Richard Suvak, MSF, CFA
Why We Don’t Time the Market
Consider a game of roulette. In America, with zero and double-zero on the wheel, the odds of the ball landing on red is 47.4%. Likewise for black. With a 1-to-1 payout, it’s easy to see why the casino wins in the long-run and why betting on either red or black is a losing strategy. Despite our poor odds, one might be tempted to devise a strategy to shift the odds in our favor and try to beat the casino. Rather than blindly betting on red or black, we could bet on the opposite color after three consecutive results of one color. For example, if results of the previous three spins were red, red and red, we would bet on black for the fourth spin. Similarly, we would bet on red after three consecutive black results, otherwise we would enjoy the atmosphere and free drinks. Some might recognize this strategy as foolish given the law of independent probability. A result of black does not increase the chances of a red result in the next spin. The probabilities do not change for any given spin of the wheel. That’s true regardless of whether we use one spin, two spins, three spins or ten as our betting signal. Nonetheless, let’s see what would happen in our roulette timing strategy.
Running a Monte Carlo simulation of 1,000 spins with an initial bankroll of $1,000 and $25 bets each time our signal produces the following results. By a 3-to-1 margin, we mostly walk away with less money than we had when we arrived. We win less than 25% of the time, lose somewhere between 0-50% about 50% of the time and suffer catastrophic losses 25% of the time. Our signal, based on the false pretext of dependent events, has us leave in a sour mood vowing never to return – possibly the best long-term strategy of any casino gambler.
Like our poor gambler, there are a number of reasons amateur and professional investors avoid trying to time the market. Similar to a casino’s odds, most have to do with the low likelihood of success and negative consequences of being out of the market just when you shouldn’t. Some of the reasons we fail at our attempts are:
We are too emotional about our money – this well documented psychological barrier inhibits our ability to make rational decisions just when the need is most necessary.
As a result of these reasons, we often make bad decisions at the worst possible time, inflicting damage, sometimes unrecoverable damage, to our investment objectives. Therefore, most conclude, the right course of action is to avoid attempts to time the market and invest for the long-term. History generally agrees with this approach; however, results are highly dependent upon the period you invest, the frequency of subsequent investments and the period you sell. Intentionally or not, we time the market by the very nature of our cash-flows. We invest when we have excess liquidity and we sell when we need liquidity to buy a house, retire or have other unexpected expenses. Often, these are not investment decisions as much as they are cash flow requirements.
Professional investors claim not to time the market, but don’t be fooled, they do. Whether they are value, growth, small-cap, large-cap or asset allocation managers, professional investors develop signals to designate an investment as attractive, unattractive or neutral. Based on these designations, they buy, sell or hold a stock, bond or entire asset class. Their signals not only indicate which investments to buy and sell, but when to do so.
Amateur or professional, investors are perpetually searching for a systematic way to beat the market. These systems take the form of smarter stock picking, better insights, more sophisticated analysis, better portfolio management and risk software and better and more consist signals as to the future direction of the market.
Timing the Market – An Example
Whether focused on the long-term or in high-frequency trading strategies, timing the market largely boils down to predicting inflection points – those points where the market changes direction to begin a new trend. To find these inflection points, most models follow one of two approaches, mean-reversion or trend following.
Mean-reversion is a value-oriented strategy. It says that there is a longer-term price which is the anchor for the asset. Deviation from this anchor (often the long-term mean) will eventually revert back to the average. Therefore, any significant deviation from this average is often a signal which indicates a mis-pricing, which can then be taken advantage of. My prior employer was a well-known advocate of this approach and made a few significant market calls which set the firm apart as a serious thinker in market analysis. The firm’s reputation was largely made based on calls of small-cap stocks in the early 80’s, Japan overvaluation in the late 80’s and the breaking of the internet bubble in the early 00’s. All of these bets were made based on the theory of mean-reversion.
Trend following is a momentum strategy. It says that recent returns will continue in the same direction. Followers believe, like a stone rolling down a hill, an asset’s momentum needs an external force, or reason, to change direction. Without a reason, the trend will continue. Often, momentum can be seen immediately prior to a reversal. The internet bubble, before the break, was a case in point. It ran beyond most logical measures of market valuation, leading to Alan Greenspan’s famous utterance of “irrational exuberance” to describe investor behavior at the time.
