Understanding the Employment Situation in Maine
On Friday May 22, the Center for Workforce Research and Information (CWRI) and U.S. Bureau of Labor Statistics (BLS) released April workforce estimates for Maine. These include estimates of the unemployment rate and the change in nonfarm payroll jobs. The COVID-19 pandemic caused unprecedented job loss in Maine and the nation in the last two months. It also created significant data collection and estimation challenges. This document provides information on how these estimates are derived, what some of the challenges were for collection and estimation, and then answers some questions we expect to arise.
How this data is collected
The U.S. Bureau of Labor Statistics enters into cooperative agreements with state workforce agencies to collaboratively produce and disseminate workforce data. Two programs, Current Employment Statistics (CES) and Local Area Unemployment Statistics (LAUS), are designed to provide a monthly indication of workforce conditions and trends. Data from those programs is collected and produced through highly sophistical surveying and modeling techniques. CES data comes from a survey of employer payrolls; statewide LAUS data comes from a survey of households.
The payroll survey provides estimates of the number of jobs in workplaces in many industries and in metropolitan areas. The household survey provides estimates of labor force status of residents of the state, including whether they are employed, unemployed, or not in the labor force. The payroll survey has a larger sample size and representation of all jobs than the household survey has of the population of the state. The payroll survey often provides a better indication of employment trends, but in any given month it may over- or understate the change in jobs, or even the direction of change in jobs. More information on CES is here. More information on LAUS is here.
Data collection challenges during the pandemic
The COVID-19 pandemic presented data collection challenges for both surveys. Response rates to the payroll survey were lower than normal in April. Some establishments did not respond because they were closed, either temporarily or permanently, and others did not respond because they could not or because they had other priorities during a challenging time. BLS has attempted to estimate the magnitude of business closures during this period, but it is clear that there will be substantial revisions to this data over the course of the next year as more data emerges and the picture becomes clearer.
The challenge for the household survey was determining labor force status of many people, especially those who recently lost employment. The definition of unemployment has long been that it includes those who were actively engaged in work search in the previous month and were available for work, while retirees and others who did not look for work are not in the labor force. The pandemic disrupted this situation by preventing people who otherwise would have looked for work if not for personal safety concerns or stay-at-home orders that were issued by many states. Furthermore, many people did not look for work because they expect to return to their job. People meeting this description should have been counted as unemployed, rather than not in the labor force. It appears that many of the survey interviewers did not adequately take this into consideration, which caused the number of people considered not in the labor force to rise too much and the number of people who were unemployed not to rise enough. This was the case throughout the nation.
Timing of the surveys
The survey collects information on jobs and labor force status for the week or pay period that includes the 12th day of the month. Loss of employment that occurred later in April will be reflected in May estimates, which will be published June 19.
These data are not as accurate as many people believe. The 90 percent confidence interval for the statewide seasonally-adjusted unemployment rate is usually plus or minus 0.7 percentage points from the published rate. Nonfarm jobs estimates can be off by 2,000 or more for a month, even in periods of economic stability, as existed in 2018 and 2019. Earlier this year we published a blog on annual revisions, which shows how much these data were revised for the last two years. There is little doubt that revisions to 2020 data, which will be published in early 2021, will be much more substantial.
Why do the indications of employment decline differ between the two surveys?
Nonfarm job estimates from the payroll survey and resident employment estimates from the household survey generally move in the same direction over long periods of time. Over short periods they may diverge in the rate of change, and sometimes even in direction of change. Estimates from these surveys have different coverage and both have some margin of error. Both series are subject to annual revisions, which incorporate more complete information that tends to make them more correlated than preliminary data indicates.
The payroll survey counts the number of jobs in workplaces in the state, so it includes non-residents who work in the state, and it excludes residents who work in other states. As a count of jobs, a person with two or three jobs appears two or three times in this data. The largest job declines during the pandemic were in hospitality industries, which is the source of second or supplementary jobs for many people. It is likely that more jobs were lost than people who lost all employment.
