Stock market volume is a sadly drooping southward curve, interrupted by violent spikes. Volumes and turnover rates are now back to levels of the late 1990s. If high-frequency trades and ETF-arbitrage are taken out of the equation, turnover might well be back to where it was in the 1980s. So algorithmic trading – for example, an instant burst of computer-generated buy orders based on a headline that hit a fraction of a millisecond earlier – can have a big impact on the market overall.
Art Cashin, in his “Market Commentary” for UBS on yesterday’s market action – “Bulls Turn Tide, Maybe Helped By Artificial Intelligence” – put his finger on the power of those algo trades:
U.S. stocks followed the lead of European markets and moved lower after a mixed opening. A key negative influence was further weakness in crude, which had Chevron and Exxon subtracting over 20 points from the Dow by mid-morning. Also hurting was a continued ratcheting up of interest rates around the globe.
As Europeans moved toward their close, the bulls on this side of the pond regrouped. They got help from several sources. The losses in Europe had narrowed. The selling in crude stabilized and buyers began to show up in Apple. Those things and others turned U.S. markets higher and shortly before noon, the Dow and Nasdaq Comp had punched into plus territory.
Then around 1:00, the indices all spiked higher. The spike came with virtually no trader buzz. The buying seemed virtually all electronic and was almost instantaneous.
Traders scrambled about for a few minutes, trying to find the trigger. It was simultaneous with the ten year Treasury auction but that was unlikely to cause stock buying, since rates actually went up on the auction.
At the time that spike sprang from nowhere, there was a headline about a poll indicating that Scotland would not secede. Given the silent burst nature of the spike, traders theorized that it might have been the result of one of the new headline reading algorithms, which scan the newswires and act when they see a key word or phrase. Welcome to the new world of trading.
In a bullish environment, where momentum is more or less consistently up and the default action is “buy” and “buy the dip,” these algo trades drive the market relentlessly higher. It works as long as the bullish case is being made everywhere to maintain the momentum, which it is, even if the data they trot out to support their bullish case is negative, and sometimes even scary.
Michael Hartnett, Chief Investment Strategist at BofA Merrill Lynch, for example, explained his bullish stance by pointing out that 81% of free-float equity market cap and 56% of world GDP were “currently supported by zero rates” at a time when the Fed is contemplating the end of zero rates; and among other things, “War and conflict currently affect 11% of the world’s population.” So buy, buy, buy.
But if the momentum turns south and the environment turns bearish, the default action by algos becomes “sell.”
Algos will have a field day cherry-picking negative headlines out of the myriad of mix of positive and negative headlines hitting the internet – just like they’re cherry-picking positive headlines now. They’ll be inundating the market with wave after wave of instant sell orders. Just like algos create spikes in a bull market for no apparent reason, they’ll pull the rug out from under a down market, also for no apparent reason. The mechanics are the same, the effects are similar, only the direction is different.
As Art Cashin said, “Welcome to the new world of trading.”
Algo trading makes up a big part of the otherwise puny trading volume. Corporate buying is another factor, both for share buybacks and for M&As (up 60% this year). But LBOs? That’s when private equity firms, the ultimate smart money, buy the shares of publicly traded companies to take them private. And LBO volume has plunged to the lowest level since crisis year 2009. Read… ‘Smart Money’ Unloads, Sits on Cash, Waits for Stocks to Swoon
Enjoy reading WOLF STREET and want to support it? You can donate. I appreciate it immensely. Click on the beer and iced-tea mug to find out how:
Would you like to be notified via email when WOLF STREET publishes a new article? Sign up here.
Somethings that was drilled to me from Investment class I took as part of MBA was that volume proceeds price and the big “mo” works both ways.
Suspect the algos are programmed in classic momentum driven mantra which has worked for them. Wonder how algos will behave once indices nose dive few %. Will the algos in unison exit for the same door in rather heated and no doubt crowded selling at market or even lower market price in frenzy to “preserve” the capital or hold tight (NOT)? Hear that corporate buyback throttled down in last few months too removing some demand to boot.
And when the market swoons down in alarming volume then the algos boys will be blamed AKA scapegoat followed by the Fed investigating them. Of course the Government Sacks of the world who has been front running and trading against their customers won’t be touched thanks to TBTF. Yep and they will indeed made whole via TARP II.
Bulls – please exit in an orderly manner!
I don’t believe they can exit in an orderly fashion. When a trigger is tripped the resulting buy or sell is automatic without any human interaction. That’s what has produced flash crashes. The major exchanges are supposed to have circuit breakers to stop catastrophic sell-offs, but sometimes the sell-off is well underway before the circuit breaker can respond. The so-called “dark pools” don’t have circuit breakers which seems to be at least part of their attraction.
It was the fear of this exact problem that caused Reagan to institute the President’s Working Committee on Money and Markets (aka the Plunge Protection Team). After the infamous five hundred point drop in October of 1987, the Gipper swore that wouldn’t happen again. The problem is that the markets are much larger now than they were then and whether anyone could stop a panic remains to be seen.
The algos in any financial company are changed constantly sometimes on a daily basis. It is a necessity when you mark to model to change the model so you don’t have any down days.
I think constant tweaking of the algos may create more problems down the line. Sometimes one should allow or give time to prove itself. Something I learned from feedback/feedforward process controls and mistake one may make on coding without thorough testing.
Fine tuned and frequent changes to the math/stat formulas, hypothetical models, etc. may be creating a bigger algo monster designed to extract the most on rising market in frenzied attempts to outbid/front run fraction of cent in split second. Concern is how does this fine tuned algo fare when market tanks?
They only care about what the position is at the end of the day. They can’t afford to have down days.