January 2, 2009
I was lucky enough to recieve a copy of Joel Hasbrouck‘s Empirical Market Microstructure this Christmas; I’ve now worked my way through the first few chapters. Hasbrouck opens with a general survey of trading in the first two chapters, covering similar ground to Harris, and then follows with the Roll model. So far I’m favourably impressed, but will post more later…
December 5, 2008
In his discussion of continuous order matching systems Harris describes the discriminatory pricing rule on p126: “the limit price of the standing order determines the price of each trade. If the market matches a large incoming order with several standing limit orders placed at different prices trades will take place at the various limit order prices.”
Which means no price improvement for standing orders at best bid if an incoming marketable sell limit order matches one or two levels off best bid.
December 1, 2008
Christmas is coming, and I’ve been figuring out my reading list for the break. I’ve pulled my old copy of Godel, Escher, Bach off the shelf, and have started the first re-read for 20 years. A couple of things hit me right off the bat: how could I have missed the GEB/EGB (Godel,Escher,Bach/Eternal Golden Braid) sequencing pun ? And it seems obvious that GEB is a big influence on briarpig‘s style. I’m also looking forward to reading Hasbrouck‘s Empirical Market Microstructure. Looks like a natural successor to Harris’ Trading and Exchanges.
Thought for the day: to be is to be the value of a bound variable.
September 19, 2008
Lately I’ve been taking an interest in the maths of limit order book behaviour. Electronic exchanges, and many ECNs, are organised as limit order book trading systems. Stock exchanges, like the LSE, and futures exchanges, like Eurex, are organised as limit order books. Larry Harris has interesting comments to make on the interplay between limit and market orders on limit order books. It’s an interesting area for trading system developers, since broker dealer banks and hedge funds sometimes sponsor system development aimed at exploiting order book behaviour. Here’s some linkage…
March 29, 2007
In ch 19 Harris distinguishes between fundamental and transitory volatility. Fundamental volatility is due to unexpected changes in the fundamental values of instruments, and transitory volatility is due to impatient traders trading market orders. Bid/ask bounce is a form of transitory volatility. When you see bid/ask bounce in a market you’ll see the prices at the top of book tick up,down,up,down constantly, usually in a one tick range.
A good place to observe this in fixed income is Eurex futures. The front month contracts for Schatz, Bobl, Bund and Buxl (2,5,10,30 yrs) are the most heavily traded contracts in the Euro rates world. A lot of the trades on these contracts are hedges on cash bond positions. The traders hedging their cash positions will take whatever prices are showing at the top N levels in order to hedge, so they match Harris’ characterisation of impatient market order traders.
March 28, 2007
Some time ago JP proposed that all systems could be reduced to four aspects or capabilities: syndication, search, fulfillment and collaboration. At the time I thought it an interesting point, but not one that applies to etrading systems. Today I found myself reading Harris chapter 19 on liquidity. Harris characterises liquidity as the object of the bilateral search in which buyers and sellers look for each other. Which reminded me of JP’s point, and prompted the thought that ECNs like Liffe, which are organised as electronic limit order books are indeed search engine. They’re constantly performing a highly constrained search for matching orders.
So let’s have a go at mapping JP’s four pillars to etrading…
- Syndication: market data syndication as per Reuters
- Search: exchange order matching, or systems that reaggregate liquidity fragmented across multiple venues
- Fulfillment: T+2 – what else !
- Collaboration: squawk boxes, IM, Bloomberg chat
January 22, 2007
More enlightenment, courtesy of Harris: dealers and arbitrageurs are both liquidity suppliers. However, dealers use their inventories to match buyers and sellers arriving at different times in the same market. Arbitrageurs use their hedge portfolios to match buyers and sellers arriving at the same time in different markets.
January 9, 2007
Back in 1990, Brad Cox wrote a far sighted paper on software components, static and dynamic binding technology, and their implications for software reuse: Planning the Software Industrial Revolution. Mainstream software practitioners still haven’t assimilated Cox’s contention that static and dynamic binding are equally valid techniques appropriate at different places in software architecture. Many seem to think the two are polar opposites, and one must pick sides, rather than being complementary. Which is probably a symptom of the fact that it’s still early days for mixed language development environments.
In 80s Britain there was a similar debate on market economies versus comand and control, socialist, economies. Market economics won out of course, and that result was validated by the collapse of the Soviet Union at the end of the decade. More recently, the debate on the applicability of markets has moved from the national macroeconomic scale of the 80s debate to the level of individual business organisations. Some advocate prediction markets as a better means of internal organisation than traditional command and control. Apparently Google and Microsoft use them internally. The NHS introduced internal markets, with mixed results.
In chapter 9 of Trading & Exchanges, Harris points out that “not all economic decisions are best made in the marketplace. Markets work well only when the costs of negotiating are small relative to the costs of the goods or services that trade there. Markets work poorly when transaction costs are large or when activities need to be highly coordinated….Companies are the most important command organizations within a market economy. People form companies that managers control in order to avoid the excessive market negotiation costs.”
Markets are like dynamic binding or loose coupling. And command and control is like static binding or tight coupling. They’re both useful, applicable and appropriate in different circumstances. Monopolies and socialist economies impose static binding where dynamic binding does the job better. And I suspect some of the outsourcing deals impose loose coupling where tight works better.
Recently Holky has covered the contention by Icap and others that MTS is a monopoly. And he’s also discussed new transaction types and the shortcomings of the quote driven RFQ model. Quote driven markets like Bloomberg and TradeWeb impose a kind of tight binding in that those raising RFQs can only trade with dealers. Order driven markets, which show form executable prices, are more loosely coupled. Any participant can trade with any other. I suggest that any new trading model beyond the established quote and order driven ECNs must be loosely coupled to succeed. I wonder how loosely coupled LiqudityHub will be ?
October 14, 2006
One of the differences between order driven and quote driven markets is that all liquidity is supplied by dealers on quote driven markets. When you trade on a quote driven market like Bloomberg, you must trade with a dealer. By contrast, on an order driven market like Liffe or Eurex, any participant can trade with any other. All that is necessary is that orders from opposite sides of the market match.
This article discusses the role of designated market makers in order driven markets for stocks, futures and options. The DMMs role is to stop the screen going dark, to be the liquidity provider of last resort when no one else is trading. This is especially important for ETOs – exchange traded options. Unlike a futures market, there isn’t natural interest on both sides. Sellers protect themselves from future price drops, and buyers from rises in the underlying. The analogous hedges with options are to buy puts to hedge against price drops in the underlying, and to buy calls to hedge against rises. Selling puts and calls isn’t a natural hedge on the underlying. But an options exchange needs sellers, hence the importance of designated market makers.
September 26, 2006
Accrued Interest posts on a study claiming that the TRACE corporate bond trade reporting systems had collapsed spreads to the tune of $1 billion. There are many other factors that could affect spreads beyond TRACE, so I question the attribution of all of that billion to the reporting effect. Accrued points to increased competition – a very plausible explanation as the study period coincides with the rise of electronic trading for bonds, which has surely driven down spreads.
Reporting could cause spreads to widen for block trades – especially for the less liquid securities. If an institutional trader wants to sell a block of an illiquid corporate bond, then any dealer offering liquidity will have to price in the cost of front runners taking the trades he’ll need to make to unwind his inventory when they see the block reported. Any front running trader who’s long will see the block reported, sell his long position, and buy it back cheaper when the dealer drives prices down selling off his block.