[nindalf]

What?

> America’s economy added 130,000 jobs in January, almost double the number that analysts had been expecting, indicating that the labour market might have picked up after months of apparent stagnation. Unemployment also fell to 4.3%, a slight dip from the previous month. The figures may put off the Federal Reserve from lowering interest rates as quickly as Donald Trump would like.

[rybosworld]

Yeah not sure where the 600k figure is coming from. But that 130k that's being reported is almost certainly not correct, and will be revised down.

The ADP is only reporting 22k for January. Which lines up very closely with the monthly pace of job creation for 2025 (~15k, on average).

The downward revisions on the BLS numbers for 2025 are the largest on record. There's a sentiment that the monthly numbers were purposely inflated for 2025.

[mandevil]

> There's a sentiment that the monthly numbers were purposely inflated for 2025.

Jobs data (which comes from a survey of employers) is always subject to revisions, and the revisions have gotten worse. But it's not necessarily political. Back when Trump unprecedentedly fired Erika McEntarfer (former head of the BLS) I looked into this and tried to understand where the data comes from, how it works.

Basically, the BLS surveys 50k out of the several million companies in the US. But response rates have dropped rather strongly since the Pandemic in particular. If you give them 3 months, it gets up to 90%, but after 1 month it is only 50%. This would be okay, except that the response rates turn out to be skewed: large companies almost always respond quickly, small companies respond more slowly. And small companies tend to be much more responsive to the state of the overall economy than large companies- for good and for bad. And even the response times of the small companies turn out to be sensitive to economic trends, for good and for bad: the companies that respond more slowly to surveys, it turns out, tend to be the most sensitive to the broader economy, the first to fire in a recession, but also the first to hire when the economy is getting better.

The final thing is that these companies can be highly correlated with each other and the broader economy. So there are always corrections, but in placid economic weather the corrections cancel out- some of the extrapolations are high, some are low, it ends up being close to 0. And the financial markets understand how to read these corrections. They understand that when the economy is shifting you will get all of the misses in the same direction- either high or low- and to pay attention these corrections data.

All of these misses low in 2025 are a very bad sign for the economy overall, not necessarily a sign of political pressure. That would probably show up in large, sustained differences to the ADP numbers or other privately reported numbers, rather than revisions announced by the BLS itself.

Both Trump I's appointee to head the BLS, William Beach (who served from 2019-2023) and Biden's appointee Erika McEntarfer (whom Trump fired in 2025) wanted to modernize the survey system, make it easier for companies to respond, and get better and faster data collection. But, because this system is so incredibly important to the markets, they wanted a larger budget to run the new system and the old system together for a significant portion of time (several years) so that everyone would be comfortable understanding the intricacies of the system, what revisions would mean, etc. And since they didn't get that budget (in fact it got cut) they decided to prioritize the old system rather than throw it away and experiment with something different.

[jeffbee]

Good explanation. I think it's sad that people find it easier to adopt imaginary theory of why labor statistics are flawed than they are to simply go read their methodology. People don't want to believe that there are BLS statisticians working the phones and knocking on doors. They don't want to know that there are hundreds of USDA employees tasked with writing down the retail price of carrots in every major American city every day. They aren't interested in how Census takers judge the population by counting gas and electric meters. They'd rather believe that bureaucrats are evil and incompetent.

[mandevil]

I think it's the rise of authoritarianism: we are losing trust in truth, in what is said, the day to day social trust is dissolving. It worries me an awful lot, but it seems like we're (1) destroying the basis for much of our society and prosperity. All of the social technology we've amassed over the past century and a half seems to be in decline.

1: American English language culture I can read as an insider, and I pay some attention to the UK culture as well. I'm sadly monolingual so I have vague understanding of other countries, but they seem on a broadly similar path.

[taeric]

I'm assuming this is talking about the revised numbers for 2025? And if that revised so heavily downward, hard to imagine the latest numbers are somehow more stable?

[dataflow]

How does BLS even know if each person in the country is unemployed or seeking a job?

[mandevil]

The same way that anyone would: they conduct a statistical sample and report unemployment numbers based on that. Source: 15 years ago I was one of the people they contacted. As I recall, a letter came to our apartment, asking for the adult (18+) who had their birthday closest to that date to contact them and provide a phone number and some basic demographic data. That was me, not my roommate, so for the next 12 months I got a call once a month asking me for my employment details about a specific week that month- had I been looking for a job, was I still employed, etc.

[rybosworld]

Mostly through payroll data/surveys.