This is the single biggest problem with the Michael Lewis book is that it implies that this is happening without ever walking through it.
What is actually happening is that:
1. HFT is quoting both buy and sell prices on every exchange.
2. A big buyer sends orders to all of the exchanges trying to buy (or sell) all of the available inventory at a given price.
3. On 1 exchange the big buyers order gets there earlier than on other exchanges, triggering a transaction with the HFT.
4. The HFT uses that transaction as a price signal to change their prices on all the other exchanges and if they are faster than the big buyer, the big buyer cannot take advantage of the lower price.
There is no middleman buying from 1 place and selling to another and driving up the price. What this sort of price signaling does is allow the HFT to quote smaller spreads in the more common case, when 1 participant is not trying to wipe out all of the liquidity in the market at a given price point.
I'm ignorant and only working from first principles, while you seem knowledgeable about actual practice. In your explanation, the fastest HFT knows before anyone else that the someone is buying large quantities of a stock at Exchange A at a price higher than the ask at Exchange B. From this, they know to raise their asking price at B, since they (and no one else) know that there is a buyer willing to pay a higher price.
This makes sense, and they would make some money doing this. But why wouldn't they also buy up all the stock at B that is priced less than transaction price they observed at A? If they do, they (hopefully) get to quickly resell it for a slightly higher price. If they don't, much of the order from the big buyer they are counting on will be filled with lower priced stock from their slower competitors, and they will make less money.
There is no middleman buying from 1 place and selling to another and driving up the price.
If they have the knowledge and they ability, why wouldn't they? I'd think it would be in their financial interest to do so. Or is your point that they are indeed doing what would normally be called "frontrunning", but that the knowledge comes from one exchange while the transactions all take place at another?
Let me first clarify, there very well may be certain HFT that try to do this, but it isn't a large class, because the sophistication and speed required would mean that you could market make more profitably in most instances.
The short answer to your question, is that other HFT prevent them from doing this. It is very rare for a single HFT entity to represent the entire order level at a price. Therefore just because they can react to their own transactions (the only ones that aren't on a public feed) faster doesn't mean they can react to everyone else. Further, just because demand on one exchange implies demand on others, it doesn't require it. So in the case where you read the demand wrong, and you are just changing your price you lose priority at the old price point and therefore there is an opportunity cost, but if you are actually making a transaction you lose both priority and the bid/ask spread which is a real cash cost.
Finally, we need to be very careful about the term frontrunning. It is a specific thing. Acting on the same information available to a counterparty faster than them is never frontrunning. The only time it is frontrunning is if you have a fiduciary duty to be acting on a parties behalf and you don't live up to that duty by trading ahead of orders they placed with you.
In chollida's top comment:
bigger players like Citadel pay the retail brokers for their retail flow so they can trade against the "dumb" money before anyone else. I've heard that most Canadian banks sell their flow to a big US fund that has a chance to trade against it before it reaches the market.
This sounds a lot like frontrunning, but with the behaviour split between two parties. The bank sells the information but doesn't make the trade. The HFT makes the trade, but has no fiduciary duty. It would seem that if the HFT is benefitting, and if this any any way causes the banks client to pay more, it the bank is in breach of fiduciary duty.
I'd want to be a little broader about 'frontrunning' so that it includes any case where you gain knowledge about an upcoming trade before it happens, and manage to change the market in your favor so that you profit and the original buyer pays more than they otherwise would.
If this isn't frontrunning, is there a different term for that more general case?
I am not an expert on equities law or order flow trading, but I believe that the fiduciary duty extends to the order flow purchaser (or at least the broker maintains the liability).
I know for a fact that order flow purchasers are required to conform to the NBBO prices, so in theory they can't adversely impact the price of the security.
The problem with your broader definition of front running is that it undermines basic market mechanics. If a participant expresses a desire before entering the market that desire needs to be taken into account for accurate pricing. This is true of buying cars, houses, and cheeseburgers. Why shouldn't it be true of equities?
The problem with your broader definition of front running is that it undermines basic market mechanics. If a participant expresses a desire before entering the market that desire needs to be taken into account for accurate pricing. This is true of buying cars, houses, and cheeseburgers. Why shouldn't it be true of equities?
I think it makes a large difference how that desire is expressed. If you go to a price comparison website, and click on the link that takes you to the seller's website for the item that is priced $100 less than all others, you've expressed some desire to purchase the item. If the comparison website were to sell their real time click flow to the seller, who then raised their price in the milliseconds before servicing your web page request, you'd probably feel duped. Even if there was no breach of fiduciary duty, it would probably be classified as false advertising.
Still, that's not quite parallel. I think a closer example might be domain names. You go to a site to search for whether a $10 domain name is available. It is, but you aren't ready to buy. The search site sells the search information to an HFT who buys the domain for $10 (or better, just reserves it for a month without paying anyone), and then offers to sell it to you for $100. Have they have increased market liquidity in much the same manner as the HFT for a similar degree of public benefit?
What more positive parallel story do see for cars, houses, and cheeseburgers?
"The search site sells the search information to an HFT who buys the domain for $10 (or better, just reserves it for a month without paying anyone), and then offers to sell it to you for $100. Have they have increased market liquidity in much the same manner as the HFT for a similar degree of public benefit?"
In general, I don't like speaking in analogies about HFT as it is a highly technical area that depends on details. That said, this is a particularly bad analogy because unlike equity shares, domain names are not in any way fungible. pets.com is worth way more than akjalfdj89898afdsfyyy834u384734.sexy and there is only 1 pets.com, so if a market participant makes pets.com harder to acquire they are certainly not providing liquidity.
If you really want a stretched analogy, lets assume that McDonalds has contracted with a purchasing company, to secure enough beef to support their SUPER SAD SEPT SALE of cheeseburgers where they will need an extra amount of beef. If that purchasing agent then tells his brother to go out and buy all the beef he can, and they then collude to defraud McDonalds that is probably out of bounds. Conversely, if the purchasing agent finds that the average price of beef on the market is $1 a lb, but he himself has bought some for $.90 a lb, most people would view him selling it to his client for the market rate as fair. If the purchasing agents brother has some $.93 a lb beef and pays the purchasing agent $.02 a lb to sell to McDonals for $.98 we also usually don't have a problem. Further, if the purchasing agent stocks up on Coke because he assumes the sale will drive up the price of sugar water for his other clients, that also doesn't seem out of bounds to most people. Finally, in our super stretched analogy, no one would expect a savvy farmer who notices the purchasing agent buying up all the beef around to not raise his prices.
What is actually happening is that: 1. HFT is quoting both buy and sell prices on every exchange. 2. A big buyer sends orders to all of the exchanges trying to buy (or sell) all of the available inventory at a given price. 3. On 1 exchange the big buyers order gets there earlier than on other exchanges, triggering a transaction with the HFT. 4. The HFT uses that transaction as a price signal to change their prices on all the other exchanges and if they are faster than the big buyer, the big buyer cannot take advantage of the lower price.
There is no middleman buying from 1 place and selling to another and driving up the price. What this sort of price signaling does is allow the HFT to quote smaller spreads in the more common case, when 1 participant is not trying to wipe out all of the liquidity in the market at a given price point.