CEO Stress, Aging, and Death

Ulrike M. Malmendier is Edward J. and Mollie Arnold Professor of Finance at University of California Berkeley Haas School of Business, and Professor of Economics at the University of California Berkeley. This post is based on a recent paper authored by Prof. Malmendier; Mark Borgschulte, Assistant Professor of Economics at the University of Illinois Urbana-Champaign;  Marius Guenzel, Assistant Professor of Finance at the Wharton School of the University of Pennsylvania; and Canyao Liu, PhD student at Yale School of Management.

CEOs work long hours, frequently make high-stakes decisions, such as layoffs and plant closures, and face uncertainty in times of crisis (Bandiera et al. 2020, Porter and Nohria 2018). They are closely monitored and criticized when their firm is underperforming, and the notion that CEOs are “overworked [and] overstressed” is prominently discussed in the media (cf. CNN’s Route to the Top segment from 3/12/2010).

In this paper, we estimate the long-term effects of experiencing high levels of job demands on the mortality and aging of CEOs. Despite the fact that job demands and work-related stress are increasingly recognized to be key determinants of health (see, e.g., Marmot 2005 and Ganster and Rosen 2013), there is little quasi-experimental evidence that links job demands and stressors at work directly to health outcomes in the general population, let alone in the CEO context. We use new data on the lifespan of CEOs and recent machine learning (ML) apparent-age estimation techniques, combined with photographs of CEOs’ faces, to provide evidence on how varying levels of job demands shorten managers’ lifespan and induce visible signs of aging.

Our analysis has three main parts. In the first part, we relate variation in the intensity of CEO monitoring due to corporate-governance legislation to CEO mortality. In the second part, we exploit variation in job demands due to industry-level distress shocks, and also study the effect on CEO mortality. In the third part, we continue to exploit industry-level distress shocks, here from the Great Recession, and relate them to visible signs of accelerated aging, identified by neural-network based ML estimations.

In the first part of the analysis, the source of identifying variation is the staggered passage of anti-takeover laws across U.S. states in the mid-1980s. The laws shielded CEOs from market discipline by making hostile takeovers more difficult. Prior research has documented that they reduced CEOs’ job demands and allowed them to “enjoy the quiet life” (Bertrand and Mullainathan 2003). The prevailing view in law and economics at the time of the passage of the laws was that the “continuous threat of takeover” is an important means to counteract lagging managerial performance (Easterbrook and Fischel 1981). While some later studies question whether the passage of anti-takeover laws in fact reduced hostile takeover activity (Cain et al. 2017), it arguably constituted a significant shift in managers’ perception of their job environment.

For this analysis, we construct an augmented version of the data from Gibbons and Murphy (1992) and hand-collect the exact dates of birth and death of more than 1,600 CEOs of large U.S. firms included in the 1970-1991 Forbes Executive Compensation Surveys. Using a hazard regression model and controlling for CEO age, time trends, industry affiliation, and firm location, we find that anti-takeover laws significantly increase the life expectancy of incumbent CEOs. One additional year under lenient governance lowers mortality rates by four to five percent for an average CEO in the sample. The estimated effect sizes are large, both on their own and relative to other variables affecting life expectancy. For example, in our CEO data, one additional year of age increases a CEO’s estimated mortality hazard by roughly twelve percent. Thus, for a typical CEO, the effect of the anti-takeover laws is roughly equivalent to making the CEO two years younger. The effect size is comparable to removing known health threats. For example, smoking until age 30 is associated with a reduction in longevity by roughly one year, and lifelong smoking with a reduction by ten years and more (General 2014, Jha et al. 2013).

In the second part of the analysis, we consider industry distress shocks. Typically defined based on a 30% median firm stock-price decline over a two-year horizon, industry shocks have been used to study effects on, e.g., market concentration, creditor recoveries, and employee exit (Opler and Titman 1994, Acharya et al. 2007, Babina 2020). In our analyses, they constitute a separate and oppositely-signed change in job demands compared to anti-takeover law passage. About 40% of CEOs in our sample experience at least one such industry-wide downturn during their tenure. We find that distress exposure significantly increases a CEO’s mortality risk. The estimated mortality effect is equivalent to increasing age by 1.5 years, and comparable to serving three fewer years under lenient monitoring.

In the final part, we document more immediate health implications in the form of visible signs of aging in the faces of CEOs. We utilize machine-learning algorithms designed to estimate a person’s apparent age, i.e., how old a person looks rather than a person’s biological age from Antipov et al. (2016). We collect a sample of 3086 pictures of the 2006 Fortune 500 CEOs from different points during their tenure to estimate differential apparent aging in response to industry-level exposure to the financial crisis. Using a difference-in-differences design, we estimate that CEOs look about one year older in post-crisis years if their industry experienced a severe decline in 2007-2008 relative to CEOs in other industries. The estimated difference between distressed and non-distressed CEOs increases over time and amounts to 1.18 years for pictures taken five years and more after the onset of the crisis. To the best of our knowledge, this represents the first application of visual machine learning to a quasi-experimental research design.

Our findings both pave the way toward a better understanding of how work-related stress affects the working population in general and have important implications for high-profile corporate positions specifically. CEOs bear the ultimate responsibility for the success of the firm and satisfaction of employees. Given their overarching importance within their firms, it matters how incentives and performance affect CEOs personally. Additionally, the health implications of CEOs’ job demands affect their ability to stay on the job and, if anticipated, their willingness to select into the CEO job.  Our analysis might thus speak to the prevalence of certain CEO characteristics and possible feedback effects: Are aspiring CEOs (over-)confident about their health? Are women vastly underrepresented in the C-suite not only because of discrimination but also because they (correctly) anticipate the health costs of assuming such positions?

Our findings suggest further avenues of investigation. Besides the question of whether managers fully account for personal costs as they progress in their careers and how this affects selection into CEO service, one important question is which jobs and hierarchy levels come with the largest adverse health consequences, also in light of looming financial hardships. Another promising avenue is the more fine-grained identification of stressors. What aspects of individual job situations and which decisions tend to have the largest adverse health consequences, for either management or regular employees? Likewise, heightened workplace stress can also adversely affect other aspects of life, including marriage, divorce rates, parenting, and alcohol consumption.

The complete paper is available for download here.

Trackbacks are closed, but you can post a comment.

Post a Comment

Your email is never published nor shared. Required fields are marked *

*
*

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>