What part of the day offers the best stock returns? Does this sweet spot vary by day of the week, time of the month or calendar month? In their July 2007 paper entitled “Return Differences between Trading and Non-trading Hours: Like Night and Day”, Michael Cliff, Michael Cooper and Huseyin Gulen use transaction-level data to decompose returns for individual stocks and exchange-traded funds (ETF) into four time intervals: Night (4:00 PM to 9:30 AM), AM (9:30 AM to 10:30 AM), Mid-day (10:30 AM to 3:00 PM), and PM (3:00 PM – 4:00 PM). Using intraday price data for the period 1993-2006, they conclude that:
- Over the last decade, overnight returns are strongly positive, and daytime returns are close to zero and sometimes negative. Specifically, during 1993-2006, average overnight returns for the individual S&P 500 stocks range (depending on the averaging method) from 0.028% to 0.048%, and average daytime returns range from -0.028% to 0.002%.
- This overnight/daytime effect holds for individual stocks, equity indexes, futures contracts on equity indexes and ETFs. It is robust across exchanges, with the average overnight-minus-daytime return spread 0.18% for QQQQ, 0.061% for DIA and 0.068% for SPY. (See the table below.)
- Average overnight returns are consistently higher than daytime returns across days of the week, days of the month and months of the year. Average overnight (daytime) returns for the S&P 500 index are positive (negative or close to zero) for all five days of the week. Overnight returns are strongly positive from Friday close to Monday open, reversing results from older research. For example, the average return from Friday close to Monday open for the individual S&P 500 stocks is 0.042%.
- High opening prices that decline in the first hour of trading drive much of the effect. For example, the average return for the individual S&P 500 stocks during the first hour of trading ranges (depending on averaging method) from -0.012% to -0.036%.
- The overnight/daytime effect is not due to portfolio risk (the standard deviation of overnight returns is consistently less than the standard deviation of daytime returns). Nor do timing of earnings announcements, the growth of ECNs or decimalization explain the anomaly.
The following table, taken from the paper, lists the average return (in basis points) for 14 high-volume ETFs during the four intraday periods defined above. It shows that average overnight returns consistently beat average daytime returns for all 14 ETFs. In fact, average daytime returns are negative for 12 of 14 ETFs. The table also shows that the first hour of trading is the least profitable of the four periods.
The paper offers similar tables summarizing the overnight/daytime effect by day of the week, by day around the turn of the month, by calendar month and by year for 1993 through 2006.
In summary, both individual stocks and broad funds have, on average, appreciated overnight and stalled or declined during the trading day over the past 14 years. The first hour of trading may be the worst hour.
