Editor's Note: In late February, Financial Engines Chairman William Sharpe talked with Editor-in-Chief Evan Simonoff about the equity risk premium controversy, different perceptions of risk, criticisms of modern portfolio theory and the need to look at retirement investment from a multi-dimensional, nonlinear perspective. What follows is that interview.
Simonoff: In the last year or two, many Americans have started to question their ability to retire. In the New York area, you hear Wall Street investment bankers now joke their retirement plan is to become a Wal-Mart greeter. Do you think that the post-World War II notion of retirement is one, financially feasible, and two, socially desirable?
Sharpe: The huge increase in longevity is a major driver. There's a radical difference in the number of years people who are born now can expect to live as opposed to people who were adults pre-World War II. That raises both the necessity to either work longer or save more during your working years. Over the last three or four decades, the retirement age has been lowering-people are retiring earlier while they're also living longer. If you do the math, that puts a terrible burden on people to save when they're in their working years.
To some extent, the social structure is such that some of the burden is on the people who are now working to support the people who are retired. That worked very well with my generation because we were Depression children, there weren't very many of us, and we had all the baby boomers to support us. But it won't be that way in the future.
As you stand looking forward from age 21 to your life span, you have the choice of how long you're going to work, how much you're going to save during your working years and how well you're going to live when you retire. Those three things are tightly interrelated. I think people, who must pay for their own retirement, will choose to either work longer at their regular job or work part-time at least through some of the "retirement years," whether it's as a Wal-Mart greeter, at McDonald's or as a consultant for their former firms. I just don't think people are going to want to only live poorly in retirement or consume small amounts during their working years. At least in part, I think they're going to choose to work longer.
Simonoff: Over the last year, many academics and institutional asset management types, are coming out with pretty gloomy predictions about the returns we can expect from equities in the next decade. One of the most optimistic among them who I've discovered, Jeremy Siegel, says don't expect much more than 5% to 6% a year. Do you share their views?
Sharpe: I like to look at this issue in terms of the risk premium-the amount of added return over and above putting your money in the bank. This avoids issues having to do with whether there will be high inflation or low inflation. There are technical differences that always bedevil these discussions, so it's important you get your terms straight-to make sure you are clear on whether you're talking about "arithmetic means" or "geometric means." Very often in a dispute about the risk premium after a half an hour of heated discussion, the parties find out that one of them is talking about an arithmetic value, the other about a geometric one. That alone can make a difference of 150 to 200 basis points. So you need to get your terms right. Moreover, are you talking about large U.S. stocks? All U.S. stocks? All global stocks? These technical issues become important when you're debating whether the premium is say 5% or 7%.
When I think about the risk premium, I concentrate on the arithmetic mean. At Financial Engines, we do not believe that you ought to take, say the Ibbotson data that starts in 1926 and look at the historical premium and say that will be the risk premium in the future. We believe that looking forward, the premium is going to be lower than this. Why? For three reasons: first, because it's plausible to believe that there is less risk associated with investing than there was during the '20s and '30s. With less risk, people require and get lower risk premiums. Second, we're better at sharing the risk. We have global financial markets and new financial instruments. This makes it possible to bear risk more efficiently, so the premium doesn't have to be as high. The third reason is that investors are richer than they were 20, 30, 40 or 50 years ago. People who are richer are generally willing to bear risk for a lower reward. So for all three reasons, we think the risk premium going forward is going to be lower than it was in the long historic period. How much lower? If you talk to the consultants and the institutions, you generally find a forward-looking risk premium in the 5.5% to 6.5% range (assuming this is defined as the arithmetic premium of all U.S. stocks over short-term cash.) I'm comfortable with numbers in this range.
Simonoff: Do you think another reason why the risk premium on stocks has come down is because large segments of the general public have concluded that they'd rather own stocks than bonds?
Sharpe: To an extent, yes. The question is: Is this part of the market equilibrium process or evidence of irrationality? Perhaps people said, "Historically, we got 8% or 9% over cash by holding stocks. That's a great deal, let's go out and buy some." Now if everybody tries to buy something, the result is its price will go up. Then at some point, people say, "We're only going to get 5.5% to 6.5% over cash. Maybe we don't want any more stocks than we've already got," and the market happily settles into that equilibrium. On the other hand, some people tell the irrationality story that people look at or looked at, say, the last five years and thought stocks would deliver 21% in the future. What a deal. And they bid the prices up to the point where stocks were wildly overpriced. That view leads some people to argue the risk premium in the next few years will be very low or negative. Frankly, I don't believe the markets are that crazy.
