A reader wants more information on the GDPNow model. Let’s discuss with examples how the model works. 
The reader question, shown below, is in response to my post today GDPNow Had a Significant Bounce from its June 3 Low, What Happened?
The Atlanta Fed GDPNow nowcast is significantly below the forecast high of 4.2 percent on May 14, but also higher than the June 3 nowcast of 1.8 percent.
Understanding the Model
Having followed GDPNow for years, I can now usually make a reasonable advance guess as to how the model will react to the data.
Still, it’s not always that easy. The key is not how good or bad the data is, but rather anticipating what the model expects.
Reader Kaya Asks ….
“The key is not how good or bad the data is, but rather anticipating what the model expects. “
I like the phrase, although I don’t understand it completely.
Thank you
My suspect that if one person does not understand, many others also don’t understand but are too afraid to ask.
Understanding the Initial Forecast
The initial forecast for the quarter (nowcast actually), is based entirely on expectations. That nowcast is typically the day after the first GDP release for a new quarter.
It is based on what the model expects of future data. And what it expects is based on historical analysis of prior trends. Some of the key components can be found in a 2017 post Building a Better Model: Introducing Changes to GDPNow.
I cannot explain the initial forecast, and could never predict it. I strongly suspect that Pat Higgins, the GDPNow creator, cannot predict the initial forecast with high accuracy either.
His model makes the nowcast. Adjustments to the model are very infrequent, and I do not believe ever mid-cycle.
I will ask Pat to comment, then post an addendum if he answers.
Understanding Reactions to the Data
Frequently, the GDPNow nowcast increases on bad data and drops on good data.
This happens because the nowcast changes not on the data itself but rather what the model expected vs the incoming data.
Today provides a good example. The trade deficit went from $-69.4 billion last month to $-74.6 billion this month.
Q: Is that a good result?
A: On the surface the answer is no.
Yet, GDPNow rose 0.2 percentage points because the model expected worse.
Q: Why did the model expect worse?
A: Likely because it overreacted to the advance trade numbers.
The advance trade numbers are for goods only. The full trade report out today was goods and services.
You also have a hint of this in the Bloomberg consensus estimate of $-75.2 billion. The deficit was worse than last month but a better than expected $-74.6 billion this month.
The model does not look at other estimates, it makes its own then reacts to the incoming data.
After looking at the full trade numbers, I estimated the impact on GDPNow would be positive. Thus, I made an educated guess that worsening data (the deficit increasing) would result in a rise in the nowcast.
Similarly …
My May 30 Comment
I made an educated guess that the advance numbers were worse than the model expected.
I discussed the trade deficit dip on June 1 in Soaring US Trade Deficit Smacks the Atlanta Fed GDPNow Forecast
It now appears the model overreacted to the Advance Trade Data.
By overreacted, I mean that it appears the model, based on the advance data, set a future expectation that was worse than the full trade report delivered.
To summarize, it’s not the data (good or bad) that matters. What matters is whether an economic report is better or worse than the model expects.
To estimate the GDPNow reaction, one needs to make an educated guess as to what the model expects rather than looking at just the data. It’s a totally different mindset.
The past 24 years can be viewed as an era in which risk declined due to the dynamics of globalization and financialization.
The ascent of China as “workshop of the world” generated a deflationary wave of lower prices for products (due to lower labor costs and lower quality components) that blunted the inflationary impact of the global economies adding $150 trillion in debt since 2000. Global debt, public and private, now tops $315 trillion, 333% of global GDP.
Absent the deflationary impact of globalization, this vast increase in money sloshing around would have sparked inflation. Absent the vast expansion of money via financialization, the expansion of production and consumption enabled by globalization could not have occurred.
At the same time, central banks coordinated policies to steadily reduce interest rates, reaching effectively zero or negative rates (when adjusted for inflation) in 2009 and beyond. This reduction of rates far below historic norms enabled creditors to borrow more even as their debt service costs fell.
