The Cross-Section of Intraday and Overnight Returns: Night and Day
Do intraday and overnight returns contribute equally across individual stocks?
Separate close-to-open and open-to-close returns for the same dated US-equity universe, with corporate actions and transaction times aligned. Compare anomalies within each interval; the cited result is a sample finding, and extra round trips require an explicit trading schedule.
US equity intraday anomaly variation is sample-specific; round-trip count requires an explicit trading schedule.
Evidence map
| Aspect | Finding |
|---|---|
| What it is | Two studies establishing that the night-day split appears across multiple return predictors in the studied cross-section, and that at the index level the two periods have behaved very differently over long histories. |
| Key result / formula | Cooper, Cliff and Gulen make the simplest version of the point on broad United States indices over several decades: the return earned between the close and the next open and the return earned between the open and the close accumulate to strikingly different totals, with the non-trading period accounting for a disproportionate share of the long-run gain, a pattern stable enough across sub-periods and indices to rule out a single episode. |
| Why it matters for backtesting | The practical use in Stochastly is diagnostic. |
Key result / formula
Bogousslavsky takes the same decomposition to the cross-section and shows that the premia of many documented anomalies load very unevenly on the two components: some are almost entirely overnight effects, others reverse between the two periods, and the pattern relates systematically to characteristics such as size, liquidity and past return. The conclusion both share is that the daily return is a composite of two series with different participants, different information arrival and different risk, and that studying the composite hides more than it shows.
Why it matters for backtesting
Any strategy on daily bars can be decomposed into its overnight and intraday parts using the open and close already present in the data, and the attribution is worth running before any conclusion is drawn, because the two parts are tradable under very different conditions. An edge that lives overnight requires holding across the close, which means gap exposure that no intrabar stop can limit and a fill at the open rather than at a price the strategy chose. An edge that lives intraday requires being flat overnight, which doubles the number of round trips and therefore the cost. A strategy whose small positive gross edge is the net of two halves that nearly cancel is fragile in a specific way: a small change in the cost of either half flips the result. Continuously traded instruments have no such split, and applying the decomposition there produces an artefact of the data provider's day boundary.
Source
Bogousslavsky, "The cross-section of intraday and overnight returns", Journal of Financial Economics 141(1), 2021, 172-194; Cooper, Cliff & Gulen, "Return Differences between Trading and Non-trading Hours: Like Night and Day", SSRN 1004081, 2008. Primary source