America Added 178,000 New Jobs Last Month... But How Is That Possible?!

0:00 Late last week, the Bureau of Labor Statistics released

0:02 its monthly jobs report with some really good news.

0:05 We had added an extra 178,000 jobs in the month

0:09 before and unemployment had fallen by a modest.1%.

0:12 But uh this raised the obvious question of how could that possibly be true.

0:19 Consumer sentiment has hit its lowest level since the data was first collected,

0:23 beating out the height of COVID, the global financial crisis,

0:26 and even the oil shock of the late '7s.

0:28 The Conference Board survey found that the highest

0:31 number of consumers ever considered jobs hard to get.

0:34 And Indeed's data team found that overall hiring was

0:37 in an extremely defensive posture amongst the companies they service.

0:40 Now, perhaps all of these sentiments could be written off as bad

0:44 vibes amongst people who don't truly appreciate how hot this economy is,

0:48 except that on top of all of this, there is

0:51 a mountain of data that just straight up contradicts this number.

0:54 In March alone, when we apparently added 178,000 new jobs,

0:57 companies announced over 60,000 job cuts,

1:00 the Fed's internal data showed low hiring.

1:03 Weekly unemployment claims increased to 219,000.

1:06 And in the exact same report from the BLS themselves,

1:11 they noted that the labor force had shrunk by 396,000.

1:15 So, just to make things clear, unemployment claims rose,

1:18 companies cut jobs, and labor force shrunk.

1:21 And yet the official jobs figure went up.

1:24 The overall trust in these figures has been waning for some time now as massive

1:29 revisions become commonplace and we swing between

1:31 massive gains and massive losses every single month.

1:33 Even here on this channel,

1:35 I've spent more time than I would like to admit methodically reverse engineering

1:39 the statistical technicalities that these government

1:41 departments use to produce their numbers because,

1:44 well, it's really important to our everyday lives.

1:47 But in this case, just basic arithmetic

1:49 should prove that something is going wrong here.

1:52 Either some combination of these numbers are just plain wrong or they

1:56 are so unintuitive or unreliable that they may as well be disregarded entirely.

2:01 Breaking numbers right now.

2:03 Economists had forecasted 65,000 jobs at it.

2:06 Major rebound for American workers after

2:08 shedding more than 130,000 jobs in February.

2:11 A blowout jobs number.

2:13 How much momentum is in this economy?

2:15 David Intel with its research and development hub

2:18 in Hillsboro is laying off workers yet again.

2:20 Right now, federal workers are bracing for potential mass

2:23 layoffs as we enter day three of a government shutdown.

2:26 Trump administration's beginning mass layoffs.

2:28 The economy added 172,000 new private sector jobs to kick off the year.

2:35 The United States added 172,000 jobs.

2:38 The private sector.

2:39 Okay.

2:39 So, when you read a headline or listen to a politician

2:43 talking about all of the jobs they have just created,

2:46 the number they are talking about is almost always

2:48 the Bureau of Labor Statistics total non-farm payroll count.

2:51 It is the number that moves bond yields the morning it gets released,

2:54 the number that sets the tone for Federal Reserve meetings,

2:57 and the number that politicians on both sides of the aisle will use to claim

3:01 that the economy is either roaring or collapsing

3:03 depending on what suits them that week.

3:05 The problem is that more and more over time,

3:07 it has slowly stopped describing the thing most

3:10 people actually mean when they hear the word jobs.

3:14 When the BLS says 178,000 jobs were added,

3:17 most people hear 178 previously unemployed

3:20 Americans waking up to a new paycheck, what it actually means is something much,

3:26 much more complicated and uh significantly less reassuring.

3:29 To understand why these numbers have drifted

3:32 so far away from what they sound like,

3:34 you need to look at where they actually come from.

3:37 The Bureau of Labor Statistics doesn't have a magic

3:40 real-time feed of every paycheck in the United States.

3:42 Despite what you might assume, in 2026,

3:44 with the IRS sitting on a real time withholding database that could

3:48 in theory tell you exactly how many workers got paid in any given week,

3:53 that's not the data we use.

3:54 Instead, every single month, the BLS runs two completely separate surveys.

3:59 the establishment survey and the household survey.

4:01 The former surveys individual business sites and offices and gathers

4:05 data on how many people they have on their payroll,

4:07 what those people do for work,

4:09 and where that job site is actually physically located.

4:11 The latter, as the name might suggest,

4:14 surveys hundreds of thousands of households to ask them if they lost their job,

4:18 got a job, remained employed, or unemployed.