The household survey counts employed residents of the state, regardless of what state they work in. It does not include non-residents who work in Maine. A person is counted once, no matter how many jobs they may have.
If nonfarm jobs decreased at a faster rate than the nation since February, why did Maine’s unemployment rate increase by so much less?
The change in nonfarm jobs comes from a survey of employer establishments. The unemployment rate is derived from a survey of households. They have different coverage and both involve some margin of error. Both are revised over time. Any divergence between them often becomes smaller when they are revised based on more complete information. Households in Maine reported a lower rate of employment loss compared to the national rate. The two surveys differ in concept, coverage, and sample size. The employer survey is much larger and tends to have a lower margin of error than the household survey.
Under normal circumstances, there are differences between estimates from the household and the payroll survey. The COVID-19 pandemic lead to disruptions in data collection that could have exacerbated these differences. Response rates in the household survey declined in April. This can be problematic in a state with a small population as the number of respondents to this survey is relatively low, typically around 500 households per month. Responses to the payroll survey were also impacted by the pandemic. The Bureau of Labor Statistics uses statistical models that do not expect such dramatic changes in such a short period.
Why did the number of unemployed increase so much less than employment decreased?
The labor force is comprised of employed and unemployed people. Those considered to be unemployed were engaged in work search during the month and were available for work. Other jobless people such as retirees, full-time students, homemakers, inmates, and others who did not search for work or were not available for work are not counted as unemployed – they are not in the labor force.
The household survey found that labor force participation in Maine decreased from 62.4 percent in February to 59.8 percent in April. The 2.6 percentage point decrease is historically large for such a short period of time. This indicates that many people who lost their job did not engage in work search or were not available to work, so they were not counted as unemployed. Given the personal safety concerns and stay-at-home order in Maine in March and April, many who lost a job and wanted a new job did not or could not search for work. If those barriers to work were not present, about 28,600 more people would have been counted as unemployed, and the reported unemployment rate would be close to 14.3 percent. (The same calculation yields an 18.9 percent unemployment rate for the U.S.)
The Department of Labor reported 139,400 initial claims for state unemployment benefits in the nine weeks since the middle of March. Why is the estimated number of unemployed people so much lower?
The reference week for unemployment estimates is the second week of the month. There were 89,500 initial claims between the middle of March and the middle of April. That is closer to, but still higher than the estimate of 70,800 unemployed in April. There are a number of reasons that the number of initial claims for unemployment benefits differs from the number of unemployed people. In the regular state unemployment insurance program, a person who is determined ineligible in one quarter may be eligible in another. Because April began a new quarter, many people who were deemed ineligible in March needed to file a new initial claim in April.
Additionally, rules for the Pandemic Unemployment Assistance program required that self-employed and others covered by this new program first be deemed ineligible for regular state UI, so many of those people had initial claims in both programs. These and some other technical issues caused the number of initial claims to exceed the number of people applying for unemployment benefits.
Some people who made initial claims were called back to work or gained a new job. The Paycheck Protection Program (PPP) in the CARES Act passed by Congress incentivized employers through loan guarantees to call back to work those they had laid off. The PPP program became effective weeks after many people had lost their job and applied for unemployment benefits, so many people who initially applied for unemployment benefits regained employment later. People who are on employer payrolls, whether they are working or not, are considered to be employed.
Finally, unemployment benefits are for people who lost a job. Though unemployment usually stems from job loss, those who left a job by choice, or who entered or reentered the labor force by beginning work search or by becoming available to work also are counted as unemployed.
This information was prepared by:
Mark McInerney, the new Director of the Center for Workforce Research and Information. Mark has a PhD in economics from the University of Connecticut.
Glenn Mills, Deputy Director of CWRI. Glenn oversees CES and LAUS program activities, and he has primary responsibility for analyzing and reporting on workforce conditions and trends in Maine.