A study of limiting HFT

From Philip Delves Broughton:

These advantages were demonstrated in a recent natural experiment set off by Canada’s stock market regulators. In April 2012 they limited the activity of high-frequency traders by increasing the fees on market messages sent by all broker-dealers, such as trades, order submissions and cancellations. This affected high-frequency traders the most, since they issue many more messages than other traders.

The effect, as measured by a group of Canadian academics, was swift and startling. The number of messages sent to the Toronto Stock Exchange dropped by 30 percent, and the bid-ask spread rose by 9 percent, an indicator of lower liquidity and higher transaction costs.

But the effects were not evenly distributed among investors. Retail investors, who tend to place more limit orders — i.e., orders to buy or sell stocks at fixed prices — experienced lower intraday returns. Institutional investors, who placed more market orders, buying and selling at whatever the market price happened to be, did better. In other words, the less high-frequency trading, the worse the small investors did.

…In a paper published last year, Terry Hendershott of Berkeley, Jonathan Brogaard of the University of Washington and Ryan Riordan of the University of Ontario Institute of Technology concluded that, “Over all, HFTs facilitate price efficiency by trading in the direction of permanent price changes and in the opposite direction of transitory price errors, both on average and on the highest volatility days.”

The pdf of the paper is here.  Here is the conclusion of a Charles M. Jones survey paper on HFT (pdf):

Based on the vast majority of the empirical work to date, HFT and automated, competing markets improve market liquidity, reduce trading costs, and make stock prices more efficient. Better liquidity lowers the cost of equity capital for firms, which is an important positive for the real economy. Minor regulatory tweaks may be in order, but those formulating policy should be especially careful not to reverse the liquidity improvements of thelast twenty years.
There are a variety of significant problems on Wall Street, but this really isn’t one of them.


'There are a variety of significant problems on Wall Street'

The lack of criminal convictions comes instantly to mind. A subject this web site will get around to, as soon as the statue of limitations expires, I'm sure.

Whining liberals love to say that.

Who should be convicted of what?

Is your hypothesis that left wing fame seeking DAs do not want these convictions? Is so you are insane.

But keep the lefty whining up without specifics.

And this site in in on the conspiracy right?

Socialist fool.

Wow. You, Silly, really are an angry little thing, aren't you?..."silly" says it all. Keep drinking that kool-aid...that's it...drink up.

That's what eople l like myself have been saying since the onset of this debate: HFT has more positives than negatives in terms of lower costs for retail investors. HFT simply means trades occurring at a high frequency. There is nothing inherently malicious about that.

The stock market equilibrium equation says that volume components are invariant, meaning that increasing the volume on both sides of the equation by a scalar factor doesn't change anything. A person selling 10 blocks of 100 shares and the another person buying 10 blocks of 100 is no different than someone buying 1000 and selling 1000

I may be coming around to HFT being not so bad, but if it was only "trades occurring at a high frequency" it wouldn't involve so many "cancels," would it?

What does paying for preferred access/speed have to do with "trading at high frequency?" Two separate issues. Trade as quickly as you like, but everyone should have access to the same info, at the same time. Much like corporate press releases (or at least that's the goal).

Everybody has equal access to information if they're willing to pay for it. To get a live feed from an exchange, you have to pay for it. To get articles from Seeking Alpha, you have to pay. If you want to co-locate a server close to the exchange, no problem. You just have to pay for it. If you want free access to all of those valuable things, you're just dreaming.

That's the reality.

I'm just wondering, and not even sarcastically, just where do we draw the line about paid premium access.

Say, NYSE were to implement some sort of QoS on their packet routing network so that certain flagged priority messages (from traders that paid a fee) got routing priority, would that be kosher too?


I would say it's the same place you draw the line on paid access to doctors, lawyers, drivers, food, housing, etc. Information is valuable and not costless to produce, why should it be made available to everyone for free?

If real-time market data, servers located near exchanges, high-speed internet connections and dedicated lines are valuable to you, you'll pay for them. If not, you won't. Why should any of those things be available to everyone for free? I'm always puzzled by why people who don't think they should have free access to, say, DEXA scans and cable TV channels suddenly think that trading resources should be available to everyone at no charge.


Trading resources like buying immensely powerful servers, or large clusters or hiring super duper IQ people, designing custom made ASICs or FPGAs or even employing clairvoyants is all fine in my book.

But in my mind that's fundamentally different from an exchange charging a very high premium for a direct feed, or allowing subscription only access to flash quotes or stuff of that nature.

Sure an exchange isn't cost-less to run: that's why they charge a per transaction fee. Yes Information is valuable, but we don't allow your doctor to sell your medical test results to the highest bidder.

What's a super high price? You can buy a live data feed from NYSE for under $200 per month per trader. Other exchanges are cheaper. Why is a dat feed a product any different from any other product? A market data feed is nothing like a doctor selling personal information to the highest bidder without the patients permission. All the information in a market data feed is public information. None of it is proprietary.

