Richard Olsen: You're one of the world's experts in the field of market microstructure. What do you think are the big issues now in terms of market liquidity?

Joel Hasbrouck: First, I have to qualify being an expert. Sometimes being an expert really requires you to acknowledge your ignorance. This is an area in which we really know very, very little and are just starting to come to grips with some of the issues. I think one of the most important things you asked me to think about was the idea of rules of liquidity, and why do they change over time.

When we think about liquidity and how to trade in a market, what we usually try to characterize - both for purposes of research and in practice - are traders who are actively in the market. That's where their focus is and they stand ready to transact or potentially transact at reasonably short notice. What distinguishes these people is that they have, to a certain extent, thought about what they might like to trade, the circumstances under which they'd like to trade, and prices. In other words, these are people for whom it's a very small leap to bring them really into active participation in the market.

When we speak of a "pool of liquidity", though, we're speaking much more broadly, across the entire economy, of people who might be drawn into trading if the prices and the circumstances were right. Now, when people think about trading costs, they think of commissions, they think of price impacts. Nowadays, we also think of asymmetric information and the costs of being bagged - but I think one of the largest costs of trading actually occurs far before that. It's at the point where somebody thinks or tries to start thinking about whether they should enter the market at all.

As individuals, think of all the property, think of all the securities that we own. Because we're not actively trying to buy or sell, we don't have to think about how we would buy or sell. We don't have to think about how we would put in orders, for example, for things that we're not likely to trade, and all of that thinking is very, very costly. We're not going to think about a price, for example, if we don't think there's a very good chance that our order will be acted upon. For example, in all markets, you don't observe very many limit orders priced away from where the market currently is. From a classical economic viewpoint, this is sort of surprising in that you would think, for example, that as the price of a security or an asset or a good drops, more people would be demanding that good. So you would think that as the price declined, you would see the bids on the buy side of the book simply grow and grow and grow, getting deeper and deeper and deeper. Well, this happens with a narrow range right around where the market price currently is, but if you go much beyond that, the limit order book gets very, very empty.

R: So, in a sense, you've made a dramatic statement, if I understand you correctly. You say the supply and demand curve which we typically associate with markets only holds a very narrow band around the price.

JH: I wouldn't say that the laws of supply and demand have been repealed here. It's just that once you move beyond ... let me say that the "expressed" demand and supply in terms of participants who are actually there, actually willing to trade - yes, that does thin out. And it 's not because these other people don't exist; it's just because they might be willing to buy, say, $5 down from the market price, but they view the eventuality of that order being executed as so small that they don't really want to think about it. They don't want to have to think about putting in such an order and keeping an order current, so, as a result, they're not actively there in the market.

R: And even worse, from my point of view, is that they might readjust their model of the economy, because when it has actually dropped $5, they may have revised their own expectations.

JH: Absolutely. The reason I'm not going to put in a limit order, say $5 down from the market, say to buy, is that if the price really has moved far enough that my limit order will execute, I should be thinking about why. That is, if there is a big price movement, there's likely to be some information that is going to cause me to reevaluate not just my orders in this particular security but my entire portfolio strategy.

R: So, if I understand you correctly, you imply that liquidity is far more elusive than one initially would expect.

JH: Oh, absolutely. A market practitioner - I think it was Ken Pasternak - jokes about these people who put in limit orders, and thereby represent that they're supplying liquidity, but when the price actually moves towards those limit orders, they get canceled. So, the kind of liquidity that we see in markets is a very transient, very sensitive, very ephemeral sort of thing that, during normal market circumstances when everything runs smoothly, you can do a lot of volume, et cetera, et cetera. But it's not always there.

Now, you mentioned some of your concerns about risk management. I think one of the problems of building in liquidity for risk management is that liquidity failure is itself an extreme event, and I don't think looking at normal liquidity movements tells us much about those extreme events. For example, as a statistician I can look over bid/ask spreads on the foreign exchange market or the stock market, and I can come up with a statistical model, a time series model, of how those bid/ask spreads are going to move, and that model is going to be correct most of the time. The problem is, of course, that it doesn't really characterize those events where there are very large price movements, where there's major information, and it's not like liquidity, however measured, deteriorates by 10 percent or 20 percent. It's not that you have to pay a little bit more to get your trades done, it's that all of a sudden you find, in a liquidity catastrophe, that nobody is standing, willing to buy. This is essentially what happened to Long-Term Capital with their Russian bonds. No modeling beforehand of liquidity movements could reasonably have predicted that.

R: If I understand you correctly, there are two interesting elements. On the one hand, the behavioral element - the people participating in the market - is extremely volatile in the sense that how they view the market depends on the actual price action. On the other hand, the kind of correlation structure that you observe in the market is dependent on the different levels of liquidity in the market.

