Link : The Journal of Finance( Sep, 1995 )

As early as 1997, the US financial markets comprised blue chip stocks traded by specialists at NYSE , other stocks traded at NASDAQ by specialists and a small scale electronic system. Fast forward to 2012, the US market comprises 40 trading destinations. There are four public exchanges – NYSE, NASDAQ, Direct Edge and BATS. Inside each of these exchanges there are various destinations. NYSE has NYSE Arca, NYSE Amex, NYSE Euro next and NYSE Alternext, NASDAQ has three markets, BATS and Direct Edge have two market destinations with in themselves. There are toxic Dark pools.There are Internalizers – Citadels of the world that execute trades with in their trading pools. The system, as you can see, has become extremely complex. Dark pools and internalizers accounted for 40 % of all trading volume in 2012. The pace of developments have been unbelievable.

The paper was published in 1995 and hence refers to a less complicated scenario than today’s world. However one can read about a nice application of VECM model. I have come across VECM in a stat arb setting.This paper showed me an application of VECM in a market microstructure setting.

What’s this paper about ?

With the market fragmentation, a security is listed in multiple venues. Hence there are quote processes, trade processes in various venues. The paper tries to answer the following question :

Where is the price discovering happening ?

It is assumed that all the securities have a common stochastic component. This means that any long short combination of security listed in two different markets is cointegrated. If there are n markets where a security is traded, one can build a VECM model to explain the behavior of a n dimensional vector. The paper presents the cointegrated model in two forms. The first form is the Stocks and Watson representation that is good for interpretation. The second form is VECM form that is convenient for estimation. In the first representation, the variance of common stochastic random walk can be decomposed in to variance arising from various markets. Identification issue plagues VECM models too. For a vector Moving average representation, there can be a many VECM models. Similar to how one imposes causal ordering on a Standard VAR to retrieve Structural VAR, innovation accounting in a VECM entails imposing restrictions. Since the variables are cointegrated, one does not choose a specific causal ordering. One usually reports the upper bound and lower bound for the % explained in a forecast error decomposition technique.

The paper defines “Information share of a market” as the proportion of innovation variance that can be attributed to it. Upper and lower bounds of the % variance explained of various markets are reported and this serves as a proxy for price discovery. Whichever market place figures as a dominant % variance explained venue, that’s the place where price discovery happens or to put it in a better way,”that’s the market where stock moves first”. The author applies the above technique to 30 stocks and concludes that price discovery appears to  be concentrated at NYSE.

If the author were to somehow get the data across all the fragmented venues and run this type of analysis again, given today’s market place, it is likely that NYSE might not be the place where price discovery is happening.