As for flashes, HFT Trader did an excellent job explaining that the order had to marked as flashable. In other words, the traders whose orders were flashed chose for them to be flashed because they thought they would be advantaged by that. If it's a voluntary action by the trader sending the order why should you have a problem with it?

what matters is percentages, not the absolute number of trades or cancellations

When I say "so many" I am talking about scratch rate.

Data is fed in real time into the valuation models that drive order generation. If any input changes enough (and since we're talking about incredibly small edge, the change can be miniscule), the previous quote is canceled and replaced with a different one. If information changes often, then quotes are changed often. It's ridiculous to try to force people to trade on stale information. All that'll happen is spreads will widen to account for the additional risk.

Most "cancels" are market makers re-adjusting their prices to account for new information. Imagine you're making a market in a stock, like virtually every stock it's value is correlated with the market (i.e. it has a non-zero beta). Every time the price of the overall market moves it affects the value of stock you're making a market in. If the S&P index futures tick up you want to re-adjust your bids and offers upwards and vice versa if it moves down. If you don't then an arbitrageur is going to come along and buy your now stale offer. You'll be selling below value and will lose money. (Incidentally since the S&P index futures lead the overall market this is why having a fast connection from Chicago to NYC is important).

Modified are treated in the data feed as two separate events a cancel followed by an add (to protect the anonymity of the trader). If you look at the feed you'll see tons of orders that are cancelled, many within a short period of when they were placed. In reality many of these orders represent much longer-lived "meta-orders" that are continuously modified. Forcing market makers to keep their orders in place for some threshold like 500ms will expose them to stale prices. That means they'll have to widen their spreads. Overall it would serve as a subsidy to arbitrageurs at the expense of general liquidity consumers.

There are traders (market makers, and implementation "algos") who don't really care so much about the actual price they quote at - because
they exist to make the spread (off liquidity) or are trying to fulfill an larger order at the market's "average price" - because that's what they've promised their client. When these algorithms become very simpleminded, they degenerate into "pegging" - they will willingly offer liquidity along with other people, but they don't want to stand on their own (they don't know enough). You won't make money doing this so simply, but what you are doing is still probably net-beneficial (you are offering more real liquidity, just a bit timidly) and pegging-like behavior of some sort is surprisingly common. Some exchanges even offer built-in pegging support; they will move your quotes for you to follow the market according to a simple rule.

But anything like pegging leads to a LOT of cancels - market moves up (maybe because the participants who are actually listening to new information believe it is the right thing to do) and a lot of other people shift their quotes to follow the market. Hence a lot of harmless cancels, in the service of moving valuable but rather dumb liquidity to new prices where it can be helpful.

There is a drawback. If you were malicious, you could use pegging-like behavior to magnify the market message rate well beyond your own order changes (maybe this assists "quote-stuffing" if this is a real phenomenon, which is unclear), and/or set yourself up for other profitable trades by leading peggers to do unwise things ("gaming" - this definitely happens). You move the market around, plenty of people will follow. That could be dangerous (initiating a price move can be unprofitable unless you are careful) and absolutely, unequivocally, illegal. But would it be caught, if you were a bit smart? The risk you face is the probability of detection, and it doesn't matter how high or low that is, because you multiply this risk by the chance that someone on Wall Street will actually suffer any consequence for bad behavior even once detected. We pretty much know what the latter number is.

When they write that the number of messages sent to the Toronto Stock Exchange dropped by 30 percent, I'm guessing a huge portion of this is purely the HFT-to-HFT traffic?

I get the impression from all the debates since Flash Boys started getting publicity that HFT defenders are talking around the argument. Yes, HFT trading can and does increase liquidity and lower transaction costs for the average investor. It also can and does involve a lot of questionable strategies that do nothing to improve market effeciency, increase costs for institutional investors, and interfere with price signals. It would be nice to see some balance in Tyler's defenses, some acknowledgement that not all HFT practices are helpful, and that with the appropriate changes we could remove harmful practices without giving up the main benefits.

if high frequency front-running is occurring the amount is so small that it wouldn't be noticeable (<.01% per 100 lot)

overt front running is illegal though

"if high frequency front-running is occurring the amount is so small that it wouldn’t be noticeable"

That seems very unlikely. The premise of Flash Boys is that non-HFT traders noticed odd behaviors in their order execution, and figured out what was happening when they started to dig into the details of market operations. If the amount is so small it "wouldn't be noticeable", then how was the issue identified in the first place? At least provide a link with some evidence to back up your assertion.

"odd behaviors in their order execution"

like what

You place a market order and it gets filled instantly. The HFT can at best make a few cents from a 100 share block assuming the book is transparent and he can front-run it before someone else does or the order otherwise matches. For illiquid stocks, the trader get very bad fills because of the wide bid/ask spreads. The few pennies being skimmed pales in comparison to how much a buyer would lose from the spread if the liquidity didn't exist.

