ft.com/alphaville All times are London time
News
Or: do wide high yield spreads over US Treasuries predict a recession?
We ask because this is an interpretation we’ve seen a fair bit in the media over the last few weeks.
And if true, it’s especially alarming since on Tuesday the BofA Merrill Lynch High Yield Master II Index, which measures option-adjusted spreads of high yield bonds over Treasuries, closed at 754 basis points on Tuesday — a new high for this cycle. (It fell back slightly yesterday.)
This is where spreads are now (chart via Fred):
The widening spreads came in the same month that issuance all but evaporated and high yield funds had big redemptions. According to Barcap, August was the eighth-worst month for high yield returns in twenty years, with the bank’s High Yield Index declining 4 per cent and erasing nearly all of its year-to-date gains.
But back to our original question: does junk predict a funk?
If the answer is yes, very accurately, then we’re in trouble.
But having taken a look at a few attempts at using these spreads as a forecasting tool, our answer is no, not very accurately. Or at least, the metric is of limited use.
The BNP PIP model
Martin Fridson of BNP Paribas Investment Partners, in a note out this week, puts a number on the High Yield-implied probability of recession, which he updates regularly and which gets a fair amount of attention:
The present spread indicates a 63% probability of recession, according to our methodology. … The heightened recession risk justifies caution, but any investor who is absolutely convinced there will be no recession should consider this an excellent entry point to buy high yield bonds.
We called Fridson to ask for detail on how he arrived at this number, and after getting a better grasp of how the model works, we think the wording here is imprecise, or at least should be more comprehensive.
BNP’s econometric model involves four variables — industrial production, capacity utilisation, high yield default rates, and a question about lending standards from the Fed’s Senior Loan Officer Survey — that historically account for about 84 per cent of the variance in high yield spreads.
If you plug in what those four variables are now showing, according to Fridson you would expect high yield spreads to be about 347bps above US Treasuries. Of course, we’re now at a whopping 754bps.
One might conclude from this discrepancy that high yield spreads are too high — in other words, that the market has run away from fundamentals — but that’s the opposite of the conclusion drawn in the note.
Why? Here’s where things get complicated…
Fridson plugs in what each of those four variables would show if we were in a recession. Where does he get the inputs for this assumption? From the averages of those four variables during the period between December 2007 and June 2009 (the last recession). The model then spits out a number of 988 bps.
So to recap, we have:
– What the model estimates spreads should be: 347bps – Where spreads actually are right now (as of earlier this week): 754bps – What the model would show if we use 2007-09 inputs: 988bps
Well, 754 is 63 per cent of the way between 347 and 988; hence the conclusion that spreads are showing a 63 per cent chance of recession.
But this is clearly a selective interpretation, and it’s entirely contingent on economic expectations that have already been established.
In other words, spreads are only predicting a 63 per cent chance of a recession if you already assume that we’re in a recession.
But if you think we’re not in a recession — or to be precise, if you think the four variables are accurately reflecting the state of the economy as per the BNP analysis — then you actually think spreads are likely to contract.
And that’s the thing. This model isn’t meant to predict a recession — it’s meant to give investors a guide to whether high yield bonds are overpriced or underpriced once they’ve already decided for themselves whether we’re headed for a recession.
To be clear, Fridson is forthcoming about all of this. He told us that a) the 2007-09 recession was arbitrarily chosen and the model would spit out different numbers if we chose a previous, shallower recession, b) the model has a necessarily limited sample because the required data only goes back to the early 90s (a few years after the advent of junk bonds), and c) when backtested, the model has previously shown conflicting numbers at the start of prior recessions.
To elaborate on point c), the model found that spreads were almost precisely where they should be (near 600bps) in December 2007 — that is, the model was projecting a zero per cent chance of recession. And when applied to March 2001, the estimated spreads were actually well above actual spreads. By the same logic, one would have concluded that spreads were actually predicting a negative probability of recession. Of course, the point is that the same logic can’t be applied because it relies entirely on the viewpoint of whomever is interpreting it.