JH: Yes.

R: So, basically, on two separate dimensions, you have powerful forces that at any time can change the overall structural relationship within the financial markets. Is that correct?

JH: That's correct. There is research now that shows that the correlations that we observe in securities markets are much more dramatic with large price movements. That is, for large price movements, particularly negative price movements, you don't get the diversification advantage that you appear to get in normal times. And I think the same thing is true of liquidity. When there is a liquidity failure for risk management purposes, it's going to fail at exactly the time when you need it most badly; that is, when you will most badly want to implement your dynamic hedging strategies.

R: A further question is how does the trend of creating ECNs, electronic communication networks, affect the issue of liquidity from your point of view?

JH: Although we have a large number of ECNs right now, they all follow, at least in the United States, a very standard model; that is, they are open electronic limit order books. In some respects, the performance of these ECNs is very, very impressive. One of these ECNs - Island - follows a very farsighted practice of putting their electronic limit order book in all the stocks up on the Web so you can dynamically watch the order book changing. And it's clear that you get a lot of information from it; you get a very dynamic picture of the market. For stocks like Microsoft and Cisco and a handful of other large cap stocks in which the natural trading activity seems to be very large, the ECNs seem to work wonderfully for modestly-sized orders.

There are three areas of concern, and these are not specific to these ECNs. Rather they're generally about electronic limit order books in general. Specifically, we have some concerns how well they do with very large trades of the sort that institutional money managers seem to require. The Paris Bourse has used an electronic limit order book for years, and there is some evidence that the large trades are still negotiated trades and that, when they transact, they go through the limit order book in a way that it is not completely integrated with the limit order book and the smaller orders on the book. So we don't know if electronic limit order books can really provide liquidity for large orders. We're also concerned about how they might function for very small firms. The most dramatic pictures on Island's website are those for the large capitalization stocks. As you go to the smaller stocks, you find in many cases that the limit order book is empty. And I'm not picking on Island here. I've looked at the lower cap tiers at the Tokyo Stock Exchange, and there, too, the question is who's going to provide liquidity? The public, in the form of their electronic limit orders, is not stepping up to the plate, is not providing the liquidity, even under normal market conditions. So, that's a concern.

The final concern is, how well do these electronic limit order books function in periods of market stress? There's some research evidence out there - and this is more speculative - suggesting that when volatility hits, limit order traders withdraw their orders in large numbers, so that spreads on the book thin out dramatically. In some comparisons of limit order books versus floor markets, there are people who believe that a floor market in a period of market stress has advantages.

So, the ECNs look great; they function well in very many respects, but we're not quite there yet in being confident that this is the universally optimal market architecture.

R: The ECNs as I know them today basically publish one price, but what one should really look at is what is the price for a transaction of $10,000 worth or a price for $100,000 or a million or ten million or a hundred million? And depending on the size of the transaction, basically the prices are different places?

JH: Well, yes. Absolutely right. With full information on Island's book, for example, you can figure out what, at least on Island, a large order would transact at if it had to go all the way through the book.

The problem, of course, is that Island is not the entire market, and it's very difficult to build up a picture of the entire market and figure out how much liquidity there is, or how much is available - liquidity is a more abstract term than I need here. It's very difficult to figure out what is available and what is sought at prices far away from the market.

Now I should point out that this ties into another theme related to ECNs; namely, transparency. One aspect of the standard ECN model is a very high degree of transparency; that is, if your order is out there, it will be shown to absolutely everybody in the market. If I'm a retail trader, this doesn't concern me very much. If I'm trying to buy or sell a thousand shares, frankly the more people who know about my bid or offer, the better. But if I'm a pension fund trying to buy or sell a hundred thousand shares, I may want to be a lot more selective about whom I disclose my buying interest to. And this need for selective disclosure of trading intent is absolutely counter to the ECN principle of transparency. It's another reason why the ECNs are not likely to be attractive to this particular segment of the market.

R: So you think there is a big market segment that will just not trade over ECNs, or will only trade over ECNs if their liquidity has improved tremendously?

JH: There would have to be some very dramatic improvement in liquidity. I'm not sure that that is really possible. I'm not sure that there are, for example, enough limit order buyers in the world that they are willing to post their desires on ...

R: To pre-commit.

JH: To pre-commit. Even, I might add, many of these electronic limit order systems offer the possibility of hidden orders and that's useful. But the only way you can find out about a hidden order is to actually go and try to trade. Compare that with the kind of negotiation that can go on face-to-face. You and I, Richard, might agree on a price at which I'm going to buy 10,000 shares, and after we've agreed on that deal, as long as it's just ourselves talking, I might say "Well, would you be willing to sell another 10,000 shares at that price?" You might be willing to. And, in this way, we might talk our way up to a larger trade. We can do that because we're selectively negotiating with each other and we're selectively revealing to each other the possibilities at which we'd be willing to trade. That, too, that negotiation feature - you can build anything into an electronic system, but that negotiation is not currently a part of the standard ECN electronic limit book paradigm.

