Managing Portfolio Risk With Tactical Asset Allocation

Tactical asset allocation (TAA) is the solution and the problem. The solution because dynamically managing the asset mix offers the potential for superior risk control and perhaps even higher returns relative to a passive strategy. But TAA is also a problem in the sense that no one’s really sure which set of rules for managing a portfolio’s asset allocation in real time will shine in the future. That alone isn’t a reason to shun TAA, although it’s a reminder that the hazards may be higher in this niche compared with a simple rebalancing regimen such as moving allocations back to target weights every Dec. 31, for instance.

TAA opens the door to something better… maybe. The details matter in this space, and there’s no shortage of details. Indeed, TAA models come in a rainbow of possibilities, although the one aspect that unites them is the focus on opportunistically adjusting the asset weights based on perceived changes in expected return and risk. The challenge, of course, is developing a model that’s reasonably accurate. Even for TAA applications that excel, it’s not obvious that the results will exceed a relatively passive asset allocation over the long haul. But perhaps that’s missing the point. The motivation for many TAA applications, advocates say, is managing risk in the short term.

The wake-up call on this topic was the market meltdown in 2008. The quasi-passive strategy of periodic rebalancing and otherwise leaving the portfolio to fluctuate based on Mr. Market’s whim came under attack in the wake of steep drawdowns in the final months of that fateful year. If your pre-set rebalancing date was the end of the year in 2008, for example, you paid a heavy price in the preceding months for sticking to that schedule.

In theory, TAA offers a way to dodge the bullets. In practice, results vary, as they say, and by more than trivial degrees. But as an academic exercise, it’s intriguing to consider the possibilities. Where to begin? One approach is using market volatility as a signal for dynamically adjusting weights.

It’s well established that volatility cycles through time. History also shows that transitions from low vol to high vol are associated with low/negative returns and vice versa. Let’s harness this information as the basis for a dynamic rules-based system for managing asset weights. As vol rises for an asset, the allocation will fall; as vol falls, the weight for the asset will rise. The logic here is that the risk contribution of each asset will vary, perhaps radically so, when asset allocation is left to the market. In order to manage this risk, we’ll limit the portfolio’s exposure to volatility for each asset to a constant 2%. The formula for determining asset weight in our test: the ratio of the risk factor to trailing 100-day volatility (standard deviation) of daily return. For instance, if the trailing volatility is 10%, the resulting weight is 20% (2%/10%). The basic idea is to systematically raise (lower) weights as volatility falls (rises).

As a further level of refinement for risk management, let’s also incorporate a momentum factor into the model for determining if there’s any risk exposure (as determined by the rules above) or not. The basic setup: buy when the trailing return is positive and the 50-day exponential moving average (EMA) is above its 100-day EMA; otherwise, we’re in cash for the asset in question.

To keep things simple, we’ll use only two ETFs to represent US stocks and bonds: the SPDR S&P 500 (SPY) and the iShares Core US Aggregate Bond (AGG), respectively. Given that this is a two-asset portfolio, we’ll impose a constraint of 50% as the maximum allocation for the model. Meantime, weights are allowed to fall to zero, i.e., holding cash for a given asset bucket until a buy signal is triggered.

Before we look at the model’s results, consider how the asset weights fluctuate based on an unmanaged portfolio that’s initially set to equal weight as of Dec. 31, 2004:


Note the dramatic changes at roughly the midway point in the chart, which reflects the extreme market conditions in late-2008. The bond market’s weight quickly soared to over 60% while the allocation to stocks slumped below 40%. Clearly, Mr. Market’s decisions on asset allocation can suffer sudden shocks. It’s not obvious that dramatic changes to asset allocation are optimal, although it’s a safe bet that this is what you’ll endure if you leave Mr. Market in charge of portfolio design.

Now let’s consider how the asset weights change with the model outlined above. As the next chart shows, the results are quite different. In particular, the model has re-engineered the portfolio rules so that rising volatility triggers falling weights and vice versa. The portfolio design is also driven by the momentum factor, i.e., we own the asset if the trailing return is positive and the 50-day EMA is above the 100-day EMA; otherwise, we’re in cash for that particular asset. Over the 10-year period in this test, there were 25 rebalancing events.

Keep in mind too that the dynamic weighting rule will limit how much we invest in each asset and so there’s usually a varying amount of cash for each asset bucket, even during periods of a buy signal. One other point: the initial weights this time are based on the trailing data for the parameters as of Dec. 31, 2004. Based on those inputs, here’s how the SPY/AGG portfolio’s weights evolved over the past decade:

In contrast with the first chart, the second graph shows that the model is in control of the asset weights (as opposed to letting Mr. Market’s erratic choices run the show). As a result, the cash level went up sharply as market volatility in both ETFs increased in 2008.

