Thanks to Eddy Elfenbein for sending over the data on how the market does over multiple nights when the market is closed. Unfortunately, the data is skewed because of 9/11, where the market was closed for seven days, and the change from the close to the open was -4.59%. What should be done with that data point? When the market closed on Monday 9/10/01, traders expected that the market would reopen as normal on Tuesday, but it didn’t. The seven day hiatus was not planned, so traders treated it as a one night gap on Monday, but it opened as a seven night gap the next Monday, with negative results.
Now, if you throw out the 9/11 data point, the average price return over a one night gap is 0.005% over the last eight years. For a multiple night gap, the return is higher — 0.012%. If you include in 9/11, it is lower — 0.002%.
But what of dividends? Where do they belong? They belong to the nighttime returns, because on the morning that a stock goes ex-dividend, on average the price drops at the open to reflect that. Now, assume a 1.5%/year dividend rate (rounding, the actual is a little higher). Now the returns for a one night gap are 0.010%, and for a multiple night gap it is 0.024%. Even counting in 9/11, the result is 0.014%, higher than the single night gap.
One commenter on last night’s post commented that it might not be the risk of holding stock overnight as much as the possibility or occurrence of news flow. Before the fact, risk and potential news flow are similar concepts. After all, how does risk shift, but often through news flow changing the opinions that people hold regarding assets?
For a long term investor like me, this all doesn’t matter much. I’m not going to buy a bunch of futures contracts or ETFs near the close and sell them into the open. Still, this could be another example of a market anomaly that stems from the perception of a risk which does not occur on average.
My premise for the discussion of day trading is that the statistical majority of day trades are leveraged positions in liquid stocks where risk control is done through selling with tight stops. For that type of trading, holding positions across potential gaps in the market (always possible when the market is closed) is extra risky (compared with, for example, holding a position across a Fed meeting that happens in the middle of trading hours). So I took the pure risk/daytrading hypothesis to be one that prices are slightly discounted at the end of the day because day traders withdraw their liquidity at that time, while the news flow hypothesis is that the positive returns from outside trading hours simply reflect a large portion of material information released outside of market hours. To test the difference one needs to somehow apportion material news flow to within market and outside of market hours and then see whether the proportion of positive returns outside of market hours exceed or match that proportion.
In an abstract confluence, The Aleph Blog Comments sounds like ABC, but in references to long term versus short term rewards, the core theoretical reference is actually the Aleph set, a term used by mathematician Cantor (the Aleph set is also called the Cantor set). Hence the confluence.