With just one month left to go in Q4 2019, we can finally begin to look at the US GDP estimates created by economic forecasting algorithms at the Atlanta, New York and Cleveland Federal Reserve branches. Over the years of writing about these novel models we’ve noticed that they are typically very noisy until they see 2 months of data. On the downside, that means they can give false readings in months 1 and 2. On the plus side, we like the fact that the models are 100% automated around economic data inputs or capital market signals and leave nothing to human judgment.
First, let’s look at the Atlanta Fed’s GDPNow model, which currently calls for 1.3% growth in Q4 2019:
A few points on this:
- Estimates here have been volatile all of Q4, ranging from 0.3% to 1.7% in just the last month.
- No prizes for what sorts of data push the number around. The weak Industrial Production number released on November 15th helped create that low estimate. Consumer-centric data releases like Personal Spending (out on November 27th) lifted it higher.
- Bottom line: US economic growth is currently a tug of war between a still strong US consumer and sluggish manufacturing/business confidence.
Next up is the New York Fed Staff Nowcast, which is more pessimistic and calling for just 0.8% GDP growth in the current quarter:
Here is why the NY Fed model differs from the Atlanta Fed’s:
- At the start of Q4 both models were in the same ballpark: 1.3% at the NY Fed and 1.5% for Atlanta.
- Weak Industrial Production, Capacity Utilization, and ISM Manufacturing data have taken a large toll since then, shaving 0.6 points off the NY Fed estimate. Offsetting positives from the housing market and consumer economy have not been enough to balance those disappointing numbers.
- Bottom line: the difference between the 2 models comes down to how much of an effect the ongoing manufacturing slowdown will have on Q4 GDP. Atlanta says “not much, but noticeable”. New York says “a lot”.
Finally, here is the Cleveland Fed’s model that, unlike the multi-factor economic data models at New York and Atlanta, just uses the difference between 3-month and 10-year Treasury yields:
Yes, that red line is this model’s forecast, and it calls for 2% GDP growth over the next 4 quarters. A few thoughts on that surprising number:
- Looking back to 2002, the model does show that yield curve spreads broadly predict GDP growth. And the relationship is especially tight from 2010 – 2016.
- Since 2017, however, the two datasets have decoupled. Credit the 2017 change in tax rates that boosted US growth in 2018 along with negative long rates abroad and still-low yields at home for that dichotomy.
- Bottom line: the Cleveland Fed’s model seems to assume this trend will continue into 2020. That’s odd, considering that the NY Fed’s Recession Probabilities model uses the same data and currently gives the odds of a downturn in the next 12 months of 29%.
Summing up: average the NY and Atlanta Fed models (better assessments in our book because they move as fresh data comes in) and you get 1.0% Q4 GDP growth, below the Blue Chip Economists’ estimate of 1.5%. Should that come to pass, it would be the slowest reading since Q3 2011’s 0.9% reading. No wonder that markets are laser-focused on the US and China resolving at least some of their trade dispute. Or that Fed Fund Futures stubbornly discount 50-50 odds that the Fed will have to cut rates in the first half of next year…
Atlanta Fed GDPNow: https://www.frbatlanta.org/cqer/research/gdpnow
New York Fed Staff Nowcast: https://www.newyorkfed.org/research/policy/nowcast