How does the model compare in terms of performance? Not surprisingly, it underperforms a simple buy-and-hold strategy for an equal weight mix of SPY and AGG. Ditto for a year-end rebalancing strategy that moves the initial 50%/50% allocation back to equal weights every Dec. 31.

Is the model’s dramatic underperformance a sign of failure? No, not at all. Recall that the model is designed to control each asset’s risk contribution to 2%, which is a rather heavy bias on the side of conservative risk management. We could easily adjust the parameters to reflect a more aggressive risk tolerance, in which case the associated return would be higher.

The point of this toy example is to illustrate some of the tools at our disposal for managing risk. Ultimately, there’s a tradeoff between risk exposure and performance. If you’re intent on earning a higher return, you’ll have to tolerate higher risk. In the long run, an extreme strategy of letting Mr. Market choose the asset weights could end up generating returns in the upper quartile of results for similar strategies… maybe. In any case, it’s going to be a rough ride with Mr. Market at the helm.

Most investors can’t tolerate high levels of volatility in the short run and so the case for managing risk to engineer a smoother ride is well founded. The challenge, as always, is engineering risk down to a level that’s tolerable without giving up too much performance.

In the examples above I purposely looked at extremes–an unmanaged market-based portfolio vs. a strategy that sharply lowers risk. In the real world, the goal is usually a compromise: designing portfolios that exist somewhere in the middle of these extremes. Our task, by the way, is a quite a bit easier if we use a wider spectrum of asset classes rather than the two noted above.

Meantime, the test model excels in controlling risk. Consider how the drawdowns compare. The worst drawdown in our simple model was a light 2.97%:

Model portfolio Top-5 drawdowns

        From     Trough         To   Depth Length To Trough Recovery
1 2008-03-24 2008-10-08 2010-04-14 -0.0297    520       140      380
2 2011-07-25 2011-08-08 2012-02-27 -0.0216    150        11      139
3 2006-03-17 2006-06-13 2006-09-20 -0.0180    130        61       69
4 2013-05-22 2013-06-24 2013-09-18 -0.0176     83        23       60
5 2005-03-08 2005-03-29 2005-06-01 -0.0167     60        15       45

By contrast, letting Mr. Market’s strategy dominate resulted in a dramatically deeper drawdown in the worst-case period—a hefty 27.2% decline! A similar result is found with the year-end rebalanced strategy.

Buy & Hold 50%/50% portfolio Top-5 drawdowns

        From     Trough         To   Depth Length To Trough Recovery
1 2007-10-10 2009-03-09 2010-10-25 -0.2720    767       355      412
2 2011-07-25 2011-08-08 2012-01-03 -0.0727    113        11      102
3 2013-05-22 2013-06-24 2013-07-22 -0.0460     42        23       19
4 2007-07-20 2007-08-15 2007-09-18 -0.0457     42        19       23
5 2012-05-02 2012-06-04 2012-07-18 -0.0373     54        23       31

It’s important to think of TAA as a tool rather than a silver bullet solution. Much depends on the design details and the asset mix. Results can be quite sensitive to small changes in a given model and so it’s crucial to run stress tests before going live with an actual portfolio. Among those tests is estimating the higher taxes and trading costs that are inevitable with a TAA-based portfolio vs. Mr. Market’s strategy.

Nonetheless, the basic framework of controlling volatility at the input level offers an encouraging methodology for engineering outcomes that are closer to what investors can tolerate. It’s all about keeping short-term pain to a minimum without sacrificing too much of the long-run expected return. It’s not a free lunch by any means, in part because successful results require a fair amount of testing to create a model that matches specific requirements for long-term investment goals and short-term risk tolerance. The good news is that there’s a reasonably high degree of confidence for assuming that we can improve on Mr. Market’s chaotic management style.


2 thoughts on “Managing Portfolio Risk With Tactical Asset Allocation

  1. James Picerno Post author

    Thanks, Ron. The performance history for this particular strategy is shown in the wealth index chart. The cumulative return is roughly 25% for the period 12/31/2004 through 3/6/2015. Not very impressive, but the intention is to illustrate the methodology with a simple example. If we were building a real world portfolio we’d use more asset classes and probably dial up the risk exposure to make it more competitive. In short, we’d customize the portfolio design and risk management overlay to match a particular goal. That’s all in a day’s work, so to speak. The point of this toy example, meanwhile, is simply to highlight some of the tools at our disposal for “engineering” a portfolio outcome.

Comments are closed.