IvyVest's Rules-Based Dynamic Asset Allocation Model: A White Paper

by Alex Frey, CFA


This white paper summarizes the motivation, design, and execution of the IvyVest Dynamic Asset Allocation Model. Its intent is to explain the principles of the investing approach that the model encapsulates in a way that is rigorous, but that assumes little or no prior knowledge about dynamic asset allocation.

If you are a complete investing beginner, it is recommended that you start by reading A Beginner's Guide to Investing. In that book, it is argued that a rational approach to investing is one that follows three key principles:

  1. Focus on asset allocation. Its importance dwarfs everything else.
  2. Minimize costs. In a highly uncertain world, they are the one thing that can be completely controlled.
  3. Follow a system. Taking emotions out of the investing process prevents costly mistakes.

The IvyVest approach is not the only method of investing that meets these requirements. But of the approaches that do, it is the one that the authors find to be most optimal for individual investors.

This paper will proceed in three main sections: it will start by providing a high level view of what the IvyVest approach is and what a “rules-based dynamic asset allocation model” means; it will then compare and contrast rules-based dynamic asset allocation with the more common approach of static asset allocation; finally, it will look at the specifics of how the IvyVest rules for asset allocation were researched and constructed.

The IvyVest Model in a Nutshell

The goal of the IvyVest Model is to find the single mix of assets that most optimally balances risk and reward for the future. For an investor with perfect foresight and an iron stomach, this portfolio would be composed of whichever single stock or bond would offer the highest return over the investor's investment lifetime. But since nobody has perfect foresight nor a totally-iron stomach, the optimal portfolio can be defined as the mix of assets which has the highest risk-adjusted expected return. Due to the diversification benefits that come from combining imperfectly correlated assets, this portfolio will likely be composed of a variety of different assets.

Expected returns are not directly observable, so determining optimal portfolio composition requires dealing with uncertainty, and making some assumptions and estimates. There are competing approaches for going about this. The approach that the model takes is one known as rules-based dynamic asset allocation.

The “asset allocation” piece of this description indicates that the approach involves selecting from large categories of assets, rather than selecting individual stocks or bonds. Among other things, this assures that the portfolio is highly diversified. The asset classes featured in IvyVest's approach are:

  • US Stocks
  • European Stocks
  • Pacific Stocks (from Japan, Australia, Singapore, Hong Kong, etc.)
  • Emerging Markets Stocks (like China, India, Brazil, Chile, South Africa, etc.)
  • Treasury Bonds
  • Treasury Inflation Protected Securities (TIPs)
  • Commodities
  • Gold
  • Real Estate Investment Trusts (REITs)
  • International Real Estate

What makes an asset allocation model actionable is that it is easy to get exposure to each of these asset-classes by purchasing a single exchange traded fund (ETF). ETFs are traded on major stock exchanges just like individual stocks, so they can be cheaply and easily purchased in any discount brokerage account. IvyVest subscribers can trade most of these ETFs free of any commissions if they have a brokerage account at Charles Schwab, Fidelity, Vanguard, or TD Ameritrade .

An asset-allocation centered approach is not unique. In fact, it is the kind of approach that most financial advisors and advisory firms follow. This is true both of traditional firms that operate on an in-person basis and new ones that operate primarily online. What sets the IvyVest approach apart from most these of the other firms is that the IvyVest model is that it is both rules-based and dynamic.

Under a philosophy of dynamic asset allocation, the optimal amount to put in each asset-class may change from one time period to the next. The philosophy of the vast majority of the industry (including all of the new online firms) is static asset allocation. Under a “static” approach, the optimal asset allocation is initially based on a measure of a client's risk tolerance, and it never changes going forward, except to adjust to changes in that risk tolerance. With dynamic asset allocation, changes in market prices may drive changes in the expectations of future returns, which then may drive a change in asset allocation. A key point is that changes in the market are driving changes in asset allocation — if the market always performed “as expected” then a dynamic asset allocation would be identical to a static asset allocation. But history shows this is rarely the case.

IvyVest uses a “rules-based” approach to dynamic asset allocation that automatically changes based on market data. The minority of firms that advocate dynamic asset allocation predominantly rely instead of personal judgement. An advantage of a “rule-based” approach rather than one that relies on personal judgement is that judgement can be highly distorted by the emotions of the moment, while objective data cannot.

The next two sections will review each of these principle. First, a brief look at what the model is not.

What the Model is Not

Before delving into exactly how the IvyVest model works, it is important to understand a few things that is not:

  • The IvyVest model is not a stock-picking model. The IvyVest model does not attempt to find individual securities that are going to “beat” the overall market. Instead of owning Coke or Pepsi, the model owns small pieces of both — along with the 3743 other securities that are included in a US Total Stock Market ETF.
  • The IvyVest model is not a short-term trading or market-timing model. The model was not built to “predict” what direction the market is going to go in the short-term. What this means is that in years when the overall stock market declines, a portfolio that follows the IvyVest model will also likely decline, but hopefully it will do so by a smaller amount than the overall market. In 2008, for instance, a portfolio following the IvyVest rules would have decreased in value by a bit more than 15%, versus a 50%+ fall in the US stock market.
  • The IvyVest model is not a discretionary model that relies on any one person's skill or intuition. Most investment-managers make decisions in a reactionary way. Something happens, and the investment-manager reacts to it by buying or selling depending on how it makes them feel. IvyVest takes nearly the opposite approach. Rules are determined in advance, and then are mechanically followed once “something happens.” This eliminates destructive emotional responses to market developments.
  • IvyVest will not invest your money for you. IvyVest's service was built for do-it-yourself investors. Those willing to devote an hour or two every quarter can easily follow the strategy from any discount brokerage account (we recommend Vanguard or TD Ameritrade in order to minimize trading costs). But IvyVest does not take your money and manage it for you.

