Is today’s “sophisticated” advisor actually that sophisticated after all? A growing number of academics and leading investment thinkers believe they are not.
Some, like Bill Jahnke and Peter Bernstein, criticize methods commonly used by advisors. These folks eschew econometric modeling and mean variance optimization to build portfolios in favor of fundamental analysis relying on GDP growth, the yield curve, corporate earnings growth, and matching investments to needed cash flows. Other leading thinkers, like Roger Ibbotson and Mark Kritzman, rely on optimizers and econometrics. But they say it is common for advisors to misuse these methods, and add that advisors who don’t use econometrics and optimizers at all are basically guessing about investments with other people’s money.
No matter which side is right, advisors don’t look so smart in the eyes of the great thinkers of our time. The pro-optimization believers like Ibbotson and Kritzman say advisors using optimization to create model portfolios or individual client portfolios often do it wrong and misapply the tool. Not using optimization is tantamount to guessing, in their view. Then there are opponents of optimizers, like Jahnke. He says portfolios should be matched against long-term client goals and the cash flows needed to fund them, not optimized for a one-year risk and return horizon.
Take the typical sophisticated advisor’s approach to giving investment advice. When a client walks in the door, she is given a questionnaire asking about assets, income, and debts. Then to assess risk tolerance, the client fills out a questionnaire asking how she feels about skydiving and motorcycle riding.
Then, the sophisticated advisor takes a model portfolio he has built or that he has assembled in his mind, and customizes it to the client’s individual needs. If the advisor is really sophisticated, he shows his client an efficient frontier–a continuum of portfolios trading off annualized standard deviation for annualized return–and picks the best asset mix given that particular client’s acceptable risk level.
Armed with the asset class breakdown, the advisor then selects particular investment vehicles–stocks, bonds, mutual funds, ETFs, annuities, and other pro-ducts–to fulfill the optimized asset mix picked for that client.
According to the brightest financial minds of our day, practicing this way raises lots of questions. Some of the big ones:
o When you use a risk tolerance questionnaire to select a point on the efficient frontier for your client, what does skydiving have to do with his ability to tolerate risk? Should you be relying on these proxies for risk and accept that they are reliable?
o Should you be so focused on figuring out a client’s aversion to short-term volatility–one year’s standard deviation–when the client is trying to fund a long-term goal?
o What effort was made to ensure the inputs in your optimizer are smart?
o Have you thought about weighting your predictions for expected returns on specific assets against history while factoring in how confident you are about your prediction, or using optimization software that does this calculation for you?
o Are your return and standard deviation inputs based on history when they’re really supposed to be based on expected returns
o When telling a client about the optimization process, do you explain that the point you pick on the efficient frontier for his portfolio is a crude estimation of the future and far from scientific?
o Have you used software tools to show yourself and your clients that there are many efficient frontiers that might allow them to reach their goals?
o Rather than using a risk tolerance questionnaire to select a single point on the efficient frontier for a client’s asset allocation, should you perhaps be using software to find the precise point on the efficient frontier that is most likely to help a client reach his goals?
Having spent the better part of two weeks researching new enhancements to optimizers and speaking with the people researching and writing in scholarly investment journals about cutting-edge techniques, I’m now certain of two facts. First, I’m not smart enough to tell you whether Roger Ibbotson, Dick Purcell, Mark Kritzman, or Bill Jahnke are right about their respective approaches to using an optimizer, building a better optimizer, and applying historical returns to predict the future. It’s kind of like religion, and only you can decide who’s found the right answer.
Second and more important, there is another world of investing out there that financial planners and most other independent investment advisors are not living in and don’t know about. It is a world that looks down at most of the work done by independent advisors as crude, naive, and misguided. It is a world of ideas and tools that independent advisors have ignored, been too frugal to consider, or, perhaps, found too complex and new. But these new ideas and tools are bound to come your way over the next few years. If you choose not to learn about them, your competitors may soon learn to build better portfolios than you do, and your business could suffer.
Two companies that make enhanced optimizers hosted Web demonstrations at my request: Ibbotson Associates, a respected advisor toolmaker based in Chicago, and PlanScan, a tiny company out of Boulder, Colorado.
Let’s look first at Ibbotson’s new tool for resampling. To be honest, it’s not all that new, having been introduced as the major new enhancement to Ibbotson’s EnCorr suite of products a little over a year ago. EnCorr is Ibbotson’s institutional-strength analytics software package. It is mostly used by pension funds, banks, money managers, and other professional investors, along with a small number of advisors willing to plunk down the $10,000 to $15,000 annual licensing fee. However, Ibbotson does sell a trimmed-down version of this optimizer program with its resampling engine for a more reasonable $5,000 (currently discounted to $4,000), plus an annual $4,000 license fee.
Resampling, according to Ibbotson Research Director Peng Chen, has been around for years in the academic world. But it has pretty much been ignored among professional investors. It required too much computing power until the advent of speedier processors in the last few years made it possible to run resampled portfolios. “Resampling is a systematic way to account for the potential estimation error in the inputs you use for mean variance optimization,” he says. MVO requires forward-looking returns. Historical returns provide only a sample for estimating forward-looking returns. This creates room for error.
