But how good exactly? Reuters:
Jobless claims drop steeply, skewed by autos | Reuters: WASHINGTON (Reuters) - The number of U.S. workers filing new claims for jobless benefits fell sharply last week but the data was distorted by an unusual pattern of layoffs in the automotive industry, which amplified the decline. The Labor Department said on Thursday that initial claims for state unemployment insurance fell 52,000, the largest drop since December, to a much lower-than-expected seasonally adjusted 565,000 in the week ended July 4, from 617,000 the prior week. It was the lowest reading since January. Analysts polled by Reuters had forecast claims to drop to 605,000 from a previously reported 614,000. However, in a sign of ongoing employment weakness, so-called continued claims of people still on jobless aid after an initial week of benefits rose by 159,000 to a record 6.883 million in the week ending June 27, the latest for which data is available.
A Labor Department official said that there had been far fewer automotive and other manufacturing layoffs last week than anticipated on the basis of past experience of claims over July, when many plants are commonly idled. The "seasonal factors" the department uses to adjust the data to provide a better sense of the underlying trend had expected a large increase in claims in the latest week. Actual claims in fact rose by a much smaller amount, which when seasonally adjusted, generated a large fall. A number of states said that auto sector layoffs apparently had already happened, reflecting closures in the battered U.S. automotive industry, while other states said they did not get the layoffs they had anticipated. "I would expect the underlying trend (in claims) is probably diminishing but it's hard to tell from this number how much is noise," said Keith Hembre, chief economist at First American Funds in Minneapolis.
The 4-week moving average for new claims declined by 10,000 to 606,000, the lowest reading since February. This measure is closely watched because it irons out weekly volatility, and it has now declined in four out of the last five weeks.
Looking at graphs like this:
reminds me of how very much in the high-frequency data we typically examine rests on the seasonal adjustment process--and how important it is to get that seasonal adjustment right. Of course, to the extent that the true seasonal adjustment factors are not absolutely invariant to the phase of the business cycle it very quickly becomes impossible to estimate them accurately.
What the standard seasonal adjustment factors for initial unemployment claims tell us is: "don't worry if new claims spike in January or July. But does that mean we should be greatly encouraged if they don't spike in January or July? In January 2008, July 2008, and January 2009 the seasonally-adjusted claims number improved because unadjusted claims did not spike enough. But anyone who took those declines as evidence of an improving economy--well, I have a bridge they might be interested in purchasing...