These two strategies can be implemented using many metrics and with many measurement horizons and many holding periods. Those who implement asset allocation strategies typically use years as their horizon, those who implement a high-frequency trading strategy can use seconds and minutes as their horizon. For sake of simplicity and ease of demonstration, we will focus on the longer periods and readily available data sets for our analysis.
Consider valuation of the S&P 500. Price-to-Earnings and Shiller Price-to-10-Year-Average-Earnings provide commonly used market valuation metrics. Due to the 10-year averaging, Shiller’s P/E is slower moving than trailing twelve-month (TTM) P/E, but they roughly indicate similar market valuation. From these metrics, we can see the 1929 crash, nifty-fifties, 1988 crash, internet bubble and crash and 2008 global financial crisis.
Converting these valuation metrics into a mean-reversion value signal shows how far, above or below, valuation has deviated from its long-term average.
Using these valuation signals to predict forward returns shows the power of value-investing. That is, when value is low relative to its long-term average, the expected return is higher than average, and certainly higher than when valuations are high relative to the long-term average. In this case, we can see that deciles one through four have returns in excess of the long-term average for each of the holding horizons (1, 2 & 3-years). When the market flashes “cheap” and is in these deciles, odds for better-than-average returns are in investor’s favor. When the market is expensive, the odds may be better elsewhere. This is the typical thought process of most asset-allocation managers.
Buying cheap US stocks works when using P/E (TTM) as well as with Shiller’s P/E. What you can also see, is that buying stocks when they’re expensive (deciles 8-10) generally underperforms as stocks typically need to pause, or decline in order for earnings to catch up.
Similar to our roulette system, we can build a framework for when we will be invested as well as when we will not be invested. We are either invested in the S&P 500, or have our cash on the sidelines as we wait for better opportunities. The chart below demonstrates these in or out periods (blue is in, white is out). As you can see, using a simple less-than 90th percentile valuation trigger, there are four periods when we are out of the market. We could have used less-than 40th percentile (or any other number), but you’ll see why I chose 90 shortly.
Similarly, with Shiller P/E, we are out of the market five times since 1872.
Herein lies the problem with timing models… when you’re out of the market, the market often continues ahead without you. Had we used a more stringent percentile trigger, we would have been out of the market more often, leading to bigger deficits relative to the S&P 500.
Zooming into a bit closer, we can see what happens when we use value exclusively as our timing mechanism. During the internet bubble, our signal forced us out of the market, but the market wasn’t done yet. Contrary to all financial logic, investors continued to speculate that internet click-through rate was the “new” market valuation metric and that traditional metrics like earnings, earnings growth and valuation were outdated and did not reflect the new economy we were in. Our traditional earnings-based valuation metric scared us into moving into cash only to see the market continue on without us. P/E was right, briefly, as the market fell back below our portfolio value, but we didn’t get back in quick enough to take advantage. Shiller P/E was shaken to its core and exited well below the market peak, re-entered almost perfectly only to get scared again and then re-enter just as the global financial crisis hit.
Value investors, like Dr. Michael Burry of The Big Short fame, are often quoted saying the same thing, “I may have been early, but I’m not wrong.” To be fair, they’re usually not wrong in the long-term, but the difference between early and wrong can be very difficult to disentangle.
The big external events (the Fed flooding the market with liquidity leading up to the year 2000 bug, and quantitative easing during the global financial crisis) both fooled the traditional valuation metrics of the market. The Fed has such an outsized influence on investor expectations that they ignored the valuation warning signs and sent the market ahead in expectation of future sustained growth through easy monetary policy.
Valuation by itself will have these moments where something outside the market is telling investors to keep buying. Sitting on the sidelines while this happens is a losing strategy.
Perhaps momentum can be the savior. Turning to two oft-used strategies (12-month return and Relative Strength) we can develop similar signals to our valuation metrics.
Which can then be converted into a signal based on the long-term average of that signal.
