Tuesday, December 11, 2012

Sandy's impact on initial jobless claims

- by New Deal democrat

First things first: last Thursday I wrote a post that turned out to include an erroneous claim about the latest initial jobless claims report; namely, that ex-Sandy, last week's number would have been 339,000. Unfortunately, I had compared a week-ending 11/27 apple with a week-ending 12/04 orange. I won't know until this Thursday what last week's number would have been ex-Sandy.

Fortunately, the New York Times didn't run this information on their front page. My colleague Jeff Miller, however, did mention it in his Sunday column. So I apologize to Jeff and to you. Everybody makes mistakes and while I can't promise not to make mistakes again, I can promise that if I make a mistake and find out about it, I'll let you know.

That being said, what I decided to do after Thursday's column was to go back and look at all of the weeks after Sandy, and run two control groups - one consisting of the four weeks before Sandy, and the other of weekly YoY comparisons. If, for example, typically the other 48 states outside of NJ and NY include 90% of all jobless claims, but in week X after Snady, they only included 80% of all claims, converting the week X figure to a 90% weighting of nonseasonally adjusted claims ought to give us a good estimate ex-Sandy. Tnen all we need to do is use the seasonal adjustment and we have a very good seasonally adjusted ex-Sandy number as well.

It turns out that, regardless of whether I use a month over month or YoY comparison for the control, the numbers are quite close. In the first chart below, I give the % of all claims that the other 48 states outside of NJ and NY constituted for each post-Sandy week, plus the four weeks before Sandy, and also YoY from the equivalent weeks last November.

week ending48 state % of total4 weeks prior % of totalYr ago % of total
11/0390.5 89.5 91.1
11/1077.1 91.890.8
11/1780.6 90.7 91.2
11/2486.7 90.991.3

In the 4 weeks just before Sandy, the 48 states besides NJ and NY were responsible for 90.7% on average of all claims. One year before, they were responisble on average for 91.1% of all claims.

In the second table below, I've taken the raw jobless claims total for the 48 states ex- NJ and NY, and figured what the total 50 state nonseasonally adjusted number would be if the % of total claims were 90.7% or 91.1% instead of the actual post-Sandy percent. Then I've mutlipled by the seasonal adjustment for each week. As you can see, the results are very close. Finally, I averaged the two results for a final estimate that is probably within 1000 or so of what the actual number each week in November of this year would have been had Sandy not occurred.

week ending48 state nonseasonal #weekly seasonal adjustmentseasonally adjusted at 91.1%seasonally adjusted at 90.7%average after seasonal adjustment

The final column to the right, in bold, is the final estimate for what initial jobless claims would have been in November without Sandy.

Finally, here's what the 4 week moving average would have been, based on the above best estimates. I've included the 4 weeks average for the immediate 4 preceding weeks for comparison:

10/06 364,750
10/13 366,500
10/20 368,750
10/27 367,250
11/03 371,750
11/10 369,500
11/17 368,750
11/24 372,250

Had Sandy not occurred, it appears the 4 week average of initial claims would have risen slightly during November compared with October.

I'll update this with new information so long as it appears Sandy is having and effect on the claims number.