought I was teaching them pie in the sky theory that would never become viable because the instruments did not exist. They had difficulty thinking that holding 8 or 9 or 10 different stocks could proxy for holding the market portfolio. What did you teach back in those days? 21 A. At that time, I did a couple of things. First, I would point students towards the most diversified equity mutual fund I could find and say, well that’s at least pretty diversified. Then I would draw a graph, which I am a little sad that I promulgated in the world, which had portfolio risk on the vertical axis and number of securities on the horizontal axis. The curve went down and flattened out. You can look at it and say, ah 20 is enough, and 30 is certainly enough. So I would make the argument that the thing to do is to be as highly diversified as you could be and that kind of took care of the market portfolio aspect. In retrospect, of course, that graph is very deceptive because it assumes equal proportions and no industry factors, etc., etc. So it substantially overstates the amount of diversification you get from 20 or 30 stocks. I think most of us would now say 20 or 30 is not enough, and equal weighting is not the right way to do it. But at that time, the idea was that there is a way that allows you to get tolerably close to a market portfolio, perhaps. As for borrowing and lending, you could certainly put money in the bank, even if it was not a great interest rate. You could also certainly borrow on margin, although I am not sure at the time you could do it for mutual funds. In some sense, there is is a constant tension between a positive theory and a normative theory. We now know you can argue, fairly convincingly perhaps, for the security market line. You can argue that you ought to use the market portfolio, possibly with leverage of one sort or the other. But we certainly can not argue that people do this because most people do not. So if you want to hold your positive theory up as a theory not only of asset pricing, but also of how people, in fact, choose portfolios, it obviously fails right out the chute. It is hard, as you know and as we all know, to convince students to adopt the asset pricing aspect of the theory when the portfolio choice implications are clearly inconsistent with observed behavior. In my current work, I build equilibria with people who have different predictions, different this, different that, different other things. What is interesting, and what I think will be helpful to the extent people adopt this as a pedagogical device, is showing that asset pricing theory results can hold up even in a world in which people’s portfolio choices are very diverse and only a minority buy the market portfolio. 22 Q. On the normative side, are you still a defender of the market portfolio? Is that something people should pursue if they have no specialized expertise in security markets? A. It is more subtle than that. I believe it is very helpful, at least to a first approximation, to assume that asset prices probably reasonably reflect publicly available information, that is that there is semi-strong form market efficiency. You do not have to hold a hundred percent of the market portfolio. 90 percent would be good. Then you have to ask yourself if, at least for the bulk of my portfolio, am I going to adopt the premise that prices reflect people’s predictions? If so, what does that imply for what I ought to do? The way in which I like to think about it now is that people come to the market with what I call positions, your house, your job, your preferences, your utility function, and much more. If you are really trying to be broad about it, horizon, what have you. predictions. Then you make It seems to me that it is useful to counsel people to not make decisions based on predictions that are wildly different from those of the average investor as reflected in market prices. I am being very loose here. I have tried to be much more careful as I have written about this. But you live in Ohio, I live in California. You have one kind of income, I have a different kind of income. So there is no reason we should hold the same portfolio, even if we are four square with each other in terms of predictions – about the probabilities of different outcomes in the future. You need to take the other differences into account, however. For that reason alone, I am not going to hold what you hold, and probably neither of us will hold the market portfolio. Similarly with preferences, I may have one horizon. You may have a different horizon. There are all kinds of reasons for people to hold different portfolios despite the almost religious belief that you should not expect to get something for nothing. If you think something is wildly mispriced in the market in an equilibrium sense, you may well be wrong. Q. If someone were to ask you for advice about how to construct the market portfolio, is that something you could do for them? A. Only with difficulty and error. I am on this big campaign to get people in the industry to provide ongoing, at least monthly, estimates of the market capitalizations of US stocks, French stocks, US