Rules-Based Dynamic Asset Allocation vs. Other Investing Frameworks

A necessary first step in building an investing model is to decide what approach it will follow. IvyVest's model follows a rules-based, dynamic asset-allocation framework. This can be composed into three components: (1) Rules-based, (2) Dynamic, (3) Asset Allocation. The section below will explain each, starting with the last.

Why focus on Asset Allocation?

A basic feature of the IvyVest model is that it focused entirely on asset allocation, rather than on picking the best individual stocks or bonds from the bottom up. This was an intentional choice. This section will look at why that choice was made.

A starting point in approaching any investment is to understand that financial markets have trended strongly upwards throughout history. The US Stock Market, for instance, has a compounded annual average return of more than 10% since 1900. Though returns going forward are likely to be somewhat lower than they have been in the past (for a variety of reasons that go beyond the scope of this white paper), there are strong economic reasons to expect that broad markets will continue to go up over time. That is, there are compelling reasons to believe that under a capitalist system, expected returns will always be positive (if the market is operating correctly).

The most basic reason this is true is that investors require the expectation of a positive nominal return in order to hand their money over to someone else. A rational actor would never make an investment that he expected to lose money on; he would instead choose to either consume things now, or keep his money in cash (“under the pillow”).

In order to persuade capital suppliers to hand over their money, capital users must entice them with the expectation that they will eventually get out more money than they put in. Companies can do this by using capital to build useful things that will generate a real return over time. For instance, if a new automobile company like Tesla uses investors' capital to build a new factory, it will create value for its shareholders if, over time, it is able to generate more in profits from manufacturing the resulting cars than it cost to build the factory.

Investing, therefore, serves a useful purpose, as companies are able to tap the capital markets for money that they use to build businesses that in the future will produce things that consumers want to buy. The resulting profits are then given back to shareholders as dividends and bondholders as interest payments. Value is created.

So long as capitalism continues to function, this equilibrium should continue to hold: investors will demand a positive return in exchange for investing their capital in risky projects, and companies will be able to deliver it (over time) by building things that satisfy the future demand of consumers. This is why investing in aggregate is a positive-sum activity. One investor can do well without taking money from another investor, because the markets as a whole are expected to up over time. Everyone can win.

The easiest way for an investor to capture the positive-sum nature of investing is to purchase an index fund. An index fund is a special kind of mutual fund or ETF that attempts to exactly match the performance of an overall market by buying a small piece of every individual security in the market. For example, there are index funds that purchase a small piece of every US-based stock so that they will match the performance of the overall stock market. An investor who buys such an index fund will be guaranteed to receive the market-average return, minus the (usually very small) fees paid to the administrator of the fund.

Of course, it would be better if an individual investor could select only the individual stocks that are going to achieve greater returns than the overall market. The problem is that this is difficult to do. One of the reasons that it is difficult to do is that beating the market is a fundamentally different kind of game than merely investing in the market.

Just as it is impossible for everyone to be a better-than-average student, basketball player, or driver, it is impossible for everyone to be a better-than-average stock-picker, since in order for an average to be an average, about half of participants are required to do worse than average, and about half are required to do better than average. But in investing, the situation is actually worse than in most other pursuits, because there are significant costs involved. Some of these costs are explicit, like money that is spent on trading commissions and management fees; others are implicit, like the opportunity cost of time spent researching investment, or the stress of worrying about how an investment selection is performing. If these costs are properly accounted for, it becomes virtually guaranteed that the average “active” investor's portfolio will see lower net returns than those of the market average, perhaps by a significant margin.

It follows that it is possible to say with a high degree of mathematical certainty that the average active investor would be better off buying a cheap index fund instead. This sounds like a powerful indictment of active investing, but it is actually just middle-school level logic.

Stock-pickers, then, must believe themselves to be significantly better than average at the activity. It is important to note that to be better than average at something is actually a statement about two different things: the skills of the person picking stocks, and the skills of everyone else. Because professionals today dominate the investment industry, the "average" investor in the market today (at least on a volume-adjusted basis) is likely someone who invests as a full-time job. Professional stock-pickers have a number of obvious structural advantages over do-it-yourself individual stock-pickers: they can spend 60 hours a week or more researching stocks, they have privileged access to company management (and potentially other inside information), and they have relationships with a host of other research firms. It seems exceedingly unlikely that many individual investors with a job/kids/life will be able to compete. The conclusion is that most individuals are making their stock selections at a significant informational disadvantage. They are playing a zero-sum or negative-sum game that they are unlikely to win.

One approach to mitigating an information disadvantage is to diversify a portfolio, to buy more individual stocks so that if one or two perform quite badly, others will hold up. What is interesting about this is that as the number of securities in a portfolio increases, the returns tend to converge more and more on the market average. Partly this is just the definition of an average – an investor who took diversification to the extreme would just own every stock in the market, which is what an index fund does, and guarantees a market average return. But partly it is also because investments within the same asset class tend to be highly correlated (move in the same direction), especially during crises. This is why in 2008 it did not matter much what individual stocks an investor was in. So long as the investor was in stocks, his or her portfolio suffered.

This effect is why academic studies have show that asset allocation accounts for 90% or more of the returns of most large real-money portfolios. Even for the best professional investors, the choice of what asset classes to invest in is more important than the choice of what individual securities to purchase. The huge importance of asset allocation and relatively minor importance of security selection (within the context of a diversified portfolio) is another reason to concentrate foremost on asset allocation.

Passive Funds vs. Active Funds

Of course, individual investors who cannot beat the professionals might choose to join them: there are large and well-developed markets for professional asset management in the form of mutual funds and hedge funds. These are known as “active” funds, while ETFs and index funds are “passive.”

