Diversification return: when rebalancing adds growth
On this page
Key takeaways
- A mix of holdings that do not move in lockstep is calmer than its parts, and a calmer path loses less to volatility drag, so a rebalanced mix can grow faster than the average of its holdings' growth rates.
- That extra growth is called the diversification return or rebalancing bonus. The more the mix calms its holdings' swings, the larger it is.
- It is largest when the holdings are volatile, weakly correlated and grow at similar long-run rates. When one holding grows much faster than the others, a portfolio left alone drifts toward the winner and can end with more money than the rebalanced one.
- Rebalancing is what keeps the weights fixed and so collects the bonus; the reset sells some of whatever has just outrun the rest and buys what has lagged.
- MarketHeist shows it as the gap between the growth rate of a mix rebalanced on a calendar and the same mix never rebalanced, over the same months, with no trading costs or taxes.
Definition
The diversification return is the extra compound growth a portfolio rebalanced to fixed weights earns over the weighted average of its holdings' own compound growth rates. It arises because the rebalanced mix is less volatile than its parts and so loses less to volatility drag; it is largest for volatile holdings that move independently and grow at similar rates.
Also known as: rebalancing bonus, rebalancing premium, diversification bonus, excess growth rate, volatility harvesting, volatility pumping, Shannon's demon.
How it works
Compound growth sits below the average return by roughly half the square of volatility, the volatility drag described in the CAGR guide. Put two volatile holdings together and keep them at fixed weights, and the mix's average return is simply the weighted average of theirs. Its volatility, though, is lower than the weighted average of theirs whenever they do not move in lockstep. Less volatility means less drag, so the mix's compound growth rate comes out above the weighted average of the two compound growth rates.
Rebalancing is the mechanism. Each reset trims the holding that has just risen relative to the other and tops up the one that has lagged. When the two keep taking turns, with neither pulling away for good, the portfolio repeatedly sells some of a holding after a good stretch and buys more of the other after a poor one, and the mix compounds faster than either kept alone could suggest. Rebalancing is like two rowers taking turns at the oars of one boat: each tires and recovers at different times, and the boat keeps a steadier pace than either rower alone.
The bonus has conditions. It grows with the holdings' volatility and shrinks with their correlation, so two calm bond funds or two near-identical stock funds offer almost none. And it competes with drift: if one holding grows much faster than the other over the whole period, never rebalancing lets the faster one take over the portfolio, and that drift can be worth more than the bonus. Stocks against bonds over recent decades are the usual example.
The diversification return is not free money. It is the growth that a fixed-weight portfolio keeps by not letting its volatility rise, and it comes with the trades, costs and taxes of every reset. Measured against the weighted average of the holdings' growth rates it is always positive when the weights are fixed and the holdings are not perfectly correlated; measured against simply never rebalancing, which is the comparison an investor faces, it can go either way.
Formula
- gₚ
- compound (geometric) growth rate of the portfolio, rebalanced to fixed weights
- gᵢ
- compound growth rate of holding i on its own
- wᵢ
- fixed target weight of holding i (the weights sum to 1)
- σᵢ
- volatility of holding i
- σₚ
- volatility of the portfolio
It follows from g ≈ μ − σ²/2 applied to each holding and to the portfolio, with the portfolio's average return μₚ equal to the weighted average of the holdings' averages. The second term is the diversification return (Booth and Fama, 1992).
- DR
- diversification return per year
- w₁, w₂
- the two fixed weights, w₁ + w₂ = 1
- σ₁, σ₂
- the two holdings' volatilities
- ρ
- the correlation of their returns
- σₚ
- volatility of the mix
The bracket on the right is the variance (uncertainty) of the difference between the two holdings' returns. Half and half, two holdings with a volatility of 0.15 each and a correlation of 0 give 0.5 × 0.25 × 0.045, about 0.0056 a year; at a correlation of 1 the bonus is 0.
- Δ
- growth gained (or lost) by rebalancing, per year
- V_never
- final value of one unit invested at the target weights and never traded again
- wᵢ*
- target weight of holding i
- rᵢ,ₜ
- return of holding i in month t
- N
- number of months in the window
A never-rebalanced portfolio ends at the weighted average of its holdings' final values, which is at least the value implied by the weighted average of their growth rates. So Δ is a stricter test than the diversification return itself: it is positive only when the bonus outweighs the drift toward the faster-growing holding.
Worked example
Half in the S&P 500 (SPY) and half in gold (GLD), monthly total returns from Dec 2004 to Sep 2026, about 22 years, either left alone or reset to half and half every December, every quarter-end or every month.
