by Sanjukta Chaudhuri
September 2020
A recurring issue in labor market statistics during the pandemic period has been the wide gap between Unemployment Insurance (UI) continued claims and the number of unemployed labor force participants and, as a result, a lower unemployment rate than would be expected based on continued claims data.
The sources of the claim counts and the number of unemployed are different. UI continued claims are a count of people who request a UI benefit payment weekly. On the other hand, the raw count of unemployed participants is from the monthly Current Population Survey (CPS), a survey of 60,000 households nationwide (900 in Minnesota) designed to extract information on labor force status, employment status, unemployment status, occupation, and various other labor market, education, and demographic information. Although the CPS produces a raw or "CPS unemployment rate", nonetheless, this is not the official monthly unemployment rate, which is calculated using additional data input.
An estimation process done by the Local Area Unemployment Statistics (LAUS) program produces the official unemployment rate. Under this program, the CPS, together with state level inputs on UI claims and payroll employment are used to estimate the "official" unemployment rate. The CPS count of the number of unemployed persons is a key input into the unemployment estimate. If the count of unemployed is lower than the count of UI claimants, the resulting official unemployment rate will potentially be lower than expected.
Given the dramatic impact of pandemic containment measures on workers, a reoccurring question since March has been: Why is the official unemployment rate lower than an "expected" unemployment rate, with the expected being based on the huge surge in UI claims? With continued claims representing ongoing unemployment, why is the official unemployment rate consistently lagging the surge in claims?
Table 1 shows the gap between the CPS count of unemployed and initial and continued claims from January through July. While there was only a small gap in the pre-pandemic months, the gap has since increased with UI numbers much larger than the count of unemployed. For example, the UI claims exceeded the count of unemployed by 134,726 in April.
Table 1. UI Claims vs. CPS count of unemployed in Minnesota
| Month in 2020 | UI Continued Claims | CPS unemployed* |
|---|---|---|
| January | 63,316 | 93,900 |
| February | 63,837 | 106,000 |
| March | 70,118 | 130,240 |
| April | 399,767 | 265,041 |
| May | 395,180 | 329,889 |
| June | 333,555 | 277,577 |
| July | 292,008 | 203,411 |
Source: Minnesota PROMIS file. These numbers exclude Pandemic Unemployment Assistance (self-employed) claims. DEED data tools and Current Population Survey. All numbers are for Minnesota
It is important to note that Pandemic Unemployment Assistance (PUA) claimants are not included in these numbers. PUA provides for up to 39 weeks of benefits to individuals who are self-employed under the CARES Act. Including PUA claimants would increase the UI continued claims number by close to 100,000 individuals or close to 25% more claimants. We are currently working on creating an accurate weekly and monthly report that includes PUA claimants so that we can have a broader understanding of the impact of the COVID recession.
An earlier article published by DEED's Labor Market Information office summarized some of the reasons that, taken together, explain why this is happening, with the main takeaway being that the CPS survey is designed to capture more predictable changes in the labor market, not the wild swings that have happened during the pandemic.
As questions remain, the BLS has provided additional information on the discrepancy between CPS unemployed and UI claims numbers, including some numerical experiments, in an attempt to further understand why these numbers diverge. This article examines this discrepancy utilizing data for Minnesota.
Some unemployed workers erroneously reported being employed but not at work in the CPS. These workers were misclassified as employed (rather than unemployed), thus creating an underestimate of the number of unemployed. This misclassification has to do with the way "employed" and "unemployed" are defined in the survey.
The CPS counts the employed as either "employed, at work", or "employed, absent from work". Those who are employed but absent from work are further sub-divided by reason for not being at work, such as illness or vacation. Those who did not specify why they were not at work are categorized as "employed, absent from work for other reasons".
BLS has reported that there is a likelihood that some workers who were actually unemployed were misclassified as "employed but not at work for other reasons" during the pandemic period. It is easy to see why this would be an easy mistake for workers to make considering how quickly so many people lost jobs and the uncertainly surrounding the amount of time they might be out of work. The lack of the standard in-person follow up questioning (due to social distancing protocols) that is normally intended to clarify the true labor market status of a respondent greatly exacerbated this problem by giving the CPS fewer tools to use to clarify responses.
Moreover, workers who are employed part-time due to economic reasons, but usually work fulltime, are counted as part-time employed workers in the CPS survey. Many of these workers were eligible for UI benefits; data from Minnesota's UI program indicates that 20,000 to 30,000 part-time underemployed workers were drawing benefits in April and May, for example. This is the second reason for higher UI numbers compared to CPS unemployed number. The official count of unemployed, a measure called U-3, does not include part-time underemployed workers, even those who are drawing UI benefits, since they are, in fact, employed.
In this section we add the part-time underemployed and those whose employment status was misclassified into the unemployed number and remove them from the employed number to provide an estimate of the extent to which they may account for the discrepancy between the count of UI claims and the number of unemployed. These do not in any way provide an alternative unemployment rate since they include individuals who are working and therefore not unemployed under the U-3 definition. Instead, this is simply a way to explore the reasons behind the gap between the number of UI claimants and CPS unemployed, and should be taken purely in the spirit of a statistical experiment meant to further our understanding of the gap.
