Eric Falkenstein notes that virtually all returns come from overnight price movements. Beyond the implications related to the low volatility anomaly, a consequence I find interesting is that it suggests the continuous pricing mechanism of the day session actually suppresses rather than enables the price discovery process. Basically, it seems day sessions are, on average, almost pure noise, with the closing price being pretty much like the opening price.
It could mean it takes closed markets to absorb new information, which is actually not that unbelievable. The way trading sessions are organised would look truly bizarre to a martian, or anyone discovering the markets from a non financial viewpoint. It is a relic from pre-computer times when markets were run as an outcry auction with people shouting at each other in a big room. While this was a remarkably efficient arrangement in the absence of computers, the model was obsoleted by electronic trading, which, after a few decades of rearguard action by incumbents with vested interests, has come to dominate.
Now despite almost all important markets being electronic, they have kept artefacts from the manual trading era: a trading session which corresponds to working hours at the location of the exchange, which is mainly nonsensical given that people trading financial instruments can be located anywhere, news happen at any time, and the companies or instruments quoted often have global scope. If continuous trading — the ability to trade at any point in time — was particularly useful everything should trade continuously, 24 hours a day, the way foreign exchange markets do. There are no technical difficulties: computer program that run markets and market makers only need to be kept running, and whatever human based monitoring is required can follow the sun as most financial institutions have offices in all countries. It works that way for foreign exchange markets, which trade continuously from Monday to Friday. They still stop at weekends, for no particular reason either, as people don’t stop using currencies at weekends.
If markets were truly 24h continuous, the “all return is overnight” phenomenon could not happen. Why may it be so?
One reason can be that most of the news happens overnight. Given trading days for stocks are typically around 8 hours or so, that covers 1/3 of 5/7 of the days, aka 23% of the time. Add up the habit of publishing company news outside (e.g. after) trading hours to avoid short term funny effects, and that can come to explain some of it.
Another factor I suspect is relevant is that market making algorithms and/or high frequency trading — in so far as they can be distinguished, most HFT is a form or market making — have become so dominant that they often suppress meaningful price moves. It could be at the stock level, where statistical arbitrage will tend to push back information, because it expects the stock to track the most important index it is a member of; or that, simply, enough of the trades are market making (market-neutral agents can trade with each other) that they drown the (single-shot) informed traders in the short term. As many such algorithm will be calibrated during or after the opening of the session and shut down at or before close, this is consistent with the phenomenon observed.
Whatever the reason, it seems the whole sophisticated infrastructure behind continuous trading is redundant. If all return is overnight, we may as well have a daily (or even less for low liquidity instruments) auction and be done with it. This would eliminate the role of (intraday) market makers and their cut — a point auction has no spread — at no loss to investors and final users of capital. The only thing that would be possibly lost is the entertainment value for people following and imagining order in the randomness of intraday moves.
Still, it is vastly improbable that any exchange is going to announce the closing down of their continuous day session to save costs and benefit investors and issuers. Shall we conclude that markets are not (externally) Pareto-efficient?