However the data shows that even most professional investors are unable to consistently deliver market-beating returns. To cite just one example, the Standard & Poor's SPIVA report showed that over the last five years, 72.72% of large-cap funds lost to the benchmark S&P 500 on a net basis. Small and mid-cap funds that were operating in supposedly "less efficient" sectors of the market fared even worse: 83.94% of mid-cap funds lost to their benchmark, and 74.73% of small-cap core funds lost to theirs. Standard and Poor's publishes this analysis every year, and the conclusion is always broadly the same.

Many mainstream financial commentators have interpreted this to mean that mutual fund managers must be stupid. But in fact, it is just another variant of the phenomenon discussed above: since professionals make up most of the market, the market average return will closely approximate the return that the average professional achieves before deducting fees. After deducting fees, it becomes highly probable that the average professional investor will perform worse than the overall market, especially since professional investors are so well compensated.

This would not matter so much if it was easy to find a "better than average" mutual fund. Empirically, however, the frequent disclaimer that “past performance is not indicative of future results” seems to be literally true: academic studies show that the top funds from one period rarely repeat their feat in another (http://onlinelibrary.wiley.com/doi/10.1111/j.1540-6261.1997.tb03808.x/full#b5).

Some have taken this to mean that the market must be “efficient”, that nobody is able to “beat it” except by random chance. A host of evidence now suggests that this is not true. But there are a myriad of reasons why the top funds in one period may underperform in the next, beyond the possibility that they just got lucky in the first.

First, employees of investment-management firms are “free agents.” Those who do well frequently leave for better paying or more prestigious titles or firms. The manager that replaces them is not guaranteed to be as good.

Managers that do stay in place may still “suffer” from their success. As mutual funds take in more and more money as a result of great performance, it becomes harder and harder for them to find enough attractive investment opportunities to keep generating great returns. Entire classes of investing strategies that might work with $100 million might not work with $10 billion. Client-friendly fund companies will eventually close their funds to new investments before this happens, but there is always going to be an incentive to take in more money in order to charge more fees.

Finally, mutual funds have several institutional constraints operating against them. They are required by regulation to provide unlimited daily “liquidity” to their customers, which means they must process both withdrawals and purchases as soon as they come in. This can prevent the purchase of certain types of illiquid investments that might offer higher returns, and it can require a fund to sell investments at a distressed price in order to meet redemption requests.

But the most rational reason to avoid active management is simply that it is expensive. Typical actively-managed funds can charge an annual fee of 1% of assets or more. Simulations show that at the end of a twenty-five year period, a typical investor will have paid more in fees than he initially invested. In the context of a diversified portfolio, it is hard for even a skilled manager to generate enough “alpha” to make up for this fee-drag (in investment jargon, “alpha” is the return on an investment that exceeds the return of the market average).

All this said, even investors who do believe that they have an approach that works for selecting the few mutual-funds that will beat their averages should still care the most about asset allocation, for all the reasons mentioned above. These investors could still easily follow the IvyVest approach and simply substitute their chosen funds for ETFs. For example, an individual investor could decide to own the T. Rowe Price Growth Stock Fund instead of VTI for US stocks exposure.

What to Do if You Love Picking Stocks

Many people (authors included) understand all of this, but choose to purchase some individual stocks because they enjoy it, and because maybe over time they can become “significantly above average” at it. There are a couple ways to balance pursuing this activity while still responsibly and reasonably investing for the future:

  1. Setup a “sandbox” or “play money” account with a small portion of an overall portfolio. Invest the rest according to the IvyVest approach.
  2. Use IvyVest to set overall asset allocation, but use your own stock picks for the first 10-20% (or more if you really want) of your US stocks allocation. For instance, if the IvyVest model is in 30% US stocks, you could put 15% of your portfolio in your stock picks and another 15% in the VTI ETF to get 30% overall allocation to US stocks.

Most people, though, will be better off focusing 100% on asset allocation and just buying cheap index funds to achieve exposure to each asset-class. By focusing on asset allocation, you are spending your time on the positive-sum activity that is going to drive most of your results, not the zero-sum activity that isn't.

Why use a rules-based system to dynamically adjust asset allocation?

IvyVest's model is a dynamic asset allocation model. As previously mentioned, what differentiates a “dynamic” approach from the more common “static” approach is that allocations change along with the market, rather than always remaining constant. Practically, this means that the IvyVest model's “weights” in each of the ten asset classes change periodically (monthly in this case) based on what is happening in the market. The recommended asset allocation in 2014 will be different than what was recommended in 2012 or 2013, which will be different from what was recommended in 2008.

Investors who use a financial advisor or another investing service are most likely invested in a “static” portfolio. A static asset allocation, much like it's name sounds, does not change. An investor that starts with 60% of his money in stocks, will always have 60% of his money in stocks, no matter whether it's 1999 or 2009.

This section will start by reviewing why many have rejected any kind of 'market timing.' It will then discuss why a “prediction-free” approach to investing is a practical impossibility.

The conventional wisdom that individual investors should “buy and hold” (or, more accurately, “buy and re-balance”) is largely justified by the apparent futility of any efforts to change the allocation of a portfolio based on a prediction of the near-term direction of the stock market. For instance, an often-cited study showed that forecasters with investment newsletters possessed no ability to predict future returns. The newsletter predictions were just noise. Other studies have shown the market-timing abilities of mutual fund managers, pension-fund managers, individual traders, and economists to be similarly lacking.

While most human prognosticators have been shown to have little ability to discern future market movements, a few surprisingly simple systematic rules have been much more successful. For instance, in a widely-downloaded paper, quantitative analyst Meb Faber showed that an investor who followed the simple strategy of owning stocks when the market is above its 200 day moving average, and selling them and moving into cash when it falls below it, would have achieved stock-like returns with a mere fraction of the volatility of investing purely in the stock market. Later work showed that this result was relatively insensitive with regard to length of the moving-average window used.

