Key takeaways
- A 60/40 portfolio last set to target at the October 2007 peak had drifted to 47/53 by October 2008 in Robert Shiller's monthly real total-return data. Rebalancing it meant moving 13.3 percentage points out of bonds and into stocks.
- That October 2008 rebalance was 4.3% behind the frozen portfolio five months later at the March 2009 low, and still only 0.7% ahead after three years. By June 2026 it was 21.1% ahead.
- The same trade in March 2020 was 3.3% ahead within a year and 6.5% ahead after five, and was never behind at any month-end — the S&P 500 fell 33.9% in 23 trading days and closed back above its old high 148 days after the low.
- Almost none of that payoff came from buying stocks cheaply. From October 2008 to June 2026, real stock total returns ran +587% against −3.3% for 10-year Treasuries. The gap is the 13.3-point shift into the better-performing asset, not a rebalancing bonus.
- Acting a month early mattered more than acting at all: the same rebalance done in September 2008 rather than October did not move permanently ahead of the frozen portfolio for 52 months.
The answer turns on how much further it falls, which is the one thing nobody knows at the time
Equities are down 30%. The 60/40 you chose is now something like 50/50. Buy more stocks, or leave it alone?
Here is what the two episodes people reach for actually show. In both October 2008 and March 2020, moving the portfolio back to 60/40 eventually beat leaving it alone. But "eventually" did very different work in each case. In March 2020 the trade was ahead inside a year and never once behind. In October 2008 it was 4.3% behind within five months, and for three years it was worth less than one percent. What separates those two outcomes has nothing to do with the strategy and everything to do with how far the market fell after the decision was made.
That makes this a poor question to answer with a rule, and a good one to answer with a distribution of outcomes. Two crashes is a tiny sample, both happened to end in strong recoveries, and the whole payoff in both cases came from a source that has nothing to do with buying low. What follows is the arithmetic, the counter-case, and the places where it breaks.
The method, stated plainly, so the numbers can be checked
Stock and bond returns both come from Robert Shiller's monthly US dataset — the spreadsheet behind Irrational Exuberance, updated through July 2026. The equity leg is his real total-return series for the S&P Composite, which includes reinvested dividends. The bond leg is his real total-return series for 10-year US Treasuries, which includes coupons. Both are inflation-adjusted, so the comparison between two portfolios is unaffected by which measure you prefer.
The convention that matters most: Shiller's monthly price is the average of that month's daily closes, not the month-end close. His figure for March 2020 is 2,652.39, which is exactly the average of the 22 daily S&P 500 closes that month in the St. Louis Fed's series. So this data smooths a fast crash. It shows March 2020 as an 18.9% real equity drawdown, when the daily index touched −33.9%. For October 2008 the smoothing matters far less, because that decline was spread over 17 months.
The experiment is the reader's actual choice. Assume the portfolio sat at exactly 60/40 at the market's last peak. Let it drift with real returns. At the decision month, compare two portfolios: one moved back to 60/40, one left untouched. Neither is rebalanced again, neither receives contributions, and no trading costs or taxes are charged to either. Those omissions all favour rebalancing, and they are dealt with further down.
October 2008: the rebalance bought five more months of falling, then three years of almost nothing
Real equity total returns peaked in October 2007. Twelve months later they were down 38.0%, while 10-year Treasuries had returned +6.3% in real terms. The 60/40 was worth 79.7% of its peak value and had drifted to 46.7% equities — past the 50/50 in the reader's question, because the bond leg had gained while the equity leg collapsed.
Rebalancing there meant shifting 13.3 points of the portfolio from Treasuries into stocks, in the month after Lehman Brothers failed. Then the market fell another 19.3% in real terms while Treasuries added a further 11.8%.
By March 2009 the rebalanced portfolio was 4.3% below the frozen one. Measured from the October 2007 peak, it was down 25.7% against the frozen portfolio's 22.4%. It took ten months to move permanently ahead, and the lead was thin for years: +0.8% after one year, +0.7% after three. Only at the five-year mark did it reach +6.0%. Today, in June 2026, the gap is +21.1%.
Read the same row twice and you get two honest headlines. Rebalancing in October 2008 paid 21%. Rebalancing in October 2008 paid nothing at all for three years and cost 4% first. Both are the same number series.
March 2020: the same trade paid quickly because the crash lasted 23 trading days
The 2020 episode barely qualifies as the same kind of event. The S&P 500 peaked at 3,386.15 on 19 February 2020, bottomed at 2,237.40 on 23 March — 33.9% down in 23 trading days — and closed back above the old high on 18 August, 148 calendar days after the low.
At monthly resolution the drawdown looks much milder. Real equity total returns fell 18.9% from the January 2020 peak, Treasuries gained 8.7%, and the 60/40 drifted only to 52.8% equities. Rebalancing meant a 7.2-point shift, roughly half the size of the 2008 trade. On daily prices it briefly looked like the reader's premise: after a 33.9% equity fall, a 60/40 with an unchanged bond leg sits at 49.8% stocks. That was true for about a week.
