High-Frequency Traders are Eating Investors’ Brains. For Free!
It can be tough to articulate a cogent argument against high-frequency trading in the context of highly liquid, efficiently functioning securities markets, but I think Rajiv Sethi has done so (riffing off Michael Lewis’s typically scathing and revealing article on Goldman’s recent bad behavior).
In brief, in my words:
1. Value investors are trying to buy shares in truly productive firms, which activity indirectly, over the long term, allocates real resources to those firms. This is A Good Thing.
2. When these investors trade in any volume, those trades are broken into multiple batches. A buy order for 1,000, 10,000, or 100,000 shares rarely finds a single sell order for the same amount at the same time.
3. A micro-instant after the buy order appears, HFTers jump in front of a bunch of the value investor’s trades, pushing up the price that the value investor pays.
In Sethi’s words:
their private information is effectively extracted early in this process
Here’s his explanation at more length, emphasis mine:
Effective prediction of price movements, even over such very short horizons, is … essentially a problem of information extraction, based on rapid processing of incoming market data. The important point is that this information would have found its way into prices sooner or later in any case. By anticipating the process by a fraction of a second, the new market makers are able to generate a great deal of private value. But they are not responsible for the informational content of prices, and their profits, as well as the substantial cost of their operations, therefore must come at the expense of those investors who are actually trading on fundamental information.
It is commonly argued that high frequency trading benefits institutional and retail investors because it has resulted in a sharp decline in bid-ask spreads. But this spread is a highly imperfect measure of the value to investors of the change in regime. What matters, especially for institutional investors placing large orders based on fundamental research, is not the marginal price at which the first few shares trade but the average price over the entire transaction.
Since value investors have no choice but to place buy and sell orders that HFT algorithms can see and react to instantly, they are inevitably giving away their private value/fundamental research for free to the HFTers. They can’t not. This cost/extraction inevitably (incentives matter) results in less fundamental research, so poorer allocation of resources.
The HFTers aren’t paying the cost of that externality.
If you’re looking for solid economic, theoretical, justification for a Financial Transactions Tax, I don’t think you’ll do much better than this one from Sethi.
Cross-posted at Asymptosis.
Maybe there is an argument to made that HFT is effectively micro-insider-trading. They are collocated with the market databases, and therefore are able to trade on information before it could reach the remainder of the world, even at light speeds.
@J.Goodwin:
I think that’s an excellent way of viewing it. It’s quite literally “insider” trading.
The archetypal “efficient” “free market” requires that “all information is known.” That information must include all the information that the market itself provides.
But that’s impossible if all information is not “known” at the same time, by all market players.
Here’s the counter-argument: a fire-eating efficient-marketer might say: “Well why don’t all those value investors just invest in HFT funds.”
If that happened, all the information entering the market would then be delivered by these algorithms based on whatever data sources they use. It’s possible that those algos would do a better job of sniffing out true value than traditional methods.
Always trying to challenge my own thinking… I’ll stop on that right now, but I do think it’s a compelling comeback that merits careful thought.
IT is dishonest, ban it, end of story.
This type of skimming is rent-seeking in its purist form.
I don’t think your HFT fund value-seeking idea holds up. These algorithms can’t be VALUE based. They have to be market action, quick reaction based. They add no value of any kind to genuine market activity. The average holding period on the NYSE is less than 1/2 minute. Value has absolutely nothing to do with it.
Look at trade volumes today compared to 1950. For the S&P 500, daily volume Jan 1950 was 1.2 to 2 million shares. That was not a choked off market. Volume so far this month was 2,529,300,000 to 3,775,170,000. That’s 1000 times higher. Does it make any rational sense?
I have no problem with a transaction tax. But I think a randomly generated trade delay period of from 5 minutes to some arbitrary upper limit less than, say, 4 hours, would be far more effective.
No reason not to do both, of course. I think the transaction tax has merit even without a deterring effect on instant trades.
Cheers!
JzB
@Jazz and J.Goodwin
I’ve thought more about this.
