This weekly roundup thread is intended for all culture war posts. 'Culture war' is vaguely defined, but it basically means controversial issues that fall along set tribal lines. Arguments over culture war issues generate a lot of heat and little light, and few deeply entrenched people ever change their minds. This thread is for voicing opinions and analyzing the state of the discussion while trying to optimize for light over heat.
Optimistically, we think that engaging with people you disagree with is worth your time, and so is being nice! Pessimistically, there are many dynamics that can lead discussions on Culture War topics to become unproductive. There's a human tendency to divide along tribal lines, praising your ingroup and vilifying your outgroup - and if you think you find it easy to criticize your ingroup, then it may be that your outgroup is not who you think it is. Extremists with opposing positions can feed off each other, highlighting each other's worst points to justify their own angry rhetoric, which becomes in turn a new example of bad behavior for the other side to highlight.
We would like to avoid these negative dynamics. Accordingly, we ask that you do not use this thread for waging the Culture War. Examples of waging the Culture War:
-
Shaming.
-
Attempting to 'build consensus' or enforce ideological conformity.
-
Making sweeping generalizations to vilify a group you dislike.
-
Recruiting for a cause.
-
Posting links that could be summarized as 'Boo outgroup!' Basically, if your content is 'Can you believe what Those People did this week?' then you should either refrain from posting, or do some very patient work to contextualize and/or steel-man the relevant viewpoint.
In general, you should argue to understand, not to win. This thread is not territory to be claimed by one group or another; indeed, the aim is to have many different viewpoints represented here. Thus, we also ask that you follow some guidelines:
-
Speak plainly. Avoid sarcasm and mockery. When disagreeing with someone, state your objections explicitly.
-
Be as precise and charitable as you can. Don't paraphrase unflatteringly.
-
Don't imply that someone said something they did not say, even if you think it follows from what they said.
-
Write like everyone is reading and you want them to be included in the discussion.
On an ad hoc basis, the mods will try to compile a list of the best posts/comments from the previous week, posted in Quality Contribution threads and archived at /r/TheThread. You may nominate a comment for this list by clicking on 'report' at the bottom of the post and typing 'Actually a quality contribution' as the report reason.
Jump in the discussion.
No email address required.
Notes -
"Trump fires Bureau of Labor Statistics chief without evidence for political reasons" says the news radio I wake up to, then continues to say he removed the Democrat appointee "without concrete evidence." Since COVID-19 caused lockdowns, the BLS numbers have been revised downward from initial reports regularly, sometimes ridiculously so, which Axios says has justifiable reasons.
So why are the initial numbers even reported if we know the algorithm they use will be wildly inaccurate?
Getting the exact numbers 3 months later is just not as useful as getting directionally-correct ones fast.
That's good when the numbers are directionally correct, not when they're completely wrong
Right, well, the FBI stats are not the BLS stats.
The BLS stats have been generally correct (and getting better) and, more importantly, have erred both upwards and downwards approximately equally.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
Reuters has a nice graph showing that large revisions have been made quite often, both downward and upward, for decades.
I feel like you're burying some of Reuters own commentary there.
Yes, revisions have been made quite often, but this one is noteworthy and unusual. I'm not yet willing to ascribe it to malice, but we should acknowledge that it's peculiar.
More options
Context Copy link
More options
Context Copy link
This seems like focusing on the wrong part of the story.
"Moreover, the BLS and other federal agencies are essentially trying to serve two purposes at once. On the one hand, they’re hoping to provide actionable information as fast as they reasonably can to employers, investors, job-seekers, policymakers and the Federal Reserve. On the other hand, they’re trying to formulate a “permanent record”, data that is treated as ground truth in future economic analysis. Those imply taking different positions in the unavoidable trade-off between speed and precision."
Biased estimators can still be useful. If you know an estimator is consistently high, you can account for that in your planning. On the other hand, if political leadership is putting their thumb on the scale to make themselves look good (or salve dear leader's ego), trustworthiness goes out the window. It's one thing to be wrong occasionally, it's another to be bullshit.
