Seasonal Distortion In Jobless Claims: The Sequel

Last week’s sharp rise in new jobless claims looks ominous, but there’s a good reason to reserve judgment at this point. Recall that the previous weekly update in new filings for unemployment benefits tumbled dramatically to a new four-year low. But as several analysts warned, the fall was probably due to a seasonal factor linked to summer shutdowns of auto plants. Fast forward a week and it appears that the seasonal distortion has been effectively corrected. In sum: the latest reading on jobless claims (for the week through July 14) is more or less unchanged from the late-June figures, which implies that the roller coaster ride in the interim was a lot of noise about nothing.


As the first chart below shows, weekly claims have been unusually volatile lately—even by the standards of what is inherently a volatile data series. Taken at face value, new claims on a seasonally adjusted basis have generally remained flat through last week’s tally of 386,000. In other words, there’s not much progress to report on this leading indicator for the labor market. That’s bad.

On the other hand, if we look for deeper clarity in the raw figures, and compare the numbers on a year-over-year basis, the trend looks brighter. As the second chart reminds, the annual pace of unadjusted claims continues to decline relative to its year-earlier level. That’s good.

But, wait—even the annual pace of claims has suddenly moved closer to zero. Is that a sign of trouble? Perhaps. If and when the annual pace of claims is flat (it hasn’t been for more than a year) or, even worse, begins rising on a sustained basis, we’ll have a strong signal that we’re in trouble (much deeper trouble) with regards to the labor market and, by extension, the economy overall.
For now, however, it’s premature to say what all the back and forth means. As you’ve read on these pages many times, it’s usually a mistake to read too much into any one data point when it comes to claims, as last week’s head fake reminds.
Yes, there’s darkness on the edge of town via the latest batch of monthly numbers for June (retail sales, the ISM Manufacturing Index, and payrolls). But optimists can also point to yesterday’s encouraging numbers on housing starts and the resilience in the June update for industrial production. Feeling a bit whipsawed? Welcome to the club. But if we’re looking for clarity on estimating recession risk, based on the full set of numbers available so far through May, it’s not yet obvious that the cycle is destined to crumble. My recession risk index for the preliminary June profile is holding up quite well too (I’ll have an update soon).
But let’s be clear: there’s plenty of weakness all around to inspire worries about what comes next. We may be at the tipping point… or is the trouble merely a batch of statistical noise that’s temporarily infecting the numbers? In any case, all will be sorted out soon. Meanwhile, it’s premature to rush to judgment that the cycle is doomed. In short, don’t let emotion, driven by the latest numbers, overwhelm strategic perspective.
“The snapback in the pace of claims should not be particularly surprising as last week’s favorable seasonal unwinds,” says Millan Mulraine, senior macro strategist at TD Securities, via Reuters. “However, we believe that the current level of initial jobless filings overestimates the true pace of jobless claims, and should see claims fall back to around 370,000 in the coming weeks.”
If he’s wrong, and there’s more pain to come, the evidence will become clearer in the near term. For now, assuming the worst still relies heavily on speculating rather than interpreting the data in hand. Itchy trigger fingers play well on the big screen, but the habit of shooting first and asking questions later isn’t helpful for dissecting the cross currents that typically swirl in the seas of macro.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>