In a different vein, Yale Professor and Nobel-Prize winning economist Dr. Robert Shiller has shown that a simple measure of the smoothed long-term price-to-earnings ratio of the stock market (now dubbed the "Shiller PE Ratio") is surprisingly successful at predicting the relative level of future stock market returns. When the “Shiller PE Ratio”is lower, subsequent market returns tend to be higher. While the measure is not a crystal ball, it has predicted when returns will be higher or lower than their historical norms with a high degree of statistical significance.

The below charts update Shiller's research with the latest data:

Returns Vs Cape

Shiller 10 Year Returns

The theoretical explanation for these interesting empirical results may come from the emerging field of behavioral finance, which argues that the many of the assumptions of classical economists -- such as the belief that "the market" is composed of totally rational and profit-maximizing actors that possess something approaching perfect information -- are false.

Strong evidence indicates that individual investors are susceptible to over-relying on emotions, and to getting too greedy and too fearful at the wrong times. And the market -- which is nothing but an aggregation of individual investors -- is similarly susceptible to boom-bust swings of the kind that many market observers have noticed and commented on over the past twenty years.

If these results are taken at face value, the case against a static allocation is relatively simple:

  • It would be irrational to hold stocks at a time when their expected return was negative.
  • An investor who follows a static asset allocation will by definition always own the same percentage of stocks.
  • During extreme periods of mania, the market will sometimes go into such a bubble that stocks are priced to have negative expected returns (Japan in 1989 and US tech stocks in 1999 are good examples).

The results above present an apparent paradox:

  • Data shows that the vast majority of market timers fail badly, making any kind of market timing seem like a futile exercise.
  • On the other hand, a wealth of evidence shows that a) the stock market is not “perfectly efficient”, b) Simple measures like moving averages and long-term valuation ratios can predict future returns with a statistical significant degree of accuracy.

Or, to put it more succinctly: it is irrational to time the market, and irrational not too.

This apparent paradox is broken when it is realized that the vast majority of these “market timers” are human beings, who are susceptible to the same behavioral biases that create bubbles and crashes, the same biases that are responsible for the inefficient market. The distinction that matters is likely not between those who try to time the market and those who do not, but between those who follow a systematic and rules-based approach, and those that rely on their human “intuition” to tell them when to invest or not.

Apart from the debate about whether “market timing” is ever feasible or not is the issue of whether it is even possible to practically invest money without making some kind of judgement about the future. Firms that practice static asset allocation through something called “Modern Portfolio Theory” (which, somewhat ironically given its name, comes from the 1970s) would suggest “yes”; basic logic suggests “no.”

The issue is that while it is feasible to invest in an individual asset-class in a “prediction free” manner by purchasing an index fund or ETF that will simply buy the entire asset-class, it is not feasible to do the same across asset-classes. There is no one accepted way of combining US stocks, European Stocks, Emerging Markets, Bonds, Commodities, REITs, Gold, etc. into a portfolio without making implicit predictions about how the individual asset classes are going to perform over some future time period.

To allocate assets without making any kind of market prediction, one might think to use their relative market capitalizations as weights. This might be a valid approach in theory, but in practice almost nobody does this, because it results in a weird allocation that has less stocks than most people find appropriate, and because it is difficult to assign a “market cap” to asset classes like gold and commodities. Under most calculations, this would result in a much lower weight in stocks than most growth-oriented investors would find prudent.

The more typical approach to so-called “prediction-free” investing is to rely on something called “mean-variance optimization” or “MVO.” “MVO” is often lumped with another acronymn: “MPT”, which stands for “Modern Portfolio Theory.” MVO is a mathematical way to determine the optimal static portfolio for any acceptable level of risk. The advantage of this approach is that it calculates a single optimal portfolio, given a few inputs. The disadvantage is that one of the inputs is the expected returns and covariances for every asset class. There is simply no way to divine what these expected returns are solely from looking at the market. To do so requires a prediction.

The way that many firms try to talk their way around this is to use some measure of historical returns as expected returns, so that they can say they are not actually predicting anything, just measuring the past. This is both an inaccurate statement, and a provably suboptimal approach:

  • It is inaccurate, because it is still a prediction: in this case, the prediction is that returns over the future period in question are expected to precisely mirror returns over the past measurement period. And it still requires judgement — the answer will vary greatly depending on what past period is used to calculate returns.
  • It can be proven suboptimal by looking at the case of Treasury bonds. Because the future returns of government bonds are guaranteed (providing no defaults), it is trivial to see that their future returns do not approximate their past returns, because bond yields today are lower than they have been in the past. Government bonds are not unique: stocks and REITs are also priced to deliver very different future returns today than they did in the past. In fact, the approach is not only suboptimal, but it almost guaranteed to be wrong at key turning points. Historical returns will always be highest when the market is in an unsustainable bubble – but that is also precisely the time when future returns will be the lowest.

There's one more problem: even many firms that practice static asset allocation will update their estimates of “long-run market returns” yearly, and may change the composition of their client's portfolios at that time. So for all effective purposes, they are already following a somewhat “dynamic” approach, even if they profess to avoid “market timing.”

Building the IvyVest Rules-Based Model

There are countless different approaches to dynamic asset allocation. As has already been mentioned, the IvyVest approach is a rules-based approach. That means allocations are systematically determined by following a set of pre-conceived rules that have been programmed into a computer. Allocations change when the underlying data changes; they do not change based on anyone's guess or “hunch.”

The IvyVest rules adjust allocations relative to a “neutral portfolio.” Assets that look attractive are given greater than their neutral weights; those that seem less attractive have their weights cut. Thus, a first step in model construction was to determine the composition of the neutral portfolio. Then, rules to dynamically adjust the portfolio were examined.

Constructing a “Neutral Portfolio”

Determining an appropriate “neutral” portfolio is a similar exercise to selecting the optimal asset allocation for a static portfolio. This exercise reduces to answering the question: what portfolio would I own if I had to own it forever (allowing only re-balances)?

