Market Making
Being a dealer is not an overly complicated business in principle. Yes there are tons of technicalities that make the whole system quite complex, but at a high level the function of a dealer is pretty simple. A dealer is just buying securities at a low price while selling the same security to someone else at a higher price. For example, dealers will tell everyone that they will buy a bond for $99.90 and sell the bond for $100.10. If one investor hits they dealer’s bid and another hits their offer, dealers are picking up 20 cents a bond.
Sometimes dealers will buy more than they sell and build an inventory that’s too large. If that’s the case, often they can sell those bonds in an interdealer market (market where dealers exchange inventory). Spreads are tighter because volumes are higher. Perhaps the spread in the interdealer market is $99.95/$100.05. If that’s the case, the dealer will have bought bonds at $99.90 from investors and then hit the bid of another dealer in the interdealer market for $99.95, or 5 cents a bonds. Not nearly as good, but still pretty great if you’re trading billions of dollars a day.
Being a dealer is so simple that robots are doing the work for dealers. Moreover, there are a number of high-frequency trading (HFT) firms that are acting as pseudo-dealers by providing the market with liquidity. They will buy an issue and sell it fractions of a second later for fractions of a penny. These HFT firms, like dealers, are not into taking any sort of market risk. They just want to deal.
This is a great business when markets are docile. When prices begin to move, it gets a bit trickier. Dealers/HFT’s don’t want to get caught holding a bunch of bonds that are declining in value. If a bunch of investors or dealers hit their bid, that’s a pretty good indication that the value of the bonds are dropping. If the value is dropping, it may mean that they have to sell these bonds at a loss. If you are a dealer and just want to deal (you don’t have an opinion on value and don’t want to speculate on future price movements) you aren’t into taking directional market exposure. You will do your best to hedge in futures or swap markets, but if the price is moving fast, this is tough to do. Dealers also don’t have the balance sheet to take large positions because of Basel III regulations. This often leads dealers to widen their spreads, decrease volume of bid/offers, or just stop trading altogether until there’s stability in the market. Most trading algorithms have automatic shutoffs for when volatility gets too high.
This has a major effect on the liquidity of the bond market. Volatility typically strikes in times of market stress. These are also times when many financial institutions are looking to sell their bonds, as we’ve discussed in the past. If there’s no liquidity in the market and you are trying to sell your bond, you either have to not sell it, or sell it for less than its worth. Unfortunately for many, we saw some of the most volatile fixed income markets ever this year. This chart from NY Fed shows volatility for 5 year, 10 year, and 30 year US treasuries.
US treasuries are the most liquid fixed income instruments available, but even treasuries saw their bid ask spreads on the interdealer market explode in March.
Not only did spreads blow out, but the depth of the order book (how much people want to buy/sell on a given exchange at any given time) on the interdealer market also plummeted.
One of the main culprits for this sudden drop in liquidity is the trading algos that we mentioned above. Josh Younger, an analyst at JP Morgan, recently shared some data that made that abundantly clear. It shows the amount of liquidity provided by both humans and robots. You can see that from 2015 and onward, robots have dominated the dealer business in treasuries. You can also see that in 2020, the amount of liquidity provided by algos fell off of a cliff.
Algorithms are incredible at doing the thing they are told to do quickly and efficiently. As we saw, being a dealer in normal times is pretty easy. Lots of “if this, do that”. Perfect job for some code. When being a dealer gets more complicated, and all of the sudden there are many conflicting factors to the problem (interest rate expectations, balance sheet considerations, regulatory actions, etc., etc.), algorithms have no idea what to do. They can’t process, synthesize, and draw conclusions with all of the new and relevant information like humans can. If they aren’t told to just stop trading, bad stuff can happen. We have seen this occur in numerous flash crashes in the past.
As we move into a market structure that involves A) More regulation restricting dealers ability to expand their balance or take on market exposure (ie. prop trading), and B) Algorithms continuing to dominate trading in over the counter (OTC) markets, we may see more of these liquidity issues arise more. That said, perhaps we could see central banks like the Fed leave liquidity facilities open so that there is a place to sell bonds in liquidity events. The former would suggest more crashes in what can be described as “risk-free assets” in stressful times (we didn’t see a crash, per se, but US 10 years did see their yield go from 0.54% on March 9 to 1.18% on March 18 due to rapid deleveraging). The latter would suggest that perhaps there is a floor on bonds at some price.
