Federal Reserve Economic Data

The FRED Blog has used US Bureau of Economic Analysis data to compare changes in the source of household income over time. That dataset showed, overall, that wages and salaries represent the single largest source of earnings for the average household. Today we tap into the Consumer Expenditure Survey (CES) dataset from the US Bureau of Labor Statistics to compare the very different sources of income for the lowest- and highest-earning households.

Our FRED graph above focuses on the lowest-earning 20% of households surveyed and tracks the percent of annual income (before taxes) from nine reported sources of earnings.*

For this group of households, social security and retirement income (the red area) has represented a larger share of earnings than wages and salaries (the blue area). Also, during most years between 1984 and 2014, the lowest-earning households reported negative income from self-employment (the light green area at the bottom of the graph). (The CES glossary defines self-employment income as the net difference between gross receipts and operating expenses, so it appears as if these households operate unincorporated businesses or farms at a loss. See this quick St. Louis Fed video for more background on recent self-employment trends.)

Our FRED graph below focuses on the highest-earning 20% of households surveyed and tracks the percent of annual income (before taxes) from the same nine reported sources of earnings.*

For this group of households, wages and salaries (the blue area) has represented the largest share of total earnings. Also, income from self-employment (the light green area at the bottom of the graph) wasn’t negative at any time between 1984 and the time of this writing.

What can explain these differences in the sources of income? The CES dataset reports several demographic characteristics of the households responding to the survey that can shed some light here. For example, the proportion of adults 65 and older in the lowest-earning households has historically been at least twice that recorded in the highest-earning households. Because the compostions of the two groups of households differ and include people at different stages of their earning lives, a direct comparison of their sources of income has its limitations.

* We removed the graph legends to leave more room for the data, but the legends are shown here for the first graph and here for the second graph.

How these graphs were created: In FRED, navigate the list of releases for the “Consumer Expenditure Surveys.” Next, click on “Tables by Different Characteristics” and select “Quintiles of Income Before Taxes.” Next, select “Lowest 20 Percent (1st to 20th Percentile)” and then click on “Income and Taxes.” Next, click on the boxes to the left of the nine data series names listed directly below the heading “Money Income Before Taxes.” Use the “Format” tab in the graph to change the graph type to “Area” and the stacking option to “Normal.” Repeat the steps described above to build the FRED graph for the “Highest 20 Percent (81st to 100th Percentile).”

Suggested by DeAndre Johnson and Diego Mendez-Carbajo.

The FRED Blog has previously examined the history of job openings data from the Bureau of Labor Statistics (BLS). In short, a job is considered open if a specific position exists and there is work available for it, the job can be started within 30 days, and there is active recruiting for the position. Job recruitment can take place online, and Indeed.com is one such internet-based job-posting platform whose data can be accessed in FRED.

The FRED graph above plots the recent changes in the monthly level of total nonfarm job openings reported by the BLS (the red line) and daily job postings reported by Indeed.com (the blue line) as a seven-day trailing average. Both data series are presented as an index with a value of 100 on February 1, 2020. That’s the start of the COVID-19-induced recession, the most recent large-scale shock to labor markets at the time of this writing.

The number of both job openings and job postings swiftly decreased during the economic contraction, but eventually bounced back to pre-recession levels about 12 months later. Both series generally move in sync, and changes in job postings tend to reflect overall changes in job openings. However, it’s important to keep in mind that the BLS samples around 21,000 business establishments to produce its data and that Indeed.com doesn’t report the actual number of job postings listed on its website.

While reports from Indeed.com are a useful sample of trends in job postings, data from job postings aggregators may not wholistically reflect market dynamics: Differences in data sample sizes and data-collection methodology, as well as self-selection bias in the publication of help-wanted ads, should be considered when drawing inferences about broad labor market conditions.

Comparing the average monthly changes in both series between February 2020 and the time of this writing reveals differences ranging from +26% to –11%. Over the past two years, the cumulative differences between both data series are relatively small; but significant differences can still be observed.

