Public and Private Labor Market Data: Insights From the Government Shutdown

Private datasets present opportunities and pitfalls in economic analysis. Last year's federal government shutdown demonstrated the value of private sector data sources but also highlighted the difficulty in interpreting them and reconciling potential differences between them and the official sources.

In general, private data excel at capturing real-time, granular behavior, while government surveys (despite their size and timeliness limitations) provide systematic coverage of the entire economy. The most rigorous economic analysis treats these sources not as competitors but as complementary, but the devil is in the details. In this article, I discuss these issues in two specific contexts:

• A dataset used for researching online job search behavior of low-wage workers
• Labor market data sources that became popular during the government shutdown

The Advantages and Disadvantages of Private Data Sources: a Case-Study Using Data on Job Search

Advantages of Private Datasets

Large private datasets offer something government statistics cannot: real-time, large-sample observations of economic behavior. When the 2025 shutdown occurred, government agencies could not update labor statistics. Private data sources, though, continued without interruption.

The advantages of private data sources extend beyond continuity during government lapses. Because of their size and level of granularity, private data sources often capture detailed economic behavior that official sources can only approximate in smaller samples.

Consider data from a large job search online platform (JSOP) — containing over 130 million job applications from all 50 states, D.C., Puerto Rico and Guam — for the period January 2020-November 2025. I compare these data directly to the Job Openings and Labor Turnover Survey (JOLTS), a product of the Bureau of Labor Statistics (BLS) aimed at counting how many open jobs exist in the U.S. economy every month.

In JSOP, I observe not just employers' job postings but also all applications submitted at a daily frequency, with precise geolocation data for both the jobs and the workers as well as very detailed occupational classifications. These details are unavailable in public data sources on vacancies like JOLTS.

Challenges of Private Datasets

But these advantages come with complications. Validation and benchmarking to representative samples is key, but comparison with official data is far from straightforward. Specific to labor market data and research, most data sources rely on information from only one platform. For instance, in the case of JSOP, individuals may simultaneously apply on other job search sites. Thus, my data, however granular, is a partial and selected description of workers' search behavior.

An example illustrates the depth of the problem: Individuals might have, say, five applications in JSOP over five years (a relatively low number), but they could have submitted many more applications across other platforms. Or they might stop applying in JSOP not because they stopped searching, but because they migrated to a different platform after finding JSOP not productive.

We cannot observe behavior across platforms or adjust for selection into this particular platform. An individual's decision to use this platform versus others likely depends on factors we cannot measure — such as prior experience, network effects or awareness — and, worse, these factors affect workers' search behavior.

Even if one had data from all platforms, the bias would persist. It is not only that official sources are meant to represent the aggregate of the U.S. labor market — while private sources necessarily represent a sliver of that market — but the two sources often measure different objects. For instance, JOLTS omits job postings that cannot start within 30 days of the recruiting activity. On the other hand, data from online platforms allow us to measure all open jobs regardless of their expected fill date, but I only observe job postings that have attracted at least one application. Which one is the correct definition of a job opening? Neither, of course, and both. But economists and policymakers need to be aware of the differences.

Additional Differences Between Public and Private Datasets

These challenges point to broader lessons about using large private datasets for economic analysis. First, transparency about measurement matters. In my example, JSOP is a timely and granular data source that represents platform-mediated job search and job posting but does not necessarily reflect the aggregate state of the labor market. Studying how behavior changes on different platforms under different economic environments and conditions helps in forming a more complete picture of the economy, and public data sources play a vital role in benchmarking this exercise.

Second, different sources measure different constructs. Private data sources are generally more responsive to technological change affecting the definitions of economic variables. Public sources guarantee the best comparability across time. JSOP's algorithm changes in response to employers' innovations in recruiting. Thus, these changes produce a modern product that is not entirely comparable across years. On the other hand, JOLTS reflects hiring constraints from its inception in the early 2000s, such as the 30-day-to-hire restriction, which made more sense in an era when candidate screening was more time intensive. JOLTS has not changed over the years as much as the economy has, which means we can interpret differences across time as differences in economic fundamentals, not differences in how often the software is updated.

Private Data's Role When Public Data Are Not Available

These measurement challenges — platform selection, definitional differences and benchmarking limitations — proved especially consequential during the recent lapse in government appropriations that halted the collection and publication of economic data at most public agencies in fall 2025. The 43-day federal government shutdown that began Oct. 1 suspended the release of official BLS employment data, prompting researchers, policymakers and market participants to rely on private-sector alternatives for real-time labor market monitoring. Three prominent data sources — each with different methodology, coverage and academic validation — emerged as substitutes during this period:

• ADP National Employment Report
• Revelio Labs Public Labor Statistics (RPLS)
• Gusto Small Business Data

ADP National Employment Report

The ADP National Employment Report — produced by the ADP Research Institute in collaboration with the Stanford Digital Economy Lab — derives employment and wage measures from administrative payroll records rather than surveys. ADP processes payroll for about 26 million U.S. workers across more than 500,000 business establishments, representing roughly one-fifth of the private-sector workforce (ADP Research, 2025).¹ The methodology relies on anonymized and aggregated payroll transaction data, capturing both the timing and amount of worker compensation, with employment counts benchmarked annually to the BLS's Quarterly Census of Employment and Wages (QCEW), which covers over 95 percent of U.S. employment.²

