Hacker News Reader

Comment by mrleiter | Hacker News Reader

Just a quick reminder what median means in this case:

You have a finite set of wages/pay - and $197,000 should be in the middle of this set. Meaning that half the wages at Alphabet are below, the other half above this wage.

It is a rather robust measure, because it is insensitive to extreme values (like the arithmetic average would be), while also giving you a good overview, e.g. that 50% of all the wages are above $197,000.

Vice versa, it tells you nothing about how low or how high the wages are aside the median (e.g. the lower 45% of the wages could be <$45,000). So take it with a grain of salt.

View on HN

Also, as with the Facebook news that their median is $240k (https://news.ycombinator.com/item?id=16955393), it could well be the case that Alphabet's lower paid staff are mostly contractors who wouldn't be counted in this metric.

Everyone seems to report either median or average, both of which leave out a lot of interesting information. If people simply reported both values, it would go a long way to painting a more accurate picture.

I was just thinking about this. Ideally, the median and mean salaries should account for added value but not vary so much as to make inequality unbearable. Having CEOs paid 100 or 200 or even 1000 times more than other employees is bad for everyone in the long run. It leads to so much inequality that people end up tearing each other appart, looking for culprits (immigrants, strangers, neighbouring countries, etc). Having a CEO paid more than others is fine as long as everyone can benefit from the added value, not just a few.

For an accurate picture, you can't beat a picture. A graph of the distribution of employees over salaries is easy to understand and information rich.

Median is the only one you need a disclosure for. Mean is simply the compensation expense divided by the number of employees. But the mean salary doesn't really say anything, it'll always be higher than median and at a company the size of Alphabet you don't really get additional insights from that.

Reporting median is a good step, most companies don't disclose that.

The difference between mean and median provides information. It's always safe to assume the distribution is right-skewed, but having both median and mean gives you a sense for how much.

Just the difference doesn't tell you much. It's always right-skewed. But that could be because it's bimodal with developers earning a lot and support staff little or it's because everyone makes the same except for some really well paid executives.

What would be really interesting would be to see the actual distribution. But I don't think we'll ever see that.

I think at this point we're quibbling over the definition of "how much" it can tell you. I think "median of $70k, mean of $150k" vs. "median of $70k, mean of $1.5M" carries a lot more information than just mean or just median, even if it doesn't approach a full summary of the distribution's shape.

Amazon's medium pay is $28,446. But they also employ nearly 500,000 people. The linked article also has salaries for other tech companies.

https://www.wsj.com/articles/amazons-typical-worker-is-in-a-...

Reporting and pushing median pay now is smart for a company like Amazon.

When you have a Democratic Congress, $15/hr minimum wage is coming. Many of the grunts will be getting a big raise, which will increase that median significantly in a few years.

These companies are reporting it now because they don't have a choice, it's a federal law that most large public companies have to report median pay and CEO pay.

Most of them work for Integrity Staffing not Amazon.

Nope. https://www.theverge.com/2017/2/3/14504526/amazon-job-growth...

That story from a year calls out that it excludes contractors. It does include full-time and part-time employees.

Totally possible. I am not acquainted with the regulatory statutes regarding who you may or can count as FTE in such a statistic.

But sure sounds like a feasible strategy to prop up median pay.

I'm pretty sure most companies would want to make median pay appear as low as possible when making SEC filings. Investors hate high labor costs.

I would agree in regards to SEC filings, but I'm not sure this comes from an SEC filing?

Also, do you have to state median pay in such a filing?

You're right, I suspect this article is based off of new federal requirements that companies report employee vs. CEO pay disparities [0].

0: https://www.sfgate.com/business/networth/article/Yes-median-...

I don't think companies are required to and this statement from the WSj article:

>the fourth-highest pay among the hundreds of companies in the S&P 500 index that have disclosed those figures.

seems to imply that it is not.

This article seems to suggest it's mandatory for public companies:

>As Congress chips away at bank regulations established by the 2010 Dodd-Frank law, another part of the measure is exposing the extreme income inequality between bosses and their workers.

>The law requires publicly traded companies to calculate the ratio of their chief executive officer’s compensation to the median pay of the companies’ employees. After a series of delays, firms are finally disclosing their pay ratios in filings with the Securities and Exchange Commission. [0]

[0] https://www.huffingtonpost.com/entry/ceo-pay-disclosure_us_5...

It is the case. Almost all of their scaled support is from overseas most of which are not Google employees. They pay other companies to manage these folks. That 85K employee headcount is only the FTE count. The number of non-FTEs is crazy.

If the underlying data is normally distributed, then there's no difference between median and mean for a sufficiently large sample size. Which is why mu can be median or mean when calculating the probability density function.

It's safe to say salaries at FB, Apple, and Alphabet are normally distributed. Thus, reporting average or median doesn't matter.

Uh, no, it is not safe to say that salaries are normally distributed, and they certainly aren't.

The salary distribution graph for a large corporation is likely to look something like this:

http://www.visier.com/wp-content/uploads/2016/05/Graph-showi...

with a very, very, very long tail to the right representing the salaries of the CEO and senior executives. Depending on the corporation, there may be multiple humps ("salaries for Amazon warehouse workers" representing one hump, and "salaries for Amazon programmers" representing another, for example).

It's nothing like a normal curve.

> It's nothing like a normal curve.

If the data comes from a sample, then it's going to normal, regardless of the underlying distribution. So the mean and median will be approximately the same.

You may be confusing the distribution of the means of the samples (which will be normally distributed if the Central Limit Theorem holds) with the sample itself (which will not generally be normally distributed).

Reply on news.ycombinator.com