You are implying that the liquidity can't exist without the front-running. That is false.

Besides which, you haven't really addressed the question: these amounts are clearly noticeable, or nobody would have noticed them. Fairly small amounts on very high volumes become important over time.

It's actually true. If they can't charge institutional/information-based traders appropriate prices, then they have to charge everyone a higher price resulting in higher spreads.

It's not technically front-running because they can't see the books and have no fiduciary relationship with the traders. But they figure out by various signals that a large trade is taking place with informational value, and they trade based on that knowledge, which adds to price discovery. Of course to the detriment of the trader, who is making less money on the backs of the market makers. But why should we favor the institutional investor over the market maker?

Great question. Not to mention that all market makers since the dawn of time have been reading order flow because it's key information. Why has that activity suddenly become evil?

I understand re: front-running. We can use whatever term you prefer ... it's just an easy way to communicate that market makers are acting based on advance knowledge of incoming orders. In this case that appears to be legal; whether it is ethical or efficient, those are separate questions.

I don't think we should favor the institutional investor over the market maker -- I think they should be on a level playing field. It's clear that the markets in question were not designed with the intent that market makers would have advance knowledge of the orders that are coming in. The fact that HFT traders can employ this strategy is a historical accident, it is not the result of a well-designed market.


Why not start identifying trader quotes with explicit identifying flags? So you could cumulatively tally even visible portions of hidden / iceberg orders? Why all this subterfuge?


I don't know if you've ever played poker, but if you have, you know you don't want to show your hand so that the other players can take advantage of the information you give up. If you had to show your hand every time you traded the market would use that information against you. The more information you are forced to give up, the worse it is for you. That's one reason you have so many 100 up markets. You may be willing to do much larger size, but you don't want the market to know that so that you becoming a leaning station (where participants, knowing your hand, will simply jump in front of you by a fraction of a penny, knowing that if they want they can turn around and hit your bid. You do no business at all and your competitors pick your pockets all day.).

You'd be adversely selected all day, that is.


I was being sarcastic. I mean sure order flow is key information for a market maker, but it's a slippery slope. Already some exchanges have proposals to flag retail orders but you can't expect to keep going and keep "bribing" the exchanges to reveal more & more details about the order flows of the informed investors?

Jon Rodney,

We are on a "level playing field". Market makers for time immemorial have tried to suss out information about order flow because you can go out of business very quickly if you don't move your market in response to order flow. Customers (non-MM's) have for time immemorial tried to hide information about their orders so as not to move the price against themselves. There's nothing unethical going on and large institutions are not some poor, helpless shlubs at the mercy of MM's. They have their own bag of tricks.

Moreover, I think it has been pointed out before that not all orders contain valuable information and aren't worth trading in front of. One thing that all market participants should understand fully is that the moment anyone (ANYONE - small investors, large investors and market makers alike) sends in an order or a quote they are giving up information to the market and market participants will use that information to their advantage.


Not to offend anyone, but retail orders aren't generally smart orders, so all that such a tag would mean is "execute without fear. Provide all the liquidity you want. No adverse selection to worry about with this order."

"the moment anyone (ANYONE – small investors, large investors and market makers alike) sends in an order or a quote they are giving up information to the market and market participants will use that information to their advantage"

This is the historical accident I'm talking about. Most investors have no idea this is happening, and they have no way to execute a trade without giving up this information (unless they invest in the appropriate data feed, hardware, and HFT software). This situation happens not because there was thoughtful consideration of whether it makes sense for a few players to have advance knowledge of everyone else's orders -- it happens because it didn't occur to anyone that the infrastructure could be gamed in this way.

You could argue that there's nothing wrong with this accident, but I'd argue otherwise. Yes, market makers have always tried to suss out what kind of order flow will be coming in. In the past they've done this in a variety of ways ... they might take special interest in the trading patterns of key players, or keep an ear out for rumors of large investors dumping a position. Those behaviors are very different in nature and in scale from having advance knowledge of every last market order. Too much market power gets concentrated in the hands of people with that knowledge, and setting up markets where you have to build servers inside the exchange in order to compete on an equal footing is wasteful of resources. Markets should be designed such that a participant who wants to lift an offer or hit a bid doesn't have to signal their intent ahead of time. This is so basic to it seems axiomatic to me.

every trader who knows anything knows his trade broadcasts information to the public. And that this is true regardless of how much he has invested in various technology. That's why trade execution is so important.

The only ones who don't know that is Joe Average retail investors. That's okay because retail guys don't know anything special and their little trades don't convey any information worth paying attention to. Professionals know how to execute a trade so they give up as little as possible to the market.

Also, this is not an "accident". This is simply the nature of the game and just as it is the nature of any negotiation. It has always been this way since the first voluntary exchange was negotiated between two homo-sapiens. The average Joe just doesn't understand the market and thus imagines all sorts of fantastic things to do with modern technology.

