Google, robots.txt and the “Search for Liquidity”

I had an interesting conversation a few days back with a buy-side figure who will remain nameless.  The content of the conversation was wide ranging but focussed on the need for buy-sides to gain access to liquidity in order to operate in an efficient manner.  Let’s start from first principles here and establish a paradigm that explains what we see...
[Note: this is a post on the equity market only for the time being.  I may follow up with f/x and fixed income at a later date if there’s interest]
An institutional active asset manager will have a heavily constrained investment strategy.  Typically any investment mandate will include numerous restrictions such as:
  • No investment in derivatives
  • No investment in unlisted securities
  • No investment in a listed security can be greater than x% of the entire listed amount of the security
  • No investment in instruments issued by Company X (perhaps where company X is a competitor to the firm that owns the assets)
So, there’s no way for an investment manager to take on an aggressive maximisation strategy.  Take a crude example.  If an investment manager believes that one company will outperform any other within the appropriate investment horizon then it would make sense for the investment manager to investment the entire fund in that instrument.  Now, this is crude, since it ignores simple portfolio diversification, it’s a high risk strategy that would not be appropriate for any publicly available institutionally managed investment.
So, let’s say that an investment manager does indeed believe that a stock will soar in value.  It makes sense to buy as much as possible within the risk/reward structure and the mandate, firm and market specific compliance and regulatory regimes.  So perhaps at most 8% of a reasonable sized fund would be invested in one stock.  Now, let’s imagine that the stock does indeed perform as expected within the investment timeframe and the investment manager decides to sell the position. 
Here’s the problem that the investment manager will encounter.  How do you sell a large position without adversely affecting the market price that you will receive for the sale?  In other words, if you are selling into a market the bid/ask spread may widen and/or skew against you.  So if you start selling at a time when the market is 150p/151p you may find that the spread widens to 149p/152p or skews 149p/150p.  Now, the problem here is actually a lot more subtle than it looks.  If the investment manager bought the stock at 10p and sells one year later then a difference to the sale price of 1p or 2p is meaningless since the stock has clearly been a massive gainer over the year.  But in the real world where investment gains are much smaller the difference of 1p or 2p per share can be huge.
If the position to be sold was a small one relative to the average daily volume (the number of shares traded in that stock averaged over a specific time frame – say three months or one year) then the market impact will be minimal –selling 1000 shares in a stock that trades 1,500,000 per day will not have a meaningfully measureable impact.  It would be like throwing a grenade into an exploding volcano.
But, if the investment manager is selling a large position then this will have an impact on the price that is realised in the market.  This is intuitively appealing in pure supply and demand terms.  As a secondary consideration, there is always a degree of diversity in opinion on the fair price for an equity.  So if there is a big seller then this may lead sophisticated market participants, acting from a position of rational ignorance, to make an assumption that a better informed participant has made a value decision and is selling the stock.  As such, this could lead to further falls in the market price.  None of the above is new or innovative and every trading desk I have seen has a market-impact-cost-model to try and establish what will be the impact of selling a large position.
This leads to a premise – that in a market with sophisticated participants the act of selling a “large” position has swift impact, since the distribution of market data for trades on a lit exchange is real-time.
So – what’s the rational course of action for an investment manager?  Well, one way to operate would be to trade in a way that people do not see what you are doing.  That explains the growth of dark or off-market trading.  Again, this is not new or unique.
This leads to a conclusion that the market structure has an impact on trading performance while can be material.  Again, this is intuitively appealing since if doing anything apart from trade on a primary exchange it’s reasonable to assume that this is for a good reason and improved trading performance is a good reason.
So, the market structure in equities is pretty complex and a complete taxonomy and analysis is beyond the realm of this blog.  But let’s simplify to draw out some pertinent points and see if we can create a few conclusions and discern some future developments.  In essence, the equity market has a number of key participants:
  • Inter-dealer brokers – pure brokers who intermediate between sell-sides
  • Sell-sides – broker/dealers with an exchange membership
  • Buy-sides – institutional asset managers who manage large funds within a risk/reward structure and the mandate, firm and market specific compliance and regulatory regimes.
  • Hedge funds – a form of buy-side that may short-sell on a regular basis and may have a lighter regulatory overhead and a more aggressive risk/reward structure than a regular buy-side firm
  • Private individuals – the “man on the Clapham omnibus”
  • Exchanges – the place where “lit” orders are matched – where buyers and sellers meet and trade. 
