Sheet: bad decade in history

All the research sheets

Translated from the Spanish original. Where they differ, the Spanish one holds.

• App: Horalis Growth · Reviewed: 3 October 2026
• Code (in projects/wealth-lens/): src/lib/decade.ts (historicalDecade, decadeYears, FAMOUS_DECADE, DECADE_YEARS), src/lib/what-if.ts (badStartHead), src/lib/findings.ts (sequenceFinding), src/lib/calculator.ts (valueAt with head)

The question it answers

"What if I start just before a bad run?" It is the "What if…?" "If your first 10 years were like 2000–2009" and the bad first decade finding: how much less there would be at the end if the first years were like a real bad decade.

The formula, in plain words

The plan's first 10 years (or all of them, if fewer) grow year by year as that asset really grew in a particular decade; after that, the money grows at its average again. Which decade:

• 2000–2009, the one almost every stock market lost, if for that asset it was worse than its 1988–2022 average;
• if it was not (for it, it was a good decade), its worst decade within 1988–2022, named with its years.

Exact: with X what there is today and Y the contribution, for each year k of the decade, balance_k = value of (balance_{k−1}, Y, r_k, 1 year) with the growth formula and r_k that year's real return; after that, the normal formula follows with the average g. What the decade is made of:

• an asset: its own years;
• a mix: the years of its parts, with their weights (rebalanced every January or left to run, as the mix says);
• own growth: the years of US shares;
• written growth or swing: the same years shifted to that growth and stretched to that swing, so the decade keeps its shape: log(1 + r') = log(1 + g') + (σ'/σ)·(log(1 + r) − log(1 + g)).
• Without swings (savings, own growth with a swing of 0) there is no bad decade.

Assumptions

• A real decade is a clear way to imagine a bad run: "like 2000–2009" is easier to understand than "the 10th percentile".
• After the decade, everything returns to the average, with no rebound and no penalty.

Sources (dated)

• The yearly 1988–2022 data of growth.
• Sequence-of-returns risk: what matters is not only the average but the order of the years; see Bengen (1994) in the withdrawal rate.

Validation (tests)

The contribution needed for a distant goal uses the same active decade. That decade's path is calculated from zero with €1/month; by linearity, what comes from the starting capital is separated from what comes from the contributions. For the term T, if A is the value of the capital without contributions and B the value of contributing €1/month, the amount needed is max(0, (goal − A) / B). If B is zero and the goal is not covered, there is no finite contribution that reaches it. Without a decade, the original monthly formula is kept. This corrects the scenario's inverse; it does not change its growth or select other historical years.

src/lib/what-if.test.ts checks that applying the proposed contribution, with the bad decade still selected, reaches the goal exactly at 30 years, both from zero contributions and from an existing contribution.

• src/lib/decade.test.ts (8): 2000–2009 for US shares; the worst decade, with its years, where 2000–2009 was good (bonds 2013–2022, gold 1988–1997); the plan's years if fewer than ten; without swings there is no decade; an asset's real years, one after another; one's own, with US shares' years shifted to the written growth; a larger swing, a worse decade; the parts of a mix in the same year, with their weights.
• src/lib/what-if.test.ts (18, "A bad first decade"): the "What if…?" equal to applying it, its effects on goals and countries, and that it does not exist without ups and downs.
• src/lib/findings.test.ts: the finding uses the same decade and gives the same figure as the "What if…?".

Limits

• It is a scenario, not a probability: it says how much, not how likely.
• Only one decade; a longer run (Japan since 1990) is outside the data.

What is debatable

1. Choosing the worst decade is cherry-picking. For bonds and gold the worst of 35 years is searched; for shares, 2000–2009, although 1999–2008 was somewhat worse (S&P 500: −4.41 % a year). It is a decision for clarity ("2000–2009" is recognisable), but it is not "the worst" for everyone. Cost: say "one of the worst" or always use the worst.
2. Returning to the average after the crisis. After a bad decade, valuations are usually low and the later returns better (2010–2019 was very good for shares). Simply returning to the average may exaggerate the long-term damage; there is no guarantee of a rebound either. Cost: nothing immediate; say it in How it works.
3. Stretching to another swing. The transformation keeps the decade's shape, but applying US shares' sequence to any written growth is an analogy, not a fact.

History

• 2026-10-03: first sheet (no change to the calculation).
• 2026-10-05: the contribution needed for goals out of reach includes the same historical decade as the time to reach them. Regression test: €1,000 to start and a goal of a million; before, the proposed figure left about €790,322 in 30 years, now it reaches the goal under that scenario.
• 2026-10-09 (phase A2): no World, Nasdaq-100, single shares or My portfolio. Own growth uses US shares' years (before, the World's). Gold, from the World Bank's Pink Sheet: the same decade, 1988–1997.