This is a standalone version of an article first published on quantstr.at. The Python and R code, the data files and the line-by-line walk-through are in the original.
TL;DR
- Strategy: Section 4.5 of 151 Trading Strategies shorts a leveraged ETF and its inverse and holds Treasury bills against them, to collect the volatility decay of leveraged funds.
- What we did: an attempt to replicate the volatility-decay formula of Avellaneda and Zhang (2010) and test the strategy on 16 ETF pairs from Yahoo Finance, under daily, monthly and annual resets and borrow fees of 0% to 5%.
- The decay formula holds: across 542 full-year ETF observations the fitted slope is −0.99 against a predicted −1 (R-squared 0.90, standard error 0.027 clustered by year).
- The strategy: 2.8% a year with daily resets, against 1.6% for a Treasury bill ETF. Monthly resets: 7.9% with no borrow fee, 4.8% at 3% and 2.8% at 5%.
- With frictions: margin calls (81 forced closes), no interest on short proceeds and trading costs together cut the monthly rule to 4.1%, and to 1.1% at a 3% borrow fee. The energy pair lost 99% in March 2020, and annual resets wiped out two pairs.
The idea
A 2x leveraged ETF targets twice the daily return of an index. Over longer periods its return is not twice the index return, because the exposure is reset every day and volatility erodes it. For an ETF with leverage L, the log return over a period is approximately
where S is the index and ru its daily returns. The drag grows with volatility. Shorting both the fund and its inverse should in principle collect the drag while index moves cancel. In practice the shorts must be reset, borrowed and kept within margin.
Method
- Pairs: 16 leveraged ETFs and their inverses (S&P 500, Nasdaq-100, Russell 2000, Dow 30, S&P MidCap, financials, energy, technology, semiconductors, 20+ year Treasuries), 2x and 3x.
- Substitutes, disclosed: unleveraged ETFs stand in for the indexes, SHV for Treasury bills; the 2x financials pair uses IYF as its index proxy before 20 March 2023 and XLF after; the energy pair changed from 3x to 2x on 1 April 2020 and 2020 is left out of the regression.
- Strategy per $1 of equity: short $0.50 of the leveraged ETF and $0.50 of the inverse, hold $1 of SHV, with $1 of collateral earning nothing.
- Resets: daily, monthly or annual. Borrow fees of 0%, 1%, 3% and 5% a year are scenarios, not observed rates.
- Margin: 60% maintenance on short value for 2x pairs and 90% for 3x; a breach closes the shorts at the close and holds cash until the next reset.
Results
The realised loss tracks the formula closely, and the slope is statistically indistinguishable from −1.
- Reset frequency matters. Daily resetting earns little. Monthly resetting earns the most, but borrow fees take much of it.
- Margin rules bite. With the margin rule the monthly strategy returned 5.9%; with no interest on short proceeds 6.3%; with trading costs 7.9% (before borrow fees). All three together: 4.1%.
- Tail risk is real. The energy pair's drawdown reached −99% in March 2020, and annual resetting wiped out two pairs.
Over the common period from 2010 the monthly portfolio returned 5.5% with no borrow fee and 2.4% with a 3% fee, against 1.5% for bills.
Limitations
- Unleveraged ETFs replace the true indexes; borrow fees are scenarios; survivors only (closed funds are missing).
- Margin is modelled in a simple form; prime-broker terms differ.
- This is an attempt to replicate; the strategy results extend the paper, which gives the formula and no results. Not investment advice.
Conclusion
The volatility-decay formula is confirmed in the data. The strategy built on it earns a modest premium over Treasury bills before frictions, and borrow fees, margin and trading costs consume much of it. The result is highly sensitive to how often the shorts are reset, and a single volatile month can destroy a pair.
References: Kakushadze, Z. and Serur, J. A. (2018). 151 Trading Strategies. SSRN: https://ssrn.com/abstract=3247865, Chapter 4, Section 4.5. Avellaneda, M. and Zhang, S. (2010). Path-dependence of leveraged ETF returns. SIAM Journal on Financial Mathematics, 1(1), 586-603.
The full article, with the Python and R code: https://quantstr.at/quant-strategies/leveraged-etfs/?utm_source=devto&utm_medium=syndication
If you want the next strategies from the 151 Trading Strategies series, follow @quantstrat here.
Educational content, not investment advice. Backtested results are not a forecast of future returns.



Top comments (2)
Here's the Python code: quantstr.at/quant-strategies/lever...
Here's the R code: quantstr.at/quant-strategies/lever...