The Wall Street Journal Touts Dubious Research (CEO Performance Edition)

Will someone, please, teach the reporters at the Wall Street Journal the basics about scientific research? I know it’s hard finding stuff to write about day in, day out. But the story “Scholars Link Success of Firms To Lives of CEOs” is a travesty.

The centerpiece of the article is a study by Morten Bennedsen, Francisco Pérez-González and Daniel Wolfenzon. Let’s note first that this paper has not yet been published in any recognized academic journal (it’s posted on the University of Texas website, where Francisco Pérez-González is a member of the faculty), so it is not yet clear whether it will be deemed to pass muster in respectable circles.

The study took the records of 75,000 Danish businesses from 1992 to 2003 and looked at whether events in the CEO’s life affected performance. The chart summarizes some of their findings. The death of child had the greatest negative impact on performance, followed by the death of a spouse. Conversely, the death of a mother in law was a plus.

Now this study has a certain intuitive appeal. Someone who is grieving might devote less time to his business, or make worse decisions than he would otherwise.

Nevertheless, correlation is not causation. A study of this sort at most highlights a possible connection.

Even at the gross level, the findings may not add up. Psychologists regard the death of a spouse as the most traumatic event one can suffer. One measure, the Holmes-Rahe scale (which has limitations of its own since the underlying research was done retrospectively, and was developed to see how stress levels translate into physical illness), rates death of a spouse as 100 while death of a close relative (which is where “death of a child: would fall) as 63. Yet this study found greatest drop in performance from the death of a child.

There are also differences between the demographics of the firms where the CEOs suffered losses compared to the ones that didn’t. This also makes it difficult to give too much credence to the findings:

Table I shows that event firms are larger, more profitable, older and grow faster than non CEO-shock firms, in all cases with differences that are statistically significant at the one-percent level. On average, event firms’ age is 15.5 years, while it is only 11.2 years for non-event firms. The differences in age between event and non-event firms are expected as CEO shocks are more likely to occur in relatively older firms as CEOs family size and age are larger for older firms.

Even if this research does have some validity for smallish companies in Denmark, it’s not clear it can be generalized to bigger corporations in the US. Large companies in fact may provide a refuge from painful events. As the Journal notes:

Gerald M. Levin was chief executive of Time Warner Inc. in 1997 when his grown son was murdered. “Of course I went into a tailspin,” he said. “I made…I won’t call it a mistake. I returned to what for me was a narcotic, I returned to work. I worked 25 hours a day.” He said he couldn’t judge whether his performance was affected but notes that he felt drained of emotion, as though “nothing that happened could affect me anymore.” Mr. Levin’s grief didn’t correlate with a drop in Time Warner’s stock price, which greatly outperformed the broader market during the three years after his son’s murder.

It’s only one datapoint, but someone betting against Time Warner after the death of Levin’s son based on the study above would have made a mistake.

The Journal article reports on other correlation studies that are at least amusing, like the one by David Yermack and Crocker Liu of Arizona State University that found that CEOs that built or bought a particularly large house had stocks that underperformed the S&P for the next three years.

But for others cited, it’s too easy to come up with alternate explanations:

Two Penn State professors recently attempted to rate CEOs of technology companies on their degree of narcissism. They looked at things like the size of executives’ photos in annual reports and how often they use the first person singular in press interviews. The authors concluded that narcissistic executives tended to take greater risks, leading to bigger swings in profitability of their companies. The study, called “It’s All About Me,” is to be published in Administrative Science Quarterly.

Gee, that result could just as easily show that CEO that are bad at hiring (or supervising) corporate communications firms are also bad at hiring/managing other staff, and that lax control is the cause of the swings. There’s no proof here that the narcissistic-seeming annual report was the CEO’s doing.


Other academics have found underperformance, in both profits and stock prices, at companies led by executives who received awards such as best-manager kudos from the business press. The theory: Once they become stars, some CEOs may pay more attention to writing memoirs and sitting on outside boards and a little less to running their companies.

Ever heard of reversion to the mean?

It’s one thing to present this sort of fluffy research as a curiosity, quite another to treat it, as the Journal does, with unwarranted reverence.

Print Friendly, PDF & Email


  1. Lune

    How long before compensation committees use these studies to justify paying for full-time bodyguards for CEO’s family members? And an even larger corner office, the better to increase their CEO’s narcissim. :-)

    Of course, the studies showing underperformance when a CEO buys a bigger house or gets cited in the business press will be quickly dismissed as quack research done by a bunch of hacks with no merit whatsoever. :-)

  2. Yves Smith

    Dear Anon of 10:31AM,

    I read the abstract of the paper (not the full version, since it’s not free, otherwise I would). I don’t buy their methodology for identifying the comparison group CEOs. They chose ones “who had similar characteristics and performance to the award winners in terms of company size, book-to-market and returns over the year leading up to an award.”

    First, “book to market” is not a good measure. “Book” can and often is non-comparable by virtue of goodwill. It’s a peculiar measure of stock price performance. Similarly, they don’t mention what other measures they used for “performance”.

    This begs the question that these “peers” may not have been close peers at all. And how were the industries defined? Some companies have no close comparables. How would you find a comparable to GE, a multi-industry manufacturer which gets about 40% of its income from financial services? Some have criticized Jim Collin’s Good to Great peer analysis because he had only one peer for each of his chosen companies (the MIT work is silent as to whether they had one or more peers).

    Phil Rosenzweig, in his excellent book The Halo Effect, discusses how companies that are identified by the media and experts as high performers are seen as doing everything right, until performance slips and they are suddenly depicted as doing everything wrong. The Halo Effect is a well-documented psychological tendency of humans to fail to discriminate among performance vairables.

    One example Rosenzweig discusses in detail is Cisco, which suffered badly in the dot com bust. Cisco was portrayed as brilliantly run when it was doing well, and terribly run when the dot coms came apart. But nothing had changed in the way it was run. Its market had fallen apart. It was the attributions about its performance that had changed.

    So I’m not convinced by what I read about the peer analysis. And note they did use a stock market price measure as one of their measures. Stock prices move in parallel with media coverage. It may be that the CEOs who got awards were in companies that were media darlings (that’s why they got the awards) and then the media turned on them.

    A lot of people similarly use magazine covers as the sign of when to sell a stock, particularly a Business Week cover. The notion is by the time Business Week is on to a trend, it’s past its sell-by date. The same may be true for the award givers in enough cases to skew the results: they only select CEOs with a well established run of outperformance, one that is soon due to revert to the mean.

    The point is it doesn’t matter whether any of my theories are correct. The point is there are alternate theories to explain the phenomenon. Saying or suggesting that correlation is causation is bogus. Anyone who is a decent researcher, including the guys who wrote these papers, should know that, as should the Journal.

  3. Anonymous

    So here is another way to boost economic growth, sequester CEO children and provide them with the best of facilities for their health and well being.

    CEO child deaths as a leading economic indicator? Come on Universtity of Michigan!!

  4. A Reader

    There are 5.5 million people in Denmark. Let’s say 3 million are working in these 75000 companies. That makes an average company size of 40 employees, and probably a median size smaller than that, let’s say 20.
    Of course you would expect a company of 10-20 employees to have a bad year if the CEO has deep emotional trauma due to the death of a relative. But that shouldn’t apply to most of the huge companies in the S&P 500, it’s common sense.

Comments are closed.