Opening the door or closing it?
President Donald Trump just fired the US Commissioner of Labor Statistics, Erika McEntarfer, for revisions that substantially reduced this summer’s employment estimates. For May and June, the reduction was 258,000. That’s a big change indeed. Trump alleges, unencumbered as usual by evidence, that McEntarfer overestimated employment just before the November 2024 election to help Kamala Harris win.
Trump continues: “Important numbers like this must be fair
and accurate, they can't be manipulated for political purposes.” He’s right
about that. But the question is who is manipulating what. Below are the real job numbers, not Trump’s fantasies.
Figure 1 shows that employment has been remarkably stable for more than 20
years. We focus here on July in each year because employment varies with the
season. For example, construction jobs abound in the summer but not in the
winter. By concentrating on July, we get a good picture of what’s been
happening over the years.
Employment fell sharply in the Covid-19 lockdowns, of course. But it recovered
in two years. In Figure 1, you can see that employment has resumed the long-run
pace of growth that prevailed before the pandemic. That growth is modest. This
is not Trump bashing. It’s a simple fact that preceded Trump.
Figure 1
https://www.bls.gov/charts/employment-situation/civilian-employment.htm
The Bureau of Labor Statistics calculated Figure 1 from a monthly survey of
60,000 households by the Census Bureau. The Bureau of Labor Statistics also
surveys 631,000 worksites -- a third of all nonfarm payroll jobs -- monthly to
double-check on its estimates. Figure 2 shows the establishment results.
The pattern is like the one observed from the household
survey and shown in Figure 1. After a
sharp decline during the lockdowns, employment returned to its long-run
pace. Incidentally, note that in both charts, employment drops in recessions,
shown by the vertical gray bars. That’s the Keynesian reason that the
government borrows and spends in slowdowns – to create jobs until the economy
recovers to the point that it can generate jobs on its own. But the main point
is that recent employment is consistent with the trend of the past 20 years. If
Trump was correct – that the Bureau is shaving the numbers to make him look bad
– we should see a break in the data, like what happened in the deep recessions
of 2020 and 2008.
Figure 2
https://www.bls.gov/charts/employment-situation/employment-levels-by-industry.htm
Figure 3 suggests why employment is not growing as rapidly as Trump pretends:
Most adults who are able and willing to work already have jobs. In fact, the
rate at which people participate in the labor force has been falling for two
decades. Why? Well, for one thing, Americans are aging. Most old folks would
rather retire than work.
By definition, the labor force includes both those who work and those who are
looking for work. But the unemployment rate – the share of the labor force of
people looking for work – is only 4%, the lowest it’s been for 20 years, as
Figure 4 shows. So, a long-run reason for why employment grows slowly may be
that firms have trouble finding folks to hire. In that case, deporting
immigrants won’t help, because most of them work.
Figure 3
https://www.bls.gov/charts/employment-situation/civilian-unemployment-rate.ht
Figure 4
https://www.bls.gov/charts/employment-situation/civilian-unemployment-rate.htm
Figure 5 illustrates what the slow growth in jobs is doing to us. As wages
rise, employment slows. Industries that are desperate for workers, like
utilities, are bidding up their wages to the point where they can no longer
afford to hire them. Jobs are growing more rapidly in lower-wage industries
like retail, health care, and education.
My untutored hunch is that the long-run labor market is
tight because of a lack of workers. Goosing the market by stimulating demand
for workers -- for example, by cutting interest rates so that firms can borrow
and spend more – may raise wages and prices rather than employment, in my
opinion.
There’s lots of room for disagreement. In my view, the basic problem is excess
aggregate demand. The economy is already producing about all that it
comfortably can – hence the low rate of unemployment. People are trying to buy
more output than that, so they bid up prices. If I’m right, prices should rise
and output should slow down until the excess demand disappears. But I may not
be right. Another view is that the labor market is cooling because of a fall in
aggregate demand. People are demanding less output, not more. In that case, both
inflation and output might slow down. And it is indeed true that inflation has
halved in about a year.
Always fightin’
Well, OK. Economists love to argue, even if you don’t love
to listen. But one thing is clear: Trump is shooting the messenger. There is no
evidence that the Bureau is faking the employment numbers. Instead, it is
describing the economy with dispassionate accuracy. If you would like the
lowdown yourself, ignore the White House and read The Wall Street Journal.
It is natural to ask why job updates differ from the initial estimates. I will
explain. The Bureau of Labor Statistics surveys more than a half-million firms
each month about payroll employment; the deadline for the reports is the 12th.
Many firms haven't finished their payrolls by the 12th. So the Bureau includes
their estimates in two updates over the next two months. The third and final
estimate includes 95% of the worksites. The first estimate may include 60% or
70% -- the share has been falling for more than a decade. https://www.bls.gov/opub/btn/volume-2/revisions-to-jobs-numbers.htm
If economic conditions haven't changed between the first two estimates, then
the second estimate will look like the first. That is, it will be predictable.