Fridson knows all this — he’s the one who told us — and plans to continue tweaking the model while backtesting it further. When we asked him, he doesn’t say that people should use the model to predict a recession or a recovery, nor does he say that we’re necessarily heading for one now. He says that the number can help investors determine whether high yield bonds are overpriced or underpriced for varying assumptions about the economy. But those assumptions are up to investors to find for themselves.
Anyways, all of this matters not so much for how BNPPIP clients use it (unless you’re a BNPPIP client) but for how it gets disseminated elsewhere, especially in the media. We’re not here to point fingers, only to caution journalists and others who pass along the BNPPIP model (or any other) to include the necessary caveats.
But since we’re also trying to find out if actual spreads are a useful leading indicator, let’s move on.
The 700bps rule
This is a guideline we hear now and again — that spreads of more than 700bps signal a recession. But at best it seems to be more of a coincident rather than a leading signal, and not always a reliable one.
Moody’s Analytics writes that spreads above these levels are “associated” with the “nearness” of a recession. Goldman, in a note out this week, writes that they are “consistent with mild recessions of the past.”
Yet even in the limited history we have to go on, there has been one instance where a recession failed to follow after spreads penetrated 700bps and another instance where a recession had begun before spreads reached 700bps.
We already mentioned the latter: spreads were below 700bps at the end of December 2007, though they were climbing.
As for the former:
Look right in the middle of the graph — the second half of 2002 — and you’ll see that like our current situation, the uptick in spreads arrived quickly, followed soon after the end of a recent downturn and came at a time when equity markets didn’t quite trust the strong profits that companies were displaying:
For example, the high yield spread averaged 722 bp during January-August 2008, which included a very wide 804 bp for August. However, despite the even wider 965 bp average spread of 2002's second half, a recession did not materialize.
There are plenty of differences between now and then — the spike in spreads then was triggered by Worldcom-led worries about corporate balance sheets, which eventually eased. And there have also been other instances where rapid rises (such as after the downfall of Long Term Capital Management) didn’t bring about a recession. Mostly these instances were related to market freakouts, but the point is that an accurate reflection of a downturn is only one possible reason for spreads to climb; there are others, both fundamental and technical.
It’s the US Treasuries what did it
To get back to the current situation, here’s a point from Citigroup that’s been echoed by others we’ve spoken with:
To be certain, the gap has widened out sharply as investors have adjusted to potentially slower global economic growth and European sovereign debt worries. However, we note: 1) nearly half the "delta" of the widening reflects further decline in 10-year treasuries and half from higher HY yields; and 2) from a historical perspective, HY spreads remain well contained relative to 2007-09, a more problematic macro period and sharp sell-downs for alternative and traditional managers alike.
Richard Smith, head of high yield capital markets at RBS, explains it this way:
I think those models are wrong because the spread is not true. The 10-year is being held down at that low 2% by [monetary policy], and by a flight to quality, not necessarily by what’s going on with commerce. Rates should be higher, but they’re not. It’s absolutely about fear, but it doesn’t necessarily mean anything about what’s going to happen next.
Of course, the move in 10-year US Treasuries is itself a signal of growth expectations, so this isn’t a perfect criticism. But the 10-year is also subject to technical moves, and Smith’s point here is that the moves in high yield bonds themselves haven’t been as dramatic as the spreads would indicate.
And if you look at the companies that have actually issued these bonds, their balance sheets are fine after a couple of years of Fed-assisted refinancing. Here is BlackRock’s Jim Keenan, who runs the asset manager’s leveraged finance group:
Nobody is looking to the high yield market to start having a lot of defaults. The corporate credit market generally is in pretty good shape. Most of these companies have used the recovery in the last three years to extend liability and take down interest expense. So a lot of companies that had near-term maturities have extended them out by 7-10 years, which would put you in the 2018-2020 range. The wall of maturities that everyone was worried about has been dealt with, or at least a lot of it.
We are NOT suggesting that we’re going to avoid a recession. Honestly, we have no idea. It’s only to cast doubt about an indicator that seems to get a lot of attention as a useful precursor of recessions with only questionable justification.
Read Full Article »