Meb Faber

Faber's 10-month timing model: five asset classes, each with its own switch

Take the slow trend filter most people know from a single stock index, and run it separately on US stocks, foreign stocks, bonds, real estate and commodities. That is the entire system. The paper's finding was not that this beats buy-and-hold — it was that a rule simple enough to execute by hand delivered most of the drawdown protection investors were paying hedge-fund fees to get.

Style: Macro & allocation
Approach: Mechanical
Difficulty: Beginner
Horizon: Long term (years)
Holding period: Months to years
Time needed: 20 minutes at each month end
Markets: Broad ETFs · Asset-class index funds
Source: Meb Faber, 'A Quantitative Approach to Tactical Asset Allocation' (2007, updated 2013) — Meb Faber

The rule set

Split the portfolio equally across US stocks, foreign stocks, bonds, real estate and commodities
Once a month, at the close, compare each sleeve's price with its own 10-month simple moving average
Hold a sleeve while its monthly close is above that average
Move that sleeve — and only that sleeve — to cash when its monthly close is below it
No leverage, no shorting, and no decisions of any kind between the monthly checks

What makes it distinctive

One decision per asset class per month — the whole system fits on an index card and survives being executed by hand
Five sleeves that rarely leave the market in the same month, so exposure steps down in fifths rather than jumping to zero
The published result was not higher returns but materially smaller drawdowns, and that is the claim the evidence supports

When it works

Long bear markets, where being out of an asset class for months at a stretch is exactly the point. The rule stepped aside from the worst of 2000–2002 and 2008, sleeve by sleeve, as each broke down.

When it fails

Sharp V-shaped falls and recoveries whipsaw it — it sells near the low and rebuys higher, in as many sleeves as crossed. In a long grinding bull market it simply trails buy-and-hold while charging you the trading costs of five switches instead of one.

How a decision moves through it

Input
Monthly bars for five asset classes
One price series per sleeve — the shipped version uses SPY, EFA, IEF, VNQ and DBC as the ETF stand-ins for the paper's five indices. Monthly bars are the input, not a summary of it: the system never looks at a daily price.
Measure
A 10-month simple average, per sleeve
Ten monthly closes averaged — roughly the same line as a 200-day average, maintained from twelve numbers a year. Each sleeve gets its own; the average of one asset class says nothing about another.
Decide
Each sleeve's close against its own average
Five independent comparisons, run once at each month end. There is no portfolio-level signal anywhere in the system — no sleeve's answer depends on any other sleeve's.
Act
Hold the sleeve, or move that sleeve to cash
The unit of action is the sleeve, never the portfolio. A month can end with any mix — all five in, all five out, or anything between — and each of those mixes is the system working as designed.

What the paper actually showed

Faber's claim was deliberately modest: applying a 10-month trend filter to each of five asset classes produced returns in the neighbourhood of buy-and-hold with far smaller drawdowns — roughly the risk-adjusted profile investors were paying tactical managers two-and-twenty to deliver.

The paper tested one rule on five indices — US stocks, foreign stocks, government bonds, real estate and commodities — holding each while its monthly close sat above its own 10-month average and parking that sleeve in cash otherwise. Nothing is forecast. The filter reacts to declines that have already persisted for months, which is what distinguishes a bear market from a correction.

One sleeve's year. On monthly bars the 10-month average is the 200-day line under another name — and this exact picture runs five times in parallel, one per asset class, each on its own clock.

It has become the reference implementation of tactical asset allocation precisely because there is nothing in it to argue with: one parameter, one cadence, five liquid markets. Anyone can reproduce it, and many have — which is also why its failure modes are unusually well documented.

Five switches are a different machine from one

A trend filter on a single index is binary: fully invested or fully out, and every whipsaw hits the whole portfolio. Running the same test on five sleeves changes the arithmetic. Exposure moves in steps of 20%, a whipsaw in commodities leaves the bond sleeve untouched, and in an ordinary year the five switches disagree with each other more often than they agree.

The portfolio has six exposure states — 0%, 20%, 40%, 60%, 80%, 100% — instead of two, so risk comes off gradually as the evidence accumulates.
Whipsaw risk is diversified: five smaller, uncorrelated whipsaws are easier to sit through than one large one, even if there are more of them in total.
Each sleeve exits on its own schedule in a slow-building bear market, which is what stepping aside from 2008 actually looked like — not one clean exit, but five staggered ones.

The honest caveat, developed fully in the failure-modes page: the independence of the five switches is a fair-weather property. In a genuine crisis, asset classes fall together and the switches fire together — the diversification helps most in ordinary years and least in the month you would most like it to.

Five ways into this system

The ideas behind it

This system assumes you already know these. Each one is explained from scratch in Investing 101.

These are documented methods described for study. Nothing here is investment advice, a recommendation, or a claim about future returns — every system on this page has losing periods, and the pages say where.

Reading about a system is not having one.

Plutux is where you write your own rules down, test them against real data, and keep the record your memory would otherwise rewrite. Join the waitlist for early access.