Federal Reserve Economic Data

The FRED® Blog

How to read Indeed job posting data

This blog post is obsolete. Indeed changed the methodology for its data in January 2021.

Tracking the availability of new jobs is no easy task. But FRED recently added online job postings data from Indeed. These data are presented in an interesting way, so some explanation is in order.

First, the data cover a 7-day moving average of job postings on Indeed.com as well as other online platforms. Indeed makes every effort to remove duplicate job postings from these counts, but doesn’t include job postings that are not found online. The proportion of online postings has been steadily increasing, and this brings us to our second point.

If you measure only some of the job postings—in this case, online only—and know that this proportion is increasing, it’s unrealistic to compare the data from previous years with the data from the current year without making any adjustments. An unadjusted measurement would likely systematically show an increase every year simply because the proportion of online postings is increasing.

So one needs to reset every year of statistics, and that’s what happens with this dataset: Every February 1 is set to a value of 100 for every year of data. This adjustment allows us to see how the postings are evolving.

For example, in the graph above, we see that on April 6, 2020, the 7-day moving average of new job postings was at –51.2%. This means that postings on April 6, 2020, as compared with February 1, 2020, were 51.2% lower than postings on April 6, 2019, as compared with February 1, 2019. But this figure of 51.2% doesn’t include any changes that may have happened between 2019 and 2020.

Despite our best efforts here, we know this may still be a bit confusing. So, a short explanation is that the data are useful for looking at patterns within the year, but not as useful for looking at patterns across years. Of course, 2020 has been pretty special, with all its dramatic changes, which show up in the graph. At present, we have less than a year of data, so we’ll need to wait a bit for all those interesting yearly patterns to show up. As with any FRED data series, you can check in at any time to see what’s new.

How this graph was created: Start from the release table, check the series you wish to display, and click “Add to Graph.”

Suggested by Christian Zimmermann.

View on FRED, series used in this post: IHLCHGNEWUS, IHLCHGUS

The impact of COVID-19 on U.S. states’ economic activity

State-level GDP data show the second quarter was much worse than the first

GDP comes in various forms—for the nation as a whole and also for individual U.S. states. The map above shows the change in real GDP in each U.S. state for the first quarter of 2020. This was at the start of the pandemic, and some states were hit hard.

The worst declines were in Louisiana (-11.91%), Delaware (-11.43%), Wyoming (-10.53%), Hawaii (-8.92%), Wisconsin (-8.76%), South Carolina (-8.24%), and Michigan (-7.94%). The hope at the time was that this slump would be temporary. We now have the data for the second quarter:

While the colors of the map are similar, the actual values in the second quarter are much worse. To put this in perspective, consider that the decline in the worst state in the first-quarter map (-11.91% in Louisiana) was nowhere near the decline in the best state in the second-quarter map (-21.91% in Delaware). The numbers are astounding, with Hawaii, Nevada, and Tennessee losing over 40% of their economic activity.

To see more details and find your home state, click the link below each map in this post to reach the interactive maps on GeoFRED.

How these maps were created: The original post referenced interactive maps from our now discontinued GeoFRED site. The revised post provides replacement maps from FRED’s new mapping tool. To create FRED maps, go to the data series page in question and look for the green “VIEW MAP” button at the top right of the graph. See this post for instructions to edit a FRED map. Only series with a green map button can be mapped.

Suggested by Christian Zimmermann.

Employment losses are largest for the least educated

The FRED Blog has discussed how unemployment rates are inversely related to educational attainment and how they change during recessions. In short: Workers with more education are richer in so-called human capital and tend to be able to adapt more easily to changes in large-scale labor market conditions.

The FRED graph above shows employment levels after the COVID-19-related recession began. The length of the bars represents the percent change,  relative to a year ago, in the number of employed people 25 years and older. And these workers are divided into groups according to educational attainment.

Workers who didn’t graduate from high school had the largest losses in employment. Workers who did graduate from high school, including those  with some college or an associate degree, also experienced significant losses in employment but fared a bit better. Workers with a bachelor’s degree or higher were able, for the most part, to remain employed.

The second FRED graph shows the same four groups of workers but for the previous recession, from December 2007 to June 2009. Although these bars don’t go as far into negative territory, we see a similar pattern: At least initially, the more-educated labor force was more resilient. As the recession passed the 12-month mark, however, all education groups started to report losses in employment.

Low educational attainment isn’t necessarily a permanent trait, so it’s possible for workers who are laid off to exit the labor force, gain more human capital through formal education, and re-enter the labor force as more-educated workers. When they do this, they can expect to enjoy steadier employment. To learn more about education’s effects on employment stability, read the work of Isabel Cairo and Tomaz Cajner.

How these graphs were created: Start from Table A-4 of the Current Population Survey, select the series you want shown, and click “Add to Graph.” From the “Edit Graph” panel, select units “Percent Change from Year Ago” and click on “Copy to all.” From the “Format” tab, select “Graph type: Bar.” Adjust the sample period to match the dates displayed in each graph.

Suggested by Diego Mendez-Carbajo.

View on FRED, series used in this post: LNS12027659, LNS12027660, LNS12027662, LNS12027689


Back to Top