The Evolution of Credit Trading
The bond market was traditionally a voice trading market. Up until about 20 years ago, the majority of the bond trades were conducted over the telephone, later by Bloomberg messages and maybe email in some cases. All of this would still be classed as voice trading, and this is how a lot of the trades are still conducted. With the arrival of the internet, we started seeing the development of electronic trading. Initially, this was just replacing some of the more straightforward voice trading, whereby the dealers created single-dealer platforms and clients were able to explore prices and execute trades with the dealer. Then with the advent of multi-dealer platforms such as Bloomberg, MarketAxess, TradeWeb and others, clients where are able to send requests to multiple dealers, receive multiple quotes back, and execute on the best price. More recently, we are seeing the advances in algorithmic trading where algorithms and computers place orders on both the client and dealer sides. We have also seen the development of systematic market-makers where computers respond to clients with prices. Today, there are a number of combinations of all these different forms of trading and are increasing in complexity every day.
Electronification of Credit Trading
Figure 2 puts into perspective how advanced electronic trading in bonds has become, as well as which segments of the market are most effective. This chart was produced by WBR/Flow Traders: “Beyond Inflection Point: The Future of Credit Trading” at the end of 2021. The chart shows that some segments of the fixed income market have been almost completely electronified, i.e., US Treasury trading and European government bonds. In some cases up to 90% of Treasuries are trading electronically now, and overall it is probably around 70 to 80% of the government bond market in the developed markets. Investment grade corporate bonds, both in Europe and US, are in what has been referred to as an inflection point, which is to say that roughly 50% of trading is now happening electronically by volume and the other 50% is still by voice. Emerging markets, the municipal bond market, the high yield markets, and others are lagging behind, but steadily increasing. Some estimates place about 70% of all the bond trades, by trade count, are now happening electronically, but only about 40 or 45% by volume. This means that a lot of the clients remain more comfortable executing larger trades via voice and smaller trades electronically. There is a trend for this to change and every year the execution venues and platforms are reporting an increase in the sizes of the tickets they trade. So, gradually there will be a shift to the right in this chart for all market segments.
Whilst algo trading has been around for a long time in equities, it is only now appearing more in fixed income, primarily in the more liquid markets like UST and European government bonds, as well as other markets too.
What are the reasons for this electronification of the market?
Firstly, regulation, compliance and transparency rules, pre- and post-trade, became more stringent after the financial crisis of 2008. These rules are easier to implement in the electronic trading world than in voice trading. Another driving force is the new trading protocols designed to increase efficiency of execution for both the buy-side and sell-side. There have also been advancements in algo trading. Whilst algo trading has been around for a long time in equities, it is only now appearing more in fixed income, primarily in the more liquid markets like UST and European government bonds, as well as other markets too. A host of non-banking participants are entering this space too, such as electronic market-makers in exchange-traded funds (ETFs), which account for a huge proportion of electronic trades on the trading venues. Their processes are almost fully automated in fixed income as well.
Electronic Trading Platform Features
Modern electronic bond trading platforms deliver advanced tools for users to control all aspects of the trade lifecycle, pre and post execution.
This includes:
- Live two-way trade level negotiation
- Automatic hedging of interest rate and FX risk
- Ability to fine-tune the number and type of recipients of a Request-For-Quote (RFQ) (ranging from traditional client-to-dealer RFQs to All-to-All inquiries, and every combination in between)
- Anonymity of requests and sensitive order information
- Simultaneous execution of multiple orders (lists, portfolio trading, etc)
There are many different types of electronic trading platforms available:
- Most common are the traditional RFQ-based platforms (BBG ALLQ, MA, TradeWeb), where the client sends a request to a dealer offering a bond and receives a price back from a dealer.
- There are many other types of platforms which have been developed or adopted to fixed income recently, for example, central limit order books (CLOB), which is essentially an exchange model. Many exchanges now list bonds, and many bonds are trading on exchanges. This is similar to the equity world but was less common in fixed income until recently.
- There are some inter-dealer broker matching platforms which are opening up to clients. Here, trades take place if there is interest in a specific level, from both the buyer and seller, when the broker posts.
- Finally, there are dark pool models, which again are more common in equities. There are however now dark pool models for fixed income, where orders are entered anonymously and trade negotiations happen if there is matching buying and selling interest in specific bonds. Dark pool information is not available before the trade and therefore does not affect the market.
All of these platforms have been steadily developing over the last 15-20 years, with volumes growing every year. Dealers are handing higher and higher numbers of RFQs every day using aggregation tools, which allow them to monitor multiple platforms at the same time. This allows the dissemination of prices to many platforms simultaneously. It also allows clients who have order management systems to send orders or requests quotes from multiple platforms concurrently. Despite this, the market is in a situation now, where these tools may not be enough. Traders on a busy trading desk may receive hundreds, if not thousands, of RFQs every day and this becomes almost unmanageable for a human to do efficiently.
