The Job Numbers Will Probably Get Revised—That’s Not A Scandal | The Professor Is In

Yeah, wow, I love how nerdy these comments. Platypus Economics
has the biggest nerds on the internet.
Can we actually believe the jobs report? You released a
video arguing that we should, and I want to hear
more about that. I'm Megan Connor's.
And I'm justin Wolfus. This is the professor is in
where I'm here to take your questions.
So, as you noted, we've now had three solid months
of jobs growth, and yet the unemployment rate hasn't really
budged over that time. Can you explain what's going on
there and what we know?
Yeah? Really good question. And what's nice about that is
you're already showing. One of the ways we check whether
we can believe data is say, is one set of
data jobs growth consistent with another set of data, what's
going on with the unemployment rate? Let me introduce a concept.
Economists call it the break even job's growth rate. Essentially,
this says, how many jobs do we have to create
each month to keep the unemployment rate stable? Now, it
turns out it's quite a few, and it's quite a
few because the American population is growing, and so therefore
the number of workers is growing. So to keep the
same number of people, the same share of us in work.
We've got to keep running fast. We're on a treadmill now.
Historically that number was actually very high. We had to
create one hundred and fifty thousand jobs a month. One
thing that's really important is we used to think we
understood what that number was with some precision. But the
Trump administration, by virtue of pulling back on immigration, has
fundamentally changed the rate of population growth very very quickly,
and we may actually be flatlining for the first time
in generations. And so we are less certain what the
break even rate is. So let me come back now
and try and make sense of everything that we've seen.
You could say job's growth was pretty strong, unemployment didn't
move much, and I might say, well, that tells us
that the break even jobs growth number is maybe a
little higher than we thought. It's definitely going to be
above zero. Let me give you a very quick bit
of mental arithmetic. Ish, instead of looking over for three months,
are going to look over the entire Trump administration. Over
that period, We've created about seven hundred thousand jobs. So
that's over about seventeen months. That's around about forty thousand
jobs a month. And over that period, the unemployment rate
was rising for a bit and now it started the
flat line. Reason maybe a quarter of points, so that
says break even jobs growth is probably round about forty thousand.
Some thought it was lower, some thought it was higher.
So the last three months we've been creating an absolute
bucket load of jobs. If that were to continue, you
should expect the unemployment rate to decline.
Okay, So we've also we had a lot of questions
and comments and kind of speculation about what type of
jobs these might actually be, so kind of bell into
two buckets. One group asking about like whether they are
low or high paid jobs, and we also had people
speculating whether these might be second or third jobs that
people are taking on in order to deal with the
increased cost of living. What can this report tell us
in that regard and what can it not?
Great question? And I'm laugh at you a little bit
because I think it's a really good version of this
question and one that I kind of want to frown
and say, be better. Every time there's a jobs report
that people don't like, they'll say, and I see this
on social media all the time, well these were crappy
jobs or people have to take it at second jobs,
or on and on and on a case. The reality
is they're almost always making that up. So the most
important numbers come from account of non fun pay rolls.
We basically call a bunch of companies and they say,
how many checks did you print last month for workers
right or last fortnight. Now realize that doesn't tell you
much at all about the quality of a job. It's
literally the checks they printed. So it doesn't tell you
if it's a high wage job or a low wage job,
if the worker is black or white, tall or short.
It doesn't tell you whether it's a forty hour a
week job or a twenty hour a week job. It
doesn't tell you if it's a first job or a
second job. We know what industry it's. That's helpful. We
know what's happening to average wages. There are some industries
if you're seeing job growth, you know, in say retail
and restaurants, you might have your suspicions that these are
not going to be really incredibly important jobs. Whereas if
we're seeing stuff in professional and business services, these feel
like jobs you might want your kids to grow up
and take. So there's a little bit of detail but
not very much.
Are there other data sources, federal data sources that can
illuminate any of this.