It's a given that when someone executes a trade, it broadcasts information to the public *after it has been executed*. That's obviously one of the main mechanisms for the market to absorb new information. If you're trying to say every trader worth his salt knows that his order will be intercepted and acted on between the time he clicks the execute button and the time the order is carried out ... that's flat out wrong.

Many of those practices have nothing to do with HFT per se, but simply relate to market fragmentation, payment for order flow, etc.

I agree -- which is why I don't think the book is a condemnation of HFT in general, but rather a condemnation of specific practices that HFT traders have adopted due to various correctable market flaws.


This Michael Lewis saga might end up being a classic case of a bad argument hurting what might have been otherwise a potentially valid cause.

Agreed, Tyler's posts seem to analyze HFT categorically without making any effort to distinguish between the various types of HFT techniques. I wish there was less debate on whether HFT is categorically desirable or harmful, and more conversation on the prevalence & effects of particular types of HFT strategies.

There's Credit Suisse report on HFT that does a good job of trying to distinguish between HFT strategies that improve liquidity, price efficiency, reduce spreads, etc. vs. HFT strategies, such as quote stuffing, layering, etc., that may lead to increased short term volatility, higher trading costs, and decreased liquidity. While the latter kind of trading isn't exclusive to HFT, it may be more prevalent b/c of HFT.

The "evil" HFT strategies you cite are a de minims proportion of its overall activity. Below is a paper that looks at quote stuffing and finds that in 2010 the average stock was quote stuffed for less than five minutes in an entire trading year. That comes out to quote stuffing affecting %0.005 of market activity. In contrast it's well known that HFT makes up 50% or more of the trading activity of the equity markets. Similarly with layering, commonsense will tell you that it only works with illiquid stocks that have thinly traded books. A liquid stock like MSFT is never going to have an empty bid or ask size. Again there's no way that layering in micro cap stocks is going to account for any significant percentage of market activity that we know HFT constitutes.

So when people criticize HFT, but then later defend it by saying they're only talking about "bad HFT", it would be like me criticizing the pharmaceutical industry by saying they chop people's heads off in the desert and kill police. Then I would defend myself saying that I'm not necessarily talking about all drug companies, just the bad ones like Juarez Cartel.


Great! If it is de mimimus, what's the argument for not stopping it for once?

Because the "good" HFT is just for show, the "bad" HFT is where all the money is made. Hence the defense lawyer tactics ("my client is not a thief, and besides, he only stole a little bit.")

@HFT Trader

But Pharma Companies get criticized all the time! For lobbying FDA, botching up drug trials, sponsoring shaky research, all kinds of reasons.

So also, we should be allowed to analyse HFT & single out whatever crap we find there. e.g. Your statistical arbitrage toolkits sound all great to me, but in the purely latency derived or volume / rebate focused strategies might be good candidates for deeper scrutiny.

The stat arb stuff is just window-dressing. HFT isn't chasing price discrepancies, they're chasing order flow. Which is why you hear obvious bullshit about such strategies playing a tiny role on the one hand, but on the other hand any restrictions on the ability to enter into/back out of trades within microseconds would introduce gross inefficiencies to the market. It's the financial equivalent of cutting in line, nothing more.

Also as an aside Credit Suisse runs the largest dark pool in the country. So there natural predilection is to be biased against HFT to scare investors away from the lit exchanges.

Well, that's rich.

I'm not well read in these issues, but how do we reconcile this evidence in the post with this UChicago paper's conclusion that HFT is more of an arms race due to market design:

"We argue that this high-frequency trading “arms race” is a manifestation of a basic flaw in financial market design: financial markets operate continuously. That is, it is possible to buy or sell stocks or other securities at literally any instant during the trading day. We argue that the continuous limit order book market design that is currently predominant in financial markets should be replaced by frequent batch auctions – uniform-price sealed-bid double auctions conducted at frequent but discrete time intervals, e.g., every 1 second."


My impulse is to think that 1 second auctions are both fair and efficient. Not bad if you can achieve both.

1 second auctions would blunt a lot of the methods that HFT uses to reduce cost. The behavior of order flow on the millisecond level is highly indicative of its informed or uninformed nature. For example order flow that follows the 3-10 ms after the index futures tick tends to be highly predictive of the movement of individual equities. Market makers respond by widening spreads during these periods to avoid being picked off, which allows them tighten spreads during other periods.

If you resorted to 1 second auctions you would no longer have anyway to distinguish this informed flow from the uninformed type. As a consequence ordinary investors would pay higher transaction costs to substitute the type of arbitrageur traders in the above scenario.

That part I don't get. Say transaction costs increased. OK. Spreads are a few basis points worse. For an "ordinary investor" (who we've earlier defined as a typical retail investor, infrequently trading & in for the long run) how much does this matter? As compared to his other transaction costs is this a meaningful loss of utility?

