GDPNow data from the Atlanta Fed, chart by Mish

The lead chart is from my October 14 post-GDPNow Forecast Unfazed by Weak Retail Sales and Strong Inflation

Weak retail sales did not cause a dip in the forecast. I asked Pat Higgins if he could shed any more light on that early October surge. Here's his reply:

Hi Mish,

I don’t think I can speak too much regarding how the model reacted after any particular data release, but in rows 33 and 34 of the tab TrackingHistory, you’ll see “Final Sales to Domestic Purchasers” and “Final Sales to Private Domestic Purchasers”. The former is final sales excluding imports and exports and the latter is final sales excluding imports, exports, and government consumption/investment expenditures. You can see the increase in the forecasted growth rate of real “Final Sales to Domestic Purchasers” and “Final Sales to Private Domestic Purchasers” in late September to early October is smaller than the increase in real final sales so, as a matter of arithmetic, a fair amount of the increase in the model nowcast of final sales in late September and early October was concentrated in net exports. In terms of the mechanics of how GDPNow forecasts imports and exports growth, the primary data source for services exports/imports is the international trade report released jointly early in the month by the Census Bureau and the BEA. For imports and exports of goods, the model also uses data from the Census Bureau’s “Advance Economic Indicators” released late in the month: https://www.census.gov/econ/indicators/index.html . That release gives an early snapshot of goods trade before the full report is released. The model treats the imports/exports data on foreign goods trade from the Advance Economic Indicators release as hard data [i.e. the same way it would tread the monthly data from the full trade report] and incorporates the revised goods data when the full international report is release roughly 7-10 days later. The model does break down imports and exports into goods and services separately so that the change in their forecasts can be traced after each release.

Best regards,

Pat 

Essentially, weak retail sales did not matter because the model expected them or perhaps they no longer mattered. 

Q: Why is that?

A: This goes back to September 30 when I noted GDPNow for the 3rd Quarter Surges Following Strong Trade, Inventory, and Income Reports

I even posted the answer 

Spotlight on Current Real Final Sales (RFS) Estimate - September 30

• Base GDP Estimate: 2.4 Percent 
• RFS Total: 2.4 Percent 
• RFS Domestic: +0.2 Percent (Report Details)
• RFS Private Domestic: -0.2 Percent (Report Details)

A relative surge in exports vs imports drove the September 30 surge and it stuck despite a very poor construction spending report and poor retail sales.

Honing in on the latter, if retail sales are weak because consumers are not buying imports, then the model likely expected poor retail sales. 

Spotlight on Current Real Final Sales (RFS) Estimate - October 14

• Base GDP Estimate: 2.8 Percent (Lead Chart)
• RFS Total: 2.9 Percent (Lead Chart)
• RFS Domestic: +0.6 Percent (Report Details)
• RFS Private Domestic: +0.2 Percent (Report Details)

Real Final Sales to private domestic consumers is a barely positive 0.2 percent. 

Government spending tacks on another 0.4 percentage points.

All the rest is net exports. 

That does not mean the model is correct. But the explanation is reasonable. 

Blue Chip Forecast

Evolution chart from GDPNow, arrow annotations by Mish

The Blue Chip forecasters have not honed in on that export surge. But we still will have skirted recession even if the much lower Blue Chip assessment is correct.

More By This Author: