We use cookies on our website to give you the best online experience. Please know that if you continue to browse on our site, you agree to this use. You can always block or disable cookies using your browser settings. To find out more, please review our privacy policy.

COVID-19 RESOURCES AND INFORMATION: See the Atlanta Fed's list of publications, information, and resources; listen to our Pandemic Response webinar series.


Policy Hub: Macroblog provides concise commentary and analysis on economic topics including monetary policy, macroeconomic developments, inflation, labor economics, and financial issues for a broad audience.

Authors for Policy Hub: Macroblog are Dave Altig, John Robertson, and other Atlanta Fed economists and researchers.

Comment Standards:
Comments are moderated and will not appear until the moderator has approved them.

Please submit appropriate comments. Inappropriate comments include content that is abusive, harassing, or threatening; obscene, vulgar, or profane; an attack of a personal nature; or overtly political.

In addition, no off-topic remarks or spam is permitted.

April 13, 2020

Unpaid Absence from Work Because of COVID-19

As has been widely reported, the March employment report shed light on the early impact of the COVID-19 pandemic on the U.S. labor market. One telling number is the official unemployment rate, which tallies the share of the labor force made up of people out of work (but who are looking for a job) plus those who have been laid off and expect to be recalled. The official unemployment rate increased from 3.5 percent in February to 4.4 percent in March. My own preferred measure is the share of the working-age population who are unemployed, working part-time because of economic conditions, or outside the labor force but still want to work. This underutilization rate is featured in the Atlanta Fed's Labor Report First Look analysis of the U.S. Bureau of Labor Statistics (BLS) labor report. The First Look shows that this measure increased from 5.8 percent in February to 7.1 percent in March.

According to the BLS, people who did not work "because of the coronavirus" were supposed to also be classified as unemployed (on temporary layoff). However, relative to February, it appears that around 1 million more people were classified as employed but absent from work and not getting paid for "other reasons" (reasons other than illness, childcare problems, bad weather, being in school, etc.). If those people were instead counted as unemployed, then the unemployment rate in March would have been closer to 5.0 percent, and the underutilization rate would have been 7.5 percent. The data in the table below break down the various components of these measures, if you want to perform your own calculations.


March 2020

One-month change

Labor force






Part-time for economic reasons



Unpaid absence from work for other reasons



Want a job



Don't want a job



Note: Data represent thousands of people and are seasonally adjusted except for unpaid absence. People in the "Want a job" category include those who currently want a job but are not counted as unemployed. The "Don't want a job" category includes people not in the labor force who say they don't currently want a job (see here for more details). "Unpaid absence" includes those who have a job but are on an unpaid absence from work for an unspecified reason.

Source: BLS, Federal Reserve Bank of Kansas City's Center for the Advancement of Data and Research in Economics, and author's calculations

A few observations about the absence-from-work data for March. First, a marked increase in absence from work across occupations took place, in both paid and unpaid absences. However, being paid while absent from work for "other reasons" was disproportionately prevalent among professional occupations, likely reflecting those workers' relatively greater ability to continue to work remotely. Second, unpaid absence from work was disproportionately common among food-preparation and building-maintenance occupations, as well as jobs providing personal services—the types of occupations most directly affected by social distancing mandates.

The April BLS labor report (due on May 8) will provide a clearer picture of the depth and breadth of the impact of COVID-19 on the labor market, and it will be important to include unpaid absences from work in the analysis. In the meantime, I'm keeping a close eye on the weekly unemployment insurance claims data as well as other high-frequency surveys (such as this one from Gallup).

April 9, 2020

Amid the COVID-19 Crisis, a Tale of Two Cities

The COVID-19 pandemic has resulted in both a major public health crisis and a major economic crisis. The economic impact is coming primarily via social distancing (or stay-in-place) restrictions that have resulted in the temporary closing of nonessential businesses in many parts of the country. As a result, millions of workers have been laid off or furloughed. Unemployment claims for the week ending March 21 totaled more than 3 million—the highest number of seasonally adjusted initial claims in the history of the series at that time. That record was broken quickly. For the week ending March 28, the number of seasonally adjusted initial claims increased to an adjusted 6.9 million. For the week ending April 4 there were an additional 6.6 million new unemployment insurance claims.