Sorry Rahul, but the real question is "does the small investor matter?" The answer is that he accounts for such a tiny percentage of the volume that he doesn't, yet he is enormously benefited. However, if you're dying for me to go back to the crowds on the exchange floor where we ripped you off in ways you did not know about because nobody wrote about them, then I'm pleased to oblige. These HFT guys work harder, take more risks and improve liquidity better than floor traders providing liquidity at prices that you didn't know how to calculate but that were many multiples greater than any HFT shop.

I suppose the issues are mostly orthogonal.

If HFT is an arms race, then that might be wasteful but it doesn't necessarily rig things against the small investor more than the status quo ante. The resources to fund the arms race might come mostly from bigger orgs -- indeed we would expect that because the bigger orgs account for more of the market.

I take it from the blurb that the "arms race" is the fight for decreasing margins? How is competing down the price of liquidity bad for the market exactly? I don't see how their suggestion will improve on the current model.

Because to achieve you have to spend money on computers, electricity network connectivity geek-hours etc. These costs must be paid for by someones lost profits. In principle that could be a large cost, although somehow it doesn't seem so to me.

That's not bad for the market. That's bad for inefficient competitors. It's very good for consumers of liquidity and for the efficiency of the market and that's all that matters

While 1-second batched auctions may not be the right solution, the Chicago paper has the right diagnosis. This is an arms race, wasting approximately $1 trillion NPV globally, across equities, commodities, derivatives, and currencies. My own preferred solution is randomizing temporal buffers, now being used in fx markets. My paper explaining it is at the link below. Full disclosure: I believe this is the only academic paper discussed in "Flash Boys."

I can never read an anti-HFT screed without suspecting that the author has called in exterminators to deal with his earthworm infestation-- they've got to be harming the lawn somehow.

Maybe HFT is not a significant problem, but these studies cannot address the argument that the benefits are small compared to the cost in terms of talent and resources spent in an (almost) zero-sum arms race.

The estimated aggregated revenue for the entire HFT industry was $2.2 billion in 2013. (Link below) Overall $20 trillion in equity (which doesn't count even count derivatives) is traded a year. If HFT increases liquidity and reduces transaction cost as a proportion of value traded by just 0.01% then the benefits certainly outweigh the cost.

The reality is that a large, developed post-industrial economy needs financial professionals creating liquid and efficient capital markets just as much it needs engineers and doctors.


On the final point I'm not sure I agree that the amount of resources dedicated to financial professionals is rational. The efficient markets hypothesis relies on equal information and ability. But financial professionals have both a greater understanding and greater control of capital flow, it would be expected that they would divert a disproportionate quantity of resources towards themselves resulting in an inefficient allocation of capital.

Was the Canadian policy experiment effectively a unit tax? If so, does the Alchian-Allen effect imply a quality composition shift of market messages?

So spreads for small investors go down. That seems clear. And data driven institutional investors are the ones who are hurt. But I guess one thing you can't measure is, if you can't make as much money as a data driven human trader, that means fewer resources are put into that problem, right? And perhaps that means that prices deviate further from underlying value. That implies that capital is being priced incorrectly and that capital investment will be less efficient. Maybe we're at a margin where that doesn't matter. But shutting out data driven institutional investors is not without social costs (at least in principle).

"And data driven institutional investors are the ones who are hurt."

How does more available liquidity hurt liquidity takers? The entities that get hurt the most are lousy competitors to HFT shops. Also, if you can't manage your large order, you'll get hurt, but that's not because HF traders are doing something new or more nefarious than market makers have always done but because you don't understand the market and you're terrible at managing your order execution. That's not on HFT.

For example, let's take index re-balancing. The indexes publish the stocks going in and the stocks exiting well in advance of the actual trades. Standard procedure used to be that they executed all of their orders on the closing print. That's still standard for most of them. So, everyone would buy the stocks to be included and offer them on the final print. Thing is, so many market participants started doing this that all the edge was sucked out of the trade because when the index fund sent its buy-on-close orders they were met with an avalanche of offers so large that in some cases the final print was a large downtick from where the securities were trading just before the close. So, the customer (index fund) wins and the liquidity providers lose. If the index fund finds itself at a disadvantage it can change its strategy to rebalancing the index over a longer period of time, for instance.

" if you can’t make as much money as a data driven human trader, that means fewer resources are put into that problem, right?"

No, that's not right. What that means is that the machines are much better at providing liquidity than human traders - and more liquidity is better. The machines cost less than human traders, so vast amounts of resources are actually freed up and there's huge investment in the technology.

"And perhaps that means that prices deviate further from underlying value."

Price only deviates far from fair in illiquid securities. Where there is a lot of liquidity price deviations are much much smaller and last for a much much smaller period of time. In fact, over the period since alogrithms have been deployed to make markets, spreads have shrunk and price discovery has improved.

machines are much better at providing liquidity than human traders

To put this in perspective, this study shows that the "much better" in terms of bid-ask spreads was a 0.005 percentage point difference. That's less than a cent for most stocks.