One way to determine the best portfolio would be to just look at the historical data and pick the mix of assets that has produced the best returns. This runs into the retrospect bias problem discussed at the start of this paper: this portfolio would likely consist of a single asset (US stocks if certain periods are used, emerging markets stocks if other periods are used). But it is not clear that these will continue to be the best performing assets going forward, and even if they are, many investors might lack the stomach to be able to stick with such highly concentrated portfolios. This approach is too risky and uncertain to be practical.

A better alternative is to pick the portfolio with the best risk-adjusted performance. A risk-adjusted measure like the “Sharpe Ratio” (return per unit of volatility) will favor holding a diversified portfolio because holding asset-classes with similar returns and risks that are imperfectly correlated with each other will reduce the risk of the resultant portfolio without reducing its return.

For a single historical period, calculating the risk-optimal portfolio is a trivial exercise using modern computers. An “efficient frontier” that measures the best portfolio to hold for every level of risk is also easy to find. An example for illustrative purposes is given below, though the authors would stress that this should not be taken too literally since the results are dependent on the time period used, and in this case different time-periods were used to estimate the expected returns of US and European stocks, due to issues of data availability.

Efficient Frontier

This approach, however, has a few practical problems as a means of selecting the composition of the neutral portfolio.

The first is that different time periods can provide starkly different “answers”, and it is not clear which is “best.” A reasonable answer might be to use the longest time period possible in order to get the most data. But though data exists back to the 1800s, it is not clear how relevant that data is to today, since the market was structurally quite different then (there were no mutual funds, ownership was much more concentrated, the US was not a global superpower, etc.). But if only more recent data is used, then the statistical significance of the estimate is sharply reduced, and it may over-emphasize data from a particular kind of period.

In selecting a neutral portfolio for the IvyVest model, historical analysis was used as a complimentary tool. But another source of potential wisdom was also consulted: the rules of thumb that successful investment practitioners have passed down through the generations.

Work by the financial author, risk management professional, and philosopher Nassim Taleb suggests that rules of thumb have value because they are subjected to a kind of evolution by natural selection: ones that work well get passed on, those that fail do not. In this sense, they may be less fragile than a highly optimized mathematical approach that is geared towards a particular historical environment that may not be repeated in the future.

Other risk managers have suggested that rules of thumb mitigate a big limitation of mathematical analysis. That limitation is that it is quite difficult to know from mathematical analysis alone how difficult it would be to stick to an approach over time.

An often-cited rule of thumb in the investing world says is that a “balanced” portfolio should consist of about 60% growth assets (like stocks) and 40% defensive ones (like bonds). For instance, the portfolio that Yale endowment manager David Swensen recommends for individual investors in his book Uncommon Wisdom follows this template.

Using US stocks as the risky asset and US long-term Treasury Bonds as the risk-free asset, various combinations were anlayzed over the 1900 – 2015 period. As is the convention, risk was taken as the standard deviation of annual returns, and return is the compounded annual growth rate (CAGR) over the period. This type of chart is known as an “efficient frontier” since it plots the portfolio with the highest return for each level of volatility.

Though it is not the optimal portfolio in terms of producing the most return per unit of risk, based on this analysis, a 60-40 mix seemed as reasonable a choice for the future as any.

With the mix of growth and defensive assets selected, the next step was to fill in each category. A further “rule of thumb” was added at this point was to diversify each category across different economic environments, with half the portfolio set aside for assets that will do well in periods of rising inflation, and half devoted to assets that will do well when inflation is falling. The idea for this came from Roy Dalio, the founder of BridgeWater, the largest hedge-fund group in the world, and one of only a handful of managers that are able to produce statistically significant “alpha” (market outperformance) while managing tens of billions in assets.

The other principle employed was that “more is better.” In his book Antifragile, Nassim Taleb says that a simple “1 / N” approach, in which the assets of a portfolio are split into equal pieces, is relatively robust since it relies on no fragile historical optimizations. This principle influenced the decision to include ten different assets classes and to hold at least a five percent neutral weight in each: US stocks, European Stocks, Pacific Stocks, Emerging Markets Stocks, US Real Estate Investment Trusts, International Real Estate, Commodities, Gold, Treasury Bonds, and Treasury Inflation-Protected Securities (TIPs).

Within each quadrant, weights were based on the relative market-caps of different markets (for stocks), and on a combination of using rules of thumb and historical mean-variance optimization.

The resulting neutral portfolio contains:

  • 5% Commodities
  • 8% Emerging Markets Stocks
  • 10% Europe Stocks
  • 5% Gold
  • 5% International Real Estate
  • 5% Pacific Stocks
  • 5% REITs
  • 10% TIPS
  • 20% Treasury Bonds
  • 27% US Stocks

Building a set of dynamic rules

Constructing the neutral portfolio requires answering the question: “What portfolio would I own if I had to own it forever.” Constructing a set of dynamic rules for adjusting the weights of that portfolio requires answering that question: “What makes an asset-class attractive or unattractive at a particular point in time.” To answer the latter question, it helps to understand the source of any investment's returns.

Vanguard founder John Bogle likes to break the returns of an investment over a period into two pieces: a speculative return, and a fundamental return. The speculative piece is a result of changes in how much investors decide they are wiling to pay for a stream of future cash flows; the fundamental piece is a change in the expectations of the future cash flows themselves. It is well established that over the long run, the speculative return of an investment should go to 0, since investors cannot indefinitely bid prices up or down with no changes in fundamentals.

The first focus of IvyVest's research efforts was spent on estimating the likely size of the fundamental return. We then later looked at how to adjust this fundamental starting point based on predictions of the speculative return.

