Assertive when conditions reward risk. Defensive when the cycle becomes fragile.
This is not a promise to predict the next crash. It is a disciplined way to participate in long bull markets without assuming that every late-stage excess will resolve gently.
Growth wins over long periods. The journey is the price.
From 1926 to 2025, shares produced higher returns than bonds or cash, and portfolios holding more shares finished far ahead. That conclusion is unsurprising: investors are normally rewarded for accepting more uncertainty.
The important distinction is between an arithmetic average and a compounded return. An arithmetic average describes the typical year. Compounding describes what actually happened to a dollar through every gain and loss. A 50% fall requires a 100% gain to recover, so volatility pulls the compounded result below the simple average — for shares, by well over a full percentage point per year.
| Asset class | Average yearly return (arithmetic) | Compounded annual return | Year-to-year volatility | Worst calendar year |
|---|---|---|---|---|
| Australian shares | 11.50% | 9.94% | 18.5% | −40.4% (2008) |
| US shares (S&P 500, in USD) | 12.35% | 10.49% | 19.5% | −43.3% (1931) |
| Australian bonds | 6.12% | 5.94% | 6.3% | −10.1% (1973) |
| Australian cash | 5.68% | 5.60% | 4.1% | 0.0% (2021) |
Basis: 100 audited calendar-year returns, 1926–2025, nominal (not adjusted for inflation), income reinvested where the source series is a total-return series. Long-run histories require proxies: Australian data before 1970 comes from academic research series, Australian bonds before 1970 are a yield-based proxy rather than a true total-return index, and US figures are in US dollars with no currency conversion. Every year’s figure, source and status label is in the quant appendix. Past performance is not a reliable indicator of future performance.
Different amounts of growth risk, not different beliefs about the future.
All three model portfolios hold the same four asset classes. The difference is how much short-term loss each investor is prepared to tolerate in exchange for higher long-term return.
Defensive
50% growth assets
20% Australian shares · 30% US shares · 47% bonds · 3% cash
100-year compounded return: 8.7% p.a.
$1 grew to roughly $4,100
Worst calendar year: −14.3% (1973)
Balanced
70% growth assets
28% Australian shares · 42% US shares · 27% bonds · 3% cash
100-year compounded return: 9.6% p.a.
$1 grew to roughly $9,300
Worst calendar year: −22.6% (2008)
Growth
85% growth assets
34% Australian shares · 51% US shares · 12% bonds · 3% cash
100-year compounded return: 10.2% p.a.
$1 grew to roughly $15,900
Worst calendar year: −30.6% (2008)
These are hypothetical model portfolios rebalanced to target weights each year, calculated from the audited annual series described above. No such portfolio existed or could have been invested in over this period; index investing itself only became practical in recent decades. Figures are before fees, transaction costs and tax, with US share returns in US dollars (currency movements between the Australian and US dollar, which were at times very large, are not reflected). Past performance is not a reliable indicator of future performance.
Participate first. Protect later.
Early and mid-stage bull markets are usually periods when earnings, liquidity and confidence are still supporting risk-taking. Our default posture is therefore assertive: hold the Growth profile and allow compounding to work.
Late-stage markets are different. Valuations can detach from cash flows, leverage can rise, market breadth can narrow and good news can begin producing less upside. None of those indicators announces a crash date. Together, they may justify accepting less market risk.
The greatest practical danger is moving too early. A defensive portfolio can protect capital during a crash, but it can also miss a powerful final advance. The worst outcome is not simply suffering a drawdown; it is abandoning a sound long-term portfolio for a crash that never arrives.
Stay assertive
Use the Growth profile while the investment case remains supported. Do not manufacture caution merely because markets have risen.
Reduce risk deliberately
Move toward Defensive only when the agreed evidence accumulates. The live decision may be staged; the historical study below uses a single switch so the arithmetic is understandable.
Rebuild growth exposure
The historical study returns to Growth three months after the trough. In practice, the committee would require improving liquidity, stabilising credit and functioning markets — not clairvoyance.
What being early or late would have cost.
For each of five major US market crashes, we compare the three static profiles against a portfolio that switched from Growth to Defensive 12, 24 or 36 months before the market peak — using peak and trough dates that are only knowable after the event.
- The rows marked Hindsight assume the switch dates were chosen with perfect knowledge of when each crash began and ended. No investor, including us, could have timed these switches. They measure the economics of being early or late, not a result anyone earned or a result we expect to achieve.
- Dollar figures are from a modelled monthly dataset. Australian monthly returns are spread evenly within each year, US crash-window returns are set to match the reported peak-to-trough price fall, and the headline decline shown for each crash is the fall in the US share market, not in these portfolios. Real month-by-month outcomes would have differed.