To address the CPS misclassification issue we calculated the average number of "employed but not at work" workers for pre-pandemic years 2016 through 2019 for March through June, and then examined if the pandemic numbers for March through June of 2020 vary significantly. The last column of Table 2 shows an increasing divergence between the pre-pandemic and pandemic periods. The gap was only 2,556 in March, but increased sharply to 60,918 in April, and remained high through May and June, finally dropping to only 4,249 in July.
Based on this, what would the adjusted number of unemployed be, assuming that all of the excess during the pandemic period over the pre-pandemic average was misclassified? We added the "excess" in Table 2 to the number of CPS "unemployed" count shown in Table 3, simultaneously subtracting it from the CPS count of employed (see Table 3). Adding the substantial excess, indicating a misclassification, increases the June unemployed numbers, for example, from 277,577 to 312,567, or an addition of 34,990, while the number of employed drops from 2,883,201 to 2,848,211.
Table 2. Employed, Not at Work for Other Reasons
| Month | Avg. 2016-2019 | 2020 | Excess |
|---|---|---|---|
| March | 14,937 | 17,493 | 2,556 |
| April | 7,463 | 68,381 | 60,918 |
| May | 8,048 | 57,513 | 49,465 |
| June | 14,934 | 49,924 | 34,990 |
| July | 23,464 | 27,714 | 4,249 |
Source: Current Population Survey; all numbers are for Minnesota
Table 3. Adjusted (Hypothetical) CPS unemployment number after adding difference in those classified as "Employed but not at work for other reason"
| Month in 2020 | CPS Employed | CPS Unemployed | Adjustment | Employed Adjusted | Unemployed Adjusted |
|---|---|---|---|---|---|
| March | 2,954,492 | 130,240 | 2,556 | 2,951,936 | 132,796 |
| April | 2,719,272 | 265,041 | 60,918 | 2,658,354 | 325,959 |
| May | 2,732,098 | 329,889 | 49,465 | 2,682,633 | 379,354 |
| June | 2,883,201 | 277,577 | 34,990 | 2,848,211 | 312,567 |
| July | 2,886,693 | 203,411 | 4,249 | 2,882,444 | 207,660 |
Source: Current Population Survey, all numbers are for Minnesota
In addressing the second possible source of discrepancy, we asked a similar question: for the count of those "working part-time for economic reasons", is there an excess of these workers during the pandemic period compared to previous years? If so, then assuming all of these excess part-timers were drawing UI, what would the total number of unemployed plus part-time for economic reasons be if we added them to the CPS number of unemployed? In Table 4 we see a wide divergence in this measure in April through July of 2020 over the previous years' averages. For example, in April, the number of persons working part-time for economic reasons was 188,800, an excess of 112,275 over the average of April 2016 through 2019.
Table 4. Working part-time for economic reasons
| Month | Avg. 2016-2019 | 2020 | Excess |
|---|---|---|---|
| March | 85,925 | 75,700 | -10,225 |
| April | 76,525 | 188,800 | 112,275 |
| May | 76,800 | 190,500 | 113,700 |
| June | 82,750 | 162,600 | 79,850 |
| July | 71,925 | 133,800 | 61,875 |
Source: Current Population Survey, all numbers are for Minnesota
Assuming that all of these workers received UI due to a significant reduction in work hours, Table 4 adds them to the CPS unemployed and simultaneously subtracts them from the CPS employed. This gives us the last two columns of Table 5. As an example, if we add both sources of discrepancy (employed but not at work + part-time for economic reasons), the June unemployment number increases from 277,577 to 392,417.
Table 5. Adjusted number after adding difference in those classified as "Employed but not at work for other reason" and "Working part-time for economic reasons"
| Month in 2020 | CPS Employed | CPS Unemployed | Adjustment | Employed Adjusted | Unemployed Adjusted |
|---|---|---|---|---|---|
| March | 2,954,492 | 130,240 | -7,669 | 2,962,161 | 122,571 |
| April | 2,719,272 | 265,041 | 173,193 | 2,546,079 | 438,234 |
| May | 2,732,098 | 329,889 | 163,165 | 2,568,933 | 493,054 |
| June | 2,883,201 | 277,577 | 114,840 | 2,768,361 | 392,417 |
| July | 2,886,693 | 203,411 | 66,124 | 2,820,569 | 269,535 |
Finally, Table 6 shows what the rate would be with these adjustments. This cannot be considered an unemployment rate since those who are working part-time for economic reasons are employed under the U-3 definition. Clearly, these experimental numbers produce a much higher rate during the pandemic period than the CPS data produce. These experimental rates show why there is such a gap between the rate we may expect due to UI claims and the actual official unemployment rate.
Table 6. CPS and experimental unemployment rate for Minnesota
| Month in 2020 | **CPS Unemployment Rate | Experimental Rate |
|---|---|---|
| March | 4.2% | 4.0% |
| April | 8.9% | 14.7% |
| May | 10.8% | 16.1% |
| June | 8.8% | 12.4% |
| July | 6.6% | 8.7% |
** The CPS unemployment rate is the survey based unemployment rate. This is not the same as the official, estimated unemployment rate, although the CPS is a critical component of the estimation process that produces the official unemployment rate.
