Graveyard / Fund discounts / ETFs
Closed-end fund discounts
Do closed-end funds trading far below NAV snap back?
The idea
A closed-end fund can trade below the value of what it owns. When the discount is wider than usual, it tends to narrow. Buy the fund, short a liquid ETF that matches its mandate, and wait.
How we tested it
| Data | 50 US closed-end funds (equity, sector, muni, high yield, loans, preferred, emerging-market debt). Daily price and daily NAV. Enter when the discount z-score is below −2, exit at zero, 126-day maximum hold. Hedge sized by a rolling beta. |
|---|---|
| Period | January 2006 to October 2026. |
| Costs | 10 bp per side on the fund, 1 bp on the hedge, 0.5-1% a year borrow. Stress case 30 bp per side on the fund. |
| Holdout | Full, in-sample and out-of-sample Sharpe, all three required to pass. |
| Control | No random-entry control was run for this test. |
Result
| Trades, 252-day window / 126-day window | 940 / 1,436 |
|---|---|
| Sharpe full / in-sample / out-of-sample (252-day) | 0.26 / 0.34 / −0.31 |
| Sharpe full / in-sample / out-of-sample (126-day) | 0.57 / 0.65 / 0.09 |
| Profit factor | 1.92 / 1.89 |
| Max drawdown | −31% / −29% |
| Time-stop exits (252-day) | 22% |
Why it failed
The win rate and per-trade profit factor look good, but the hedged result is volatile. Muni and high-yield NAVs are smoothed and lag their ETF hedges, and in 2008 and 2020 discounts widened further before they reverted. Out-of-sample from August 2022 the Sharpe is at most 0.1, and 2024-26 is flat or negative. Only funds alive in 2026 were tested, so the sample is survivor-biased.
What would make it worth retesting
With a structurally better hedge, for example the fund's own holdings basket, which is not available for free.
Test id: cefdisc-cefbook-event-v1. Past results of a backtest say nothing certain about the future.