Finally! Major drop in jobless claims, well below 600K.
Weekly Claims fell from a revised 617K to 565K last week. My preferred measure, the four week moving average, also fell from 613K to 598K. This is the kind of drop I have been looking for, for the past few weeks.
We have to be a little cautious in that this was a Holiday. Benefit offices were likely to be closed. Some potential filers were likely to be out of town. The statistics compensate for that somewhat but can never be perfect.
Regardless, here is a look at the near term chart.
Pretty clear downward trajectory at this point. What we need to make us feel much better going forward is for next weeks number not to go above 600K. We should expect some rise. There is significant noise in this series and it is unlikely that all of the good news is real news.
However, a number below 600K will keep the four week moving in the right direction.
5 comments
Comments feed for this article
Thursday ~ July 9th, 2009 at 12:50 pm
dWj
Other people seem to have four-week sliding window average as their preferred measure as well. The weekly change in the sliding-window average is one quarter of the new number minus that from four weeks ago, so that it is 1) fairly noisy, and 2) on some level predictable; if there happened to be a high outlier 3 weeks ago, you have a guess that the average will go down next week with no economic content.
A three-week exponential moving average makes more sense to me. Following Yahoo’s “consensus forecast” of weekly claims for a few years, I found that my three week exponential moving average actually did a slightly better job of predicting the numbers than consensus except for the immediate post-Katrina period. Changes in the number therefore represent primarily new information.
Anyway, my average dropped from 620000 to 602000. As you note, the seasonal adjustment isn’t perfect in terms of holiday effects, but a number like 565000 is hard to be disappointed with right now.
Thursday ~ July 9th, 2009 at 3:50 pm
Karl Smith
Interesting, I’ll have to look into that. Did you play around with longer lags?
Sunday ~ July 12th, 2009 at 10:48 pm
Twitted by _victor_
[...] This post was Twitted by _victor_ [...]
Wednesday ~ July 15th, 2009 at 2:44 am
dWj
I played a little bit with a variety of lags, and numbers near 3 were where I found my optimum when I did it. I have a bigger data set and better software now — and I’ve taken an honest-to-God econometrics class — and it might be worth redoing some of this.
Thursday ~ June 23rd, 2011 at 8:21 am
Liza
Hey, that’s pwoerufl. Thanks for the news.