This week for the second week in a row we got extremely low initial jobless claims, but marred by computer glitches, mainly in California. That doesn't mean we can't make a reasonable approximation of what those claims would be had there been no glitches.
Last year I was able to show that, ex-Superstorm Sandy, initial claims were probably still declining, by digging into the internals of the state by state reports. It looks like the same thing is happening this year, i.e., there is a real decline, but it is being masked by the computer problems.
Here's the way the calculation works. One week after the initial report, the DOL gives a state-by-state breakdown of non-seasonally adjusted, and seasonally-adjusted, claims for the prior week. So, this week the DOL gave the breakdown behind last week's 292,000 number (revised to 294,000 this week). What we do is, we take out the claims from the affected state, in this case California, and make the assumption that had there not been a glitch, California's claims would have reflected the same trend as in the other 49 states.
Last year in the affected week 299,700 claims were reported on an NSA basis. The SA number was 381,000. Of the NSA claims, 42,900 were from California and 256,800 were from the other 49 states. This year in the same week the other 49 states reported 210,400 initial claims NSA. So, this year compared with last year, the number of NSA claims dropped by 18.07% for those 49 states (or, put another way, were 0.8193% of last year's number)
If California actually followed the same trend as the other 49 states, then all we need to do is multiply last year's SA number of 381,000 by 0.8193, which gives us a little under 318,000. That's probably pretty close to what last week's number of initial jobless claims would have been had there been no computer glitches.
Since the DOL told us that California and Nevada were still "working through" their computer issues, that probably means that they have caught up on some but not all of their backlog. In other words, this week's number of 309,000 is probably a lot closer to the real number than last week's. This adds to my confidence that initial claims probably have continued their recent declining trend, even if not by as much as the raw reports suggest.