GS 360-US Daily: Making Sense of the Drop in Jobless Claims (Tilton)

·        The dramatic improvement in claims could be seen as a measure of rapid labor market improvement. However, a much more important explanation is the technical impact ofseasonal adjustment combined with the unusual timing of auto sectorlayoffs this year.  As these unwind over the nextfew weeks, new claims are likely to move back towards 600,000 and theimprovement in (seasonally adjusted) continuing claims will slowdramatically, and quite possibly reverse.

·        Thebottom line: though we do believe the worst of the labor marketdeterioration has passed, this morning’s numbers substantiallyoverstate the rate of the improvement, and some backsliding is likelyover the next few weeks.  Just as we aredownplaying the improvement we have seen, we will also downplay thesignificance of modest increases over the next month.

Joblessclaims dropped substantially in this morning’s weekly report. Newclaims fell to 522,000 on a seasonally adjusted basis, the lowest levelsince the first week of the year.  While we andsome others had expected a drop in new claims (we explained the basicrationale in our daily “US Events” data preview this morning), the bigsurprise was in continuing claims: the total number of benefitrecipients plummeted to 6.27 million on a seasonally adjusted basis, adrop of more than 600,000 on the week – more than half again as largeas the biggest previous drop on record.

The dramatic improvement in claims could be seen as a measure of rapid labor market improvement.  However, several factors have conspired to distort the data considerably in recent weeks: 1) auto sector plant shutdowns, which began earlier and have lastedlonger this year than usual, 2) the seasonal adjustment process, which“expects” a higher level of jobless claims in the early weeks of Julyduring the normal summer plant retooling period, 3) for continuingclaims only, the expiration of benefits for many recipients, and theconsequent “dropping out” of those people from the continuing claimsdata set.  Below we attempt to explain how thesefactors have affected the new and continuing claims data over the lasttwo months and how they are likely to continue to do so for theremainder of the summer. 

First, a quick review of the seasonal adjustment process.  Each week, the Labor Department tallies the total number of new claims for jobless benefits from state unemployment offices. Then, it divides this result by a seasonal factor that reflects the“normal” ebb and flow of jobless claims throughout the year. For example, workers are more likely to be laid off around the end ofthe year, so seasonal factors are highest in December and especiallyearly January.  The seasonal factors “expect” theraw (unadjusted) number of jobless claims to be higher. At the peak inthe second week of January, raw claims are typically about 80% higherthan the average week, and about 20% higher than the first week ofJanuary.  So the seasonally adjusted level ofjobless claims will only rise from the first week of January to thesecond week if raw claims rise by more than the 20% increase in theseasonal factor.

New claims:  Fornew jobless claims, the weekly seasonal adjustment factor increased by41% between the final week of June and the second week of July. In other words, the past behavior of new claims in this season of theyear would suggest that non-seasonally adjusted claims would normallyincrease by about 41% over those two weeks.  (Theactual increase last year was 31.3%; unadjusted initial claims rosefrom 368,500 in the week of Jun 27, 2008 to 484,000 in the week of July11, 2008.)  The main explanation for thisseasonal pattern is the summertime shutdown and retooling of autoplants, which cause temporary layoffs of many thousands of auto sectorworkers.

However, this year was different in this respect, as it has been in so many others. Chrysler and GM began shutting down plants in early and mid-Mayrespectively and idling them for longer periods of time, in order toreduce bloated inventories.  This has meant thatmany of the layoffs that would have normally occurred in July havealready occurred over the past two months. Therefore, there were fewerplants remaining to go idle in early July, and the raw level of initialclaims increased “only” 19% over the past two weeks (to 667,500 in theweek of July 10 from 559,900 in the week of June 26). In our daily “US Events” preview this morning, we noted this phenomenonas a reason why new claims were likely to decline in today’s release. After seasonal adjustment, the level of seasonally adjusted claimsreported this morning dropped by more than 15% from two weeks ago (to522,000 new claims on a seasonally adjusted basis for the week of July10 from 617,000 in the week of June 26).

The story doesn’t end here. The seasonal factor reverses course over the next couple of weeks andends up slightly lower than where it began (i.e. it “expects” evenlower raw initial claims by early August than it did in late June).  This corresponds to the normal period of re-hiring of furloughed auto workers. Although some of this re-hiring should in fact occur on schedule, a fewplants have already restarted, so the drop in unadjusted claims may notbe as large as the seasonal factors “expect”, causing seasonallyadjusted claims to rise.