R: I have another question. I've always thought that just the fact that an ECN feeds off the process of buying and selling so much, that it has a further impact. By beating out the speed of the transaction, it requires more people to participate to sustain a minimal level of liquidity.

JH: I think there is something to that viewpoint. When markets were slower, I think there was a tendency to focus more on the total volume of trade rather than, say, the rate of order flow - that the time dimension was less important. Now, with markets going faster, people will learn how to trade in such a way that they don't send all their orders to the market at once but rather they try to distribute them over time. And people supplying liquidity will also supply it in much more measured doses over time. This does confirm your viewpoint that a large order dropped onto the market is going to have a very large impact if it's dropped all at once.

R: How do you assess the current market environment, and what is your prediction regarding the adoption of ECNs and where liquidity will move?

JH: Predictions? I don't. It is very tough to generalize here because the needs of the market participants are so diverse and there are segments that ECNs do not appear to serve very well. One of the striking things about the ECNs is how advanced and transparent the equity markets are relative to other securities. And I'm not entirely sure why that is. Why, for example, has electronic trading been so slow to take off in the bond markets and the foreign exchange markets, or to a lesser extent in the futures markets? To a certain extent, this has been blamed on the existing dealers. The story goes that the existing dealers do not want electronic systems to arise because that will disintermediate the market and they will no longer be useful and no longer make profits. I think this is probably true, but it's only part of the story when you look at how quickly some existing markets have made the switch - and the classic example here is the German government bond futures contract moving from the London floor to the German electronic limit order book. I find it surprising that we haven't seen a similar movement in the bond market, and I'm not sure why that is true.

R: Isn't that because ECNs are basically pure communication networks; that is, you need another buyer and a seller, and there are far too many instruments for there to be one person already placing an order?

JH: If that is the answer, that imposes very strong and obvious limits on how far we can go with ECNs. If that is true, if these communications networks can't deal with the problem of infrequently-traded securities, then they're going to have a very limited application. And we may be at those limits already. That's a defensible viewpoint, and it might well be true. What it highlights, though, is a general problem that I've heard referred to as "the near match problem." This arises in portfolio trading, in bond trading, and in many kinds of goods and services where there are potential traders out there. But because what they express a desire to buy or sell is not letter-for-letter identical, a computer system can't pick it up. If you and I want to buy and sell Microsoft stock, it's very clear what we're buying and selling, and we know where to go to get a quick market in it. But if you and I are trying to more generally buy into technology - to put together a portfolio of smaller technology companies - we'd be willing, most likely, to buy and sell and accept some substitution there. A question is, what kind of trading network can you build that will recognize the potential for substitution? How can we express our orders in such a way that they are firm and yet flexible?

R: This multidimensionality of real life is hard to reflect in a traditional ECN.

JH: Oh, it's very hard and the suggestion may turn out to be an insurmountable problem.

R: How do you view the recent market developments with the meltdown of the dot-com companies? Is that any reference to the basic issue of liquidity?

JH: I don't think so. When I think about liquidity, I really think of it narrowly-defined in terms of issues related to the trading process and the trading mechanism. Dot-com valuation is important, both financially and also for the real economy. But the issues there I do not believe are fundamentally related to the trading process in the sense that even if markets in these stocks were an economist's ideal, you would still have the enthusiasts out there bidding up the prices of these securities based on sentiment.

R: In the back of my eye, I always see an atomic model of atoms. Basically you have these planetary orbits, and then depending on the excitement state, you go into a different orbit. I very much see the same kind of model for financial market prices. Depending on the excitement state, you go to a different price level. Now, personally, I always see a close relationship between liquidity and on what kind of planetary orbit you get into.

JH: The way I would separate those things, drawing on that analogy, is that the market structure can facilitate the transition from one price level to another. If the market is doing its job, it can facilitate that transition in the quickest, most transparent, most visible way possible. But the market structure can't affect, or a desk has only a second-order effect, as to which state you're going to go to. I'm sure when we start to do autopsies on this time, we'll be able to look back and see all the warning signs of the dot-com bust. Frankly, many of these we are aware of now. Companies raising capital, only selling a very small fraction of their total stock float, companies using capital that they've raised purely for advertising - there will be very thoughtful and comprehensive studies. But I don't think the market, per se, the market structure, I don't think of these as liquidity issues.

R: Joel Hasbrouck, thank you for the time you've spent educating me.

JH: Richard, thank you.