Our financial system is crumbling this week.

[heyzeus]

No. Absolutely not.

[cardinalkarp]

The data can’t be trusted for ANY administration.

[Joe Shlabotnik]

The data can’t be trusted for ANY administration.

You can think that and, to some extent, its probably true. But the USA has rung up huge debts not allowed to any other country and, of course, the dollar has been the world's reserve currency since WWII because of its perceived stability. A lot of that perception was based on that fact that, warranted or not, the data about the dollar and the US economy has been trusted. Trump is taking an axe to that trust through shenanigans like firing the labor statistics director for one report he didn't like. It can't end well.

[heyzeus]

The data can’t be trusted for ANY administration.

Only one administration has fired the commissioner of the Bureau of Labor Statistics for daring to publish employment figures that didn't flatter the president. We are a banana republic now. We are North Korea.

[cardinalkarp]

The data can’t be trusted for ANY administration.

Only one administration has fired the commissioner of the Bureau of Labor Statistics for daring to publish employment figures that didn't flatter the president. We are a banana republic now. We are North Korea.

Well, that really wasn’t the question at hand. Yes, Trumps firing of the Bureau of Labor Statistics head was ridiculous. The misleading statistics for job #’s have gone back as long as I’ve been paying attention (and granted that’s only been a handful of years). The job reports come out and they regularly act as though jobs are great, and then quietly adjust the numbers down after a few months when no one is paying a lick of attention and show that the job reports which came out was grossly exaggerated.

So I stand by my statement, you can’t trust the data from ANY administration.

[ghostrunner]

Well, that really wasn’t the question at hand. Yes, Trumps firing of the Bureau of Labor Statistics head was ridiculous. The misleading statistics for job #’s have gone back as long as I’ve been paying attention (and granted that’s only been a handful of years). The job reports come out and they regularly act as though jobs are great, and then quietly adjust the numbers down after a few months when no one is paying a lick of attention and show that the job reports which came out was grossly exaggerated.

This is not true.

The jobs numbers are understood by everyone to be incorrect when they first come out and they've all been revised every time, in both directions, by varying amounts and in every administration. You can't trust the initial estimate because they're inherently premature and operating off of incomplete information. I'm sure there are people qualified to judge whether or not this latest person in charge was or wasn't competent, but I bet they aren't the ones getting attention over those claims.

Up until now, I think it's pretty well understood that these numbers aren't intentionally cooked, precisely because they've been positive and negative for admins of both parties. AND you do not want to lie about those numbers because if you do, now you're asking your whole economic system (govt and private sector) to work off of numbers that aren't real. The same reason it's bad to only pay attention to polls that look good for you and ignore polls that are bad. You want something close to the truth, so you can try to correctly adjust to it.

You can go look at every monthly estimate for every year since 1979, you can also see the 1st, 2nd and 3rd revisions for each one, read the monthly reports - all of that is here.

https://www.bls.gov/web/empsit/cesnaicsrev.htm#Summary

If you look just at the 1st to 3rd revision columns - which is the initial estimate and final revision - they go up and down all over the place. No correlation to elections either, which is what you'd expect if numbers were cooked.

Looking at it myself it certainly doesn't seem as if the person who was fired was doing significantly worse at estimating these numbers than the previous 45 years. So this looks pretty bad, and hopefully the replacement understands what their job is and doesn't just respond to whatever Trump wants. What would be most concerning is if we stop getting these reports altogether, because it offers everyone some way to check their work.

[GeddyWrox]

Excellent explanation, ghost. Thank you.

[cardinalkarp]

That’s definitely interesting and I appreciate the explanation.

I guess my only question would be what’s the point of releasing job #’s if they are ALWAYS initially wrong? Call me dumb, but what good is a job report if they have to go back in and correct numbers that are sometimes off by 20, 30, 40%.

Really it just seems like some sort of market manipulation.

[ghostrunner]

Seems fair to ask why the numbers are released so early, but I can’t see how you look at all that and see manipulation.

We base a lot of our economic behavior and our retirements on a massive betting market.

[Arthur Dent]

There is an accuracy/promptness tradeoff. You want to have some sense as early as possible what's happening in the economy, so you have a chance to react to developments. Super accurate data about what happened a year ago would be of little use. Because of the trade-off, the policy of releasing an initial number and then scheduled subsequent refinements as more data comes in has been the approach. It's expected that people can behave as adults and understand uncertainty.

Also, the errors are not actually big at all if you understand what they are actually measuring, which is the total number of jobs. That number, around 160 million, is estimated each month but then the published headline is the difference in the estimates between subsequent months. A big 100k+ revision of the monthly change represents a like 0.1% estimation error.

They're only able to get this degree of precision by doing an absolutely massive survey of businesses. It's really pretty incredible. The weakness of the approach is that it is hard to survey just formed businesses, so those jobs have to be estimated using modeling and historical data. This doesn't work as well when the situation is changing as it is now, so it's well known that you get larger revisions at economic inflection points. I certainly hope that this isn't happening now, but the signs are concerning.