by Corey Hoffstein, Newfound Research
Summary
- After 2008, tactical ETF strategies rose in popularity.
- We often come across advisors who self-implement their own tactical strategies, using simple measures of momentum and trend.
- We believe that thoughtful implementation of tactical strategies requires admission that tactical choices will be wrong from time-to-time, and therefore advocate for the prudent use of diversification to manage uncertainty risk.
- The greatest risk we see in self-implemented strategies is often a failure of adequate diversification.
- We outline three ways in which we see this manifest: a failure to diversify what, how, and when.
Post-2008, we witnessed a significant rise in the popularity of tactical ETF strategies and the number of strategists offering them. Our ex-post narrative is that while the rise in exchange-traded funds provided investors with a way to implement low-cost tactical trades, it was the traumatic losses of 2008 that gave investors the motivation to make them. “I do not ever want to go through that again,” was a phrase we heard often repeated.
While the popularity of tactical ETF strategists has waned in recent years, our anecdotal experience is that self-implemented tactical strategies are just as popular as ever among the advisor community.
Painting with broad strokes, there are a few common signal categories that drive most tactical trades: value, momentum, carry, trend, defensive, volatility, economic, and thematic. The easiest of the signals to calculate – requiring only (adjusted) closing prices – are momentum and trend.
Ostensibly, anyone with the ability to enter prices into an Excel document and calculate an average can create a trend-based, tactical ETF model.
Unfortunately, the ease of implementation belies the ease of implementation. Just about any naïve implementation of momentum and trend will have looked good during a backtest of the 2000s. The Devil, however, forever runs amok and wreaks havoc in the details.
A Failure to Diversify
We’re going to offer you the option to play one of two games.
In the first game, we’re going to flip 100 fair coins all at once. For each coin that comes up heads, we will give you $1. For each coin that comes up tails, you will pay us $1.
In the second game, we are again going to flip 100 fair coins, but this time we will do it in sequence. Starting with $1, for every heads that comes up in a row, we will double the pot (e.g. one heads is $2, two heads is $4, three heads is $8, et cetera). If tails ever comes up, the game is over and you owe us $1. If you end up with 100 heads in a row, however, the entire pot is yours.
Our guess is that you would never play the second game. Yet – ignoring the default risk of us actually being able to pay you such a gargantuan sum of money – the second game has a higher expected value than the first.[1]
So why would you not elect to play? Whether you know the math or not, you probably intuitively understand that expected value only really matters if you can play the game an infinite number of times. Indeed, with just a single chance to play, a gambler’s wealth should approach the median, not the mean of forecasted growth.[2] In the second game, the median is quite clearly losing $1.
This math does not make tactical investing any easier. The “game” may not be over with a single wrong tactical call, but bad calls compound upon themselves.
Having the fortitude to stick with your process is important, but trying to diversify your risk over time simply does not work the same way as diversifying your risk across assets.
Put another way, the effects of compounding can mean that experiencing whipsaw with 100% of your wealth with a 10% probability can potentially be more harmful than experiencing whipsaw with 10% of your wealth with 100% certainty.
This risk rarely manifests itself in a backtest. But backtests have little-to-no meaning when they were designed with the benefit of hindsight. We know momentum and trend worked in the 2000s: just about any simple moving average on the S&P 500 driving exposure to levered long and short ETFs is going to look outrageously good.
Unless you live and breathe these choices, it can difficult to understand why you might want to embrace a backtest that demonstrates worse results. Yet worse can be better when worse is actually achievable going forward.
The driving wedge is uncertainty. The past is written in stone while the future is clouded. Looking backward, diversification is sub-optimal for maximizing wealth. Looking forward, however, diversification is necessary for avoiding catastrophe.
In our experience, it is a lack of adequate diversification that we believe is the biggest risk to most do-it-yourself tactical strategies.
A Failure to Diversify What Bets We Take
We know that when diversification is abundant, the accuracy rate required for tactical strategies to add value goes up significantly.[3] This is one reason why tactical strategies can be so much more successful in environments like 2008, when asset correlations crash to 1, but less so in environments like 2015. Even if the accuracy rate of the tactical strategy does not change, the required accuracy rate is a moving target.
To be clear, a well-diversified tactical strategy is not one that simply has a large pool of assets to select from. Having 20 assets to choose from, but allocating fully to a mix of 5 equity indices leaves you in a rather concentrated position. One, which if wrong, can cause significant negative tracking error.
That is not to say that highly concentrated, tactical strategies do not have a place. After all, we do offer a number of them here at Newfound. Yet advisors are sometimes surprised when we argue that these portfolios should play a smaller, satellite role in an investor’s allocation profile. It is not a lack of conviction in our approach that leads us to this conclusion: just the mathematics of diversification.
