Why U.S. Jobs Data Misses the Real Labor Market

Name: How Unemployment AND Employment Are Rising... At The Same Time
Uploaded: 2026-04-23T18:57:51+00:00
Duration: 17 min 9 s
Channel: How Money Works
Description: Summary and key takeaways on Why U.S. Jobs Data Misses the Real Labor Market, covering The Disconnect in Jobs Reporting Recent Bureau of Labor Statistics (BLS)

How Money Works

Apr 23, 2026

•

17 min video

•

2 min read

YouTube video ID: 3UO-gl-N6rY

Source: YouTube video by How Money Works — Watch original video

PDF

Recent Bureau of Labor Statistics (BLS) releases show a paradox: headline job growth and falling unemployment coexist with a shrinking labor force, 60,000 announced job cuts, and a 396,000‑person decline in the labor pool. Consumer sentiment, measured by the Conference Board, sits at record lows, directly contradicting the “hot” economy narrative. Indeed’s hiring data further suggests that companies are adopting a defensive posture rather than expanding aggressively.

The Two‑Survey System

The BLS relies on two separate surveys. The Establishment Survey counts payrolls at business locations and supplies the headline number that moves bond yields and shapes policy. The Household Survey records individual employment status and captures workers outside traditional payrolls. Since 2020 the two surveys have diverged sharply; the Establishment Survey remains consistently more optimistic, while the Household Survey paints a bleaker picture of labor market health.

Structural Flaws in Measurement

Payrolls vs. People

Because the Establishment Survey tallies payroll entries, a single worker holding multiple jobs is counted multiple times. A nurse who also works part‑time in retail, for example, adds two jobs to the total without adding a new worker.

“Births and Deaths” Model Limitations

The statistical “Births and Deaths” model fills gaps in the Establishment Survey by estimating business formation and closure. It now overcounts gig‑economy participants—an Uber driver who started last week registers as a new business—skewing the perception of entrepreneurial activity.

The Invisible Workforce

Contractors, freelancers, consultants, and creators belong to the “invisible” segment that the Establishment Survey misses entirely. They appear only in the Household Survey, which markets and policymakers largely ignore. When downturns hit, these workers are the first to be cut, yet their layoffs never surface in payroll‑based layoff statistics.

Political and Economic Stakes

Changing the methodology would create data discontinuities that markets dislike, prompting agencies to cling to the status quo. Admitting long‑term inaccuracies could invite a Congress eager to cut agency funding. Consequently, the flawed numbers persist as a convenient “noise generator” that politicians on both sides can spin to claim either a roaring or collapsing economy.

Mechanisms Behind the Numbers

Establishment Survey Methodology: Designed for a traditional, rigid workforce, it surveys individual business sites and struggles with remote work, multi‑location payrolls, and the rise of flexible employment.
Payroll vs. Person Discrepancy: Counting payrolls inflates job totals because each job, not each worker, is recorded.
Invisible Workforce Gap: Contractors and gig workers are captured only by the Household Survey, leaving a blind spot in the primary economic indicator.

The combined effect of these flaws is a labor market portrait that moves bond yields the morning it is released, yet fails to reflect the lived reality of millions of American workers.

Takeaways

Recent BLS reports show job growth and falling unemployment while the labor force shrank and job cuts rose, revealing a stark divergence between headline numbers and underlying labor conditions.
The Establishment Survey, which counts payrolls at business sites, has become increasingly optimistic since 2020, whereas the Household Survey shows a weaker labor market, creating two conflicting narratives.
Counting payrolls rather than people inflates job totals because a single worker with multiple jobs is counted multiple times, and the survey omits contractors, freelancers, and gig workers who are often the first to be laid off.
The “Births and Deaths” model used to estimate business formation now overstates new businesses by treating gig‑economy participants as startups, further distorting the picture of economic dynamism.
Agencies resist methodological changes because revisions would disrupt markets and expose long‑standing inaccuracies to a Congress that could threaten funding, allowing the flawed data to persist as a convenient political tool.

Frequently Asked Questions

How does the Establishment Survey inflate job numbers?

The Establishment Survey inflates job numbers by counting each payroll entry rather than individual workers. When one person holds several jobs—such as a nurse who also works part‑time in retail—the survey records each position as a separate job, so the total job count rises without adding new workers. This method also overlooks contractors and gig workers who are not tied to a traditional payroll.