The payoff arrived almost immediately and never reversed: +3.3% after one year, +4.4% after three, +6.5% after five, and +7.7% by June 2026. The rebalanced portfolio's worst month was the month of the decision itself. The frozen portfolio's worst month wasn't March 2020 at all — it was October 2022, when stocks and bonds fell together.
Getting the month right mattered more than getting the decision right
The chart above shows the gap between the rebalanced and frozen portfolios at four horizons in each episode. It understates how much of that gap is luck.
Running the same experiment for every month from September 2008 to April 2009 gives a spread that dwarfs the decision itself. Rebalance in September 2008, when equities were only 23% off their peak, and the portfolio was 4.2% behind at the March 2009 low, still behind after one year (−1.4%) and after three (−1.6%), and didn't move permanently ahead for 52 months. Its lead today is +10.7%, half the October figure. Rebalance in March 2009, at the low, and it was 10.7% ahead within a year and 42.5% ahead today.
One month either side of October changed the three-year result from −1.6% to +2.9%. Nothing in the 2020 grid comes close: the March, April and May 2020 decisions all land between +2.3% and +3.3% after a year. A tracker that shows drift continuously — LedgerTouch does this, and so does a spreadsheet updated once a month — tells you the weight you're holding. It can't tell you whether the next leg down has happened yet.
Nearly all of the payoff came from bonds doing badly afterwards
This is the part that should shape how much weight the two episodes carry.
The rebalanced portfolio wins for one reason: it holds more equities than the frozen one, and equities beat bonds over the measurement window. Nothing about the crash itself contributes. From October 2008 to June 2026, real stock total returns compounded to +587% while real 10-year Treasury total returns came to −3.3% — nearly eighteen years of holding an asset that lost purchasing power. From March 2020 to June 2026 the split is starker still: +136% for stocks against −32.3% for Treasuries, most of that bond damage inflicted by the 2021–2023 inflation.
Shorten the window and the effect shrinks with it. Over the three years from October 2008, stocks returned 27.6% real and Treasuries 20.6% — a 7-point gap, which is why the rebalanced portfolio was ahead by only 0.75%. The "reward for rebalancing into a crash" is really the reward for having held more equities through a long bull market and a long bond bear market. Anyone who doesn't want that exposure is reading this evidence correctly by declining it.
The counter-argument: you're buying an asset that just fell and may keep falling
The serious objection to rebalancing into a drawdown has little to do with whether it works. The objection is that it concentrates risk into exactly the moment when the range of outcomes is widest, and the October 2008 numbers show what that felt like. Someone who acted then bought before a further 19.3% real decline, deepened their peak-to-trough loss from 22.4% to 25.7%, and waited three years for a lead of less than one percentage point.
The bond leg deserves its own scepticism. All of the above uses Treasuries, which rallied hard as equities fell. A real-world 60/40 usually holds a broad bond fund with corporate credit in it, and corporate credit didn't rally. Moody's seasoned Baa corporate bond yield rose from 7.31% in September 2008 to 9.21% in November — yields up means prices down — while the 10-year Treasury yield fell from 3.69% to 3.53% over the same two months and to 2.42% by December. A portfolio whose bond sleeve was falling alongside stocks drifted less, so the rebalancing trade was smaller, and the asset being sold to fund it was not sitting on a gain.
There is also the plain argument from sequence. Both episodes are drawn from a period in which US equities recovered and went on to new highs. A method that depends on that ending is really a description of the sample it was drawn from.
The trade was most expensive exactly when it looked most attractive
Costs are usually waved away in this debate. In a crash they are not small.
Vanguard's 2022 rebalancing paper models transaction costs as a function of volatility rather than a fixed number, and finds that they "are not static through time but are sensitive to market volatility and can increase tenfold during periods of market turmoil." The same paper also notes that in turbulent markets one may rebalance in one direction and then have to reverse the transaction.
March 2020 put numbers on it. A Federal Reserve FEDS Note by Steven Sharpe and Alex Zhou records investment-grade corporate bond transaction costs rising from around 30 basis points per dollar of principal traded in February to nearly 90 basis points in mid-March, with block trades going from 24 basis points to over 150 by 23 March — the day of the low. A BIS bulletin by Sirio Aramonte and Fernando Avalos found that some of the largest investment-grade and high-yield bond ETFs traded at discounts to net asset value in excess of 5% in mid-March. Selling the bond side to buy the equity side was, at that moment, considerably worse than the index prices implied.
Tax is the other leak, and it only applies outside a sheltered account. Rebalancing after an equity crash usually means selling the asset that went up. Under IRS Topic 409, long-term gains are taxed at 0%, 15% or 20% depending on income, and losses can only offset ordinary income at $3,000 a year. A 13.3-point sale of appreciated Treasuries in a taxable account in 2008 realised a gain in the same year the portfolio lost a fifth of its value. In an ISA, a 401(k) or a SIPP, that consideration disappears.
