Continuing claims point to 10% to 13% unemployment
admin | May 1, 2020
This document is intended for institutional investors and is not subject to all of the independence and disclosure standards applicable to debt research reports prepared for retail investors.
Financial markets tend to focus on initial jobless claims each Thursday, especially since the virus-induced economic collapse. The lesser known continuing claims, however, may be a more meaningful statistic. Initial claims reflect applications to collect benefits. Some applications may be rejected, while others are rescinded when the filer finds a new job or is recalled from furlough. Continuing claims come with less noise, and the latest readings suggest the jobless rate in April is on its way to 10% to 13%.
With state systems overwhelmed in recent weeks, there is a very real chance of a significant duplication in initial claims as desperate filers repeatedly called into phone systems or logged into web sites. Moreover, a number of states have started estimating initial claims, so that some of the states that have had difficulty processing filings are essentially assuming their way out of the problem in their reported initial claims figures.
The continuing claims data, in contrast, represent the number of people who are actually collecting benefits. This should be a more reliable and easily measured variable, and, no states are reporting estimated continuing claims. One drawback is that these data are reported with a 1-week lag, so the numbers are always a week behind initial claims.
Comparing initial and continuing claims
Initial and continuing claims may evolve differently for a few reasons: individuals with duplicate filings, individuals who filed but later found a job and are not or no longer collecting benefits, and processing delays. Duplicate and rescinded filings would lower continuing claims relative to the cumulative tally of initial claims, while processing delays could result in an undercount of continuing claims.
The issue of processing delays is actually quite complicated. Delays may lead to lags of a filing being processed at all (in which case, there is neither an initial claim or a continuing claim yet), or a lag between an initial claim translating to a continuing claim (if someone has an application registered but not approved immediately), or a third possibility that seems counterintuitive but shows up in the data (as you will see below) where someone is collecting benefits but an initial claim was never registered.
Commentators lately have focused on the number of initial claims from March 15 to April 18 and have tried to translate that directly to an unemployment rate estimate. That is no better than a loose relationship. The continuing claims numbers, perhaps with an adjustment to account for processing delays, offer a better input to project the unemployment rate figure.
Over the 5-week survey period, 24.4 million NSA initial claims came in. Meanwhile, the continuing claims tally for the week of April 18 was only 17.8 million. With more than two million people collecting benefits before mid-March, the change in the number of people collecting benefits over the survey period was “only” 15.7 million.
As a result, the change in those collecting benefits represents 64.4% of the reported initial claims over that period. At first glance, this seems awfully low, but the three sources of discrepancy listed above may be quite large. The difficulty is parsing out the gap among the three.
With no hard data to make an objective estimate, that leaves a few educated guesses. In the chaos during late March and early April, it is likely that duplication was a serious problem, much more so than during normal conditions. Let’s assume that duplication accounts for about 5% of initial claims. Further, the proportion of filers who find work might be somewhere in the neighborhood of 10% to 15% of claims. That relies on anecdotal reports of hiring by the select group of firms who have seen their businesses pick up due to the virus such as Wal-Mart and Amazon. That leaves somewhere in the neighborhood of 15% to 20% of claims that have been registered that may be still stuck in limbo.
State-by-state data on the initial and continuing claims gap
For the nation as a whole, the rise in continuing claims added up to 64.4% of cumulative initial claims over the employment survey period. However, there is considerable dispersion state-by-state.
Amazingly, there are five states that had more continuing claims than initial claims (Exhibit 1). At first glance, this would seem impossible, but it speaks to just how chaotic the last six weeks have been. Some state apparently got really bogged down, and New York and Connecticut definitely had major issues. If initial claims were delayed by more than a week or two, they would not have necessarily ever shown up in initial claims (the data are only revised going back one week). If a state office classifies an initial claim by the date it was filed rather than when it was processed, then any applications that took two weeks or more to process may never show up in initial claims. Note that this does not rule out the possibility that there is still a backlog of applications to be processed, which is likely the case in New York.
Exhibit 1: States with the highest continuing claims relative to initial claims
Meanwhile, other states have unusually low percentages (Exhibit 2). A few cases stand out. Florida has had a disastrous time processing claims, so it is not especially surprising that it shows up in the Bottom 10, especially because Florida over the past few weeks has begun to estimate its initial claims. Pennsylvania has been estimating initial claims since mid-March, so it was a prime candidate to show up in the Bottom 10. The percentage would suggest that the state has substantially overestimated the number of initial filers or it has struggled to process the filings.
Exhibit 2: States with the Lowest Continuing Claims Relative to Initial Claims
Before running these numbers, my first guess would have been that the states with well-known processing delays would have ended up in the Bottom 10, but New York and Connecticut actually show up near the top of the Top 10. I would also have imagined that states without processing issues in the Bottom 10 would reflect places where there has been a heavy amount of rehiring, a hypothesis that I am not sure is borne out.
My takeaway from the state-by-state data is that the claims data are even messier than I would have imagined. There are a number of potential distortions, moving in opposite directions, and the flippant estimates that the cumulative number of initial claims translates cleanly to an unemployment rate estimate is even further off-base than I would have first imagined.
Unemployment rate projection
The continuing claims figures can at least offer a ballpark guess about unemployment in April. The increase in continuing claims during the survey period was 15.7 million. Adding 15% of the cumulative number of initial claims to account for processing delays pushes the figure to just over 19 million. Meanwhile, the household survey measure of employment already dropped by 3 million in March. So, a simple translation from continuing claims to household employment would imply a drop of around 16 million.
It does get a little more complicated because furloughed workers now get put into three different buckets in the household survey results:
- Employed but not at work,
- Not employed but out of the labor force, and
My assumption is that 1 million or 2 million may be misclassified as employed but not at work in April, so I am using a guess of 14 million for the drop in employment in April. Of that, I have 2.5 million as not employed but out of the work force—the labor force contracted by 1.6 million in March, so the April figure could be much bigger than my estimate. That leaves a rise of about 11.5 million for the ranks of the unemployed. These estimates work out to an 11.6% unemployment rate. Just for reference, adding another 1 million to the drop in employment, all else equal, pushes the unemployment rate up by six tenths of a percentage point. So, if I bracket my estimates by 2 million on each side, then a plausible range for the April unemployment rate reading might be around 10% to 13%.