The Atlanta Fed GDPNow Model was on updated June 27. The previous release was June 16. The intermediate points and movements in the chart above were just updated today.
Key Points
- Real Final Sales (RFS) is the true bottom line forecast of the economy.
- The difference between GDP and RFS is inventory adjustment that nets to zero over time.
- RFS is 1.7 percent, unchanged since June 26.
- Base GDP is up to 0.3 percent from June 26, a meaningless rise in size but more importantly because RFS is the number to watch.
Housing
- Since June 16 the most important data releases were existing and new home sales.
- June 24: New Home Sales Jump 10.7 Percent in Big Upward Surprise
- June 21: Existing Home Sales Skid Another 3.4 Percent in May, Down Fourth Month
The net impact of housing was essentially nothing.
Important Notes About the Model
This might sound odd, but it’s very important to note that the data itself does not necessarily matter.
What does matter is what the data does vs what the model predicted.
Thus, seemingly good surprises to the upside might cause the forecast to decline and seemingly bad surprises might cause the forecast to rise.
All that matters is how good or bad the data is relative to the forecast. Every data point changes the model’s expectation of what’s coming.
Looking Ahead
The next GDPNow forecast is Thursday, June 30. Looking ahead the data points are
- International Trade in Goods: June 28
- Q1 GDP Final: June 29
- Jobless Claims: June 30
- Personal Incomes and Outlays: June 30
- ISM Manufacturing: July 1
- Construction Spending: July 1
Unless there is a shocking revision to Q1 GDP the influence on Q2 will be meaningless. Jobless claims are unlikely to matter much unless that data is shocking. Construction spending is enormously lagging and outside the date range besides.
In terms of swings to the GDPNow forecast, the most meaningful reports are the three in italics above. But ISM is outside the date range of the next forecast.
International Trade in Goods for May 2022
Imports down and exports up seems like a good result, but on June 7 International Trade knocked down the GDPNow forecast because the model predicted more.
I suspect the model will have this correct unless there are some big revisions.
Personal Income and Outlays Estimate for May 2022
There is plenty of scope for big surprises in personal incomes and outlays, both to the economists and to the GDPNow model.
May retail sales numbers and April revisions took the GDPNow estimate down from
On June 15, I noted Retail Sales Flounder in May With Negative Revisions in April.
- The March 2022 to April 2022 percent change was revised from up 0.9 percent to up 0.7 percent.
- In nominal terms retail sales fell 0.3 percent.
- In real terms sales fell from 233,724 to 230,852 from April to May.
- That’s a month-over-month decline of 1.2 percent, using the CPI as a deflator. It’s real, not nominal spending that’s an input to GDP.
What’s the Model Thinking?
I asked Pat Higgins, the Atlanta Fed creator of GDPNow that question on June 3 regarding ISM.
I asked because the model took a big hit on June 1 when the forecast of RFS declined from 3.4 percent to 2.9 percent on what seemed to be an ISM surprise to the upside.
Hi Mike,
There is not really an easy way to translate what the model was expecting for the ISM Manufacturing data with the actual release [in addition to the composite index, the model also includes the employment, inventories, new orders, production, and supplier deliveries subindexes]. But I have attached the estimates of the factor on May 27th (pre ISM) and on June 1st (post ISM). On May 27th, the May and June 2022 values are forecasted based on the factor estimates through April. On June 1st, the model uses the ISM Manufacturing data [and all of the lagged data] to estimate the factor for May and then uses the factor estimates through May to forecast the June factor value. Here’s an email on Real Final Sales
Best regards,
Pat
Guessing Game
- So not only are we guessing the data, we are guessing what the model is predicting in relation to the data.
- Based on May retail sales, I suspect the model envisions a downward revision to April.
- There easily could be another adjustment lower to April and a miss for May.
- There could easily be something else.
My guess is the data will be weaker than what economists project. But that might not matter if it’s not what the model thinks.
To go out on a limb, I think May will be worse than the model projects and June much worse than the model predicts at this time.
If incoming data changes to the downside, then the model will reflect that data in later updates.
US Manufacturing Output Contracts For First time Since 2020, Service Growth Dramatically Slows
The next GDPNow release on Friday will not incorporate ISM. And the model does not look at manufacturing PMI at all.