It was established above that investments have a positive return because they return cash to shareholders, whether in the form of interest payments (bonds), lease payments (REITs), or dividends (stocks). Thus, an effective way to get a sense of likely future returns is to look at the size of these cash flows relative to the price that investors are paying for them. This is most straightforward in evaluating bonds, which are nothing more than promises by a borrower to pay fixed-sums to a creditor at set future dates. The yield to maturity on a bond is nothing more or less than the internal rate of the bond that an investor who holds it to maturity will receive if it does not go into default. A good rule, then, might be that it in the absence of any other differences, bonds will produce higher future returns when their yield to maturity is high than when it is low. In fact, this rule is certain to be true when looking at risk-free Treasury bonds and measuring the return over the lifetime of the bond.

When building a system to estimate the future return of stocks, a few adjustments needed to be made to this framework. First, the future cash flows that investors will receive are variable over time — there is a growth factor in addition to a current-return factor. Past academic work has suggested measuring returns through the “dividend discount model”, which combines the current dividend yield (dividends / price) with the expected growth rate of dividends. The expected growth rate of dividends is a number that is highly variable and uncertain for any individual company, but when averaging across all the companies in a large market, it tends to be fairly constant over time, as the below chart shows.

Dividend Growth

However, companies today have very different dividend policies than companies in the past. In particular, today companies are more likely to reinvest profits in the business in order to produce higher rates of growth, and they are more likely to return excess profits in the form of share repurchases rather than dividends (it is speculated that this is largely for tax purposes). A wide array of research and basic economic logic suggest that measuring prices relative to earnings is an equally effective way to estimate future return levels. The following shows earnings growth rates over the past century:

Earnings Growth

Both research and the experiences of practitioners shows that one of the best ways to measure future stock market returns is to look at a long-term cyclically-adjusted price-to-earnings ratio that has come to be called the “Shiller PE Ratio”, since Nobel Laureate Dr. Robert Shiller has used it extensively. Shiller actually adopted the measure from Benjamin Graham, who was in turn a teacher and mentor to Warren Buffet, so the measure is sometimes called the “Graham-Dodd PE” (Dodd was Graham's business partner). The measure is also sometimes referred to by its acronym, CAPE.

The CAPE divides the current price of a market by a ten year inflation-adjusted average of its earnings. This mitigates a key problem of the USUAL P/E ratio: that the ratio can tend to look cheapest at times when earnings are the most ahead of their long-term trend, and therefore the most likely to mean-revert, and most expensive when earnings are depressed.

The relative valuations of different markets can change drastically from year to year and decade to decade, as the chart below shows.

Shiller Capes

The neutral portfolio implicitly assumes that the expected return of each asset class in the future is equal to its long-run average past return. But from above, we know that this is not true. Assets that are currently selling for bargain Shiller PE's should see returns that are higher than their long-term average, while those that are very expensive should see returns that are lower than their long-term average.

This suggested that a good starting point for a dynamic model would be to increase weights in cheap asset-classes, and reduce them in expensive ones. The actual way that the model does this is to “flip over” the Shiller PE Ratio in order to measure an “earnings yield” in the same way that bond yields are measured. For instance, a PE Ratio of 20 is equivalent to an “earnings yield” of 5% (1/20). The intuition is that every $1 invested will produce 5 cents in earnings.

The "risk premium" is a measure of how much stocks outperform bonds over time. From the cyclically-adjusted earnings yield (CAEY), a predicted risk premium can be calculated as: CAEY + long-term-earnings-growth - 10 Year Treasury Bond Yield.

The following plot measures the predicted risk premium since 1900, versus the subsequent ten year return of stocks.

Risk Premium

While there is a lot of noise (as there is in every plot of real financial data), there is also a clear relationship. This result is used to adjust the relative weights of US Stocks, European Stocks, and Pacific Stocks, adding weight to regions that are cheap and taking it away from regions that are expensive.

A key parameter in the model evaluates the attractiveness of stocks relative to bonds by calculating a predicted “risk premium” (amount that risky stocks will exceed risk-free bonds). This is calculated as the Shiller earnings yield plus the long-run growth in earnings minus the current yield on ten year TIPs bonds.

For a time horizon that is long enough that the “speculative return” is insignificant, the above approach might be optimal with nothing else added. However, research shows that while valuation is an excellent foundation for thinking about long-term returns, it is a relatively poor indicator of market turning points, as has been confirmed from practitioner experience over the last two decades: when a bull market is going on, it can keep going for longer than reason might indicate. In the time-tested language of traders: the market can stay insane longer than you can stay solvent.

As a result, tests of a pure-valuation approach showed a slight increase in average returns vs. a purely static approach; but the inclusion of valuation factors alone failed to meaningfully reduce drawdowns or risk, and the resulting portfolio tended to be out of sync at key market turning points, shifting into and out of risk assets prematurely. This is most likely due to the fact that the "speculative" return can be very meaningful at key market turning points.

To take the trending nature of markets into account, the model also considers an asset classes' recent price trend. This is commonly known as its “momentum.”

Momentum is one of the most accepted “market anomalies” in all of finance. It has been shown to exist across most markets, geographies, and time periods.

Our work shows that the synergistic combination of value and momentum that works better than either separately. The best returns happen in time periods when stocks are cheap relative to their fundamentals, and the price trend is positive. The worst returns happen in time periods when stocks are expensive relative to their fundamentals, and when they have a negative price-trend. Therefore, the IvyVest model pays special attention to the product of our momentum measure and our value measure.

Value Momentum

Testing the model on past data

A major advantage of using a rules-based approach to asset allocation is that the rules can be programmed into a computer. This provides the benefit of being able to go back and “replay” history in order to see how a particular set of rules would have worked in the past.

The results for the IvyVest set of rules are shown back to 1980 (the start of the dataset used). Both plots show the growth of $1 invested in 1980. The first plot uses a normal scale, while the second uses a logarithmic scale.