If conditions have changed, the second estimate can differ substantially. In the
case at hand, the estimate was unexpectedly small. Perhaps the labor market is
cooling. Firms might have delayed hiring until they knew what was up with Trump’s
whimsical tariffs.
Another factor in the revisions is seasonal adjustment. For
instance, if the weather is changing, its effect on employment will also
change. We want to adjust for such factors because we want to know whether
employment is changing because of them or because of a permanent shift in the
economy itself, such as in the number of women who work.
For instance, in June, youths are no longer in school, so
they look for temporary jobs. This can obscure more subtle changes in the labor
force if we do not adjust for it. https://www.bls.gov/news.release/empsit.nr0.htm
What’s up with an update
In short, the difference in the updates is not evidence that the Bureau is
incompetent or corrupt, as Trump claims. To the contrary. The Bureau is
providing the latest data so that we can understand how the economy is
changing.
Why does Trump conclude that the Bureau is lying? Truth Social:
“The Economy is BOOMING under 'TRUMP' despite a Fed that also plays games, this
time with Interest Rates, where they lowered them twice, and substantially,
just before the Presidential Election, I assume in the hopes of getting
'Kamala' elected – How did that work out? Jerome 'Too Late' Powell should also
be put ‘out to pasture.’” As usual, when Trump wants to make an iffy point, he changes the subject as fast as he can. Keep your eye on the ball: His basic claim is that the economy is booming
because the Trump brand is magic. Therefore it will always spin off jobs. Um, he forgot to mention the tariffs.
But how do we know that the Bureau is not lying?
Well, by statistical measures, its estimates are quite accurate. It uses
samples of households and establishments, because surveying every American
worker every month would cost too much. DOGE would have a fit. No sample is exactly
like the population from which it is drawn. However, the sampling errors in the
Bureau's estimates are small. One can have 90% confidence in them, in the sense
that the estimates come from an interval that includes the true estimates 90%
of the time. In other words, the Bureau's estimate may not be exactly
correct, but it is pretty darn close. https://www.bls.gov/news.release/empsit.tn.htm
Moreover, the employment numbers would be hard for anyone to
fake, Democrat or Republican. The
financial markets follow those numbers religiously. The bloggers would vet any
unexpected change in employment. It could be cross-checked against changes in
unemployment compensation filings in the 50 states.
Kurtosis and all that
It would also
create statistical aberrations. In samples as large as the BLS uses for
employment, random changes follow a normal probability distribution with
well-known properties. Skewness is close to zero, and the height of the
distribution (which relates to a characteristic called kurtosis) is not
exaggerated (its kurtosis is close to 4). Inflating the employment numbers
would skew the distribution to the right and increase the relative frequency of
those values, which affects kurtosis.
The chances that skewness and kurtosis change by accident can be precisely estimated with simple tests of the distribution's normality, such as the Jarque-Bera test. To pass the tests, the scammer would have to fake the whole dataset, not only for total civilian employment but for at least a dozen related statistics (employment to population, the labor force participation rate, etc.). The whistleblowers would have a field day. At bls.gov, you will see that the Bureau publishes datasets as well as graphs.
For those reasons, I don't stay up nights worrying about fudging. I do worry that Trump's meddling will cause the Bureau's highly competent statisticians to leave. After all, they would have no trouble finding more gainful employment elsewhere. The quality of the government labor statistics could take a nose dive. The global implications could be momentous. There would be greater uncertainty about economic forecasts, for states budgeting unemployment compensation, for businesses predicting market demand for their products...you name it.
In principle, the President, as head of the executive
branch, has great latitude in hiring or firing executive officials. But the
statistical agencies, like the BLS, the Bureau of Economic Analysis in the
Commerce Department, the Economic Research Service in the Agriculture
Department, the Census Bureau, and so on, are supposed to be sacrosanct. They
are not political. They just provide the hard data that drive policy and political
decisions. As the econometrician Mark Kennet says, it's hard to see how those
decisions can be correct if they can't draw upon accurate statistics.
Notes: For useful comments, I thank but do not
implicate Mark Kennet, Paul Higgins, Barry Lenk, Steve Knott, and Kevin Morgan.
Figure 5
https://www.bls.gov/charts/employment-situation/employment-and-average-hourly-earnings-by-industry-bubble.htm
https://www.msn.com/en-us/money/markets/donald-trump-fires-person-behind-jobs-numbers-after-they-re-revised-down/ar-AA1JKiRA?ocid=msedgdhp&pc=EDGEXST&cvid=89abf8747fa44907acc9552b13e36c2a&ei=21