Simonoff: Roger Ibbotson recently observed that one reason stocks might surprise the doubters is that companies have reduced their dividend payouts and are reinvesting a much larger share of their profits back in their businesses. Is that reasonable to you?
Sharpe: Certainly this should be taken into account. Some simplistic calculations that take the dividend yield, add on the projected growth rate of real GDP, fail to take this into account. Moreover most stocks are issued by companies that are levered-they have bonds as well. A number of simplistic calculations leave out that element. A company can grow faster, other things equal, if it doesn't pay out a lot of money, so you need to factor both those elements into your calculations.
Simonoff: Last week at the TD Waterhouse conference, Bill Seidman said that if 3% productivity growth continued like it has for the last seven years or so, it might justify high price/earnings multiples on equities. Does that sound reasonable?
Sharpe: I don't tend to look at the issue quite that way. You have to look at the prices of the securities of the whole company or the whole corporate sector. If you're projecting a plausible growth rate for overall earnings to be similar to GDP growth, then you have to analyze the company as if it had nothing but equity. Put together the bonds and the equity and estimate the return on that total capital structure. This will be divided up so that in the long run you'll get less on the bonds and more on the equities. There are a lot of important ingredients that have to go into those calculations, and people very often proceed simplistically as if firms were financed entirely by financial equity, which they are not. Roughly 40% of the marketable securities in this country are bonds, government and corporate. At Financial Engines, we start with the view that expected returns are related to risk and to correlations and that market prices are efficient in the sense that securities are priced in accordance with a long-term overall expected return that implies equity returns of the kinds of numbers we were talking about.
So when we forecast our expected returns for asset classes, we use our projections of risks, correlations and the current market values of asset classes. We construct expected returns that are macro-consistent, meaning that if all the investors in the world followed our advice, the amount of securities demanded in Japan would equal the amount supplied there, similarly for U.S. equities, value stocks, growth stocks, and so on. It's very much a market equilibrium, macro-consistent view.
The overall levels of these expected returns are driven by what we believe the long-term risk premium will be across asset classes. The implied equity premium is lower than the Ibbotson historic value for the U.S. and somewhat lower than the historic across the world over the past 101 years, but not a lot lower than that number.
By using this procedure, we avoid telling our clients to time the market or make bets on some sectors or some countries. We take the market prices as being efficient at the asset level.
Simonoff: OK, what would you be telling a 50-year-old, who two years ago had, thought he was in a good position to retire, with say, $1.5 million in assets and was hoping say to retire at 60. But now he's lost about 45% of his portfolio.
Sharpe: If in this market you lost 45%, you were taking a lot of risk. If you'd known the implications, you would not have been in that situation. Diversification could have made his life so much better. And diversification is painless, as our users can see very graphically.
Simonoff: There are a lot of people out there like this today.
Sharpe: Well let me say first, the only way I can answer that is to race to my computer and set up a little account and run forecasts. But I'd say more generically that it's hard and dangerous to generalize because people are different. Consider two 50-year-olds with the same 401(k) portfolio who want to retire at 60. They may have different assets in other accounts, different retirement income goals, different chances of meeting their goals and different planned savings rates. One needs to take all these factors into account, as we do. But let me try to deal with your question by adding some more details. Let's say this person would like to retire at 60, with 70% as much income before tax as he had just before retiring.
At age 50, he was in a position where he had an 85% chance of reaching or exceeding that goal and was happy with that. And now the advisor calls and says, "Guess what? I have bad news. The chance of your reaching that goal if you retire at 60 now is 40%."
The technology we will be implementing with TD Waterhouse could alert the advisor about this client's situation and flags the advisor to call her client. Things have changed, and it's dramatic enough and material enough that they need to talk.
The client has a number of alternatives. He can decide to increase his savings rate enough to get his chance of retiring at age 60 with a comfortable retirement back up to 85%. He can choose to lower his goals. He can plan to retire later or change his portfolio to either take less risk and lock in some lower bound or take more risk and shoot for the stars.
He's got to reassess his plans, as well as possibly the current investment portfolio. In the case you mentioned, he undoubtedly wasn't diversified enough. Probably the advisor using our system had been screaming at him for months to diversify out of that heavy exposure to the tech sector. But perhaps he said, "Well look, I work in Silicon Valley. I know a lot about this. This is the new economy." By now, it may not be as hard to convince him that diversification is a good thing. This should have been done before, but it certainly ought to be done now.
