Philosophical Musings

August 23, 2007

Some More Shoddy Statistics

Filed under: business,investing,statistics — Elad Kehat @ 10:47 am

Founders know best, says USA Today: “Firms tend to prosper with founders at the helm“.
As a startup founder myself, I was naturally intrigued – what data supports this conclusion?

“Going back 15 years, stocks in founder/CEO companies have surged an average 970%, vs. a 222% gain for the S&P 500, according to data from S&P’s Capital IQ.”

Oh. See the problem? If not, read on…
The article goes on to try and explain (unsurprisingly, with no further hard data) why it is that founders are so successful at managing their own companies. No further consideration is given to the idea that this may just be a false correlation.
So here’s the problem: Del Jones and Matt Krantz, the USA Today reporters, compare the 15 year performance of companies with founder at the helm vs. companies in general.
What about an alternative explanation: public companies that have the same CEO at the helm for 15 years must have a damn good CEO. Otherwise they wouldn’t be successful and the board would find someone else for the job.
It’s nice that they use a system called “Capital IQ”, but a little more of it (or a passing understanding of statistics, or a minimum level of critical thought) would have prompted them to dig further and compare those founder/CEO companies to other companies that had the same CEO for the past 15 year. Would that study yield the same result? Or would that just lead to no article and an angry editor?

Or maybe it isn’t reporters fault:
“Ohio State University finance professor Rudi Fahlenbrach”. This is from a top-60 university?

And the “experts” seem to fall for it with no trouble at all:
“I should’ve attached more attention to it over the years,” says Rob Sellar, a money manager of Aberdeen Asset Management.
Really, Mr. Sellar? How about “attaching” more attention to what’s missing from the data? After all, there no telling in this “research” whether all companies headed by their founders succeed, just that public companies that had a leader successful enough to stay at the helm for the past 15 years are successful…

July 26, 2007

Lies, Damned Lies, and ComScore

Filed under: statistics — Elad Kehat @ 10:38 am

Andrew Lipsman, a senior analyst at ComScore, and “young adult” blogger has decided to check whether young adults like him (and me) aren’t reading the news as much as older persons. An interesting question no doubt. In order to check it, he naturally chose to use ComScore’s own data. The results are posted in his blog.

And now comes the usual blunder that results from mixing carelessness, otiosity and statistics: you reach the wrong conclusions, and you don’t even know it.

So here’s Andrew’s summary of his results:

“As you can see, nearly the same percentage of 18-34 year olds (59%) are reading news online each month as 35-54 year olds (61%). Not only that, but they are also going online to get their news nearly as many times each month (12 visits) as 35-54 year olds (13 visits).”

And here’s the wrong in oh-so-subtle way conclusion:

“So it’s not that young adults aren’t reading the news, they’re just doing it online instead of in newspapers.”

See the problem?

It’s entirely possible (even likely, though I haven’t researched it myself) that a far larger percentage of 35-54 year olds read dead-tree newspapers than 18-34 year olds. If that’s true, and you look at news-reading behavior in general, both online and off, then 18-34 year olds read far less news.

Further, it’s unclear whether the numbers refer to the entire 35-54 segment in the population, or only to those who regularly go online. I guess it’s the latter. I’ll make another educated guess and assume that as a percentage of the entire population there are far more 18-34 year olds that regularly go online than 35-54 year olds, so the number of online news-readers in the younger population represents a larger portion of that total population (online and off) than does the number of online news readers in the older population. Combined with what I explained in the above paragraph, that could mean that in absolute numbers, young adults do indeed read far less news than the older generation.

I guess it’s clear why Mr. Lipsman chose to suffice himself with the online-only data – it’s what is readily available to him as a ComScore employee. However, a senior analyst should have a better understanding of the process of extracting conclusions from statistic results. Let ComScore customers beware…

Blog at