The question everyone then has to ask is, whether in their world view, this is a significant gain in liquidity or an irrelevant one.

Why is "less expensive is better" so difficult to comprehend? Ir you're occasionally messing about with some little retail portfolio, you personally may not think that's much, but if you're actively managing a large portfolio, you're trading millions of shares per year and a small difference in transactions costs translates into substantial differences in returns.

Duh. Yes, but these large guys "actively managing a large portfolio" are the ones crying themselves hoarse about not wanting you HFT guys. Right?

Thank you for the link to the paper. On page 19 it reads:
"realized spread decreased following the reduction in HMAT activity. Consequently, even
though the quoted spread increases, liquidity providers receive a smaller portion of the spread."

This would imply that the reduction in HFT, when evaluated for the market as whole rather than separating out two subgroups that comprise only a fraction of the market, reduced trading costs for market orders. (Note "retail traders" refers to trades that were processed by brokerage firms' retail trading desks, p. 14, and "institutional traders" refers to the 5% of trade identifiers with the largest holding of any one security over the time period assuming no transactions took place outside the dataset. The latter may just capture routers that are designated to deal with "buy" orders. p. 15)

Even this conclusion is, however, suspect. Their results are driven by a mistaken interpretation of what they call "price impact." What they measure by "price impact" is the gain (or loss) received by the party that took the liquidity offered by a limit order five minutes after the trade. By calling this gain "price impact" they imply, and indeed interpret their results as showing, that it was the order that took place five minutes earlier that caused the change in price (which is, of course, total nonsense). Price impact is unmeasurable and this is how data analysts get around the problem.

The authors find that the five minute gain increases from pre-tax to post-tax and claim that this is a problem because it implies that orders are causing more price movements. The reason orders executed through retail brokers "lose" is solely due to the fact that five minutes after the trade retail orders have lost money (by something on the order of .008%) relative to how they performed prior to the tax change. p. 20.

I haven't finished reading the paper, but rule number one of reading research is this: Never confuse what a paper claims with what it actually says.

One question: When the authors say "bid-ask spread rose by 9 percent" that really means (I'm looking at Fig. 1) that the bid ask spread rose by about 0.0005 percentage points, right?

If so, isn't that exceedingly tiny?

The quoted spread increases from about 0.064% to about 0.069%. Note that the quoted spread differs from the realized spread, because the realized spread takes price impact into account.

Thanks! I typed in an extra zero. My bad.

So effectively what they are saying is the Canadian anti-HFT regulation would cost a typical infrequently trading retail investor around a cent on every share he trades? Doesn't sound so bad.

The way I read the paper a retail trader who places market orders that are executed as market orders by his broker will have better overall executions after the regulation because the "price impact" effect outweighs the increase in spread. The problem for retail traders is that retail orders are being executed as limit orders (which is somewhat bizarre, but maybe they're sophisticated retail traders), and limit orders get the spread, but lose due to "price impact."

To be honest all these effects are so small, one wonders whether competition didn't cause the spread to increase just enough to compensate from the actual price impact effect with the result that its all a wash. Liquidity providers are just being paid what their services are worth in both regimes.

Another point that confused me was when they talk about these regulations "significantly reduced the intraday returns of retail investors."

What exactly is an intraday return? Are they talking of someone akin to a day trader that bought and sold a stock within a market day?

I stopped reading before I reached the definition of "intraday order." I would guess from the intro that it's either the actual intra day trading return or, if the trader's orders are net buys/sells, the end of day return. I think these are also adjusted somehow based on "price impact."

"By calling this gain “price impact” they imply, and indeed interpret their results as showing, that it was the order that took place five minutes earlier that caused the change in price (which is, of course, total nonsense)."

This isn't nonsense at all. Price formation occurs because of order flow. Prices don't adjust exogenously they respond to trades. This is the basis of modern microstructure theories. Furthermore the degree to which the market price responds to an order is the very definition of trader informedness. Quite as you'd expect the paper shows that higher price impact is a sign of higher proportion of informed flow, and that in turn costs uninformed investors more, as in the classic Akerloff market for lemons.

Here's a paper that demonstrates quite the opposite of what you're saying that order directionality has a long and slow-decaying impact on asset returns.


"Price formation occurs because of order flow. Prices don’t adjust exogenously they respond to trades."

This statement is obviously true.

But it is nonsense to measure the price impact over the course of five minutes for every single trade. You need to take into account all the other trading activity that took place in that stock in the meanwhile. Why not choose one-minute or 30 seconds or 10 minutes. The measure itself is chosen completely at random.

Also in this paper the same price impact is implied to a small trade as too a large trade. The only variable in price impact in this paper is when the trade took place. Which makes no sense whatsoever.

The question isn't whether trades have price impact, but how to measure it. So you need to defend the accuracy of the measure, not the concept of price impact.

