Post

Conversation

Betsey Stevenson
@BetseyStevenson
The number of people who think that (a) it's easy to get accurate data in a timely fashion and (b) if there are revisions it must be "mistakes" or "faking data" is really worrisome. Let me explain. 🧵
Quote
Betsey Stevenson
@BetseyStevenson
My fear is that ordinary Americans are not experiencing the firing of an experienced, non-partisan, leader of a statistical agency as the major earthquake for the United States that it is.
406.1K
Views
David Watson 🥑
Post your reply

Betsey Stevenson
@BetseyStevenson
The jobs numbers for July reflect all the people on a companies payroll in the week that includes the 12th of the month. Companies need to send that information to the BLS. Many are delayed in sending it in, either because they are busy or because their pay period that includes
13K
Betsey Stevenson
@BetseyStevenson
The BLS also needs a representative survey of companies and every day new companies are formed and existing companies shut down. When a company doens't respond on time is it becuase they are busy or because they shut down?
10K
Betsey Stevenson
@BetseyStevenson
Should the data be held until all companies in the survey respond? That would mean releasing May data now, in August. It would be more accurate (as the latest revision is more accurate), but it would mean waiting months for the data.
9.8K
Betsey Stevenson
@BetseyStevenson
Another issue is new companies forming. It's hard to find companies that just opened, but assuming none opened is wrong. Should the data be held until an accurate measure of company births can be gathered? To be completely accurate we might want to wait until the UI tax data is
12K
Betsey Stevenson
@BetseyStevenson
Finally, we want seasonally adjusted data so we aren't just seeing slumps and surges due to normal seasonal patterns. The problem is that "normal seasonal" changes and got particularly shook up with Covid. That might mean only releasing annual data to avoid seasonal patterns or
12K
RealEconomist
@JeffreyVButler
It's a problem when revisions tend to go in one direction. It suggests systematic errors rather than "mistakes," which should be random and hence equally likely to be up or down.
3.2K
0xJMG
@0xJMG
Why get the data from laggy analog era manual surveys when it can be gotten digitally in realtime from the major payroll companies?
3.5K
BabbleBee
@ThankElon
Come on, man. There are thousands of us out here, doing statistics, managing statisticians, for companies and organizations. If your early estimates have far more variance than the time series you’re estimating, you’d better run a project to fix your estimates.
1.2K
Mike Howton
@MikeHowton
People who think it is okay to publish statistics the effect trillions of $’s, without a confidence interval, should be fired.
1K
AJR
@Rethans
Nice Gaslighting Betsey. Let's make this really really clear - the issue is not that there are 'revisions'. The issue if the incredible degree of the revisions. As I've said, if a private sector employee delivered 'revisions' such as these at work, they'd be fired. A 'long
3.5K
Brett Yokom
@BrettYokom
No one thinks it’s easy and no one thinks there won’t be adjustments. In 2025 though, with the technology available to her and her team, one would expect accuracy to be dramatically IMPROVING, not worsening. What was she doing all day? Did she care?
1.4K
InstructorJohn (O-Ga)
@jej117
To assume that labor statistics are immune to manipulation is misguided. It’s clear that truly "non-partisan" political nominees are rare to the point of non-existent. Dismissing this reality is not only shortsighted but disconnected from evidence.
918
baystreetwiseguy
@Bay_St_Wiseguy
Quote
James E. Thorne
@DrJStrategy
Shelter component of CPI as measured by the BLS is the largest contributor to inflation, 3.8% YoY. The latency in the shelter variable is well known yet the BLS is unwilling to fix, and the Fed unwilling to disregard. According to the latest available data, the Cleveland
Show more
Image
Image
1.6K
CaptFligh
@CaptFligh
We don’t believe the ‘non-partisan’ part. Focus on that argument. Just using the label doesn’t make it so.
1.4K
eMon
@0xemon
Citizens have lost trust in the institutions Citizens know trump will deliver for them
Image
1.5K
Rebel A. Cole
@RebelACole
You utterly FAILED to explain why the revisions have been so large and consistently in the same direction. Downward. One can, and should, model these regularities. BLS failed. Just like you.
207
Boom 💥
@LoveCodeTrade
The number of people who think governments don't ALWAYS cook the books is worrisome. Let me explain... If you believe anything the government reports is accurate, and not heavily influenced by politics, you're f'ing retarded! And it worries me that this many retards have a vote
821
Lightning_Shade
@LightningShade0
As a non-expert who doesn't understand how they get the job numbers at all -- is there a good explanation for why revisions always have to go down in one direction? Intuitively, it seems there's at least some level of systematic overoptimism in the initial estimates.
975
z
@cptzc
it makes more sense when you realize there are a lot of jobs in the US, and you have to count stocks to measure flows. They were only like .1% off which is pretty impressive
61
The Hube 🇺🇸
@TheHube
The reason the change looks high is presentation. Comparing the revision to the XXX,000 initial change instead of the XXX,000,000 estimated jobs shows it is essentially a rounding error.
53
GlibAtreides
@GlibAtreides
Nobody thinks “it's easy to get accurate data in a timely fashion.” When you start with a straw man like that we know we can ignore the rest.
34
aimeeL
@love2read78
