As firms are moving more and more into making data-driven decisions, there is one market that has been lagging behind: the bond market. The bond market has for long been characterized as old fashioned and large bond trades are still happening over the phone. When it comes to the timing and volume of bond issuance, it is still dependent on (guess)estimates of demand. This can be done through for example analyzing the liquidity of the bond market, or simply having a chat with an asset manager. Of these options, the latter is not uncommon.
Its accuracy is high as the predictions on the yield of new bond issues have been off by only 0.02 percentage points on average.
This is now changing. One example of this is the Canadian fintech startup Overbond. Overbond was founded in 2015, and has since then grown rapidly. Its platform is based on a form of artificial intelligence, more specifically, machine learning algorithms. One thing it does is predict timing and prices of bond issuances based on for example data such as credit ratings and secondary trading, which it gathers. Its accuracy is high as the predictions on the yield of new bond issues have been off by only 0.02 percentage points on average. So far, Overbond is only available in Canada.
Moreover, liquidity has been falling over the past decade. This is well captured by Tony Rodriquez who said: "if average market liquidity was a 5 on a scale of 1 to 10, today's markets might be something like a 3 or 3½". This also has a negative impact on the bond market as the illiquidity premium becomes larger. The introduction of easily and cost-efficient technologies could be the key in solving this liquidity problem.
Eighty percent of the deals is still done over phone, email or some form of chat.
In the US, which has an $8 trillion corporate bond market, thus being the largest in the world, about 80 percent of the deals is still done over phone, email or some form of chat. However, over the past few years there has been an increase in electronic trades. Currently, a new platform for ordering bonds is emerging. Ipreo, which is jointly owned by Goldman Sachs Group Inc. and Blackstone Group LP, is going to Wall Street this year. Banks such as Commerzbank, Unicredit, NatWest Markets, UBS, ABN AMRO and Natixis, have already announced that they will take orders from investors through Ipreo’s platform.
Compared to other branches of the financial industry, the bonds market had been quite slow at adapting its technology. In retail and asset management, fintech has been far more utilized. Some think it is because of the fact that investment banking depends on relationships. Regulatory requirements also put restrictions on the development and surprisingly much so through simply using up a large part of the tech budget of investment banks.
At the moment most companies and governments need the assistance of investment banks to determine the timing and price of a bond issuance.
These developments are likely to be monitored very closely by investment bankers. At the moment most companies and governments need the assistance of investment banks to determine the timing and price of a bond issuance. Furthermore, once the bank underwrites an issue, it bears the risk of failing to get good prices for the bonds. The developments in machine learning could potentially harm the business of investment banks as new platforms utilizing it can assist companies and governments, at a lower fee. On the other hand, several investment banks have mentioned that they intend to use these technologies to actively approach companies when it is a good time to place a bond issuance.
Not only companies have taken an interest in machine learning in the bond market. Ganguli and Dunnmon (2017) have written an academic article on which type of model performs best in predicting bond prices. They found promising results on the accuracy and speed. However, there has already been some critique as the model overestimates volatility. This shows that there is still room for improvement in the logarithms currently being used. A collaboration between academics and bankers would be interesting in terms of creating very accurate logarithms.
Incumbent companies likely have to actively participate in this development or face the risk of losing market share.
The development that has been happening in the stock market is finally picking up in the bond market as well. Incumbent companies likely have to actively participate in this development or face the risk of losing market share. We see different paths in the way this is going. On one hand do the already more established electronic bond order platforms, such as Ipreo and Overbond, have a first-mover advantage and might therefore settle down with a considerable market share, and being backed up by the multinational investment banking giants certainly helps. On the other hand, more and more fintech companies are popping up and they are likely to play a large role in the future of the bond market. As more and more is discovered of machine learning and AI, startups definitely stand a chance against the innovations of the more rigid tech departments of investment banks.