Simonoff: Decades ago, when you were one of the pioneers developing modern portfolio theory and the capital asset pricing model, did you ever imagine that words like diversification and asset allocation would move into the mainstream lexicon?
Sharpe: Not by any means. When I was first doing some of my early empirical work after I finished my dissertation in the early '60s, it was terribly hard to even get information on mutual funds. Nobody knew much about mutual funds, even though they had been around since the '30s. I remember my first study of mutual fund performance. It took me forever to get reasonable quarterly data over a fairly long period for 50 or 60 funds. It was like pulling teeth. Some people knew about stocks, but very few knew about mutual funds. The vast majority knew nothing about them.
A few years back, I was standing in line at one of our big local electronics stores. The computer jocks in front of me were talking about whether they were going to take job A or job B, and almost the entire conversation was about the specifics of the 401(k) plan, the number of investment choices, which mutual fund companies were being used, not about the programming projects. I thought "Wow" because I would not have anticipated this.
Simonoff: In recent years, modern portfolio theory has come under a degree of criticism. Some people have noticed that globalization has seemingly increased the correlation between different national stock markets and in particular as it pertains to the financial advisory business. Many have questioned whether an investment philosophy that was developed for large pension funds with 40-year time horizons could be successfully adapted and applied to individuals with one, shorter-type horizon of the 10-, 20-year variety, typically. Do you see that as a valid criticism?
Sharpe: Yes and no.
First, on the increasing correlation across countries, there is to some extent an increase in correlation. It's certainly most dramatic in the European Union countries, especially in the euro zone. On the other hand, the correlations aren't perfect by any means, and there are important exceptions. Japan has for many years gone its own way. There are still significant advantages in global diversification. Perhaps not as many as there were or as we thought there were, but that doesn't mean that the advantages aren't there. I would certainly not recommend a fortress America investment strategy for the majority of people. But this doesn't mean that you need to have 50% of your portfolio abroad, either.
Financial Engines' advice technology deals with the specific instruments that are available, their costs, turnover, any tax issues and so on. To generalize is dangerous, but it is true that international funds tend to be considerably more expensive than U.S. funds, and this tends to put a damper on the amounts we would recommend investing if that's the only way to invest abroad.
Your second question concerned the applicability of things that may have been fine for institutions with 40-year horizons, for individuals with a shorter horizons? Two factual issues are important here. Pension funds and endowments ought to have long horizons, but my experience over many years working with many investment committees, boards of companies, endowments and foundations is that the members are human, too. If you can get them to look out to 10 or 15 years, you can consider yourself successful. Sometimes they worry about short-term effects more than they should.
Now, it is true that some individual investors are subject to undue myopia, as the behaviorists have shown. But every advisor knows that it is important to work with clients and help them realize that if they're 50, they've probably got 40 or 50 years left. That's a long horizon, and it's very important to not worry about what's going to happen in the next month or year.
Our experience is that people do need to understand what could happen to them next year or else they'll react very badly if something goes awry, but you want to try to focus most of their attention on what really matters. We use the term "outcomes based investing" to describe our approach to doing this at Financial Engines. We try to enable advisors to focus their clients on an outcome that is central to their decision-making. So if you are 50 or 40 or 30, what we try to do is focus you on how you're going to live after you retire as a key outcome.
What's the range of that living standard? What's the chance it'll be greater than your goal? It's very important for all advisors to try to keep people from being unduly myopic. In this sense I think probably the horizons are not highly dissimilar in the two domains.
Moreover, even if you have a short horizon, issues about diversification become important. If you're worried about short-term risk, it's very important to not take unnecessary and unrewarded risk.
Simonoff: In recent years, there's been some disillusionment with the fund business. You see more investors and advisors looking more at the alternatives, whether it's separate accounts, individual stocks, hedge funds, all sorts of different vehicles. What's your take on that?
Sharpe: It depends. The goal is to give your client an overall portfolio that meets his or her needs, that is highly diversified and will fit with their lives, financial and otherwise. To do this, you need efficient ways of getting large numbers of stocks. It is more efficient to have somebody in Boston put together a portfolio of stocks and call it a large-cap growth fund and use that as a component of the overall plan or to go out and buy a number of large-cap growth stocks on your own for your client or possibly to have a separate account manager do it for you? It's really an issue of economies of scale. As the cost of building portfolios and individual securities comes down, it brings down the break-even point where it's more sensible to do it on your own. The right answer depends partly on the size of the client's account, the way in which you organize the process and how the costs are covered.
