Courtesy of Michael Mayberry
Shareholders are increasingly interested in corporate social responsibility (CSR). However, as with nearly all corporate actions, shareholders are not the decision makers. Managers are. Therefore, shareholders must rely on compensation to incentivize managers towards the level of CSR they desire. To date, researchers know very little about how compensation incentivizes executives towards a particular level of CSR. In a recent study, I examine how one important dimension of compensation—a CEO’s stock options—influences a firm’s CSR strategy. A stock option is a contract that gives the CEO the right, but not the obligation, to buy or sell shares of stock by a certain date. Stock options compensate executives for taking risks. This dimension of options that rewards for risk-taking is known as vega. The greater the stock option’s vega, the greater the reward for an executive’s risk-taking.
Vega and CSR
I hypothesize that stock option vega should be negatively related to CSR. Recent research suggests that CSR generates reputational and moral capital that reduces firm risk. This link is quite intuitive. When times are bad, consumers are more willing to buy high CSR goods and services (i.e. CSR can be a product differentiation strategy). Similarly, when bad things happen to firms with high CSR (e.g. restatements, OSHA violations, product recalls), investors and regulators penalize these firms less because they are more likely to believe the event to be “bad luck” rather than malicious. All else equal, executives should pursue CSR strategies to enjoy more highly valued shares because of this form of reputational insurance. However, remember vega. CEOs with higher vegas benefit from greater stock volatility. Therefore, all else equal, executives with higher vega should maintain a lower level of CSR as they personally benefit from more volatile stock prices. That is, vega should discourage CSR because it discourages hedging.
Correlating CSR measures with stock option vega is easy. But correlation is not causation. One concern with a simple correlation is that more risk tolerant CEOs might match to firms offering stock options with greater vega. These more risk tolerant CEOs would arguably maintain a lower level of CSR even without vega. Understanding causation—rather than mere correlation—is important to boards of directors and regulators. If vega causes poor CSR, then CSR-sensitive boards and regulators should monitor and carefully choose vega. If vega is merely correlated with poor CSR, then curtailing vega will not actually change firms’ CSR strategies.
To overcome this challenge, I exploit an exogenous change in the financial accounting for stock options called FAS 123R. This change in financial accounting represents a change in stock option vega that does not directly influence a firm’s CSR strategy. Thus, disentangling cause and effect is easier. Prior to FAS 123R, firms did not have to expense stock options granted with zero intrinsic value. After FAS 123R, firms were required to expense stock options at their fair market value. Following FAS 123R, firms granted fewer options to executives because such options would now lower earnings. Thus, FAS 123R should lower many executive’s vegas. Importantly, FAS 123R did not impact two types of firms: (1) those that did not grant stock options and (2) those that voluntarily expensed stock options at the fair market value before FAS 123R became effective. Using these two types of firms as “control firms” allows me to capture any change in CSR for reasons other than vega.
I first begin by documenting that firms impacted by FAS 123R (“treated firms”) decrease their stock option vega following the new accounting standard to a greater extent than firms that were not impacted by FAS 123R. I do not find evidence that treated firms alter other dimensions of executive compensation, such as cash compensation or incentives to maximize stock price. Second, I also find that treated firms’ stock return volatility decreases more than control firms following FAS 123R, suggesting that treated firms decrease their risk-taking when vega declines. Third, I confirm that CSR weaknesses are positively associated with future stock return volatility, suggesting that CSR weaknesses increase firms’ risk.
Having confirmed that FAS 123R represents exogenous changes in risk-taking incentives, I investigate vega’s impact on CSR. I measure CSR using scores from MSCI, an independent third party ratings agency. Following prior research, I consider CSR strengths (e.g. installing emissions reducing technology) and CSR weaknesses (e.g. having a history of toxic spills) separately. I find that the treated firms’ CSR weaknesses decline following FAS 123R’s reduction in vega to a greater extent than control firms. This result is consistent with risk-taking incentives encouraging socially irresponsiblebehavior. I fail to find an association between CSR strengths and risk-taking incentives, on average. These results are consistent with the theory expressed in prior research that “avoiding of harm” provides greater insurance-like benefits than the active “doing good” CSR. Moreover, my results are also consistent with some studies criticizing CSR strengths as mere “window dressing.”
I rule out many concerns that I am capturing something other than FAS 123R. First, I randomly assign firms to treatment and control using Monte Carlo analysis.[1] I do not find results, suggesting that random chance is unlikely to drive my results. Second, I also verify that my results do not relate to alternative, unknown explanations by running my model using two years before FAS 123R. If treated firms behave differently before FAS 123R, then it is likely that changes in vega did not cause my results. However, I do not find results prior to FAS 123R. Changes in CSR only happen following FAS 123R.
Next, I consider how the decline in CSR for treated firms varies with how much firms are likely to gain from the insurance-like benefits of CSR. All else equal, firms that are more likely to benefit from CSR through decreased stock return volatility should exhibit a greater decline in CSR weaknesses. I find firms in litigious industries, firms selling differentiated goods and services, and firms in industries that are more reliant on employees, experience greater reductions in CSR weaknesses. Thus, vega appears to encourage socially irresponsible behaviors to a greater extent in these settings.
Conclusion
My study is the first to consider how the risk-taking incentives provided by stock options (i.e. vega) influence a firm’s CSR. I find evidence suggesting that vega encourages firms to behave in socially irresponsible ways. Multiple parties can benefit from my paper. First, compensation committees of boards of directors should find my study interesting because it highlights a possibly unforeseen consequences of stock option incentives. Given that many shareholders and stakeholders express a desire to increase firms’ overall social responsibility, understanding how stock options encourage CSR is of high importance. While investors likely benefit from vega’s incentives for CEOs to increase innovation and encourage business expansion, my paper also implies boards and shareholders should cautiously consider both the costsand benefits of stock options when crafting compensation contracts.
Second, my study should also benefit regulators and academics. Both parties have suggested that executive compensation may be broken within corporations. Other studies suggest that stock option vega may encourage accounting fraud and other negative contagions. My study suggests that regulators and academics should also be cognizant of vega’s impact on society and that it may not strictly be limited to financial misbehavior.
[1] Monte Carlo analysis randomly assigns firms to treatment or control. For example, my sample has 1,686 treatment firms. Using the entire sample, I randomly designate 1,686 firms to be “placebo” or fake treatment firms. I re-estimate my model using these placebo treatment firms and save the coefficient one thousand times. I then compare my actual model to the results generated through randomness. My actual results exceed the randomly generated results 987 out of one thousand times. These results suggest that random luck is unlikely to drive my results.
