Thursday, July 30, 2009

Follow up on high frequency trading

My last blog post on high frequency trading got a surprisingly wide reception – and was reprinted for several different audiences. The comments at Talking Points Memo were muted and did not touch the general angst about this issue. That is not surprising – that is a politics driven audience. The commentary at Business Insider was outright hostile with the general consensus being that I am an idiot. The reception on my own blog was balanced – with several people pointing out mistakes I made (and I made a few) and with most talking about the argument on its merits.

Still – one lesson – if you write an article that is not outright critical of Wall Street practice then you should expect to be called an idiot. I got endless emails asserting my stupidity.

All I wanted to really understand what the risks Goldman is taking to make all those trading profits. Sure I know most of them are fixed income – but the balance sheet is still in the trillion dollar range and this crisis has proved that ultimately these balance sheets get socialised. If taxpayers are ultimately on the hook then it is incumbent on taxpayers (at a minimum) to understand and manage the risks that they are taking through their regulatory agencies.

Anyway back to the hot-button issue which is electronic and high frequency trading.

First – let’s discuss front running.

I gave an example of a stock that was bid $129.50, sell $131.50. I bid $129.55 and immediately a computer bid $129.60 over me. I called this front running.

I stand corrected – if someone (even proprietary traders associated with my own broker) bids $129.60 over my bid of $129.55 they are not committing the crime of “front running” – but using the public information in my bid to make a different bid.

It would only be front running if they (a) worked for my broker and (b) organised to have their bid filled preferentially at $129.55.

Of course from my perspective it makes not a rats difference whether they got in front of me at $129.55 or $129.56. That difference is less than 0.01 percent and if you are in my business (strict fundamental investing for medium and long term) it makes no sense to be worried about that sort of percentage. Front running as a crime might have mattered when notional spreads were wide and we did not have 1c pricing. But now the issue is just pricing over me and my clients - not whether those orders are legal or otherwise.

The issue is not front-running per-se (except maybe using my broader – non-criminal use of the term). The issue is that more often than not, as a smallish institution, we are forced to go to the middle of the spread or often cross the spread to get filled. Traders – whether they be bots or the old fashioned screen addicted traders of yore make a good proportion of their profits simply by consistently earning spreads. One old Sydney Futures Exchange trader (open outcry) admitted that (unsurprisingly) was how he made his living. Traders make profits on their short term positions (and sometimes make surprisingly good returns on equity). Those profits come out of somewhere – and I am not averse to the notion that Bronte and its clients pay a small share of them. I am not averse to the notion that clever algorithmic traders effectively “tax” other market participants (and I am not particularly scared of that loaded word “tax”).

Nonetheless the tax is small and the issue is not and should not be a target for major reform. The attention played in the blogosphere – driven in part by Zero Hedge – is simply not warranted.

To quantify: these days the spreads are often 1c on $15 shares (say 0.07 percent) and I can easily get brokerage at a tenth of a percent. I can often buy more than enough stock at the high end of spreads. Even if I lose to the algorithmic traders every time (and I accept that I do a fair bit) then – hey – I am paying 0.15 percent for trading. That is a lot better than it was in the old days when brokerage fees were fixed, large and non-negotiable.

Moreover as a fundamental driven investor the turnover in my portfolio is low. Ideally we will turn over less than once per three years – but as market volatility is high we seem to be getting more opportunities for good switches than that. I can’t imagine how the switching costs for Bronte Capital’s portfolio are higher than they were before brokerage reform even with the “tax” that high frequency traders impose on the rest of the market. [Likewise we trade currency with spreads of 1 point (ie down to the hundredth of a US cent). Only a few years ago we traded with 5 points of spread.]

This issue is – as I suggested – a distraction. There are plenty of real issues for financial reform – and one of the most important in my view being the large and seemingly wholesale funded balance sheets of investment banks. Nobody really understands these balance sheets but ultimately we (though the tax base) are guaranteeing them. That is what the slogan “no more Lehmans” means.

Can we focus where it matters? HFT remains a distraction.


PS. Thomas Peterffy pointed out in the comments to my last post that fairly good algorithmic trading is available to small institutions like Bronte at very low fees. (He should know – he is probably the single most important driving force behind electronic trading globally. However he is selling his own service.) If anything these programs reduce the “tax” paid by ordinary institutions such as Bronte even further. Peterffy thinks his algorithm tends to beat the electronic traders. We have not experimented enough to back that conclusion.

PPS. A reasonable summary is that old-fashioned traders earnt big spreads with quite a deal of sweat and only a modest degree of certainty. New fangled high frequency traders earn small spreads with high frequency and ruthless efficiency. But they are still smaller spreads.

PPS. Quantifications of this as a $20 billion issue are insane. Felix Salmon once took umbrage at my assertion that the pre-tax pre-provision profits of the financial sector in the US were at least $300 billion. That looks like an underestimate. He however swallows this insane number without too much question.


Tom Lindmark said...

Thanks for the sanity.

You may have seen this but Tyler Cowen had some interesting comments.

