The short version
- We ran this site's retirement withdrawal calculator on eight fund portfolios, at four withdrawal rates, over 30 years, 2,000 trials each. The engine resamples real monthly fund returns in 24-month blocks, so real crashes and recoveries stay intact, and it draws the same months for every fund in a mix, so real correlations survive.
- At 4%, a fixed-dollar withdrawal survived 30 years in 98.5% of paths for VTI alone and 99.8% for a 60/40 VTI/AGG mix. The exceptions are QQQ alone (86.0%, because its history starts in 1999 and includes the dot-com collapse) and AGG alone (93.2%).
- The cliff is between 5% and 7%. Bonds alone collapse first: AGG-only success falls from 93.2% at 4% to 56.0% at 5% and 9.2% at 6%. Stock-heavy mixes hold 91-92% at 6% and 75-86% at 7%.
- One caveat carries the whole essay: these are nominal returns and a fixed dollar withdrawal. The classic 4% rule raises the withdrawal with inflation every year; this test does not. At 2.5% inflation a fixed sum buys about half as much after 30 years (0.975³⁰ ≈ 0.47). For a stricter read, look at the 5% and 6% columns, not the 4% one.
- Nothing here is a forecast. It is what these specific histories, resampled, would have done.
What was tested
Portfolios were chosen for history length: every fund below has monthly returns back to at least 2003, so the block bootstrap can draw the 2008 crash, and QQQ and IWM reach back to 1999-2000, so it can draw the dot-com collapse too. Each mix is rebalanced to target weights monthly. Starting balance $1,000,000; a fixed annual withdrawal equal to the stated rate of the starting balance; 30-year horizon; 2,000 trials; 24-month blocks of contiguous calendar history, never wrapped.
| Portfolio | History used | Months |
|---|---|---|
| VTI 100% | 2001-07 → 2026-09 | 303 |
| QQQ 100% | 1999-04 → 2026-09 | 330 |
| IWM 100% | 2000-06 → 2026-09 | 316 |
| AGG 100% | 2003-10 → 2026-09 | 276 |
| VTI 60 / AGG 40 | 2003-10 → 2026-09 (AGG's start) | 276 |
| VTI 60 / TLT 40 | 2002-08 → 2026-09 (TLT's start) | 290 |
| VTI 80 / AGG 20 | 2003-10 → 2026-09 | 276 |
| VTI 40 / VEA 15 / VWO 5 / AGG 40 | 2007-08 → 2026-09 (VEA's start) | 230 |
A mix can only use the months every fund in it has, so the mixes containing AGG do not see 2000-02 and the four-fund mix sees 2008 but not 2001. The single-fund rows with the longest histories are therefore the harshest tests.
Results: share of 30-year paths that never ran out
| Portfolio | 4% | 5% | 6% | 7% |
|---|---|---|---|---|
| VTI 100% | 98.5% | 96.6% | 92.3% | 85.5% |
| QQQ 100% | 86.0% | 78.0% | 73.5% | 69.5% |
| IWM 100% | 94.6% | 88.4% | 79.3% | 67.2% |
| AGG 100% | 93.2% | 56.0% | 9.2% | 0.1% |
| VTI 60 / AGG 40 | 99.8% | 97.6% | 91.9% | 80.0% |
| VTI 60 / TLT 40 | 99.6% | 98.4% | 92.2% | 82.6% |
| VTI 80 / AGG 20 | 99.7% | 97.3% | 91.5% | 85.2% |
| VTI 40 / VEA 15 / VWO 5 / AGG 40 | 99.9% | 98.1% | 91.6% | 75.1% |
The tenth-percentile ending balance at 4% — the outcome one unlucky retiree in ten did worse than — was $3.03M for VTI alone, $2.51M for 80/20, $1.93M for 60/40 with AGG, $2.31M for 60/40 with TLT, $1.82M for the four-fund mix, $0.55M for IWM alone, $0.07M for AGG alone, and $0 for QQQ alone, whose bad decade is the one it started with.
How to read this
Why 4% looks so safe here. Two reasons, and only one of them is good news. The good one: over these histories, diversified US equity compounded fast enough that a 4% nominal draw was rarely threatened even by 2008. The other: the withdrawal is not inflation-adjusted. A retiree who took $40,000 in year one and $40,000 in year thirty would have lived on roughly half the goods by the end. The classic rule's inflation step-ups are what make 4% a genuine constraint, and they are not in this engine. Read the 5% column as the closer analogue of a real-terms 4%, and the 6% column as a stress test.
Why bonds alone fail so fast. AGG's history averages a low single-digit nominal return and includes the 2022 drawdown. A 5% draw from a portfolio earning less than 5% is arithmetic, not bad luck; the 56% at 5% is the resampler occasionally landing enough of the better years in a row.
Why QQQ alone is the worst row at every rate. Not because the Nasdaq-100 compounded poorly over the period — 86% of paths finish above zero at 4% — but because sequence matters. The engine can, and in one path in seven does, put 2000-02 near the start, when the portfolio is largest and the withdrawal bites hardest. IWM's 94.6% and VTI's 98.5% at 4% are the same lesson at lower volatility.
Why the mixes beat VTI alone at 4% but not at 7%. At 4% the bonds' job is to blunt an early crash, and they do: every mix is above 99.6%. At 7% the withdrawal is large enough that return, not smoothness, decides survival, and the 40% bond sleeve costs more than it protects.
What this can and cannot tell you
- Nominal, fixed withdrawal. Stated above; it is the single largest limitation and the reason the 4% column flatters.
- History is short and specific. The longest series here is 27 years. It contains 2000-02, 2008 and 2022, and no 1970s inflation, no 1930s. A resampler cannot produce a decade it has never seen.
- Returns are Yahoo Finance monthly total returns, dividends reinvested, before tax and before any advisory fee. The fund expense ratios are already inside the returns.
- Correlation is preserved, regime is not. Drawing the same months for every fund keeps 2008 as 2008; every historical block is equally likely to be drawn, so recent years carry no more weight than older ones.
- Rebalancing is monthly to target. Real retirees rebalance less often, or not at all; that changes the paths, and we have not tested by how much.
- The engine, its method and its limits are documented on the calculator page. Run it on your own funds and weights; the 564 funds with return history are listed there.
Related: Monte Carlo vs the 4% rule on why a resampled test and a formula give different answers, and the Monte Carlo measure page for the definition.