As we did with our value signal, combing the signal with forward returns, we can see that high-momentum (deciles 8-10) outperform the long-term averages, while low-momentum underperforms the averages.
Converting our signal into in- versus out- of the market, we create memberships. Unlike valuation, momentum, which is generally shorter-lived (has a shorter holding period) trades with more frequency.
The impact of this more frequent trading strategy can be readily seen with our three portfolios. Momentum misses most of the crash of ’29 to gain a sizeable lead, but loses that lead by the mid sixties as it was out of the market too often during the nifty-fifties. It fell behind in the ‘80s and only managed to catch up by missing the breaking of the internet bubble and the global financial crisis. Not too bad as a market signal actually, but you can see the long-term momentum portfolio slightly trails the S&P 500 over the entire period.
Relative strength (at least the version I chose) is helpless altogether.
Looking at the signal since 1989 shows that momentum has indeed done well in missing most of the two major market retracements (breaking of the internet bubble and the global financial crisis). Even relative strength fares better (mostly by avoiding the GFC).
Since we know valuation signals take you out of the market just when the market is heating-up for its last gasp before a fall (melt-up) and momentum signals trade too often, perhaps there’s merit in using both in combination. Doing so, shows that there is only one period that we are out of the market, near the height of the internet bubble.
Avoiding this one period saves us from one dramatic melt-down of the market, albeit a big one!
As we can see from the annual returns to each of the strategies, only the Combination strategy works better than simply being fully, and perpetually invested in the S&P 500.
Of course, as we have seen before, the return since 1989 is more favorable to our combination approach.
Not only does the Combination portfolio beat the S&P 500, Momentum does as well, and valuation is not too far behind.
On a return-per-unit-o-risk basis, Momentum wins but the Combination portfolio beats the S&P 500 and has a higher annualized return. This has also been true since 1989.
So, what’s the lesson from all this? Avoid the pitfalls of a market crash. In fact, if it were possible to create a portfolio which foresaw negative returns over the next 12 months and sat on the sidelines during those periods, we would have something quite special.
Of course, this would be something quite special indeed. Forget market timing, valuation or momentum metrics, we’re into pure black magic at this point. That’s more the purview of science fiction and time travel than it is investment management.
How to Actually Time the Market
As we are not time travelers and don’t live in the realm of future-seers, we must find another way to survive the pitfalls of bear markets. Additionally, since our simplified approach has only marginal benefits relative to a buy and hold strategy, we need an alternative approach to beating the market.
To do so, most in the professional money management business taut their superior stock picking ability, sophisticated research, alpha models, portfolio construction technique and/or risk management process as a means to beating the S&P 500. Untold hours are spent building and tweaking models which are all basically the same. All money managers use valuation in their portfolios. All money managers use growth/momentum in their portfolios. All perform detailed research in the quest to finding the hidden potential in a stock or group of stocks. All manage risk. All optimize the output in an effort to gain the most bang for the buck. Unfortunately, they’re all basically doing the same thing. Most are strong in one area or another. All are weak in other areas. In years when they beat the market, they beat the drum of their competence and shout to the world about the superiority of their approach. In years they lose to the market, they maintain their resolve, list reasons why the market is behaving irrationally and provide reassurance that things will turn in their favor shortly.
All along, they’re charging you fees in defiance of their ability to beat the market. It’s no wonder investors are turning to low-cost ETFs in droves.
As active managers push investors away by their inability to deliver market-beating alpha, investors benefit from a plethora of investment options previously unavailable. In the US equity market alone, we can buy many different subsets of the market which suit our pleasure. Size bands, value bands, growth bands, high and low momentum, subsets of growth, value or quality based on company size, different levels of industries along with combinations of all of the above. Investors can be nimble and easily, and cheaply, move into different parts of the market as professional money managers stay stagnant in their approach and their portfolio objectives.
We can do something they can’t. We can beat the market using an approach that goes against the very fabric of their beings. We can invest in sectors of the market that their portfolios are contractually prohibited from touching.
