Most retirement advice assumes that as you get older, you should hold more bonds. Bonds are less volatile than stocks, so a mix of the two is supposed to make your savings last more reliably.
We wanted to see whether history agrees, so we asked Ignigrade. Ignigrade is a historical backtesting tool for financial planning: you describe your finances and your plan in plain English, and it runs that plan through every stretch of U.S. market history since 1928, showing you how it would have turned out in each one.
The short version: a little in bonds makes a retirement slightly safer, and it costs a lot of upside. Over both 30 and 50 years, holding 20% in bonds lowered the odds of running out of money by a hair compared with 100% stocks, but it left far less money at the end. Holding half in bonds made things worse on both counts.
Here's the scenario we typed in:
I'm 60, single, and retired. I have $1,000,000 in a Roth IRA and plan to withdraw $40,000 a year, adjusted for inflation. Compare the performance of a 100% stock portfolio, an 80% stock / 20% bond portfolio, and a 50% stock / 50% bond portfolio, over 30 years and over 50 years.
That's the classic "4% rule": withdraw 4% of your starting balance in the first year, then raise the amount with inflation every year after. We used a Roth IRA on purpose, because Roth withdrawals aren't taxed, which keeps taxes out of the comparison (more on that below). Stocks are the S&P 500 with dividends reinvested, bonds are U.S. Treasury bonds, and the portfolio is rebalanced back to its target every year.
Ignigrade ran that plan starting in every year from 1928 on that has a full 30 (or 50) years of history after it. A plan "fails" if the money runs out before the end, and the probability of ruin is the share of starting years that failed.
Over 30 years, which is roughly a traditional retirement starting at 60 to 65, the 80% stock/20% bond portfolio had the lowest probability of ruin of the three: 1.4%, against 4.3% for 100% stocks. Going further, to half bonds, brought the probability of ruin up to 5.8%, because the portfolio doesn't grow fast enough to support the 4% withdrawal rate as reliably.
Probability of ruin over 30 years:
| 100% stocks | 80% stocks / 20% bonds | 50% stocks / 50% bonds |
|---|---|---|
| 4.3% | 1.4% | 5.8% |
The 30-year comparison in Ignigrade. The chart shows each portfolio's average wealth (solid line) and the range across every starting year (shaded), in today's dollars.
These 30-year results line up with the famous Trinity study behind the 4% retirement spending rule. In 1998, three Trinity University professors, Philip Cooley, Carl Hubbard, and Daniel Walz, backtested withdrawal rates and found that a 75% stock/25% bond portfolio, close to our 80/20, does best with the 4% withdrawal rate assuming the 30 year retirement horizon.¹
Over 50 years, the kind of horizon an early retiree has to plan for, the ranking is the same: 80% stocks again had the lowest probability of ruin, 12.2% against 14.3% for 100% stocks. But the 4% rule itself is much less safe over 50 years: even the safest allocation reached 12.2%, up from 1.4% over 30 years. If your retirement may last 50 years, lowering your withdrawal rate to 3.5% makes a big difference: no 50-year retirement in our history ran out of money with an 80% stock portfolio.
Probability of ruin over 50 years:
| 100% stocks | 80% stocks / 20% bonds | 50% stocks / 50% bonds |
|---|---|---|
| 14.3% | 12.2% | 36.7% |
The same comparison over 50 years.
To get a complete picture, we ran every stock allocation from 0% to 100% in 5% steps for the two retirement horizons. The safest allocation for both time horizons was around 80%-85%. Below about 60% stocks, the probability of ruin climbs steeply.
Probability of ruin at a 4% withdrawal rate, for every stock allocation from 0% to 100% in 5% steps. A modest bond allocation lowers it slightly; past that, more bonds raise it, and more so over 50 years. Small bumps along the curves are a single starting year each.
20% in bonds bought a little safety. The worst starting year for 100% stocks was 1929, straight into the Great Depression. It ran out of money in 22 years. The 80% stock/20% bond portfolio weathered the Depression and survived the 30-year horizon, but over 50 years it ran out in year 41.
It cost a lot of upside. The average ending balance in today's dollars was about $2.9 million at 80% stocks versus $4.0 million at 100% over 30 years. Over 50 years the gap widened: $6.1 million versus $12.1 million, so all-stock retirees ended up with about twice as much on average.
More than 60% bonds made things worse. At 50% stocks, the probability of ruin rose to 5.8% over 30 years and 36.7% over 50, and the average ending balance fell to $1.6 million and $1.2 million.
By owning bonds in your portfolio, you trade a slightly lower chance of running out of money for a much smaller nest egg. With 20% in bonds, one or two fewer starting years ran out of money. With 100% in stocks, the typical retirement ended with far more left over, for yourself, your heirs, or the unexpected.
Retirement worries tend to focus on sequence-of-returns risk: retiring right before a crash like 1929 and selling stocks at the bottom. A modest bond allocation is a real, if small, defense against that. But it's a defense with a price, and the longer the retirement, the higher the price, because stocks compound for longer. And with half the portfolio in bonds, they stopped protecting you at all: a portfolio that doesn't grow enough to keep up with withdrawals is its own way to run out of money.
A few caveats:
Your situation almost certainly isn't a single 60-year-old with everything in a Roth. You might have Social Security, a pension, a mix of account types, a spouse, or a mortgage. Ignigrade models all of those, taxes included. Describe your plan in the app, then change the allocation, the withdrawal amount, or the time horizon and see how the whole range of historical outcomes shifts.
This blog is where we'll write about questions like this one, how the simulation works, and what we learn along the way. If there's something you'd like us to look into, use the feedback button in the app. We read every one.
¹ Our results point the same way as the Trinity study's, but the exact percentages differ. We traced most of that difference to the data: the Trinity study's market data ends in 1995, while ours has more recent data which runs through 2025.