Thanks to a great question from one of my premium subscribers, we are embarking on a 5-week journey through the research behind market timing. Last week, we talked about the evidence against market timing. This week, we examine the research supporting market timing and, specifically, which type of market timing studies support.

As we will explore below, there is plenty of evidence that market timing methods can and do work. Yet, there is also substantial evidence that timing methods have not worked for investors in the past. So, what is it?

Subjective vs. Quantitative Timing Methods

The missing link between both arguments is in the application of market timing. Most efforts at market timing are subjective. Meaning, they are based on opinions rather than facts; gut feels rather than a testable strategy; and, often, emotional rather than rational. Based on the evidence we observed last week, timing the market with subjective reasoning does not work.

There can be no subjective market timing system that works consistently and reliably through time. To the extent such a system does exist, it is impossible to know who possesses this information or who can be relied upon ahead of time. Therefore, the only timing system you should consider using is one that is based solely on quantitative data with a long history of success throughout all kinds of markets. Fortunately, there is evidence that such a system can offer benefits to investors. More specifically, it can benefit investors who care more about protecting the downside than maximizing their upside.

Before we get into specific timing strategies, we must understand why a quantitative model would be superior to the subjective analysis made by us or "experts" we watch on TV.

Simple Models > Human Beings at Making Complex Decisions

Why is a quantitative system superior to a subjective one? In all types of fields, quantitative models — even simple ones — have been proven to make better decisions than human beings. A typical case is written about in Michael Lewis' book "The Undoing Project" where he recalls a study of radiologists done at the Oregon Research Institute. The results of this study were — in the words of the researcher — "generally terrifying."

The researchers were attempting to create an algorithm that would mimic the decision making of doctors. They started with a simple algorithm based on seven factors the doctors had mentioned were signs that a tumor was malignant or benign. The researchers assumed this model would be too simple to be effective. After all, they hadn't applied advanced mathematics; there was no weighting assigned to factors that might be more important than others; there were no prior models used as a starting point.

So they started with 96 different stomach ulcer patients. The doctors were then asked to rank the stomach ulcers on a 7-point scale from "definitely malignant" to "definitely benign." The simple 7-factor algorithm also ranked each case based on quantifiable observations. Their findings were shocking. As Lewis puts it:

If you wanted to know whether you had cancer or not, you were better off using the algorithm that the researchers had created than you were asking the radiologist to study the X-ray. The simple algorithm had outperformed not merely the group of doctors; it had outperformed even the single best doctor. —Michael Lewis [emphasis mine]

This is just one of many studies demonstrating how much better a simple algorithm is at making decisions than human beings. They range from psychiatrists to wine tasting to credit risk to the judicial system. In all walks of life, complex decisions are better suited to the emotionless and consistent application of a model, not the nuance (and emotions) of a human.

Of course, just using a quantitative system doesn't guarantee a good outcome. The system you adopt and the underlying data that it relies on is of utmost importance. Investing based on last quarter's GDP growth, for example, would not improve your investment outcome. Neither will investing based on whether Democrats or Republicans are in control.

Quantitative systems may not provide an effective timing decision, but they do have the distinct advantage of being insulated from the #1 reason people fail at investing: decisions based on emotions rather than logic. If we use a strategy that's steeped in evidence, stick to that strategy through thick-and-thin, and ignore our own emotions (which always steer us wrong), then market timing does have some reasonable basis. Let's take a look.

Quantitative Timing: Momentum or Valuation?

If we are to use a quantitative system, we now need to determine what kind of inputs we will use. There are thousands of data points we could use, but the two most steeped in research and economic reality are valuation and momentum.

Valuations simply refer to the aggregate value of the market at any given time. The most common way to measure value is with the price-to-earnings (P/E ratio). You will commonly read that the market is "overvalued" because the P/E ratio is currently 18 whereas it has historically averaged 16. We could make our timing decisions based on this or something like it. When stocks are "too expensive" we would sell stocks and buy bonds or vise versa.