The PMI took a big smack on June 23 as noted by my post US Manufacturing Output Contracts For First time Since 2020, Service Growth Dramatically Slows
Since that is not reflected in the model, there’s a good chance of a sizable model miss on ISM. However, ISM tends to be way higher than the PMI.
Place your bets.
I suspect the model estimates too much from consumer spending and way too much out of the ISM.
If so, the GDPNow Real Final Sales forecast will plunge.
Yet, that still is not conclusive. The GDPNow model itself may be inaccurate. The final forecast of GDPNow has been quiet good for many quarters, but that does not imply the next one will be as good.
The model can miss to the upside or downside.
Models Cannot Think
Models cannot think. We can. The data looks bad and feels bad. More importantly, the direction is bad.
Strong April retail sales took a dive in May, housing is pretty much going to hell, the Fed is tightening at a record pace on a percentage basis, QT just started, and the wealth impact from the stock market and crypto collapse is huge.
Add it all up and you get a recession that started in May.
I’ve Seen Enough, the US is in Recession Now, Q&A on Why
For more discussion of the above data points, with many charts, please see I’ve Seen Enough, the US is in Recession Now, Q&A on Why
This post originated at MishTalk.Com.
Thanks for Tuning In!
Please Subscribe to MishTalk Email Alerts.
Subscribers get an email alert of each post as they happen. Read the ones you like and you can unsubscribe at any time.
If you have subscribed and do not get email alerts, please check your spam folder.
Mish
inflation, he would re-institute reserve requirements against his new, inclusive, definition of transaction accounts (Regulation D), or what Nobel Laureate Dr.
Milton Friedman advocated, December 16, 1959. Powell doesn’t give a damn about
inflation. Powell doesn’t give a damn about the economy.
increasing? Or do I go because R-gDp is falling? is due to an increasing volume and proportion
of savings being impounded and ensconced in the payment’s system as rates rise.
Banks aren’t intermediaries. Banks don’t lend deposits. Deposits are the result
of lending. Ergo, all bank-held savings are frozen. That is the cause of
secular stagnation, a deceleration in the transaction’s velocity of funds.
Hi Mike,
There is not really an easy way to translate what the model
was expecting for the ISM Manufacturing data with the actual release [in
addition to the composite index, the model also includes the employment,
inventories, new orders, production, and supplier deliveries subindexes].
But I have attached the estimates of the factor on May 27th (pre
ISM) and on June 1st (post ISM). On May 27th, the
May and June 2022 values are forecasted based on the factor estimates through
April. On June 1st, the model uses the ISM Manufacturing data
[and all of the lagged data] to estimate the factor for May and then uses the
factor estimates through May to forecast the June factor value. The factor
is standardized to have a mean of 0 (since 1967) and standard deviation of
1. I took these datas from the files we posted at https://www.atlantafed.org/-/media/documents/cqer/researchcq/gdpnow/GDPTrackingModelDataAndForecasts.xlsx . You can get the values in the tab Factor of this spreadsheet (3rd most from the left).
There was also a construction spending release on June 1st that directly impacted the residential, nonresidential structures and
government spending forecasts. The construction spending data does not
impact consumer spending directly, so the change in that forecast gives a
partial, probably understated, estimate of how the change in the factor
estimate impacted the forecast.
The light blue filled cells in row 8, columns FU and FV, of
the tab Consumption (4 tabs to the right of Table) in the
spreadsheet (see above link) gives the forecast for the growth rates of real
“retail control” goods in May and June. These sales excludes sales from
motor vehicle and parts dealers, building supply stores and gasoline
stations. The numbers are annualized logarithmic growth rates, so that a
non-annualized 1% increase would show up as 1200*ln(1.01) = 11.94 in the
spreadsheet.
The same columns in row 41 gives the forecasts for the real
growth rate of sales from food services and drinking places. Revisions to
the past 2 months of retail sales data also factor into the calculation (i.e.
the model tries to anticipate the revision to relevant consumption categories
based on the retail sales revision).
Best regards,
Pat
https://www.bea.gov/news/blog/2022-04-28/gross-domestic-product-first-quarter-2022