Flying Blind: ADP Stops Giving the Fed Access to Its Private Jobs Data

After a speech by Fed Governor Chris Waller in August, ADP stopped giving the Fed data.

Lost Access

The data in question is not the public data from which I produce monthly charts such as the lead chart.

Rather, the Fed lost access to proprietary subsets.

The Wall Street Journal reports Fed Lost Access to Private Jobs Data Ahead of Government Shutdown

Federal Reserve officials, who are unable to receive U.S. economic statistics because of the continuing government shutdown, recently lost access to a separate measure of employment data from a third-party provider.

Since at least 2018, payroll-processing company ADP has provided the Fed with access to a data set that includes anonymized information on employment and earnings for millions of workers. The data, which covers 20% of the nation’s private workforce, had been available to the Fed with a roughly one-week delay—making it both a timely and comprehensive gauge of job-market conditions.

ADP stopped providing its data to the Fed shortly after a speech by Fed governor Christopher Waller in late August drew attention to the central bank’s longstanding use of its weekly payroll data, according to a person familiar with the matter.

The Fed’s data-sharing relationship with ADP has been public for years. Minutes of Fed policy meetings as recently as 2023 included generalized descriptions of the central bank’s analysis of ADP data similar to Waller’s August speech.

Fed Chair Jerome Powell first publicized the central bank’s collaboration with ADP in a 2019 speech. Powell explained how the central bank’s staff economists had developed a method that used underlying data to predict official government data, including after revisions, on payroll growth reported every month by the Bureau of Labor Statistics.

Powell has attempted to persuade ADP to restore the data sharing, but so far, those efforts haven’t succeeded, according to the person familiar with the matter. The loss of the data could take on new significance after the government shutdown prompted statistical agencies to furlough workers on Oct. 1 and suspend most data releases this month.

During a question-and-answer session at an economics conference last week, Powell said the Fed doesn’t expect private data sources to fully replace the government statistics it isn’t receiving. But Powell noted the availability of “some pretty good substitutes” for employment data, and he cited ADP as one example.

Waller’s Let’s Get On with It Speech

On August 28, Fed Governor Chris Waller made a speech Let’s Get On with It

…. I have heard the argument that declining labor supply due to lower immigration means that low jobs numbers aren’t as bad as they look and that the “breakeven” number for keeping the unemployment rate steady is declining. To those people, I would say: Yes, these estimates have been falling, but I haven’t heard anyone say the breakeven level is negative. Supply-side changes can’t account for the ugly jobs numbers of the past three months.

Another sign of a weakening in the labor market is that the unemployment rates for some cyclically sensitive groups of workers, such as teenagers, are up sharply this year, to levels, again, similar to those in the softer job market of the mid-2010s. Supplementing this hard evidence of falling labor demand is the consistent story from my business contacts that they are not hiring. The mix of reasons is not only uncertainty over tariff policy and slowing demand for their goods and services, but also, increasingly, uncertainty over how to use artificial intelligence, which is especially freezing hiring for some entry-level jobs.5

This softening in labor demand is occurring at the same time, as I noted earlier, that there is a decline in labor supply. I often hear the argument that the unemployment rate is the key labor market indicator because it reflects what is happening in both demand and supply. And given that the current unemployment rate is 4.2 percent, close to most estimates of maximum employment and barely changed over the last year, this suggests that the labor market is solid. While I would normally agree with this statement, I do not think it is accurate now. I believe that any decline in labor supply is only masking weakening demand in the labor market. Whether or not supply is down, weakening demand is not good, and it is specifically what monetary policy is intended to address.

Before I move on from the labor market, let me address a question of some public debate recently, which is the quality of the payroll data. As should be obvious from these remarks and others I have delivered, I consider the monthly jobs report, including its estimate of payroll employment, an absolutely indispensable tool in evaluating the economy and setting monetary policy, and that view is close to unanimous among economic forecasters and policymakers. Considering how important this data is in getting an accurate and reliable read on the state of the labor market, it is entirely appropriate to examine the quality of the jobs numbers and the process by which they are collected. There is always room for improvement in data collection and analysis. In fact, the Bureau of Labor Statistics (BLS) and other agencies that collect economic data have benefited from expert committees of outside advisers who have provided advice on ways to improve data quality.

There has been substantial commentary since the last jobs report regarding the declining response rate of establishments and individuals in the employment survey and the outsized revisions of the data. This commentary has raised questions about the quality of the employment survey data. So I want to address this issue and clarify some facts about the survey response and the revisions. When the BLS sends out the establishment survey for a given month, it gives firms three months to return the survey—they can return it in the first month, the second month, or the third month. Some firms do not respond at all. Firms that submit their responses in the first month are the basis for the initial, or preliminary, payroll estimate. Those firms that respond in the second month get added to the first-month responders to get an updated number, which yields the first revision that is published a month later. Then the firms that respond in the final month provide the BLS with the final numbers for the original month. This is the second revision and is the official payroll estimate for that month.

A decade ago, the response rate for the first month was 75 percent, meaning three-fourths of the firms surveyed responded immediately. But the initial response rate has fallen to 60 percent recently. Does that imply that the overall response rate for the entire three-month period has fallen? No, response rates by the final estimate have not fallen—about 95 percent of firms responded a decade ago, and that is how many are responding now. What has happened is that U.S. firms are slower in returning the survey. Ten years ago, if the survey had been sent to 100 firms, 75 would respond in the first month, and 20 firms would respond in the next two months. But now only 60 firms respond in the first month, and 35 firms respond in the second and third months. But in both cases, 95 out of 100 firms surveyed responded.