The other method utilizes momentum or trend-following. This relies on the near-term movement in stock prices to predict future stock prices. Timing systems based on this assume that stocks have physical properties akin to physics. An object in motion tends to stay in motion unless acted upon by an outside force. Investments generally do the same. A stock or stock market moving higher generally keeps going higher and vise versa.

Which should we use for our quantitative timing model: valuations or momentum?

Market Timing with Valuations

Methods for timing the market based on valuations have shown promising in-sample results, but fail when considering out-of-sample time periods. In other words, valuation techniques work to time the market, but only when you have access to data that you would not have in real-time.

Asness, Ilmanen, and Maloney (2017) studied timing models based on the CAPE valuation from 1900-2015. They comment on the disappointing results of valuation-based timing methods:

It does not prove that contrarian timing strategies won’t work in the future. But it does illustrate a fundamental difficulty faced by such strategies: valuations can drift higher or lower for years or decades, making it difficult to categorize the current market confidently as “cheap” or “expensive” without hindsight calibration, and therefore difficult to profit from such categorizations.

Not only would a value-based investor not had the benefit of data in future years, but they wouldn't have the benefit of valuation metrics. The dividend yield was the most common valuation signal in past decades. The CAPE ratio was only popularized in recent decades by Robert Shiller. An investor in 1930 would have used a different metric than CAPE, which further adds to the evidence against using valuation-based timing methods.

Timing with Momentum

In the same paper, Asness et al found that adding a simple momentum timing strategy to the valuation strategy improved the value-only timing model by 0.4% annually with 2% less annual volatility. Not only that, but the momentum model worked so well that it outperformed even the model using both value and momentum. And not only did the trend-following strategy outperform all other strategies, but it also did so with less volatility. This result was consistent across the entire sample period (1900-2015) and in more recent times (1958-2015).

Which Momentum Should We Use?

If momentum shows more promising and consistent results across time, then what specifically should be used as the quantitative signal? There are all kinds of ways to measure momentum or trends. Let's explore just a few terms here so it's clear what we're talking about in the next section when we examine specific timing systems that have been studied. If you are familiar with these terms, proceed to the next section.

• What do we mean by a "moving average"? A moving average is simply an average of a past number of stock market prices. For example, the 200-day moving average is simply the average price over the last 200 trading days. A moving average could also be calculated using end-of-month or beginning-of-month prices. For example, a 12-month moving average would be defined as the average over the course of the last 12 months using the same day each month (usually the last trading day).

• What specifically do we mean by "momentum"? Momentum in this context is simply the past total return (price appreciation plus dividends) defined over some time period. Most of the research on momentum concludes that the past 3 to 12 months is the most useful for predicting future returns.

A Sampling of Effective Strategies, Part I: Moving Averages

And now we get into the actual systems and their results.

Faber (2007) introduced a simple timing strategy that went long the S&P 500 Index when the monthly price was greater than the 10-month simple moving average. He updated the model only once per month on the last trading day. Taxes, commissions, and trading slippage were excluded from his model. His signals were as follows:

• When the price was above the 10-day moving average, the strategy was 100% stocks.

• When below the moving average, he moved to cash. Returns for cash assumed the 90-day Treasury bills.

The results of his study were quite impressive. The timing solution improved compounded returns and reduced risk while being invested in stocks 70% of the time. Turnover was also quite low with less than one round-trip trade per year — equivalent to less than 100% turnover or roughly in-line with the average mutual fund.

A non-discretionary, trend following model acts as a risk-reduction technique with no adverse impact on return. Utilizing a monthly system since 1973, an investor would have been able to increase risk adjusted returns by diversifying portfolio assets and employing a market-timing solution. In addition, the investor would have also been able to sidestep many of the protracted bear markets in various asset classes. Avoiding these massive losses would have resulted in equity-like returns with bond-like volatility and drawdown. — Meb Faber

A Sampling of Effective Strategies, Part II: Time Series Momentum

Gary Antonacci (2012) introduced the concept of combining absolute and relative momentum in his excellent book, "Dual Momentum: An Innovative Strategy for Higher Returns with Lower Risk." He used a simple 12-month absolute momentum signal to determine whether or not to invest in risky assets (either US stocks or international stocks) or Treasury bills.

