GDPNow Drops to 3.8%: My Email Exchange With Pat Higgins on Volatility
Mike Mish Shedlock
GDPNow Latest Forecast: 3.8 Percent June 29, 2018
The GDPNow model estimate for real GDP growth (seasonally adjusted annual rate) in the second quarter of 2018 is 3.8 percent on June 29, down from 4.5 percent on June 27. The nowcast of second-quarter real personal consumption expenditures growth declined from 3.7 percent to 2.7 percent after this morning's personal income and outlays report from the U.S. Bureau of Economic Analysis.
Nowcast Latest Forecast: 2.9 Percent June 29, 2018
- The New York Fed Staff Nowcast stands at 2.8% for 2018:Q2 and 2.5% for 2018:Q3.
- News from this week's data releases decreased the nowcast for both quarters by 0.1 percentage point.
- Negative surprises from the advance durable goods report and personal consumption expenditures data were only partially offset by a positive surprise from new home sales data.
Nowcast vs GDPnow Reaction to Income and Outlays Report
- Nowcast: -0.04
- GDPNow: -0.70
That is quite a remarkable difference.
Email Exchange With Pat Higgins
On June 25, I emailed Pat Higgins, creator of GDPNow.
I note a Massive 1.8 Percentage Point Spread Between GDPNow and Nowcast What's going on?
Is it possible to break out the impacts individually when multiple reports come out on one day?
The key day in question is May 31.
I knew that Pat could not comment on GDPNow, but I expected and received an answer to my other question.
Following is the reply from Pat Higgins. First, an explanation from me.
On June 1, GDPNow had a misleading line on its spreadsheet. The ISM was on a line dated May 31. However, I consider the error mine. I should have known the ISM report came out June 1. A check today shows Pat made a fix to the spreadsheet.
For ease in reading, I dispense with normal blockquotes. It's easy enough to spot the end of his comments.
Reply From Pat Higgins
I can’t really comment directly on the difference between GDPNow and the FRBNY Nowcast. The only mistake I see in the table is that on May 31, the following reports were released: personal income and outlays and NIPA underlying detail tables. On June 1, the employment situation, ISM Manufacturing, and construction spending reports were all released. Sorry this isn’t completely clear from the table in the GDPNow slide deck you were looking at – it looks like you read ISM Manufacturing and the employment report as being released on May 31. I’ll try and see if there is a way to make the table more clear.
If one came up with an ordering of the data releases, one could come up with a quasi-estimate of the impact of the data releases. For example, on June 1, one could do the following three runs
0.) Data through May 31
1.) Data through May 31 plus data from June 1 employment report.
2.) Data through May 31 plus data from June 1 employment and construction spending reports.
3.) Data through May 31 plus data from June 1 employment, construction spending, and ISM Manufacturing reports.
and take the difference between 1.) and 0.); 2.) and 1.) and 3.) and 2.) to come up contributions of releases. However, if you used the alternative release ordering
0’.) Data through May 31
1’.) Data through May 31 plus data from June 1 ISM Manufacturing report.
2’.) Data through May 31 plus data from June 1 ISM Manufacturing and construction spending reports.
3.’) Data through May 31 plus data from June 1 ISM Manufacturing, construction spending, and employment reports.
it probably won’t be the case that the “employment contribution” in the first list [1.) minus 0.)] is the same as it is on the second list [3’.) minus 2.’)].
Returning to the Question
My question: Is it possible to break out the impacts individually when multiple reports come out on one day?
Based on the above discussion, it does not seem possible because one might get different results depending on the order the individual reports were applied.
So, here we are once again.
GDPNow continues to be far more volatile to the same data releases than Nowcast.
Implications and Observations
Volatility implies that the GDPNow model did not predict the impact of today's Income and Outlays report but the Nowcast model did.
This is an observation, not a slam of either model. I would not know which one to slam, if either. Yet, once again, we have a possibility of a very wide spread converging as the reporting period progresses.
The once 1.8 percentage point spread is now down to 1.0 percentage points.
If we get more bad economic reports that GDPNow does not expect but Nowcast does, the spread will narrow further.
For discussion of today's unexpected Income and Outlays report, please see Real Spending Flat in May, Revisions Lower April Income and Spending.
Meanwhile toss those 4.5% to 5.0% GDP estimates for the second quarter into the ash can.
Mike "Mish" Shedlock