The paper you link to, which I only skimmed, gives on page 42 two different estimated functional forms for market impact, depending on how hidden order executions are modeled. It most definitely does not give theoretic support for measuring market impact as the change in price five minutes after trade.

The assumption that market impact is constant over time or over different assets -- which is necessary in order to use any functional form to measure market impact -- while it may be useful for short-term trading purposes, would be a pretty questionable foundation for regulatory decisions that shape the long-term evolution of markets.

In short, you have in no way rebutted my statement: it is total nonsense to measure price impact by the five minute gain.

"The assumption that market impact is constant over time or over different assets" should read:
The assumption that the functional form of market impact is constant over time or over different assets . . .

So how is this not worth discussing? If this question is at least partly about who should win out: institutional investors or retail investors, then shouldn't we favor institutional investors? Far more Americans have money in a 401(k) than are trading individual stocks in hopes of stocking a college fund. From a public policy point of view, it seems pretty clear to me that you would care more about the needs of institutional investors in that context than about HFT's or retail investors' needs.

I'm not myself a huge fan of the big banks, but our clear public-policy choice in this country has been to make institutional investors the primary sources of non-government retirement income for salaried people. And the stock market is governed by rules that at any given time clearly create winners and losers. So if we're going to fight about the rules, why not have a fight about HFTs, if the HFT business model is at least partly predicated on taking gains at the expense of institutional investors, AKA our preferred public policy choice for dealing with the problem of white collar workers living beyond their ability to provide for themselves through their own labor?

401(k)s are retail investors. They are not trading based on superior information.

Ouch. Good catch. I stand by the larger point, though: this isn't an argument about how to make the market efficient. It's an argument about policy. To act as if opponents of HFT would back down if only they understood market efficiency is, at a minimum, obtuse. Institutional investors want to benefit because they want to benefit. Period. Ditto for HFT guys. If they wanted to make the world a better place, they'd be founding charities, instead of HFT shops.

I disagree. HFTs make markets and that is a valuable activity which radiates throughout the economy. It reduces the risk of investing and the cost of equity capital for some poor guy with an idea but no money and facilitating the guy's ability to create wealth is far better than keeping him idle and dependent on charity. That's the great thing about capitalism: everyone works in the own self- interest and ends up inadvertently benefitting everyone more than they benefit themselves.

You are either high or stupid if you think leap-frogging the order book between existing transactions is "market-making". Or maybe you're just dishonest.

You note the Jones paper did not reference the Broughton paper, which found HFT curtailment actually helped people who use "market orders" as opposed to "limit orders" (that is, most of us including institutions). Further, neither paper seems to discuss HFT extreme effects, such as the Flash Crash of 2010 and others. They limit their analysis to when times are 'normal' and conclude HFT is OK because it slightly lowers transaction costs, which btw for most people are now trivial anyway, regardless of HFT.

"and others"

Such as? When are all these times you keep repeating when market liquidity disappeared. There was one technical glitch occupying 20 minutes in over six years of the post Reg NMS, HFT world. So unless you have all these other examples your argument is extremely weak.

@HFTT and others:

You ignore, as most do, the Jones paper referenced by TC. It shows those that trade at "market" rather than "limit" are actually HELPED by a HFT ban. Only limit order sellers, which are tiny retail customers, are hurt when there is no HFT.

As to your second question, there are dozens of mini-flash crashes EVERY DAY, says this article below. Note some have quit trading due to them. Sounds bad to me, agreed?



There may not have been any major market malfunctions recently, but mini flash crashes still happen nearly every day.

Stock exchanges don't publicly release data about these mini crashes -- when a stock rapidly plunges then rebounds -- but most active traders say there are at least a dozen a day.

Dennis Dick, a proprietary trader at Bright Trading in Detroit, said he stopped tracking them because they happen so frequently.

While none have been as disruptive as the "flash crash" of 2010, or the ones that marred the IPOs of the BATS exchange and Facebook in 2012, they highlight the fragility of markets increasingly dominated by high frequency traders who count on fancy algorithms to make a quick profit.

@ myself: the CNN article does not clearly state some people have stayed out of the market due to flash crashes, so I retract that inference. But mini-crashes nearly every day, as the article states, sounds bad to me.

While not as compelling as an experiment, it's worth noting that the median bid/ask spread for a typical Hong Kong stocks is about twice that of stocks trading in Tokyo or Sydney. Hong Kong is not a small market - it's very big, with lots of listed companies.

One might wonder why the spread is so much bigger in Hong Kong. One plausible explanation that Hong Kong, unlike Tokyo or Sydney or the US or Canada or Europe (the UK imposes a stamp tax, but unlike in HK it does not apply to trading firms), imposes a stamp tax of 10 bps on the buyer and seller of shares. This is a Tobin tax. 10bps, .1%, doesn't sound like much, but it is large compared to the profits that most HF firms make. The tax therefore discourages HFtrading, which may, plausibly, be the reason why the spreads are so large. Large spreads are obviously bad for investors in HK stocks.