bonds, French bonds, venture capital, private equity, institutional, quality real estate not already reflected in the 23 stock values, etc. Some of those data are available. Some you can tease out on the Internet from one place or the other. But most of the data you have to get by begging and pleading with various people in the industry who do not make the inforrmation public. They may make the returns public. Wilshire makes the return on the Wilshire 5000 stock index public every month. But they do not make available the market capitalizations for those securities, although they have them, of course. So that’s my current campaign. There are other big issues, such as what to do about] human capital and owner occupied real estate, or whether government debt is part of the market portfolio, or if it just an asset for you and a liability for me as a taxpayer. These are huge conceptual issues. But certainly the market portfolio is not just the S&P 500; it is not just the Wilshire 5000; and it is not just the US stock and bond markets. There is a lot more there. Once you get to that point, to go back to some of our earlier discussion, you have got to think about an equilibrium in which there are French people and American people, there are Euro people and dollar people. You need to think about a much more complex equilibrium in which people’s positions have to do with, among other things, not only where they live, but in what currency they tend to buy most of their goods, and what their consumption baskets are. Increasingly we need to think much more broadly than the wildly simple version of the Capital Asset Pricing Model which assumed away huge swaths of the real issue. That being said, all theories assume away huge swaths of real issues. The test of the theory, as always, is have you included enough of what really matters to give you something that leads to better decisions and better understanding than you would have had otherwise. We have obviously come a long way since the original Capital Asset Pricing Model. Q. Staying with practical advice, have you thought about might be appropriate for a young investor with $100,000 to invest, who has no specialized knowledge about securities and just wants to have a reasonably diversified portfolio? A. The first practical advice, of course, I would say go get help. Until not too many years ago, it would make no sense for somebody with $100,000 to get help because it would cost them $50,000 to get the help, or at least $10,000. Now 24 you can get good help electronically from a number of sources for very nominal amounts of money. That is what I would start with. But that is not the way you wanted to frame your question. So let me restate your question. Assume that someone can’t go get help, and I have to tell him in a paragraph or two what to do. That’s it, and I can’t ask any questions. All I know is here is this young person who is 30, you said young, so we will call it 30. He has $100,000, and I really don’t know much more. If you had said this is a person who is average in every respect— Q. Let’s change the question to average in every respect. A. Oh that’s easy. World market portfolio, thank you very much, I’m done. Q. But how do you do that? A. Oh, then you buy as best you can estimate what the market proportions are for the major assets that are out there, and you can find an index fund typically for each of the broad classes. For example, you can get the Vanguard total stock market fund, which covers the US stock market very well. You can get the total market value bond fund, or whatever it is called, which covers the US bond market very well. There is also a Vanguard fund that is a 60/40 fund. So you can either do that or invest in the two funds separately. Do the same for international stocks. Buy highly diversified low cost index funds with low turnover and low management fees, to give yourself exposure in those areas, and do it in basically world market proportions. Of course, a young person is not average. An average person is middle aged If you weight by dollars or weight by dollars, Euros, and such, an average person is probably more than middle age because after all, in the market, dollars vote, not people. There are more dollars representing old people. So market prices tend to be set by older people and richer people, partly because richer people are older. Price are also set by both Euro people and dollar people. So it is a bit of a cop out to say that an average person should hold the world market portfolio because I have never met an average person. So unless you happen to be a citizen of the world, and you are just the right age the market portfolio is not likely to be the best choice. For example if this person is younger than average maybe that involves a tilt with a little bit more aggressive stance. I would also] 25 like to know about his or her job. Is this a person with tenure or a person who works in high tech in Silicon Valley? It makes a difference. Q. Fair enough. So you are really going back to the advice, and you are uncomfortable giving advice without knowing the particulars. A. Precisely. Just like a doctor. Without