Leverage and volatility drag: why 2x is not twice the growth
On this page
Key takeaways
- Leverage multiplies both the average return and the swings, but the growth lost to volatility drag rises with the square of the multiple: twice the exposure brings about four times the drag, before borrowing costs.
- So doubling exposure less than doubles the growth rate, and beyond a certain multiple, known as the Kelly fraction, more leverage reduces growth while risk keeps rising.
- Leverage leaves the Sharpe ratio unchanged before costs, since return and volatility scale together; it deepens every decline, and at twice the exposure a fall of half the holding's value wipes the position out.
- Leveraged funds that reset to a fixed multiple every day show the same effect, often called volatility decay: over long periods they can return far less than the multiple of the index.
- MarketHeist's lab borrows for free with no trading costs, which flatters leverage; the backtester applies one constant multiple, either fixed or set from the whole backtest's volatility or drawdown.
Definition
Leverage is holding more exposure to an asset than the capital invested, by borrowing or with derivatives, so that gains and losses are multiplied. Because volatility drag grows with the square of exposure while the average return grows only in proportion, a leveraged portfolio compounds at less than the multiple of its unleveraged growth rate, and past some level more leverage lowers growth.
Also known as: leverage, leveraged portfolio, volatility decay, leveraged ETF decay, 2x leverage, margin, Kelly criterion, constant leverage.
How it works
Suppose an asset rises by a tenth one month and falls by a tenth the next. Held once, a value of one becomes 1.1 and then 0.99, a loss of one percent. Held at twice the exposure, it becomes 1.2 and then 0.96, a loss of four percent: twice the exposure, four times the damage from the same up-and-down. That is volatility drag, and it grows with the square of the exposure.
Over many months the arithmetic is simple. The average return scales with the leverage multiple, and so does volatility; the drag on compound growth, about half the square of volatility, scales with the multiple squared. For a moderately leveraged portfolio of a volatile asset, the drag eats a noticeable part of the extra return. Push the multiple high enough and the drag outgrows the extra return: the level at which growth peaks is the Kelly fraction, the average excess return divided by the square of volatility, and beyond it more leverage means slower growth with larger losses.
Borrowing has a price that this arithmetic leaves out. Margin loans, futures and leveraged funds all pay a financing rate on the borrowed part, roughly a short-term interest rate plus a spread. With borrowing at a few percent a year, the cost on the borrowed half of a 2x portfolio takes a direct bite out of the gain before any drag.
Leverage is like driving faster on a winding road: the trip can be shorter, but every bend costs more, and past a certain speed you arrive later because of the time spent recovering from skids, or not at all. A decline of half wipes out a 2x position that is never reduced; a portfolio held at a constant multiple is rebalanced after every fall, so it survives but sells into the decline and buys back higher.
Formula
- g_L
- compound (log) growth rate per year at leverage L
- L
- exposure as a multiple of capital (1 = unleveraged, 2 = twice the capital)
- μ
- the asset's average (arithmetic) return per year
- σ
- the asset's volatility per year
- r_b
- the borrowing rate paid on the borrowed part, L − 1 (0 in this page's lab)
From the drag rule g ≈ μ − σ²/2 of the CAGR guide, applied to returns multiplied by L and rebalanced to a constant L. The approximation is for moderate monthly returns; with large or fat-tailed moves the drag is larger.
- L*
- the growth-maximising leverage (Kelly fraction) in this continuous-time approximation
- g at L*
- the highest growth rate reachable by leverage
At 2L* the expected growth falls back to the borrowing rate. The estimate is fragile: μ is measured with large error, so leverage set from a backtested μ often overshoots. Many practitioners use a fraction of L* for that reason.
- rₜ
- the asset's return over the period between rebalancings
- rₜᴸ
- the leveraged portfolio's return over the same period
- cₜ
- the financing rate for the period
At 2x, a fall of half between rebalancings ends the position; at 3x, a fall of a third. Rebalancing more often (daily for leveraged ETFs) avoids the wipe-out but locks in the drag.
Worked example
The S&P 500 (SPY, extended back with its index mutual fund), from Feb 1988 to Sep 2026, held fully invested or resized each month to a volatility target of 20%, well above the index's own, using the previous 12 months and capped at 2 times exposure. Borrowing is free and nothing is charged for trading.
The rule was leveraged most of the time: exposure averaged 151% of the portfolio's value, ranged from 66% to 200%, and sat at the 2x cap in 29% of months.
