Tactical allocation: momentum, dual momentum and trend following

In depth7 min read · 9 sources

Tactical allocation changes the weights according to fixed rules: into assets in an uptrend, out of falling markets, or less risk when volatility rises. The best-known rules are momentum, dual momentum, moving averages such as the 200-day line and target volatility. They rest on some of the best-documented effects in financial research, but also on backtests that are particularly prone to looking too good.

Momentum rotation

Assets that have performed best over the last 6 to 12 months tend, on average, to keep doing so for a while. Jegadeesh and Titman (1993) showed this for US stocks: buying past winners and selling past losers produces clearly positive returns over holding periods of 3 to 12 months; part of this reverses over the following two years. Asness et al. (2013) found momentum in eight markets and asset classes, from equities and bonds to currencies and commodities. A rotation strategy holds the best one to three assets from a list each month.

Dual momentum

Antonacci (2014) combines two tests: relative momentum (which asset did better?) and absolute momentum (did it do better than the money market at all?). Absolute momentum acts as an emergency exit: if all equity markets fall, the strategy switches into bonds or the money market.

In research, absolute momentum is called time-series momentum. Moskowitz et al. (2012) document it for all 58 futures they studied on equity indices, currencies, commodities and bonds: returns over the last 1 to 12 months persist and partly reverse over longer horizons. A broadly diversified portfolio of such signals performed well precisely in extreme market phases. More on this in the article trend following.

Moving averages and the 200-day line

If the price is above its average over the last 200 trading days (or 10 months), you stay invested; otherwise you move into the money market. Faber (2007) tested a 10-month rule across five asset classes since 1972, with the result of equity-like returns at the volatility and drawdowns of bonds:

The rule often avoids most of a long bear market, but costs return in sideways markets through false signals (“whipsaws”): the price falls just below the average, you sell, it rises again, and you buy back at a higher price.

Target volatility

Because high volatility often goes hand in hand with falling prices, the rule usually reduces risk in bad phases. Moreira and Muir (2017) show that portfolios which take less risk when volatility is high achieved higher Sharpe ratios for the market and many factors, because the expected return in turbulent phases does not rise to the same extent as risk. Harvey et al. (2018) find mainly fewer extreme losses for equities and hardly any benefit for bonds. A similar principle with valuation instead of volatility: setting the equity share by the Shiller P/E.

For advanced readers: crashes, combinations and risk weighting

Momentum crashes: momentum rarely loses, but when it does, it loses heavily. Daniel and Moskowitz (2016) show that such crashes occur in panic phases after market declines and when volatility is high, precisely when the market suddenly rebounds sharply and the stocks that had fallen before rise the most. In their study, a strategy that scales its risk to expected volatility roughly doubled alpha and the Sharpe ratio compared with static momentum.

Combinations: according to Asness et al. (2013), value and momentum are negatively correlated. Combining both smooths the dry spells of each individual strategy. In a similar way, a trend rule complements an equity-bond portfolio: it helps in long downturns, in which rebalancing alone achieves little.

Adaptive asset allocation: Butler et al. (2012) combine selection by momentum with weighting by risk: a minimum variance portfolio is formed from the assets with the best momentum. This prevents three highly correlated winners from ending up in the portfolio at equal weight.

Short-term reversal: many studies leave the most recent month out of the momentum window, because stocks tend to reverse over one month. For broad indices and asset classes this effect is weaker; rules such as dual momentum usually use the full 12 months.

Which assets are suitable?

Tactical rules need assets that genuinely behave differently. There is hardly anything to rotate between two global equity ETFs; between equities, government bonds, gold and the money market, there is. In the literature, broad indices of individual asset classes and regions have proved their worth, but individual stocks have not, because their momentum comes with high trading costs and strong reversals.

• Good candidates: world, US, European and emerging market equities, government bonds of various maturities, gold, broad commodity indices.
• Fallback asset: money market or short-dated government bonds. Long-dated bonds are often a good safe haven in equity crashes, but not when interest rates rise, as in 2022.
• Too few assets: with two assets there are only two states; a single false signal can then decide the outcome for years.
• Long history: for a rule with a 12-month window, the backtest needs at least a few decades to include several bear markets. ETFs alone rarely suffice for this. Asset classes with long series help.

Backtest pitfalls

• Signal and trade: a month-end signal may only affect the returns that follow. Applying the new weights to the return of the month in which the signal arose means looking into the future.
• Overfitting: if you test 50 parameters and show the best one, you are finding chance. Better to test a few common parameters and several periods (see backtest overfitting).
• Costs and taxes: frequent switches trigger taxes, which are often missing from backtests. Every switch into the money market realises the gains on the equity position.
• Long dry spells: in strong bull markets, trend following often lags buy-and-hold for years.
• Small asset list: with two or three assets, a single switch often decides the result. The range of results with slightly different parameters says more than the best value.

How Portfolio Lab calculates it

The Tactical page offers eleven models: moving average (per asset, default 10 months, selectable from 2 to 24), dual momentum (default 12 months, with absolute momentum against the money market), relative strength over averaged windows of 1, 3, 6, 9 or 12 months, target volatility (default 10% p.a.), adaptive allocation (top N by momentum, weighted by minimum variance, inverse volatility or equally), the equity share by the percentile of the Shiller CAPE, four rotation models with breadth momentum (Protective, Vigilant and Defensive Asset Allocation (PAA, VAA, DAA) as well as Accelerating Dual Momentum) and a regime model (hidden Markov model). All signals are calculated at month-end using only prices up to that day; the portfolio is rebalanced at that closing price, and the new weights apply from the following month. In real trading, the closing price can only be hit approximately. The benchmark is buy-and-hold of the base weights with monthly rebalancing; the page does not calculate taxes, and trading costs only if you set a rate per order (the same for model and benchmark). Optionally, the portfolio is rebalanced only one month after the signal. A fallback asset can be used instead of the money market. Details are in the tactical models methodology.

This article explains a method; it is not investment or tax advice. Worked examples are simplified and rounded. All terms in the glossary.