Who is How Money Works on YouTube?

How Money Works is a YouTube channel that publishes videos on a range of topics. Browse more summaries from this channel below.

Does this page include the full transcript of the video?

Yes, the full transcript for this video is available on this page. Click 'Show transcript' in the sidebar to read it.

Helpful resources related to this video

If you want to practice or explore the concepts discussed in the video, these commonly used tools may help.

Books On Economic Data Analysis Recommended

Provides foundational knowledge on how economic indicators are calculated and interpreted, helping the reader critically evaluate government reports.

Amazon →

The Misbehavior Of Markets By Benoit Mandelbrot

Explores the limitations of traditional economic models and the reality of market volatility, mirroring the critique of flawed government measurement systems.

Amazon →

Lying With Statistics By Darrell Huff

A classic guide on how data can be manipulated or misinterpreted, directly relevant to the speaker's analysis of how jobs data is framed.

Amazon →

Links may be affiliate links. We only include resources that are genuinely relevant to the topic.

Summarize another video

Full Transcript YouTube

Late last week, the Bureau of Labor Statistics 
released its monthly jobs report with some really  
good news. We had added an extra 178,000 jobs in 
the month before and unemployment had fallen by a  
modest.1%. But uh this raised the obvious question 
of how could that possibly be true. Consumer  
sentiment has hit its lowest level since the data 
was first collected, beating out the height of  
COVID, the global financial crisis, and even the 
oil shock of the late '7s. The Conference Board  
survey found that the highest number of consumers 
ever considered jobs hard to get. And Indeed's  
data team found that overall hiring was in an 
extremely defensive posture amongst the companies  
they service. Now, perhaps all of these sentiments 
could be written off as bad vibes amongst  
people who don't truly appreciate how hot this 
economy is, except that on top of all of this,  
there is a mountain of data that just straight 
up contradicts this number. In March alone, when  
we apparently added 178,000 new jobs, companies 
announced over 60,000 job cuts, the Fed's internal  
data showed low hiring. Weekly unemployment 
claims increased to 219,000. And in the exact  
same report from the BLS themselves, they noted 
that the labor force had shrunk by 396,000. So,  
just to make things clear, unemployment claims 
rose, companies cut jobs, and labor force shrunk.  
And yet the official jobs figure went up. The 
overall trust in these figures has been waning  
for some time now as massive revisions become 
commonplace and we swing between massive gains  
and massive losses every single month. Even here 
on this channel, I've spent more time than I would  
like to admit methodically reverse engineering the 
statistical technicalities that these government  
departments use to produce their numbers because, 
well, it's really important to our everyday lives.  
But in this case, just basic arithmetic should 
prove that something is going wrong here. Either  
some combination of these numbers are just plain 
wrong or they are so unintuitive or unreliable  
that they may as well be disregarded entirely. 
Breaking numbers right now. Economists had  
forecasted 65,000 jobs at it. Major rebound for 
American workers after shedding more than 130,000  
jobs in February. A blowout jobs number. How much 
momentum is in this economy? David Intel with  
its research and development hub in Hillsboro is 
laying off workers yet again. Right now, federal  
workers are bracing for potential mass layoffs 
as we enter day three of a government shutdown.  
Trump administration's beginning mass layoffs. The 
economy added 172,000 new private sector jobs to  
kick off the year. The United States added 172,000 
jobs. The private sector. Okay. So, when you read  
a headline or listen to a politician talking 
about all of the jobs they have just created,  
the number they are talking about is almost always 
the Bureau of Labor Statistics total non-farm  
payroll count. It is the number that moves bond 
yields the morning it gets released, the number  
that sets the tone for Federal Reserve meetings, 
and the number that politicians on both sides of  
the aisle will use to claim that the economy is 
either roaring or collapsing depending on what  
suits them that week. The problem is that more and 
more over time, it has slowly stopped describing  
the thing most people actually mean when they 
hear the word jobs. When the BLS says 178,000  
jobs were added, most people hear 178 previously 
unemployed Americans waking up to a new paycheck,  
what it actually means is something much, much 
more complicated and uh significantly less  
reassuring. To understand why these numbers have 
drifted so far away from what they sound like,  
you need to look at where they actually come 
from. The Bureau of Labor Statistics doesn't  
have a magic real-time feed of every paycheck in 
the United States. Despite what you might assume,  
in 2026, with the IRS sitting on a real time 
