from Liberty Street Economics
— this post authored by Jason Bram and Fatih Karahan
News headlines highlighting the loss of at least 30 million jobs (so far) underscore the massive shock that has hit the U.S. economy and the dislocation, hardship, and stress it has caused for so many American workers. But how accurately does this number actually capture the number of net job losses?
In this post, we look at some of the statistical anomalies and quirks in the weekly claims series and offer a guide to interpreting these numbers. What we find is that the relationship between jobless claims and payroll employment for the month can vary substantially, depending on the nature, timing, and persistence of the disaster.
What Exactly Are Initial Jobless Claims?
When a worker is laid off, that person has the option to file a claim to receive unemployment benefits. These applications count as initial claims and are approved based on several eligibility criteria regarding the former employer, work history, and so on. Since 1967, the Department of Labor has produced a weekly report on the number of initial claims for unemployment insurance. The report, released each Thursday, also shows how many unemployed individuals are currently receiving benefits under unemployment insurance. While not all unemployed workers qualify for unemployment benefits, the initial claims number still provides insight into the health of the labor market.
How Appropriate Is Seasonal Adjustment?
The weekly initial jobless claims series is probably the most timely, accurate indicator of employment trends. However, there are some issues that analysts need to be aware of. One is that the national data are seasonally adjusted. This is a standard statistical procedure to discount normal seasonal swings. For example, jobless claims typically surge in early January, as millions of holiday-season positions come to an end. So to get a bead on what’s fundamentally happening in the economy, we need to adjust for this. For instance, if jobless claims are typically about 70 percent above normal in the first week of January (which they are), then it is necessary to scale down that figure so that it is comparable with levels in December, November, and every other month. Conversely, if jobless claims tend to be lower than usual at a certain time of year, as they are in the spring, then they should be scaled up appropriately.
Around this time of the year (spring), initial claims get scaled up by around 10 percent. This is a reasonable practice when the economy is running its regular course and the swings have a seasonal component like every other year. But applying this type of multiplicative seasonal adjustment now would basically imply that pandemics typically cause far fewer jobless claims in the spring than at other times . . . which, of course, is meaningless for a once-in-a-century type event. Clearly, the massive dislocation brought about by the COVID-19 outbreak has little to do with seasonal factors.
Initial claims are currently being used by many analysts to construct a “nowcast” for payroll employment and the unemployment rate – an estimate of the real-time figures for the current month. Using adjusted versus unadjusted figures may not be important when initial claims are around 200,000 – 300,000 (meaning that 20,000 – 30,000 claims can be attributed to seasonality). But when you get 3 – 6 million people filing claims in one week, as was the case for a number of weeks, the seasonal adjustment can make a sizable difference for the unemployment nowcast. In fact, for the five-week period from mid-March to mid-April, using the seasonally adjusted figures increases initial claims by two million – from 24 million to 26 million. It is also important to keep this in mind because in the weeks ahead, the seasonal adjustment will continue to inflate the claims numbers by about the same amount. So as we’re trying to figure out exactly what is happening in the economy, it is important to follow the right numbers.
What Do Initial Jobless Claims Tell Us About Net Job Loss?
Another issue in understanding initial jobless claims pertains to how these claims actually translate into net job losses. When we look at monthly payroll employment data, the headline number is typically the net change in employment; a negative figure is what we call a “net job loss.” In 2019, the average monthly net gain was about 175,000. Meanwhile, the average number of weekly initial jobless claims was about 217,000, which comes to about 940,000 per month. How can this be? Well, in large part, it is because initial claims is a gross, not net, measure of job loss. It counts the number of people who have lost jobs but not the number who have found jobs; during 2019, more than a million people found jobs in the typical month. In the current environment, of course, it is unlikely that as many people are finding new jobs. However, it is plausible that some essential businesses have had to hire people to fill in for absent workers or to enforce safe distancing practices, for example. It is also plausible that some businesses that were able to get Paycheck Protection Program (PPP) funding called back workers who had been laid off a few weeks prior. Thus, we will have to wait for the April employment report, out tomorrow (May 8), to accurately gauge the net effect of the widespread closing of businesses.
Looking at past major surges in jobless claims resulting from broad economic shocks illustrates how variable the relationship between initial claims and net job loss can be. After Hurricane Katrina devastated New Orleans in 2005, Louisiana’s statewide jobless claims climbed by 250,000 (which is proportionately close to the recent U.S. surge) and employment subsequently fell by 180,000, or about three-quarters of the initial claims number. In the two weeks after Hurricane Sandy hit the New York City metro area in late 2012, jobless claims surged by almost 60,000 in New Jersey – not even close, proportionately, to the current situation nationwide, but still sizable. But based on the final November employment report, the monthly job loss wound up being 16,000, or just over one-quarter of the initial claims.
At the other end of the spectrum, in the weeks after Hurricane Irma struck virtually the entire state of Florida in early September 2017, jobless claims climbed by about 35,000, whereas the monthly job loss figure was upward of 150,000 – more than four times as high – and all of it occurred in September, with a full bounce-back in October. This stark difference likely reflects that Irma hit right at the beginning of the monthly survey week (specifically, the week of September 10-16) and suggests that the timing of the disaster within the month can have a huge effect on how to interpret these statistics. This divergence also underscores how short-lived even a steep disruption in employment can be; it seems likely that many of the people who lost their jobs in early-to-mid September may have been called back before they even had a chance to file for unemployment.
This, of course, is not likely to be the case during the current pandemic, since its disruptions are clearly lasting for many weeks, not just a few days. Given the closer parallels to Hurricane Katrina in Louisiana – in both magnitude and duration – the relationship between jobless claims and job losses for that storm would seem to be the most relevant in trying to get a rough handle on the current situation. Of course, there still are many differences between the nationwide situation amid the coronavirus pandemic and Louisiana’s situation during the post-Katrina flooding, so it remains to be seen how many of the 24 million nationwide jobless claims filed between mid-March and mid-April translate into net monthly job losses.
About the Authors
Jason Bram is a research officer in the Federal Reserve Bank of New York’s Research and Statistics Group.
Fatih Karahan is a senior economist in the Bank’s Research and Statistics Group.
Source
Disclaimer
The views expressed in this post are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York or the Federal Reserve System. Any errors or omissions are the responsibility of the authors.