Finally, is Tyler Durden (aka zero hedge) really Chuck Shumer?

Steve said...

Didn't see you on Business Insider or I would have supported you there.

I'm still reluctant to agree on the "spread tax" concept. Here's my thinking: In the "old" days spread tax was a minimum of an eighth. That meant crossing the spread was much more expensive than today. Then decimalization and ECN's (competition with specialist) were introduced and then bidders could compete with other bidders and sellers could compete with other sellers. This competition is what reduced the spread down to (often) a penny.

So is this competition *causing* the tax or reducing it to a penny? From my perspective it is reducing it.

Does this sound right?

For the record, I never (intentionally) cross the spread. (And I don't do HFT.)

rahuldeodhar said...

"My last blog post on high frequency trading got a surprisingly wide reception"

- I am not surprised both post were insightful. thanx for clarifying - I guess it was needed

Anonymous said...


When did you start your fund? How has it been performing so far?

John Hempton said...

Thanks for the support.

Fund formally started on 1 June this year. Performance so far is embarrassingly good - +14 percent. There are several reasons this cannot continue so I hardly want to market it.


Anonymous said...

The $ 300 million call on bank operating profits was one of your more insightful and valuable posts.

elartistamadridista said...

There's another beef with HFT, which has nothing to do with frontrunning or fairness, and that is systemic risk. I understand Wilmott is a respected name in these matters, and he seems to think they do pose a threat.

Alexandra said...

It sounds a bit like you're saying that if it benefits you (your business) so it must be ok...

Algorithmic trading, in my view, leads to more volatility in the market, which is good for profits, because profits are based on price changes. The faster they change, up or down, the bigger the profit you can make.
Your personal risk is reduced if you either can influence the market enough, so that it will go your way.
The game is even more skewed, when you happen to know what others are going to do and thus have a picture of what's in the pipeline micro/milliseconds before others do.
I think you need to view algorithmic trading, dark pools and front running as one single problem.

Anonymous said...

Returning to the original topic of how does GS derive it's trading profits -
HFT aside, I'm sure you do have an opinion on the source of their revenue. Care to share ?

But What do I Know? said...

I like reading Zero Hedge too, but more for the tenor of the accusations and the general response to market moves--the more strident it becomes the more desperate the shorts, it seems (in the same way that the longs became more and more insistent last year, until they finally through in the towel in Q1. When I short, I want to have as little company as possible.

In regards your HFT comments, you did bring some sanity to the dicussion, and I think you are right about the spreads being much tighter and therefore not minding if someone is taking you for a penny or two on your trades--in the old days it was at least an eighth. The concern that the HFT's and dark pools don't provide liquidity is more troubling, IMHO. Do you think that the market makers and specialists did provide liquidity in the old days (say 10 to 20 years ago), or was this a much trumpeted cover for their pilfering spreads? Do you think there will be a bid in a crisis if there is no NYSE around to make sure that there is. . .

I would love to hear your opinion on this. . .

Brick said...

Whilst I would agree with most of what you have said I wonder if your focus is to narrow. HFT applies to lots of market activity with client interaction being only part of the story.The volume in prop trading in relation to client trades does worry me. What happens after your trade when the momentum HFT's possibly kick in? What happens when ETF rebalancing goes through HFT.
The concern I think is shifting from flash orders and front running to whether the strategies used through HFT are becoming aligned and shifting markets away from fundamentals. The question I think should be whether HFT's facilitate a convergence of strategies with poorly modelled risk consequences. Volume, convergence and unknown risk feedbacks are ringing alrm bells for me.

Josh Kalish said...

It's just a side point, but when you are manually working an order and see you the effects show up in someone else's system, the odds are just as well that someone else was trying to use an algo on the same side as you. That's really just good, old-fashioned supply/demand.

If you talk to your prime broker they probably have a good analytics group that can show you how closely they track the markets and the true impacts of your orders. The algo exececution does a good job smoothing it out.

jupiter said...

what is the last name of --PENTAGON--US Defense Secretary ?

what is the last name of --made founded as MICROSOFT company's head--?

what this tells to you !

Anonymous said...

I'm not familiar with what is avaialble off the shelf for equity traders as far as bots go. I can tell you that for futures there is a company called Ninja Trader that has a futures product that is basically "bots for the little guy". Used to be $40 dollars/month.

You could instruct it to become the best bid as the market was going down and remain the best bid. If someone jumped in front of you you would jump in front of them. In the Russell 200 futures it was obvious when there was Ninja short covering at the bottom of a down move. They were all jumpinng ahead of each other!

Also you could wait until most of the bid had traded out before you sold. So if there was a bid for 1000 s&p minis and you didn't want to sell until there was 100 left on the bid you could instruct it to do this.

All of this also led to companies trying to counteract what ninja users were doing. They would enter bids incrementally instead of 1000 at once, quickly jump in front of new bids to see if they could cause a big chain reaction ect...

This was all available for $40/month back in '04. I can only imagine what has gone on since. Ninja still exists and there is likely something similar in individual equity issues.

