Front-running the 60/40 rebalance: gross Sharpe 0.38 on a futures pair. Predictable is not the same as profitable.

The idea worth investigating

Every allocator I have worked with rebalances on a calendar, a threshold, or both. When stocks outrun bonds, the portfolio sells stocks and buys bonds to get back to target, and it does so on a schedule anyone can read. The trade that follows from this is the oldest idea on the desk: get there first.

This week we tested "The Unintended Consequences of Rebalancing" by Campbell R. Harvey, Michele G. Mazzoleni and Alessandro Melone, NBER working paper 33554, March 2025. The paper's claim is structural: institutional investors run trillions of dollars of calendar- and threshold-driven rebalancing, and a portion of those trades is predictable from public information.

The mechanism is plausible. The investment question is whether an implementable trade captures enough of it after costs.

Here's the plan:

• 1. Rebuild a 60/40 tracker that combines threshold-driven and month-end signals, and apply each signal to the following observed stock-minus-bond return.
• 2. Run it on S&P 500 and 10-year Treasury futures, ES and ZN, the instruments a desk would actually use.
• 3. Run a second construction on ES against TLT, the long-duration Treasury ETF, which has a longer history but a different duration and, in our retained series, no dividend adjustment.
• 4. Charge 1.5 basis points per unit of turnover and compare gross, net and profit factor across both constructions.

One result we can share here. On the futures pair, over 1,256 observations after a 2020 burn-in and through April 17, 2026, the gross Sharpe ratio was 0.38. That is before the turnover charge, on a house implementation whose declared parameters were rerun without re-optimization, and it is a comparison of one research construction, not a controlled experiment isolating instrument choice.

Subscribers get the full two-construction table (gross Sharpe, net Sharpe, net profit factor), why the ETF proxy looks better and why that cannot be attributed to the rebalancing mechanism, a date correction that changes how the "out of sample" period should be read, a separate 277-observation slice after the paper's March 2025 issue date, and the desk assessment of what would have to be verified before treating either construction as executable.

Let's get started.

AI assists curation and drafting. The research status distinguishes paper reviews from house tests. Read our research approach.