Thursday, November 29, 2012

A few comments on GDP Revision and Unemployment Claims

by Bill McBride on 11/29/2012 02:21:00 PM

• GDP Revision: Although Q3 real GDP growth was revised up from 2.0% annualized to 2.7%, the underlying details were disappointing. There were three main sources for the revision: 1) Personal consumption expenditures (PCE) increased at a 1.4% annualized rate,  revised down from 2.0%.  This means PCE contributed 0.99 percentage points to real growth in Q3 (revised down from a 1.42 percentage point contribution in the advance release), and 2) the change in private inventories added 0.77 percentage point contribution to growth (revised up from -0.12), and 3) exports were revised up to a 0.16 percentage point contribution (revised up from -0.23).

This suggests weaker final demand in the US than originally estimated.

Also Justin Wolfers at Bloomberg discusses the weak Gross Domestic Income (GDI) data: The Bad News in Today's Happy Growth Report.  Sluggish growth continues.

• Unemployment Claims: A reader sent me some "analysis" on the initial weekly unemployment claims report released this morning that was incorrect. The writer wrote that the 1) the 4-week moving average was at the highest level this year, 2) that there were 30,603 fewer layoffs in New York "last week", so 3) the recent increase in the 4-week average can't be blamed on Hurricane Sandy.

The first point is correct. The 4-week average is at the highest level since October 2011, but the conclusion about not blaming Sandy is incorrect.

First, the initial claims data is very noisy, so most analysts use the 4-week average to smooth out the noise. When an event happens - like Hurricanes Katrina in 2005 or Sandy this year - the 4-week average lags the event. Here is the unemployment claims data for the last 10 weeks:

Week EndingInitial Claims (SA)4-Week Average
9/22/2012363,000375,000
9/29/2012369,000375,500
10/6/2012342,000364,750
10/13/2012392,000366,500
10/20/2012372,000368,750
10/27/2012363,000367,250
11/3/2012361,000372,000
11/10/2012451,000386,750
11/17/2012416,000397,750
11/24/2012393,000405,250

It is no surprise that the 4-week average increased this week. The 363,000 claims for the week ending Oct 27th were dropped out of the average and replaced with the 393,000 initial claims this week - so the 4-week average increased even though initial unemployment claims are declining.

The 4-week average will probably increase again next week as the 361,000 claims for the week ending Nov 3rd will be replaced with the claims for this week. Note: There are some large seasonal adjustment this time of year - especially the week after Thanksgiving - so it is hard to predict the level of claims.  But the math is simple.

The good news is in two weeks the 451,000 claims for the week of Nov 10th will be dropped out of the 4-week average.

Key point: the 4-week average is intended to smooth out noise, but it lags events.

The writer's conclusion about 30,603 fewer layoffs in New York "last week" so the increase in the 4-week average can't be blamed on hurricane Sandy are incorrect. As part of the weekly release, the DOL notes the UNADJUSTED state data for the PREVIOUS week. The headline number was for the week ending Nov 24th, but the unadjusted state data was for Nov 17th.

The state data for New York showed a large decline, but the week before the New York data showed an even large increase. Since this data is unadjusted, we can't tell if claims are still elevated in New York, but since the increase for the week ending Nov 10th was much larger than the decrease for the week ending Nov 17th, my guess would be that claims are still above normal.

The bottom line is the recent increase in unemployment claims is most likely due to Hurricane Sandy, and there is nothing in the data that would suggest otherwise. And using simple arithmetic, we'd expect the 4-week average to lag the event. The state data supports this view, and I expect the 4-week average to increase again next week, and then start declining the following week (although there can be large seasonal effects this time of year, so we could be off a week or two).

Last 10 Posts