Model Wealth

Log Log

While requiring only moderate trading (an investor following it would have traded 3-4 times a year), the model outperformed a “buy and re-balance” approach by more than 2% a year, while cutting the maximum drawdown that an investor would have faced from 38% to a little more than 15%. The objective of the dynamic portfolio is to reduce the losses that investors will face in bear markets, since it is at these times of the greatest vulnerability when so many make the most costly mistakes. The following table shows the performance statistics since 1980. The "benchmark" is just the IvyVest "neutral portfolio", re-balanced monthly.

IvyVest Model Benchmark
Avg. Return (CAGR) 13.2% 10.5%
Volatility (Annualized) 9.0% 9.6%
Maximum Drawdown 15.2% 36.0%

It is important to what this kind of “backtest” can and cannot say about the model. Because a portfolio that followed the IvyVest model would have returned more than 13% a year since 1980, does not mean that it should be expected to return 13% from following the same rules in the future. In fact, it is highly unlikely future returns will be so high, because the starting point today is quite different. In particular, observable yields (the amount of income you get from a $1 initial investment) are much lower across the board now then they were in 1980. So it should be expected that the returns on almost any investment will be lower over the next decade than they have been in the previous thirty-four years.

This does not mean that nothing can be learned from backtests. The intent of IvyVest's rules-based strategy is not to “beat” the market in a good year, it's to protect investors' money from the ravages of bubbles and bear markets. So particular attention was paid, for instance, to how well the portfolio stood up during the 2008 financial crisis, when it was down 15% versus a more than 50% fall in the S&P 500 and a 38% drop in a risk-equivalent static asset allocation approach.

As the disclaimer says, past performance is no guarantee of future results. But all else equal it's probably better to be invested in a strategy that works in the past, rather than one that hasn't (or, as is more frequently the case, one where nobody knows whether it has worked in the past or not).

Appendix 1: Doing It Yourself, vs. Hiring An Advisor

IvyVest is a service primarily intended for do-it-yourself investors.

The alternative to managing your own finances is to pay someone to do it for you. Before deciding to hire an advisor, you need to be clear what you are actually paying him to do, and how much you are paying him to do it. Then you can decide for yourself whether it is worth it.

What an advisor will do

There is a misconception that financial advisors are “stock market experts.” In fact, the personality characteristics and skillset that is needed to be a good financial advisor are quite different than the ones needed to be a good investment analyst or portfolio manager. A good advisor needs to be a people person who is personable, confident, and convincing. Advisors will spend most of their time prospecting for new clients (for instance by giving seminars at a local small business or community organization) and meeting with existing ones.

A good research analyst or Portfolio Manager does not need to be an especially good people person (many of them, in fact, are not). He or she is going to spend the majority of his or her time reading through reports, meeting with company management, and getting deep into the weeds understanding the businesses of companies they are invested in.

A smart, honest, fee-based advisor will realise that:

  1. He is outgunned trying to play the zero-sum stock-picking game. To try to time the market by picking individual securities in between sales calls, client meetings, etc. would be a losing game.
  2. All the data suggests that it is all but impossible to pick winning mutual funds over the long term.
  3. The one thing he can control without a doubt: the fees of the underlying funds that he puts you in.

The conclusion is pretty clear: a smart, honest, fee-based advisor will put you in low-cost index-funds or ETFs from Vanguard. Those happen to be the exact same funds we recommend.

The main philosophical difference between the approach that an honest fee-based financial advisor would take and the approach that Ivyvest takes is that most advisors practice static asset allocation, whereas IvyVest take a dynamic approach in an attempt to avoid bubbles and bear markets.

How much an advisor will cost

The second factor to consider when deciding whether to hire an advisor is his or her cost.

Most fee-based advisors will require assets of $500,000, or maybe even $1,000,000 to get in the door, and will typically charge 1% of assets a year, which will come to between $5,000 and $10,000 a year.

An important point to note about these fees, is that their effect compounds over time, as the below chart shows.

That is assuming that you end up with a fee-based Registered Investment Advisor.

Brokers (who, confusingly, are still allowed to call themselves advisors or “financial consultants”) do not charge explicit fees, but are instead paid by the companies that manufacture the financial products that they sell (these are called “commissions”). Brokers*appear* to be less expensive (in some cases they can appear to be free), but it should be obvious that this comes with a tradeoff: it creates a clear and obvious conflict of interest. A broker cannot truly give a client impartial advice, because he has to sell them on a particular variety of product in order to make any money. It should also be obvious that a broker is never going to put clients in the lowest-cost products out there. Because if they were the lowest-cost products out there, they wouldn't be charging enough to afford to pay a commission. And if he weren't putting clients in products that paid him a commission, he wouldn't be making any money. So if you're invested with a broker, you are not in low-cost products.

Summing Up

Informed customers should know first what they are purchasing, and second how much they are paying for it. Unfortunately, when it comes to buying investment advice, the average purchasers has little idea of either.

Academic evidence suggests that financial advisors add value by preventing their clients from trading too much and making bad decisions (see http://www.iijournals.com/doi/abs/10.3905/jwm.2011.13.4.034). Advisors catering to individual with a very high (north of $1 million) net worth may also provide valuable ancillary services like customized estate planning, tax-planning, and overall financial planning.

Beyond that, neither data nor any kind of logic suggest that brokers or advisors possess much in the way of market-beating investment skill. What you are getting is hand-holding. And in the financial world, hand-holding services don't come cheap.

Appendix 2: Other Questions

The below section answers some of the more frequent questions that IvyVest receives.

How much could you stand to lose by investing in the IvyVest strategy?

There are very few free lunches in the world, so it is critical to understand that any investment in the market may go down. You can think about an investment in the IvyVest strategy as having two components:

  • A highly diversified investment in a collection of ten global markets, made through extremely low-cost ETFs
  • A set of rules for adjusting the allocations to those markets

It has already been written that the most you could have lost using our rules-based approach in the past thirty-four years would have been 15%. To be clear, you only would have lost that amount had you purchased a portfolio at its 2008 peak and sold it at the low in early 2009. In this particular time, the “neutral diversified portfolio” was down 38%, meaning the rules-based strategy “added” 21% by steering a hypothetical investor out of stocks and into Treasury Bonds ahead of the very worst of the bear market.