Left alone, the stocks-and-gold mix grew at 10.61% a year. Reset every December it grew at 11.27% a year, and with quarterly resets at 11.25%: rebalancing added growth here, the diversification return at work.
It did so with less volatility, not more: 11.8% a year with yearly resets against 12.5% left alone. The deepest decline was -26.1% with yearly resets and -26.1% without, about the same.
The two funds took turns. Left alone, gold's weight wandered as far as 26.9% (percentage points) from its half and ended at 46.6%; with yearly resets it strayed at most 14.9% before being set back. Each December the reset sold some of whichever fund had just done better and bought the other, and over this window the turns kept coming, so neither fund ran away with the portfolio.
Resetting more often did not add more: monthly resets gave 11.25% a year. A month is too short for the two funds to drift far apart, so there is little to harvest at each reset, while every reset is a trade.
This pair is a favourable case: two volatile holdings with little correlation and, over this window, growth rates not far apart. The rebalancing guide runs the same test on a stocks-and-bonds portfolio, where stocks outgrew bonds for decades; in a case like that the drift toward the faster holding can be worth more than the bonus, and never rebalancing can come out ahead.
Data through Sep 2026. Reproduce id 9f80f4771fe74ac7 (this guide, its lab setup and the data month).
Interactive lab
Gold's actual weight month by month in a half-stocks, half-gold portfolio under each rebalancing rule, with growth, volatility, deepest decline and Sharpe ratio per rule. Compare the growth of never against the calendar resets.
| Rebalance | CAGR | Volatility | Max drawdown | Sharpe | Largest drift of GLD | GLD weight at the end |
|---|---|---|---|---|---|---|
| Never | 10.6% | 12.5% | -26.1% | 0.87 | 26.9 pts | 46.6% |
| Yearly | 11.3% | 11.8% | -26.1% | 0.97 | 14.9 pts | 46.0% |
| Quarterly | 11.3% | 11.8% | -26.4% | 0.96 | 7.4 pts | 50.2% |
| Monthly | 11.2% | 11.9% | -25.6% | 0.96 | 4.8 pts | 48.3% |
Target weight of GLD: 50.0%. Drift is the largest gap between the month-end weight and the target. Dec 2004 to Sep 2026.
How MarketHeist computes it
Every choice an implementation has to make, as the code makes it today. The page is checked against the code each time it is built.
| Choice | How it is computed | In the code |
|---|---|---|
| Data | Monthly total returns (dividends reinvested). Funds younger than the period shown are extended backward with older funds, spliced on returns, not prices. | |
| Window | The worked example and the lab use the common window of the holdings, Dec 2004 to Sep 2026 (262 months), ending at the data month of this page. | |
| Rebalance dates | Yearly resets in December, quarterly in March, June, September and December, monthly every month; between resets the holdings drift with their own returns. The drift chart shows each month-end weight before the rebalancing trade. | |
| Simulation | Both holdings start at their target weights in the first month of the common window. Each month every holding's value grows by its own total return, and the portfolio's return is the change in total value. Never means buy and hold from the first month. | simulatePortfoliopackages/portfolio/engine/portfolio.ts |
| Reset timing | On a rebalance month (December for yearly; March, June, September and December for quarterly; every month for monthly) the values are set back to target weight times total value at the end of the month, after that month's returns. | isRebalanceBoundarypackages/portfolio/engine/portfolio.ts |
| Diversification return | Not printed as a separate figure. The page shows the gap that matters to an investor: the compound growth rate of the rebalanced portfolio minus that of the same holdings never rebalanced, over the same months. The textbook diversification return, measured against the weighted average of the holdings' own growth rates, is at least as large as that gap. | |
| Growth and volatility | CAGR is final value to the power 1/years, minus 1, with years = months / 12. Volatility is the sample standard deviation (divides by N − 1) of monthly returns times √12. Sharpe ratio uses a risk-free rate of 0. | cagrpackages/portfolio/engine/metrics.ts |
| Drift figures | The tracked holding's weight is read at each month-end after that month's returns and before any reset; the largest gap from target over the window is the maximum drift. The build checks that this loop reproduces the engine's value curve. | driftWeightscontent-engine/lab/compute.ts |
| Costs and taxes | None. Resets are frictionless, so the figures are the bonus before the spreads, commissions and taxes that each reset would pay in a real account. | simulatePortfoliopackages/portfolio/engine/portfolio.ts |
| Engine functions | The functions the portfolio app and the MarketHeist MCP tools run, called here on the same monthly returns. | simulatePortfoliopackages/portfolio/engine/portfolio.tsisRebalanceBoundarypackages/portfolio/engine/portfolio.ts |
Diversification return (rebalancing bonus) compared with its neighbours
| Idea | What it compares | Sign | Depends on |
|---|---|---|---|
| Diversification return (Booth and Fama) | Rebalanced portfolio's growth vs the weighted average of its holdings' growth rates | Positive whenever weights are fixed and correlation is below 1 | Volatilities and correlation |
| Rebalancing bonus vs buy and hold | Rebalanced portfolio vs the same holdings never rebalanced | Either | Whether the holdings take turns or one pulls away |
| Volatility drag | A holding's average return vs its compound growth rate | Always a cost | The square of volatility |
| Volatility pumping (Shannon's demon) | Repeated rebalancing between a volatile asset and cash in a model with no trend | Positive in the model | The volatility of the asset |
Limitations
- One pair over one window. The bonus depends on the holdings taking turns; a period in which one of them keeps outgrowing the other can turn it into a loss against buy and hold, and gold's history in this test is only as long as the fund's.