There's a lot to be said for to the economy of scale of mutual funds, ETFs and the like. For example, ETFs allow you to do some things, especially on the tax front, that can be very useful for a client. If you have somebody who's worried about being laid off in Silicon Valley you can craft a portfolio that's light on technology but nonetheless covers the rest of the market with ETFs quite handily and at low cost.
Simonoff: OK. Earlier we talked about outcomes based investing. Is this really a reaction to what a lot of people have, either intentionally or unintentionally done, to extrapolate past returns into the future? If you applied that principle to individual stocks you would have had Cisco representing a large percent of the U.S. economy.
Sharpe: No, not in any direct sense. First of all, the earliest writings, at least mine and that of others like Markowitz, emphasized the need to do analysis and forecasting based on your best estimates of the future risks, returns and correlations. It was always the future. Now what a number of people did often to excess was to assume the future will be just like the past in terms of risks, returns, and so on. That was a path of least resistance but often highly questionable. It is outrageous to do this for individual stocks, and academics figured this out quite a while ago. It doesn't make any sense to just project past returns and risks stock by stock. If you put such numbers in a Markowitz type of optimizer, you'd get a portfolio that nobody in his right mind would invest in.
So it was always about the future and it was always about how to use the past, along with other information, to make sensible estimates of the future. At Financial Engines, we use the past information where it is relevant-for risks and correlations-weighting the more recent past more heavily.
But as I have said, we use macro-consistency considerations for expected returns, and we take into account a lower risk premium than long-term history might suggest. When we deal with mutual funds, we do not use their historic returns in isolation, but we project future returns based on their historic returns, expenses, turnover and the variability of their returns. So we have a much more complex but sensible and realistic view of the future than if we just projected the past. For stocks, we basically pay almost no attention to the things that made one stock do well compared to another in its industry. Our assumption is, in a way, based on a return to normality, which anybody trained academically would consider appropriate.
Simonoff: I'm told, for instance, that two years ago, your model would have shown you had a decent chance of a 75% down side on certain high-flying tech stocks.
Sharpe: Yes, we had those conversations. Of course, many said, "But this is my company, which is the future of the world. You're telling me there's a 75% chance that if I put all my money in this stock, I could be destitute"-and we said yes.
Simonoff: What are some of the possible problems that you see with this approach to outcomes based investing?
Sharpe: Fewer concerning risk forecasting than some other aspects. If a stock is in high technology, we've got a pretty good idea that it's very risky, and we project that into the future. If you have a portfolio of stocks all in technology, all in Silicon Valley, we know that they've moved together heavily in the past. We can reasonably assume they're going to move together reasonably in the future. The past tells us a lot about risks and correlations, and we use that information in a very direct way.
As far as returns, the simplest way to say it is, we're projecting a return to normality, whether a stock has done especially well or poorly in the past, we know that it may not repeat this performance in the future. The center of our probability distribution is based on a normal return, given the attributes of the stock, not on the past average performance.
What are some issues? Needless to say, it's tempting to say there are none. One problem we wrestle with every day is this. We try to tell people, "Don't worry about needing to know the nuances of the investment industry, mutual funds, Sharpe ratios or standard deviations. Your goal is to figure out what range of outcomes is best for you and how much you need to save, when you're planning to retire and how much risk you want to take. These are the decisions that need to be made. What matters is the range of outcomes that flow from various alternative sets of decisions you could make." As much as you say this, people may still fixate unreasonably on the investment vehicles. Do they have the fund run by this really hot manager that I just saw on the cover of something?
Simonoff: Garrett Von Waggoner.
Sharpe: That's a pitfall.
More generically, people need as much help as they can get. We do everything we can in the services we provide through employers to make these issues transparent and obvious and as friendly and as simple as possible. It can be far better in many cases to combine this kind of technology with the services of an advisor who knows the client, who knows how to communicate with that client, and who knows the circumstances. If an account is large enough to make that kind of human involvement economic, it's by far the best alternative.
Simonoff: One criticism of Monte Carlo simulation techniques is that they typically project financial outcomes according to a normal distribution when in fact the distribution of returns and financial assets is hardly normal. How can that problem be mitigated?
Sharpe: First of all, Monte Carlo simulation is powerful enough to take into account non-normality, log normality, serial correlation, you name it. It's not a technique that requires overly simplifying assumptions about distributions.