4:21 The results of one of them gets printed

4:22 on the front page of the Wall Street Journal,

4:24 while the other one gets quietly buried in the back of the same report.

4:28 Now, logically, in a healthy data ecosystem,

4:30 the two surveys should be moving in roughly the same

4:33 direction most of the time with small disagreements at the margins.

4:36 That used to be the case.

4:38 For most of the post-war period, the establishment number and the household

4:42 number tracked each other reasonably well.

4:44 When one said the labor market was strong, the other almost always agreed.

4:49 However, over the last decade, that relationship has visibly broken down.

4:52 Economists at the San Francisco Fed have

4:55 published research showing that the gap between

4:57 the two surveys has widened to its

4:58 largest level since they started keeping records,

5:01 and the divergence has accelerated specifically in the period after 2020.

5:05 In plain English, the two numbers that are supposed to be measuring the same

5:09 thing are increasingly telling us two completely

5:12 different stories about the American labor market.

5:14 And the one we have decided to focus on is the one

5:17 that structurally is set up to look

5:19 better than reality almost every single month.

5:21 Again, just last month, the payroll survey said we gained 178,000 new jobs,

5:26 while the back of the report,

5:29 households claimed to have lost 64,000, which really just shouldn't be possible.

5:33 Now, you might be thinking, this is the kind of thing the BLS

5:36 would be quietly working to fix in the background.

5:39 And to their credit, in their own technical notes,

5:41 they actually emit a lot of these problems.

5:44 They publish the household number.

5:46 They publish the revisions.

5:47 They publish footnotes on methodology.

5:49 And they publish their own confidence intervals.

5:51 The catch is that nobody who isn't paid to do

5:54 this for a living ever reads any of that.

5:56 Markets react to the headline.

5:58 Politicians react to the headline.

6:00 Business owners deciding whether to hire react to the headline.

6:03 Workers deciding whether their job is safe react to the headline.

6:06 And the headline keeps coming from the survey that has the worst

6:10 structural fit for what the modern American economy actually looks like.

6:14 Which raises the obvious question,

6:16 how has this number become so divorced from reality?

6:19 And why do we even pay attention to it anymore?

6:22 Well, it's time to learn how many works to find

6:24 out what the monthly jobs report is actually measuring,

6:27 why it keeps contradicting itself by hundreds

6:29 of thousands of workers in the exact same report,

6:32 and why nobody actually wants to fix it.

6:35 Every week there's another massive data breach,

6:37 another AI model scraping personal data,

6:39 or another company rewriting who owns yours.

6:42 If that makes you uneasy, you're not alone.

6:44 And that's why this video is sponsored by Proton.

6:47 Proton is an endto-end encrypted platform built

6:49 by CERN scientists who wanted a more private internet.

6:52 Over 100 million people use it now,

6:54 and the free plan gets you Proton Mail for secure

6:57 email and Proton Drive for encrypted cloud storage in one account.

7:00 What I actually like is how easy it makes privacy.

7:03 Your emails are encrypted so only you and the recipient can read them.

7:07 Your files are encrypted before they leave your device.

7:10 And because Proton is based in Switzerland,

7:12 your data is covered by some of the strongest privacy laws in the world.

7:16 You can even create email aliases to cut down on spam and fishing.

7:20 And Proton Drive works like any other cloud storage,

7:23 just without anyone, not even Proton, seeing your stuff.

7:25 If you want to take control of your digital life,

7:28 you can get Proton Mail and Proton Drive together for free.

7:31 Check out the link in my description to get started.

7:33 Okay, so it is easy and honestly a little

7:36 bit understandable to look at these consistently inaccurate economic metrics

7:40 and conclude that the agencies putting them together are just

7:43 trying to make the politicians they answer to look good.

7:46 However, at the very least, before we attribute this all to malice,

7:49 we should understand how it is much

7:51 more likely to just be statistical stupidity.

7:53 And to that end, there are really three big

7:56 problems with how we come up with this number.

7:58 And the first simple problem is that payrolls are not people.

8:02 The establishment survey does exactly what the name suggests.

8:06 It surveys establishments and asks them

8:08 how many payrolls they process that month.

8:10 At the time this data source was first conducted,

8:13 it was a perfectly reasonable proxy for how many people have jobs.

8:18 One person, one employer, one payroll.

8:20 This is still the norm for most employees,

8:23 but it's getting less and less normal every year.

8:25 These days, a single person can be attached to multiple payrolls.

8:28 And the number going up doesn't necessarily mean a new person got a new job.