Anyone who has been in investment arena understands the data gets revised. It’s only since these Johnny come lately maga so called “experts” now use X as their platform it’s an issue. Too many people with not enough knowledge.
259
Christian Smithson
@cds1989oct
Not enough people get it. What makes matter worse is response rates and timeliness has gone down a TON since COVID. There was a volunteer committee comprised of very smart people that was supposed to meet 2x/yr to figure out how to fix the response problem. Nutlick ended it.
521
Taxpayer at large
@MikeBoatCPA
It's not mistakes with the data. It is bias in seasonal adjustments and the birth/death model. These items contribute to revisions, why not take a harder look at their use?
112
Make Music Not War
@musicman4203
Non compliance is the problem. Many companies are not responding to the survey in a timely manner. Either force them too or don’t release the data until you have 90%
212
TraderSven
@sj_bennett
I am curious why the BLS hasn't changed the methods you use to calculate the numbers. Payroll tax should be fairly easy to track. It seems that the adjustments to correct things actually mess them up.
92
Biard MacGuineas
@BiardMacGuineas
@grok
"The number of people who think that (a) it's easy to get accurate data in a timely fashion and (b) if there are revisions it must be 'mistakes' or 'faking data' is really worrisome."
35
Lenny Eusebi
@lennyeusebi
Or, if there’s significant uncertainty, they could report a range rather than single number and narrow the range as their certainty increases.
49
Curtis Hayes
@hayes_curtis
If it’s wrong every month then it’s no value. Only and economist can’t see this issue. Economists and weathermen keep their jobs when they are wrong 50% of the time or more.
69
Patriotic
@Patriotic126274
Remember those 3 months last year with identical 222,000 numbers. Then massive revisions after. Yeah those. Statistically impossible
12
David
@David_V_eng
The scale of revisions during this woman’s tenure were several times greater than during any of her predecessor’s. I don’t know how to interpret that other than incompetence
54
NthInfusion🔻
@nth_infusion
I understand revisions are normal and necessary, but from 230,000 to 14,000? 99% of the original count was off. They don't do any kind of estimation given what they know about the reliability of the data?
120
John Espey
@espey12_john
You miss the point (common with government employees). Why are we spending $700,000,000 per year gathering inaccurate data?
31
Nikki DeFilippo
@DefilippoNikki
Her numbers were GROSSLY off. For at least the whole last year. Which means there are indicators in the markets she’s unable to pinpoint. Which makes her incompetent at this point. She may have been great at one time but she is no longer competent
27
JackJones
@77AugustWest72
Anybody who thinks that government run data collection, processing and reporting is accurate, unbiased and well managed is from another planet. These are government hacks collecting data with obsolete methods and using software written in COBOL. Give me a break.
144
Mario Furious
@InfinitumSolus
it may not be easy to get 100% accurate data but it sure was easy for her to overshoot every single time for Biden and then try to quietly revise it downward when no one was paying attention the Biden economy lie was in part perpetuated by her consistently poor data analysis
62
chris repke
@Chris_Repke
Nonsense. You miss by 80% you’re either lying or retarded. Either way, you should have been fired a LOOOOOOONG time ago. To keep them on the job implies there’s no one in the world who could’ve gotten the numbers closer to reality sooner. Dumb govt people need to be fired.
101
JoshJelsing
@josh_jelsing
Why is it that ADP, a private company, gets it so much better than government? Why, if so many markets around the world depend on information from the BLS reports, has the government not taken the time to update their systems to get away from waiting on companies to complete
51
Be the Light
@Gr8ter_is_He
Your argument is that it impossible to get accurate data fast, but we must have data fast even if it is inaccurate. That's just stupid. If it can't be fast AND accurate, it needs to be accurate whenever it can be gotten.
45
Marcus of Ockham
@HouTwit
Natural mistakes would be made in each direction though. If you’re always making the same mistakes in the same direction, then you’d need to factor in a contingency to correct for bias in one direction.
62
MostlyHarmless
@harmless_human
She had to revise her own numbers by 90% multiple times in two administrations. She was terrible at her job. Long established procedures that are off by hundreds of thousands of jobs should be in the trash. Why do you like trash data?
26
Madisonian
@amorales866
Blah blah blah. No one believes anyone who was in the Obama Administration. All a bunch of lying cheating people. That’s why. All your credibility was flushed down the toilet, by your ex boss. Will be remembered as one of the worst presidents in American history.
29
on5.info
@On5info
You describe just a few examples where these statistics can, and most likely are manipulated. This is an inherent problem with large numbers and statistical manipulation.
19
Jean 🏖
@jks417
No. Let me explain. Antiquated systems collecting unreliable data manipulated by a political appointee = GIGO - garbage in, garbage out.
163
Richard Talaber
@yankeerick1
It’s the job. The severe/historic revisions have been happening for two years. Do a better job, get better tech, whatever. Stop with the excuses.
86
Rufus
@Moldymaltquaff
Your points are valid. The true definition state of the economy can never be known until significantly afterwards. But you’re missing the forest for the trees. The errors consistently skew one direction when a Republican is in office, but the other when a Democrat is in office.
21