Roughly seven in ten people did nothing, and most of those who traded sold equities
The behavioural half of this is often asserted and rarely measured. It was measured in 2020.
Giglio, Maggiori, Stroebel and Utkus matched a survey of Vanguard clients to their actual daily trading through the crash (NBER working paper 27272, May 2020). Between 31 January and 31 March, 67% of the optimists, 73% of the neutrals and 70% of the pessimists made no portfolio change at all. Among the minority who did trade, every group moved the same way — out of equities. Measured at constant prices, optimists actively cut their equity share by 1.05%, neutrals by 0.98%, pessimists by 0.63%. Investors who traded and had been most optimistic in February took their equity weight from 68% down to 64% by the end of March.
That sits awkwardly beside a second Vanguard measurement, quoted in our piece on risk tolerance and risk capacity: across more than five million retail households in the first half of 2020, 62% of those who traded moved money into equities. Both figures are Vanguard's and both are right. They count different things. One is a headcount of households and the direction of their cash across six months. The other is the change in a surveyed group's equity share, at constant prices, across two. A purchase too small to offset the drift still shows up as a fall in equity weight, so neither number settles what the other measures.
Two things follow. First, rebalancing into the drawdown was a minority act, and the average trade went the opposite way. Second, and more usefully: the authors point out that "'not trading' to rebalance a portfolio after market changes is also an endogenous decision" — freezing is a position, not the absence of one. It quietly lowers the equity weight and locks in a more defensive portfolio at the worst prices of the cycle. Whether that is a bug depends entirely on whether the original 60/40 still described the risk the household could actually carry. Our guide to the behaviour gap covers what that pattern has cost in aggregate.
The uncomfortable implication cuts against the rebalancer. If someone reaches for a rebalance only because a rule tells them to, and abandons the rule at the March 2009 low when it is 4.3% underwater, they've engineered the worst of both. The 2008 path punished exactly that.
Two crashes that both recovered can't tell you about one that doesn't
The limits here are severe and worth stating without hedging.
Two episodes is not a sample. Both are US. Both ended in recoveries that were, by historical standards, fast and complete. Shiller's monthly averaging hides the extremes, which is why March 2020 shows an 18.9% equity drawdown here against 33.9% on daily prices. The bond leg is 10-year Treasuries, not a broad bond fund, and 2008 shows how much that choice matters. And the whole result is a single historical path, not a distribution — the same objection that applies to every backtest.
Japan is the standing counterexample. The Nikkei 225 closed at 38,915.87 on 29 December 1989. It fell to 7,054.98 by 10 March 2009, 81.9% below the peak, and didn't close above the 1989 level again until 22 February 2024 — a gap of 34.2 years. That series is price-only and in yen, so dividends and currency would change the arithmetic for a foreign holder. On a broader measure the wait was longer still: the OECD's monthly share-price index for Japan bottomed 74.5% below its December 1989 level in November 2011 and didn't regain that level until July 2025. What it can't change is the shape: an investor who kept rebalancing into that drawdown through the 1990s was adding to a position that hadn't recovered two decades later, and didn't recover within an ordinary investing lifetime. Nothing in the 2008 or 2020 data speaks to that case, because neither episode contains it.
What would change the conclusion
If bonds had performed well afterwards. The payoff in both episodes is the equity-minus-bond return over the following years, applied to a 7-to-13-point shift. Over the three years after October 2008 that gap was 7 points and the reward was 0.75%. Over the seventeen and a half years to June 2026 it was 590 points. A post-crash decade that was kind to bonds and unkind to stocks reverses the sign, and the frozen portfolio — which holds more bonds — comes out ahead.
If the drawdown had more legs than these two did. The September-2008 result is the warning: 52 months before moving permanently ahead, from a decision taken one month too early. In a decline with four legs rather than two, the early rebalance carries a much longer penalty, and the 2008–09 numbers only bound how bad that gets in a recovery that arrived.
If the account is taxable and the bond leg is sitting on gains. Every figure above is pre-tax and pre-cost. The March 2020 cost data suggests the execution leak alone was measured in tens of basis points, against a first-year payoff of 3.3%.
If the target weight was wrong to begin with. Rebalancing restores an allocation. It's only worth restoring if the number still reflects what the household can hold through another 25% fall. A 60/40 chosen in calm markets and abandoned at the low was never a 60/40. That question sits upstream of this one, and it's separate again from how often to rebalance in ordinary markets.
The thing to watch during a drawdown isn't the index. It's the weight — how far the portfolio has actually drifted, whether that drift is 3 points or 13, and whether the resulting mix is one that can be held if the market falls another 20% from here. The October 2008 investor who rebalanced and held ended 21.1% ahead of the one who froze. The investor who rebalanced and then capitulated at the March 2009 low took the 4.3% shortfall and none of what followed.