To start, let’s discuss the broad market. Like the Dow Jones Industrial Average, the S&P 500 is a large-cap representation of the broad market. The DJIA had its day as the most popular and most followed index as the S&P 500 does today. Over time, as we have seen with the P/E and Shiller P/E charts, it waxes and wanes in average valuation and size, but it largely occupies the same space in the spectrum of market valuation and size.
By way of comparison, we can look at the Fama-French portfolios (which are created by segregating size and valuation into six separate portfolios – Large & Expensive, Large & Middle, Large & Cheap, Small & Expensive, Small & Middle, and Small & Cheap). These demonstrate the restrictive nature of broad market indices. When we invest in an index like the S&P 500, the market has to favor that particular segment of the market for us to do well.
What happens when Large & Expensive (as is often the case during the melt-up at the end of bull market runs) is in favor and we’re not invested? Or, what happens when Small & Cheap (as is often the case at the beginning of bull market runs) is in favor? While not on this chart, what about Large & High Quality, or Small & High Quality? Limiting our exposure to one area of the market limits our ability to participate when other areas of the market are in favor.
The inability of professional money managers to react and move, due to contractual objectives of their particular fund, as necessary limits their ability to invest in the area of the market that is moving at any given time. We, on the other hand, can move freely as we find better opportunities.
As John Lennon might encourage, imagine another way. Imagine the market as a sphere. On the surface of the sphere lie all the factors which influence the current price of the market. Valuation measures are clustered in one area as each has a similar influence on the market. Growth metrics lie clustered at the antipodal point of their value counterparts. High quality factors lie adjacent to value while low quality is closer to growth. While the surface begins to get crowded with market specific factors, macro-economic factors have an influence as well. The Federal Reserve Board gets a large space, tangential to interest rates and interest rate spreads. Fiscal policy, tax policy, the debt, deficit and current talking points of the President quickly fill in the holes left by market specific factors. Global competitors, exchange rates and relative valuation and growth of other countries squeeze their way onto the surface as well. The last piece, and perhaps a very significant piece is the mindset of current investors. The way in which they digest information, collate it and collectively push markets in one direction or another might be the biggest factor of them all.
Eventually, the surface is covered with the relevant factors, grouped, sized and spaced according to their importance on the price of the market. From each factor, we can imagine a rubber band stretching from it to the point within the sphere where the market lies. The pull, or stretchiness, of each rubber band being dependent upon the importance of the factor to which it is connected. For those factors which have a significant bearing on the market, the rubber band pulls with greater force. For those with less bearing, less force.
At steady state, the market is positioned directly in the middle of the sphere. Growth is offsetting value, economic factors are balanced, the Fed is not intervening and Congress has been on vacation (not too hard to imagine). As the factors change over time, the tension on each particular rubber band changes in conjunction. More influence brings a bigger tug, less allows more slack. As the price of the market is determined by rubber bands connected to the surface in all directions, the pull and slack of each factor determines the price of the market. For example, as economic growth increases, it pulls the market toward itself and away from valuation as the critical factor. As interest rates change so too does the impact of interest rates on the price of the market within the sphere. Factors pull and fight against the other’s pull with the stronger having more influence on the price of the market. However, the relative strength of these pulling forces change over time as different factors take prominence and pull harder. This cycle continues in perpetuity as the influencers of the market change with each new data release, each new public statement, each new investment which changes market sentiment.
Imagine the market as it moves away from its steady-state price, temporarily settling in to another price until new information or data is presented. New GDP growth, unemployment, new product deployments, ectara all have their say in the movement and price of the market. Of course, this movement causes the valuation of the market to fluctuate as it gets pulled in multiple directions around our 3D sphere. However, the movement can only get so far away from its value factor before the tautness of the rubber band increases, preventing it from moving any further. Eventually, this becomes the dominant force, dragging the market back toward the center as all other factors become secondary. Consistent with the valuation factor above, mean-reversion within the sphere happens as well. As value only explains about 34% of the market’s movement, the sphere also allows for the explanation of the remaining 66% as other factors take precedence in moving the market.
