Following the second consecutive ISM spike I had the following Email exchange with Pat Higgins.
Mish ISM Question
How much of today’s jump was due to ISM and how much construction?
If you replace today’s values of GDPNow’s dynamic factor with the values as of February 27, the growth forecasts for consumer spending (2.0%), residential investment (-5.9%), and nonresidential structures investment (-0.8%) were all below their February 27 readings. The counterfactual forecasts are in parentheses. The rise in the factor values due to incorporating today’s ISM Manufacturing release, and other previously released data for February, played a large role in the increase in the nowcast. The data in this morning’s personal income and outlays release was incorporated in today’s update.
If you look at row 4 of the tabs Consumption, Equipment, IntellPropProd, NonresStructures, Residential, ExportsImportsGoods, ExportsImportsServices, FederalGovt and StateLocal you’ll see the factors in row 4. If you change the factor values, you’ll see the forecasts change for each of the subcomponent forecasts. The factors are also in the tab Factor all the way to the left. The other tabs read in the values from this tab, so you can change them here and they will change in the other tabs as well.
I replaced the factors with the ones from February 27 but the bottom line calculation did not change. Higgins explained.
Hi Mike – the spreadsheet is not going to make that calculation. That cell is hard-coded – it’s the published forecast on March 1 as is and it is not going to change as you change the factors. The only thing that will change are the subcomponent forecasts in the tabs Consumption, Equipment, etc. To derive the GDP forecast would take a little work. I was referring you to the tab Consumption. Thanks for forwarding me the link – I’ll take a look at it.
I will explain the bit about the "forwarding the link" in a moment. For now, let's continue with the discussion of the dynamic factor changes and various subcomponents.
PCE Estimates With and Without ISM Factor Changes
As noted above, it is a bit more complex than just changing the factors. Yet one can see that had it not been for a change in the factors, there would not have been another ISM spike.
I asked Higgins "Is there some justification to do use the factors from Feb 27?"
He responded: The justification is that the factor values in the February 27th file do not incorporate the data in the March 1 ISM release, or any of the other released February data like Consumer Sentiment, while the factor values in the March 1 file at https://www.frbatlanta.org/-/media/documents/cqer/researchcq/gdpnow/GDPTrackingModelDataAndForecasts.xlsx do incorporate the February data. So you can get a sense of how the ISM, and other February data, is influencing the estimates of the factor values."
ISM Discussion With Jim Bianco
I had lunch last week with Jim Bianco, head of Bianco research. He talked about survivor bias and how it has impacted ISM surveys.
ISM has not been a reliable indicator of anything for quite some time. All of the diffusion reports from the Fed regions suffer similar flaws.
Bianco mentioned a Fed study that came to the same conclusion about ISM but he does not recall which Fed region did the report or precisely when he saw it.
GDPNow Next Report Tomorrow
GDPNow reports again tomorrow. On March 1 (still current), the report contained this discussion on ISM and the dynamic factor.
The model's estimate of the dynamic factor for February—normalized to have mean 0 and standard deviation 1 and used to forecast the yet-to-be released monthly GDP source data—increased from 0.23 to 1.68 after this morning's Manufacturing ISM Report On Business from the Institute for Supply Management.
I suggest the ISM factors are a bit messed up but I am not certain we that tomorrow.
Mike "Mish" Shedlock