I don't think that's really the danger here. If the BLS statistics aren't trusted, some actors are going to do their best to fill in a trustworthy answer. The problem there won't be their honesty, but that the data are not going to be evenly available. We'll go back to information asymmetries rather than public knowledge.
More options
Context Copy link
This estimator isn't biased though.
More options
Context Copy link
If the estimator knows that they're consistently high, why aren't they adjusting the model they're using to produce estimates with to account for that?
If the estimator is wrong consistently but in a predictable way... they should be able to be wrong less often?
When I say "account for that in planning", I don't mean you adjust your forecasts downward X% from the report because they always overestimate by the same margin. Consistently high is not the same thing as 'always high' or 'consistently high by the same amount'. It just means that on average the estimator is greater than the true value (or, really, the quick estimate tends to be higher than the slow estimate).
Not necessarily. Estimation is always dealing with real world constraints liked limited resources and time frame for gathering and analyzing data, sampling bias, unknown unknowns, etc...
I encourage you to read the Nate Silver article I linked. He talks about this significantly more articulately than I can.
I am somewhat familiar with Nate Silver's approach to modelling and prediction.
And I'll reiterate the general critique.
If you damn well know your model is going to be inaccurate, include error bars, express how much irreducible uncertainty there is. At least acknowledge that the number is most likely incorrect and is subject to large revisions, downplay confidence.
Actually, it looks like they DO have that option on display and HOLY CRAP the bars are really large on some of these.
Maybe its not a particularly useful estimate if businesses are looking for something something reliable to act upon.
More options
Context Copy link
More options
Context Copy link
Do you want to be the guy in the office who adds in the unprincipled empirical fudge factor to a statutorily mandated report? The verbs in 29 USC 2 are “collect, collate, report, and publish”, nowhere does it say “estimate”, “calculate”, or “determine”.
I'd be the guy in the office suggesting "hey we can publish the report, and show both the standard estimate and the estimate with the empirical fudge factor side-by-side so its clear we're not hiding the ball."
More options
Context Copy link
More options
Context Copy link
At least in this case I think there is an added political dimension of reluctance to update the model: "You were happy enough to overestimate [measurement] for my opponent last term, and now you want to publish lower estimates, maybe even underestimates on my watch. Are you trying to display partisan bias?"
In addition to the value of "we've at least measured it consistently for the last century, even if there are known issues with it it's easier to fix those in post", which also has some value.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
The whole "without evidence" tic is pretty played out at this point. Of course, Trump does have evidence -- the revisions are higher than usual. It's pretty bad evidence (so "without concrete evidence" is true), but it's enough to make "without evidence" naked editorializing.
BLS has been putting out these numbers monthly for many years; I am sure if they proposed delaying the releases two months there would be all sorts of complaints about that too.
Trumps own former BLS chief himself doesn’t like it. And it includes this very damning quote:
Other articles note that usually, initial estimates are based on larger employers, and smaller ones take longer to report. Savvy consumers of the stats know this. Also, what size company has been hit hardest by recent market uncertainties including tariffs? Small employers. The variance is higher.
If a number feels off is your evidence, and it’s plausible or even likely that the explanation could be explained by either malice OR the underlying stats actually being off, it’s still “no evidence” in a statistical sense. We need DETAILS to be able to assess the claim, and Trump provided none, and furthermore if his own former guy says that the chief doesn’t even see the numbers until they are nearly fully assembled, we have strong reason to be skeptical and zero actual reason to trust him (beyond a baseline level of trust in Trump himself).
What does the commissioner... do, then? This feels like the scene from Office Space with the Bobs.
The commissioner is a people person. He takes the figures from the statisticians to the politicians.
Does he physically hand them over?
No, he faxes it. Sometimes the secretary sends the fax.
Staff organization stuff, I presume. Charitably, she (it was a woman) was in charge of approving methodological changes but presuming no such change happened, there's no reason to cast blame. That's what's so frustrating, there wasn't any specific allegation like, at all, that they released.