ADP microdata have been used in peer-reviewed academic research, including a 2021 paper documenting new facts about nominal wage rigidity³ and a 2022 book chapter demonstrating that combining ADP and BLS data reduces measurement error in real-time employment estimation.⁴ However, a 2025 analysis during the shutdown found only modest correlation between monthly ADP and BLS employment changes, noting that methodological differences in coverage and revision timing complicate direct comparisons.⁵

Revelio Labs Public Labor Statistics

RPLS takes a fundamentally different approach, constructing workforce measures by aggregating and standardizing over 100 million U.S. professional profiles collected from online platforms. This method covers about two-thirds of employed individuals, which is substantially larger than the BLS establishment survey's 27 percent coverage or the household survey's 0.03 percent sample.⁶ The methodology involves collecting publicly available data from professional networking sites, job postings aggregators, employee sentiment reviews and WARN Act layoff notices, then applying machine learning models to standardize job titles, skills, company mappings and demographic characteristics.

RPLS data are distributed academically through the Wharton Research Data Services (WRDS) platform and have been employed in recent NBER working papers, including a 2023 paper on generative AI and firm values, which mapped occupational exposure to AI using RPLS firm-level employment composition data.⁷

During the October 2025 shutdown, RPLS estimates diverged notably from ADP's: While ADP reported a decline of 32,000 jobs in September 2025, RPLS estimated the economy added 60,000 jobs, highlighting the methodological differences between payroll-based and profile-based measurement approaches.⁸

Gusto Small Business Data

Gusto Small Business Data provides real-time insights specifically focused on firms with fewer than 50 employees, a segment systematically underrepresented in both ADP's coverage (which historically has emphasized firms with more than 50 employees) and traditional BLS surveys. Gusto processes payroll for over 400,000 U.S. small businesses, and the company's research team produces monthly reports tracking hiring rates, termination rates (both voluntary and involuntary since January 2020), average hourly earnings and hours worked. The methodology involves calculating hiring rates as the number of employees hired in a given month as a percentage of total workers employed, with termination rates similarly computed. The inaugural Gusto Small Business Jobs Report, launched in November 2025 during the shutdown, showed that small businesses experienced net job losses of 5,900 positions in October 2025.⁹

Limitations of Private Data Sources

Several important caveats apply to these private data sources as substitutes for official statistics. A 2025 article emphasized that, while private firms can estimate overall employment from payroll records or job postings, they struggle to measure unemployment rates and other household survey indicators—including labor force participation.¹⁰ This is because companies that observe who is on payroll or searching for a job have difficulty identifying individuals who are not working, not searching or have left the labor force entirely.

A 2025 FAQ noted that private estimates are byproducts of business activities rather than purpose-built statistical programs.¹¹ They also lack the methodological consistency and transparency of official statistics and may change access, methods or pricing without notice. Moreover, many private estimates rely on BLS data for benchmarking, validation and weighting, creating dependency on the very official statistics they seek to replace.

Despite these limitations, the 2025 shutdown demonstrated that private labor market data — particularly when multiple sources are triangulated — can provide essential (if imperfect) bridges for economic monitoring during disruptions to federal statistical operations.

Conclusion

The fall 2025 government shutdown offered a natural experiment in the strengths and limitations of private labor market data. Two lessons emerge from the analysis. First, private data sources are valuable complements to (but not substitutes for) official government statistics. The granularity and timeliness of platforms like JSOP, ADP, RPLS and Gusto enable research questions that official surveys cannot address, from real-time tracking of job search behavior to high-frequency employment monitoring during data blackouts. Yet each source captures only a partial view of the labor market, shaped by platform-specific selection, definitional choices and coverage gaps that differ fundamentally from the representative sampling frameworks underlying BLS surveys.

Second, the integrity of economic measurement depends on maintaining robust public statistical infrastructure. Private data sources often rely on official statistics for benchmarking and validation. Thus, when those official sources go dark as they did during the shutdown, private estimates lose their anchor. The divergence between ADP and RPLS estimates in September 2025 — one showing job losses, the other job gains — illustrates the interpretive challenges that arise when no authoritative reference point exists. For researchers and policymakers alike, the episode underscores that investment in federal statistical agencies is not merely a matter of government bookkeeping but a foundation for informed economic decision-making across the public and private sectors.

Claudia Macaluso is a senior research economist in the Research Department at the Federal Reserve Bank of Richmond.

To cite this Economic Brief, please use the following format: Macaluso, Claudia. (February 2026) "Public and Private Labor Market Data: Insights From the Government Shutdown." Federal Reserve Bank of Richmond Economic Brief, No. 26-04.

This article may be photocopied or reprinted in its entirety. Please credit the author, source, and the Federal Reserve Bank of Richmond and include the italicized statement below.

Views expressed in this article are those of the author and not necessarily those of the Federal Reserve Bank of Richmond or the Federal Reserve System.