There’s quite a bit of chatter about the Bureau of Labor Statistics’s (BLS) most recent Labor Situation Report yesterday, President Trump’s remarks, and the attendant firing of the BLS’s chief. Fairly representative is Sylvan Lane’s and Tobias Burns’s report at The Hill:
The U.S. added only 73,000 jobs in July and the unemployment rate ticked slightly higher to 4.2 percent, according to data released Friday by the Labor Department.
The July jobs report showed the labor market stalling out as consumers and businesses navigated President Trump’s ever-evolving trade policies and steep new tariffs.
The report came in well below the expectations of economists, most of whom projected job gains of at least 100,000 in July, according to consensus estimates.
Job growth in May and June was also far lower than first reported, according to the Labor Department, which shaved 258,000 jobs off of its past two reports.
Add the “natural increase” in the labor force and legal immigration and that means that the number of jobs added has fallen below the number of people entering the labor force for the last two quarters. It’s no wonder the labor force participation rate is declining.
Twenty years ago I followed the BLS’s monthly labor situation reports rather breathlessly. The problems I found was that the reports depended less on actual empirical data and increasingly on the several adjustment factors applied, e.g. the birth-death ratio adjustment, the seasonal adjustment, and the population adjustment. I find that methodologically suspect. Each of those adjustment factors depends on certain historical assumptions. If the assumptions no longer hold true, then the reports will be incorrect not just in detail but possibly directionally. That’s true whenever the data that are actually being measured are outweighed by the adjustment factors being applied as has been the case for some time.
I suspect that under present circumstances the population adjustment is particularly suspect.
Furthermore, the data used for the situation report are derived from two distinct surveys, the household survey and the establishment survey. It is an attempted accommodation of the two and they have deviated considerably from each other for some time:
Consequently, while I think that President Trump is correct to be suspicious of the monthly report, he was wrong to discharge the BLS chief over the report because to whatever extent the report was “fudged” it was actually fudged in his favor. And it wasn’t the most recent month that was the main problem. The main problem was the very large adjustments to prior months. These reports are supposed to aid in policy formation. It doesn’t help when the response to prior quarter’s numbers are “nevermind”.
The folks at Zerohedge seem to think that the divergence of the surveys is a good thing and reflects the Trump Administration’s crackdown on illegal immigration and deportation of migrants. That may be a factor but I doubt it’s the only one in the labor situation.
Six of the “Magnificent Seven” have conducted substantial reductions in force over the last two quarter to a total of nearly 50,000 employees. When there’s that might RIF’ing in the technology sector over such a short period and companies, essentially, aren’t hiring, that is bound to create substantial uncertainty. In the latest sitrep only the most highly subsidized sectors (government and healthcare) were hiring.
Add to that the tariff situation which fluctuates nearly on a daily basis and other sources of uncertainty and it’s a lot of uncertainty. Businesses are unlikely to take on new employees under situations of such uncertainty. And as things look now that uncertainty is likely to persist for the next three and a half years at least.
Dave Schuler: It doesn’t help when the response to prior quarter’s numbers are “nevermind”.
You mean the initial reports of a given quarter are less accurate than later reports based on more complete data. That’s inevitable.
https://pbs.twimg.com/media/GxR3RraXQAATxYn?format=jpg
Dave Schuler: Consequently, while I think that President Trump is correct to be suspicious of the monthly report, he was wrong to discharge the BLS chief over the report because to whatever extent the report was “fudged” it was actually fudged in his favor.
It doesn’t matter in whose favor it was adjusted. That would just make the result political rather than analytical.
Of primary importance, the process is transparent. Nor is it “fudged” to make the president look good or bad, but statistical techniques are used to create the most accurate results possible with necessarily limited data.
If there is a better way to determine the required information, then have at it. But Trump actually got rid of the expert advisory group that had been looking at how to improve the process. And, as pointed out before, the initial report based on incomplete information will necessarily be less accurate—and it will be even more inaccurate when the economy is in a state of rapid flux. Any statistical process will have a larger margin of error when the derivative of change is on a significantly shorter timescale than the survey itself. The current rate of chaotic change is the rate of Trump’s whims.
Dave Schuler: Furthermore, the data used for the situation report are derived from two distinct surveys, the household survey and the establishment survey. It is an attempted accommodation of the two and they have deviated considerably from each other for some time:
Let’s start with the basics: some data is better than no data, as long as we understand the limitations of the data. Hand waving that the data is useless means ignoring what we can reasonably glean from the data. Firing people because you don’t like the data won’t make the data better, but worse.