The current market model in place in every country in which I have worked is one where there are tiers of hierarchy in the market structure.  Think of an onion, the exchange is at the core of the system. One layer out from the core is where the sell-side firms operate.  To reach the exchange an order has to pass through the sell-side. The next layer out comprises buy-side firms, hedge funds and private individuals.
The onion metaphor is instructive in that it should be clear that the only way to place an order to the exchange for a buy-side, hedge fund or individual is to go through a sell-side firm.  This means that there is a problem at a conceptual level with this structure.  It’s inefficient.  How?  If buy-side A wants to sell stock X and buy-side B wants to buy stock X then what happens?
  1. Buy-side A sends sell X order A1 to sell-side C
  2. Sell-side C sends sell X order C1 to exchange
  3. Buy-side B sends buy X order B1 to sell-side D
  4. Sell-side D sends buy X order D1 to exchange E
  5. Exchange E matches order D1 with order C1
  6. In an anonymous market model with a central counterparty (CCP)the completed orders D1 and C1 are novated to a CCP. So D1 settles against the CCP and C1 settles against the CCP.
  7. Order A1 settles with sell-side C, sell-side C makes a commission on order A1
  8. Order B1 settles with sell-side D, sell-side D makes a commission on order B1
So – what did sell-side C and sell-side D do here in order to earn a commission?  Well, they did not really bear any risk, since the delivery-versus-payment model and a CCP mean there is minimal risk.  In a post-Volcker rule world where banks cannot engage in proprietary trading there is very little that sell-sides C and D have actually done beyond providing order routing services between the buy-side and the exchange.
So, all roads lead to the exchange before the trade is done and then afterwards the information flow is a little more complex.  The information transmitted to and from the exchange is sent over leased lines or financial extranets and may use a proprietary message format such as ITCH/OUCH or open protocols such as FIX.
Returning to the title, let’s look at the search for liquidity part first.  Buy-side A is seeking liquidity in stock X – it wants to ensure it can sell the stock at the best possible price.  The other side of the coin is true for buy-side B which wants to buy the stock at the best possible price.  In both cases of liquidity search the buy-side has had to use a broker to access the exchange. And in both cases the search may not be reaching market participants at different stages of the decision making process.  By this I mean that in some cases the decision to place an order in the market is different to the process of creating an order internally within a buy-side OMS.  Consider this as a form of limit order but rather than conditional on the market reaching a certain price level it may instead be a way for the investment manager to grant some discretion to the buy-side dealing desk.
The role of the IOI can be seen in a similar light – to try and flush out interest that may not be firm enough to express within the strict constructs of the exchange order types.
So where does google fit into this?  Consider that pretty much everyone in the world uses google to search the internet for keywords and phrases. Google will also answer direct questions – type in “how tall is David Beckham” and you’ll get the answer 6’0” (1.83m).  How does this work?  Well, libraries have been written about how the Google platform operates so I will not try and answer directly. What I will note is that Google operates software agents called Spiders that crawl the world-wide-web and extract data for analysis in order to provide answers to queries made.
Extend that paradigm – a buy-side creates a secured website that allows spiders to search for liquidity. That liquidity may be strong or weak, in other words, an order or an indication of interest. Many different search engines will scan websites and a robots.txt file can be deployed to request that the website is in part or entirety excluded from the search engine.
So, expand the analogy – rather than the current exchange based trading paradigm, imagine a world where Google decided to get involved.  Replace the hub-and-spoke paradigm of information distribution with a peer-to-peer model with multiple agents conforming to a standard for negotiation for trading (see earlier post on FIIDL).  Robots.txt then becomes a counterparty restriction tool.  Add in a CCP with full buy-side participation and the world of financial services looks very different.
A few stages of development for this new paradigm could be:
  1. Sell-sides commence to use the spider model to search for liquidity
  2. Tech savvy buy-sides and hedge funds use the spider model and FIIDL to scour multiple sell-sides for liquidity
  3. This technology breaks out to the sophisticated private investor and then to universal adoption. 

NYSE spent a huge amount of money on building out industrial strength datacentre space in Essex for the various exchanges they ran in Europe. In Australia the ASX built a world class facility in Gore Hill.  In both cases rackspace can be rented to allow trading or other applications to co-locate close to the matching engine.  This is the world of exchanges crossing over into the world of datacentres.
So what happens if the world of datacentres tried to cross over into the exchanges world?