Indeed job postings correlate strongly with total job openings from the Bureau of Labor Statistics’ Job Openings and Labor Turnover Survey (JOLTS). JOLTS job openings tend to fall in recessions and rise in expansionary periods.

The Indeed Hiring Lab, the research arm of job search website Indeed.com, publishes a job postings data index that aggregates a complete count of all job postings on the Indeed website. This includes a measure of new job postings, which only counts job postings the first time they are visible. The data—available starting in February 2020—are collected daily and reported as a seven-day average. For this comparison, we take monthly averages of the seasonally adjusted daily data.

In the figure, we show that Indeed new job postings track closely with job openings from JOLTS. The correlation between these measures between February 2020 and August 2025 is quite high (0.95).

While Indeed data are only available for the last five years, JOLTS data go back to 2000 and show a clear relationship between job openings and business cycles. Job openings tend to fall in recessions and rise in expansionary periods. That relationship is clearest in the 2001 and 2007 recessions and expansions, where a fall in job openings preceded the start of the recession.

In the wake of the government shutdown, the Indeed new job postings can give use some information on the state of new positions needed by employers. Moving forward, a strong and sustained divergence of trends between the Indeed and BLS measures could provide a worrying signal of degraded data quality or integrity in BLS reports.

Monthly changes in ADP employment levels tend to track changes in private-sector payroll employment from the BLS Current Employment Statistics, and both measures show that private-sector employment growth falls in recessions and rises in expansionary periods.

The ADP National Employment Report is a monthly measure of the private-sector labor market based on aggregated payroll data of more than 26 million U.S. workers. ADP Research, a research arm of the ADP payroll processing firm, releases the report monthly. While the report provides much detail of employment by firm characteristics, the topline number in the report is private-sector employment changes, available since January 2010.

ADP employment changes track closely with the Bureau of Labor Statistics (BLS) measure of private-sector payroll employment in the Current Employment Statistics survey, data that is published every month as part of the Employment Situation Summary, also known as jobs day. In the figure, we compare seasonally adjusted monthly changes in employment for each measure smoothed to three-month moving averages. These smoothed changes are highly correlated (0.76).

While ADP data are only available as of 2010, BLS private-sector employment data goes back to 1938 and shows a clear relationship between the number of jobs and business cycles. Employment falls in recessions and rises in expansionary periods. The close relationship between these measures allows us to track any notable and sustained divergence, which would indicate a concern about the quality of government data.

Revelio Labs generates monthly U.S. employment estimates by counting social networking profiles on sites like LinkedIn. Total nonfarm employment changes from Revelio generally track BLS-based estimates.

Seasonally adjusted and nonseasonally adjusted estimates from Revelio are released monthly by industry, occupation, and geographic region. National data are available as of 2021. The data are normally published the day before the BLS employment report is released. Historically, Revelio’s monthly change usually falls below the BLS estimate, but over relatively short periods of time, the data series do tend to rise and fall together.

The Challenger job cut report data are somewhat noisy on a month-to-month basis, but spike noticeably during recessions.

Challenger, Gray & Christmas collects its job cut data by tracking public announcements made by U.S. companies in both the private and public sectors. The job cuts can include cuts in multinational plants (i.e., outside the United States). The report has been published monthly since 1994 and is typically released at the end of every month or the first week of the following month. The figure shows that large increases in announced job cuts occur in recessions (and when the unemployment rate spikes).

Of note, the spike in layoffs in spring 2025 was partially driven by announced layoffs in the federal government.

Monthly changes in unemployment insurance claims track the unemployment rate and provide a useful indication of recessionary periods.

While the U.S. Department of Labor (DOL) aggregates unemployment insurance (UI) claims data and is the source of the national data, it relies on administrative data processed at the state level that are generally posted on state government websites, potentially allowing a real-time accuracy check if concerns are raised about the accuracy of data released through the DOL portal. During the shutdown, the number of initial and continued claims has been updated weekly on the 539 report, a DOL-compiled dataset based on information submitted by state unemployment insurance offices. Right now, the data are only available on a nonseasonally adjusted basis because of the shutdown. The figure displays continued (or insured) UI claims and the unemployment rate, both as 12-month moving averages to remove some volatility and seasonality.

Continued UI claims closely track the unemployment rate: They spike during recessions and fall during economic recoveries. The close relationship between UI claims and the unemployment rate allows us to track any notable and sustained divergence, which could indicate a concern about the quality of government data. More information on the UI claims data, including initial claims, federal claims, and state-specific data, is updated weekly on EPI’s UI claims page.

