So, for clarity and comparison, let's look at the TCA process per asset class in the "old world", although the focus of this post is only on Fixed Income...
Relatively simple. Take the order book data for the instrument traded during the time of the trade for every exchange on which the instrument is traded. Then process that data to look at implementation shortfall or whatever other metric in which you are interested. This is something that many firms offer right now:
- ITG Analytics
- Markit TCA
- Elkins McSherry TCA
- LiquidMetrix TCA
- Abel Noser Trade Analytics
[There are more - there's a nice list on the Greyspark website here.]
The main issue with this right now is that consolidating multiple markets is not a 2+2 =4 problem. What I mean here is that a market maker may quote in size on however many markets an instrument trades, but once they are hit or lifted in one market they may pull their quotes in other markets. Without turning this into a great discussion of algorithmic trading, the construction of a true depth of market representation is not a trivial exercise and requires interpretation of data. In some cases, the approach taken by TCA firms is to simply ignore all non-primary markets and just look at the primary listing exchange of a particular equity.
The idea is that a firm uses the output of the TCA reports to be able to show that their trading process is yielding optimal results.
Fixed Income: RFQ
Request quotes from n number of banks where n is usually between 3 and 5. Record the quotes, take the best quote and job done.
Fixed Income: RFS
The RFS basis is a little more complex, as there has to be a process to record a range of streamed quotes rather than simply one quote per bank at execution time.
Take the text above for Fixed Income and read it with a Foreign Exchange hat on.
Now, let's look at some options on how Fixed Income may work in the new world...
Same as now
Same as now
Fixed Income:Central Limit Order Book
If Fixed Income instruments move to trade on a central limit order book then that is more or less the same as equities now. As such, I would expect that the existing model of equity TCA could be used with some tweaks. But, the main problem is in this statement "ignore all non-primary markets and just look at the primary listing exchange of a particular equity". Since there may be many CLOBs and no-one venue was a "primary listing exchange" a firm will not be able to use the shortcut that some equity TCA firms use.
So, how will a firm work out the true market depth?
This is the big problem - a market participant may be willing to take on risk by quoting to buy a bond in a particular size but then issue this quote to multiple markets. Once hit, it pulls back from the other markets. But since there is no primary, which exchange should a TCA firm use as the benchmark? Will that end up like the split between Tradeweb (MBS and UST) and MarketAxess (Corps) or more like the Indian equity market where there is an ebb and flow between NSE and BSE in a particular name?
Initially it may be as simple as seeing that a quote was published at time X in size Y to CLOB Q and at time A in size B to CLOB T and since there is a degree of similarity in X and A and Y is equal to B it makes sense to suppose that there is really only a true quote of Y. Ok, so how long before market participants get smart to that and try and hit both quotes in order to get the trade done quickly? Pretty soon we are in the world of anti-gaming logic and algorithm combat that characterises the equity market....
I digress somewhat from the main point, of TCA for Fixed Income in a world of new trading venues. But I believe that this all can be done, but it's going to cost a lot more than many folks think or are willing to pay...