So what happens on the first Friday of each month
is the payroll survey. We all pay a lot of
attention to that. The reason is it's more signal than noise.
There's a second survey where they go and talk to
sixty thousand households and ask them a bunch of questions. Megan,
what's your gender, what's your age, what's your education, did
your work last week? How many hours did you work?
Do you have more than one job? And so on?
They asked that of sixty thousand. The problem is for
month to month movements, that's more noise than signal. If
we package a whole lot of those surveys together, they say,
and look at a whole year worth of data. Now
I've got enough to say something quite meaningful. And so
you can look at those numbers and say things like
jobs are going to highly educated people, less so to
high school dropouts. You can look at differences and occupations,
you can look at multiple job holding and so on.
Anytime someone's mentioning any of these details on a month
to month basis, they're basically looking at noise rather than signal.
A simple rule of thumb is almost anything you read
on social media about these details turns out to be false.
The labor market this year looks a lot like what
it did last year. Trends tend to be longer run,
rather than month to month or year to year, or
even presidential term to presidential term.
Okay, we also had a lot of people asking about revisions,
why revisions occur, whether they should impact the way we
think about these numbers or trust these numbers. And I
was hoping you could talk more about revisions.
Yeah, wow, I love how nerdy these comments. Platypus Economics
has the biggest nerds on the internet. Love that, Love
that for all of us. Okay. Revisions are when you
say I have an initial first guess, like by the
way weather forecasters have revisions. They say, next week, I
think it's going to be sunny, and then as we
get closer to next Friday, they say, well, I'm not
so sure, and then by next Thursday they say maybe
it'll rain. They're revising their forecasts. You might say, well,
there's a forecast here, we're talking about data, but actually
we never actually know what's happening right now, because we
don't know what every person in the economy is ever doing.
So one way of thinking about measuring the current state
of economy the economy is, let me use a different
word and now cast. Instead of forecast in the future,
we're casting what's happening right now, and as we get
more information, we should change our views. It's as simple
as that. Now, this became highly politicized because the President
would say, look, the BLS Bureau of Labor Statistics has
changed its best estimate for what happened last month. That
must mean they're corrupt. No, not corrupt. The opposite means
they're committed to accuracy. Let me explain the way we
put together these numbers is we send out surveys. This
all done digitally, by the way, but it's easy to
think about it if it's physical to a whole bunch
of firms, and then they fill in a form again,
they link their payroll records to those of the BLS,
which says, this is how many people we employed last month.
But it turns out some people have I know this
will surprise you. Some people have better things to do
than fill out forms. And so what will happen is
they need to publish the numbers, but they've only got
three quarters of the forms back. Well, one thing they
could do is just not publish a number and wait
another few weeks before they publish, But then the data
is not going to be very timely, and the FED
needs to make decisions, and heck, you've got three quarters
of the forms. Tell us what's going on, And so
that's what they do, and that's what's called the first estimate.
And then over the next month a few more forms
come in, so now you've got all the earlier data
plus a bit more. So this definitely is a better
estimate of what's going on, but it means you have
to change your estimate of what you think happened, say
in May, and then a month after that, even more
forms come in. And so that's the process of revisions.
You can see hands are above the table. The computer
programs are already written. By the way, there's no human
intervention involved in any part of this. The revised numbers
are always tend to be more accurate. Sometimes, if the
revisions are up three times in a row, some conspiracy
theorists will say it's because this side of politics or
that side of politics dabbled. But do you know what,
I've flipped a coin three times in a row before,
and sometimes it's come up heads three times in a row.
And that's pretty much how that works. That's the first
set of revisions. Now, the data we look at versus
the data we collect. The data we all spend our
time looking AT's what's called seasonally adjusted data. Where what
we do is we say, how is May is hiring
compared to what May normally looks like. You see, if
we didn't do that, we'd normally say there's an enormous
boom every Christmas because of Christmas shopping, and then the
FED would raise rates every Christmas and you know it
would be ridiculous. So what we want to do is say,
how are we doing relative to what you would expect
given the current time of year. Now again figuring that out,
how big of a boom should we expect around Christmas?