Google Trends provides real-time data on the relative frequency of search terms for a given keyword, category of keywords, or a topic. With the right search terms, we can construct an index that tracks the overall unemployment rate

Google Trends is a publicly available database maintained by Google that provides Search Volume Indices. Search Volume Indices are a measure from 0–100 of the relative search intensity (number of searches for a given keyword divided by total searches) by geographic location and period. The data are updated in real time and are available from 2004 onward.

Google Trends data can be benchmarked to several different economic indicators. In fact, academics have used Google Trends data to “nowcast” GDP, private consumption, unemployment, recessions, and health outcomes. The figure shows the monthly average of select labor market-related search terms and the unemployment rate. The Google Trends relative search frequency for these terms tracks the unemployment rate and spikes during recessions. The close relationship between these measures allows us to track any notable and sustained divergence. Any prolonged increase in the search frequency of these select search terms could indicate a recessionary period.

The economic policy uncertainty (EPU) index clearly responds to business cycle downturns—spiking sharply during the recessions of the early 1990s, early 2000s, late 2000s, and the COVID-19 pandemic.

The EPU index is calculated by Baker, Bloom, and Davis, and their full methodology can be found here. The index aggregates information from three basic components: search results from 10 large national newspapers measuring the volume of news articles discussing economic policy uncertainty; the number of federal tax code provisions set to expire over the next year; and the degree of disagreement among economic forecasters about future levels of key economic variables.

While the EPU is clearly cyclical, it does rise during some non-recessionary periods, most notably in the early 2000s and from 2011–-2013. The early 2000s increase is almost surely driven by the Iraq War. From 2011–2013, a rolling series of economic crises in the Eurozone, as well as short-term extensions of expiring provisions from the tax cuts passed in the first George W. Bush administration explain the spikes. The very large spike in early 2025 was related to the “Liberation Day” tariff announcements.

Sharp falls in consumer sentiment are often followed shortly by recessions.

The University of Michigan consumer sentiment index has been calculated since 1960 (and has been calculated monthly since 1978). The index is derived from answers to five survey questions that have categorical answers (that is, their questions provide survey respondents with two to three possible responses to choose from, for example, “Would you say that your (and your family) are better off, or worse off financially than you were a year ago?”).

For each question, the University of Michigan researchers calculate a relative score for answers that subtracts the unfavorable responses from favorable responses. The sum of responses is then compared with a base value from 1966.

The index reached lows in 2022 that were comparable with most recessionary periods, despite a very strong economy. The clear explanation for that was the sharp inflation spike in late 2021 through mid-2022. Early 2025 saw similarly low measures, likely driven by concerns over the potential effects of tariffs.

Because the 1966 base year was a long time ago, it seems fair to ask if there are structural changes that might reliably change the level of consumer sentiment over time. For example, if rising inequality (or anything else) made U.S. households consistently less happy about their relative economic situation over time, they might generally have a lower “baseline” level of favorable sentiment.

To assess this, and to evaluate whether the University of Michigan consumer sentiment index might be useful for assessing the broader health of the economy, the figure below shows changes in the level of consumer sentiment and changes in inflation-adjusted personal consumption expenditures (the broadest level of household spending). The theory is that consumers’ sentiment about the economy should correlate positively with their actual spending patterns. The pattern is clear: Changes in consumer sentiment do indeed coincide tightly with changes in consumer spending.

Data Accountability Dashboard

The Chicago Federal Reserve flow-consistent unemployment rate (FCR) combines several sources of labor market data to predict monthly unemployment rates. The Chicago Fed FCR closely predicts the official unemployment rate published by the Bureau of Labor Statistics (BLS).

Twice a month, the Chicago Federal Reserve publishes the Real-Time Unemployment Rate Forecast, which predicts the BLS unemployment rate in advance of its official release. The key ingredient in this prediction is the Chicago Fed’s flow-consistent unemployment rate, which is in turn a prediction of Current Population Survey job finding and separation rates using real-time data like UI claims; Google Trends index for unemployment; Bloomberg consensus unemployment rate forecasts; Indeed and Lightcast job openings; ADP employment levels; weekly Morning Consult unemployment and job search activity; Conference Board labor market differentials; and JOLTS layoffs, discharges, and hiring rates. As a result, the Chicago Fed labor market indicators are partially based on U.S. government data, but they can be extended even in the absence of government data releases.