Gold's positive return over a crisis does not mean its holder avoided a loss along the way. From September 2007 to September 2010, spot gold in euros gained 82.8%, yet suffered a 12.5% month-end drawdown inside that period. Whether it helped depends on the entry date, the spending date and the alternative being compared.
The relevant comparison is compounded wealth after ownership costs and the return forgone on alternatives. Gold’s volatility or correlation does not independently establish a growth benefit.
Using the asset-class crash study, we separate three questions: how far gold fell, how long it took to regain its own peak, and what else the same starting capital could have earned.
Key takeaways
- Gold in euros gained 82.8% over September 2007–September 2010 but still had a 12.5% month-end drawdown within that window.
- A positive crisis-window return and the time needed to regain an earlier peak answer different questions.
- Compounded wealth comparisons also depend on ownership costs, inflation and returns forgone on alternatives.
Four positive windows, four different experiences
These are nominal euro spot-price changes, without storage, fund charges, dealing spreads or tax. They are neither USD gold returns nor returns on a particular gold product. Peaks and troughs use month-end observations.
| Exact window | Gold's window return | Largest fall within window | Required gain from that trough |
|---|
| 29 February 2000–28 February 2003 | +6.6% | −10.5% | +11.7% |
| 30 September 2007–30 September 2010 | +82.8% | −12.5% | +14.3% |
| 31 January 2020–31 January 2023 | +23.9% | −13.3% | +15.3% |
| 31 December 2021–31 December 2024 | +56.4% | −8.8% | +9.6% |
Source: reproducible crisis calculations, derived from the parent's rounded observations. The recovery hurdle is 1 / (1 − loss) − 1, rather than simply the size of the loss.
The 2020–2023 window's deepest gold decline began in July 2020 and ended in February 2021. It was not the loss during the initial March 2020 equity panic. Fixed windows can contain more than one market episode.
Recovery depends on which clock starts first
In the 2008 window, gold's February 2008 peak was regained in January 2009: 11 months peak to recovery, comprising two months to the April trough and nine months back up.
After the July 2020 peak, recovery came in February 2022: 19 months in total, including 12 months after the trough. After the April 2022 peak, recovery came in March 2023: 11 months, including five after the October 2022 trough.
The dot-com window does not report a recovery of gold's May 2002 peak by its February 2003 endpoint. A blank recovery field is not evidence of an immediate rebound or a date outside the observed window.
Those recent experiences sit beside a much longer record. The supplied whole-history series has a 54.0% decline from January 1983 to July 1999, followed by recovery in September 2007. That is 198 months descending and 98 months recovering, or 296 months peak to recovery. A single long drawdown can span multiple economic regimes; it is not evidence of one continuous crisis. Euro values before 1999 are provider reconstructions.
Opportunity cost changes with the comparator
For an equal €10,000 starting amount in February 2000, gold ended the three-year window at €10,660. The compounded ECB EONIA cash series ended at €11,260. Gold gained, but finished €600 behind cash, or 5.3% lower in relative wealth.
The comparison reverses in the 2008 window: gold ended at €18,280, versus €10,620 in the cash series. That cash benchmark is an overnight-rate proxy, not an insured deposit or a particular money-market fund.
Against US equities in euros, with dividends reinvested, gold's 2008-window terminal value exceeded the equity value by €9,970 per €10,000 initially invested. In January 2020–January 2023, it finished €1,150 behind: €12,390 versus €13,540. These endpoint comparisons share dates and initial capital. They do not measure a blended portfolio's drawdown or the benefit from rebalancing.
There is consequently no single historical “cost of holding gold”. The cost depends on what was forgone and when. Four selected, partly overlapping crisis windows cannot establish gold's expected return or an optimal allocation.
Nominal break-even can still leave an economic shortfall
The 296-month recovery offers a useful hypothetical cost calculation. If an investment exactly tracked the spot price before an annual 0.5% ownership cost, €10,000 at the old peak would be worth about €8,837 when spot merely returned to that peak: 10,000 × 0.995^(296/12). That is an 11.6% cost drag. It is not the realised cost of any historical product.
Separately, assuming constant 2% annual inflation, preserving €10,000's purchasing power over 296 months would require €16,298, a 63.0% nominal gain. This is a sensitivity illustration, not measured inflation for 1983–2007. Both assumptions and formulas are in the long-recovery calculation.
Gold can show resilience in one sample and require considerable patience in another. The recovery guide explains the two clocks across asset classes; the cash and withdrawals guide addresses money that cannot wait for either clock.
FAQ
Did gold avoid losses in the 2008 financial crisis?
No. Spot gold in euros gained 82.8% from September 2007 to September 2010, but fell 12.5% from its February 2008 month-end peak to April 2008. That peak was regained in January 2009.
How long did gold take to recover its 1983 euro peak?
In the supplied month-end spot-price history, gold regained its January 1983 peak in September 2007: 296 months. The trough was July 1999, so the final recovery from the trough took 98 months. Pre-1999 euro values are reconstructed by the provider.
Did gold beat cash in every crisis window?
No. From February 2000 to February 2003, gold in euros returned 6.6% and the euro overnight cash series returned 12.6%. Equal €10,000 starting amounts therefore finished at €10,660 and €11,260 before costs and tax.
Do gold spot-price recoveries include storage costs or inflation?
No. A nominal spot-price recovery excludes storage, fund charges, dealing spreads and tax, and does not show restored purchasing power. Constant-cost and inflation examples in this article are hypothetical, not historical estimates.
