Financing and Margin Cost Modeling
Leverage isn't free money, it's borrowed money, and borrowed money carries interest. A backtest that scales up returns by a leverage multiple without subtracting the financing cost is reporting a return nobody could actually collect.
Prerequisites: Transaction Costs
Running a strategy at 2x leverage in a spreadsheet is one keystroke: multiply the return series by two. Running it at 2x leverage in a real account means borrowing an amount equal to the account's own equity, and that borrowed cash costs interest — margin interest from a broker, or repo financing from a prime broker, depending on the setup. A backtest that applies the leverage multiplier to returns but not to the financing side is describing a strategy that borrows money for free, which is not a strategy anyone can actually run.
Worked example: the levered strategy that gives half its edge to the lender
An unlevered long strategy earns 8% a year with 10% volatility, a Sharpe of about 0.8 assuming zero risk-free rate. Levering it 2x is attractive because it doubles the return and the volatility, keeping the Sharpe the same — in theory.
| Unlevered | 2x levered, no financing cost | 2x levered, with financing | |
|---|---|---|---|
| Asset return | 8% | 16% | 16% |
| Financing cost on borrowed leg (5.5%, borrowed = 1x equity) | — | — | −5.5% |
| Net return | 8% | 16% | 10.5% |
| Volatility | 10% | 20% | 20% |
| Sharpe | 0.80 | 0.80 | 0.53 |
The naive doubling shows a levered Sharpe identical to the unlevered one — leverage looks like a free scaling knob. Once the actual cost of the borrowed capital is subtracted, the levered Sharpe drops by a third, because volatility scaled perfectly with leverage but return did not — the financing cost is a fixed drag that doesn't grow with market exposure the way the borrowed capital's return does.
Every unit of leverage above 1x should be paired with a financing charge on the borrowed capital, applied continuously against the exact borrowed notional at whatever rate the account actually pays, not just against the headline return figure at the end.
Financing rates aren't fixed either — margin and repo rates move with the broader interest-rate cycle, so a strategy backtested through a near-zero-rate period and later run through a higher-rate one will see its levered Sharpe compress even with identical asset returns, purely from the financing line.
The correction is mechanical: track borrowed notional day by day, apply the period's actual financing rate to it, and subtract that carry from gross P&L before computing any leverage-adjusted Sharpe or drawdown figure.
Related concepts
Practice in interviews
Further reading
- Frazzini & Pedersen, Betting Against Beta