by Calculated Risk on 9/02/2020 01:25:00 PM

UPDATE: The NY Times has an article on this too: The Labor Department will start counting unemployment claims in a new way.

The DOL has announced that they are changing the seasonal adjustment for unemployment claims:

Beginning with the Unemployment Insurance (UI) Weekly Claims News Release issued Thursday, September 3, 2020, the methodology used to seasonally adjust the national initial claims and continued claims will reflect additive factors as opposed to multiplicative factors.

This makes sense since the huge increase in weekly claims has been due to the pandemic.

However, the DOL will not revise prior weeks, from Ben Casselman:

@USDOL says initial claims for 8/22 will NOT be revised to reflect the new adjustment methodology. So we should expect to see a big, artificial drop in claims from 8/22 to 8/29.

Currently this is the a low season for initial claims, so the seasonal factor is below 100 (see data and seasonal factors here).

For example, for the week ending August 22nd, initial weekly claims were 1,006,000 Seasonally Adjusted (SA), and 821,591 NSA. The seasonal factor was 81.7.

The formula is (NSA Claims) * 100 / Seasonal Factor = Seasonally Adjusted Claims. So 821,591 * 100 / 81.7 = 1,006,000 (rounded to nearest 1,000).

Switching to additive factors, my guess is the DOL would have added about 40,000 to the NSA number to report the SA number.  So instead of reporting 1,006,000 for the week ending August 22nd, the DOL would have reported 862,000.

As Ben Casselman notes, the DOL will not revise prior weeks using the additive factor.   Since the seasonal factor for the week ending Aug 29th (to be reported tomorrow) is also 81.7, this means the number of SA claims would decline significantly, even with the same level of NSA claims.

What this means is we cannot compare the SA number to previous weeks.   However, since the seasonal factor was the same for both the week ending August 22nd, and the week ending Aug 29th, we can directly compare the NSA number for both weeks to see if claims were increasing or decreasing.