by Wolf Richter, Wolf Street:
The Atlanta Fed’s GDPNow could be all over the place, but nothing compares to how gold imports blew it up this time.
By Wolf Richter for WOLF STREET.
The U.S. Bureau of Economic Analysis (BEA) will release its first estimate of GDP for Q1 on April 30. The Atlanta Fed designed and maintains a model – an algo essentially – that takes in all kinds of economic data for each quarter and provides a running estimate of the BEA’s first estimate of “real” (inflation-adjusted) quarter-to-quarter annualized GDP growth.
TRUTH LIVES on at https://sgtreport.tv/
For Q1, GDPNow began providing estimates in late January, and they looked OK, with growth of over 2% annualized. But on February 28, when imports for January were released and taken into the GDPNow model, GDPNow plunged by 5 percentage points to nearly -3%, and went viral. The depression was here, if you believed the clickbait on the internet.
The problem was the spike in imports of gold that investors buy. Imports are a negative in the GDP calculations by the BEA. But gold that investors buy and sell, import or export, is treated like stocks or cryptos that investors buy and sell: It doesn’t enter into the GDP calculation. So imports of gold that investors buy have always been removed from the BEA’s GDP calculation and don’t reduce GDP growth.
But the designers of the Atlanta Fed’s GDPNow model ignored the issue of gold imports. The amounts were small, and it didn’t matter much, until the amounts of gold imports spiked in January, and since monthly amounts are “annualized” (multiplied by 12) to feed into annualized quarterly GDP growth, that one-month spike became a 12-month spike that blew up its GDPNow model, made worse by an additional large amount of gold imports in February, causing the model to forecast a collapse of the US economy.
On March 6, the Atlanta Fed began calculating an alternate GDPNow (the green dotted line) that treats gold imports and exports correctly, by removing them from the GDPNow model. That “Gold adjusted GDPNow” has ranged from slightly positive to slightly negative. The last estimate, released on April 24, was -0.4%. The final estimate will be released tomorrow, April 29, one day before the BEA will release its first estimate of Q1 GDP.
Ignoring gold imports in its GDPNow model went unnoticed until there was an unusual spike in gold imports. And that happens with models: They’re working fine, until they aren’t. The pandemic blew up all kinds of models, including seasonal adjustments and of course infamously on WOLF STREE the CPI for Health Insurance. And those models then have to be fixed in some way.
Last week, the Atlanta Fed published an explanation of what blew up its GDPNow model and how it fixed that in its “New GDPNow Model.” The current GDPNow model will die on April 29. The “New GDPNow Model” (described in detail in this 10-page PDF) will start on April 30.