withholding database that could in theory  
tell you exactly how many workers got paid in any 
given week, that's not the data we use. Instead,  
every single month, the BLS runs two completely 
separate surveys. the establishment survey and the  
household survey. The former surveys individual 
business sites and offices and gathers data on  
how many people they have on their payroll, what 
those people do for work, and where that job site  
is actually physically located. The latter, as the 
name might suggest, surveys hundreds of thousands  
of households to ask them if they lost their 
job, got a job, remained employed, or unemployed.  
The results of one of them gets printed on 
the front page of the Wall Street Journal,  
while the other one gets quietly buried in 
the back of the same report. Now, logically,  
in a healthy data ecosystem, the two surveys 
should be moving in roughly the same direction  
most of the time with small disagreements at the 
margins. That used to be the case. For most of  
the post-war period, the establishment number and 
the household number tracked each other reasonably  
well. When one said the labor market was 
strong, the other almost always agreed. However,  
over the last decade, that relationship has 
visibly broken down. Economists at the San  
Francisco Fed have published research showing that 
the gap between the two surveys has widened to its  
largest level since they started keeping records, 
and the divergence has accelerated specifically in  
the period after 2020. In plain English, the two 
numbers that are supposed to be measuring the same  
thing are increasingly telling us two completely 
different stories about the American labor market.  
And the one we have decided to focus on is the 
one that structurally is set up to look better  
than reality almost every single month. Again, 
just last month, the payroll survey said we  
gained 178,000 new jobs, while the back of the 
report, households claimed to have lost 64,000,  
which really just shouldn't be possible. Now, 
you might be thinking, this is the kind of thing  
the BLS would be quietly working to fix in the 
background. And to their credit, in their own  
technical notes, they actually emit a lot of these 
problems. They publish the household number. They  
publish the revisions. They publish footnotes on 
methodology. And they publish their own confidence  
intervals. The catch is that nobody who isn't 
paid to do this for a living ever reads any of  
that. Markets react to the headline. Politicians 
react to the headline. Business owners deciding  
whether to hire react to the headline. Workers 
deciding whether their job is safe react to the  
headline. And the headline keeps coming from the 
survey that has the worst structural fit for what  
the modern American economy actually looks like. 
Which raises the obvious question, how has this  
number become so divorced from reality? And why 
do we even pay attention to it anymore? Well,  
it's time to learn how many works to find out what 
the monthly jobs report is actually measuring,  
why it keeps contradicting itself by hundreds of 
thousands of workers in the exact same report,  
and why nobody actually wants to fix it. Every 
week there's another massive data breach,  
another AI model scraping personal data, or 
another company rewriting who owns yours. If that  
makes you uneasy, you're not alone. And that's 
why this video is sponsored by Proton. Proton  
is an endto-end encrypted platform built by CERN 
scientists who wanted a more private internet.  
Over 100 million people use it now, and the free 
plan gets you Proton Mail for secure email and  
Proton Drive for encrypted cloud storage in one 
account. What I actually like is how easy it  
makes privacy. Your emails are encrypted so only 
you and the recipient can read them. Your files  
are encrypted before they leave your device. And 
because Proton is based in Switzerland, your data  
is covered by some of the strongest privacy laws 
in the world. You can even create email aliases  
to cut down on spam and fishing. And Proton Drive 
works like any other cloud storage, just without  
anyone, not even Proton, seeing your stuff. If 
you want to take control of your digital life,  
you can get Proton Mail and Proton Drive together 
for free. Check out the link in my description to  
get started. Okay, so it is easy and honestly 
a little bit understandable to look at these  
consistently inaccurate economic metrics and 
conclude that the agencies putting them together  
are just trying to make the politicians they 
answer to look good. However, at the very least,  
before we attribute this all to malice, we should 
understand how it is much more likely to just be  
statistical stupidity. And to that end, there 
are really three big problems with how we come  
up with this number. And the first simple problem 
is that payrolls are not people. The establishment  
survey does exactly what the name suggests. It 
surveys establishments and asks them how many  
payrolls they process that month. At the time 
this data source was first conducted, it was  
a perfectly reasonable proxy for how many people 
have jobs. One person, one employer, one payroll.  