This makes intuitive sense as we have seen the market ignore valuations for long periods of time as other factors take precedence. Consider any market peak or trough as evidence. Tulip mania ran well past reasonableness, lasting until enough investors started asking why they were paying so much for a flower. The late 1990 internet bubble was highly influenced by Federal Reserve policy and growth expectations, ignoring valuation for several years, before investors questioned whether irrational exuberance was rational – as well as Greenspan reversing course on flooding the market with money in order to stave off catastrophe were Y2K to cause havoc. The subsequent crash also brought its own version of negative exuberance as investors panicked and brought the market to its knees. Fresher in everyone’s mind would be the 2008 Global Financial Crisis where poor loan practices seized the flow of money around the world, making that particular rubber band pull tighter than all the rest combined. It crushed the market as it was the only factor that mattered, influencing not only the market, but economic activity as well. The Fed, in an attempt to flex its own muscles, opened the flood gates and let every dollar it had, and trillions it didn’t have, spill out into the world in an effort to stem this particular tide and out-muscle a particularly tough and taut rubber band.
In each of these occasions, the market inside the sphere was pulled in different directions as the influence of each factor and the tightness of its accompanying rubber band was pulled or loosened in turn.
The more quantitatively inclined among us might recognize this description as a multi-factor model. You’re not wrong either. This is exactly how one might describe the influence several factors have on a particular variable – in this case, market returns. The difference however is due to the periodic nature and strength of the factors. In a multi-factor model, the factors and regression period remain constant. For example, one might create a multi-factor model using 6 factors – market valuation, market momentum, market quality, economic growth, monetary policy and interest rates. These factors are regressed, either over the full history or some rolling basis in order to explain the next 1, 2, 5 or 10-years of market returns. Based on the influence of each factor in the regression, you simply multiply each factor’s coefficient by the current value of the factor, add them up and bingo-bongo Bob’s your uncle, you have a forecast for the market’s return.
A decent idea, and one that is employed by many well-meaning managers, but you must start asking some questions. Such as, why are these factors the most important? Why are the coefficients the same over time? What happens to your model when the coefficients change, or their t-stats change? What do you do when another factor, one not specified by your model, takes precedence in the market? Why would 6 factors be enough, or not enough, to explain future returns?
To answer these questions, serious quants begin to introduce Markov-switching models, non-linear regressions, stepwise regression and a multitude of other techniques in an attempt to eliminate or reduce the limitations while retaining the additional explanatory power multiple factors have on their models. All good, and reasonable, solutions to a very difficult problem. However, the spanner-in-the-works lies in the assumption of regression models themselves. Regressions are a mathematical calculation used to relate one, or many, factors to another. As such, they require a great deal of data before a relationship can be robustly explained. In the specific case of financial regressions, they require a lot of historic data to robustly explain future returns. Not only do we not have a great deal of historic data, 30 years is about the maximum for most financial series, it’s complicated by differing frequencies (daily, monthly, quarterly, etc.), different update schedules, multiple revisions and changing definitions. Regressions are further complicated by inflection points. That is, how does one incorporate a change of direction of a factor to explain another factor? If they both change at the same time, great, but what happens when they don’t? How does the coefficient for this factor, over this inflection period, get calculated by a regression? Well, if you’re using a full history and ignoring inflection points, a regression will effectively average all the points, potentially wiping out the information you were hoping to glean. If you run a rolling regression you may have a bit more success for this period, but lose information for the “normal” periods. Running a multi-factor regression to get the “right” answer is as difficult, and possibly more difficult to disentangle and understand the results, as is figuring out how to combine multiple factors without regression.
The inescapable problems inherent in single-factor and multi-factor models led this author to a proprietary State Model methodology. It borrows the positive aspects from the known and accepted modeling techniques while removing the negatives such as timing the market, under- and over-specification of regression and handling of inflection points. In essence, the State Model accepts what is known and what is not known in financial forecasting and, like our spherical model of the world, adopts a best-case approach at any given time.
A State Model allows an investor to move away from simple timing models forcing you to move in and out of one (or more) assets based on forecasted returns. Similar to a multi-factor model, it allows investors to rely on history to determine the best asset to be invested in at a given time without relying on the peculiar insecurities of multi-factor coefficient calculations. Also, like our spherical representation of the world, a State Model allows for a time-varying number of factors and a time-varying influence of those factors without pre-specifying how that calculation should be done.
