Micheal Kitces coined the concept. Traditional advice is to reduce equities as you get older, using "age in bonds" or some other curve. For example, since we generally think we should match asset allocation to the time until we need the money. Say you keep 5 years worth of expenses in fixed income. At the start of a 30 year retirement, that's 5/30 = 17% of your assets, with the rest in some stock/bond mix. Later, when only 5 years remain, that's 100% fixed income, no stocks.
However, when you backtest this on historical data and compare it to a fixed asset allocation, fixed gives you a lower chance of running out of money in retirement. And, in fact, increasing equities over time does better than either.
There are a couple factors at play here:
1. "Safe" assets aren't. Fixed income assets are considered safer than equities, and in fact are often referred to as safe assets. This is certainly true over the span of days, months and years. But in the U.S., stocks have always outperformed Treasuries over all 20 year periods. Internationally, there are cases where they haven't, such as Japan's lost decades. Still, stocks are much, much more likely to outperform bonds over any 20 year period. This matters, because generally people retire before they have accumulated the sum total of all their future expenses. For example, if your expenses stayed the same in real terms, a 30 year retirement would need a portfolio that's 30x your annual expenses, or a 3.3% withdrawl rate. If you retire before this, say when you need a 4% withdrawl rate, then 100% fixed income will guarantee you run out of money, forcing you to play the Squid Game, which is very dangerous. In fact, fixed income doesn't keep up with inflation in the long term, so even if you had 30x you'd still run out. If you need positive real growth to succeed, then assets that provide negative real growth aren't safe in this sense.
2. The future performance of stocks doesn't just depend on their current price. Even in a bear market, it's safe to assume people will need food, clothing, shelter, transportation, entertainment, etc for decades to come, so companies that provide these will have profits for decades to come. In fact, there's an idea that the value of a company is the discounted sum of future dividends. Of course, we can't know the future dividends or the discount rate, but the point remains that stock markets oscillate between under valuing and over valuing, and if we've just had a bear market, stocks are more likely to do better.
To some, the idea that the stock price reflects all public information, including valuations, is the essence of the efficient market hypothesis. The capital asset pricing model assumes this and, as the wikipedia article says, "fails many historical back tests." It may be true when looking at minute by minute returns, or day by day, even year by year. But things are different when you look at decades.
This idea, that $1000 in stocks is the same whether we've just had a long bull or bear market, is very pernicious and baked into a lot of thinking. For example, people panic when the stock market crashes, because they think they've lost a lot of money. They're implicitly thinking the market is as likely to go down as it is to go up. But historically this isn't true.
3. Sequence of return risk. Traditional theory more or less ignores the effect of adding funds during accumulation, or withdrawing them during retirement. In a world where you neither add nor withdraw from your portfolio for 20+ years, historical data says you're best with 100% stocks, even in the worst case, and only need to hold bonds if you don't have diamond hands and will panic in a market crash. However, withdrawing money during retirement changes this. If you start with $1M in all stocks, have $40k annual expenses, and the stock market crashes 75%, well now your $40k withdrawl is 16% of your portfolio, and have a smaller portfolio to take advantage of the recovery. So the need for bonds, and the amount, is driven by sequence of return risk, not ability to take risk.
Putting it all together
Why have a lot of bonds at the start of retirement, and then go to 80% or even 100% equities after 5 to 15 years?
- If there's a bull market in the first decade of retirement, your equity portion has grown nicely. Your portfolio is now higher than when you started, even after 10 years of withdraws, and you only have 20 years of expenses ahead of you, not 30. Any asset allocation will succeed, in the sense of not running out of money in 20 years, even 100% equities.
- If there's a bear market in the first decade, you haven't been touching your equities, taking your living expenses exclusively from your bonds. When the following bull market comes, you're better off being stock heavy, to take best advantage of equity growth.