Phantom Employment: How Employment Figures Were Inflated in the Biden Era

Employment Rates in the United States: Revision, Corrections, and the Dispute Over “Artificial” Jobs

Employment statistics are—or should be—one of the most rigorous public metrics and least prone to political rumors: governments, businesses, and markets make decisions based on these numbers. However, recent years have exposed vulnerabilities in the United States’ employment measurement system that transform what should be a technical matter into a matter of enormous political consequence. Two recent events illustrate the problem: massive corrections to payroll data and debates over whether part of the reported growth was actually the result of model estimates that overestimated actual job creation.

What happened to the employment figures?

The Bureau of Labor Statistics (BLS) announced significant revisions to its employment data covering recent periods: preliminary estimates and benchmark revisions led to a downward adjustment of the net number of jobs created in a year by hundreds of thousands, with public estimates indicating job cuts of nearly 911,000 fewer jobs than originally reported for the period April 2024–March 2025. These corrections result from comparing monthly surveys with the most comprehensive administrative records—the Quarterly Census of Employment and Wages (QCEW)—and force a recalculation of what was thought to be the evolution of the labor market.

This is not the first time the BLS has adjusted figures: previous revisions had already counted hundreds of thousands of fewer jobs (for example, one revision that left 818,000 fewer jobs relative to previous estimates). These corrections are often explained by the fact that monthly surveys rely on estimates and models that can only later be validated against administrative accounting.

The Role of the “Birth-and-Death Model” and Why It Generates Controversy

A key technical element is the so-called birth-and-death model, which the BLS uses to estimate jobs resulting from new or closing firms not covered by the monthly survey sample. When the economy changes rapidly—due to crises, booms, or structural transformations—this model can overreact or underreact, producing a material difference between the real-time estimate and what administrative records subsequently show. In recent adjustments, several analysts have pointed out that the model may have overestimated job creation in certain sectors, contributing to the subsequent large cuts.

Political critics and some columnists have gone further by describing these reported jobs as “artificial” or “inorganic,” suggesting that statistical methods and, in some cases, administrative intervention artificially inflated the figures during politically sensitive periods. This interpretation, however, conflates two things: on the one hand, there is a technical problem (estimates that need revision); on the other, there is a political interpretation that attributes deliberate intent. To support the second claim, direct evidence of political manipulation of data is needed, something that has not been publicly proven so far, but which is strange because it always favors a political group. The debate has included everyone from critical legislative offices to academic economists who call for strengthening the resources and independence of the BLS.

From the initial figures to the revised figures: a comparison

Comparing the figures reported month by month with those revised after cross-checking with administrative records, significant differences emerge. A reasonable summary of the data set would be:

Initial figures (monthly surveys): They showed a higher pace of job creation in different parts of the 2023–2025 period, with sectors such as professional services, leisure/hospitality, and trade reporting strong progress.

Revised figures (benchmark and QCEW): They cut hundreds of thousands of jobs overall; in some months, the average annual growth rate slowed significantly. An estimated job cut of close to 818,000 (in a previous revision) and the estimate to which media and analysts have most recently reacted—almost 911,000 fewer jobs over 12 months in the latest announcement—show the magnitude of the adjustment.

This gap does not imply that all the jobs initially reported “did not exist” in absolute terms. In many cases, it simply reflects that the estimation methodologies, subject to sampling noise and assumptions about business creation, produced a cumulative overestimate. But for the public and voters, the difference has practical consequences: the perception of a strong economy can depend on those monthly headlines, even if the data is later severely adjusted downward.

Was there political intent behind the figures?

This is where the discussion becomes entirely political. Some actors and media outlets have argued that the corrections were necessary because the data were, in practice, inflated to present a better image of the outgoing administration—a serious accusation if proven to be the case. Critical notes have been published by some legislators’ offices, for example, linking the narrative of “maximum job creation” to institutional messages from the White House. At the same time, left-wing economists and partisan think tanks have argued that what happened is, in essence, a technical and methodological failure that requires more resources and transparency from the BLS, and that it was most likely an organized conspiracy.

It is also important to emphasize that the data review process is public and documented: the corrections are accompanied by technical notes and explanations about the seasonal adjustments, which can easily be manipulated since they are always estimates, and the benchmark is based on administrative records. This does not rule out malpractice, but it does require that accusations of manipulation be accompanied by concrete evidence—for example, internal communications or political instructions—which until now have not been made public to a standard that conclusively establishes intentional manipulation.

What does this mean for politics and elections?

The economy, and in particular the labor market, are central issues in elections. In this context, the narrative of jobs “created” or “lost” can be used for communication purposes by parties and candidates. Therefore, it is not surprising that political and media circles are asking the question many citizens are asking: to what extent did these data help build a favorable economic image during election periods? Direct hypotheses and forceful assertions can be read on social media and in columns; one of the versions circulating verbatim in certain forums, which we reproduce due to its relevance to the public debate, is: “Just as they did to try to give Kamala Harris a positive image during the election.” This phrase reflects a political suspicion that, I repeat, requires additional evidence to establish it as fact.

Conclusion and Journalistic Recommendations

Separate technique from policy. There are clear technical reasons (birth-and-death model, benchmark adjustments) to explain why the numbers change. Of course, techniques based on estimates clearly lend themselves to manipulation, although confusing this with intentional manipulation may not be accurate.

Demand more transparency and resources. The BLS does not need stable resources or autonomy to improve its samples, reduce error, and communicate methodological limitations so that the public understands what preliminary data can be reported and what only definitive data can be responsibly reported. A variety of voices, from economists to congressional offices, have not called for exactly this.

Monitor the political use of data. Parties and officials will always use favorable figures; journalistic work in this context consists of contextualizing and comparing metrics, which is why preliminary data with administrative reviews should not be allowed.

Read headlines and official releases with caution. Monthly figures are useful but not infallible; The full story comes after the benchmark with the administrative payrolls.

The recent adjustments to employment indices reveal flaws in real-time estimates that may have created misperceptions about the strength of the labor market. This fuels political perceptions—and accusations of “artificial jobs”—that should be rigorously investigated. Distinguishing between technical errors, misinterpretations, and intentional manipulation is a central task for journalism and society in a context where economic numbers weigh so heavily on public opinion and at the ballot box.

Month-by-month comparison: initial figures vs. revised figures

Below is an illustrative table with actual data published by the BLS, reflecting how in certain months there have been significant downward revisions (fewer jobs than initially reported).

Month/Year Initial Job Creation Estimate Revision (Total Jobs Fewer) Source and Context
January 2025 +125,000 jobs Revision: –14,000 → Final: +111,000
Bureau of Labor Statistics

February 2025 +151,000 jobs Revision: –34,000 → Final: +117,000
Bureau of Labor Statistics

March 2025 +185,000 jobs Revision: –65,000 → Final: +120,000
Bureau of Labor Statistics

May 2025 +139,000 jobs (preliminary) Revision: –120,000 → Final: +19,000
KCRA

June 2025 +147,000 jobs (Preliminary) Revision: –133,000 → Final: +14,000
KCRA

April 2024–March 2025 (annual period) ≈ +147,000 average jobs per month

April 2024–March 2025 (annual period) ≈ +147,000 average jobs per month Cumulative revision: –911,000 jobs (~–76,000/month) → Final average: ≈ +71,000/month
The Real Economy Blog