Faber GTAA 5

Who it is for and who it is not for

It is a reasonable rule for anyone who wants to understand how a tactical strategy works before moving on to more complex ones. It needs five ETFs, a little cash and a review once a month.

It does not suit anyone who needs the certainty of beating the stock market in every period, who cannot absorb the tax cost of rotating or who doubts they would keep their discipline when the rule tells them to sell just before a rebound.

At the close of the last business day of each month, using the total-return monthly price (that is, with dividends reinvested):

1. For each of the five assets, calculate the 10-month simple moving average: the average of the last 10 monthly closes, including that of the month that has just ended.
2. If the closing price is above its average, that asset keeps its 20% of the portfolio.
3. If it is below, that 20% is set aside in Treasury bills. In practice anywhere from 0 to 5 parts can be in cash.
4. Fluctuations within the month do not count; you only act on the monthly close.
5. Hold until the close of the following month and repeat. Faber indicates monthly rebalancing in his answers to frequently asked questions (p. 55).

The five assets in the paper and the US ETFs this site uses to calculate them:

No ETF replicates the paper's index perfectly. To understand the rule that does not matter; to compare figures with the paper, it does.

Decisions made by this site

The paper leaves some details open and the site has had to choose:

• The average includes the current close. It is the last 10 monthly closes counting the month being evaluated.
• If the price is exactly equal to the average, the engine treats it as "below" and sends that part to cash. The case is theoretical.
• Cash is BIL, an ETF of 1-3 month bills, not the 90-day bills in the paper.
• A 0.1% cost per trade is applied, the same assumption used by Keller's papers. Faber, by contrast, publishes his figures without commissions, taxes or slippage (p. 22). That difference explains part of the gap between his numbers and ours.
• Each month, the parts that are invested are reset to 20% per asset; weights are not carried over from one month to the next.

The data is calculated with the close of the last business day and the order is assumed to be executed at that same close, as in the paper. Someone trading real money will do so, at the earliest, the next day.

With real ETFs from Apr 2008 to Sep 2026 (since data exists for all the assets and the 10 closes the average requires), our engine returns 4.5% a year, with volatility of 6.8% and a maximum monthly drawdown of -11.1%. The Sharpe ratio (return above cash divided by volatility) comes out at 0.50. On average, 33% of the portfolio is in cash.

Average annual turnover is 194%: that is the total purchases in a year as a percentage of the portfolio. It looks like a lot for a slow rule, but these are five independent decisions: it is enough for one asset to cross its average for there to be a trade, and with five assets that happens often.

A useful detail: the paper was circulated in May 2006 and our ETF history starts later (Apr 2008), so the whole backtest is after publication. The weights and the rule cannot have been tuned with this data. That does not prove it will work in future, but it is a cleaner check than a rule that was tested and published on the same past.

The "Year by year" table below and the period selector let you compare with the S&P 500 and the 60/40 portfolio over whichever stretch interests you. It is worth looking at individual years: this kind of rule tends to lag the stock market in good years and make up for it in bad ones.

What the paper says

These are the figures Faber publishes, copied from his charts and not recalculated by us:

The 1973-2012 Sharpe is calculated with an average risk-free rate of 5.41%. The 2006-2012 stretch is the one Faber considers out-of-sample, because the original study ran to 2005 (p. 33). The results are total-return series with no costs or taxes.

What Faber's figures show is a return similar to buy and hold, with much lower volatility and maximum drawdown. Our figures with ETFs and costs, of 4.5% a year over Apr 2008 - Sep 2026, are not comparable one to one with 1973-2012: the period is different, the assets are not identical and there are costs. But they do confirm the shape of the result, a maximum drawdown far more contained than equities, in exchange for a return that does not stand out.

With average annual turnover of 194% in the history, there will be several sales a year, although each affects a fifth of the portfolio. Every sale of an ETF may realise a taxable gain or loss, depending on your country and the type of account. The figures on this page are before taxes.

Strengths

• A rule anyone can verify. Five averages and five comparisons. No hidden parameters and no ranking to reorder.
• It exits what loses its trend. When an asset falls below its average, that part moves to cash without waiting for another signal to confirm it.
• Diversified from the start. US equities, international equities, bonds, real estate and commodities; each part is decided without depending on the others.
• Little work. One review a month, done in a few minutes.

Real criticisms and limits

• Possible optimisation of the average. Ten months is not a number with a theoretical basis. Faber shows the results are stable with averages from 3 to 12 months (Fig. 15, p. 33), which supports that it is not an isolated case, but does not remove the risk of having chosen the period that worked best.
• No costs or taxes. The paper's figures exclude them (p. 22). In taxable accounts, the author himself recommends other ways of rebalancing (pp. 22 and 55).
• It won in few years. Between 2006 and 2012 it beat buy and hold in only 3 of 7 years (p. 34), even though the total result was better. The advantage is concentrated in crisis years.
• Sideways markets and whipsaw. When an asset crosses its average up and down in consecutive months, you buy high and sell low. With five independent assets there are more chances for one of them to do so.
• Turnover. Our history gives annual turnover of 194%. In a taxable account each ETF sale may realise a taxable gain (see the taxes section).
• ETFs are not the indices. The paper uses series from 1973; with ETFs there is only data from Apr 2008. Under 20 years, with a crisis (2008) at the start, is a small sample.
• Five assets, a single mechanism. There is no protection if the trend signal fails, for example in a fast drop that reverses before the monthly close.

1. Wait for the close of the last business day of the month. This page's signal is calculated with that close.
2. Check in the "Current signal" section which assets are above their average and which go to cash.
3. Map each ETF in the signal to the fund you actually hold.
4. Work out the orders with the rebalancing calculator.
5. Trade on the first business day of the following month. The rule assumes execution at the close, so there will be a small difference from the backtest.
6. Keep a record of each trade for your tax return.