In response, Congress has passed two laws that contain measures designed to help individuals affected by this shock: The Families First Coronavirus Response Act (FFCR Act), signed into law on March 18, and the Coronavirus Aid, Relief, and Economic Security Act (CARES Act), signed into law on March 27. (In addition, the Coronavirus Preparedness and Response Supplemental Appropriations Act was signed into law on March 6 and provided $8.3 billion in emergency funds to state and local governments to fight the outbreak.)

In this macroblog post, we summarize how the assistance offered in the acts supplement the preexisting social safety net to financially support a hypothetical displaced American working in a restaurant prior to the outbreak. We present our case study in two locations: Boston, Massachusetts, and Birmingham, Alabama. Overall, the potential impact of these acts depends on the ultimate length of the unemployment spell, variation in state unemployment insurance laws, and variation in cost of living.

Our case study: Introducing Chef Andrew

Chef Andrew is single adult 25 years old without children. He works full-time in a restaurant, earning the area-specific typical entry-level wage for chefs. On April 1, Andrew's restaurant employer closes. He is laid off but is unable to find another chef job since many other restaurants are also closed or operating in limited capacity.

What the acts and existing social safety net do

Both acts have a variety of provisions to help those affected by COVID-19, but we focus only on the key unemployment-related components of the acts that benefit workers directly. By focusing only on workers, the analysis excludes components that incentivize employers to retain or rehire workers. In addition to the acts, the preexisting social safety net provides assistance to low-income individuals, including those who encounter short-term periods of financial distress. We review the key elements of the acts and the social safety net that are relevant to Chef Andrew.

  • Individual receipt of up to $1,200 in tax rebates and an additional $500 per qualifying child. The rebate begins phasing out when a single filer's annual income exceeds $75,000.
    • We assume Chef Andrew makes $34,472 per year in Boston and $37,000 in Birmingham. In either city, he receives the full rebate of $1,200.
  • A suspension of all payments on federal student loans through September 30 and a pause on the accrual of interest.
    • We assume Chef Andrew pays $100 per month in student loan payments. The suspension of payments suspension will allow him to reduce monthly expenses between April and September.
  • Expansion of unemployment insurance (UI). On top of the regular state UI benefits, the federal government provides an additional $600 per week in federal pandemic unemployment compensation through July 31. After an individual reaches the state's maximum number of weeks on UI, emergency federal UI benefits extend the state UI benefits for up to an additional 13 weeks.
    • In Birmingham, Chef Andrew's regular state UI payment is $275 per week for 14 weeks. Boston provides a higher dollar amount of unemployment compensation: $353 per week for 26 weeks.
  • The FFCR Act suspends work requirements for determining Supplemental Nutrition Assistance Program (SNAP) eligibility until the declared end of the national public health emergency.
    • Without the act, Andrew would have access to SNAP for three months only while unemployed. The act provides possible extended eligibility to the program. The initial amount he receives is small (less than $20 per month). However, if he is still unemployed after the maximum duration of UI, the size of the support will be larger (about $200 per month).
  • Medicaid and health insurance subsidies provided by the Affordable Care Act (ACA) help low-income individuals afford health insurance, but the laws vary significantly from state to state.
    • When Andrew becomes unemployed, we assume he loses access to employer-sponsored health insurance. In Boston, Andrew is eligible for either Medicaid or ACA subsidies, depending on the amount of his UI income. This significantly reduces his health insurance costs compared to Alabama, which has not expanded Medicaid under the ACA. Therefore, Andrew in Alabama has to pay the full cost of health insurance on the private market when he is unemployed. We estimate this cost to be about $400 per month.

The accompanying analysis describes the assumptions and specifics of how each program assists Andrew during his period of unemployment.

Putting it all together: Birmingham and Boston

Here, we summarize how the provisions in the FFCR and CARES acts, along with the preexisting social safety net, combine to support Andrew and help him pay his living expenses. The vertical axis in the following chart shows net resources, which we define as the sum of after-tax income, SNAP, Medicaid, ACA subsidies, and assistance from the two acts, minus the basic expenses affected by the acts and safety net provisions (food, health insurance, rent, and student loan payments). A value of net resources above zero can be thought of as the excess amount of money that can be used to cover additional expenditures.

