Next up, the May employment report
admin | May 29, 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.
Many analysts have talked about the unemployment rate jumping to 20% or even 25% in the coming months. But with most states reopening in earnest, workers should begin to return to their jobs in large numbers soon. The jobless rate will likely peak in May or, at the latest, June. Based in part on unemployment claims, the jobless rate should fall somewhat short of 20% in May.
The last two months has been an unprecedented period for the labor market. The US has never seen tens of millions of layoffs over such a short period. Many states saw their unemployment claims systems overwhelmed for a while, but almost all states appear to have finally worked through the processing backlogs, making the weekly observations somewhat more reliable as an indicator of the labor market.
Initial claims readings over the last month point to two broad takeaways, each somewhat at odds with the other. First, initial claims have steadily declined from the late March peak. The number of new filers has dropped by more than 200,000 weekly in each of the last eight periods. At the same time, however, initial claims surprisingly still stand above 2 million in the week ended May 23.
In any case, initial claims over the 5-week April employment survey reference period totaled more than 24 million. Over the 4-week May survey reference period, the regular program registered about 12 million, while another 4 million or so filed under the pandemic assistance program, allows workers who would ordinarily be ineligible, such as gig workers, to collect federal benefits. In total, about 16 million new claims came in over the May employment survey reference period.
However, it is becoming increasingly clear that initial claims are inflated in certain states. In Georgia and Kentucky, states not particularly hard hit by lockdowns, initial claims since mid-March total almost half of the February payrolls compared to a national average of about 25%. The initial claims tally through the mid-May period consequently may be around 2 million “too high,” making the right number may be closer to 14 million.
Meanwhile, the continuing claims figures measure the total number of people collecting benefits. Thus, in theory, this gauge should be more closely correlated with the unemployment rate. However, that relationship did not work out so well in April. Initial claims totaled about 25 million in the five-week April survey reference period, while continuing claims “only” rose by 17 million over that span.
As it turns out, the cumulative drop in the household survey gauge of employment in March and April was just over 25 million, while payrolls sank by over 21 million over the same timeframe.
The continuing claims figures may have been grossly underestimated as of mid-April due to processing delays that left many workers waiting for benefits for far too long. Presumably, those processing delays had been predominantly ironed out by the middle of May. However, there are still a few anomalies in these data. For the readings as of May 16, the week coinciding with the May survey reference period, continuing claims sank precipitously in both California and Florida. The California figures have been gyrating up and down in a seesaw pattern, which is clearly erroneous but may reflect the fact that California requires workers to recertify their status every other week. The number of people collecting benefits jumped to 3.6 million in the week ended May 9 and then receded to 2.1 million in the following period. Meanwhile, in Florida, continuing claims spiked from 1.0 million to 2.2 million in the period ended May 9, which I thought meant that the infamous delays in Florida’s processing had finally been resolved. However, the reading sank back to 509,000 in the week ended May 16, an absurdly low reading for a state with almost 9 million workers prior to March.
In all, continuing claims in the regular program rose by about 3 million over the 4-week May survey period, while the pandemic program may have added about 8 million claimants (7 million so far with one week to go), for a combined total of about 11 million. Adjusting for the anomalies described above, the “right” number might be closer to 12 million. However, the trouble with this calculation is that the states were greatly delayed in getting their pandemic programs up and running. In most states, state labor offices were unable to accept applications from workers who were ineligible for the regular program until late April or early May, even though the bulk of the underlying layoffs probably occurred in late March.
To avoid such timing issues, one way to examine the data is to look at the combined increase from mid-March and mid-May. By that filter, initial claims rose by 39 million, while continuing claims jumped by 28 million. Adjusting for some of the anomalies in the data, the better figures are 37 million and 29 million respectively.
Translating initial and continuing claims to unemployment
Initial claims only cover one side of the employment equation, layoffs. In April, that probably did not matter much because very few people were being hired, but as we moved to May, the flows in the labor market were likely beginning to work in both directions (hiring and layoffs). In retrospect, the continuing claims for April appear to have been too low relative to the payroll and household survey results, but I would argue that mainly reflects the delays noted above. I expect the May household survey results to come in somewhere in between the 29 million and 37 million figures derived above from the two claims measures. In fact, as a best guess, I am assuming that the household gauge of employment may show a 33 million drop through May, which would imply a decline of about 8 million in May alone (vs. 3 million in March and 22.4 million in April).
Even so, getting to an unemployment rate estimate is a little more complicated. There are several different categories that furloughed workers can fall into. First, the BLS has reported that several million workers who were furloughed classified themselves as employed but not at work, which means that the number of jobs lost was likely millions higher than even the official household survey figures for April. Second, of those that lost their jobs, some would be classified as falling out of the labor force. In March and April, the labor force contracted by 8 million, which means that the 25 million loss of jobs only translated into a spike of about 17 million in the ranks of the unemployed.
For the purposes of the May forecast, assume an 8 million loss of jobs as measured by the household survey, while the labor force may have contracted by just over 2 million, consistent with the relationship between those two variables in March and April combined. This would push the total number of unemployed up by another 5.8 million to almost 29 million and would send the unemployment rate to about 18.7%, about a percentage point below the current consensus forecast.
I should note that if it remains the case in May that several million furloughed workers are erroneously classified as still employed but just absent from work, then the “true” unemployment rate could be several percentage points higher than the stated figure. In that sense, even my relatively low forecast would be consistent with the 20% to 25% range that has been frequently cited by Fed officials and economists.