According to the Bureau of Labor Statistics, surveys are the backbone of the federal statistical community. Response rates for federal surveys have been declining for many years. In summer 2019, response rates to JOLTS (the Job Openings and Labor Turnover Survey) averaged 58%. In summer 2023, those rates averaged 32%. The BLS and other federal statistical agencies are tackling this problem by changing their methods of data collection so that they continue meeting data quality standards, but data from the private sector may also be a useful complement.

Today we compare job posting data from Indeed (privately sourced) with job openings data from JOLTS (government sourced). The graph above shows that there’s a very close relationship between these two data sources.

One advantage of looking to private-sector Indeed data is that they are published much more frequently. Job postings on Indeed are updated weekly, while the equivalent JOLTS series is updated monthly. With increased frequency of data collection, there’s a timelier release of those data. For example, at the time of this writing, job postings data from Indeed are available all the way up to November 2023, while the similar JOLTS series has data only up to September 2023.

One advantage of the JOLTS dataset is that it reports on employment, layoffs, and quits, with the intent of covering the entire economy, and thus a more accurate picture of the economy. There’s no equivalent in the Indeed dataset, which is limited to its own listings and a few other sources.

Complementing governmentally sourced data with related private-sector data can add to overall data credibility and trust, which is a major part of FRED’s mission.

How this graph was created: in FRED, search for and select the series “Job Postings on Indeed in the United States” (IHLIDXUS). From the Edit Graph panel, use the Edit Line option to set units to “Index Feb, 1 2020=100” and frequency to “Daily, 7-Day.” Use the Add Line option to add the “JTSJOL” series and set its units to a custom index and select the date for the index base to be 2020-02-01. Set date range to 2020-01-31 to present.

Suggested by Alexander Bick and Kevin Bloodworth.

The national economic accounts are the foundation for calculating, among other things, US gross domestic product. The US Bureau of Economic Analysis, which collects and releases GDP data, periodically updates these national economic accounts. On September 28, the BEA released its 2023 comprehensive (or benchmark) update.

Now, FRED has these new data. So, today we turn to Archival FRED (ALFRED) to compare these new data with the old data. The ALFRED graph above shows two different vintages of annual real GDP data between 2017 and 2022. The vintage dates included in the series names are associated with the dates of the comprehensive updates: The red bars represent data from the 2023 update, the latest available at the time of this writing; and the blue bars represent data from the previous comprehensive update, released in 2018.

There are relatively small differences in the annual growth rates of inflation-adjusted GDP between data vintages because comprehensive updates involve changes in the underlying methodology used to measure economic activity. As outlined in this BEA briefing, the latest update also synced the schedule for releasing national GDP and related industry and state statistics within the same timeframe. Lastly, the reference year for inflation adjustment and price measures has been updated to 2017 from 2012.

Additional information about the scope and impact of the 2023 comprehensive updates to the national, industry, and state economic accounts can be found here.

How this graph was created: Search ALFRED for “Real Gross Domestic Product.” By default, ALFRED shows a graph with two sets of bars: the most recent vintage and the prior vintage. Add additional vintages by using the “Add Line” tab and select the date of the desired vintage from the “or select a vintage” dropdown menu. Change the start date and the end date above the graph to customize the number of data points shown.

Suggested by Diego Mendez-Carbajo.