The US economy grew at a 3.8% rate in the second quarter, significantly stronger than previously reported

[Wiz Imp]

47. The process makes it virtually impossible.

As someone extremely familiar with the process, I'm going to copy the summary from Google's AI overview. In this case, I can attest to it being completely accurate. This overview summarizes it more clearly and succinctly than I could do myself.

The commissioner of the Bureau of Labor Statistics (BLS) does not even see the jobs numbers before they are finalized. The process is intentionally designed to keep the commissioner, who is a political appointee, at arm's length from the data collection and estimation process to protect the integrity of the statistics.

Key details about the process:

Final numbers are locked in The commissioner is typically briefed on the final numbers on the Wednesday afternoon before the public release on Friday. By this point, the data is prepared, finalized, and locked into the system.

No role in estimation The commissioner plays no role in estimating or manipulating the figures. Their involvement is limited to reviewing the accompanying narrative for the public release, not altering the underlying data.

Internal safeguards The BLS maintains a strong culture of independence and has systems in place to prevent any single person, including the commissioner, from interfering with the data. Former commissioners have stated it would be impossible to manipulate the data without being detected, which would likely trigger whistleblowers and resignations.

Career staff handles data The numbers are collected and processed by nonpartisan career civil servants who have no connection to the political administration.

These safeguards came under scrutiny in August 2025 when then-President Donald Trump fired Commissioner Erika McEntarfer after a weak jobs report and baselessly alleged the numbers were "rigged". However, multiple former BLS officials, including those appointed by Trump, rejected the idea that the commissioner could manipulate the data.