8:33 It might just mean an existing worker picked

8:35 up a second one to keep up with rent.

8:37 The most extreme version of this shows up in healthcare,

8:40 where it's pretty common place for doctors

8:42 and nurses to work across multiple different establishments.

8:44 A single doctor working at a hospital, a private practice,

8:47 and doing some research on the side could technically add three

8:50 payrolls to the official number while only being one single working person.

8:55 This technicality was particularly important for last

8:57 month's job numbers because it also counted

8:59 35,000 healthcare professionals returning to work after

9:02 having been on strike the month before.

9:05 Hopefully, it should go without saying that this isn't the same

9:07 thing as the economy genuinely providing

9:09 employment opportunities to 35,000 new people.

9:11 But all this data does is look at payrolls every month.

9:15 So, the establishment survey says we added 178,000 payrolls.

9:19 The household survey in the same report on the same

9:23 day says 64,000 fewer humans actually have work.

9:25 Both of these numbers are technically true.

9:28 They are just measuring different things and almost nobody outside

9:31 of macronerds ever bothers to look at the second one.

9:34 Which brings us to problem number two.

9:37 The methodology.

9:37 Collecting payroll data at the scale

9:40 of the entire American economy is genuinely impossible.

9:43 And the model the BLS uses to do it is becoming increasingly outdated.

9:47 Payrolls are not counted firm to firm but rather location to location.

9:52 Again, the establishment in the establishment survey

9:54 means a survey of individual well establishments.

9:57 So, as more and more people move into jobs that are work from home,

10:02 jobs that move between cities or jobs that run out

10:04 of a head office on the other side of the country,

10:07 it has become much harder to accurately assign

10:09 those payrolls to a real workplace and monitor them monthtomonth.

10:12 Double counting, undercounting,

10:14 and plain old misunderstandings about what payroll should count have just

10:17 become far more common as our working arrangements have become less rigid.

10:21 This is part of the reason why revisions have gotten so embarrassingly large.

10:25 Last year alone, the BLS quietly revised away

10:29 800,000 jobs that they had previously claimed were added,

10:31 meaning the headline number for an entire chunk of the year was just wrong.

10:36 You can actually go and look up the exact collection

10:39 methodology that the BLS uses on their website and you

10:41 can see it was designed for a time when

10:44 most people either went to a factory or an office,

10:46 did their hours, and went home every day.

10:48 To try and balance this out,

10:50 the BLS also uses something called the bursts and death model,

10:53 which is essentially a statistical estimate of how

10:56 many businesses they think were created or shut

10:58 down in any given month used to fill

11:00 in the gaps that the survey itself can't capture.

11:02 The births and deaths model has its own problems,

11:05 but the biggest one is that it is

11:07 heavily skewed by the proliferation of gig work,

11:09 which we have mentioned a few times in previous videos.

11:12 A new business in this model can be

11:14 a guy who started driving for Uber last week.

11:17 That gets counted as a new job creating enterprise,

11:20 even though it is functionally a person

11:22 without stable employment trying to make rent.

11:24 So, the survey itself misses the modern workforce.

11:26 And the model designed to patch the survey is

11:29 also calibrated for a workforce that doesn't really exist anymore.

11:32 Now, these growing blind spots in one of the most important

11:36 metrics in our economy is not great, but it gets worse.

11:40 The third problem is by far the biggest one,

11:42 and that is the people the payroll survey doesn't see at all.

11:46 The establishment survey by its very nature

11:49 tends to count people in traditional payroll jobs,

11:51 which on average means more secure positions with more

11:55 established employers who actually respond to voluntary BLS requests.

11:58 An increasingly large share of the American

12:00 workforce does not fit any of those descriptions.

12:03 They are private contractors.

12:05 And this is not just gig workers.

12:08 This includes everyone from a guy delivering groceries through Door

12:10 Dash all the way up to extremely high-end technical roles,

12:14 consultants, freelance software engineers, and well,

12:17 yeah, basically the entire creator economy.

12:19 Can't forget about the real heroes.

12:21 These people get counted as employed or unemployed by the household survey.

12:25 And if they lose their work,

12:27 it still sucks just as much as anybody else losing their job.

12:30 But they are completely invisible to payroll.

12:32 They are also conveniently the people most likely to lose

12:36 their work before anybody else in more secure formal roles.

12:39 When a company starts to feel a downturn coming,

12:41 the first line items to get cut are

12:44 not the staff with employment contracts and severance packages.