1. It seems perfectly plausible, even likely, that good algos could do a better job at determining company values, allocating financial capital, than people do. They could draw on all the same sorts of information, from earnings to twitter sentiments, to do that.
2. But: that does not mean that HFT, though it uses these algorithms, has any economic value. It takes very few trades and traders to “get the price right.” At the very least, all the HFT machinery and staffing is wasteful overhead, throwing away massive resources on ventures that are, in aggregate, zero sum.
3. An FTT or other method to discourage HFT does nothing to discourage algorithms.
So you can throw out the dirty bathwater with no harm to the baby. Quite the contrary, in fact.
So here’s the question:
Does HFT penalize value-seeking algos at the expense of other algos?
If yes, that would result in inferior resource allocation.
Steve –
Sounds about right.
Cheers!
JzB
A delay of 5 minutes, uniformly implemented, and a standard for how long an offer is required to stay open that is longer than 1 minute might prevent HFTs from working effectively. Any span that permits external analysts/computers that are not on site to be able to see the offers and respond to them ought to prevent the insider-style information that is created here from being leverageable (apparently not a word).
I always had the impression that HFT was about bait and switch in that it involves offering transactions that cannot be accepted, but I guess they also provide a means of front running. Basically, HFT provides a mechanism like the old NYSE floor specialists, except in reverse. Does anyone else here remember specialists? They were exchange members who specialized in specific companies, so they’d hang around the appropriate trading areas. Their goal was to maintain an orderly market, balancing supply and demand. In exchange, they were able to pocket a piece of the spread. You could make good money as a specialist whether the market was going up or going down, and you could go home at night satisfied that you were doing something important and useful. The goal of HFT is to introduce disorder into the markets, hiding real bid and ask information and profiting from the resulting volatility.
Do all trades have to be run through markets? Say I want to sell AAPL in a Fidelity account, and someone else wants to buy. Do they have to go to the NASDAQ to do the trade or can they put their two clients together, and then report the price?
Say a stock owner has a problem and sues. Who is making sure that their suit even has a day? The government. That function has a cost and should be borne by those that profit off it.
Winstongator, that’s what is known as dark pool trading. You can agree the price privately, preventing that transaction’s future existence from affecting the price until the trade is actually completed. Once the trade happens, it has to be reported.
Say you want to sell 10,000 units at prices of at least $10, if you put that offer out there, it will fill in little bits, maybe 1,000 units will be bought at 10.50, another 2000 at 10.25, and some more at 10.00, but you still end up with a number of unsold units, and your price may need to be dropped. The average price of the transaction could be less than $10, and you also have HFT algos that will see that you want to sell 10k units, and they will attempt to buy your units for 10.24 (or some other price lower than the bid but higher than the ask) then sell them to someone who is willing to pay 10.25 for them (because they can see that bid before you can see it).
If you are willing to sell at $10 and are aware that someone else with is willing to buy at $10 (who hasn’t listed an offer to buy), you can just directly agree to sell them. In some cases, the broker might buy them from you as well at that price, then assume they can clear them.
Markets are run in such a way as to be as non-transparent as possible, to allow “professionals” to make money. Meanwhile they are barely as transparent as required by law to give the impression of fair dealing.
Is it only the location of respective computors, HFT’s and Exchange’s, that allow the HFTs to gain their nano second of price information on trades? I vaguely recall some other reason for the slight information advantage that results in far greater than slight and assured profits.
Due to their proximity they can also put out microsecond almost immediately cancelled offers to make price discovery possible without committing to trades. They can go up and down the price scale and see who bites without actually buying or selling.
Proximity is probably the biggest part of what makes it work at least in today’s context of what happens. If they weren’t allowed to be on site, they’d park themselves next door instead, or in the same city, or whatever…if they weren’t allowed in that city, they’d put themselves on the backbone just outside it.
If they were required to leave offers open for more than a fraction of a second, all the sudden you’re in totally different territory where they have to really change their strategy, and the proximity effect becomes less important.