Thereis an additional subtlety which pushes in the same direction: theseasonal factor is multiplicative rather than additive. In a normalyear, with initial claims in the low 300,000 range, the 41% rise in theseasonal factor would imply an increase of perhaps 125,000-150,000 inthe weekly claims total.  But this year the underlying level of claims is much higher, closer to 600,000. This means that the implied increase in claims over the past two weeks– and the implied decrease over the next two weeks – is in the200,000-250,000 range.  Yet the auto industry isnot any larger than it was last year, so it is not at all obvious thatthe plant shutdown effect should be larger (it lasts for a longerperiod of time, but does not affect a substantially greater number ofpeople).  Therefore, unless there is a largedropoff in new claims from other industries, we would expect to see thenew claims figures move back up at least to the high 500,000s andlikely over the 600,000 mark over the next two weeks.  A smaller increase than this would suggest underlying labor market improvement.

Continuing claims: All the same issues are in play with continuing claims, with two additional layers of complication:

1. Immaculate continuing claims. In industries and states with institutionalized seasonal layoffs, it ispossible for temporarily laid-off employees to bypass the initialfiling process and go directly into the pool of continuing claims (i.e.they become a continuing claimant without ever having filed an initialclaim).  Auto industry seasonal layoffs are the most important example of this phenomenon.  One might call these “immaculate continuing claims,” as they never were “born” via an initial claim filing.

2. Benefit expirations.  The widely-reported series on continuing claims includes only standard jobless benefits, which expire after 26 weeks. After that time, workers may qualify for 13-20 weeks of additionalunemployment compensation under the EUC 2008 program (eligibility forthe 14^th-20^th week requires the state labor market to be particularly weak based on unemployment measures) passed by Congress last year. Also, another longstanding extended benefits program can offer anadditional 13-20 weeks, with all of this contingent on stateunemployment figures surpassing certain levels. As of the week of June 26, the latest for which figures were available,about 2.5 million Americans were receiving benefits under the EUC 2008program and nearly 300,000 under the extended benefits program. (Acorrection: our US Daily on July 9, “Hope that the Worst has Passed forLayoffs,” included only the latter series when tallying the number ofworkers on extended benefits.)  The number ofclaimants under these extended programs has increased rapidly in recentweeks as unemployed persons exhaust their regular benefits – so thestandard series on continuing claims is increasingly underreporting thetotal number of people receiving benefits, which is nearing 9 million.  Onaverage, over the last two months, the exhaustion of benefits hasremoved close to 100,000 people per week from the continuing claimsrolls, although this amount has varied significantly from week to week.

Aswith new claims, the continuing claims seasonal factors rise sharplyfrom the end of June until the second week of July—this year, by 14%. Before the recession began, with continuing claims in the 2-3 millionrange, this would have implied an unadjusted increase in continuingclaims of about 300,000-400,000 over the two-week period. Most of theincrease occurs in the first week of July (which corresponds to thecontinuing claims data reported today; continuing claims for a givenweek are released one week later than initial claims).  Notethat this increase of 300,000-400,000 unadjusted continuing claims in atypical year is much larger than the increase of “only” 125,000-150,000unadjusted initial claims that we noted above.   Themain reason for the difference is “immaculate continuing claims,” thoseworkers who appear on benefit rolls without having filed an initialclaim. 

Again, as with new claims, the seasonal factors are multiplicative. With continuing claims running at over 6 million this summer, the 14%seasonal factor implies an increase of more than 800,000 continuingclaims from the week of June 26 to the week of July 10, with most ofthat occurring in the first week of July (the numbers reported thismorning).  Regular claims actually did increaseon an unadjusted basis, but only by 64,000 – so after the seasonaladjustment, we observed a whopping week-on-week decline of more than600,000, by far the largest ever.

We speculated in a comment immediately after the report this morning that benefit exhaustions might explain the decline. It’s likely they played a role, but as noted above, they have tended toremove about 100,000 people per week from the rolls recently, with thepeak impact around 200,000 per week.  So, we nowthink that the seasonality issues discussed above were the primaryexplanation, with perhaps as much as 1/3 of this week’s decline incontinuing claims due to benefit exhaustions but probably less. 

Theseasonal effect for continuing claims will also unwind itself, but overa somewhat longer period of time than for initial claims. (We may see a small further decline in the seasonally adjustedcontinuing claims in the coming week, as the adjustment factor reachesits peak.)  The seasonal factor declines about 15% over the next eight weeks, or about 2% per week. This suggests that raw unadjusted continuing claims would need todecline by about 120,000 per week (2% of the current level of claims ofslightly over 6 million) for the seasonally adjusted figures to remainconstant.  This is pretty similar to the weekly average of benefit exhaustions noted above. Therefore, beyond the next week, the path of seasonally adjustedcontinuing claims should give us a somewhat cleaner signal about there-hiring of unemployed workers. 

Thebottom line: though we do believe the worst of the labor marketdeterioration has passed, this morning’s numbers substantiallyoverstate the rate of the improvement, and some backsliding in theseasonally adjusted claims numbers is likely over the next few weeks. Just as we are downplaying the improvement we have seen, we will alsodownplay the importance of modest increases over the next month.

Andrew Tilton