The divergence is due to a number of factors, among them that Household includes agriculture and informal jobs, while Institutional counts people who have multiple jobs. Household is better at showing demographics, while Institutional is generally more accurate due to greater sampling. Unaccounted immigration is also a factor. Divergence is greater during periods of economic dislocation or restructuring.
Hmm. The chart you provided, “Household vs. Establishment Employment Trends”, doesn’t seem to match the data. The Household data is 163.1 million, but the Establishment data is 159.5 million, a difference of less than 4 million. Not sure why.
Actually, what is really important in this incident is the increasing dementia and belligerence of President Trump. His ever greater belligerence makes a real disaster increasingly probable.
Yeah, I know. I voted for him. But considering the Dims nominated both the worst presidential and vice presidential candidates in US history, I don’t regret the vote.
Yes. Let’s.
The Bureau of Labor Statistics collects data. They then adjust the data in three ways: seasonal, birth-death, and population. Once they do that it’s not data anymore. It’s adjusted data. They publish the adjusted data.
The problem is the magnitude of the adjustments.
It would be irresponsible and misleading to publish the data. So the BLS doesn’t do that.
Also, guess what? Every government everywhere in the world is political. Every bureaucracy is political. That is because they are composed of people and people are political animals as Aristotle pointed out two millennia ago. If you don’t want something to become political, don’t involve government.
This vision you have of government bureaucrats as unimpassioned Platonic philosopher-kings is a bizarre fantasy. It isn’t true in the United States and never has been. It isn’t true in the European Union. It isn’t true in China. It isn’t true anywhere.
No, Trump is not correct to be suspicious of the data. While you seem obsessed over stuff like birth-death adjustments the BLS has been transparent on their methods and AFAICT they have been consistent. You miss the forest for a few, tiny trees. There is a need for data on employment and other numbers as close to real time as possible. There are probably some pretty draconian and very expensive methods to achieve better numbers closer to real time than what we do. Instead we have chosen on surveys of smaller numbers to figure out what is going on among 100s of millions of people. That means, a priori, the numbers will never actually be correct. It also means that when tracked over time you can find and adjust for natural variations, which they attempt to do.
Having read both BLS material and people who used to work for them I am pretty confident that they are using up to date statistical methods that are widely accepted, by economists from both tribes and even from libertarians and other heterodox practitioners. Note that no one, again including you, ever offers numbers or any real evidence of bias or bad practices. Maybe they need larger samples now that the population is larger, but that would need to be paid for. Maybe they need faster computers and if so I am sure DOGE will OK that.
However, the only time there are ever accusations of the numbers being faked or politically biased is when someone doesnt like the numbers and even then note that the same numbers that supposedly helped your opponent are then claimed to be harmful. Really? How can you even begin to take those seriously?
Anyway, if you want to claim they are biased, provide evidence. You never do and neither do the critics. As a separate issue, could they do better. Undoubtedly, but if that is what is wanted then just ask them what they need to improve. I am sure they (BLS) would be very happy to respond.
Steve
Dave Schuler: Once they do that it’s not data anymore. It’s adjusted data. They publish the adjusted data.
Virtually all science is adjusted data. You don’t think Neptune is actually deep blue, do you? See Smith et al., Voyager 2 at Neptune: Imaging Science Results, Science 1989.
Dave Schuler: It would be irresponsible and misleading to publish the data. So the BLS doesn’t do that.
BLS provides access to raw, unadjusted, and anonymized data. That way, other analysts can study the data.
Dave Schuler: This vision you have of government bureaucrats as unimpassioned Platonic philosopher-kings is a bizarre fantasy.
“The fact that we live at the bottom of a deep gravity well, on the surface of a gas covered planet going around a nuclear fireball 90 million miles away and think this to be normal is obviously some indication of how skewed our perspective tends to be.” ? Douglas Adams
All scientific endeavors are made up of people. That’s not an argument that a particular finding is in error.
steve: Maybe they need faster computers and if so I am sure DOGE will OK that.
Trump even fired unpaid economic and data experts who were advising on statistical methodology, including problems inherent in the changing economy. So Dave Schuler is right—it is political. Just not in the way he suggested.