As the investment world narrows it’s focus on how things are done, asking a few simple questions like; why, how, and is there a better way, have led to a new approach with better results.
The process works like this.
As I have eluded to above, there are scenarios that we (mostly) know what the “best” investment is. Usually, (everything will be qualified) markets have a melt-up at the end of a long bull run. This is psychologically based as investors become complacent about the returns the market can generate and begin thinking nothing can go wrong. Additionally, those that have been calling for an end to the bull market finally acquiesce and re-invest because they have been losing by being out of the market. Buying by both groups pushes the market to newer highs that would normally be unachievable based on the underlying fundamentals of the economy or market. So, if we can stomach the unease at which the premise for investing at this time entails, investing in high growth, high beta stocks during this final phase of a bull market, we usually will do quite well.
Once the buying ends and there is no one left to push the market further, the opposite effect occurs and panic starts to invade investors mindsets. Exiting the market often happens quickly and tends to hit the same stocks that drove the market at the end hardest. Therefore, high quality, high profit, stable, staple stocks tend to outperform the market as it crashes back to reality.
These are two well-known phenomena which propel the market in different directions. The problem, of course, is identifying these and other phenomena so that investors can take advantage of them. Momentum attempts to accomplish this task, and is largely successful at it… once enough time has passed that a new trend, or new group of stocks, can be identified. Momentum though, while able to detect trends over longer periods, struggles when the market is flat or switching between bull and bear trends. In this scenario, momentum tends to identify a trend, just when a new trend is taking stocks in a different direction. In other words, momentum is not enough. Value, on the other hand, tends to give up on bull markets too early, while also entering bear markets too early.
While most investors know of these two particular market phenomena, the trick is to know when to switch your investment focus. As these two are largely psychologically based, the timing can be very difficult to ascertain. That’s why you will see endless punditry trying to guess how the market will move in the next 6-12 months. However, there are many other “states” of the economy and market that can be objectively measured which also indicate a preferred investment for that period. While there are many yet to discover, there are several that work wonders.
The premise of this approach is going to be similar to the in- or out- membership discussed above. When the “state” is observed, we will invest in the accompanying assets (in many cases, there are multiple assets accompanying each state). When a state is not active, we will not invest in those assets. As there are multiple state and asset combinations, we will be moving in and out of assets based on which state(s) we happen to be in at any given time.
Simply put. Know the state(s), buy the accompanying asset(s).
Returning to our familiar membership chart, we can see how this works. This particular example State is indicated by the dark blue bars. When we recognize that we are in this State, we buy the accompanying assets and hold them for a period indicated with the light blue bars.
The effect of buying these assets when this particular State is active and holding them for the indicated holding period, provides a cumulative return pattern as below. As you can see, with this particular State, we are more out of the market than we are in the market. That’s okay, as you’ll see shortly. One thing you can also see, is that when we’re in the market, the assets we’ve chosen perform quite well relative to the S&P 500. That’s the beauty of this approach. We’re buying the most favorable assets based on a particular State of the economy and market.
We see the same thing happen in different States as well. We generally buy assets which outperform the S&P 500 given the particular State we happen to be in.
Once we accumulate enough States and accompanying portfolios, we can start to build our final State Portfolio. Because we have overlapping States with overlapping holding periods, we are rarely fully out of the market, and only in the most favorable assets given the State(s) we observe at any given time.
The cumulative effect of these States is a final State Portfolio which dramatically outperforms the S&P 500. The start is only hampered by the fact that there isn’t enough data to confidently identify States early in the process, but wanted to be fair in the presentation. While there are a couple early out-of-the-market periods, the final State Portfolio is fully invested the remainder of the time, only more profitably than the S&P 500. This is due to the timing effect of choosing higher performing assets during each individual State events.