- Each window runs from 60 months before the peak month to 60 months after the trough month, except the Great Depression window, which starts in January 1926 (44 months before the peak) because that is where the dataset begins.
- All figures are hypothetical, before fees, transaction costs and tax, ignore currency movements, and are rounded to the nearest $100. Past performance is not a reliable indicator of future performance.
September 1929 to July 1932
| Strategy | Starting value | Ending value | Return |
|---|---|---|---|
| Switch 12 months before peakHindsight — not achievable | $100,000 | $300,800 | +201% |
| Switch 24 months before peakHindsight — not achievable | $100,000 | $284,500 | +184% |
| Switch 36 months before peakHindsight — not achievable | $100,000 | $270,200 | +170% |
| Growth (static) | $100,000 | $220,400 | +120% |
| Balanced (static) | $100,000 | $212,900 | +113% |
| Defensive (static) | $100,000 | $201,700 | +102% |
Market fall measured on the S&P index and its predecessor (price only, in USD). Context: Federal Reserve History. Calculation and data provenance: quant appendix.
January 1973 to October 1974
| Strategy | Starting value | Ending value | Return |
|---|---|---|---|
| Switch 36 months before peakHindsight — not achievable | $100,000 | $210,000 | +110% |
| Switch 24 months before peakHindsight — not achievable | $100,000 | $205,400 | +105% |
| Switch 12 months before peakHindsight — not achievable | $100,000 | $201,300 | +101% |
| Defensive (static) | $100,000 | $180,600 | +81% |
| Balanced (static) | $100,000 | $176,900 | +77% |
| Growth (static) | $100,000 | $173,800 | +74% |
Market fall measured on the S&P 500 (price only, in USD). Context: Federal Reserve History. Calculation and data provenance: quant appendix.
August 1987 to December 1987
| Strategy | Starting value | Ending value | Return |
|---|---|---|---|
| Switch 12 months before peakHindsight — not achievable | $100,000 | $524,700 | +425% |
| Switch 24 months before peakHindsight — not achievable | $100,000 | $494,400 | +394% |
| Growth (static) | $100,000 | $484,000 | +384% |
| Balanced (static) | $100,000 | $479,400 | +379% |
| Defensive (static) | $100,000 | $472,100 | +372% |
| Switch 36 months before peakHindsight — not achievable | $100,000 | $467,300 | +367% |
Market fall measured on the S&P 500 (price only, in USD). Context: Federal Reserve History. Calculation and data provenance: quant appendix.
March 2000 to October 2002
| Strategy | Starting value | Ending value | Return |
|---|---|---|---|
| Switch 12 months before peakHindsight — not achievable | $100,000 | $413,100 | +313% |
| Switch 24 months before peakHindsight — not achievable | $100,000 | $399,800 | +300% |
| Switch 36 months before peakHindsight — not achievable | $100,000 | $387,300 | +287% |
| Growth (static) | $100,000 | $352,000 | +252% |
| Balanced (static) | $100,000 | $331,800 | +232% |
| Defensive (static) | $100,000 | $306,100 | +206% |
Market fall measured on the S&P 500 (price only, in USD). Context: FRBSF Economic Letter. Calculation and data provenance: quant appendix.
October 2007 to March 2009
| Strategy | Starting value | Ending value | Return |
|---|---|---|---|
| Switch 12 months before peakHindsight — not achievable | $100,000 | $289,500 | +189% |
| Switch 24 months before peakHindsight — not achievable | $100,000 | $279,200 | +179% |
| Switch 36 months before peakHindsight — not achievable | $100,000 | $270,600 | +171% |
| Growth (static) | $100,000 | $239,900 | +140% |
| Balanced (static) | $100,000 | $232,800 | +133% |
| Defensive (static) | $100,000 | $222,900 | +123% |
Market fall measured on the S&P 500 (price only, in USD). Context: Federal Reserve History. Calculation and data provenance: quant appendix.
The strategy demands more temperament than prediction.
Patience
Late-stage conditions can persist. The investor must tolerate underperformance while a more aggressive portfolio continues rising.
Governance
Triggers, authority, ranges and review dates must be written down. Otherwise “dynamic” becomes a polite word for improvisation.
Humility
The signal may be early, late or wrong. Position changes should be proportionate to the evidence rather than framed as certainty.
Key risks of a dynamic approach: reducing risk too early and missing further gains; acting on false signals; transaction costs and tax triggered by switching; delay between decision and implementation; markets recovering faster than the re-entry rule allows; currency movements on international assets; and abandoning the process under pressure. A dynamic approach can produce worse outcomes than simply holding a static portfolio.
Nothing important is hidden.
The separate appendix includes the 100 annual observations, all 1,200 monthly observations, every source and status field, the crash overrides, proxy definitions, calculation notes and downloadable workbooks.