• When stocks produced more than Treasury bills over a 12-month period, he invested 100% in stocks. Which stocks he chose (international or US) depended on which of those had the highest performance.

• When neither asset had a return higher than Treasury bills, he invested in 100% bills.

His findings were impressive and, like Faber, demonstrated that a trend-following strategy can be effectively used to increase returns with lower risk. Other studies have contradicted the "higher returns" portion, but we will discuss that in a minute.

A Sampling of Effective Strategies, Part III: A Combined Approach

Gray (2015) combined Antonacci's strategy of using time series momentum with Faber's moving average strategy. Here was how he constructed his study:

• 50% of the portfolio = When stocks produced more than Treasury bills over a 12-month period, he invested that portion of the portfolio in stocks. If not, this portion is in Treasury bills.

• 50% of the portfolio = When the monthly price exceeds the 12-month moving average, long stocks. Otherwise, long Treasury bills.

In effect, this diversifies the risk of one particular method failing. It also reduces the number of head-fakes (i.e.: quickly jumping in and out of stocks) and eliminates extreme movements from 100% to 0%. The portfolio can move between 0% to 50% to 100% depending on the combination of signals.

His strategy provided substantial downside protection in bad markets (equity drawdowns exceeding 15%). In the 29 drawdowns of 15%+ from 1801 to September 2015, his moving average strategy outperformed in 26 of 29 with two ties and underperformance only once (1860-1861).

If market timing strategies are more for protecting against the downside, perhaps the better comparison is not a 100% stock strategy but, rather, a 60% stock/40% bond strategy? According to Gray (2015), the timing strategy outperformed the 60/40 strategy in 18 out of 27 down markets. The worst showing of the timing strategy relative to a 60/40 was 1939-42 when the timing strategy underperformed by nearly 14%. Even in that scenario, the strategy would have outperformed the S&P 500 by just under 3%.

The Key to Timing Effectiveness

Faber (2007) admitted that his timing strategy only outperformed the market in about half of the years, which is not a particularly great outcome. So why was the strategy so effective? It was in the magnitude of the reduction in maximum drawdown. The correlation between the S&P 500 and his timing model was 0.83 in positive years; in other words, when stocks went up 10% the timing strategy would generally go up 8.3%. In down years, the correlation flipped to -0.38; in other words, when stocks decreased by 10% the timing strategy generally increased by 3.8%. Therefore, the magnitude of the outperformance in down markets was so substantial that it offset the drag on the upside.

Asness (2017) found that timing strategies based on momentum had lower maximum drawdowns (-39%) compared to buy-and-hold (-53%) and valuation-based timing strategies (-51%) from 1958-2015. He later concludes that timing strategies may not be effective for improving returns even before costs are considered, but does endorse trend-following timing for downside protection purposes.

The combined value and momentum timing strategy beats both pure value and buy and hold on all measures, in both samples. If market timing is [wrong], perhaps you could—or even should—[market time] a little. —Cliff Asness

Volatility & Returns are Skewed

By only owning stocks above the moving average, you are in effect making the assumption that stocks typically do better above their moving average than below it. Research bears this out. Faber (2007) found that volatility was 60% higher with returns 30% lower for five asset classes (stocks, bonds, commodities)

Timing May Add Value Even in a Sideways Market

If the market were to swing back and forth in a long sideways direction, then the moving average strategy should underperform without the risk reduction benefits. However, Chaves (2012) finds that trend-following still would have added significant value in Japan — a market that has gone sideways for 20 years.

Does Momentum-Based Timing Improve Return?

Based on the evidence, it is hard to deny that mechanical, trend-following techniques of several varieties would not have provided downside protection. What is far less clear, however, is whether timing systems can be considered ways to improve total returns compared to a buy-and-hold approach.

