Skip to content

Breadth Momentum and the Canary Universe: Defensive Asset Allocation (DAA): summary and key takeaways

The paper replaces the crash signal of Vigilant Asset Allocation with a separate two-asset canary universe (VWO and BND). In backtests the average cash fraction falls to less than half that of the earlier strategy, while return and return-to-risk are mostly similar or better.

Keller, W.J.; Keuning, J.W. · 2018 · SSRN Electronic JournalSummary updated 20 November 2026
Published2018Working paper
Asset focusMulti-assetMomentum / TAA
TypeBothTactical Allocation
Reading time5 minthis summary

Breadth Momentum and the Canary Universe: Defensive Asset Allocation (DAA) at a glance

The question
Can a small separate set of canary assets signal crash protection with a much lower average cash holding than breadth momentum on the full risky universe?
Dataset & period
Monthly total-return data using index and ETF proxies. Canary search on the US large-cap index and bonds, Dec 1926 to Mar 2018, with the search sample ending Dec 1970. Strategy tests from Dec 1970 to Mar 2018, optimized in-sample to Dec 1993 and evaluated out-of-sample after that.
Method
Each month, the canary assets (VWO and BND) are checked for non-positive fast momentum (the 13612W filter). The cash fraction equals the number of bad canary assets divided by a breadth parameter. The risky part holds the top-ranked assets by relative momentum, equal weight, and the cash part holds the best bond from a small bond universe. The only free parameter tuned in-sample is the number of top risky assets, optimized on a drawdown-adjusted return measure.

Key findings of Breadth Momentum and the Canary Universe: Defensive Asset Allocation (DAA)

  • On the global twelve-asset universe, out-of-sample cash fraction is 28.8% for DAA versus 60.8% for VAA.
  • For the same universe out-of-sample, DAA return is 12.9% versus 8.2% for VAA, with similar maximum drawdown.
  • Max drawdown for DAA on that universe is 9.0% out-of-sample, against 39.2% for the equal-weight passive benchmark.
  • The canary search found VWO plus BND best on the pre-1970 sample. Next-month US equity returns after both signal bad momentum are weaker than otherwise, both in average return and in share of positive months.
  • On a four-asset risky universe, DAA with the default setting did less well than VAA, although drawdown was 20.3% versus 44.9% for the passive benchmark out-of-sample.
  • An aggressive variant (all-in or all-out signal, leveraged-style bond universe) reaches a full-sample return of 22.3% versus 16.1% for the standard all-in or all-out version.

Robustness, as reported in the paper

  • The canary universe was chosen on data ending Dec 1970, then tested on later data to limit datasnooping.
  • Parameters were optimized in-sample to Dec 1993 and evaluated out-of-sample afterward.
  • Backtests include one-way transaction costs of 0.1%, and the paper reports total costs of 0.74% versus 1.45% for the single-risky-asset DAA variant versus VAA twelve-asset.
  • Results are shown for four different risky universes, plus a recent ten-year subsample.

Caveats and limits

  • Most historical prices are non-tradable index or ETF proxies; only recent years use observed ETF prices.
  • The paper itself notes the leveraged variants rely on simulated proxies and illiquid products, so those results are questionable.
  • The four-asset universe underperforms VAA, so the canary benefit is not uniform across universes.
  • Return and drawdown figures come from one monthly backtest design with in-sample parameter choices (inferred).

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

Why it matters

The paper shows how crash protection can be separated from the choice of risky assets, so cash exposure can be tuned independently. Practitioners comparing momentum rotation designs can use it to see the trade-off between cash drag, tracking error and drawdown.

Questions about this paper

What is the canary universe in DAA?

It is a separate pair of assets, VWO and BND, whose momentum sets how much of the portfolio moves to cash. It does not need to be part of the risky or cash universes.

How does DAA compare with VAA on cash holdings?

On the twelve-asset universe out-of-sample, DAA holds 28.8% cash on average versus 60.8% for VAA.

Does DAA beat VAA everywhere?

No. It did worse than VAA on the four-asset universe, and the paper reports mixed risk results on others.

Where can I read the original paper?

The original is at its publisher: https://doi.org/10.2139/ssrn.3212862. 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.