Hacker News Reader

Hacker News

Comment by adventured | Hacker News Reader

Full thread

adventured

This is an exceptionally poor premise. It's hard to believe the writer of this article missed the obvious.

That CEO pay chart tracks the value of options granted.

What it almost perfectly aligns to is the S&P 500 chart.

Here, I did a crude overlay, it's blatantly obvious.

http://i.imgur.com/S2hN29F.png

It wasn't Monday Night Football or any other ridiculous notion. It was the greatest stock market boom in the US since the roaring 1920s, which sent the value of options soaring.

View on HN

"What it almost perfectly aligns to is the S&P 500 chart."

You've displayed correlation and insisted causation. Of course, Ezra seems to have done so as well, but your correlation isn't any more impressive than his.

The fact that it so closely aligns, in both spikes and drops, over a 50 year period of time, more than proves the point.

I'm insisting that such an extreme coincidence is impossible.

It seems to decouple quite noticeably around 2008.

The decoupling in 2008 is plausibly explained by increased pressures for pay restraint following the financial crisis.

I agree, but now we're back to confusing correlation with causation.

Note that the compensation is partially (largely?) due to "options exercised." Depending on how that data is tabulated, it's possible that many options have not yet been exercised, and thus aren't included in that metric. ie, CEOs might be sitting on a bunch of granted-but-not-yet-exercised options.

You could get a better idea by looking at the value of options granted and their vest dates, but I don't care enough to dig that deeply.

Very often in statistics the common sense answer is completely wrong. That's why we insist that correlation does not equal causation. Your picture proves STRONG CORRELATION but says nothing about causation. It can't say anything about causation, because that information is missing currently without an analysis.

Technically speaking (the best kind of speaking) you have demonstrated only that these two things are correlated, but you have not proved that one caused the other.

For example, the rise in CEO pay could have caused the stock increases, not the other way around. Or, both could be caused by another even more perfect factor.

Without a causation analysis to factor for error, we cannot know, and its simply wrong to assume that one caused the other.

You're ignoring that he provides a trivial analysis that is easily tested: the claim that CEO compenstation is dominated by stock options or other stock-price-linked things. The high degree of correlation is a priori slightly weird, although it would be interesting to look more deeply into how probable it is that two such time-series would be so strongly correlated.

However, the causal account is easily tested: http://web.stanford.edu/group/knowledgebase/cgi-bin/2010/09/...

From the paper: "Based on a sample of 4,000 publicly traded U.S. companies, the average (median) CEO stands to gain roughly $58,000 in wealth for every 1 percent increase in stock price. Among the largest 100 companies, this figure approaches $640,000.6"

So the hypothesis the observed correlation was used to generate appears to be plausible given the further evidence that CEO wealth really is tied to stock price, whether directly through options or other means.

This is how we actually use statistical analysis, as part of a dialog between reality and propositions, rather than repeating the empty matra "correlation is not causation", which while true is something I too often see deployed to denigrate interesting and potentially fertile observations.

Whenever we have the urge to say "correlation is not causation" we should step back and ask, "How can I investigate the question more deeply?" Often more fruitful avenues will present themselves to the inquring mind.

I agree that it doesn't improve causation but it is more realistic to expect that their increase in pay strongly correlates with the S&P 500 as stock price is the underlying vehicle of their compensation.

>stock price is the underlying vehicle of their compensation.

Would not company performance be the vehicle, and stock price simply one measure of performance? It's not like stock price and business performance are identical, it's very possible for a company to be over or under valued.

Could we not say that company performance causes both stock price AND ceo pay, thereby showing that stock price and ceo pay are NOT in a causative relationship, only correlated because both are caused by actual business performance?

Stock price is what the market thinks of company performance, not to mention the metric of company performance CEO compensation is usually explicitly linked to.

I'm not persuaded of the existence of some more accurate measure of company performance than market cap that CEO compensation committees are aware of but Wall Street isn't.

Their bulk of their pay comes from the granting of stocks, whose value is driven by their market price. This is was the point I was basing my statement off of.

I remember reading Krugman's article on CEO pay and just shaking my head. The reality is that he is an economic writer and not an economist any longer. I'm sure there are plenty here who will disagree but I think Krugman long ago lost his credibility as "a leading economist." Anyone who remembers the "broken windows theory and iPhone upgrade cycle" article couldn't help thinking, really?? This guy won a Nobel Prize?

The reality is that he's a political writer (a.k.a. pundit) trading on economist credentials.

Krugman expressing his political opinions is worse than Chomsky. At least Chomsky doesn't dress his political opinions up as linguistics.

I remember reading the same blog post and here is what I remember: He told an anecdote told to him by someone else, wrote multiple times that he finds it ridiculous and pointed out that it is most likely a funny coincidence.

Sorry but my sympathy is with Krugman on this one - he didn't use his academic credentials as an argument on authority. It is a just a bit of brainstorming on his blog.

If anything it makes me think more highly of him - being always serious is harmful, a creative mind must dare to be silly from time to time.

By your logic, imagine a world 50 years from now when the S&P is 500 times the size it is today (mostly due to inflation), CEOs will get paid 150000 times what workers are paid. That doesn't really make sense.

I think you may be missing the point. I don't think the boom in equities escaped Krugman or other observers. I think the point is more that boom or no, boards get to approve comp packages. And when designing a comp package, it is not all that difficult to ensure that the CEO does/does not qualify for gargantuan sums of money on meeting performance targets. Further, taking this step can lead to better alignment with shareholders.

For example, one frequently recommended way to improve compensation is to compensate on alpha vs. a peer group and/or index. Clearly, every CEO in the S&P 500 didn't deserve the massive ramp in comp implied by the charts.

Bigger picture, just like with any employee, management usually has an idea of the total package value in mind and then backs into how the employee gets there. If the Board starts with an expectation that the CEO should make a gazillion dollars, they can design a package to get there. If the expectation is that the CEO should earn less, they can design a package to do that as well. The question remains why this expectation changed.

Couldn't the causation also have run the other way? Say, increased CEO salaries for companies with rising stock prices led CEOs to optimize for the short term, with layoffs, cutting research departments and pension plans, etc.?

Or: rise in CEO pay is cause of the rise stock market price, after all, you wouldn't vote to raise the CEO's wages if you didn't think it was going to improve the company now, would you?

Reply on news.ycombinator.com