You can write to the API at IB. Over on the forums at Elite Trader it isn't hard to find someone to write you a program that would allow you to do the same thing that was done to you on your bids.

Individual bots are easy to come by. The HFT stuff is at a whole different level. Some of these guys are trying to predict what the small bots will do.

To remove the systematic risk the regulators could tighten price limits. They used have a short trading halt if the indexes moved down 2%. That was expanded to 5% so that the CME and the spdr products were both at 5% (next was 10% ect...). The limits were only for downside moves. They can put them on upside moves as well.

Better speed bumps can eliminate the systematic risk.

TheDrugsDon'tWork said...

I’ve heard it said 70% of all equity trade are HFT. Now basic VaR models – as I remember – assume normal distribution (Or at least a distribution which is back tested against historical data).

You’ve got to start questioning the back testing if 70% of the trades are HFT’s and can suddenly disappear. Should the holding period used in VaR models be adjusted in market panics (w.r.t. HFT’s as well as the usual adjustment)?

Surely regulators need to understand the amount of HFT trading is going on so they can make a call on Liquidity and VaR model adjustments???

A similar point can be made about Black Scholes Models, I guess.

Anonymous said...

I could be totally wrong here, but one reason why I think HFT is a systemic risk is that it seem to provide most of the liquidity. I could be misunderstanding the issues, so bear with me.
If I understand it correctly, NYSE (and presumably other exchanges) have rules which stop machine-trading in some conditions (large moves). Would not this rule mean that HFT would have to stop as well, resulting in large reduction of liquidity at the same time as large moves occured = more volatility, more likely more large moves?

John Hempton said...

You are not totally wrong. HFT provides liquidity except in a crisis.

Liquidity in a crisis is a myth. Its part of the reason we have crises...

There was - during the crisis - a lot of mortgage paper bid 20, offer 80 worth in the end maybe 85.

Liquidity is what it is...

The HFT traders add nothing, chew resources and impose a tax on me.

Its just not a big issue. The quantification sucks.


Anonymous said...

Thanks John. I totaly agree that the quantifications suck, my point was more that HFT may not even be that good from other points, being sold as good because it provides liquidity.

I also agree on the liquidity-in-crisis, but if it's true that HTF is 70% of all liquidity in good times, if we say non-HFT liquidity would drop by 90% in crisis, we would still have more than 3 times as much liquidity w/o HFT (10% vs 3%). At the same time good-times liquidity is lower, making it harder to believe the liquidity-is-always there myth.

Maybe if we could force HFT to act as provider of liquidity in bad times as well as in good times, then the "tax" in good times would be acceptable :).

scottlocklin said...

I seem to have said a lot of similar things, in particular in the comments section of my blog. Could HF be 10 billion? Maybe. It's still not a very big number considering the size of the markets we're talking about. Mostly people seem to fear it because they don't understand it, and resent the fact that HF guys consistently make money. They seemed more comfortable with 1/8 or 1/4 spreads in the old days, as it was more obvious how the market makers made their dough.

scottlocklin said...

I'm coming to this discussion late in the game, but I just wanted to post my thanks for writing one of the few clear, reasoned discussions of high frequency trading on the entire interweb. The entire subject has degenerated into a sort of mass hysteria, which has thankfully calmed down a bit in the last two weeks.

Anonymous said...

I think your post, while interesting lacks foresight. Perhaps your situation is not injured by these traders but I would maintain that price distortions caused by front running have imbalanced the market - especially if you happen to be a long -term low turnover type participant who happens to have a stop - taken out buy runs amok altos font running each other and driving your 110 dollar stock to 1 dollar in 15 seconds and back.

Moreover a culture of theft is a very enticing thing for wall street. It is clear to me (rom working there) and just simple impossibility for JPM for have 300+ trading days in a row without losses. This represents a MUCH bigger problem than you allude to and will devastate prices in the non-long direction just as badly as the culture of front-running and fed money has pushed to ridiculous extremes in the long direction. The extremes in the markets are most certainly underpinned by malicious activity pushing prices much further and more extreme than plausible for a normal fear and greed human.

I can understand your point sabot a penny here and a penny there and I write trading systems that do not even notice the difference between a point here or a point there…but none of this changes the undercurrent in the river that reflects total breaches of ethics, information security and price coherence. This is like being a little bit pregnant - one way or the other there is no way to make excuses and the result will be a baby. In this case,. theft is theft and will only create a culture of more theft undermining the markets in ways which are guaranteed to make the little .01 offsets look very very unrewarding and expensive.

Ignorance of this kind of bejhavious is not wise healthy or excusable

General disclaimer

The content contained in this blog represents the opinions of Mr. Hempton. Mr. Hempton may hold either long or short positions in securities of various companies discussed in the blog based upon Mr. Hempton's recommendations. The commentary in this blog in no way constitutes a solicitation of business or investment advice. In fact, it should not be relied upon in making investment decisions, ever. It is intended solely for the entertainment of the reader, and the author.  In particular this blog is not directed for investment purposes at US Persons.