The chart shows monthly totals of Andrew's net resources from January 2020 to January 2021. We assume his period of unemployment and the distribution of the acts' funds begin in April 2020. We also assume he remains unemployed for 10 months, a period that includes the maximum duration of assistance under the acts in Massachusetts—39 weeks—and an additional month of unemployment simply to illustrate his financial status without any of the acts' financial support.

Since the dynamics of the net resources over time is complex and depends on changes in individuals' income and expenses over time, we provide a simplified summary here and present a detailed analysis in the accompanying analysis.

Net Resources: Birmingham, AL vs. Boston, MA

Before the crisis, Andrew has about $1,000 of slack in his monthly budget in Birmingham. In Boston, due to the slightly lower wage and higher living expenses (rent in particular, which is more than twice as high in Boston as in Birmingham), he cannot meet his covered expenses even before he becomes unemployed. Thus, in this period, he would likely have to borrow to make ends meet.

When he loses his job in April, the acts (represented by the green lines) allow Andrew to cover basic expenses in both cities from April until August. Despite the fact that the financial support package is larger in Boston than in Birmingham, net resources are higher in Birmingham for this period because of the difference in the costs of living. In September, if Andrew is still unemployed, his net resources drop below zero in both cities. The FFCR and CARES Acts improve Andrew's financial security relative to a world without the passage of the acts. Without the acts (represented by the blue lines), net resources drop below zero as soon as Andrew loses his job and remain negative for the duration of unemployment.

Our case study highlights that although the FFCR and CARES Acts provide financial stability for displaced workers in the short term, the size and duration of the acts' positive impact will depend on individual circumstances, including their income prior to unemployment, state of residency, and household composition. Duration of the crisis also matters. The longer the COVID-19 crisis continues, the greater the financial stress for many households and the greater the call for additional policy action.

March 22, 2019

A Different Type of Tax Reform

Two interesting, and important, documents crossed our desk last week. The first was the 2019 edition of the Economic Report of the President. What particularly grabbed our attention was the following statement from Chapter 3:

Fundamentally, when people opt to neither work nor look for work it is an indication that the after-tax income they expect to receive in the workforce is below their "reservation wage"—that is, the minimum value they give to time spent on activities outside the formal labor market.

That does not strike us as a controversial proposition, which makes the second of last week's documents—actually a set of documents from the U.S. Department of Health and Human Services (HHS)—especially interesting.

In that series of documents, HHS's Nina Chien and Suzanne Macartney point out a couple of things that are particularly important when thinking about the effect of tax rates on after-tax income and the incentive to work. The first, which is generally appreciated, is that the tax rates that matter with respect to incentives to work are marginal tax rates—the amount that is ceded to the government on the next $1 of income received. The second, and less often explicitly recognized, is that the amount ceded to the government includes not only payments to the government (in the form of, for example, income taxes) but also losses in benefits received from the government (in the form of, for example, Medicaid or child care assistance payments).

The fact that effective marginal tax rates are all about the sum of explicit tax payments to the government and lost transfer payments from the government applies to us all. But it is especially true for those at the lower end of the income distribution. These are the folks (of working age, anyway) who disproportionately receive means-tested benefit payments. For low-wage workers, or individuals contemplating entering the workforce into low-wage jobs, the reduction of public support payments is by far the most significant factor in effective marginal tax rates and the consequent incentive to work and acquire skills.

The implication of losing benefits for an individual's effective marginal tax rate can be eye-popping. From Chien and Macartney (Brief #2 in the series):

Among households with children just above poverty, the median marginal tax rate is high (51 percent); rates remain high (never dipping below 45 percent) as incomes approach 200 percent of poverty.

Our own work confirms the essence of this message. Consider a representative set of households, with household heads aged 30–39, living in Florida. (Because both state and local taxes and certain transfer programs vary by state, geography matters.) Now think of calculating the wealth for each household—wealth being the sum of their lifetime earnings from working and the value of their assets net of liabilities—and grouping the households into wealth quintiles. (In other words, the first quintile would the 20 percent of households with the lowest wealth, the fifth quintile would be the 20 percent of households with the highest wealth.)

