Hacker News Reader

Comment by analyte123 | Hacker News Reader

The ADP payroll report is noise. It is based on the payroll data from ADP only. The assumption of this report is that companies that use ADP to process their payroll are completely representative of the entire economy and that there is no regional, sector, or company stage bias to their customers. A firm with ADP laying employees off and 3 new firms with the same number of employees being founded and using a different payroll provider would be reported as a "loss" here. Maybe the private sector did lose jobs, but I wouldn't use this report to find out.

View on HN

ADP say that they handle payroll for one in six of all companies in America. That is both a large sample size, and probably broadly representative of the economy. There will of course be some business segments that are over or underrepresented but that is different than disregarding the entire report as noise.

ADP is predominant in large companies and has little hold in startups. It skews the stats.

ADP data does have a bias, but it is so much data it provides a valuable signal. The importance of understanding the source of your data and how it represents the largest population is something every scientist has been drilled on.

two small companies i used to work for used adp.

...You know that "startup" is not synonymous with "small company", right?

What's the situation in the vast swath of the economy that's made up of small non-startup companies?

> over 900,000 small businesses trust ADP [0]

From what I can find they have 1.1m clients and over 900k are small businesses. The vast majority. This makes sense, as there aren’t that many massive companies.

[0] https://www.adp.com/what-we-offer/payroll/payroll-for-1-49-e...

> That is both a large sample size,

It doesn't matter how large your sample size is if your sampling method is biased. This could be measuring market share gain/loss in different segments of a steady employment environment.

> disregarding the entire report as noise

Studies with bad / non public sampling methods should be, at a minimum, treated with great skepticism. Why would that not apply here?

A factual number that has less statistical noise or political bias is extremely valuable. Yes, one needs to factor for the biases but that doesn't mean the number should be ignored.

The trend is useful, since one can fairly safely assume that most of the biases haven't radically changed.

> ADP say that they handle payroll for one in six of all companies in America

that's less than 20%. If you had 10 people to interview and interviewed 2 of them i wouldn't say you interviewed a representative sample of the 10.

No, but if I had 1000 people to interview and interviewed 200 of them at random then absolutely yes.

I know right. Statistics.

The key is "at random". Businesses are not randomly assigned to use ADP

That's a separate problem. I have no idea what the sampling bias is, but the issue isn't that they "only" sample 20% of the population.

ADP reported a meager 29,000 increase in new jobs, but the BLS showed a much larger 139,000 gain. April also showed a similarly wide gap between the two reports.

Through the first five months of 2025, the difference between the two reports has averaged a whopping 63,000 a month.

source: https://www.morningstar.com/news/marketwatch/20250702107/be-...

ADP has always been out of sync with BLS numbers. Here is an article in the Atlantic all the way back in 2011 talking about it.

https://www.theatlantic.com/business/archive/2011/05/chart-o...

Elsewhere in the comments it mentions that ADP is weighted more toward the private sector, while BLS is more representative of government jobs.

These two things don’t need to be in lockstep.

BLS tracks both numbers, and the private still diverges quite a bit. We saw this again in July. ADP number was -33k private sector jobs, BLS showed growth of 74K for private sector, so massive 100k delta in just the private sector number.

That is quite the delta, but why is it assumed BLS is the more correct one?

From the BLS site on their methodology[0]:

> The Current Employment Statistics (CES) program, also known as the payroll survey or the establishment survey, is a monthly survey of approximately 121,000 businesses and government agencies representing approximately 631,000 worksites throughout the United States.

> All new samples are solicited by computer-assisted telephone interview (CATI), and data are collected for the first 5 months via this mode. After the initiation period, many sample units are transferred to one of several less costly reporting methods that are self-initiated by the respondent.

How is a survey from 121k businesses better than the actual payroll data from a company handling 460k companies[1]?

Isn’t the larger sample generally considered better? Isn’t source data generally considered better than a phone survey of self-reporting?

Looking at the two methodologies, it sounds like BLS data is noise, not ADP.

[0] https://www.bls.gov/bls/empsitquickguide.htm

[1] 2022 data https://mediacenter.adp.com/2022-03-02-ADP-National-Employme...

Noise is the absence of a signal. A biased signal is fine, just account for the bias.

ADP is huge and covers a broad range of sectors. It would be a very interesting result (and a very extraordinary claim) if the employment data from non-ADP companies went in the opposite direction of ADP. I certainly see no evidence that firms underrepresented in ADP's data are hiring prodigiously.

> assumption of this report is that companies that use ADP to process their payroll are completely representative of the entire economy

No such assumption is made except by an errant reading of the report. The ADP report [1] can be used to predict BLS numbers, but it’s also independently useful. The reason the headline is 33,000 private-sector jobs were lost is because 33,000 private-sector jobs were lost, ADP can directly count that.

[1] https://adpemploymentreport.com/

There have been many instances of ADP and BLS job reports being out of step, which can be expected because of their differing methodologies. On the other hand, nobody with a brain can take BLS surveys at face value under these circumstances.

It depends on how randomly representative it is. If it is close to random, then it's better than most poll data.

Reply on news.ycombinator.com