Broken Markets tries to be informative of the problems with modern stock markets. At times, the book is informative. Unfortunately, the authors cannot make the picture clear. Their arguments sound correct at first, but after thinking about it, they contradict themselves many times on the same page (i.e., the authors are not coherent) Examples:
- authors claim that HFT provides liquidity to the top 5% liquid stocks, but hurts the remainder 95% because they are unwilling to create a market for the less liquid stocks. On the same page, they also claim that the specialists (human market makers) are driven out because they cannot compete on speed. But if both these statements are true, why don't the human traders make markets for the 95% when HFT is not present? Makes no sense.
- authors complain about the parity priority rules. Also, they say that old specialist systems are better and praise NYSE over NASDAQ, BATS, EDGE, etc. Upon looking into the issue, only NYSE has parity priority rules, and these rules were shenanigans from the past. In other words, the authors are using bad policy from the past, claim it's used on the newest exchanges (BATS, NASDAQ, Direct EDGE) when it's not, while praising the old NYSE. This is straight up mis-representing the facts.
- authors correctly state that there's no SEC rule against flash orders, and claim they are alive and well. This is a lie. After searching on Google, I found out that NASDAQ and BATS policed themselves immediately back in september 2009 and Direct EDGE ended the practice in february 2011 (for stocks). ARCA never had flash orders. (NOTE: flash orders are present in the options market)
- on medium liquidity stocks, the authors claim that the HFT is moving the bids/asks up and down without any trades happening, and this pushes the price the mutual fund is paying in execution. Also the authors claim HFT only hold stocks for seconds. Third they claim that the previous system where the investors would trade only between themselves was offering better prices. These 3 statements create a contradiction. Here's how: say investor A is buying and investor B is selling, while you have HFT (market maker) C. Since C's holding time is so small, it must be that it trades with both A and B within a short period. This means the most C can do is make the spread. Since spreads are smaller, the system is more fair than the old stock market system. This contradicts the main claim of the book. What the authors should have said is that there are different strategies. In particular one that tries to capture the spread and minimize the holding time, and another that plays on momentum. The authors are confusing by taking some properties from some strategies and combining with properties of others. The result is incoherence.
- the authors correctly state many times, that in times of stress HFTs don't want to provide liquidity (too risky). They say the specialist system was better because they were forced to make a market. In reality no rule can force the market maker to make markets. As 1987 proved, the specialists simply didn't pick up the phone. So, it was way way worse... the market became totally illiquid. During the flash crash, the spreads were huge (about 1%) and the price move was tremendous, but there was a market (not all HFTs left). The 1987 crash took days to recover, while the flash crash recovered in 10 minutes. Also, as a magnitude, the 1987 crash was much bigger (44% from watermark, 22% in one day, and 30% in consecutive day loss); the flash crash is generally quoted at 9%. The modern market proved to be the much more robust system by comparison. The authors fail to recognize that. In any case, this should be an example on how modern market makers are better than the old time specialists. The authors draw the opposite conclusion.
- the authors keep calling HFTs as "wall street". In reality these first have little to do with banks. Yes, there are HFT groups at Goldman and Barclays, but there are maybe 100 different HFT firms, and some are not even located in NY. (Getco in Chicago, Knight in Jersey City, Tradebot in Kansas, Tradeworx in Red Bank) More important is that these firms are technology firms and in structure have nothing to do with banks. The authors are trying to piggy-back on the layman hate of banks to create hate for HFTs. Also, the HFTs are very diverse and very disorganized. The author is claiming the HFTs were using lobbying, but in reality the SEC decided to push for new rules because the old system ware extremely unfair and opaque. A couple of big HFT firms (Getco) are lobbying and try to create barriers of entry for the competition. This is nothing compared to the lobbying from banks and old brokers like the authors.
- the authors mention Kweku Adoboli, an UBS rogue trader as a bad example on HFTs. I'm not sure what they are trying to accomplish here, as this trader has nothing to do with the automatic market making of HFTs, and even if he had it'd be an example on how to lose lots of money. The authors are simply trying to link HFTs to a scandal, when no link is present. This is intentional, malicious and manipulative.
- authors claim that the fact that a hidden order matched on NASDAQ is identified with the same number if is filled multiple times (lots of partial fills). The authors claim that this is a big disadvantage because the other players can figure out there's a lot of liquidity there because the same number appears many times. This is just stupid, as the player that puts the hidden orders can just put many small orders of 100 shares, and different ID numbers would be generated. The authors are either stupid or playing stupid here.
- the authors correctly mention that the computation of indexes are based on the prices on the exchange. They claim that because of fragmentation, the regular investor doesn't see the accurate index and HFTs can compute their own based on all the prices. This is simply a stupid argument, because of reg NMS the market only has one price, and any advantage an HFT would have would be in milliseconds. The investors are not doing strategy based on changes in milliseconds. Also, since it's an aggregate, a lot of data goes into the computation of the index. Even sampling would be more than good enough to make decisions based on the index for long term investing. (unless you're doing microsecond level arbitrage, which obviously investors are not doing) In other words, the authors are claiming the investors are at a disadvantage when in reality they are not based on their example.
- colocation: the authors say it's unfair, and claim it's something new. In reality colocation appeared with the first market during the time of John Law in the 1700's when prices were different at different parts of the same square. In the 1800's the Rothschild family send their children to different markets basically using colocation. There was always a need to get close to the market. In the pit, many traders were selected from football linebackers simply because they were bigger, more visible and able to jerk others around. Needless to say we used to have a much more unfair system. Now, everybody who collocates in a datacenter has cables of the same length, (i.e., no advantage to anyone) and it's relatively cheap to colocate compared to the fees of the past.
- algos that read news: it's just stupid to complain. if they make mistakes they lose. they take risks. it's their problem. the authors claim the algos that read news destabilize the market because they can make mistakes. humans can make mistakes too. if algos make mistakes, doesn't that create opportunities for others?
- stop-loss: the authors claim many times that these orders may be triggered by temporary swings (flash crash is an extreme case). These are traditional types of instructions given to brokers. The authors never take issue with the stop-loss orders, which obviously accentuate the swings (i.e., sell when stock drops) and increase volatility. If you want to minimize volatility, stop loss orders are obviously bad to the market. No mention that these orders affected the flash crash in the sense that it made the drop bigger.
- the mention of BATS IPO is laughable at best. BATS chocked on their own technology. Nothing to do with HFTs.
- the mention of the lack of IPOs is hilarious. The authors claim that it was automatic trading the reason why we had few IPO's in the last 10 years, and also claim it costed the US economy 20 million jobs. Yeah, it had nothing to do with the fact that the banks sold a lot of snake oil during the tech bubble... and also nothing with the fact that it got expensive to go to IPO based on other regulation that has nothing to do with HFT. I'm not sure what the authors were smoking, but it must have been very potent.
On top of these the authors completely fail to recognize how much better the market is for some stocks, if not all. Say trading IBM, the spread is at times 1-2 cents and it's hard to move the price, but with the specialists, it was 100 times bigger. The costs of trading IBM went down by a factor of 100. In other words, up to 100 times reduction in costs. On trading bank of america, the spread is always 1 cent, while in the past it was force to be at least 12.5 cents, and it was at least 25 because of the gentlemen agreement they had between them.
The authors are experienced old brokers/traders from the old Instinet that cannot compete and frankly understand the market microstructure only superficially. The bottom line is that they are bitter they cannot compete with the HFT, and don't seem to have the skills to build good algos for execution.
Read more ›
