[Special thanks to longtime reader, Matthew Morrison for reviewing this post and providing some key input.]

Practically since stock markets were invented, people have looked for ways to avoid the inevitable periodic dips in stock prices.  Such attempts combine the unmindful concept of trying to beat the market, which I posted about last time, with the equally unmindful concept of trying to time the market.  For reasons I’ve noted before, a mindful approach to investing suggests that successful short-term market timing is impossible with any consistency.

Nonetheless, people have created myriad systems that attempt to successfully time the market.  The idea of day trading alone covers a vast array of potential market timing systems, signals, and economic indicators for individual stocks, bonds, funds, and just about every possible asset.  So, unlike my last post on beating the market, I’m not going to attempt to review multiple market timing systems.

Instead, I’m going to focus on one category of market timing that goes by the snazzy name of “Tactical Asset Allocation”(TAA).  Essentially, TAA uses valuation, economic, momentum, sentiment, and/or other indicators to shift portfolio allocations among two or more asset classes over time.  The goal is to favor assets that are expected to perform well in the future and avoid assets that are expected to perform poorly.  TAA shifts can involve anything from slight refinements in allocation percentages to all-or-nothing bets on particular assets.

Given the poor track record of trading and market timing, you might be surprised that I’m even addressing TAA here at Mindfully Investing.  But after doing a bit of research, I found some pretty reasonable arguments for (and against) TAA that amount to more than mere conjecture and self-interested cheerleading.

The Dream

The idea of TAA grew out of basic market-timing research dating back to at least the 1970s.  At first, the premise was simple; find the best times to buy and sell stocks.  But as finance researchers, including luminaries like Robert Shiller and Kenneth French, found ever more ways to potentially predict future asset returns, it caused a seemingly inevitable progression toward the more nuanced idea of TAA.  By the 1990s, tens of billions of dollars were already invested in TAA mutual funds of one type or another.  And yet, even today, the debate still rages on the best timing indicators, the best assets to include, and even whether any of the many flavors of TAA work.

TAA has been tried with many different potential indicators, timing systems, and combinations of assets.  So, I thought it makes sense to start with a select example that became popular in the last cycle of peak TAA interest, which perhaps predictably, occurred in the years around the 2008 financial crisis.

Moving Averages – Meb Faber published a paper in 2007 and updated it in 2013 that reviewed a simple TAA system using the 200-day moving average (MA) of stock prices.  Faber’s paper generated so much interest that at one point it was the most downloaded publication on the SSRN website.  The next year, Jeremy Siegle issued the fifth updated edition of his best-selling book Stocks for The Long Run, which added an analysis of a very similar 200-day MA system.  Faber stresses that he didn’t invent the 200-day MA idea.  It’s been around for a long time, which makes me wonder why his paper caused such a fuss.

Faber’s version of the 200-day MA system shifts to all-cash investments (3-month T-bills) when the monthly stock price falls below the 200-day MA and back to stocks when prices rise above the same MA.  Here’s a graph from Faber’s paper showing the growth of a $100 initial investment in the S&P 500 using the 200-day MA system (labeled “timing”) as compared to buying and holding the same basket of stocks.

The timing system outperformed buying-and-holding the S&P 500 in terms of both annualized absolute and risk-adjusted returns over this period.

However, these century-long growth charts can be somewhat misleading, because once one system grows a lead, it’s hard for the other system to catch up again.  So, here’s the same comparison from Faber but for the period 1990 to 2012.

You can see more clearly at this scale how the growth of the timing system flattens out when it shifts to cash.  You can also see that someone using this timing system would have appeared overly cautious from 1990 to 2000 but brilliant from 2008 to 2012.  Again, over this period, the 200-day MA timing system performed better than buy-and-hold both in terms of absolute and risk-adjusted returns.

Faber also presented his own Global Tactical Asset Allocation (GTAA) system that includes five global assets instead of just two U.S. assets, as well as a more complex system for allocating across the menu of assets.  I refer you to Faber’s paper if you want more details.  GTAA showed positive results but not quite as good as the simple 200-day MA system, as shown in this graph starting in 1973.

Other Promising Examples – In 2017, Corey Hoffestein wrote a great summary of TAA showing that several other flavors of TAA have historically performed quite well using the data from those very same luminaries, Shiller and French.  Here are graphs from his summary showing the relative performance of three TAA systems as compared to buying and holding U.S. stocks.  Again, if you want more details on these systems, I refer you to Hoffestein’s summary.

Positive values on the vertical axis indicate that an investment in the TAA system would have outperformed its buy-and-hold counterpart and negative values indicate the reverse.  As you can see, all three TAA systems outperformed buy-and-hold and stayed ahead for the entire period of the analysis.

The Reality

This all looks very compelling.  But before you invest in the dream, let’s take a more wide-awake look at these same graphs.  Here’s the Faber “timing” graph again, but I’ve highlighted in yellow the periods where the timing system underperformed buy-and-hold.