Many advisors gave up using optimizers in recent years because they relied so much on history and saw their optimizers as nothing more than a fancy rear-view mirror. Others who use MVO rely heavily or completely on historical returns and standard deviations for their inputs.
Mark Kritzman of Windham Capital Management in Boston, has pointed out that Harry Markowitz, who won the Nobel Prize in economics in 1990 for work outlined in his groundbreaking 1952 article, “Portfolio Selection,” said from the very start that he used expected and not historical returns for his optimization inputs.
The means and variances plugged into an optimizer can be guided by history but require judgment about the future, and expected returns. But Chen points out that the trouble with using expected returns is that we don’t know the future. The past only provides a sample of all the possible returns, simply because it cannot include the future. To manage this “estimation error” inherent in your inputs, Ibbotson’s optimizer resamples an efficient frontier using Monte Carlo simulation. Instead of relying on a single efficient frontier as an illustration of your possible portfolios, Ibbotson’s software lets you randomize the returns to see 150, 250, or more, efficient frontiers around your original one. It creates an average efficient frontier based on the many simulations. The result is a portfolio that tends to be better diversified and that won’t need to be constrained to avoid dominance by any one asset class.
The other tool to enhance optimization comes from Dick Purcell. An out-of-the-box thinker, Purcell spent his career writing and teaching CPAs about finance after graduating from MIT and Harvard. Purcell, a fiery critic of the education and training for CFPs, has programmed optimization software that runs Monte Carlo simulations all along an efficient frontier to find the optimal portfolio on the frontier for achieving a client’s goals. Instead of focusing on the probability of achieving an annualized return at a given level of risk, Purcell’s Portfolio Pathfinder software can show clients the likelihood that a portfolio will allow them to achieve their long-term goals. “An efficient frontier graph tells us nothing about where along the curve a client portfolio should be,” says Purcell. “It tells us nothing about which portfolio is better and which is worse for helping a client achieve his long-term goals.”
Purcell says the current method for choosing a point on the efficient frontier for a client is administering a risk-tolerance questionnaire, which he calls nonsense. “There are 141 topics in the CFP investment education module, and not one of them addresses how to find the best portfolio for your future dollar goals,” says Purcell, who taught about business graphics for the American Institute of CPAs for many years and authored Understanding a Company’s Finances–A Graphic Approach, for Houghton Mifflin in the early 1970s. “Instead, thousands of CFPs are trained on a curriculum foisted on them by more than 200 universities that diverts attention from lifetime goals to single-year volatility and returns.”
Purcell says that making a decision about which portfolio a client should get based on a one-year standard deviation versus return payoff ignores the benefits of long-term compounding on investments. “For example, at a rate of 10%, compared to the gain over 10 years, the gain over 20 years is over 3.5 times as great,” he says. “For this reason, increases in expected return rate that appear very small on the single-year efficient frontier produce far larger increases in expected long-term return. To see this advantage, you have to move beyond the efficient frontier’s single-year view, to compare the portfolios in a long-term compound return.”
By running a Monte Carlo simulation on a range of 12 or so points covering the efficient frontier, the client can see the likelihood of meeting his goals over his lifetime by using one portfolio versus another. Purcell says this is a critical difference from the way optimizers are currently used because it shows the client that one portfolio on the frontier is likely to bring him closer to achieving his long-term goals than another portfolio on the efficient frontier. In addition, it shows the client the optimal portfolio over the long term, taking into consideration the effects of compounding and shrinking deviations on asset classes over the long haul. Purcell’s software is based on using ETFs and index funds to fulfill portfolios because they can be true to the optimization of the asset classes.
Things are heating up on the portfolio management front at Fidelity Institutional and Schwab Institutional. Fidelity has signed a deal with a portfolio accounting and performance reporting software company that is already a big player in the institutional asset management business. Schwab, meanwhile, is releasing the long-delayed SQL version of its Centerpiece software.
Fidelity’s deal is with Los Angeles-based Integrated Decision Systems, which was founded in 1981 and has been serving the wrap account, prime brokerage, and asset management units of large brokerages including UBS, Morgan Stanley, and A.G. Edwards. It is also the performance reporting system for the retail version of Fidelity’s Portfolio Advisory Services program.
IDS claims $4 trillion of assets on its system, but until the Fidelity deal did not serve RIAs. Schwab’s former chief of technology marketing, Ron Lovetri, who was laid off by the discount broker two years ago amid massive bear-market cutbacks, has been hired by IDS to run the marketing program.
Lovetri says IDS “would not know what to do” if RIAs individually called to buy the IDS reporting solution. IDS, at least for now, will serve the RIA market solely through Fidelity, he says. However, Lovetri says the deal with Fidelity is not exclusive and IDS in time is likely to open the product to RIAs who do not work with the Boston-based company.