Unintended Consequences

On [Thursday], the Dow Jones industrial average rapidly fell nearly 1,000 points before rebounding 700 points. … Although regulators have not pinpointed the cause, it has become increasingly clear that the stock exchanges’ disparate rules contributed to the market chaos. … As several stocks declined sharply under heavy selling pressure, the New York Stock Exchange, one of the largest pools, stopped or slowed trading in particular stocks.

As part of that process, the NYSE held on to “buy” orders, in the hopes that it could gather enough of them to meet the selling demand. “Sell” orders that came to the NYSE were rerouted to other exchanges, which were not required to slow trading. Those other exchanges were soon overflowing with sell orders and didn’t have enough buy orders to meet them, leading to the rapid decline in prices. (more)

So an NYSE regulation intended to prevent large price drops actually caused a big price drop. Of course one must expect all regulations to have unintended consequences; the issue is whether regulators understand a situation well enough to on net to improve it via imperfect regulations. My guess is that in this case no regulation is on average better, but that savvy regulators expect that the public expects them to “do something.”

GD Star Rating
Tagged as:
Trackback URL:
  • Popeye

    Would any of Professor Hanson’s preferred policies and ideas generate any unintended consequences, or are unintended consequences only a problem for regulators?

    • Roko

      Presumably Robin thinks that his policies would have unintended consequences, since he specifically said that all regulatory policies have unintended consequences, so you should be asking him if he thinks his preferred policies will be good on net, and why.

      • Popeye

        But Hanson writes:

        My guess is that in this case no regulation is on average better, but that savvy regulators expect that the public expects them to “do something.”

        So he only has a guess, not based on evidence. What is there to talk about? We all have our ideologies and deeply held beliefs.

        Moreover, I’m not sure that Professor Hanson recognizes that his policies are regulatory policies; after all, he prefers “no regulation,” which sounds like a nice, neutral, non-political position, untainted by subjectivity.

        Meanwhile “unintended consequences” can of course attach to virtually all types of human behavior. But those who speak of the “law of unintended consequences” are much more likely than not referring to regulatory actions supported by the government. Is that a coincidence? Of course not. This is an idea that is almost always trotted out in defense of the view that certain governmental actions are fundamentally illegitimate.

        Like almost every other comment I make on this blog, this is just a complaint that “overcoming bias” isn’t really about overcoming bias.

      • Jess Riedel

        Popeye, I don’t think Robin Hanson is claiming this regulation has anything to do with bias. It’s just the name of the blog.

  • dWj

    Can I place an order that says, “execute this only on NYSE/Arca”, rather than seeking out NBBO? If not, does it even make sense for different platforms to seek out sensible trading rules? If NYSE implements a trading pause — supposing that, in a vacuum, is a good idea — and that causes my order to get routed to an ECN that doesn’t, that isn’t good for me, for NYSE, or for the stability of the market in general. There might well be some people who would always say, “NBBO”, but we need macroinstitutional rules that allow for competition among exchanges to create good microinstitutional rules. An NBBO mandate may well short-circuit that.

    • Chris

      You can place an order on ARCA and ask that it not be routed. However, if the national best order is not available on ARCA, your order will not be filled.

  • Dee

    The problem with using this incident as a data point is that you should avoid unfairly privileging this data point over times in which the regulations worked as intended. I haven’t seen enough info to decide whether the NYSE’s regulations to slow down trading has, on average, caused or prevented liquidity problems.

  • Lord

    I see this as a case of programming machines with the worst aspects of human failings so we can all reach hell together faster.

  • talisman

    I work in equity market microstructure and am quite familiar with Thursday’s fireworks. It’s flatly incorrect to blame the price drop on the NYSE’s LRP system. To be sure, the discrepancy between the LRP system and other exchanges’ lack of a similar speedbump made things worse, but it’s simply not that big of a factor.

    As a general matter, anything written in the mainstream press about equity market microstructure should be ignored. Try Kid Dynamite for a start.

    This post feels very knee-jerk—reminiscent of others’ negative reactions to Dr. Hanson’s own ideas without understanding the issues very well.

    • I don’t claim special expertise about the details of this event, and was relying on the source I quoted.

      • talisman

        You’ve read articles in the media about you or your work or other things you are expert in, right?

        And you still trust them in areas where you’re not an expert?

  • lxm

    Just for fun…

    It boils down to this: this episode exposed structural flaws in how a trade is implemented (think orphaned algo orders) and it exposed the danger of leaving market making up to a network of entities with no mandate to ensure the smooth and orderly functioning of the market (think of the electronic market makers and high freqs who can pull bids instantaneously as opposed to a specialist on the floor who has a clearly defined mandate to provide liquidity).

    from: http://www.ritholtz.com/blog/2010/05/a-closer-look-at-the-1000-dow-plunge/

    Apparently HFT now accounts for between 50 – 70% of all trades. The argument for HFT is that it provides liquidity. The 1000 point drop suggests the liquidity is provided only when it isn’t really needed. Which leaves the conclusion that HFT profit the HFT owners at the expense of everyone else in the market. I am sure there is a nice economic term for that.

    But I’m sure you’re right. Inconsistent regulation is the problem.

  • My guess is that in this case no regulation is on average better

    I’ve guessed otherwise and, based on my default assumption of neutral consequences on average, argued that the existence of unintended consequences should not be taken as an argument against regulation per se here and here. I should stress that I have low confidence in the estimate of the consequences being neutral on average.

  • Someone tell me what’s wrong with this suggestion to limit the waste and volatility associated with HFT and algorithmic trading:

    1) Any given stock trades exactly once per (say) minute, at predefined times. So any auction for stock XYZ is settled at 3:00:00 pm, 3:01:00 pm, etc. There should be no need to trade more than this often, and it eliminates the need to act at superhuman speeds in response to events, and being closer to the floor gives no advantage.