What follows are the median effective marginal tax rates that we calculate from this experiment:

Wealth percentile

Median Effective Marginal Tax Rate

Lowest quintile


Second quintile


Third quintile


Fourth quintile


Highest quintile


Note: The methodology used in these calculations is described here and here.
Source: 2016 Survey of Consumer Finances, the Fiscal Analyzer

Consistent with Chien and Macartney, the median effective marginal tax rates for the least wealthy are quite high. Perhaps more troubling, underlying this pattern of effective tax rates is one especially daunting challenge. The source of the relatively high effective rates for low-wealth individuals is the phase-out of transfer payments, some of which are so abrupt that they are referred to as benefits, or fiscal, cliffs. Because these payments differ widely across family structure, income levels within a quintile, and state law, the marginal tax rates faced by individuals in the lower quintiles are very disparate.

Note: The methodology used in these calculations is described here and here.
Source: 2016 Survey of Consumer Finances, the Fiscal Analyzer

The upshot of all of this is that "tax reform" aimed at reducing the disincentives to work at the lower end of the income scale is not straightforward. Without such reform, however, it is difficult to imagine a fully successful approach to (in the words of the Economic Report) "[increasing] the after-tax return to formal work, thereby increasing work incentives for potential entrants into the labor market."



January 18, 2018

How Low Is the Unemployment Rate, Really?

In 2017, the unemployment rate averaged 4.4 percent. That's quite low on a historical basis. In fact, it's the lowest level since 2000, when unemployment averaged 4.0 percent. But does that mean that the labor market is only 0.4 percentage points away from being as strong as it was in 2000? Probably not. Let's talk about why.

As observed by economist George Perry in 1970, although movement in the aggregate unemployment rate is mostly the result of changes in unemployment rates within demographic groups, demographic shifts can also change the overall unemployment rate even if unemployment within demographic groups has not changed. Adjusting for demographic changes makes for a better apples-to-apples comparison of unemployment today with past rates.

Three large demographic shifts underway since the early 2000s are the rise in the average age and educational attainment of the labor force, and the decline in the share who are white and non-Hispanic. These changes are potentially important because older workers and those with more education have lower rates of unemployment across age and education groups respectively, and white non-Hispanics tend to have lower rates of unemployment than other ethnicities.

The following chart shows the results of a demographic adjustment that jointly controls for year-to-year changes in two sex, three education, four race/ethnicity, and six age labor force groups, (see here for more details). Relative to the year 2000, the unemployment rate in 2017 is about 0.6 percentage points lower than it would have been otherwise simply because the demographic composition of the labor force has changed (depicted by the blue line in the chart).

In other words, even though the 2017 unemployment rate is only 0.4 percentage points higher than in 2000, the demographically adjusted unemployment rate (the green line in the chart) is 1.0 percentage points higher. In terms of unemployment, after adjusting for changes in the composition of the labor force, we are not as close to the 2000 level as you might have thought.

The demographic discrepancy is even larger for the broader U6 measure of unemployment, which includes marginally attached and involuntarily part-time workers. The 2017 demographically adjusted U6 rate is 2.5 percentage points higher than in 2000, whereas the unadjusted U6 rate is only 1.5 percentage points higher. That is, on a demographically adjusted basis, the economy had an even larger share of marginally attached and involuntarily part-time workers in 2017 than in 2000.

The point here is that when comparing unemployment rates over long periods, it's advisable to use a measure that is reasonably insulated from demographic changes. However, you should also keep in mind that demographics are only one of several factors that can cause fluctuation. Changes in labor market and social policies, the mix of industries, as well as changes in the technology of how people find work can also result in changes to how labor markets function. This is one reason why estimates of the so-called natural rate of unemployment are quite uncertain and subject to revision. For example, participants at the December 2012 Federal Open Market Committee meeting had estimates for the unemployment rate that would prevail over the longer run ranging from 5.2 to 6.0 percent. At the December 2017 meeting, the range of estimates was almost a whole percentage point lower at 4.3 to 5.0 percent.