If HK ever eliminates or reduces this tax, we would have a very clean natural experiment on the effects of HFtrading. Conversely, if another country were to levy such a tax, we would be able to study the effects on the other side.

Even though I work in a HF trading firm, I would be in favor of a say a one month experiment in the US in which the SEC choose at random half the stocks and imposed a 10 bp per side Tobin tax on them. I predict HFtrading would decline substantially, and we would very quickly see if HFtrading was a good or a bad thing for retail investors and passive low cost index funds like Vanguard's (the kind of funds that retail investors should be invested in). I have predictions on how would turn out, but I think we should run the experiment, let everyone see what happens, and then we can decide how to proceed.

It is very rare in the public realm to be able to run clean experiments that have a good chance for conclusive outcomes without inflicting much harm. So I think it would be great of the SEC seized this chance and ran the experiment. (There was something like this during the financial crises in which some set of stocks were banned from being sold short. Some people believed that short selling was the cause of the price declines. I believe subsequent research showed that stocks that had the ban and stocks that did not went down about the same amount).

@jdm--thanks for your post, good to hear from an insider who is not talking his book. FYI, there was a sort of "indirect" Tobin tax in the Canadian stock market referenced by the Jones paper, which curiously nobody in the Comments section seems to want to discuss. It found that a small indirect Tobin tax that made HFT less profitable was good for sellers relying on a 'market order' and bad for sellers relying on a 'limit order'. So there's the natural experiment you seek.

Even if the bid ask spread in HK is twice that of Sydney, for all practical purposes aren't both spreads so so small that for any infrequently trading retail investor this doesn't make any meaningful difference at all?


Why would you want to pay more rather than less? And, yes, those costs over time can diminish your return by quite a lot even if you don't trade very much. But, retail guys are not the only consumers of liquidity either. They aren't all that matters. Larger spreads also increase companies' equity cost of capital, increasing the hurdle rate of investment. Why would you want more inefficiency and higher costs rather than less even if those higher costs don't impact your personal brokerage account much? The additional costs will filter down to you in other ways. Lack of liquidity in financial markets negatively impacts the rest of the economy.

"those costs over time can diminish your return by quite a lot even if you don’t trade very much."

Someone with a quantitative idea of the profile of a typical retail investor might be able to help here. Say a spread widening of 0.005 percentage points; how much would be the impact on a typical individual investor's portfolio over a ten year period? I'd love to see some calculation, however subjective the choice of a "typical retail investor".

Hey bitch! You want to know where the value is, I'll tell you where the value is. It's up my ass, about to be farted into your skank, butt-ugly faces. You HFT guys are, as they say, going into prison as tight ends and coming out as wide receivers.

Here's an interview with Eric Scott Hunsader, an early critic of HFT: http://www.washingtonpost.com/blogs/wonkblog/wp/2014/04/04/a-veteran-programmer-explains-how-the-stock-market-became-rigged/

One of his arguments is that HFT only works because it ignores the regulations in REG NMS and is, in fact, illegal.

He says, "We don’t want this wink-wink-nod-nod business. That drives smart people out. I could quantify it--everybody is losing a small amount of money. The real cost here is that we live in a society based on rule of law. I’m sorry, but you’re not getting a free pass on this. This is wrong and it needs to be corrected."

Here's a story about other financial parasites at work or at plunder:


I guess, as they say, the best scams are legal.

Streetwise Professor just posted (4/5/2014) a thorough analysis of many of the issues being discussed in this thread. It is entitled "Pinging: Who is the Predator, and Who Is the Prey? and can be found at:


@Scooter. Excellent article, thanks for posting.

This is not relevant to the "debate" brought on by Lewis' book. Applying this paper, and conclusion, to the debate is to conflate the advent and advantages of Computerized Trading and decimalization of spreads with HFT exploiting (and violating) Reg NMS. The latter influence of HFT has arrested the decline in spreads, not reduced them and the "liqudity provision" argument is specious as well given that HFT bids/offers are largely exploratory, ie pinging, and do not trade as revealed by the quote/trade ratio.

Whilst an argument can be made that none of this affects the small investor/speculator it is an extension of the tolerance of (potentially?) illegal activities that poses the largest threat, which is to market confidence.

The SEC, DoJ and FBI are investigating. The fact that it took Lewis' book, (which very few commentators have read), to galvanize reaction will all but ensure a political response...which Reg NMS was and how this whole mess started.

For a reasoned discussion of Flash Boys and how Reg. NMS can be reformed, I suggest you take a look at this joint interview by Reuters with Manoj Narang of Tradeworx and HFT critic Haim Bodek:


Spoiler alert: they actually agree on a lot with respect to Reg. NMS.

Kid Dynamite is going to get fucked in the ass in prison!

Comments for this post are closed