This is still the norm for most employees, but 
it's getting less and less normal every year.  
These days, a single person can be attached to 
multiple payrolls. And the number going up doesn't  
necessarily mean a new person got a new job. It 
might just mean an existing worker picked up a  
second one to keep up with rent. The most extreme 
version of this shows up in healthcare, where it's  
pretty common place for doctors and nurses to work 
across multiple different establishments. A single  
doctor working at a hospital, a private practice, 
and doing some research on the side could  
technically add three payrolls to the official 
number while only being one single working person.  
This technicality was particularly important for 
last month's job numbers because it also counted  
35,000 healthcare professionals returning to work 
after having been on strike the month before.  
Hopefully, it should go without saying that this 
isn't the same thing as the economy genuinely  
providing employment opportunities to 35,000 new 
people. But all this data does is look at payrolls  
every month. So, the establishment survey says 
we added 178,000 payrolls. The household survey  
in the same report on the same day says 64,000 
fewer humans actually have work. Both of these  
numbers are technically true. They are just 
measuring different things and almost nobody  
outside of macronerds ever bothers to look at the 
second one. Which brings us to problem number two.  
The methodology. Collecting payroll data at the 
scale of the entire American economy is genuinely  
impossible. And the model the BLS uses to do it 
is becoming increasingly outdated. Payrolls are  
not counted firm to firm but rather location 
to location. Again, the establishment in the  
establishment survey means a survey of individual 
well establishments. So, as more and more people  
move into jobs that are work from home, jobs that 
move between cities or jobs that run out of a head  
office on the other side of the country, it has 
become much harder to accurately assign those  
payrolls to a real workplace and monitor them 
monthtomonth. Double counting, undercounting,  
and plain old misunderstandings about what payroll 
should count have just become far more common as  
our working arrangements have become less rigid. 
This is part of the reason why revisions have  
gotten so embarrassingly large. Last year 
alone, the BLS quietly revised away 800,000  
jobs that they had previously claimed were added, 
meaning the headline number for an entire chunk of  
the year was just wrong. You can actually go and 
look up the exact collection methodology that the  
BLS uses on their website and you can see it was 
designed for a time when most people either went  
to a factory or an office, did their hours, and 
went home every day. To try and balance this out,  
the BLS also uses something called the bursts and 
death model, which is essentially a statistical  
estimate of how many businesses they think were 
created or shut down in any given month used to  
fill in the gaps that the survey itself can't 
capture. The births and deaths model has its  
own problems, but the biggest one is that it is 
heavily skewed by the proliferation of gig work,  
which we have mentioned a few times in previous 
videos. A new business in this model can be a  
guy who started driving for Uber last week. That 
gets counted as a new job creating enterprise,  
even though it is functionally a person without 
stable employment trying to make rent. So,  
the survey itself misses the modern workforce. 
And the model designed to patch the survey is  
also calibrated for a workforce that doesn't 
really exist anymore. Now, these growing blind  
spots in one of the most important metrics in our 
economy is not great, but it gets worse. The third  
problem is by far the biggest one, and that is 
the people the payroll survey doesn't see at all.  
The establishment survey by its very nature tends 
to count people in traditional payroll jobs,  
which on average means more secure positions 
with more established employers who actually  
respond to voluntary BLS requests. An increasingly 
large share of the American workforce does not  
fit any of those descriptions. They are private 
contractors. And this is not just gig workers.  
This includes everyone from a guy delivering 
groceries through Door Dash all the way up to  
extremely high-end technical roles, consultants, 
freelance software engineers, and well, yeah,  
basically the entire creator economy. Can't 
forget about the real heroes. These people  
get counted as employed or unemployed by the 
household survey. And if they lose their work,  
it still sucks just as much as anybody else losing 
their job. But they are completely invisible to  
payroll. They are also conveniently the people 
most likely to lose their work before anybody  
else in more secure formal roles. When a company 
starts to feel a downturn coming, the first line  
items to get cut are not the staff with employment 
contracts and severance packages. The first to go  
are the contractors, the freelancers, the agency 
the marketing team was using, the consultants,  
and the uh inoff life coaches that teach the CEO 