In some cases, research suggests that timing with momentum does not lead to outperformance. In his 2008 book "Stocks for the Long Run: The Definitive Guide to Financial Market Returns & Long-Term Investment Strategies," Jeremy Siegel investigates using the 200-day moving average to time the Dow Jones Industrial Average (DJIA) from 1886 to 2006. When the DJIA price closed 1% above the 200-day moving average, he bought stocks. When it closed 1% below, he invested in Treasury bills. He concluded that market timing improved both risk-adjusted and absolute total returns. However, when you consider the costs (taxes, bid-ask spreads, commissions), the timing model does not improve absolute returns; it only improves risk-adjusted returns.

In the same book, Siegel studies the impact that a timing system would have on the Nasdaq Composite Index. He found that timing did improved absolute returns on the NASDAQ Composite Index since 1972 even after considering costs.

It Works, but Just Two Caveats

Based on the evidence, momentum-based timing models do work. However, there are several caveats that must be considered before implementing them in any investment strategy:

1. The Goal is Risk-Reduction, not Performance Enhancement

The evidence is too inconclusive to say definitively that timing improves future returns. Timing may improve absolute returns in the future, but the evidence is too weak for me to use it for that purpose. If the future contains more events like 2008-09, then timing would lead to larger future returns. However, if the future contains more sharp market drawdowns and recoveries (i.e.: the COVID crisis in 2020, Black Monday & the 1987 crash) then it will not.

Since neither you nor I (nor anyone else) knows the future, we can only study the past. And the past tells us that timing systems are not reliable ways to improve returns after costs are considered. If my goal is to achieve the highest possible rate of return and I care nothing for downside risk, then I would not personally use market timing to attempt to improve my returns. It is probably not going to help with that goal. In fact, it may result in lower returns, particularly if I am investing in a taxable account where taxes will reduce the compounded returns of my investments.

If, on the other hand, my goal was to achieve most (not all) of the upside potential of stocks and reduce downside risk, I would consider timing to be an effective way to do that. If I were, I would consider comparing my performance to a portfolio that included a permanent allocation to bonds. It would be unfair to assume that a strategy that includes bonds (like the timing one does) to keep up with an all-stock portfolio. Since stocks outperform bonds in almost all long-term scenarios, that portfolio is going to do worse. In several studies, the timing strategy spent roughly 70% of the time invested in stocks, and 30% invested in bonds. So I would probably think about my timing strategy as compared to a 70% stock / 30% bond account.

2. Quantitative, not Systematic

If you adopt momentum-based timing into your investment system, it is absolutely imperative that you have a plan for doing so and that you follow that plan through thick-and-thin. There will be moments when you do not "feel" like you should be in stocks. There are moments when you do not "feel" like you should be selling stocks. The system works because the system takes your feelings out of the equation. The moment you reintroduce them, you jeopardize the success of the entire program.

The moving average strategy works not because it is a magical strategy, but because it is a systematic one. It is not the brilliance of the strategy; it is the brilliance of its execution. It allows you - the emotional human - to make decisions based on a simple "yes or no" answer. Is the price greater than the moving average or not? If the answer is "no", then you sell. If the answer is "yes," then you buy (or hold.

Next Week's Article

The Tactical Allocator: A Systematic Strategy Using ETFs [premium subscribers only]

My original plan was to do a 5-part series, but I think my Part III and Part IV plans are now redundant with what I've already mentioned in the last two articles. So, I'm going to move straight into the practical application of all this research.

Next week, I'll introduce an experimental portfolio that I've adopted. This system hopes to improve upon existing systems by mitigating the application risk of momentum-based timing systems. That article will be available to premium subscribers only.

References

• Asness, Ilmanen, and Maloney (2017). "Market Timing: [...] Resolving the Valuation Timing Puzzle." Journal of Investment Management, Vol. 15, No. 3, (2017), pp. 23-40.

• Gray, Wesley (2015). "The World's Longest Trend-Following Backtest." Alpha Architect. Accessed on August 12, 2020 from: https://alphaarchitect.com/2015/11/09/the-worlds-longest-trend-following-backtest/.

• Antonacci, Gary (2012). "Dual Momentum Investing: An Innovative Strategy for Higher Returns with Lower Risk."