Financialization vastly increased leverage and the commodification of credit/debt, enabling emerging-market nations and enterprises and consumers globally to increase their borrowing/spending.
Globalization generated incentives for nations and their central banks to “play nice” and cooperate with other governments and banks to spur profitable (and happily deflationary) trade. These coordinated efforts enabled the global economy to avoid the potentially fatal disruptions of the Global Financial Crisis (GFC) in 2008-09.
https://charleshughsmith.substack.com/p/prepare-for-the-repricing-of-risk
GDP numbers of any kind will never be useful until they and adjusted for inflation and adjusted for Government Deficits. Since the growth of the National Debt runs at 7% range all GDP numbers should be adjusted down by that percent.
Jobs boom!
https://www.cnbc.com/2024/06/07/jobs-report-may-2024-us-job-gains-totaled-272000-in-may.html
Nonfarm payrolls expanded by 272,000 for the month, up from 165,000 in April and well ahead of the Dow Jones consensus estimate for 190,000.
Talking heads on CNBC confused, they were expecting lower numbers but not me, it’s the boomers retiring/expiring that’s the missing link that no one talks about.
If someone retires and someone else takes their place, the size of the workforce hasn’t changed. The company still has the same size payroll. In addition to the 272K hires, the unemployment rate went up .1, to 4%.
The labor force participation rate went down from 62.7 (apr) to 62.5 (may).
https://fred.stlouisfed.org/series/CIVPART/
“If someone retires and someone else…”
That’s your problem right there. When someone with 30 years experience retires, it takes 2 or 3 people to replace that 1 person. Throw in the reduced labor participation rate and you see the problem more clearly.
It’s not the jobs which will likely be adjusted lower in the future, it’s that wages came in hot too which says inflation is a LONG way from being contained.
I don’t think we are getting a rate cut this summer.
Agree, I don’t think we get rate cuts at all and if we do they will be short lived as inflation will roar back up.
It’s turtles all the way down and inflation all the way up.
The explanation is very good, I suspect that I will read these same articles several times, thank you very much.
How does the GDPnow model work? Glad you asked, pull up a seat, get yourself some hard liquor and take a hit on this bong – it will help you understand.
First, gather a group of people (herein called bureaucrats) who think they should be paid six figures for barely working six hours a day, and feel no remorse taking this money from folks who actually work for a living.
Give these @sshole bureaucrats poorly designed survey questions, and have them call unemployed people. People who have real jobs have no time to answer government surveys.
Now take the survey data and xerox it 4-5 times to save paper. Seriously. I told you to take a big bong hit, and I wasn’t kidding.
Next, kill a pigeon. This angers the bird gods. It doesn’t impact the survey or GDPnow.
Call one of the bureaucrats at animal control and tell them to bring you the next deer carcass.
Analyze the dead animals end trails, and using all your powers and all the PhD equations your marxist professors could print… devine a random number. Publish the random number as the official CPI reading, the official GDP number, it doesn’t matter. WMDs are a slam dunk, GDP is booming, and govenment bureaucrats cannot be fired even when they are caught lying on purpose.
Kill another pigeon. Take another bong hit.
And that friends is how the GDPnow number is calculated.
Do you really think that everything is done without any sense?
This is the Biden administration’s stats. Everything should be taken with a grain of salt. If their lips are moving they are lying.
Washington DC is a dying, hopelessly corrupt empire with more debt than it can ever repay. So yes, its all done without any sense.
Sorry (not!!!) that it took you so long to recognize the obvious.
That is so stupid I ought to delete it.
I will let it go for humor assuming that is what it is supposed to be.
The Model has had a very hot hand for many quarters
The marxist political agitators write comments intended to piss off readers, get massive negative feedback — Mish waves that crap right on through.
I poke fun at the absurdity of “analysing” what is at best a random number and just as likely a politically goal-seeked spew of propoganda — Mish threatens to delete the comment!!!
Typical political censorship behavior for facebook, old twitter and disney corp … and now Mish is going down the censorship drain too
I thought this was very funny.