The New Exchange

Exchange differentiation under the new paradigm is not easy. Twenty years ago NYSE traded NYSE listed securities and OTC securities were traded via NASDAQ; shares in UK based companies were traded on the LSE, shares in German companies were traded on Deutsche Boerse, and so on. Now, especially for US and European equities, shares of anything can be traded virtually anywhere. I’m over-simplifying of course, but with globalization flattening the world of stock exchanges, regulations keep everyone on a level playing field and with all measuring speed in microseconds the only obvious differentiator left is the name of the venue. Even if we assume traders naturally migrate to where liquidity is deep and spreads narrow, only through an understanding of exchange technology can one rationalize what causes that situation to occur.
It is a poorly kept secret that high frequency trading firms and proprietary trading desks at investment banks co-locate to shave off microseconds. This practice is at the heart of exchange-client connectivity. These high speed orders are generated within the servers of proprietary trading desks and hedge funds, and are sent via a high speed network into the exchange’s matching engine, all literally residing under one roof. This practice generates the majority of order volume in the US and increasingly in Europe.
Large agency orders from traditional buy-side sources are also important to an exchange’s success, but it is more often the job of the broker to ensure connectivity to the data center for their client flow. Simply put, the sell-side handles inter-data center connectivity and the exchanges handle intra-data center connectivity. TABB Group estimates that North American spending on market connectivity sits at just over $2 billion annually, with 70% of that number coming from the sell-side.
That leaves the exchange to ensure the pipes coming into their matching engine are big enough to handle the volume and fast enough to get orders to the market before prices change sending traders elsewhere. To do so exchanges have put a heavy focus on the physical networks (reducing hops, upgrading to newer standards such as 10GE, etc.), messaging middleware and message format to ensure orders travel to the matching engine as efficiently as possible. Some (such as NYSE and the London Stock Exchange) have taken this entirely into their own hands by owning and operating their data center, whereas others (such as BATS and Chi-X) prefer to build an ecosystem in a shared services data center choosing to keep their technology expenditures literally close to the exchange itself.
It is interesting to note that despite the drastic reduction in latency in the past decade, FIX continues to be the protocol of choice for most order routing. Of course FIX has undergone change since the mid-90s, but its core remains the same with simple key-value pairs. But in a nod to its original designers, those messages remain relatively small by today’s data standards leaving them suitable for high speed trading.
Once the orders are received exchanges must then do what they are famous for – match buyers and sellers. If you were lucky enough to spend time on the floor of the NYSE pre-Y2K you fundamentally understand what order matching entails. Things today are no easier or harder than they were before; they are simply automated and fast really fast. Matching engines from BATS, DirectEdge and NASDAQ all hover around 200 microseconds. NYSE will sit at a similar level when its Universal Trading Platform (UTP) is fully implemented.
We’ve become jaded over the past few years about what fast really is. What the exchanges have done with matching is nothing short of extraordinary. The human negotiation process seems straight forward, but programming a machine to handle the subtleties of a trader’s mind is quite complex. This process, now simply referred to as order matching, includes much more than just matching the next buyer and seller of the same security at the same price. Risk checks, microsecond precision time stamps, order type and a number of other factors must all be considered before a trade is executed. And what really makes this process extraordinary is that the major exchanges have whittled this process down to a few thousand lines of machine code that match orders in millionths of a second and do so millions of times a day for hours on end.
Finally, executed orders become market data quotes that must be distributed to the world. TABB Group estimates that 60% of the $2 billion spent in North America on connectivity is for gathering market data. Furthermore, as trading has become increasingly automated and latency sensitive, the use of traditional market data aggregation platforms has declined in lieu of leased lines which now account for 43% of connectivity spend. Many of the same aspects apply to market data distribution as to client connectivity. Those exchanges that can get market data back to their clients fastest will attract more traders, more volume and hence higher profits.
Interestingly, although each of these three components is core to the value proposition of the major global exchanges, NYSE Euronext (via NYSE Technologies), NASDAQ OMX, Chi-X (through Chi-Tech) and a few others are all willing and hopeful about selling the technology they use in-house to other global exchanges. Besides the obvious driver for more revenue, the thought is that if a global network of exchanges can be created with all of them running the same technology, then that technology will become a standard, thereby making global liquidity more easily available regardless of your physical location.
The trading floor model is tried and true. A few such floors still exist, and a smaller few are still thriving. However it is hard for investors, traders and exchange shareholders not to see that technology is bringing efficiency to the market that just makes dollars and sense. The exchanges of tomorrow will make most of their money not as execution venues, but as data and technology infrastructure providers.Connectivity and order matching functionality is already expanding to include pre-trade risk, latency monitoring, cross-product execution and numerous other forward looking initiatives. This exchange metamorphosis has become truly global with national exchanges worldwide looking to technology to get an edge on the world stage. So I wonder, how long until Microsoft sees NYSE Euronext as a competitor – the Bing Stock Exchange anyone?