How many trading days were there? Are people doing more
shopping with Black Friday? How did Prime Day change everything
from that very important public holiday in my house? Estimating
seasonal factors, the seasonal adjustment, that's something we redo every year,
and we should because we're learning more about the ups
and downs of the seasonal cycle of years. So that's
the second part. And there's one more thing I know
that's exciting for you. It's not called a revision. It's
called a benchmark, but for folks at home, it's going
to feel like a revision. Here's what's going on. Let's
say I send out forms to one in one hundred firms,
and then I get those forms back and it says
and those one on one hundred firms collectively they employ
million people. So therefore, if I survey all firms, I
reckon there'd be one hundred million people employed. So what
I have to do is scale up my survey responses
by what share of the population they were. Here's the
hard thing. If I send out a whole bunch of forms,
am I sending them to one percent of all firms
or one point two percent? Well, it depends how many
firms there are, which I also don't know. And so
the benchmark is when they basically go back and say, oh,
we finally figured out how many firms.
There are, and how do they where does that information
come from?
Oh, my goodness, deep in the bowels of the Census Bureau.
For that, so the Census Bureau. We all love the
Census Bureau. It's harder to keep track of the number
of firms. But basically, you figure out a sampling, you
figure out how many firms there are, and then you
keep track of how many new ones are being born
and how many die. And the problem is that's a
very difficult task to do in real time because when
a firm dies, surprisingly enough, it doesn't have a funeral
and it doesn't send out a death notice, and so
it takes a while to figure out which one's died.
And also, when a firm is born, no one tells
them right away. Actually, quick side note, Platypus Economics is
a new firm that was just born, but it's not
yet clear the federal government knows about that, which, by
the way, means this is an example of a benchmark revision.
Megan is employed by Platypus Economics, so that means Meghan
might not be counted in non farm payrolls for a
little while. Once the government discovers Platipus Economics, she'll be counted,
And that's the benchmarking process. I have never been asked
to go that deep.
Help someone out there is appreciating this. Okay, So we
also had some people really concerned about why these numbers
aren't matching perfectly our private sector data. I think some
of what you've already said is maybe illuminated that question,
but you could speak to it more directly, like ADP.
I think their number was one hundred and twenty two
thousand for May. Why should we not expect those numbers
to perfectly match?
I want folks at home to actually start by understanding
just how enormously difficult it is to track the number
of people who are employed in America. This is a
huge economy, a huge country. We have I think roughly
one hundred and thirty million people on non farm payrolls.
I hope that's right. I take that back, roughly one
hundred and sixty million. Actually, we have roughly one hundred
and sixty million people on non farm payrolls. How would
you measure that? If you could get pretty close, that
would be pretty good. Imagine that we could get within
one percent of the true number. Well, that would mean woulday,
there's one hundred and sixty million people plus or minus
one point six million we're trying to do when we
measure monthly payrolls growth is spectacularly more difficult because the
change in the number of people and non fun payrolls
from one month to the next might be one hundred thousand,
which means that we have to be able to estimate
this with an incredible degree of accuracy. It's a very
very hard to one hundred thousand as a proportion of
one hundred and sixty million is very very small, and
people get really upset when the Bureau of Labor Statistics
can't tell you whether payrolls grew by sixty thousand or
one hundred thousand. That's the most important thing. It's a
really hard task. What we then do as we throw
all the resources the federal government, the Bureau of Labor Statistics,
and the best statistical nerds in the country at it,
and we call that the BLS Estimate of non farm Payrolls.
What also happens is other people take their best guesses,
but their best guesses aren't linked to benchmarks. They don't
have the statistical wizardry of the BLS, they don't have
tens of thousands of surveys going out, and so it
turns out their estimates are way worse. Still, okay, but
way worse. The margin of error on the month month
movement non fun payrolls is round about one hundred thousand.
