Admittedly, I’m not wholly enamored with this approach to deciphering what various assets are “saying” about the odds of a recession, but then again, I can’t think of a predictive recession model that I am entirely enamored with, so I guess it’s worth highlighting.
In a note dated Friday, JPMorgan attempted to discern what equities, IG, HY and 5-year Treasury yields are predicting in terms of a US recession and the approach is pretty simplistic or “mechanical”, as the bank calls it.
Of course predicting recessions is a notoriously difficult endeavor (and please, spare me your “the yield curve always works” spiel because frankly, it doesn’t – there are unpredictable lags, questions about which curve to use, and on and on and on), so I suppose there’s an argument to be made that the more straightforward the better – there’s elegance in simplicity. Or something.
In any event, on equities, the bank’s Nikolaos Panigirtzoglou just takes the ratio of the current peak-to-trough drawdown and the average decline of the S&P during the past 11 recessions. “So far the S&P500 has declined by 16% from its peak, so equity markets price in a 16/26=62% chance of an average recession”, he writes.
(JPMorgan)
The “mild” versus “deep” distinction just refers to whether earnings dropped by more or less than the median. “On that basis, we can infer that equity markets price in a 16/18=88% chance of a mild recession or a 16/33=48% chance of a deep recession”, Panigirtzoglou says, referencing the averages in the “deep” and “mild” categories seen at the bottom of the table.
On credit, the approach is similarly straightforward. “We estimate that the average US HG spread level during recessions (over the past six) is 200bp and the average level outside of recession is 113bp”, he continues, before suggesting that based on that, the math is as follows:
US HG credit spreads have widened sharply over the past month and currently stand at 185bp over USTs. This suggests that HG credit markets price in (185-113)/(200- 125)=83% chance of US recession and are discounting tail-like market risk scenarios.
But this time is indeed “different” for high grade. Citing the bank’s HG team, Panigirtzoglou notes that between lower Treasury yields, more duration risk and the proliferation of BBBs (he just says “lower average credit quality’), everyone should go ahead and assume that HG spreads are going to be wider than in the past in a recession-like scenario. When you do the same math above while taking all of that into consideration, you end up with HG pricing in a ~55% chance of a recession.
On junk, spreads haven’t blown out nearly to levels seen during previous recessions (you can probably thank years of QE and the concurrent suppression of idiosyncratic risk for that), which mechanically means HY is the outlier, pricing in just a 12% chance of a US recession, in stark contrast to equities and HG credit and also in stark contrast to 5-year Treasury yields.
On the latter, JPMorgan notes that the math obviously needs to be adjusted to account for the fact that the low starting point means they cannot be expected to fall as much as they did during historical downturns. “In fact, an argument can be made that with a lower neutral rate in the US post Lehman crisis, around half of that seen in previous cycles, 5-year UST yields should decline by half of the 2.1% peak to trough seen on average in the previous recessions”, Panigirtzoglou says, on the way to suggesting that “this means the 70bp decline in the 5- year UST yield since early November points to two thirds rather than one third chance of US recession.”
(JPMorgan)
Ultimately, this exercise yields the following cross-asset recession probabilities, which you can take or leave, depending on your assessment of whether the approach outlined above makes sense.
(JPMorgan)
In the end, I suppose this is probably just as good a way to measure recession risk as any other. Sure, some readers will invariably cite far more “robust” models from the extant academic literature and/or purportedly “better” predictors derived by Street economists. Indeed, I’m sure JPMorgan’s resident economists have their own models. But you have to ask yourself whether any of those (always laborious) efforts has actually been able to predict a downturn with any degree of accuracy.
But hell, at the end of the day, anyone will be more accurate than the guy currently in charge of advising the Trump administration on economic policy…
(Bloomberg w/my annotations)
Read more
Egg On Your Face: Fantastic Recession Calls And How To Trade Them
I indeed see the value in simplicity, but I probably wouldn’t put too much weight into this. Modern mechanics and post-QE regime have likely changed the dynamics too much to put much faith in what they’ve presented here. That’s in addition to the stark contrast of HY.
What I found most interesting was the magnitude of decreases going from the 1950s to 2008, clearly showing bubblenomics at work and appearing to suggest the next meltdown will be even more catastrophic. My gut tells me we’ve likely seen the worst on a vol-unadjusted basis, and we’ll continue to see higher volatility and gains into mid-year until earnings collapse, and the next crisis will come from either the currency market (EM and USD) or credit bubble pop in China.
I wonder how these forecasts back-test? And after the predictors reach a trough and reverse, how does that implied change in forecast back-test?
No need to wonder about Kudlow’s competence as a forecaster. He was a hack on TV, its like appointing Judge Judy to the Supreme Court, which I’ve no doubt is next.
We can argue predictive data for recessions all we want but a recession is merely a downturn if the data is ‘cooked by the Government to achieve that particular end… Hate to be cynical, but look at inflation calculation methodology for an example!!!
I like to look at it the other way. What is priced into stocks, credit, etc. so if there is a recession how much will I lose (or possibly make on govt bonds). Or stagflation, or mild growth, stronger growth etc.
My goal is to make good risk reward bets rather than predict an event and then try to maximize that call.