It's crucial to understand though, that the IvyVest strategy is not a short-term “market timing” kind of model. There is reason to believe that the collection of assets owned will likely continue to go up over time, but it would be silly to give IvyVest “credit” when the overall market is going to down, or to “blame” the company when the market is going down. If the markets go down a lot, it is very unlikely that IvyVest's rules-based strategy will be able to totally protect you. There are no miracles here, only a low-cost, diversified, and rational approach to investing for the long-term.

If dynamic asset allocation is so smart, why isn't everyone doing it?

Do-it-yourself investors have traditionally focused on stock picking. Stocks are just sexier. People like to watch Jim Cramer shout about the latest hot stocks and imagine that they can one day get rich from buying the next Apple. They don't care that the odds are stacked against them; it's the age-old triumph of hope and excitement against reason.

Those that aren't do-it-yourself investors have typically trusted a professional to manage their money for them. And those professionals have a vested interest in perpetuating the status quo of static asset allocation. It makes their jobs a lot easier (which is important since most of them are good at sales, not investing), it removes all pressure for them to either 'beat the market' or to keep you out of the next bear market, and it keeps them employed, paid, and happy.

The premise of this question needs some thought. The idea that you should base a financial decision on what other people are doing seems kind of suspect when “other people” are, by and large, quite awful at investing. The annual DALBAR study repeatedly shows that the returns of the average investor fall embarrassingly below the returns of the overall market. So investing like other people is probably not the best strategy to take.

Shouldn't my asset allocation depend on my risk tolerance?

A common notion for many “strategic” (or static) portfolios is to vary asset allocation depending on the individual's risk tolerance. This approach is theoretically correct, and makes sense to some level, but there are a few problems.

One problem is that it's very hard to determine what risk tolerance really means, or how to measure it. There is a lot of evidence that people's risk tolerances are not static, but rather change with time, with their prior experiences, and with what is going on in the markets. Unfortunately, this makes the concept all but useless practically. The point of measuring risk tolerance is supposed to be to predict how you would react in the event of a market crash, but if your risk tolerance changes when the market crashes, then it is going to do a pretty poor job of that. You would want to know your risk tolerance at the time that the market crashes, not your risk tolerance now.

There's another problem as well. Even if you know for sure that you are more conservative or risk-seeing, it doesn't necessarily mean that you are going to get to your outcome more efficiently by holding a particular kind of portfolio. You might be very conservative when it comes to your personal life or career choices, but still decide that your best hope of meeting your goals is to hold what some would consider a pretty “aggressive” portfolio. In this situation, it might be more painful for you to look at the fluctuations of this portfolio than it would be for someone who is naturally more risk-seeking, but it could still be the right portfolio for you. After all, sometimes the right thing to do in life is not the same as the thing that comes the most easily! It doesn't make sense to us that those classified as “conservative” are automatically put into bond-heavy portfolios (especially with interest rates at rock-bottom levels...), when a more equity-heavy portfolio might have a much better chance of meeting their goals!

While the financial industry would love to justify charging you high fees for “personalizing” your portfolio, most people are better owning the diversified balanced portfolio that offers the best risk-adjusted return. It's simpler, easier to stick with, and it makes more theoretical sense.

Shouldn't my portfolio get more conservative as I get older?

A related belief is that your portfolio should get more conservative as you get older. The idea is that younger investors have more time for their portfolios to bounce back from a significant bear market.

The problem is that you would expect this reduction to show up in the data, and it doesn't. Specifically, we can look at past data and actually compare how investors who followed the typical “glidepath” of reducing equities as they got older fared on various risk and return metrics versus those that did not.

This is exactly what researcher Rob Arnott did in an important study. And he found that reducing equities did not improve either risk or return. In fact, investors who followed the exact opposite strategy — of increasing equities as they got older — were successful a greater percentage of the time in outliving their assets (which is ultimately what counts).

Some investors may still rationally decide to “take some chips” off the table if they near or meet their retirement goals. It is possible to do so without artificially tinkering with the composition of their main portfolio. An alternative approach is to build a "risk-free portfolio." Ideally, this portfolio would be composed of a "ladder" of zero-coupon TIPs bonds. The intent should be to "match" anticipated living expenses with a bond that will mature close to the date that they will be incurred.

Even with a "risk-free" portfolio, most investors should choose to hold some of their money in a growth-oriented, risky portfolio. All but a select-few super-wealthy investors will likely need some growth to meet their goals. And the super-wealthy will likely still decide to take on some risk in order to improve likely outcomes for future heirs, or for charities that they would like to contribute too.

How does IvyVest select the Individual ETFs that it recommends? Does IvyVest receive compensation from the manufacturers of these funds?

When selecting between ETFs that track the same asset classes, IvyVest selects primarily on the basis of the lowest total cost of annual ownership, a measure that includes management fees, implicit trading costs, and the trading commissions that we estimate a typical subscriber would pay. We will also pay attention to how broad of market exposure the ETF provides.

Most of the ETFs we recommend are run by Vanguard, a customer-owned financial institution that is akin to a credit union of the investment world. Vanguard is the second largest fund manager in the world, having more than $1 trillion in assets. We have no business relationship with Vanguard, and get no “kickbacks” from any recommendation that we make. Our only revenue is from customer subscriptions.

As a relatively new feature, we can also customize the ETFs that we recommend based on the discount brokerage that you use, which allows subscribers with accounts at TD Ameritrade, Vanguard, Fidelity, and Charles Schwab to benefit from the commission-free ETFs offered at each.