knowing what the symptoms are, you don’t want to prescribe a standard drug. Q. I can understand that. You mentioned that [it matters] if you live in the United States, or in France, or Euro country, or non-Euro international country. Do you think a certain amount of home bias is appropriate in investments, a tilt toward your own domestic consumption bundle, or is that not appropriate? A. If you play with the data a bit and assume either that currency risk is an issue, or that getting rid of it is costly, then what seems to come out of the data, at least as I’ve looked at it, is that home bias in the bond area makes a lot of sense. The advantage of diversifying your bond portfolio abroad is not sufficient to significantly overcome either the currency risk or the cost of getting rid of the currency risk. So in the bond area I think home bias makes a lot of sense. In the stock area, the advantages of international diversification remain pretty substantial. The advantages are possibly not as much as they used to be, with globalization, although one can read those data in different ways. But there are still some pretty big advantages. So I would think that home bias on the fixed income side, for sure. On the equity side, there probably should not be a lot of home bias. But at the end of the day, it really depends on your consumption bundle. Q. On the same general theme, assume that a young couple is probably going to buy a house in the near future but they do not own any housing right now. In effect they are short housing in some sense. Should they have a tilt toward more real estate, such a REITs? A. Well, again it depends on whether you think of a lifetime portfolio including human capital, etc. I haven’t really thought through the issue of what to do if you’re going to buy a house. I’ve thought through the human capital issue at some length, but I beg off on this question. Q. That’s fine. And you may want to beg off on this one too, the baby boomer generation is about to hit retirement age. Presumably many are looking 26 at a change in housing, and perhaps even moving into a retirement home. If they have that in contemplation, should they tilt their portfolio toward, I don’t know if there is such a thing as a retirement real estate REIT, but sunbelt type housing for people from Columbus, Ohio or even colder climates. Is that something they should be thinking about as they approached retirement? Or is that again something off the radar screen for you? A. Now you are talking about my new field. I have declared that when I get a few other things wrapped up, I am going to become a researcher in retirement economics. You can take that either way, if you will. It seems to me, and I’m sure it seems to you, that as we get to the point where we think about precisely the things you’re describing, that the economics profession hasn’t really started modeling big elements of the decisions you face at a later age. I’m thinking of annuities, long term health care issues, housing, etc., etc. There are some big issues there. I am playing around with a state preference tree structure and a pricing kernel view. I haven’t gotten to the point where I can answer your question, but I will tell you what my neighbors, next door here have done, which I find fascinating. They live in Carmel. Their daughter lives in Pittsburgh. Not Pittsburgh, California, Pittsburgh, Pennsylvania. And they are determined they are not going to end their days in Pittsburgh. So they bought a unit in a three stage retirement community out in Carmel Valley, which has independent living, assisted living and skilled nursing. They pay the dues, which cover three meals a day, but they still live here. They go out there one day, have dinner and stay overnight, and then they come back. They view this as basically an insurance policy. It is an insurance policy that is not generic, but specific. They know that they will always be able to get in there whether they go into independent living, or into assisted, or skilled nursing. They have basically bought forward for that stage of their lives, which is very interesting. Being financial types, we tend to think about buying a contract of some sort that will pay money so you can buy later. But very often there is an adverse selection issue for anything of that sort. You may have the money, but they won’t take you if you are too feeble or if you are really need heavy nursing. Q. If you have enough money they’ll take you (laugh). A. Well that’s true. That’s true. 27 Q. A feeble versus money trade-off frontier perhaps? A. Yes. And, of course, they are bearing a risk that the outfit will go under so. So I think those are big issues for people of a certain age. And partly because most of the people doing research in finance are not of that age, I don’t think those issues have been sufficiently addressed. Now with all the baby boomers coming to the point where they are going to be facing those issues, I think we are beginning to see some serious research activity in this area. I think we will see more. These are tough issues, really tough. So I plan to play around with them some more. Q. What advice would