How this graph was created: Search FRED for and select “Job Openings: Total Nonfarm.” Next, click on the “Edit Graph” button and use the “Add Line” tab to search for and add “Job Postings on Indeed in the United States.” Next, click on the “Edit Line 2” tab, change the units to “Index (scale value to 100 for chosen date)” with 2020-02-01 as the index date.

Suggested by Kyla Duggal, Natasha Swindle, and Diego Mendez-Carbajo.

The economy is susceptible to shocks such as pandemics, natural disasters, and financial panics, which can have serious implications. And it can be challenging for economists to capture these economic effects in real-time, as many key economic indicators are released with a considerable lag. For example, real GDP for the second quarter of 2023 was released on July 27, 2023, about one month after the quarter ended. Unemployment rate data, from example, is more timely: Measures for the middle of July were released last week, on August 4, 2023.

The COVID-19 pandemic and the ensuing lockdowns had immediate impacts on the economy, and economists have turned to higher-frequency data released weekly or even daily. These high-frequency data releases have historically been limited to financial market metrics such as interest rates on bonds, commodity prices, and stock prices. These data provide useful information on financial conditions, but lack detail on the real economy (i.e., the state of households and businesses). Economists have begun to explore new data series, such as TSA passenger traffic, online restaurant bookings, and weekly employment directly from payroll companies. These data, however, often lack the long histories necessary to provide a clear economic signal. For example, a commonly cited real-time metric, the Weekly Economic Index, has been tracked only since 2009.

One notable exception is the Department of Labor’s weekly unemployment insurance (UI) claims data: They began reporting these data at a weekly frequency in 1945, and the series in FRED goes back to 1967. This long-standing timely reporting and straightforward interpretation make the UI report one of the most important indicators for monitoring economic activity.

Two key series from the UI report are initial claims and continued claims. Initial claims is the count of individuals filing for unemployment insurance benefits each week. While some claims may later be rejected, initial claims is still a good measure of the flow of individuals into unemployment. Continued claims is the stock of individuals who have received unemployment insurance benefits the prior week and have again filed for continued benefits.

While people can be unemployed and not receive UI benefits (such as recent graduates searching for their first job), UI data have historically provided an accurate picture of where the national unemployment rate is headed. The FRED graph above plots weekly continued claims for unemployment (blue line) and the unemployment rate (red line) starting in January 1967.

Notice in the graph that both continued claims and the unemployment rate rise at the start of recessions (gray shaded regions) and continue to rise throughout the duration of the recession. They only begin to fall after a recession has ended. During the most recent recession, continued claims reached their highest level during the week of May 9, 2020.

Data from the FRED graph below show signs of the economy gradually recovering from the effects of the pandemic. Both initial and continued claims for unemployment insurance benefits steadily declined to their pre-pandemic levels: about 200,000 initial claims for UI benefits and 1.8 million continued claims. As the labor market became increasingly tight during the summer and fall of 2022, workers found it much easier to quickly find new jobs. As a result, continued claims dropped to as low as 1.3 million. High-profile layoffs and slower job growth during late 2022 and early 2023 led to upticks in initial claims. This trend, along with a steady rise in continued claims, prompted concern that the US economy may dip into a recession. However, while initial claims for UI benefits may be slightly elevated from recent history, continued claims appear to have reached a peak in April 2023, again suggesting unemployed workers are having an easier time finding jobs. Keeping a close eye on the UI claims has been and continues to be one of the most effective ways gauge the state of the labor market in real-time.

How these graphs were created: For the first graph, search FRED for “Continued Claims (Insured Unemployment).” Keep the units as “Number” and the frequency as “Weekly, Ending Saturday.” Click “Add Line” and add the series “Unemployment Rate.” Keep the units as “Percent” and the frequency as “Monthly.” Finally, click “Format” and change the line axis position of “Line 2” to “Right.”
For the second graph, search FRED for “Continued Claims (Insured Unemployment).” Keep the units as “Number” and the frequency as “Weekly, Ending Saturday.” Click “Add Line” and add the series “Initial Claims.” Keep the units as “Number” and the frequency as “Weekly, Ending Saturday.” Click “Format” and change the line axis position of “Line 2” to “Right” and change the date range to show the previous three years of data.

Suggested by Charles Gascon and Sean McQuade.