It's the people who don't use Monte Carlo who must use extremely simple assumptions. Now it is true that a lot of people who use Monte Carlo continue to use simple assumptions, but they don't need to do so. If you know how to use the method, you have the luxury of representing the way in which returns can evolve more realistically. What do we do to achieve more realism? First, we have a layered approach where we start with a set of core macro variables-a term structure of interest rates, inflation, overall productivity of the economy and inflation. In our simulations, these evolve in ways that are realistic in the two senses. First they are correlated contemporaneously-for example, interest rates tend to be correlated contemporaneously with inflation. Second, they also evolve so that they are correlated serially, for example, it's more likely to have high inflation following high inflation than it is to have high inflation following low inflation.
We go from those variables, to asset class returns, and then from there to funds and stocks. And for returns, we take compounding into effect, which will mean that a return distribution-even if it's normal on a year-by-year basis-will tend toward log normality. You have a 40-year horizon and some money will have been there a year, some money will have been there two years and so on. So the distribution of the final value is going to be a complex combination of normal, log normal and other distributions.
So, I can't tell you that the final distribution for a given client will be of a particular form. It's not that simple because the world isn't that simple. It's a lot harder to do all these things, but you have the luxury with Monte Carlo to do them if you want to take the time and work hard enough.
Simonoff: Reportedly, Financial Engines' 401(k) business at the corporate level has really taken off in the last 18 months. Do you think that that's correlated to the stock market?
Sharpe: Yes, we've seen tremendous growth. Currently, 700 organizations have contracted with us to provide advice to 2 million employees. I think there are two forces in terms of how the stock market impacts us.
The fact that the market fell should have helped us in the sense that we're the "risk people." We're the people you don't normally invite to the party because we keep saying, "The market could go down."
During the glory days, we were out there crying in the wilderness saying, "Watch out. If you take more risk you could lose more money." So in some sense having that "correction" should have helped us as people said, "They were right, it is important to focus on risk."
Another force in the partnership area may have slowed us down. As the market fell, financial services firms took a big hit. And a lot of them chose to defer some development projects or at least put them on a slower schedule and a smaller budget.
Which was the stronger force? It's hard to say. Fortunately we're beginning to see signs that the financial services industry is beginning to get back to business and doing some of the development that is desperately needed.
Simonoff: Are you concerned about any of the bills in Congress that are related to 401(k) advice and allow the people who provide the funds to also provide the advice?
Sharpe: I speak for the firm and for myself when I say that the company's goal is to provide really good advice to as many people as possible. That's a motherhood and apple pie message, but that's what everybody should be working toward. Where people begin to differ concerns views of what is good advice and to what extent can the recipient of advice tell whether it's good? If you want to be paternalistic and you're worried that the recipient can't tell what's good advice, you may want to put some controls on the kind of advice that's given or by whom it's given due to possible conflicts of interest. It's hard to know. In the 401(k) area, it's certainly been possible for employers to provide our advice and that of several of our competitors to their employees for some time. We like to think our advice is pretty good, and we know we don't have conflicts of interest because we don't get paid any more if people put their money in fund A rather than in fund B.
Simonoff: What are some of the other trends that you think are likely to dominate the investing and financial world in the next decade?
Sharpe: You heard my remarks at the TD Waterhouse Institutional Services conference. My crystal ball has been reasonably good for a while, except that I uniformly underestimate the amount of time it takes for something to happen. I don't know about 10 years. Simple economics argues that if you lower the price of something, you get more of it. We're lowering the price of customization in the financial services industry.
I like to make an analogy there to Dell Computer. To the extent that people differ, we will be able to increasingly tailor investment strategies and vehicles to the individual. You're going to see more of this. You're going to see closer ties between investment and insurance, both life insurance and payout annuities.
You'll be able to go to your financial advisor, craft your overall financial plan, implement it and keep it up to date and change it as conditions change. Some of the changes will be automatic. Some will be done manually from time to time.
There are some really fascinating issues as to how you deal with what derivatives people call counter-party risk. If I buy life insurance, I'd like to know that my insurance company's going to be around to pay my heirs. The company gets my money now, and my heirs don't get their money until 40 years from now if I live long enough. So how do we provide assurance? A combination of guarantees, a government backstop, credit rating agencies or reinsurance? You can go through a whole litany of ways in which we've dealt with these issues in the past. How is it all going to sort out?
I would certainly think the financial services industry is going to change at least as much in the next ten years as it has in the last ten, once everybody gets back to business and starts developing and competing again as significantly as they did in the last ten years.