12:46 The first to go are the contractors,

12:49 the freelancers, the agency the marketing team was using,

12:52 the consultants, and the uh inoff life coaches

12:54 that teach the CEO how to act human.

12:57 None of those people show up as layoffs on press releases.

13:00 None of them show up in the establishment

13:02 survey and it can take weeks or months before

13:05 they show up in unemployment claims because most

13:07 of them aren't eligible to file in the first place.

13:09 This is how the labor force can fall by almost 400,000

13:12 people in a single month while payrolls technically still go up.

13:16 The two surveys are just capturing two very different groups

13:20 of workers and those groups are drifting further apart every year.

13:23 One group is on a W2 in an HR system with their payroll

13:27 processed by a service the BLS already knows how to call.

13:31 The other is invoicing clients paying their own taxes

13:34 and disappearing from the data the second the contract dries up.

13:37 Which raises a perfectly reasonable question.

13:40 If the payroll survey is flawed, why do we still use it?

13:44 Well, two reasons.

13:45 The first is that it's fast.

13:47 It can give us monthly data with comparatively

13:50 little collection which matters for an economy

13:52 where the Fed and the government need

13:55 to make decisions on a near real-time basis.

13:57 The second is that the payroll survey is reliableishly unreliable.

14:01 Businesses are not exactly motivated to make

14:05 up payrolls to boost national statistics.

14:08 Whereas individual households can be koi about

14:11 their employment situation for all kinds of reasons.

14:13 People who have a job are less likely to answer

14:15 the phone for a BLS survey because they are well working.

14:19 People who just lost one might tailor their response out of embarrassment

14:23 or genuine confusion about what counts as actively looking for work.

14:27 There is also a less talked about timing problem.

14:30 The household survey covers the week containing the 12th of the month.

14:33 So, a person who loses their job on the 13th keeps counting as employed

14:37 for another 4 weeks before the data even has a chance to catch up.

14:40 So, the household survey is not some pristine

14:42 alternative the BLS is just refusing to use.

14:45 It has its own real problems.

14:47 Now, I know this again sounds like boring money

14:50 man rants about data collection methodologies for 15 minutes straight.

14:54 But this really does matter for two reasons.

14:56 The first is that of course these numbers

14:58 get used to make real decisions that affect us.

15:01 Federal Reserve policy on interest rates,

15:03 stimulus debates in Congress, how foreign investors price treasury yields,

15:07 and ironically, it even feeds into a business's own plans to hire.

15:11 If companies see unemployment rising and headline payrolls falling,

15:15 they become more riskaverse,

15:16 which leads to less hiring, which leads to less consumer spending,

15:20 which leads to even less hiring.

15:22 So, it's just as important to make sure that these numbers don't look

15:25 artificially pessimistic as it is to make

15:28 sure they don't look artificially optimistic.

15:30 The second reason this is so important to understand

15:33 is that if these numbers keep looking like total nonsense,

15:35 it becomes easier for people to dismiss them

15:38 as useless at best or outright propaganda at worst.

15:41 Not having faith in our economic statistics is a great

15:44 way to undermine faith in the economy as a whole.

15:47 Now, the economists at the BLS know all of this.

15:50 They have written papers about it.

15:51 Statistics agencies in other developed

15:53 economies have already started rebalancing.

15:55 The reason the United States hasn't is structural.

15:59 Changing the methodology would create

16:01 a one-time discontinuity in the data series,

16:04 which makes year-over-year comparisons look weird, which the market hates.

16:08 It would also require the BLS to admit on the record that the number

16:12 it has been publishing for decades isn't

16:14 really measuring what people thought it was,

16:16 which is the kind of emission federal agencies do not love making

16:19 in front of a Congress that is already gunning to cut their funding.

16:22 And on top of all of that, every administration in power has

16:25 a quiet incentive to leave a politically

16:27 convenient noise generator exactly where it is.

16:29 A monthly number that can swing 200,000 jobs

16:32 in either direction based on which doctors happen

16:34 to be on strike that month is incredibly useful

16:37 if you are the one writing the press release.

16:40 So the number stays, the methodology stays, the revisions get larger,

16:43 and the gap between the headline and the lived

16:46 experience of the labor market keeps widening,

16:48 which means for the foreseeable future,

16:50 the most quoted economic statistic in the United States is going

16:54 to keep telling a story that fewer and fewer working people actually recognize.

16:58 Now, if all of that wasn't enough,

17:00 go and watch this video next to further understand how to lie with statistics.

17:03 And don't forget to like and subscribe to keep on learning how money works.