What are the implications of this delayed response pattern? First, if more data come in later in the survey period, then the initial estimate is less likely to accurately represent the final estimate, which means you are more likely to get significant revisions in the second and third months. The revisions do not mean the quality of the data has declined; it simply reflects the fact that more data is coming in late in the survey period.

A second point is that revisions have been, on average, negative for the past couple of years, and while the recent revisions were large, revisions have not been getting bigger, on average, based on an analysis recently issued by economists at J.P. Morgan. They did find that large revisions are often associated with major turning points in the labor market. One possibility is that the recent large revisions in the payroll data for May and June indicate the labor market is at an inflection point and may worsen in coming months.

The reference to ADP was in footnote #4.

I also look at timely data that Federal Reserve staff maintains in collaboration with the employment services firm ADP to construct a measure of weekly payroll employment, which covers about 20 percent of the nation’s private workforce. This measure is comparable to the one ADP publishes. The current May–July contour for the staff measure of ADP-based private employment is broadly consistent with that of the Current Employment Statistics numbers. And in the weeks after the July jobs report’s reference period, preliminary estimates from ADP show continued deterioration.

Pure Speculation

In the pure speculation category, is is possible the Trump administration pressured ADP to stop giving data to the Fed?

Alternatively, ADP may have reacted in advance of Trump administration pressure, fearful of it.

Q: Is either explanation likely?
A: I don’t know, but they are the only explanations I can come up with.

I suspect direct pressure from Trump is more likely than ADP operating in advance out of fear of pressure.

And since it was widely understood that ADP gave data to the Fed, a sudden change of mind for no reason at all seems less likely than the above explanations.

A reader provided a more straight-forward explanation: ADP wants the Fed to pay for data.

Regarding Waller’s Speech

Waller is either covering up for the BLS or is foolish enough to believe BLS methodology is sound.

This is not just a matter of BLS response rates.

The BLS is known to be wrong at turns, and Covid enhanced those errors.

I have written about this many times, here’s a recap.

Related Post Recap

December 8, 2024: Few Bother to Respond to BLS Surveys Anymore (And Why That’s Important)

BLS response rates plunged after COVID and never recovered, especially job market surveys.

June 16, 2025: QCEW Report Shows Overstatement of Jobs by the BLS is Increasing

The discrepancy between QCEW and the BLS jobs report is rising.

Sample Sizes

• CES: 70,000 Establishments
• CPI: 10,500 Establishments
• JOLTS: 7,200 Establishment
• QCEW: 12.2 Million Establishments Quarterly
• BED: 9.2 Million Establishments

QCEW stands for Quarterly Census of Employment and Wages. It represents 95 percent of the data.

CES is the nonfarm payroll report.

September 3, 2025: The Unemployment Level Is Now Greater than Job Openings

For the first time since the pandemic unemployment is above openings.

September 9, 2025: New QCEW Data Indicates More Big Negative Revisions Coming to Job Reports

The discrepancy between jobs reports and quarterly data widens again.

September 24, 2025: Hoot of the Day: Blame the Public for BLS Job Reports and Revisions

Some blame the Public. Trump blames manipulation.

ADP Private Jobs Decline by 32,000 in September, Huge Negative Revisions

On October 1, 2025 I noted ADP Private Jobs Decline by 32,000 in September, Huge Negative Revisions

ADP Change in Small, Medium, Large Employment Year-Over-Year

• Small: 1-49 Employees: -50,000
• Medium: 50-499 Employees: +367,000
• Large: 500+ Employees: +776,000

Small businesses are getting clobbered by tariffs, as predicted in this corner

It’s obvious what’s going on. At the beginning of 2025, small business jobs were up ~400,000 from a year ago. As of September, year-over-year small business jobs were 50,000 in the hole.

Small businesses have fewer means of avoiding Trump tariffs and are far more at risk in a slowing economy.

Wait Till Next Year

On October 5, I noted Trump Adopts Chicago Cubs’ Perpetual Message, “Wait Till Next Year”

“One Big Beautiful Bill” did not resonate. Trump opts for “Wait Till 2026”.

But “Wait Till 2026” is a fundamental mistake. When 2026 is bad, the message will have to change.

The beauty of a more Cub-esque “Wait Till Next Year” is the slogan never has to change.

Then again, maybe all the polls are wrong, ADP is wrong, QCEW is wrong, and Trump is right.

Five New Polls Show 70% of Americans Are Pessimistic on the Economy

On October 9, I noted Five New Polls Show 70% of Americans Are Pessimistic on the Economy

Independent polls by Pew, S&P, CBS, the WSJ, and Fannie Mae all say the same thing.

How Screwed Up Are BLS Real and Nominal Median Earnings?

On October 7, I asked How Screwed Up Are BLS Real and Nominal Median Earnings?

Discrepancies between ADP and the BLS are vast. Let’s start with the BLS.

It’s the addendum that’s most relevant.

Addendum

A reader accurately points out the BLS is a sample, while ADP is a population.

This is correct.

Presumably the BLS is a valid sample. But we know that it isn’t. The BLS underestimated jobs by millions coming out of Covid and is overestimating them now.

We do not have good data. And when BLS does make revisions, it’s in one big bang. Historical errors are left intact.

If the BLS had a valid sample and knew how to weigh them, we would not have the errors that we do.

But none of this matters. The new message is “wait till next year.” That’s when miracles happen.

Addendum

One explanation for lack of access is so obvious that I missed it. A reader commented that ADP wants the Fed to pay for data. Added to the above listed reasons.