How does it compare to the government's official salary regression model based on company-internal payroll data? https://service.destatis.de/DE/gehaltsvergleich/index.html

I suspect a lot of that data is from job advertisements which go unfilled so they keep getting reposted.

Really interesting insight - something I’ve always suspected and my only hypothesis is that those likely to share will also be those on ordinary salaries. For example, your high achievers on unique salaries probably won’t post as they will be identifiable, and/or are unlikely to post because they are happy at the company and there is no need to (most reviewers do it on switching co or if they are unhappy).

this is US-specific information, but even at FAANG and adjacent, my experience with websites like levels.fyi is that the listed compensation, even contemporaneously, skewed very low compared to offers I knew of and was actually seeing.

generically, people are more likely to post on these compensation-sharing websites when they have gripes about compensation.

there's also the fact that many people take FAANG offers without negotiation (equity was often heavily negotiable) and you're looking at the average of a leftward-biased sample of a wide distribution.

I have a theory that Glassdoor, Indeed, etc., skew low on purpose. They want to retain business partners and businesses, more often than not, want to pay the lowest amount possible for labor.

There's also the growth factor consideration that probably isn't implemented by those websites. As in, salary data from several years ago might be not be adjusted for inflation and have the same weighting as recent salary data.

Half the time, the websites can't even gather salary info from job posts and rely on estimates despite having hard data.

What this can do is that less savvy users might not notice the estimated data tag and use the lower range in negotiations, which could ultimately lower starting salaries for workers moving forward.

I've written job boards before about this, but of course they aren't going to change anything as it isn't in the their best interests.

Yeah, my company previously used numbers from Mercer to argue we were above average, when numbers from the union placed us below. Both of course might have a bias, the union in our favor, and Mercer getting paid to keep the numbers low.

Absolutely this. I'm always scared by seeing these low numbers all these sites. They have no relation to reality