Expected Returns Ch. 5: Can You Rotate Between Premia? — The Honest Evidence

Investment Background

The first four chapters created a tempting situation.

If premia are observable, and they vary over time — sometimes the equity premium is high, sometimes credit spreads are wide — then the obvious next step is to overweight when cheap and underweight when expensive.

This chapter handles that honestly. And the answer is a carefully stated "mostly no."

The Wall Street Translation

The Signal Is Real

This must be granted first, because denying it would be dishonest.

Premia genuinely vary over time, and current levels genuinely contain information about future returns.

• Stocks for the Long Run ch05 demonstrated it with CAPE: starting valuation correlates robustly and negatively with the following decade's returns.
• Credit spreads widen in recessions, and buying corporate bonds at very wide spreads has historically produced higher subsequent returns.
• Term premia likewise sit at different levels in different periods.

So the intuition "buy more when cheap" is not wrong. It has evidence behind it.

The question is not whether the signal exists but whether an individual investor can actually use it.

Why It Is Extremely Hard to Use

Four reasons, each independently sufficient, and decisive in combination.

One, the signal is extremely slow.

Valuation signals operate on horizons measured in decades, not months or quarters.

A high-valuation signal may continue to be contradicted for five years — markets keep rising — and then pay off in year eight.

This is the same observation as the reflexive boom in Chapter 2 of The Alchemy of Finance: there is a very long lag between recognizing overvaluation and its correction.

And a person's patience is usually far shorter than a decade.

Two, the sample is tiny.

This follows directly from Bernoulli's law in Chapter 3 of Against the Gods.

If a signal's cycle is ten years, then two centuries of data contain twenty non-overlapping observations. And those share one country's institutional history.

Validating a strategy on twenty observations is statistically very weak.

Three, implementation costs consume most of the edge.

Rotation requires trading. Trading generates costs, taxable events, and spreads. Chapter 3 of Where Are the Customers' Yachts? established how much of that is invisible.

And those costs are certain while the signal's edge is not.

Four, and most importantly: the user abandons it at the worst moment.

Worth developing, because it is the main cause of real-world failure.

A valuation-based strategy requires reducing equity exposure when markets are expensive — that is, while everyone else is making money and you are behind.

It may require underperforming for three to five consecutive years.

The Psychology of Money and Chapter 4 of Poor Charlie's Almanack both establish what happens: in year three you abandon the strategy — usually just before it begins to pay.

So the strategy's real return is not the backtested return but "the backtest minus the loss caused by quitting partway."

And for most people that subtraction exceeds the entire edge.

So What Can Be Done

This chapter does not say the information is useless. It says the information belongs somewhere else.

This is the book's most important dividing line for you:

Use	Feasible	Reason
Setting planning assumptions from current yields	✅ Feasible and recommended	Requires no timing, no trading, no patience
Setting savings and withdrawal rates	✅ Feasible and recommended	Operates on variables you fully control
Mechanical rebalancing	✅ Feasible and recommended	Automatically buys cheap without judgment
Tactical rotation between premia	❌ Not recommended	Signal weak, slow, small-sample, and abandoned
Equalizing premia with leverage	❌ Not applicable	Requires leverage

Note the third row, because it is this chapter's most practical finding.

Mechanical rebalancing captures part of what rotation is trying to capture, and requires no judgment at all.

When equities rise above their target weight, rebalancing sells some — trimming when expensive. When equities fall, rebalancing buys — adding when cheap.

It does that automatically and mechanically, and:

• It requires no judgment about what counts as "expensive."
• It requires no conviction through three years of underperformance — because the rule asks you to believe nothing.
• Turnover is minimal, so costs stay controlled.

Which is why we recommend mechanical rebalancing across this site and do not recommend tactical allocation.

It is the executable version of the same intuition.

An Honest Note

It must be conceded: Ilmanen himself, and his institution, do practice tactical premium allocation.

And their conditions differ from yours entirely:

	Institution	You
Time horizon	Can absorb years of underperformance, protected by mandate	Will quit in year three
Costs	Very low institutional trading costs	Higher, including tax
Tools	Can short, can hedge unwanted exposure with derivatives	Usually cannot
Data	Dozens of markets, decades, a professional team	Public data
Consequences	An error is a performance issue	An error is a delayed retirement

The last row is decisive, and it has the same structure as the argument about Munger's concentration in Chapter 6 of Poor Charlie's Almanack: asymmetric consequences mean the same strategy has different correct answers for different people.

Executable Trading Rules

1. Use current yields for planning, not for trading. The book's most important dividing line and this chapter's core. The same information, two uses, entirely different evidentiary status.

2. Replace tactical rotation with mechanical rebalancing. Concretely: set target weights and a deviation threshold (say five percentage points); rebalance when it is breached and otherwise do nothing. This captures part of the valuation information without requiring judgment or patience.

3. If you do apply any tilt, keep it small and write the exit condition in advance. While calm, write: "under what conditions do I concede this tilt was wrong?" Without that condition you will abandon it at the worst moment (the fallibility method in Chapter 3 of The Alchemy of Finance).

4. Do not commit to a strategy requiring a decade of patience on the strength of a backtest. Ask honestly: if this trailed the S&P by three points for four consecutive years, would I really hold on? Most people's honest answer is no — which means the strategy does not suit them however good the backtest.

5. Accept that "use it to plan, not to trade" is a complete answer rather than a compromise. Adjusting savings and withdrawal assumptions is a real and valuable act. It is not inferior to timing — it is the practically executable use of the same information.

Relevance to a Retirement Portfolio

This chapter has a concrete, potentially outcome-changing application for retirees.

It concerns a real decision: what should someone approaching retirement do when valuations are clearly elevated or real yields clearly low?

The wrong answer: cut equity exposure and wait for a better entry. That is timing, and this chapter's four reasons explain why it fails.

The right answer: adjust the variables you control.

Variable	In a high-valuation / low-yield environment
Withdrawal rate	Lower from 4% to 3.5%
Cash buffer	Expand from two years to three
Retirement date	Consider delaying one to two years
Savings rate (if still accumulating)	Raise it
Asset allocation	Largely unchanged

Note the last row.

The first four require no forecasting and their effects are certain. The fifth requires forecasting and its effect is not.

Which is this chapter's — and this book's — final recommendation to a retirement investor:

This premium information is real and useful. It should change the numbers in your plan, not the holdings in your account.

Stocks for the Long Run ch05 reached exactly the same conclusion from the valuation side. Two books, two routes, one answer — which is itself evidence the conclusion is sound.

Chapter 6 sets the book's boundary: where it stops, and what a retirement investor should take from it.