A closer examination shows that someone who started investing in the timing system in 1990 would have achieved a lower return than buy-and-hold had they ended their investment at any time except 2002 to 2003 and 2008 to 2012.  These underperformance windows equate to 74% of the period.

Here’s a similar take on the Hoffstein graphs, but in these graphs periods of TAA underperformance are indicated when the blue line is descending.  So, I roughly highlighted those descending legs as best I could given the resolution of the published graphs.

Within each yellow-highlighted timespan in these relative performance graphs, an investor would have endured their TAA system generating lower returns than buy-and-hold.  I didn’t attempt to quantify the underperforming versus overperforming periods, but it appears that each of these three TAA systems underperformed roughly half the time.

A Reality Check

As a cross-check, I was curious how the Faber and Hoffstein results might compare with similar timing systems that can be evaluated using Portfolio Visualizer.  I ran simulations of relatively simple versions of the following “timing models” available in Portfolio Visualizer over various periods¹:

• 200-day MA, stocks/cash, from 1985, 1993, and 2012 to present (essentially the same as the Faber system).
• Shiller PE Valuation, stocks/bonds, from 1985, 1993, and 2011 to present
• Relative Strength, stocks/bonds, from 1985 to present
• Dual Momentum, stocks/bonds, from 1985 to present
• 200-day Portfolio MA System, 70% stocks/30% bonds, from 1985 to present.

I realize some of these bullet points are a bit cryptic.  For those who want more details, you can look at the FAQ answers for “Methods” in Portfolio Visualizer and play with the simulator a bit, which will make these bullets more clear.

This graph compares the total annualized return (Compound Annual Growth Rate; CAGR) of each timing system (TAA Model) for each period evaluated to a similar benchmark return.  For TAA Models that shift between 100% stocks and 100% of some other asset, the benchmark is buy-and-hold the S&P 500.  For TAA Models that shift relative allocations between stocks and bonds, the benchmark is buy-and-hold a 60% stock (S&P 500)/40% bond (aggregate bond index) portfolio.

Blue dots below the diagonal blue line indicate that the TAA Model underperformed the benchmark for that period.  Orange dots above the diagonal blue line indicate that the TAA Model outperformed the benchmark.

Only the 200-day MA model starting in 1993 outperformed its benchmark (orange dot), and in that case only slightly.  On the other hand, all these models avoided horrendous underperformance, except for the 200-day MA model for the period starting 2012 (as labeled in the graph).

And here’s a similar graph that plots the risk-adjusted returns (return/standard deviation ratio) for the same set of scenarios.

In risk-adjusted terms, all but one of the TAA Models performed better than their benchmarks.  This makes intuitive sense because all the TAA Models tend to decrease allocations of more volatile assets when market turmoil or crashes occur.  I’ll consider in the conclusions section whether this better risk-adjusted-performance constitutes success for TAA.

The Nightmare

I pretty easily identified a menu of relatively simple TAA systems that have historically performed decently, all be it with long periods of underperformance compared to a benchmark.  So, you’d reasonably think it would be relatively easy to identify some mutual funds and exchange-traded funds (ETFs) that have a successful track record of using various forms of TAA.

But you’d be wrong because at least three studies have shown that most TAA funds routinely fail to beat buy-and-hold or a relevant benchmark.

In 2021, Morningstar compared a basket of TAA funds to baskets of funds that hold 50% to 70% stocks with the rest in bonds including a smaller basket containing so-called “balanced” funds.  This graph shows that TAA funds collectively underperformed more traditional stock/bond funds over every period assessed.

Another 2021 study by Joseph McCarthy and Edward Tower compared baskets of TAA funds to baskets of index ETFs having the same investment style as well as established benchmarks.  They found:

• Baskets of TAA funds underperformed corresponding baskets of index ETFs by 1.77% to 5.15% per year.
• And TAA funds underperformed comparable benchmarks from 1.92% to 5.08% per year.

And an older 2013 Morningstar study found:

• Over the 18 months ending mid-2013, only 28 of the 142 funds (20 percent) that Morningstar tracked produced a higher total return than the Vanguard Balanced Index Fund.
• Over the prior 36 months ending mid-2013, just four TAA funds produced higher Sharpe ratios (a measure of risk-adjusted return) than the Vanguard Balanced Index Fund.

Most of these studies compared averages from baskets of TAA funds to benchmarks or covered periods of three years or less.  So, logic suggests that some subset of TAA funds performed above average and beat their benchmarks, or would have beat over a longer assessment period.

So, I decided to conduct my own informal assessment based on lists of the “best” TAA mutual funds and “best” TAA ETFs at U.S. News and World Report.  I figured that any list of “best” TAA funds would likely contain at least a few outstanding performers.

Using the funds in these lists, here are the same two cross-plot graphs as before (total return and risk-adjusted return) comparing the performance of TAA funds that have existed for five years or more to a 60% stock/40% bond buy-and-hold portfolio².

