On a regular buy-side bond trading desk a fairly standard workflow is to send an RFQ for an order to an electronic venue such as MarketAxess, Tradeweb, Bloomberg or Bondvision (other venues are available - there's a list here).
Different platforms have different procedures for how RFQs are handled:
- different disclosure to the banks on the rejection by the buy-side (was it a did not trade, traded with someone else, traded with someone else and your firm was 2nd or nth best price)
- one way or two way prices
- different numbers of competing quotes allowed on an RFQ
- one time price or streaming for a short period
It's been a while since I looked at that and with the plethora of new venues I expect an update on that analysis would be good but that can wait for another day.
This post is about a step before the RFQ is sent. There are two decisions that I want to look at:
- Use which platform?
- Select which banks to receive the RFQ?
In the past the first question was pretty simple - Tradeweb for MBS/US Govies, MarketAxess for corps and Bloomberg will do for anything (no, it was not that simple, it's just relative). Since not all banks are on all platforms there is a need to ensure that a buy-side has a number more than one to ensure coverage by major banks required for the different markets in which the firms invests. Now we have perhaps 85 platforms it's unclear. The requirements for best execution and the regulator mandated changes to the way that banks will quote mean that this area will be a minefield for some time.
Modest prediction: The functional equivalent of a smart order router will appear in the arsenal of sophisticated buy-sides that are active in the fixed income market.
The second question is complex and merits a longer discussion...
Since there is no FINRA TRACE in Europe (although it may not be the hoped for "golden source") there is no way to access a time and sales data in near-time (15 minutes reporting window) as you can see below or via this link to the FINRA website.
The question of which banks to select to send an RFQ is not one that can be answered without data, experience or preferably both. In some cases it's simple - if a bond is illiquid, try the book-runners and hope. If it's liquid, a number of banks should be willing to quote.
Of course, no point asking for a quote from a firm with which a portfolio cannot trade, so best to check any compliance restrictions before asking for quotes.
Here we get into a murky area. A sensible bank will use a technology platform such as kdb+ to track every RFQ, RFS and order they ever receive. This will then be combined with other data sets to allow a bank to fine-tune a quote for a particular client. This is simple - if you bought a large position of a bond from a bank and later on ask for a two way quote in that bond, the bank can hypothesise that you are actually a seller of the bond. Of course, you may be building a bigger position but this sort of analytic is merely guidance rather than strictly deterministic.
So you may want to consider where the position came from when asking for a quote.
In a world where banks no longer have the balance sheet to actually hold the bonds at risk for any time the market has of course seen a drastic drop off in transactions and liquidity. Where banks do trade it seems to be in one of these ways:
- Bank buys a portion of the trade and works the rest
- Bank asks for an exclusive period to work the trade without putting any capital at risk
- Bank looks to interdealer market for a price, spreads it and then proposes that to the buy-side
The issue here is one of information leakage and that's well known to folks in the equity market. So when sending an RFQ a buy-side needs to consider the impact on the market of that information hitting perhaps five banks at one time. Each of which may then look to the interdealer market to offload the risk which will then move the market.
For a project some time ago for a technology vendor I was asked to assist with designing a better way to do this. Times have moved on but the basic ideas are worth setting out:
Rather than an RFQ model, start with a small number of liquid bonds and ask banks to price them in one or more sizes on a RFS streaming basis. These streams would be consolidated into a synthetic order book, rather like the way different equity market order bokos are consolidated in a platform such as Fidessa. The streams would be persisted using kdb+ to create a time-series of data alongside a number of price feeds from reference instruments such a govies and futures. The data would then be analysed with R to attempt to reverse engineer the pricing models of the banks to effectively reverse the model and allow the buy-side to take back control from the sell-side.
One of the outputs of the proposed system was a broker selection model to allow a buy-side user to have a series of banks suggested based upon bond/size/side. In the case for an actual trade where the size was within the streaming size it would be possible to simply hit/lift the quote. The main value though would be in automatically populating a series of suggestions for brokers for an RFQ where a dealer wants to use that process. Clearly this would be a suggestion rather than mandatory but it would have provided some intellectual underpinning to the selection of brokers for an RFQ.
Now, in a world where the banks no longer apply capital to bond trading the ability to reverse engineer their pricing models is not particularly valid or useful for RFQ recipient selection.
Which brings us back to the original question - how should a buy-side dealer pick which platform and which brokers?
Which is where we get to the new world of Fixed Income platforms such as Algomi - Honeycomb and Bondcube (amongst many others)...
There is no simple answer to "what is best execution?" and oftentimes the answer is "it's a process rather than a specific outcome". As market structure changes over the next few years I expect that we will see a good deal more innovation in the buy-side decision support toolbox to assist dealers in selecting the correct way to interact with the market. Some of this has already been covered in Buy-side analytics: Equity execution venue analysis and Buy-side analytics: Beyond TCA.