Econtwitter vs BLS

· 830 words · 4 minute read

I have stopped engaging with twitter, but I do still read it. ML-twitter, econtwitter and some of the most popular data-personalities still rummage around there.

Especially econtwitter is incredibly informative, and at times funny as hell. Like this weekend, when the VC-bros from the “all in” podcast and friends decided they were way smarter than the bureau of labor statistics. I’m not interested in dunking even more on the stupid VC-bro takes, but in a way this was actually reasonable questions wrapped in a reprehensive know-it-all attitude with a complete and utter lack of even basic introspection of the kind “if this doesn’t make sense to me, might it be that there is something I don’t understand?”.

The answer to most of their questions can be found on the BLS website.

I am once again asking VC-bros to read the whole BLS publication

The background 🔗

This is more or less what started it:

Payroll survey and unemployment seems to not add up

It is important to know that even though something doesn’t make sense at first glance, it can indeed be true. And also that all statistics are wrong. They are estimates, not ground truth. In fact, response rates are sadly quite low nowadays.

The numbers mentioned come from two different surveys, so they wouldn’t necessarily add up perfectly even if they should, in principle, add up. But also, they were never meant to. First off, there is the question of what unemployment even is. And unemployment is not the same as “not working”. Students don’t work, and they aren’t unemployed. Neither are retirees. So being unemployed is actually predicated on wanting a job. As a result, it is possible for the unemployment rate to rise even though nothing at all changes in the economy. Imagine a month when nobody graduates, immigrates, no companies downsize, nothing at all happens, but unemployment can still increase because some people who were happily living their life decide they would like a job for a change.

Secondly, the payroll survey covers employees. But as VCs should know, it is possible to work without being employed. I think it is called being self-employed. Or an early-stage startup. Or freelancer. Foodora worker. So there are many ways the payroll survey can show an increase in employment but for unemployment to increase. Maybe Foodora replaces their delivery people with ChatGPT. Maybe retirees discover their savings don’t last long in a high-inflation economy, and start looking for jobs.

But wait, there’s more…

CPI data collection hot-take

Once again, with feeling: If you think it is weird that the CPI is constructed by (among other things) physically surveying prices in stores, I don’t hold it against you. After all, we have computers and stuff nowadays. But I do hold it against you if you assume “other people are stupid” is the only possible explanation for this.

With surveys like this, getting a representative sample is key - and some things are not sold online. Or maybe online prices don’t match the store price. Also, I suspect the U.S. has an aversion to government surveys, so physically sampling prices might be the more palatable alternative. And going to stores to jot down prices is far from the most eye-catching example of data collection. GDP is supposed to include all parts of the economy. And freakonomics had a great story about how workers at the british BNS had to take to the streets to ask prostitutes what they charged for their services.

Official statistics is often in this kind of situation. A lot of prices are available online. Is there anything you can’t buy on Amazon? But that doesn’t mean Amazon is the only place you need to survey prices - for the simple reason that Amazon is not the only place people purchase things. The consumer price index must reflect what people actually purchase, in the correct quantities. If Amazon sells cheap toothbrushes but everybody buys their toothbrushes at Walmart because that’s where they do the rest of their grocery shopping, the price of toothbrushes on Amazon is of limited interest to the CPI. OK, this was a contrived example, but the point is important: If you want to construct a price index representative of what people actually buy, you need to know what people buy, where they buy it, and go check the price.

Similarly, Linkedin is a treasure trove of information. But few other than Linkedin is interested in the Linkedin-unemployment. The percentage of people with a “open to work” sticker on their profile photo is not the same as unemployment. And the number of people without that sticker is not the same as employment. It might come as a surprise to VC-bros, but some people don’t have Linked. And they are not missing at random.

Analysts at Indeed have great respect for BLS

But none of this makes us stop wondering if it could be possible to do away with manual surveys or create more consistent statistics. Especially when there is such a wealth of information online. And a lot of researchers do indeed explore the possibilities - such as using data from payroll processing companies to track employment: