IHS Markit

We believe different markets have different quoting patterns/conventions and enforcing both bid and offer to a quote for every market would lead to the “loss” of many quotes that represented a commitment from the quoting party. For example, in the bond market we observe a clear distinction between dealer-to-customer (D2C) and dealer-to-dealer (D2D) quoting conventions. While the former is primarily one-sided (i.e. bid or offer) the latter can be predominately categorised as 2-sided (i.e. bid and offer). On the other hand, in the credit derivatives market we observed both D2C and D2D quotes to be predominantly 2-sided.

Moreover, we believe that the requirement for a quote to be “legally obliged” contradicts markets modus operandi where the commitment to trade on a quote is based on market conventions rather than a legal agreement between the counterparties to the quote.

Our recommendation is, therefore, to leave it to the banks to determine whether to use 1-sided or 2-sided quotes based on the different markets and their understanding of the quoting conventions of each market. One could consider that, if 1-sided quotes are being used, then banks should be looking also for supporting information within the quote, such as indication of “firmness” or “Axe” in the bond market.
Due to the vast reliance of the cash bond market on quotes to prove modellability and as we have tens of millions of quotes in our database, we focused our analysis on this market. Checking modellability of Risk Factors (RF) can potentially be misleading as different banks will have a different RF taxonomy so we decided to examine modellability on an instrument ID (ISIN or CUSIP) basis rather than other possible dimensions. Furthermore, we examined GSAC bonds and MUNI bonds separately. Finally, the analysis of the impact of imposing bid and offer is on quotes only and did not incorporate post-trade information into the overall population of RPOs.

GSAC Bonds:
• As discussed in the answer to Q1, D2C market tends to be 1-sided. Our observation shows that over 90% of the D2C quotes are 1-sided.
• The number of unique instrument IDs in scope when allowing only 2-sided quotes (vs. allowing both 1-sided and 2-sided) drops by ¬40%. However, the number of modellable unique instrument IDs drops by ¬24%. Reason being only ¬44% of the unique instrument IDs are found to be modellable when allowing both 1-sided and 2-sided while it goes up to ¬56% of the unique instrument IDs being modellable when using only 2-sided quotes.

MUNI bonds:
• Our research shows that the reliance of the MUNI bond market on quotes is larger than the general bond market with less than 25% of the quotes (per unique instrument ID) turning into an actual trade within the following week to the initial quote. Furthermore, the skew towards a 1-sided quote is even larger due to the fact these are mostly retail based, with very few D2D quotes (these also tend to be 1-sided).
• The number of unique instrument IDs in scope when allowing only 2-sided quotes (vs. allowing both 1-sided and 2-sided) drops by ¬96%. The impact of this on modellability can be looked at in 2 ways:
1. On the one hand only¬3% of unique instrument IDs are modellable to begin with, even when using both 1-sided and 2-sided quotes [when ¬80% fail both modellability criteria (24 and 4in90 days RPOs)] so any move from both 1-sided and 2-sided will have limited impact on absolute change.
2. On the other, due to the fact that there is a 96% drop of unique instrument IDs in scope, the total number of modellable bonds drops by 99.1% and therefore, they become virtually non-modellable across the board without using post-trade information.
In the FAQ published by the BCBS in January 2017 (https://www.bis.org/bcbs/publ/d395.pdf), the following Q&A was included, introducing for the first time the principle of “non-negligible volume” (paragraph 2.6; p.9):
Q4. Are all transactions and eligible committed quotes valid as real price observations, regardless of size?
Answer: Orderly transactions and eligible committed quotes with a non-negligible volume, as compared to usual transaction sizes for the bank, reflective of normal market conditions can be generally accepted as valid".

Following the introduction of the concept of “non-negligible” we decided, using IHS Markit’s resource of over 2 million Interest-Rates Derivatives RPOs, to create a sense of what would constitute a “non-negligible” volume.
To do this we developed the following principles to remove “noise” in our data:
• Only “New” trades were included (i.e. no novation, amendment etc.);
• Converted domestic notional to US$ using relevant day FX spot rates to compare like for like;
• Split the check between Dealer-to-Dealer (D2D) and Dealer-to-Customer (D2C);
• Removed dealers with less than 20 trades in period checked; and
• Excluded all trades under $1,000 as a sanity measurement.

Our immediate findings were that there is no ideal number that fits all scenarios and a distinction between D2D and D2C as well as G4 versus non-G4 must be made. After running several statistical measurements, we came to the conclusion that the “benchmark” should be a D2C 1-percentile mark.
Our findings were that non-negligible threshold should be: G4 - $291,000; Non-G4 - $37,000.

It is also important to mention that we did not distinguish between instruments (i.e. swaps versus swaptions) or give different currencies different weights based on their overall volume (as set out, for example, in the BIS triannual report). Our analysis also did not distinguish between top tier-1 banks and tier-2 banks.

Regardless, our recommendation is that the EBA should not propose a single number and/or a single methodology as different asset-classes, instruments or currencies have different characteristics and levels of liquidity. Although we would not recommend it, if the EBA believes it must insist on a more quantitative approach in defining a threshold, we believe that using a D2C 1-percentile mark on different asset-classes, instruments and currencies might provide a workable threshold."
Our advice, as set out in our answer to Q1, is that the EBA should not impose a 2-sided quote only system as markets have different quoting conventions and these conventions are not legal agreements between counterparties. Therefore, we believe defining and checking bid-offer spread as set out in this question should be somewhat redundant. If the EBA takes the view that committed quotes should only be 2-sided, we would like to emphasise that defining and checking for an “unreasonable large bid-offer spread” would be extremely difficult.

However, we thought it would be helpful to share our findings on the distribution of the spreads observed in IHS Markit’s population of Cash Bonds (committed) quotes. This should help the EBA get a sense on how spreads behave, but, as illustrated in our previous answers, we believe this should ultimately be each bank’s own decision based on the information they have in-house or provided to them by a 3rd party vendor as enrichment to the core RPO information.

Spread (S) % Cumulative %
S<0.25 22.80% 22.80%
0.25<S<0.5 20.80% 43.60%
0.5<S<0.75 19.20% 62.70%
0.75<S<1.0 16.30% 79.10%
1.0<S<2.0 18.30% 97.30%
2.0<S<3.0 1.80% 99.20%
3.0<S<4.0 0.40% 99.60%
4.0<S<5.0 0.10% 99.70%
5.0<S<7.0 0.10% 99.80%
7.0<S<10.0 0.10% 99.90%
S>10.0 0.10% 100.00%

(For a formatted table see attached letter)
We believe that all 3rd party vendors should be regularly audited by an external audit firm, using a market acceptable audit framework. Moreover, we believe the vendors should be able to provide a public attestation of the audit to the banks. It is important to coordinate such requirements across jurisdiction to avoid vendors having to undergo multiple, duplicative audits.
While the EBA is taking note of MAR31.26 of the Basel text in Q14 of this CP (in the context of calibration of banks’ internal model), we believe the EBA should also recognise these principles in the context of ensuring the verifiable prices provided by 3rd party vendors are representative of the prevailing market prices.
NA
NA
NA
NA
NA
NA
NA
NA
Gil Shefi
I