you give to a young researcher just starting out, a young Bill Sharpe? A. If you can, arrange to have been born in 1934 (. There weren’t very many of us, and good economists had not spent a lot of time on the economics of uncertainty. So there was a lot of low hanging fruit. It was rather easy. It is much harder now. You never know how much of this is just old age talking – thinking it was better when we were kids. But I think financial economics has become so sophisticated, and there are so many first class people looking at a huge range of issues. As with anything else, get lucky. That’s another obvious piece of advice. But I think it’s A) more fun, and B) potentially higher risk, with higher expected return to try to look at problems that have been under-researchedOr just look at things in totally different ways. I find that however practically is a lot of the edifice that we in academics and the industry have built on mean variance foundations, the state preference approach is superior in many ways. A) It is simpler, and B) you don’t get in some of the strange blind alleys that you get in with continuous distributions and perhaps continuous time. But it is very hard for me to know what makes sense for somebody starting out. Q. All right. Next question. You have read a lot of papers in your career. Are there any that you really enjoyed so much or just were thrilled with so much, that you secretly wished you had been the one to write them? A. Oh many, many, many . Q. What’s at the top of your list? A. I don’t know. I haven’t thought about that question. I think in many ways the most beautiful papers in the entire field, by my definition of the field, are Ken 28 Arrow’s incredible paper originally written in French and then in English. Q. Did he write it in French or was it translated for him? A. I think it was translated. He was invited to give a talk at a conference in France, which I think he told me he gave in English. Q. I have one paper in French too, but I can’t read it. A. Yes. Well, I think Ken actually has quite a bit of French, but I don’t think he wrote it completely. But in that paper, which in English was 9 pages or 11 pages, he laid out the whole basic state preference approach. Then Gerard Debreu, in a very small book did the whole theory. Then he said, well there is uncertainty to deal with, so I’ll write a chapter on uncertainty. I believe the chapter was four pages long. M of financial engineering that came, although the financial engineers may not know it, from those two bodies of work. If you think about how much of what we do every day and teach every day can be understood and taught better with that paradigm than with some of the paradigms we use, you have to conclude that those are without question some of the most brilliant pieces in terms of their implications and their sweep. Q. Well, you at least got to come close because you had a very early contribution on the binominal option pricing model, did you not? A. Well, it wasn’t as early as their work, but it certainly, it usedthe state preference approachQ. That didn’t satisfy you fully? You still wish you’d written the Arrow & Debreu work? A. Oh no. I’m happy having written what I wrote. Q. Okay. A. I was taking your question somewhat more broadly, what do I think were among the most brilliant things and the most seminal pieces in the whole area of financial economics. And it’s got to be Arrow-Debreu, and I am not going to differentiate between the two of them. Q. You answered the question very well, and I think you may have answered my next question as well. You just got to it ahead of schedule. You anticipated the question perhaps. But are there any unsolved problems that you look forward to, at least trying to solve or, or wish you could solve? A. Well I’ve been working on a number of things in the last couple of years. In more academic pursuits, I have been trying to get my arms around asset 29 pricing and portfolio choice. So what else is new? But I’ve been using a state preference approach and dealing with the equilibrium situations where people do differ on positions, preferences, and predictions. In particular I’ve tried to actually get serious about people not being in agreement on probabilities. And looking to see the extent to which some of our conclusions from models such as those using representative investors either are or are not robust. In particular, I’ve been directly looking at issues that arise when people have preferences that are not mean variance, and possibly not of the standard constant relative risk aversion variety that we sometimes use. I’ve been bringing in some of the ideas from prospect theory and behavioral finance, and using, as I mentioned briefly earlier, an equilibrium simulation approach where I create people. give them positions, preferences, predictions and initial holdings, and then let them trade with each other until they do not want to trade anymore. Then I investigate the properties of the resulting equilibrium. That is a new technology -- certainly for me, and I think pretty much for the profession. I think it has real promise as a pedagogical device. I also think it