Stocks for the Long Run Ch. 4: How Long Is "Long"? — Holding Period Risk Convergence

Investment Background

The first three chapters established the long-run numbers. This chapter takes on the question everybody actually cares about:

How long is "long run"? Can I wait that long?

This is the most-cited section of Siegel's book. It is also the most misquoted.

The Wall Street Translation

The Convergence Phenomenon

Siegel's method: across two centuries of US equity data, compute the best and worst annualized real returns for holding periods of different lengths.

The results run roughly as follows:

Holding period	Worst annualized real return (approx.)	Best annualized real return (approx.)
1 year	−38%	+67%
5 years	−11%	+27%
10 years	−4%	+17%
20 years	+1%	+12%
30 years	+2.6%	+10.6%

Note the twenty-year row.

In the US historical data Siegel examined, no twenty-year period produced a negative real equity return. In the worst twenty-year stretch, purchasing power still grew slightly.

And the change in the width of the distribution is striking: the one-year range spans more than a hundred percentage points; the thirty-year range spans about eight.

This is holding-period risk convergence.

Why It Converges

Understanding the mechanism matters more than memorizing the table — and Chapter 3 already laid the groundwork.

Recall the decomposition: total return equals dividends plus earnings growth plus valuation change.

• Short run: valuation change (sentiment) dominates, and sentiment swings enormously.
• Long run: valuation change is diluted toward zero, leaving dividends and earnings growth — both relatively stable, because both come from actual economic activity.

So convergence is not a statistical coincidence. It is a direct corollary of that decomposition. As the horizon lengthens, the dominant term shifts from "what others think" to "what businesses earned."

It also explains why bonds show no comparable convergence: the real return on bonds is dominated by inflation, and inflation does not cancel itself out over long horizons — 1940 to 1981 in Chapter 2 is the proof.

A Misreading That Must Be Corrected Immediately

This is the most important part of the chapter. Siegel's table is widely quoted, and it is nearly always restated as a guarantee.

The incorrect statement: "History proves that if you hold for twenty years, you cannot lose money in stocks."

That sentence has four independent problems.

Problem One: The Sample Is Startlingly Small

Two hundred years sounds like a lot of data. But there are only ten independent, non-overlapping twenty-year periods.

Ten observations.

Concluding "it has never happened" from ten non-overlapping samples is statistically very weak evidence. The distance between "did not occur in ten tries" and "cannot occur" is very large.

Problem Two: This Is US Data

This is the most serious criticism of Siegel, and Chapter 6 develops it fully.

The United States was the most successful economy in history across those two centuries. Using the most successful country's record to infer that "equities must revert over the long run" is inference from a winner's sample.

Other countries' records differ. The Dimson, Marsh and Staunton international dataset shows several developed markets with very long stretches of negative real returns. In 1910s Russia and 1940s China, investors lost 100% — those markets simply ceased to exist.

Problem Three: The Mean-Reversion Mechanism Is Not a Guarantee

Convergence occurs because the valuation term is diluted. But that assumes valuations eventually return to some normal range.

That assumption has held historically, but it is an empirical regularity, not a law of physics. No mechanism compels price-to-earnings multiples to revert.

Problem Four: You May Not Have Twenty Years

This is the most direct point for retirement planning.

A 65-year-old who needs a particular sum at 70 has a five-year holding period, not a twenty-year one. The worst case in that row is −11% annualized.

"It works out over the long run" is useless comfort to someone who does not have the long run.

So What Is the Table Actually Good For

With those four excluded, what remains is still valuable — and more precise:

1. The longer the holding period, the smaller the dispersion of outcomes. This directional conclusion is robust and mechanically supported by the Chapter 3 decomposition.

2. Short-run equity risk is systematically underestimated and long-run equity risk is systematically overestimated. Most people are unprepared for a one-year 38% decline while being needlessly frightened of thirty-year horizons.

3. Your holding period determines which row you should be reading. This is the most practical use: do not ask "are equities risky." Ask "over my actual holding period, what does the historical distribution look like."

Executable Trading Rules

1. Segment money by when it will be spent, then allocate by holding period. This is the only way to turn this chapter into action:
2. Money needed in 0–3 years → cash and short-term bonds. The first row tells you equities do not belong here.
3. 3–10 years → a blend.

4. 10+ years → equity-dominated, because on that scale purchasing-power risk exceeds volatility risk.

5. Never present "twenty years has never lost" to yourself or anyone else as a guarantee. The accurate statement is: "Across ten non-overlapping twenty-year periods in two centuries of US history, none produced a negative real return." That sentence is longer, but it is correct, and it preserves your ability not to fall apart when you meet a counterexample.

6. Plan against the worst case, not the average. Test whether your retirement plan still works at the thirty-year worst case of 2.6%, rather than the 6.5% average. If it survives the worst case, your plan is robust.

7. Accept that your holding period shortens over time. A 40-year-old has twenty-five years; the same person at 60 has five (for the portion about to be spent). Allocation must change accordingly — this is not market timing, it is duration matching.

8. Do not use this table to talk yourself into equity exposure beyond your tolerance. Convergence describes the dispersion of returns. It does nothing whatsoever to change whether you sell after a 40% decline in year three. That is The Psychology of Money's territory, and psychological capacity is an independent and equally binding constraint.

Relevance to a Retirement Portfolio

This chapter directly supports the bucket framework and age-layered allocation used across this site.

The concrete application runs like this:

Someone retiring at 65 and expecting to live to 90 does not have one holding period. They have an entire spectrum of them:

• Next year's living expenses → one-year horizon → cash
• Living expenses five years out → five-year horizon → short-term bonds
• Living expenses twenty years out → twenty-year horizon → equities are appropriate, because on that scale inflation is the primary enemy

This is why an 85-year-old's portfolio should still contain equities. The holding period on that portion of the money is not zero; it is ten years.

The same point appears in Chapter 3 of Retirement Decumulation Mechanics and Chapter 3 of Winning the Loser's Game — "the time edge belongs to the money, not the person."

This chapter supplies the data behind that sentence.

Chapter 5 takes on the variable most easily overlooked in this framework and most practically consequential right now: the valuation level at which you buy.