If you looked at that for these other measures, it
would be much larger, like two hundred three hundred thousand.
They're never going to match up exactly. But the reason
for that is it's not that the non fun payrolls
is bad, it's these other things are worse.
And that difference fifty thousand is within the margin of
error anyway.
Yes, yes, So when I say the alternative measures are useful,
they're much more useful, not for months to month, but
over a long run. Right, So if one president came
to power and started fiddling with the numbers, you would
start to see these drift apart, and that would cumulative
drift over several years would be quite large. Figuring out
what's going on in may look, honestly, your best betch
just looking at non farm.
Payrolls, and they're directionally similar to I would assume if
we're you know, one's positive and one's negative by a
large amount, that would be more concerning.
But again, not month to month, right, Okay, month to
month there's a lot of noise, and so sometimes payrolls,
flips hair and IDP flips tiles.
Okay, so you ended your last video about this saying
that these numbers, particularly since it's been three months in
a row, leads you to kind of want to calm
people's recession fears and panic. However, the same day we
release that video, we got, you know, a new CPI
report showing inflation being at its highest level I think
in three years, and a lot of people wanted to
know whether that changed your mind about things, And I
was wondering if you could like talk about what is
the relationship between inflation and recession risks and has your
mind shifted in any way?
Really good question, because when you look at measures of
consumer sentiment, how people are feeling about the economy. How
people say they're feeling usually reflects in some measure. Is
it easy enough to find work? Am I getting pay rises?
Stuff like that? And how upset am I about inflation?
And so when you ask people their feelings, often I say, oh,
the economy feels dreadful, even though we're creating lots of jobs.
Not to be clear, what is a recession. A recession
as a generalized decline in the level of economic activity
so it's not just one sector, it's across a whole
bunch of sectors. You know, it's various indicators, and it's
that the economy is actually shrinking. The thing to notice
about that is that it makes no reference whatsoever to prices.
So the world can suck because of high inflation. But
we have a word for that. We call that inflation.
We don't call that a recession. This was a really
big deal post COVID, remember that COVID inflationary burst. Lots
of people calling it a recession. It's not. It's an inflation.
It sucks. We just have a different name for it.
So let me now come back. I want to make
two points. One the video that we may argue that
we could believe the non farm payrolls numbers coming out
of the Bureau Labor Statistics. They also put out the
consumer price index, which was this new indicator. So for
exactly the same reasons, I believe the inflation data. And
then the other thing is to the extent that the
inflation data tells us anything about the state of the economy.
Generally speaking, inflation heats up when the economy doing moderately well.
It would be a little bit more of a puzzle.
If the economy were in a recession and inflation were
heating up, I want to be careful even on that,
I said, a little bit more of a puzzle. Most
people are used to thinking about the world in terms
of what economists called demand shocks. A demand shock is
a lot of people want to buy a lot of stuff,
so therefore employers hire a lot of people, unemployment falls.
That side feels good. But so many people want to
buy so much stuff, prices rise, so they're used to
thinking about real economic activity goes up, causing inflation to
go up, and that's what we call the Phillips curve.
But there's also something else that goes on, this different
kind of shock called a supply shock. That's when the
cost of doing business rises. For instance, if you happen
to close the strait up Homus, or if you happen
to impose tariffs, or if we happen to be in
the middle of a global pandemic. When that happens, you
can get two bad things happening at the same time,
which is slowing economic activity and rising inflation. So they
can go together. But you know there, I think the
inflation number is no real surprise to anyone who's been
watching the economy.
Well, thank you for join to me today. To our audience,
we hope you enjoyed this episode of the Professor is in.
If you'd like your question to be answered in the
future segment, just leave a comment wherever you are listening
or watching or reading all of them. Make sure to
like and subscribe Platypus Economics on YouTube, substack, and all
social media