Most of the large banks have already implemented some form of this systematic market-making for liquid markets and now are working on less liquid parts of the market.
Automating Market-Making Decision Process
So how do buy-side and sell-side firms deal with this situation? When looking at the next level of the automation of electronic trading, it helps to establish what a market-maker does on a day-to-day basis when they receive an RFQ. The market-maker needs to evaluate the requests and come up with the price for the client.
Factors that are being considered:
- Size and direction of the request
- Existing position in book
- Overall risk appetite of the desk/bank
- Client’s tiering/trading history
- Overall market conditions/factors relevant to the sector
- Inter-dealer market
- Risk limits/compliance policies
So, above is a typical list of questions a trader has to answer before responding to the RFQ. This is typically completed within 30 to 60 seconds for a standard bond trading request. Experienced traders can do this much faster, there are however some limitations. So how can the process be improved? With it being a very algorithmic process, it is something which can be encoded relatively easily. A computer can also do this job, and in some cases better. For example, a machine learning model can look at the data that needs to be processed to answer all of these questions and come up with a price. It can also go further and augment this data by using a data science platform which will take into account many more risk factors than a human brain can process. This, for example, might include the performance of relative value signals, markets and economic signals, non-financial signals like analysing Reddit or Twitter chats etc. All of these can be used as signals for a machine learning model in response to an RFQ many times faster than a human. A lot of the banks have already implemented systems like this and this is what we call a systematic market-maker.
The advantages of this are clear. Systematic market-makers can respond a lot quicker to a request than a human one. It can incorporate a lot more information into the decision-making process and it does not have the emotional component present in human market making or decision making. This may or may not be good, but generally speaking, it is probably better for the client. Most of the large banks have already implemented some form of this systematic market-making for liquid markets and now are working on less liquid parts of the market. To remain relevant and competitive in the market, most market participants, on the sell-side at least, will have to implement similar systems to automate this process.
Over the last two years, the pandemic changed the way people work and this also triggered some of the most challenging market conditions seen in history, with a lot of traders working from home.
Processing a Stream of RFQs
When automating responses to RFQs, requests will be split into “high-touch”, or more important requests, which will be routed to the human market-makers and the “low-touch” ones, which will be a higher volume, lower size, which will be handled by the systematic algo market-makers. The proportion of requests handled by the systematic market-makers is likely to change as they take on more and more of these tasks.
Over the last two years, the pandemic changed the way people work and this also triggered some of the most challenging market conditions seen in history, with a lot of traders working from home. In situations like this before, there was a tendency for most dealers to decrease electronic market making and trading and to rely more on voice trading. However, this was not the case in 2020, in fact, we saw a surge in volumes, which is a testament to the adoption of the electronic trading technology by the market.
Advances in Portfolio Trading
Portfolio trading is another driving force in the evolution of electronic trading. Portfolio trading allows clients to trade the entire portfolio of instruments in one go rather than line by line. This significantly reduces execution time and effort as well as providing additional benefits to both clients and dealers in the form of cost savings and increased efficiency. Portfolio trading is an excellent tool for ETF market-makers as well as the market participants who regularly trade portfolios of bonds. These trade protocols are now supported by most electronic execution venues.
Another point worth highlighting is the cultural shift taking place in buy-side and sell-side trading floors. Skillsets are changing too, given that a knowledge of markets and trading is as important as programing and data science. Both sales and traders need to adjust to the present-day reality and the new ways the market evaluates the performance of dealers.
Summary
The fixed income market is undergoing a shift in the way trades are executed. Amid growing competition and advances in electronic trading venues, both buy-side and sell-side participants are facing a steady increase in volumes of RFQs that have been facilitated by electronic platforms. In response, major market-making banks have started to roll out automated systems to emulate the decision-making process that a trader might follow when responding to an RFQ.
The end of LIBOR is going to have a very deep impact across the spectrum of fixed income. The shift to other reference rates, such as secured overnight finance rates (SOFR), affects products including OTC derivatives, short-term borrowing, floating-rate borrowing/lending, and other financing activities. The LIBOR transition impacts almost every part of the financial services industry, including banking, capital markets, insurance and asset management.
The fixed income market is slowly but surely moving towards automation and new technology is reconfiguring the operating landscape. This steady digitization has accelerated the percentage of fixed income trading that occurs electronically. The overall automation of front-to-back processes, use of data, and use of machine learning has also increased.
Fast, accurate analytics is a prerequisite for any automated bond trading or market-making platform. With traditional platforms relying on outdated technology that is cumbersome and inflexible, firms are turning to technology providers, like Quantifi, that provide sophisticated model libraries and pricing frameworks built on new technology.