2014 Financial Markets Conference
Are Less Active Markets Safer? Session Highlights
Joseph E. Stiglitz: There's been a significant body of macroeconomic research showing that there are large macroeconomic externalities (basically, an application of the Greenwald-Stiglitz theorem to macroeconomics) showing that firms may borrow excessively in foreign-denominated bonds, and on the basis of this there is now a broad policy consensus that it is desirable for governments to impose restrictions. When I say a broad policy consensus, it's not universal, but what is striking is that the IMF [International Monetary Fund] now supports the view that there should be capital controls.
The reason I emphasize this is there is now within the policy community and the academic community a broad sense that these cross-border flows, unrestricted, are hyperactive, and as a result of that are destabilizing and that appropriately designed interventions can help lead to a more stable macroeconomy.
High-frequency trading [HFT] has been justified in terms of two concepts: "price discovery," making markets more informative, and adding liquidity to the market.
But these terms are thrown around in a very loose way, and what I want to argue is that the real test is, does it lead to better outcomes? This is nothing more than a zero-sum gain; it's actually a negative sum gain. Why do I say it's a negative sum? It's a zero-sum because the flows of goods and services are fixed, and all that's being determined is who gets the goods and services. So it's just a question of redistribution.
Markets cannot really be fully efficient, that is to say efficient in transmitting information, because if they were fully efficient, if I gathered information and paid for it, everybody in the market would know it instantaneously, and therefore I would not be able to glean any return to my investments in information. So if markets were fully efficient in conveying information, they would convey only free information. So the equilibrium in that context would be a very uninformative market.
Well, once you understand that insight, then you understand that allowing a high-frequency trader to look at what you do, to see your actions, gives them the opportunity to take some of the rents, information rents that you would have gathered. And if they're taking your information rents, then what are your incentives to gather information? They're diminished. What we have then is that high-frequency trading results in a less informative market.
Let me now turn to the issue of liquidity. You can't measure or assess liquidity by the average volume of trade, it's really the relevant thickness is when you want to sell and particularly when you want to place a large order, and that's the relevant measure. Part of the problem is that liquidity is related to having the standing orders to buy and sell, but whenever you have a standing order you're exposed to what is called the "lemons problem," and when you have big asymmetries in speed this problem gets worse and therefore you have an incentive to decrease the amount of real liquidity in the system. There are some other adverse effects: increased focus on short termism; misallocation of resources from productive investment to rent-seeking behavior.
In general, markets can be hyperactive in a wide variety of ways—in ways that have adverse effect on the economy as a whole, and that therefore some form of intervention, I think, is desirable. There is continuing debate exactly what is the best measure, but there is a broad consensus that there are a number of different interventions that are welfare increasing.
In the context of the other areas I discussed, there is less of a consensus, and my own feeling is that one should use both tax and regulatory measures. I think you can discourage short-term trading, for instance, by a financial transaction tax. That there are regulatory interventions, some of them relatively simple like...let me say each of these has their problems, but not doing something also has a problem.
So it seems to me that what should be clear, I think that the benefits of discouraging excesses may well exceed the costs. The benefits of this excessively active market have not been established. I think the costs are apparent, including costs in the confidence in the marketplace itself, and I think the advocates have yet to make a convincing case that the benefits of these excessive transactions are worth the costs.
Andrew Kuritzkes: But more significantly for me is the comparison with Canada. So why do I choose Canada? Well, Canada is widely perceived as having had a good crisis. Canadian banks emerged from the crisis relatively unscathed, there were no Canadian bailouts for banks or other financial institutions, OSFI [Office of the Superintendent of Financial Institutions], the Canadian regulator, is often held up as an international role model, and this despite the fact that Canadian bank concentration is actually much greater than in the U.S.—the assets of the top five Canadian banks account for 89 percent of total banking assets of Canada and 179 percent of Canadian GDP [gross domestic product] versus 60 percent and 52 percent for U.S. banks, respectively.
Well, as you can see from the chart it turns out that the financial services sector share of GDP in the U.S. and Canada is virtually identical at 6.5 percent for Canada and 6.6 percent for the U.S., and what's more, the changes over time mirror each other remarkably closely. I think it's quite noteworthy that the share of GDP devoted to financial activity in the U.S. and Canada is the same despite major structural differences in the Canadian economy and Canadian financial markets. Just a few of these differences, the level of bank concentration, which I referred to a second ago, the structure of the mortgage sector, the structure of the Canadian pension system. There are huge differences, and yet the shares of GDP that are taken up by financial activity appear to be remarkably the same.
So to me an implication of this is that based on this very simplistic "good country, bad country" comparison, it's not obvious that the size or growth of the U.S. financial services sector is in itself problematic or potential destabilizing.
Rishi Narang: There is generally a perception that HFTs don't want regulation—that's dead false. Fully, just completely false. Cameron Smith is the president of Quantlab [Financial], which is one of the quietest, but also most active HFTs on the planet and particularly in U.S. equities. I think they by themselves trade something like 5 percent of U.S. equity volumes; I suspect that qualifies them as a major player. He's on the record, an op-ed in the FT [Financial Times], talking about what reforms he wants to see. My brother, cofounder of Tradeworx—who is the CEO of that company—is on the record in a bunch of places, including comments to the SEC's [Securities and Exchange Commission] open commentaries back when Reg NMS was being discussed in '07 and again in 2010 about lots of reforms that are needed. There are lots of reforms that are needed. Market structure isn't perfect, but it's pretty damn good. It's actually better for all its complexities and for all of the things that have been put in place that can be gamed, it's actually a better market structure than virtually any other market. So we have competition among brokers, competition among exchanges that's led to a really serious problem of fragmentation. Then how you unify a fragmented market is a really challenging problem, especially when the centers of the market are geographically dispersed. So information naturally needs to travel from one market to the other. If the S&P [Standard & Poor's] moves in Chicago, it needs to move in New York; and "at what speed?" has got only one good answer—the minimum possible speed. Right? You can say that one microsecond is not so great, but one second or one minute is much better, well, it's entirely arbitrary. It's like saying that I prefer sprints to marathons, or vice versa; they're both races. And there is no minimum value other than the absolute achievable minimum that makes any sense. Any other number is totally arbitrary.