Does rebalancing make you money?
No, and it does not reliably cost you either: rebalancing is a way of keeping your mix from drifting, and nothing more than that.
Over the 35 years we measured, it cost money compared with never touching the portfolio, and over longer runs of history the sign flips the other way. Neither result is big enough to plan around.
The pitch is appealing. You decide on a mix, say 60% US shares, 30% other developed markets, 10% emerging markets. One part runs ahead. You sell some of it and buy the laggard. You have sold high and bought low without predicting anything, and lots of people will tell you that is a source of extra return.
We tested it on those three regional stock indexes at that mix, every month from 1991 to the end of 2025. Every rebalancing rule we tried finished behind simply leaving the money alone. Rebalancing every month returned 9.88% a year against 10.32% for never trading at all, a gap of 0.44 points, and the gentler rules gave up between 0.24 and 0.27 points against that same comparison. Over 35 years those are not rounding errors.
Now the scope, because it decides how much that number is worth. One mix, three regions, 1991 to 2025. Run the same question over longer stretches and the sign reverses. Measured from 1871, and again from 1963, the rebalanced portfolio comes out slightly ahead of the untouched one rather than slightly behind. So the honest reading is that rebalancing is not a way of making money in either direction. It is a way of keeping your mix from drifting, and whether it costs or earns a little depends on the stretch of history you happen to look at.
Why it lost over this window
Rebalancing pays off when a region that has run ahead then falls back, so that the shares you bought cheap recover. Relative performance has to reverse.
In this data it did the opposite. A region that had beaten another tended to keep beating it, at every time span we looked at, in all three pairs of regions. The one thing that would make rebalancing profitable is missing from this sample, and its opposite is present. That is a structural reason rather than one unlucky run.
It does not protect you in a crash either. Every rebalancing rule we tested had a worst fall equal to or slightly worse than leaving the money alone, around 53% in each case. In a crisis every region falls together, so buying the one that has fallen most means buying more of something still falling.
What it does buy
Keeping the promise you made. Left completely alone, the 60/30/10 mix drifted an average of 14.8 percentage points away from where it started, and 26.4 points at its worst. For much of the period it was closer to 75/17/8. That is a different portfolio from the one anyone chose, arrived at by accident. Rebalancing held the drift to somewhere between 0.6 and 3.1 points, and which end of that range you get depends on the rule.
It is cheap to do. Reviewing once a year and acting only when a holding is a quarter away from its target comes to about 0.4 rebalances and 2.9 trades a year.
The trap that makes it look like it works
Under ordinary assumptions, rebalancing wins against leaving things alone about 68% of the time even when its average return is identical. The wins are frequent and small, the losses rare and large, and they cancel. So a study showing rebalancing won 70% of the time has shown you the result you would get from a rule that does nothing. A win rate cannot settle this question.
In our own data even 68% turned out to be generous. Nine of the twelve windows we measured came in below it.
Tax
The simulation charged no tax, because it held no purchase records and could not know what gain a sale would trigger. In a taxable account every rebalance realises gains. Adding that would move the answer further against rebalancing.
So rebalance to stay in the portfolio you chose, and do the trades in a retirement account if you have one. Do not do it expecting to be paid.
What would change our mind
Evidence that relative performance between regions reverses rather than continues, measured on data we have not already looked at. That is the mechanism the whole question turns on, and it is the one thing that would move the answer out of the range where the window decides the sign.
Where these numbers come from
- Monthly total returns for three regional stock indexes, January 1991 to December 2025, 420 months, run through our own simulation of calendar, band and cash-flow rebalancing rules.
- The standard mathematical result on how often rebalancing wins under lognormal returns with equal expected growth, which is where the 68% comes from.