Compounding over decades: why the last ten years do the work

The shape nobody expects

Everyone has seen the compounding chart: a line that crawls along the bottom, then bends upward, then appears to go nearly vertical. Most people read it as encouragement. It is more useful read as a warning about where the value sits.

Here is the arithmetic in plain terms. If a balance grows at a constant rate, each year's growth is proportional to the balance at the start of that year. Early on, the balance is small, so the growth is small in absolute terms even though the percentage is identical. Late on, the balance is large, so the same percentage produces very large absolute growth. The percentage never changes. The absolute contribution of each year changes enormously.

Suppose a balance grows at 7% a year. In year one on a balance of 10,000, growth is 700. In a year when the balance has reached 500,000, growth at the same 7% is 35,000 — fifty times larger, from the identical rate. Nothing improved. The base got bigger.

This is why the final stretch of a long accumulation produces such a disproportionate share of the total. Not because returns get better, but because the balance those returns act on is at its largest exactly then.

The corollary is uncomfortable and worth stating plainly. The final decade produces the most growth, but you cannot get that decade's growth without having built the balance first. The last ten years do the work; the first twenty years decide how much work there is to do.

Decomposing where the money comes from

A useful exercise is to split a completed accumulation into three parts:

1. **Money you put in.** The sum of contributions, undiscounted.
2. **Growth on contributions.** What those contributions earned directly.
3. **Growth on growth.** What the earnings themselves earned.

The third bucket is compounding proper. In short horizons it is nearly invisible. Over long horizons it becomes the largest of the three.

Take a hypothetical. Suppose you contribute 2,000 a month for 30 years and the balance grows at a steady 7% a year. Contributions total 720,000. The ending balance under those assumptions is roughly 2.44 million. So around 70% of the final figure never came out of your pocket, and a substantial slice of that came from growth compounding on earlier growth rather than on your deposits.

Now shorten the horizon to 20 years with the same monthly contribution. Contributions total 480,000. The ending balance is roughly 1.04 million. Contributions are two-thirds as large; the balance is well under half. You did not lose a third of the effort. You lost the decade in which the largest absolute growth would have occurred.

Reverse it. Keep the 30 years but contribute for only the first 10, then stop and leave it. Contributions total 240,000 — a third of the original — and the balance after 30 years is roughly 700,000. Compare it to someone who contributes nothing for 20 years and then contributes 2,000 a month for the final 10. Their contributions total 240,000 as well, and their balance is roughly 346,000. Same money in, half the result, and the only difference is where in the timeline it sat.

These are illustrative figures using a fixed rate, not a projection of any real product. But the ordering they demonstrate is robust: **when money arrives matters more than how much arrives, once the horizon is long.**

Why "just start" is more than a slogan

The decomposition above explains the advice, but the mechanism is worth spelling out because the slogan is often misapplied.

Starting early is valuable specifically because it buys the money more compounding periods. A contribution made in year one is exposed to growth for the entire horizon. A contribution made in year 25 of a 30-year plan is exposed for five. If the rate is 7%, the year-one contribution roughly multiplies 7.6-fold over 30 years; the year-25 contribution multiplies about 1.4-fold. Same currency, very different jobs.

This is why a delay is expensive in a way that a small contribution is not. Suppose you can only afford 300 a month today, and expect to afford 2,000 in five years. The instinct is to wait until you can "do it properly". The arithmetic says the 300 that starts now is doing a job the 2,000 later cannot do — it is buying compounding periods that expire and never come back.

But apply this carefully. Two honest counterweights:

• **Do not start a long-term accumulation on top of expensive short-term debt.** If a debt is charging more than a plausible growth assumption, repaying it is the higher-certainty use of the same money. The arithmetic of compounding works in both directions, and it works fastest against you when the rate is high and the balance is one you owe.
• **Do not start without a liquidity buffer.** Long-term money that has to be liquidated in a bad year to cover a car repair does not compound; it gets interrupted. The buffer is not an alternative to investing, it is what allows the investing to be left alone.

The three enemies of the curve

The textbook curve assumes three things that rarely hold: a constant rate, no charges, and no interruptions. Each assumption failing has a different signature.

Charges compound too

An annual charge is subtracted from the same base that growth is applied to, every year, for the entire horizon. That makes it a compounding drag, not a one-off fee.

Take the 30-year, 2,000-a-month example at 7%. Now assume a 1% annual charge, so the net rate is 6%. The ending balance falls from roughly 2.44 million to roughly 2.01 million. The difference — around 430,000 — is more than half of everything you contributed, produced by one percentage point a year.

The reason it is so large is that a charge does not merely remove 1% of one year's growth. It removes 1% of the balance, permanently, which then never compounds again for the remaining years. International pension analysis consistently identifies charges, along with contribution continuity and time in the plan, as principal determinants of final balances in defined contribution arrangements.Source^source

This does not mean the cheapest option is always right, and it does not mean charges are illegitimate. It means a charge should be compared against the horizon it will run for, not against a single year's return.