how to act human. None of those people show up as  
layoffs on press releases. None of them show up 
in the establishment survey and it can take weeks  
or months before they show up in unemployment 
claims because most of them aren't eligible to  
file in the first place. This is how the labor 
force can fall by almost 400,000 people in a  
single month while payrolls technically still go 
up. The two surveys are just capturing two very  
different groups of workers and those groups are 
drifting further apart every year. One group is on  
a W2 in an HR system with their payroll processed 
by a service the BLS already knows how to call.  
The other is invoicing clients paying their own 
taxes and disappearing from the data the second  
the contract dries up. Which raises a perfectly 
reasonable question. If the payroll survey is  
flawed, why do we still use it? Well, two reasons. 
The first is that it's fast. It can give us  
monthly data with comparatively little collection 
which matters for an economy where the Fed and  
the government need to make decisions on a near 
real-time basis. The second is that the payroll  
survey is reliableishly unreliable. Businesses 
are not exactly motivated to make up payrolls to  
boost national statistics. Whereas individual 
households can be koi about their employment  
situation for all kinds of reasons. People who 
have a job are less likely to answer the phone for  
a BLS survey because they are well working. People 
who just lost one might tailor their response out  
of embarrassment or genuine confusion about what 
counts as actively looking for work. There is also  
a less talked about timing problem. The household 
survey covers the week containing the 12th of the  
month. So, a person who loses their job on the 
13th keeps counting as employed for another 4  
weeks before the data even has a chance to catch 
up. So, the household survey is not some pristine  
alternative the BLS is just refusing to use. It 
has its own real problems. Now, I know this again  
sounds like boring money man rants about data 
collection methodologies for 15 minutes straight.  
But this really does matter for two reasons. The 
first is that of course these numbers get used  
to make real decisions that affect us. Federal 
Reserve policy on interest rates, stimulus debates  
in Congress, how foreign investors price treasury 
yields, and ironically, it even feeds into a  
business's own plans to hire. If companies see 
unemployment rising and headline payrolls falling,  
they become more riskaverse, which leads to less 
hiring, which leads to less consumer spending,  
which leads to even less hiring. So, it's just as 
important to make sure that these numbers don't  
look artificially pessimistic as it is to make 
sure they don't look artificially optimistic. The  
second reason this is so important to understand 
is that if these numbers keep looking like total  
nonsense, it becomes easier for people to dismiss 
them as useless at best or outright propaganda at  
worst. Not having faith in our economic statistics 
is a great way to undermine faith in the economy  
as a whole. Now, the economists at the BLS know 
all of this. They have written papers about it.  
Statistics agencies in other developed economies 
have already started rebalancing. The reason the  
United States hasn't is structural. Changing the 
methodology would create a one-time discontinuity  
in the data series, which makes year-over-year 
comparisons look weird, which the market hates. It  
would also require the BLS to admit on the record 
that the number it has been publishing for decades  
isn't really measuring what people thought it was, 
which is the kind of emission federal agencies do  
not love making in front of a Congress that is 
already gunning to cut their funding. And on top  
of all of that, every administration in power 
has a quiet incentive to leave a politically  
convenient noise generator exactly where it is. 
A monthly number that can swing 200,000 jobs in  
either direction based on which doctors happen to 
be on strike that month is incredibly useful if  
you are the one writing the press release. So the 
number stays, the methodology stays, the revisions  
get larger, and the gap between the headline and 
the lived experience of the labor market keeps  
widening, which means for the foreseeable future, 
the most quoted economic statistic in the United  
States is going to keep telling a story that fewer 
and fewer working people actually recognize. Now,  
if all of that wasn't enough, go and watch this 
video next to further understand how to lie  
with statistics. And don't forget to like and 
subscribe to keep on learning how money works.

Heimler's History

May 01, 2026

Watch Read Summary

PDF

The Two‑Survey System

Structural Flaws in Measurement

Payrolls vs. People

“Births and Deaths” Model Limitations

The Invisible Workforce

Political and Economic Stakes

Mechanisms Behind the Numbers

Takeaways

Frequently Asked Questions

How does the Establishment Survey inflate job numbers?

Who is How Money Works on YouTube?

Does this page include the full transcript of the video?

Helpful resources related to this video

Share This Summary

Embed This Summary