Earlier today, we received the following question from a very thoughtful [and successful] investor:
"So what are your quarterly GDP estimates for this year, and will we see a negative print and/or two quarters of negative growth? I appreciate that it is somewhat irrelevant if we have a technical recession by definition, but you guys are making the recession call, no?"
Given it's obvious relevance and our interest in continuing to broadly own the U.S. economic debate, we thought we'd share our response with a broader audience:
Regarding your question on when, where and how a recession will occur in our U.S. GDP model, it’s important to note that it’s impossible to model in negative GDP prints from where U.S. economic growth is tracking currently – either econometrically or quantitatively.
Our #LateCycle view that has since morphed into our #USRecession and #CreditCycle themes has always been centered on the high and, most importantly, rising probability of a recession commencing in/around mid-2016. It’s not about ascribing a round (or not-so-round) probability estimate to that [broadly undesired] outcome. It’s actually about whether investor, business and/or consumer expectations for an outright recession are rising or falling.
And much like expansions on the way up, a rising probably of recession is reflexive on the way down from the peak in economic growth. Recall that those peaks occurred in 2H14 for manufacturing and capex growth and in 1H15 for consumption and employment growth, respectively.
Modeling in a U.S. recession would be like modeling in +1% revenue growth for AAPL at this time last year – which would’ve obviously been too far out of the band of probable outcomes for even the most ardent bear. The point of our Bayesian inference modeling process is to constantly shock our GDP and CPI models with relevant Bayes factors (i.e. high-frequency economic and financial market data) in order to appropriately adjust our estimate from the reported base rate as implied by the base effects themselves, as well as per trends in high-frequency economic data.
What we’ve learned from Dr. Daniel Kahneman’s work on Prospect Theory is that the market prices in those adjustments (i.e. rates of change) from the base rate, rather than the final outcomes (i.e. absolute states). As such, we don’t start with a desired outcome (i.e. “GDP feels like it’s going to be 3-4%”); we let the data guide our forecasts higher or lower – purely in differential terms.
It sounds overly complicated, but we can assure you it really isn’t; it’s just overly differentiated from competing GDP and CPI models. How we model the economy is not at all unlike how any good bottom-up investor would model a company.
So to answer your question(s) specifically:
- Yes we still believe a recession commences in/around mid-2016 – if not sooner. But as we explain above, that’s actually not the point. It’s all about the rate-of-change in that probable outcome trending higher or lower, and, per the latest economic and financial market data, that probability continues to rise.
- For now (we can’t stress that enough in a Bayesian inference framework) our quarterly GDP estimates for 2016 are: +1.7%, +1.0%, +1.1% and +1.4%, which takes us to +1.3% for full-year 2016 from a previous estimate of +1.5%.
- NOTE: those are YoY deltas. For what it’s worth, our headline GDP estimate (i.e. QoQ SAAR) for 1Q16 is +0.2%. As we discussed at length in our 9/2 Early Look titled, “Do You QoQ?” it’s beyond a fool’s errand to model headline GDP any further out than ~3 months. The chart below highlights just that.
All told, we reiterate our #USRecession theme in the context of the latest data. Our refreshed U.S. GIP model, GDP and economic summary tables are below. Feel free to email with any follow-up questions."
Have a great weekend,