Inventory numbers today suggest a far bigger inventory contribution to GDP than I expected even last week. However, given poor retail sales, the inventory build will most likely result in a third quarter payback.

The BEA will post its first (advance) estimate of second-quarter GDP tomorrow. Let’s take a look at 6 predictions of what might happen.

GDPNow: 2.8% July 27

GDPNow Link

Econoday: 2.6% July 27

Econoday Link

Markit: 2.0% July 24

“The overall rate of expansion remains modest rather than impressive. The surveys are historically consistent with annualized GDP growth of approximately 2%, but the signs are that growth could accelerate further in coming months.”

Snip from IHS Markit Flash US Composite PMI

Nowcast: 2.0% July 21

“The New York Fed Staff Nowcast stands at 2.0% for 2017:Q2 and 2.0% for 2017:Q3.”

Nowcast Link

ZeroHedge: 1.9% July 27

The ZeroHedge forecast is via email from today.

Mish: 1.7% July 27

The above spreadsheet is from GDPNow. I added my calculations in the line with yellow highlights.


  • PCE: Retail sales have not been strong. Auto sales were terrible. The Personal Income and Outlays report shows PCE was up 0.4% in April and another 0.1% in June. People made more money but they did not spend it. July numbers are not out yet. We get them tomorrow. Lovely. Adding to the mix, Real Retail and Food Services Sales are up .31% for the quarter. That number is already in. Charts below.
  • Equipment: I do not track. I accept the GDPNow number.
  • Intellectual Property: I do not track. I accept the GDPNow number.
  • Nonresidential Structures: Investment has not changed much. I am comfortable with zero.
  • Residential Investment: The Residential Construction Report shows housing starts were down significantly but the number of properties under development remained unchanged. Five unit or more structures under construction were down about 1.5%. I expect a subtraction to GDP.
  • Government Spending: Government spending is generally stable.
  • Net Exports: Today’s trade data was to the plus side. I accept the GDPNow number.
  • CIPI: CIPI stands for Change in Private Inventories. I did not expect an inventory surge this quarter. Today’s advance reports (see Inventories Surge Led by Autos, Trade Deficit Narrows: A Boost to Second-Quarter GDP) tells me I am wrong. Ahead of the report, I was anticipating something on the order of 0.3%. I bumped that up to 0.9% today. There is a wild card here. Auto dealers are adding inventory, but what is it really worth? Discounts are at record highs.

Personal Income and Outlays

New Residential Construction

Retail and Food Services

Real retail and food services consumption is up 0.31%. This is a known second-quarter number.

Model Thoughts

My model is way simpler and far more seat-of-the-pants than that of Pat Higgins, creator of GDPNow.

Here is an explanation of how GDPNow figures PCE in response to a question I emailed Pat.

Hi Mish,

Each of the logarithmic growth rates of the monthly subcomponents of real PCE

-Real Core retail goods ex foods services
-Real PCE: New Motor Vehicles
-Real PCE: Net Purchases of Used Motor Vehicles
-Real PCE: Gasoline & Other Energy Gds
-Real PCE: Purchased Meals & Beverages
-Real PCE: Services less food services

is forecasted using the model’s factor for the same month, up to three lagged values of the factor, and the last 6 lagged values of logarithmic growth rates of the same subcomponent being forecasted. If you go to cell FV48, for example, and put the cursor in the formula bar it should highlight the cells that are used to make the forecast of June real PCE Services less food services growth. It will highlight the values of the factor, the lagged factors, and the lagged logarithmic growth rates of real PCE Services less food services as well as all of the regression coefficients [those are in columns GB – GT]. Some of those columns have regressions coefficients of 0.00 for lags that aren’t used in the regression. For other components, different lag lengths can be used.

The Jan – Mar factor in cells FQ4 – FR4 are the model’s estimate of its dynamic factor. It’s highly correlated with the CFNAI, but not identical to it. Those Jan – Mar values are estimates with data through July 19th and are different than what they would have been at the beginning of the forecast cycle for Q2 GDP on May 1. The very left most tab in the spreadsheet – FactorCNFAIData – compares the CFNAI with the GDPNow factor.

Pat’s explanation refers to the “Consumption” spreadsheet tab in the GDPNow Excel Download.

Final Thoughts

It is possible I am correct (or GDPNow is correct) for the wrong reasons.

I do not foresee the huge jump in PCE as does Pat Higgins. However, an offsetting error in CIPI could make either of us right or wrong in our overall number.

Sometimes errors offset and sometimes they stack up. We find out on Friday.

Thanks again to Pat Higgins for his time and effort in explaining his model.

Mike “Mish” Shedlock