THOUGHT LEADERSHIP

SIs: The good, the bad and the ugly

Henry Yegerman, global head of sales at ISS LiquidMetrix, explores recent execution trends in bank and non-bank systematic internalisers.

Over the past two decades, one of the major changes in market structure has been the proliferation of different types of trading venues. We are going to look at systematic internalisers (SIs), a trading venue (of a sort) which has become part of the European trading environment post-MiFID II.

Highlights:

  • There is significant difference in execution style, execution size and information leakage both between different types of SI’s (bank versus non-bank SIs) and within the same type of SI.
  • There is greater information leakage among non-bank SIs. Especially for less liquid names.
  • To an extent market segmentation has taken place among SIs. Some SIs are ‘full-service’ executing in multiple ways, while others are targeting a specific niche for certain types of trades. 

Across all asset classes, there are now 223 SIs registered with ESMA. Most are fixed income, but approximately 80 SIs trade equity. Although SIs originated in 2007 under MiFID I, their growth did not take off until MiFID II mandated the closure of BCN’s (broker crossing networks) as trading platforms. 

Under MiFID II a trading venue is a platform that brings together buyers and sellers in financial instruments. Broadly speaking, MiFID II divides the world into:

Multilateral trading, which is trading on trading venues, (regulated exchanges and MTFs such as the LSE, Turquoise, Liquidnet) and;

Bilateral trading which is when two counterparties trade directly with each other. SIs are a type of bilateral trading, although like exchanges and MTFs, SIs must send out quotes prior to executing.

The public policy rationale for SIs is twofold:

  • To make over the counter (OTC) trading activity – i.e. trading which takes place outside a trading venue – more transparent, and;
  • To level the playing field so that rules followed by trading venues and investment firms which trade on their own books outside a trading venue, are reasonably similar.

The issue for policymakers was that bank owned BCNs, which were bilateral trading vehicles, did not have much in the way of transparency. Instead of sending client orders to traditional exchanges, the BCN bought or sold the shares for/from their own ‘inventory’, that is, their book of stock. Policy makers were concerned there was a lack of visibility and pre-trade transparency around prices on BCNs compared to exchanges. Although SIs also engage in bilateral trading, the new regulatory regime aimed to make them more transparent than the old BCNs.

SIs, like BCNs continue a 20-year trend of blurring the line between brokers and exchanges. In many respects, an SI is not so much a trading venue as a trading counterparty. You buy from, or sell to, the SI itself. The SI uses its own capital for trading and takes on risk in trading positions. In any case, SIs have largely replaced BCNs as part of European market structure. Perhaps not surprisingly, most of the largest SIs are part of banks, just as most of the large BCNs were owned by the same banks. This has raised the concern that some SIs are merely old wine in new bottles operating much like the banks former BCNs. Yet, at the same time, there have been new non-bank SI entrants into the market. 

Do non-bank SIs perform differently than bank SIs? How different is the performance of SIs within their own category? We will look at some criteria for evaluating the performance of bank versus non-bank SIs and the variation of performance within both bank SI and non-bank SI categories.

First, let’s look at how the SIs position themselves as sources of liquidity. To do this, we will look at how the execution style that traders use interacts with the SI. By execution style, we refer to whether it is a passive execution which adds liquidity, an aggressive execution that removes liquidity from the venue or whether it is executed at the mid-point. Figure 1 indicates that almost all executions done on non-bank SIs are aggressive crossing the spread to access liquidity. Most bank SIs execute largely at the mid-point, although there are some exceptions where fills are either largely aggressive or passive. This suggests that non-bank SIs are engaged in a form of segmentation focusing on certain types of stocks or trading environments, such as high momentum situations where clients are more willing to trade aggressively and bid up the price.

So how do the SIs perform? One important metric for evaluating the execution quality of a venue is trade size. The sooner an order can be completed, the faster the risk associated with the price moving against you can be eliminated. The key to completing an order faster is to get more of it done per execution. Figure 2 shows the average executed value for our sample of bank and non-bank SIs (the blue columns). We see that a couple of bank SIs have significantly larger average execution sizes than the non-bank SIs. The rest of both bank and non-bank SIs are roughly comparable. 

The lines across the secondary Y-axis indicate the spread of the stock when the trade was executed. Bank SI 2 has the largest average value traded, and is also trading stocks with much wider spreads than any other SI. Interestingly, SI 2 is a second-tier bank who may be choosing to target a specific market segment by specialising in illiquid names. 

Three bank SIs (including the above-mentioned bank SI 2) have the largest average fill sizes, but after that the non-bank SIs have larger traded value per execution on average. Spreads (again apart from bank SI 2) are similar in the 10 – 15 bps range.

Another key criterion of execution quality is the extent of information leakage is evidenced on the trading venue. Information leakage can take two forms. An aggressive execution that crosses the spread to take shares signals demand for liquidity that can produce price impact. Adverse selection occurs when one trader has superior information and uses that information to their advantage at the expense of the counterparty in trade. For example, repeated buy orders can signal to a seller that there is still considerable demand from a buyer on the other side of a trade. The seller will adjust by raising their offer prices higher. Once the trade is executed, the price reverts lower. The buyer has executed at a higher price than the natural price level of the stock.

LiquidMetrix uses a metric where we measure the change in the mid-quote in basis points from each execution from one milli-second to one second to analyse potential information leakage. The less price change, the lower the potential for possible information leakage.

We will examine price impact and adverse selection among SIs for both highly liquid and less liquid stocks. To define what is a ‘more’ liquid and ‘less’ liquid security, we employ the MiFID II categorisation of financial instruments into tick bands used in RTS-28 reporting. We will not examine on illiquid names with an average of less than 80 executions per day, but focus on the ‘high’ liquid securities (tick bands 5 and 6 with more than 2,000 executions per day) and ‘medium’ liquidity names (tick bands 3 and 4 with 80 to 1,999 trades per day). We are using a research dataset with over 1.3 million trades for high liquidity stocks and over 994,000 trades for medium liquidity stocks.

For ‘more’ liquid names you begin to see post execution price changes at 25 milli-seconds after the fill (Figure 3). Overall price movement is similar, although there is at least one non-bank SI which evidences much larger price movement. 

Figure 4 shows price impact for less liquid stocks. For less liquid names there is clearly greater price impact after the execution among the non-bank SIs. Especially for non-bank SI 2, which also had large price impact in the more liquid names. In general, non-bank SI 2 is a weaker performer as it can be pointed out that their average value per execution was next to last among non-bank SI’s.

When we look at adverse selection on passive executions, we see the same trends only much more pronounced. Figure 5 shows significant price movement in more liquid stocks among almost all the non-bank SI’s by 50 milli-seconds after the trade. This is not the case for the bank SI’s, which for the most part, show almost no price change even up to one second after the execution.

But the greatest difference in potential information leakage is when passively executing in illiquid stocks (Figure 6). Here again, the bank SI’s show almost no change in the mis-quote even after one second after the trade. In contrast, the non-bank SI’s begin to evidence price movement 10 milli-seconds after the trade and most begin to show significant reversion from three to almost 10 basis points by 500 milli-seconds post execution.

Although the regulatory goal in creating SIs was to create an even playing field, it appears in terms of performance that some SIs may be more equal than others.