A Failure to Diversify Why We Take Our Bets
The topic of diversification is not limited only to asset classes. Diversification in process can be just as important.[4]
For example, empirical evidence suggests that 6-to-12 month measures of momentum all work over the long run. Precisely which measure you implement, however, can have a profound impact on your results in the short run.
Furthermore, the choice of which style of momentum to implement can cause a significant difference in results. Full period total return? Skip the last month? Risk-adjusted momentum? All approaches that have demonstrated success in the long-run, but significant variability in the short run.
With only two axes of choice, we have already introduced 18 possible implementations (3 versions of momentum for lengths of 6 months through 12 months). As we add further dimensions of parameters, this figure compounds.
We often see tactical implementations that try to address this problem by blending signals. For example, a portfolio might select the top 5 assets that have the highest average 3-, 6-, and 12- month total return. The problem with this approach is that it does not diversifying process risk: it simply defines a new, single process. Diversifying process risk would require creating portfolios for the 3-, 6-, and 12-month total return numbers and then allocating across those different portfolios.
This is a key distinction, especially when portfolio construction is a non-linear function of the inputs (e.g. traditional mean-variance optimization).
Unfortunately, performing process diversification correctly often requires computing dozens of potential portfolios – something that cannot be done simply in an Excel spreadsheet.
A Failure to Diversify When We Take Our Bets
Most do-it-yourself tactical strategies we come across are implemented on an end-of-month basis, with portfolios being reformed and held for the next month. The risk – as we have demonstrated many times in the past[5] – is one of timing luck.
It is obvious that a strategy’s allocations are a function of the underlying signals at the time the portfolio is rebalanced. If those can meaningfully change in a short period of time – as is the case with momentum and trend – then the portfolio selected at the end of the month may be meaningfully different than the one selected just two weeks prior.
This is not a problem if we believe our signals are always accurate. However, if we happen to rebalance when our signals are wrong, then timing luck can lead to significant whipsaw.
We’ve demonstrated in past commentaries that the solution to timing luck is simple: tactical strategies should be implemented using overlapping portfolios (“tranching”). Actually implementing this change, however, requires tracking a number of sub-portfolios and implementing smaller, more frequent changes within the portfolio.
Conclusion
We believe that tactical investing offers a significant diversification benefit to traditional buy-and-hold investing. The whipsaw risk of tactical investing, however, is not to be taken lightly; the compounding of errors can lead to permanent wealth impairment.
Thus, the presumption of eventual error is important in managing risk. While many self-implemented strategies may seek to manage uncertainty risk by diversifying across assets (the “what”), they often fail to adequately diversify across process (the “how”) and rebalance timing (the “when”).
Unfortunately, diversification in these areas is difficult to achieve without tracking a variety of sub-portfolios and implementing portfolio tranching.
That is not to say that strategies that forego these steps are necessarily doomed to failure. Indeed, they may pick the right method for the times or the right rebalance date. But skill should be distinguished from luck, and luck is not something we believe belongs in a prudent portfolio plan.
[1] In the first game, each flip is an even chance at making $1 and losing $1, so your expected value is zero. In the second game, you only win if you get 100 heads in a row, which is the infinitesimally small probability of 0.5100. However, you win the indescribably large sum of $2100. Which simply all nets out to an expected value of $1 for winning. If you lose, however – which happens with the near certain probability of (1 - 0.5100), you lose $1. Thus, the expected value is very, very, very slightly positive.
[2] See the section Multi-Period Investing: Volatility is a Drag in https://blog.thinknewfound.com/2017/07/growth-optimal-portfolios/
[3] See https://blog.thinknewfound.com/2016/10/rising-correlations-tactical-asset-allocation/
[4] See https://blog.thinknewfound.com/2017/02/embracing-conflict-asset-allocation/
[5] See https://blog.thinknewfound.com/2018/01/quantifying-timing-luck/
Corey Hoffstein
Corey is co-founder and Chief Investment Officer of Newfound Research, a quantitative asset manager offering a suite of separately managed accounts and mutual funds. At Newfound, Corey is responsible for portfolio management, investment research, strategy development, and communication of the firm's views to clients.
Prior to offering asset management services, Newfound licensed research from the quantitative investment models developed by Corey. At peak, this research helped steer the tactical allocation decisions for upwards of $10bn.
Corey is a frequent speaker on industry panels and contributes to ETF.com, ETF Trends, and Forbes.com’s Great Speculations blog. He was named a 2014 ETF All Star by ETF.com.
Corey holds a Master of Science in Computational Finance from Carnegie Mellon University and a Bachelor of Science in Computer Science, cum laude, from Cornell University.