has real promise as a research device because you can look at much more complex systems than with closed form models. It has the disadvantage that it’s very hard to know when you’ve really got a result rather than just an artifact of a particular simulation. So you have to keep giving the simulator different initial conditions and different situations until you can convince yourself perhaps that you have a general result. I have also found this kind of approach useful as a sort of a theorem producer. You play around with cases, you see results, and then you say, I wonder why that result came out? Sometimes you can figure out why and that might lead you to a stylized closed form approach. I love computers, I love computer programming, and I think we can do a lot more with them. People are doing a lot with, Monte Carlo simulations. They are doing a lot with numerical solution procedures, and I think building equilibrium simulators is a very useful undertaking. If hope to get my code to the point where somebody else can use it, Then I can try to make it available on the Internet and see if anybody agrees with me. And retirement economics. Don’t let me forget retirement economics. 30 Q. I would like to mention three names, and after each name just give me your reaction, short or you can even pass on it if you don’t have a reaction for me. The first is Harry Markowitz. A. Well as I was enthusing about Arrow-Debreu, I thought in retrospect -- whoops. Harry Markowitz basically started the field. He started all that has come from mean variance. And he started at least half of what is now standard practice in sophisticated investment shops in the world as well as in corporate finance groups and others. If you think about impact on practice as wekk as impact on academics, it is very hard to think of anybody else. Obviously Arrow, over a wide range has had a huge impact, but I Harry started the deluge. He’s up there at the top of the pantheon in finance and financial economics. Q. Next name. Merton Miller. A. Merton Miller and Franco Modigliani. In terms of their seminal work they are sort of [interchangeable]. Making an arbitrage based argument. Think about how that changed people’s thinking and teaching. Go back to the finance textbooks that preceded their work. They are downright embarrassing by today’s standards. Mert and Franco just shook the finance field as it was then, corporate finance aspect in particular, to its foundations. They made people rethink. And then, of course, Mert did huge amounts of work in a number of other areas. He was up there at top. It’s not the field in which I have done most of my work. But if I were in corporate finance I would be saying about Merton Miller, and Franco Modigliani what I just said about Harry Markowitz. Q. If only because of the big announcement today, but thinking specifically about the contribution on the equity premium puzzle, what are your thoughts about Edward Prescott? A. Well I must admit I don’t know more than a couple of his pieces. And I haven’t looked at those for a long time. And I don’t think I have ever met him. I think of him as a macro economist who has done a considerable amount of work. I looked at his bibliography just before we sat down and it is overwhelming. But I only know a small piece of his work. The piece I know is very good, and very important. I don’t think we are yet quite clear whether the puzzle exists, because it could be an artifact of the data. I go back to some of the things we said earlier about how hard it is to even tell from data, in the very long run over many 31 countries, what the premium is let alone why it is. It might also an artifact of too narrow a construction of the utility functions that drive the markets. There are a number of possible explanations. It may be that we don’t measure the complete market including human capital. Who knows? There is a plethora of purported explanations for the puzzle. But if nothing else, just raising the puzzle and saying you can’t keep using the simple models you were using in a world in which the premium seems to be in the nature of 5 or 6 percent, or 8 or 10, whatever it was when he, they wrote the paper, was a huge service. But I don’t know enough about his entire body of work to give you a good assessment. I’m sure if I did, if I were in macro economics I’d probably say what I said about Harry Q. Well that’s the last of my questions. Is there anything you would like to add for the audience, American Finance Association? A. Well only to say that finance is a wonderful field. It it is marvelous because it has some of the toughest intellectual problems that you can come up with, and yet it has all sorts of practical applications. To the extent we can understand these things better, we can help ordinary human beings figure out how to arrange their lives so that they are as well off as they can be. It’s got social value. It’s got commercial value. And it’s got intellectual value. It’s a great field. I’ve had a great ride in it, and I recommend it to whomever is watching. Q. On that note I think we will conclude the interview. A. Thank you. It has been a great pleasure. 32