4 ms·
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 coi
by adventured 12y ago
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.
- Phlarp 12y agoIt seems to decouple quite noticeably around 2008.
- notahacker 12y agoThe decoupling in 2008 is plausibly explained by increased pressures for pay restraint following the financial crisis.
- Phlarp 12y agoI agree, but now we're back to confusing correlation with causation.
- breischl 12y agoNote 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.
- criley2 12y agoVery 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.
- tjradcliffe 12y agoYou'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/2010/09/17/sensitivity-of-ceo-wealth-to-stock-price-a-new-tool-for-assessing-pay-for-performance/ 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.