    2) Automate market makers. Make it so traders just feed their supply or demand curves for a particular stock to a computer to which no one has read access (perhaps by encrypting it), which then determines the Pareto-optimal exchange (“splitting the difference” if there’s no unique solution) and then resolves the auction at the predefined time above. That way, no one has advance knowledge of anyone else’s trade (which they shouldn’t be allowing anyway). (You can still act as a market maker if you want, but there wouldn’t need to be any for trades to resolve.)

    What am I missing?

  • talisman


    The rapid market drop was not 1000 points. The SPX was down 3.5% on the day on fundamental issues (Greece, etc.) The rapid plunge was 5.5% in the S&P, and about the same in the Dow—about 600 points. At the end of the day things were still down 3.5%

    Not sure how to say this nicely: your use of the Dow rather than the S&P, and of absolute moves rather than percent moves, are clear indicators that you aren’t knowledgeable about the markets, and certainly not about market microstructure, market making, and liquidity provision, which is a pretty arcane world. I think it’s deeply contrary to the spirit of overcoming bias to form such strong opinions based on a barely passing familiarity with the field you’re discussing.

    The thing that happened was a high-speed version of October `87, with the S&P down 20.5% on the 19th, but up 5.3% on the 20th and another 9.1% on the 21st. Other than the speed with which they happened, the two events are extremely similar in mechanics, cause, and ultimate result. Blaming HFT market makers is as nonsensical as blaming New York specialists for 1987.

    That isn’t—obviously—to say that the event wasn’t a disaster, and that action shouldn’t be taken to improve the market structure. I’m closely following the CFTC’s and the SEC’s actions and discussions, feel they understand the issues fairly well, and am reasonably hopeful that they’ll move things in the right direction. (I’m fairly libertarian but the SEC has done a good, albeit slow, job with microstructure regulation over the last decade.)


    Waste, yes.

    Volatility, no, if you use shares or dollars traded as your time variable instead of regular calendar time. The former is, if you think about it, the appropriate time variable—if investors and institutions are trading more frequently, this should indeed induce greater volatility in time space.

    Your suggestion #1 is in the right direction. As a practitioner, I’ve been arguing for somewhat similar structural changes for three years. That said, any exchange is free to try such a structure—there are many marginal venues that could try it for essentially no risk—and it hasn’t caught on, despite many attempts to do various kinds of intraday auctions. There would have to be global regulation to get something like this, and as a proponent who has thought about such ideas for years I can tell you that that would be excrutiatingly difficult to get right.

    Your suggestion #2 has more of an “ivory tower” feel. It would require more space and time than I have to explain why it’s not a practicable solution. I will note that one exchange—the little-known “Arizona Stock Exchange”—did attempt a similar structure in the late 90’s.

    Silas and lxm:

    Both your comments assume that it’s easy to provide liquidity to the market, and to find people and firms able to do so. I can assure you that neither is easy.

    Neither a New York specialist nor an HFT market-making algorithm is going to be too excited to buy the S&P when it just dropped 3% at a time scale fast relative to the pace of trading (slow in 1987, fast today), nor should they. In retrospect it looks like an obvious trade but at the time one can’t be sure what’s going on. Again, in 1987 they had hours to puzzle it out and it kept falling. In 1929 the market was down 13% on 10/28 and down *another* 10% on 10/29, and was, on 11/27, still down 22.2% from its 10/27 close. The brave liquidity provider doing the “obvious” trade would have been massacred.

    • Thanks for the reply, talisman, very informative. A few points:

      1) Re: your comment to Ixm, it’s true that reporting percentage change is more meaningful than index point change, but then, the entire financial media, and (AFAICT) traders at all levels internally talk in terms of points rather than %. And it drives me nuts (as an engineer). But talking in terms of points is not much evidence of naivety, or else the entire industry is naive (maybe it is lol).

      2) Re: automatic auctions with input S/D curves: IIRC, I heard somewhere that this is how electric utilities buy fuel (or electricity) in Nordic countries, but haven’t checked into it. And I’m not sure what the biggest problem you have in mind with that idea, but the S/D input aspect already exist in terms of limit orders, which, for a buyer, is effectively inputting the demand curve of “Buy N shares of XYZ at or below price P; buy nothing otherwise.”

      3) I’m probably missing a lot, but I do understand the role of liquidity providers. I just don’t think it’s important for prices to be one microsecond old rather than one minute old.

    • lxm

      Hi talisman,

      I make no claim to be an expert in any part of the market. One look at my portfolio will prove that.

      The only point of my post was to provide another perspective from what was presented in the original post. And to suggest that blaming the event on imperfect regulations was imperfect, too.

      As to your claim that the DOW was down only 600 points, 5.5%, well others disagree. Try this on for size: http://www.ritholtz.com/blog/2010/05/dow-was-down-1250-on-may-6/

      Further, my use of the 1000 number was just a reflection of what was in the original post. Mr. Hanson cited the 1000 number. I make no claim to accuracy, but you do. Perhaps you should overcome your own biases.

      You agree that market structure needs to be improved. And I agree my few sentences on HFT were a stretch. But I have seen others, much more sophisticated than I, argue that HFT is nothing more than a tax on the market, providing no social value other than making the traders richer.

    • lxm

      Hi talisman,

      Here’s Senator Kaufman discussing HFT:

      We cannot simply react to problems after they have occurred. We need the information and resources to identify problems before they arise and stop them in their tracks…[W]e cannot allow liquidity to trump transparency and fairness, and we cannot permit the need for speed to blind us to the potentially devastating risks inherent in effectively unregulated transactions