Growth rose from 11.3% to 15.7% a year, but by a smaller multiple than the exposure: average exposure was about one and a half times the portfolio's value, while growth rose by noticeably less than half. The difference is volatility drag, which grows with the square of exposure.
Risk rose with the exposure. Volatility went from 14.5% to 20.8%, and the deepest decline from -50.8% to -63.1%: a fall of that size takes a gain of well over double to make good. The Sharpe ratio, 0.81 against 0.81, barely moved; leverage scaled return and risk together and added nothing per unit of risk.
And this is the flattering case. Borrowing a large part of capital at a typical margin or futures financing rate costs several percent a year on the borrowed amount, which would take a large share of the remaining gain, and every monthly resize would be a trade.
Data through Sep 2026. Reproduce id 3e9813405244a5a7 (this guide, its lab setup and the data month).
Interactive lab
Growth of $10,000 unleveraged and with exposure resized monthly toward the volatility target, up to the leverage cap, with the exposure below and the table of growth, volatility, deepest decline and Sharpe ratio. Raise the cap to 3 or the target to see drag and drawdowns grow.
| Measure | Plain | Targeted at 20% |
|---|---|---|
| CAGR | 11.3% | 15.7% |
| Volatility (annualized) | 14.5% | 20.8% |
| Maximum drawdown | -50.8% | -63.1% |
| Sharpe ratio | 0.81 | 0.81 |
| Growth of $10,000 | $633,728 | $2,821,395 |
| Average exposure | 100% | 151% |
| Highest and lowest exposure | 100% | 200% and 66% |
| Months at the 200% cap | none | 29% |
Exposure is the target over the trailing volatility (the sample standard deviation of the 12 months before each month, times the square root of 12), capped at 200%. Money not invested earns 0, and an exposure above 100% is borrowed at 0. Risk-free rate 0. Feb 1988 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 compare the plain portfolio and the targeted one over the same Feb 1988 to Sep 2026 (464 months), which starts once the longest lookback the lab offers (36 months) has history, so the lookback slider does not move the window. | |
| Portfolio streams | A fund is its proxy-extended monthly return stream. A built-in portfolio is read as the MarketHeist MCP reads it: its composed stream with its own rebalancing rule and overlay. A custom mix is rebalanced yearly. | |
| Volatility targeting | Each month the exposure is the target divided by the trailing volatility, capped at the leverage limit. The trailing volatility is the sample standard deviation of the monthly returns of the lookback months before the month (never the month itself), times the square root of 12. The portfolio earns the exposure times that month's return; the part not invested earns 0, and an exposure above 100% borrows at 0, so no financing cost is charged. The plain comparison is the same stream fully invested. | |
| Leverage in this lab | Exposure each month = min(cap, target ÷ trailing volatility), the trailing volatility being the sample standard deviation of the lookback months strictly before the month, times √12. The portfolio earns exposure × the month's return; the borrowed part (exposure − 1) costs 0, and resizing is free. | volTargetSidecontent-engine/lab/compute4.ts |
| Wipe-out guard | If an exposure would turn a month's return into a loss of more than 100%, the lab refuses the setting instead of continuing with a negative value. | volTargetSidecontent-engine/lab/compute4.ts |
| Leverage cap | From 1 to 3 times capital in the lab. | VT_LEVERAGE_MAXcontent-engine/lab/compute4.ts |
| Backtester: fixed leverage | One constant multiplier applied to every bar's strategy return (after transaction costs), so the position is in effect rebalanced to that multiple each bar. A value that is not a positive number is treated as 1; values above 10 are capped at 10. Buy-and-hold stays unleveraged for reference. No borrowing cost is charged. | resolveLeveragepackages/shared/engine/backtest.ts |
| Backtester: target volatility | The multiplier is target ÷ the unleveraged strategy's realised volatility over the WHOLE backtest (sample standard deviation of bar returns × √bars per year), with the same limits. It is one number set with hindsight, not a rule that adapts over time. | resolveLeveragepackages/shared/engine/backtest.ts |
| Backtester: target drawdown | The multiplier is the target drawdown ÷ the unleveraged strategy's maximum drawdown over the whole backtest, with the same limits (1 when there is no drawdown). Because losses compound, the leveraged backtest's drawdown is not exactly the target: scaling returns by L scales a drawdown by roughly, not exactly, L. | resolveLeveragepackages/shared/engine/backtest.ts |
| Engine functions | The functions the portfolio app and the MarketHeist MCP tools run, called here on the same monthly returns. | nanstdpackages/shared/engine/utils.tsmetricsFromReturnspackages/portfolio/engine/portfolio.ts |
Leverage and volatility drag compared with its neighbours
| Route | How exposure is kept | Financing cost | Path effects |
|---|---|---|---|
| Margin loan on a fixed position | Not rebalanced; leverage drifts down as prices rise and up as they fall | Broker's margin rate on the loan | Margin calls in declines |
| Constant leverage, rebalanced | Reset to the multiple each period | Financing rate on the borrowed part | Volatility drag; sells after falls, buys after rises |
| Daily-reset leveraged ETF | Reset to the multiple every day | Built into the fund's return, plus its fee | Volatility decay over long holding periods |
| Futures | Notional exposure for a margin deposit | Implied in the futures price (about the short-term rate) | Needs rolling and cash for margin |
| Volatility-targeted leverage | Multiple set by recent volatility, capped | Financing on the borrowed part when above 1 | Less leverage after turbulent stretches |
Limitations
- The lab charges nothing for borrowing or trading. Real leverage pays a financing rate on the borrowed part and costs on every resize, so the growth figures here overstate what a leveraged investor would have kept.