- The approximation in the formula assumes moderate returns; with large monthly moves and fat tails the true figures differ from it.
- Trading costs and taxes are left out. In a taxable account each reset can realise gains, which can take back much or all of a bonus of this size.
- The diversification return is not a source of return separate from risk: it is the drag a portfolio avoids by keeping its volatility down. A higher-volatility version of the same mix, with more of the riskier holding, can still grow faster.
- Past turn-taking between two assets is no indication that it continues.
Frequently asked questions
- Does rebalancing increase returns?
- Sometimes. Against the weighted average of the holdings' compound growth rates, a fixed-weight portfolio of holdings that are not perfectly correlated always gains a diversification return. Against simply never rebalancing, rebalancing adds growth only when the holdings take turns leading; when one keeps outgrowing the rest, the portfolio left alone grows faster because it ends up holding more of the winner.
- What is the rebalancing bonus?
- A name popularised by William Bernstein for the extra growth a rebalanced portfolio of volatile, weakly correlated holdings can earn over its holdings held separately. Academic work calls it the diversification return (Booth and Fama, 1992) or the excess growth rate.
- Which holdings give the largest diversification return?
- Holdings that are each volatile, have low or negative correlation with one another and grow at similar long-run rates. Two bond funds offer little because they are calm; two broad stock funds offer little because they move together.
- Why did rebalancing help stocks and gold but not stocks and bonds?
- Stocks and gold were both volatile and largely unrelated, and over the window tested neither pulled away from the other for good, so each reset harvested a turn. Stocks grew much faster than bonds over decades, so the unrebalanced portfolio drifted toward stocks and its extra stock exposure outweighed the bonus.
- Is the diversification return the same as volatility pumping?
- They are the same mechanism. Volatility pumping, or Shannon's demon, is the textbook case of rebalancing between one volatile asset and cash in a market with no trend, where the rebalanced portfolio grows even though the asset alone does not. The diversification return is the general version for any set of holdings.
References
- Booth, D. G. and Fama, E. F. (1992) Diversification Returns and Asset Contributions. Financial Analysts Journal 48(3). Defines the diversification return of a rebalanced portfolio.
- Fernholz, R. and Shay, B. (1982) Stochastic Portfolio Theory and Stock Market Equilibrium. Journal of Finance 37(2). The excess growth rate of a portfolio over its holdings.
- Bernstein, W. J. (1996) The Rebalancing Bonus: Theory and Practice. Efficient Frontier (online journal).
- Erb, C. B. and Harvey, C. R. (2006) The Strategic and Tactical Value of Commodity Futures. Financial Analysts Journal 62(2).
- Willenbrock, S. (2011) Diversification Return, Portfolio Rebalancing, and the Commodity Return Puzzle. Financial Analysts Journal 67(4).
- Luenberger, D. G. (1998) Investment Science. Oxford University Press. Volatility pumping.
Try it on your own holdings
Build a half-stocks, half-gold portfolio and switch its rebalancing schedule →
Or ask an assistant connected to the MarketHeist MCP server (https://marketheist.io/mcp), which computes the same figures:
Using MarketHeist.io, compare a 50% SPY / 50% GLD portfolio rebalanced never, yearly, quarterly and monthly over the common window: CAGR, volatility and max drawdown, and explain whether rebalancing added growth
Ask an AI assistant
The prompt opens in a new tab, already filled in. An assistant without the MarketHeist connector answers from its own knowledge, not from the figures on this page.
Historical simulation, not advice. Computed by the MarketHeist portfolio engine from monthly total returns, data through Sep 2026. Markdown version of this page: diversification-return.md.