Relative and Absolute Momentum in Times of Rising/Low Yields

Our aim is to develop a very offensive (‘aggressive’) tactical asset allocation strategy, by combining some of our previous models like Protected- (PAA), Vigilant- (VAA) and Defensive (DAA) Asset Allocation. We will call this new strategy the ‘Bold Asset Allocation’ (BAA). BAA combines a slow relative momentum with a fast absolute momentum and crash protection, based on the concept of the ‘canary’ universe, where we switch from our offensive to the defensive universe when any of the assets in the canary universe has negative absolute momentum. (Page 1)

Momentum of stocks was (re)discovered in the academic investment literature by Jegadeesh (1993). Since Faber (2010), there is also a lot of practical interest in tactical asset allocations (or strategies) based on simple momentum models, like the SMA10 trend filter. (Page 1)

The core of these momentum models is a switch from an ‘offensive’ (or risky) single asset (like SPY) or universe to a ‘defensive’ (or risk-off) single asset (like IEF) or universe, depending on the number of ‘bad’ assets in the offensive universe. This is called trendfollowing or absolute momentum. Since it is effective in limiting drawdowns, we will also call this ‘crash protection’. (Page 1)

The choice of the best assets within an universe is often based on relative momentum: the assets with the highest momentum form, say, the Top3 assets, often equally weighted. This holds for both the offensive as well as the defensive universe. When both universes consists of only one asset (eg SPY as offensive and IEF as defensive), only absolute momentum matters. (Page 1)

Our strategies PAA (Protective Asset Allocation, see Keller 2016), VAA (Vigilant Asset Allocation, see Keller 2017) and DAA (Defensive Asset Allocation, see Keller 2018) all added some new momentum filters, eg. a slower SMA12 from PAA for relative momentum and a faster filter called 13612W from VAA and DAA for absolute momentum. Plus so-called ‘breadth momentum’. With this breadth momentum, the fraction defensive assets (risk-off) is determined by the breadth of a portfolio, ie. the number of assets with a ‘bad’ momentum. It replaces the traditional absolute momentum rule, where often each of the selected offensive assets (in, say, the Top3) were replaced by a defensive asset (like IEF) when it had a negative trend. (Page 1)

With the traditional absolute momentum rule, when choosing eg. the Top3 best assets from an offensive universe of say 12 assets, switching to defensive (ie. ‘crash protection’) only occurs very late, ie. when at least 10 assets are bad. With our ‘breadth momentum’ parameter eg. B=6. (Page 1)

In addition, we introduced a ‘canary’ or protective universe in DAA, separate from the offensive universe. Now switching is determined by the number of bad assets in the canary universe (see breadth example above, now applied to the protective universe). So, we distinguish three different universes: offensive, defensive and protective. (Page 1)

There are always three universes (Offensive, Defensive and Protective), where we switch between the Offensive and the Defensive universe. This switching is based on the ‘breadth momentum’ of the Protective (or ‘canary’) universe (see Keller 2018). For PAA-G12, the Offensive and Protective universe are equal and contains 12 assets (so NO=NP=12, see SelO and SelP in fig 1). The so-called breadth parameter B equals six (B=6). (Page 3)

This implies that the switching is 100% to defensive when at least six of the canary assets show negative (or ‘bad’) absolute momentum, no switching (so 0% defensive) with no canary assets ‘bad’, and so on, proportionally for 0<#bad<6. The Defensive universe of PAA-G12 only holds two assets, BIL and IEF (see SelD), so ND=2. All our assets are ETFs, traded at month end. (Page 3)

Besides absolute momentum, we will also use relative momentum in order to determine the selection of the best assets in the Offensive and Defensive universe. (Page 3)

All our strategies rebalances at a monthly frequency with equal asset weighting per universe. To give an example, with TO=6, TD=1, B=6 and 3 ‘bad’ canary assets, the PAA-G12 strategy chooses a mix of 50% of the best defensive asset (eg. IEF as Top1) and divide the rest equally over the best Top6 offensive assets (each 8,33%), for a total of 1x50% + 6x8,33 = 100% total, for 7 assets. (Page 3)

In this section we will look at the most aggressive BAA strategy called the BAA-G4 strategy. Although most characteristics are the same as BAA-G12, we now change the offensive universe to the aggressive VAA-G4 universe (with 4 Global assets, see Keller 2017) - plus one small change - and select only the Top1 offensive asset as with VAA-G4 (instead of the Top6 for PAA-G12). The change is that we substituted QQQ for SPY10, yielding a global offensive universe of four assets: QQQ, VEA, (Page 6)

VWO, BND. We will use the same defensive and protective (canary) universes as with BAA-G12, and use again SMA(12) for the relative momentum in the offensive and defensive selection (as well as for the absolute defensive momentum), and 13612W for absolute ‘breadth’ momentum for the protective (or canary) universe. (Page 7)

In this section we consider some variations on the BAA theme, in particular wrt. the (diversification of the) offensive selection (TO) and universe, and wrt. both momentum filters used in BAA. This might give the reader some information of the ‘robustness’ of the various BAA strategies. We start with the ‘offensive’ diversification. (Page 8)

Although BAA-G4 looks very impressive, some might call it ‘too concentrated’ with only one offensive asset chosen (TO=T=1) when not in defensive mode. Something similar (but opposite) might be said for the ‘very diversified’ offensive selection (TO=T=6) for BAA-G12. (Page 8)

In fig 11 we present this simplified BAA-SPY strategy. Notice that our combination of a defensive SMA(12) absolute and relative momentum together with an enhanced defensive universe (most bonds plus DBC) and our fast 13612W protective momentum with a good canary universe, seems able to protect for nearly all of SPY’s bear markets. (Page 9)

In both momentum variations, we included the very slow RET(12) momentum filter (ie. the return over 12 months) as possible alternative, besides the very fast 13612W filter and the slow SMA(x) filter (for x=3, 6, 9, 12 months lag). Notice that our SMA(9) filter (the best L for BAA-G4) equals the often used SMA10 filter (see note 5). (Page 11)

By combining a slow relative momentum filter (from PAA, see Keller 2016) with a fast absolute momentum filter (from VAA/DAA, Keller 2017/18) and allowing for some limited risky exposure in our defensive universe, together with a very fast crash-protection (based on a separate canary universe with breadth parm B=1, see Keller 2018), we arrived at several offensive strategies which we labelled the Bold Asset Allocation (BAA).. (Page 12)