We can see the ebb and flow of these States when we look at the number of holdings across time. Each State Portfolio has a varying number of accompanying assets (Sub-Portfolios in this case) and therefore we see spikes in Sub-Portfolios whenever the number of holdings in the Portfolio spikes. Interestingly, we see a smaller number of States, and therefore Portfolios, as we approach peaks in the market. Effectively, we’re seeing fewer opportunities in the market as the States at those time offer fewer market-beating opportunities as other times. A corollary also occurs. As the peaks turn into troughs, we see more states and more market-beating opportunities and therefore our number of holdings increase.
Turnover is always important to portfolio managers as it incurs transaction costs. While the expectation was for this strategy to have untenable turnover, the average turnover is quite reasonable indeed.
Using different colors to represent the concentration and migration of holdings, we can see shifts in the State Portfolio over time. The dramatic vertical switches in portfolio positions represent new States and the transition to their accompanying assets.
Using Fama-French’s 5-Factor portfolio on a rolling 3-year basis, we can compare the differences between the State Portfolio and the S&P 500. As you can see from the two charts, the State Portfolio is more exposed to the Conservative-minus-Aggressive and Robust-minus-Weak factors, while the S&P 500 is initially more exposed to the Market-minus-Risk Free Rate and Small-minus-Big factors. More recently however, the Fama-French factors are less evident in S&P 500 returns.
We also see this in the R-Squared statistic of the Fama-French 5-Factor model’s explanatory power on the State Portfolio and S&P 500. In this case, the Fama-French model nearly fully explains the returns of the State Portfolio while only partially explaining the S&P 500’s returns. In full disclosure, neither the exposure nor the explanatory power statistics means one portfolio is necessarily better than the other, but it does make it a bit easier to understand what is driving returns. Once we know what drives returns, we can then decide whether we’re comfortable with that, or not. If we don’t know, it’s harder to get comfortable with the portfolio because it’s harder to know what makes the portfolio tick.
While the cumulative return chart is impressive, the annualized return of the State Portfolio compared to the S&P 500 is even more so. 4.9% average alpha since 1951, which includes the early periods when it was out-of-the-market, is pretty awesome. By comparison, my previous firm considered +2% alpha the goal, and we often struggled to achieve it.
Looking at annual returns is similarly impressive. We can see the early periods when the State Portfolio was out-of-the-market (mid-50s and 60s), as well as the additional return this portfolio provides above and beyond the S&P 500 is fairly consistent and punctuated by significant out-performance in certain periods. The late 70s, early 80s, early 90s and the peak and crashing of the internet bubble around year 2000 had the State Portfolio significantly leaping ahead relative to the S&P 500. Interestingly though, the State Portfolio followed the S&P 500 down during the Global Financial Crisis of 2008, only marginally outperforming in such a bad year.
This pattern of out-performance is repeated when returns are charted as a scatter plot. The State Portfolio has a slope of 0.82 relative to the S&P 500, while adding an average 0.56% per month in higher returns. As an aside, lower beta and higher alpha is a portfolio manager’s dream!
Return-per-unit of risk is also important when considering different portfolios. If you are forced to take more risk to achieve higher return, your portfolio may not be superior, and it may not be comfortable for all investors. The State Portfolio however, achieves higher return per unit of risk then the S&P 500 across a number of widely-used risk metrics.
Better risk versus return statistics can be viewed annually as well. Comparing the State Portfolio to the S&P 500 annual returns and you quickly see the different degrees of degradation of returns as volatility increases. For every additional unit of risk taken, the S&P 500 loses 1.5% in return whereas the State Portfolio loses 0.37% per year.
On the scale of difficulty, successfully timing of a single market is right up there with walking on water and turning water into wine. There may be one who can do it, but he hasn’t been around for quite a while. There is a way however. One which does not involve the tried-and-proven-to-be-average asset allocation method, one that does not involve guessing or reading the market tea-leaves. The solution to timing one market by using one or more signals is to focus on many signals and the best accompanying assets for that signal. Moving from asset to asset based on a host of signals lowers the reliance on the signal being wrong, or right at the wrong time, while also heightens the probability of success through diversification of signals and assets combined with a rotation of assets given the state of the economy and market.
Timing the market, in- or out-, should not be the goal. The goal should be to find the best assets given the State of the economy and market and invest in those. Implicitly, this is a timing model, but a more successful one.