Skip to content

Protective Asset Allocation (PAA): summary and key takeaways

The paper adds a breadth-based bond fraction to a simple moving-average dual momentum model. In the most protective setting, drawdowns stay far below those of dual momentum, a stock/bond mix and the S&P 500 ETF, while returns remain above the stock/bond mix.

Keller, W.J.; Keuning, J.W. · 2016 · SSRN Electronic JournalSummary updated 20 November 2026
Published2016Working paper
Asset focusMulti-assetTAA
TypeBothTactical Allocation
Reading time5 minthis summary

Protective Asset Allocation (PAA) at a glance

The question
Can a breadth indicator that shifts capital to safe bonds when many assets trend down turn dual momentum into an absolute-return strategy comparable to a term deposit?
Dataset & period
Twelve global multi-asset ETF proxies (US, developed and emerging equities, REITs, commodities, gold, long and credit bonds) plus a 7-10y Treasury fund as safe asset. Monthly total-return data, Dec 1970 to Dec 2015 (45 years). In-sample ends Dec 1992.
Method
Each month the model scores every asset by its price relative to its moving average. Assets with positive scores are counted as good. The bond fraction rises as the count of good assets falls, scaled by a protection factor. The best good assets are equal-weighted for the risky part and the rest goes to a safe Treasury fund. Parameters were chosen in-sample from 72 combinations, then tested out-of-sample.

Key findings of Protective Asset Allocation (PAA)

  • Over the full sample, the most protective variant (PAA2) had a maximum drawdown of about 10%, versus 25% for dual momentum, 30% for the static stock/bond mix and 51% for the S&P 500 ETF.
  • Out-of-sample, dual momentum had the highest return at 12.5%, with the low-protection PAA0 second at 11.6%. All PAA variants beat dual momentum on risk and risk-adjusted measures.
  • Out-of-sample, PAA2 was the only model meeting the absolute-return targets, with 95% of rolling 1-year returns non-negative and 99% not below -5%.
  • The average bond fraction over the whole period was 52.2%, so the model behaves like a roughly 50/50 mix with an active split.
  • The rolling 1-year return fell below -5% only in one episode in the mid-1990s, lasting three months.
  • In the early 2016 pullback PAA2 moved fully into bonds and made 3.3% in January 2016, against -1.7% for the static stock/bond mix.

Robustness, as reported in the paper

  • Parameters were fitted on an earlier sub-period and tested on a later one. Returns were lower out-of-sample, which the authors link to lower interest rates and lower equity returns.
  • Swapping the moving-average filter for a 12-month return filter worsened PAA2 on the full sample: return 13.4% and maximum drawdown 18.1%.
  • A Sharpe ratio test with a data-snooping haircut still gave a t-ratio of 2.46 in-sample.
  • Alternative safe bonds (shorter Treasuries, T-bills and momentum-chosen combinations) were tested, and most still met the absolute-return targets.

Caveats and limits

  • The in-sample years include the optimisation, so the results before 1993 carry data-snooping risk, as the authors note.
  • Index-based ETF proxies were constructed for the early decades, since the real funds did not exist (inferred from the data appendix).
  • Transaction costs use one flat one-way assumption that the authors say may be off in either direction for different eras.
  • Several results depend on the choice of safe bond and on a long period of falling yields (inferred).

Items marked (inferred) are our reading of the design, not statements by the authors.

Why it matters

The paper shows how a universe-wide breadth signal can cut drawdowns more than per-asset trend filters alone. It also shows the trade-off: lower risk came with lagging returns in strong bull markets. Practitioners can use it as a design reference for protection overlays on tactical models.

Questions about this paper

How does PAA decide how much to hold in bonds?

It counts assets in the universe with positive momentum. The fewer good assets, the larger the bond share, scaled by a protection factor.

What does the most protective setting do?

With protection factor 2, the portfolio goes fully into the safe bond fund once half or more of the assets have non-positive momentum: 100% bonds when 50% or more are bad.

Which safe bond did the authors prefer?

The default was a 7-10y Treasury fund. They suggested a momentum-chosen mix of 1-3y and 7-10y Treasuries as a possible safer alternative.

Where can I read the original paper?

The original is at its publisher: https://doi.org/10.2139/ssrn.2759734. We link to the paper and do not host the PDF.

Follow new quant research

Download MacroFinance market data as CSV or Excel, or pull it by API, and test the ideas in these papers yourself. New summaries are added as we review them.

Create a free account
Summaries are written by MacroFinance from the published papers, in our own words, for information only. They are not investment advice and may omit detail. Papers are not hosted here: each page links to the original at its publisher.