- Monthly data hide what happens within a month. A fall large enough to trigger a margin call or wipe out a position inside a month is not seen until the month ends.
- The growth-maximising multiple depends on an average return that is measured with large error; a backtest that looks good at high leverage is especially sensitive to the period chosen.
- One historical path of US stocks over a strong period. Leverage magnifies whatever the period was; a weaker period would show much worse results at the same multiple.
- Taxes, margin rules and the behaviour of leveraged funds differ by account and product; none of it is modelled.
Frequently asked questions
- Why does a 2x portfolio not earn twice the return?
- Because compound growth loses about half the square of volatility, and at twice the exposure volatility doubles, so the drag quadruples. Twice the average return minus four times the drag is less than twice the compound growth, before borrowing costs, which reduce it further.
- What is volatility decay in leveraged ETFs?
- Leveraged ETFs reset to their multiple every day, so over longer periods their return depends on the path, not just on the index's total move. In a choppy market that ends flat, the daily resets lose money; in a steady trend they can beat the multiple. The average effect is the same volatility drag that grows with the square of the multiple.
- Does leverage improve the Sharpe ratio?
- No, before costs it leaves it unchanged, since return and volatility scale by the same multiple. Borrowing costs above the risk-free rate lower it. Leverage is a way to take more of the same risk, not better risk.
- What is the Kelly criterion for portfolios?
- The leverage that maximises long-run compound growth, roughly the average excess return divided by the square of volatility. Beyond it, more leverage lowers growth while increasing risk. Because the average return is uncertain, many investors who use it take only a fraction of the computed level.
- How does the backtester's target-volatility leverage work?
- It computes one multiplier for the whole backtest, the target divided by the strategy's realised volatility over the full period, and applies it to every bar. It uses the whole history, so it is a scaling for comparison, not a rule that could have been followed in real time; the lab on this page resizes monthly from past data only.
References
- Kelly, J. L. (1956) A New Interpretation of Information Rate. Bell System Technical Journal 35(4).
- Merton, R. C. (1969) Lifetime Portfolio Selection under Uncertainty: The Continuous-Time Case. Review of Economics and Statistics 51(3).
- Thorp, E. O. (2006) The Kelly Criterion in Blackjack, Sports Betting and the Stock Market. In Handbook of Asset and Liability Management, Volume 1, Elsevier.
- Cheng, M. and Madhavan, A. (2009) The Dynamics of Leveraged and Inverse Exchange-Traded Funds. Journal of Investment Management 7(4).
- Frazzini, A. and Pedersen, L. H. (2014) Betting Against Beta. Journal of Financial Economics 111(1). Leverage constraints and pricing.
Try it on your own holdings
Open the backtester and compare fixed, target-volatility and target-drawdown leverage →
Or ask an assistant connected to the MarketHeist MCP server (https://marketheist.io/mcp), which computes the same figures:
Using MarketHeist.io's monthly S&P 500 history, compare holding the index with resizing exposure each month to a 20% volatility target capped at 2x (borrowing at 0%): CAGR, volatility, max drawdown and Sharpe, and explain why growth rose by less than the average leverage
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: leverage.md.