Graveyard / Equity events
The ex-dividend trade
Is there an ex-dividend premium in US stocks?
The idea
Stocks are said to fall by less than the dividend on the ex-date. Buy before, sell after, keep the difference. Hedge with SPY.
How we tested it
| Data | US stocks with at least 1% yield, price above $5, median daily volume above $5M. 9,762 cash dividends with daily prices. |
|---|---|
| Period | 2019 to 2026. Design on July 2019 to December 2023, holdout January 2024 to September 2026, split fixed in advance. |
| Costs | 20 bp per round trip. |
| Holdout | Frozen split, as above. |
| Control | Random twins: five dates per event from the same stock and year, with no ex-date nearby. |
Result
| Events | 9,762 |
|---|---|
| Gross per trade | about 0 bp |
| Holdout net, sell at ex-date | −17 bp (t −5.6) |
| Holdout net, hold 2 days | −25 bp |
| Holdout net, hold 5 days | −46 bp |
Why it failed
There is no premium left to collect, so costs make it negative. A first run looked spectacular (t-stat 40 to 60). It counted the dividend twice on prices that already include it. That run was thrown away.
What would make it worth retesting
With unadjusted prices and only special dividends (yield 5% or more) with known announcement dates.
Test id: exdiv-equities-event-v1. Past results of a backtest say nothing certain about the future.