Well, no, that's not quite right: we can look at Trump's statement
That's the entire allegation. Note that the claim is much stronger than a mere claim that methodologies were changed or that the standards were relaxed or whatever. Nope, "concocted" and "rigged" mean something pretty clear. M-W definition for "concoct" is "devise, fabricate"; Cambridge has "to invent an excuse, explanation, or story in order to deceive someone". In other words, intentional manipulation. This would mean something along the lines of entirely inventing a number, or deliberately skewing your sample, or spontaneously cherry-picking a methodology, or something like that.
If this were true (obviously is not) than you wouldn't be firing the commissioner, you'd be firing normal-level staff too, or doing an actual investigation, right? You might fire the commissioner only if their people-leading skills were poor or their methodological direction was faulty, but that's not the case and not what's alleged.
It's nakedly anti-truth, and that's not a TDS thing to say. No need to defend Trump in every instance, this is just straight up wrong per the info we have access to, as it seems to be a top-down doubt on the numbers rather than a bottom-up, facts based one.
More options
Context Copy link
The commissioner is the boss, in charge of all the people who do things. Or more likely in charge of several layers of sub-bosses before you get to the people who do things.
Which is why the quote isn't damning; with that authority comes the responsibility as well.
More options
Context Copy link
More options
Context Copy link
In a statistical sense, saying "datum D is no evidence for theory T" is "P(T|D) = P(D)". Here we have "P(T|D) > P(D)", which is "D is evidence".
It's not much evidence. It's not nearly enough evidence. It's outweighed by other evidence to the contrary. It's grossly outweighed by reasonable priors. But it's still evidence.
I hate to pick on anti-Trump folks about this, when Trump's own relationship with truth seems to intersect propositional logic only by random luck; forget about Bayesian statistics. But it's still a red flag to me.
Decades ago I waded into investigation of a controversial belief system, a "religion" or a "cult" depending on who you asked. I debated with folks about evidence for and evidence against many of the beliefs, and my eventual conclusion was basically "false religion" ... but the most memorable part of those discussions was, when one guy I'd been debating with was asked by another interlocutor whether there was any evidence against his religion, his answer was a flat "no". Not "yes, but there's more evidence for it", not "yes, but only if you consider evidence out of context", just "no, there's no evidence against it".
I still had lingering questions (of what I'd later start thinking of as "epistemic rationality") to resolve, but now more pragmatic ("instrumentally rational") concerns were screaming at me to be wary in a way that continuing abstract discussion of science or history couldn't have done. It might not have been his religious leaders' fault, but that guy was in a cult.
Such self-inflicted damage isn't worth it for any ideology. You might still end up at a correct belief, or a dozen, but only by random luck.
Sorry and thank you, you're very right and right to call me out, I in fact do know better. I should have written "'no evidence' in the traditional sense* and expressed myself poorly.
Two things are true: in a Bayesian statistics sense most things count as evidence, and in an everyday sense people want to see some kind of fact to support an allegation. Zero were provided, as far as I see. Not even a cogent rationale was gestured at. I do try and consistently be charitable in my comments, I tried a bit in a follow-up comment above, but Trump's method of handling this gives virtually nothing to work with. (And as I stated, the former Trump-appointed commissioner sticking up for her is pretty large evidence against Trump's claim, even if you weight Trump's claims highly on a personal prior level)
Yeah, I can't disagree with any of that. In the colloquial sense "no evidence" fits.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
So? That doesn't mean it's "without evidence".
And that quote isn't damning at all. The fact that the commissioner doesn't collect the numbers herself does not mean she is not responsible for doing so.
I'm 99% sure Trump's wrong and she wasn't cooking the numbers, and it's likely she wasn't doing a bad job.
BLS commissioner 2013-2017, an Obama-era one but still obviously a person in the know:
I think that elucidates the point a little bit more, especially the bit about how methodology changes are obvious and up-front. The operation in professional statistics orgs like this is pretty plug-and-play on the collection side and there's a lot of cross-checking that happens. Plus, anecdotally, the BLS has one of the better reputations in the stats community and worldwide.
What I mean by evidence is like, if not actual whisteblowers or a smoking gun email or edited Excel file, at least some kind of specific alleged mechanism: did she pressure data collectors to poll only certain forms? Was the sample size abnormally low? Did they go on some kind of fishing expedition? Were internal policies not followed? Something like that.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link