Some of the funds came close to beating the 60/40 benchmark on a total return basis.  But unlike the results from the Portfolio Visualizer TAA Models, none of these funds produced better risk-adjusted returns than the benchmark.  This is probably because a 60/40 benchmark already exhibits pretty high risk-adjusted returns as compared to a more volatile 100% stock benchmark.  Also, all but two of these funds have existed for less than 10 years.  So, I labeled the two with relatively long track records that both started in 2003.

And if all that doesn’t make you question the utility of TAA, I looked at one additional fund developed by Meb Faber’s company based on his GTAA system from his 2013 paper, which goes by the ticker GAA.  Here’s a graph of GAA’s performance since its 2015 inception (blue line) as compared to a 100% buy-and-hold S&P500 portfolio (red line) and a buy-and-hold 60% stock/40% aggregate bond portfolio (yellow line).

Annualized return (CAGR) and risk-adjusted return (return/standard deviation ratio) stats in this period for these same three options are shown in this table.

Scenario	Annualized Return (CAGR)	Risk-Adjusted Return (ret./st.dev.)
GAA	5.99%	0.64%
BuyHold S&P 500	13.85%	0.95%
BuyHold 60/40	10.16%	1.03%

In short, despite all the promise shown by the GTAA system in 2007 and again in 2013, the actual implementation as a fund (GAA) since then has been problematic.  In 2018, Faber published another update to his original paper.  The abstract says that since first publication, “the model has performed well in real-time, achieving equity-like returns with bond-like volatility and drawdowns”.  I could not find a free version of the paper anywhere, so I can’t tell you how exactly Faber resolves the poor performance of the GAA fund with the “performed well” conclusion in his most recent paper.

Conclusions

I see two themes emerging from this information.

First, TAA funds appear to perform worse than simple TAA systems that are applied directly.  I can’t say for sure why funds perform worse, but cost drag is one hypothesis.  Almost all of the TAA system results I reviewed here don’t include implementation costs³.  In contrast, all of the TAA fund results include costs, and most of these funds have relatively high costs, often above 1% annual.  Another hypothesis is that once embarked upon, TAA systems eliminate personal judgment from the equation; they are said to be “systematic”, not “discretionary”.  But many TAA funds likely include “discretionary” elements, where the fund managers make periodic decisions that depart from a purely systematic approach.  Paying more for worse performance is a well-known problem with more discretionary or “active” funds.

Second, TAA systems (not funds) appear to often perform well in times of stock market turmoil and relatively poorly in less turbulent times.  That fact is clear from this Portfolio Visualizer graph comparing rolling 5-year annualized returns of the 200-day MA system (blue line) to S&P 500 buy-and-hold (red line).

This TAA system worked great from 2002 to 2013 but substantially underperformed during the roaring 1990s and the last 7 to 8 years of unusually relentless positive stock market performance.  And this evidence fits well with the periodic underperformance discussed in both the Faber and Hoffestein write-ups.

Given that TAA systems can underperform buy-and-hold for years or even decades, many of the TAA proponents take refuge in “risk-adjusted” performance.  However, as shown by many of the comparisons to 60/40 benchmarks and “balanced” funds presented above, it seems like an individual investor can achieve most (if not all) of the same risk-adjusted benefits through simply buying-and-holding a balanced index fund.  And given that mindful investors don’t care much about routine volatility, benefits in terms of risk-adjusted returns feel mostly like a sideshow to me.

Despite the periodic outperformance for some TAA systems, I’m still rationally skeptical about TAA.  While multiple methods seem to work well for specific back-test periods, backtesting can only encompass a world of possibilities that have already occurred.  The unknown and unforeseen are still a mystery for any of these systems.

I also wonder if these TAA systems work so well, then why wouldn’t more of these underperforming TAA funds jettison discretion and adopt more systematic TAA?  And since some “successful” systematic TAA approaches have been around for decades (like 200-day MA), why isn’t there less debate by now?  Further, if these simple approaches work so well, why are there more than 60 well-known TAA systems that are regularly tracked?  It seems like a few successful time-the-market schemes would be sufficient for most investors.

In the end, TAA appears similar to other beat-the-market ideas I’ve reviewed over the years and even the relatively accepted idea of investing factors.  Sometimes these ideas work, and sometimes they don’t work for years or decades.  And the problem is you never know which outcome you’ll get when you finally decide to try TAA.

---

1 – Several other timing systems can be tested in Portfolio Visualizer.  And I certainly could have conducted many more variations of these systems, such as adding more assets.  But my goal was to generate data on a subjectively random subset of relatively simple TAA systems.

2 – These funds generally seem to be marketing to investors that are averse to holding volatile portfolios, which explains ticker names like “RELAX” and “DUDE”.  So, it seemed more appropriate to compare them to a relatively sedate 60/40 benchmark rather than an aggressive 100% stock buy-and-hold portfolio.  Also, some of the funds in these two lists were only a year or two old, which makes me wonder what criteria the folks at U.S. News used to evaluate such short-lived funds.

3 – Assuming zero costs seemed reasonable because I can’t imagine implementing a timing system in a taxable account due to the tax drag created by sell events.  Using a tax-advantaged account with no trading fees would virtually eliminate most TAA system costs and is an entirely feasible assumption.