Machines invading bond markets

In the AFR this weekend I write that there has been a bit of chatter about “credit”—otherwise known as the bank and corporate bond market—becoming “the next quant revolution”, as the Financial Times recently described it. There is certainly some merit to this idea.

The $52 trillion, over-the-counter (OTC) global credit market is among the most inefficient asset-classes one will find. We’ve traded many tens of billions of dollars of physical bonds this year. And yet the price and size of those transactions are not disclosed in the way that they are in, say, listed equities.

Here Australia is different to the rest of the world. If you trade bonds in the US and Europe, there is some delayed price and volume reporting. In the US, price and volume reporting within 15 minutes is mandatory. While it is voluntary in Europe, most banks report trades immediately, although they do have the option of suppressing information should they deem it sensitive.

In Australia, the regulator has for some reason not got around to insisting that the ASX, which owns the monopoly clearinghouse, Austraclear, provide comparable levels of transparency. (We’ve repeatedly pushed for this, but to no avail.)

Imagine if you bought and sold billions of dollars of equities on the ASX, but nobody ever saw the price and volume associated with those transactions. That is what Aussie credit is like. Market-makers in bonds have always opposed price and volume transparency, arguing that it hurts the profitability of their business, and their capacity to inventory risk and intermediate flows. The counter-factual is that the global equities market has managed fine using a pure broking model without the need for market-makers acting as principals.

If global bond trading shifted to exchanges, there would likely be improvements in transparency and liquidity with a reduction in transaction costs. One interesting case study in this context is the ASX hybrid market. During the shock of March 2020, liquidity in many parts of the Australian corporate bond market evaporated.

And yet liquidity in the ASX hybrid market was outstanding. We observed daily turnover in ASX hybrids rising from around $40 million per day, on average, prior to the pandemic to over $100 million per day at various points during March 2020.

The main difference between BBB- rated ASX hybrids and BBB-rated OTC corporate bonds is their trading environment. The power of a listed exchange is that it forces all buyers and sellers transparently together into a single platform. By providing immediate price and volume reporting, the exchange gives participants confidence that they can execute orders in all environments.

In contrast, the physical OTC bond market is an informational black hole. Academic research has demonstrated that when there are extreme information asymmetries, you can get outright “market failures”, which is what we observed in OTC credit in March 2020. There were some notable exceptions, such as the major banks’ senior bonds. These assets have unusually high credit ratings, which signal very low risk, and are eligible for the RBA’s repurchase facilities.

Another important feature of OTC bond markets is that a bilateral trade between two parties means that machines, or computer-based algorithms, cannot compete. If I call-up CBA, and ask, “Can you sell me $10 million NAB Tier 2 bonds”, only the two parties involved get access to this information. This is very different to loading-up a bid on to the ASX for NAB shares where everyone in the world can directly compete with your flow.

Bilateral trading in this manner makes OTC credit effectively a “walled-garden” that is inaccessible to machines. This is why systematic “quant” strategies have been historically rare in credit markets. If you are not executing electronically, it is very hard to build an algorithm to pick-up the telephone, or to jump into Bloomberg chatrooms, and engage with counterparties looking for opportunities to buy and sell.

The advent of electronic trading platforms for OTC credit, including the likes of YieldBroker, MarketAccess, TradeWeb, and Bloomberg is, however, starting to change the market. Electronic trading allows for straight-through processing of transactions and the development of bona fide algorithms for executing orders. Increasingly, there are also artificially intelligent systems that can act as digital market-makers.

The need for algorithmic market-making is being amplified by the advent of passive ETFs, and the preponderance of passive styles more generally, which want to execute large numbers of transactions across portfolios comprising thousands of positions. In practice, this is easier for a machine, rather than a human being, to manage.

In Europe, as much as 60 per cent of the total value of OTC credit trading is now executed via electronic platforms. This is greater than the circa one-third share of e-trading in US credit, possibly because of the more dispersed nature of the European market where more banks compete for business.

The emergence of liquid electronic platforms has in turn enabled the development of artificially intelligent market-makers, which are slowly cannibalising the flows that would have once been the domain of human traders.

It is not clear yet whether this is actually changing the informational efficiency of the market. The main driver of both platform and algo trading has been passive products and ETF strategies, which are both informationally agnostic: they are allocating capital based on index weights, not on the basis of what assets are rich or cheap. This unambiguously detracts from, or reduces, the price efficiency of bond markets.

One benefit of these developments has arguably been liquidity, although perhaps only during benign market conditions. The ETF revolution has given vast volumes of retail money access to the once impenetrable OTC bond market. Yet to the extent that these capital flows are herd-like in their movements, sudden shifts into and out of an asset-class can make market liquidity one-sided.

Some argue that this is what afflicted liquidity in bond markets in March 2020 when the dominant passive investment flows suddenly wanted to exit at the same time. We certainly saw passive ETF products start trading at material discounts to their claimed net asset values, which were stale.

It should not, therefore, be surprising that we generally find that OTC credit markets are much slower to react to material news events than their more informationally efficient equity cousins.

When discussing which asset-classes lead others, some claim that bonds are, in fact, smarter than equities. By bonds, folks normally mean the interest rate futures and derivative markets, which are mostly exchange traded and immensely price efficient. In our experience, it is much more difficult to identify mispricings in interest rate derivatives than it is in the opaque and often very sluggish OTC credit markets.