Interruptions cost more than they look

A pause in contributions is not simply "the contributions you missed". It is those contributions plus everything they would have earned for the remaining horizon. Pausing 2,000 a month for two years in year five of a 30-year plan removes 48,000 of contributions and, at 7%, roughly 250,000 of ending balance.

Life will produce interruptions. The practical response is not guilt but design: keep the automatic contribution small enough that it survives a bad year, and treat increases as something you add on top when things are good rather than something you must defend when they are not. A contribution you never have to cancel outperforms a larger one you cancel twice.

The rate is not a constant

This is the biggest gap between the model and reality. The chart assumes a smooth rate. Real returns arrive unevenly — long stretches below the average, sharp recoveries, occasional deep drawdowns. Long-run analysis of interest rates and asset returns shows substantial variation across decades.Source^source

Two consequences follow.

First, **a projected figure is not a promise**. Any number produced by a fixed-rate model is a scenario. Treat it as a way to compare decisions ("does starting five years earlier matter?") rather than as a forecast of a balance.

Second, **the order of returns matters when you are drawing down**, and much less when you are contributing. While you are contributing, a bad early stretch buys more units cheaply; the average return over the whole period dominates. When you are withdrawing, a bad early stretch forces you to sell more to fund the same income, which permanently reduces the base. This asymmetry — sequence risk — is why the transition from accumulating to spending deserves its own planning, and why the "last ten years do the work" framing gets dangerous if you assume the last ten years will be good ones.

Real returns, not nominal ones

A projection in nominal currency flatters itself. If the balance grows 7% a year and prices rise 3% a year, purchasing power grows around 3.9% a year, not 7%. Over 30 years that difference is not cosmetic: the same 2,000 a month at a 3.9% real rate accumulates to roughly 1.32 million in today's purchasing power, against 2.44 million nominal.

Both numbers are correct. They answer different questions. The nominal figure tells you what the statement will say; the real figure tells you what it will buy. Inflation erodes the purchasing power of a nominal balance over time, which is the entire reason long-horizon plans should be read in real terms.Source^source

A practical habit: run every long projection twice, once nominal and once with inflation subtracted from the growth rate, and make decisions on the second one. Also index the contribution. A contribution fixed in nominal terms for 30 years is a shrinking contribution in real terms — increasing it with your income roughly preserves the plan's real ambition without requiring a decision each year.

What compounding does not do

Because the curve is so often used persuasively, it is worth being explicit about its limits.

• **It does not make returns certain.** Compounding is a property of growth over time. It says nothing about whether growth will occur, at what rate, or in what order.
• **It does not recover from permanent loss.** A balance that falls by half needs a 100% gain to return to where it was. Compounding amplifies whatever base survives; it does not restore a base that is gone.
• **It does not reward complexity.** Nothing in the arithmetic requires clever selection. The variables that show up in the formula are the amount, the rate, the cost and the time — three of which you can influence directly and one of which you mostly cannot.
• **It is not a reason to take more risk than you can hold.** A higher assumed rate improves a spreadsheet instantly and a real balance only if you actually stay invested through the periods that produce it. The rate you can hold through a bad decade beats the higher rate you abandon in year three.
• **It does not run indefinitely.** Human horizons end, and so does the accumulation phase. The curve steepens right up to the point you start withdrawing, at which point a different, less flattering arithmetic takes over.

Using the shape to make decisions

If you accept the back-loaded shape, a handful of practical implications follow. None of them are exotic.

1. **Protect the horizon first.** Anything that shortens the time — a late start, a long pause, an early withdrawal — costs more than a proportionally similar reduction in contribution. Horizon is the variable with the most leverage and the one that is genuinely irreversible.
2. **Treat costs as a horizon-length decision.** A percentage sounds small; multiply it mentally by the number of years it will run.
3. **Automate the contribution, and size it to survive.** Continuity beats magnitude. The plan that never gets cancelled quietly outperforms the ambitious one that does.
4. **Index contributions to income.** This keeps the plan real rather than nominal without repeated decisions.
5. **Plan the transition separately.** The last decade of accumulation and the first decade of drawdown are the two periods where a bad sequence does lasting damage. Whatever you do about that — reducing volatility as the date approaches, holding a cash buffer to avoid forced selling, staging the transition over years — decide it in advance rather than during a drawdown.
6. **Do not confuse the model with the world.** Use fixed-rate projections to compare choices, and use ranges rather than single numbers when you are deciding whether a plan is adequate.

The honest summary

The last ten years produce the most growth because that is when the balance is largest, not because anything gets better. That fact is simultaneously an argument for starting early, an argument for not interrupting, an argument for caring about a one-percent charge, and an argument for planning carefully around the moment the accumulation stops.

It is not an argument that the outcome is assured. The curve is arithmetic applied to an assumption, and the assumption is the fragile part. What you control is the amount, the continuity, the cost and the time. Those four are enough to make a serious difference, and they are all decided long before the steep part of the curve arrives.

Source^source: Bank for International Settlements.

Source^source: OECD.

Source^source: International Monetary Fund.