Wednesday, February 07, 2007

What's the liquidity premium for corporate bonds?

Sometimes I'm a bit slow. But I've finally made a connection which I think is an important one.

A recent study suggests that liquidity in the corporate bond market has improved. I wrote about this study back in September. If this improved liquidity is real and permanent, what is the impact on spreads overall? First, some quick background on liquidity.

In a recent discussion in the comments section on TIPS, one commenter pointed to an academic study suggesting that illiquidity in TIPS explained some of spread vs. Treasuries. Liquidity, here measured by the bid/ask spread, is worth something to every investor. A tight bid/ask means that the transaction costs of entering and exiting a trade are low. The more often you trade, the more important the liquidity of your item is.

Since liquidity is valuable, a relatively illiquid bond should trade with a higher yield than a similar bond with better liquidity. On-the-run Treasury bonds have the best liquidity in the U.S. market, followed off-the-run Treasuries, TBA mortgages, large issue non-call agencies, and large issue high quality corporates. With corporates, the liquidity can be very name-specific.

Measuring the liquidity premium isn't easy, and the premium is probably not constant. For example, the difference in yield between the on-the-run Treasuries and the nearest off-the-run issues is about 9/10ths of a basis point right now. There we have two items with exactly the same credit and interest rate conditions, but a slight difference in yield entirely because of liquidity. While this is irrefutable evidence that there is a liquidity premium, the bid/ask difference between on and off-the-run issues is a mere 1/64th of a point (called a "plus").

More information can be gleaned from the agency market, e.g., Fannie Mae and Freddie Mac. Both are considered to have extremely strong credit quality and perhaps even have the implicit backing of the U.S. Treasury. The most liquid 10-year agency issues trade with around 4.5 ticks bid/ask, or about 4 ticks more than the 10-year Treasury. These issues offer about 32bps of extra yield over Treasuries. How much of this can we attribute to credit quality and how much to liquidity?

The best way I can think of is to compare TBA Ginnie Mae MBS to Fannie Mae MBS with the same coupon. In the TBA market, both have the same bid/ask, but Ginnie Mae is explicitly U.S. government guaranteed. The difference between the two is currently 17 ticks on the 5.5% coupon. The theoretical difference between an agency and Treasury measured in dollar price is 81 ticks (2.5 points). So 64 ticks, or exactly 2 points, is attributable liquidity differential between Treasuries and Agencies. In yield terms, that implies that the liquidity premium for agencies is 25bps.

So where is all this going? If all of my analysis above is correct, then a market where the bid/ask tightens by 4 ticks will see the overall spread of the issues contract by 25bps.

Back to the study I mentioned at the onset of this piece. It claims that the average bid/ask (quoted in spread) is now 8bps tighter than it was in 2002. That's about 20 ticks on a 10-year issue. If my estimation is correct, that should have allowed corporate bonds to tighten by 125bps entirely due to liquidity. Since June 2002, investment grade corporate spreads have contracted 149bps.

Now, let's step back. We've come up with 125bps of tightening by building a tower of assumptions. Notably, 2002 was period of particularly poor liquidity, as we were still coming out of the Enron/Worldcom scandal period. So the researcher's estimation of 8bps of bid/ask tightening might be biased based on the measurement period. Regardless, there is not one sell-side trader or experienced bond salesperson who does not say that the bid/ask has tightened from where it was in the late 1990's. So maybe its 8bps, maybe its 5, maybe its 3, whatever. It is still a very large influence on spreads.

What are the trading consequences of this? Take another look at the graph I published on Tuesday. High-grade spreads are only 19bps tighter than the long-term median. If the bid/ask has permanently tightened by a mere 2-3 bps, the overall spread market should have moved more than 19bps tighter. In my 2007 forecast, I said I'd be long investment grade corps vs. Treasuries, because while they might widen slightly, it wouldn't be enough to overwhelm the yield advantage. Now I'm wondering if the performance might actually be even better...


Anonymous said...

I've heard that some CB's have been buying high grade corporates

Tom G. said...

I've heard that as well. I've also heard other foreign buyers have been active in the corporate market.

Anonymous said...

comping GNMA TBA vs FNMA TBA is incorrect even at the same coupon. gnma production is significantly lower and prepay expectations are different even at the same coupon.

Tom G. said...

The theory is that GNMA pays slower than conventional. But guess what? 5% or 6.5% coupon, the price differential is 20 ticks both ways. I'm looking at a live screen as I write this, so those are good levels.

The 5% is a 3 point discount and the 6.5% is a 2 point premium. So it doesn't make much sense that GNMA would trade universally higher than conventional on prepays alone.

Not sure what you mean about there being consistently less production. I mean you're obviously correct, I'm just not sure what you're driving at.