Do Managers Do Good With Other Peoples’ Money?

The following post comes to us from Ing-Haw Cheng of the Department of Finance at the University of Michigan; Harrison Hong, Professor of Economics at Princeton University; and Kelly Shue of the Department of Finance at the University of Chicago.

In our paper, Do Managers Do Good with Other Peoples’ Money?, which was recently made publicly available on SSRN, we test the hypothesis that corporate social responsibility is due to agency problems using two quasi-experiments. First, we use the 2003 Dividend Tax Cut as a quasi-experiment that increased the marginal cost of pet projects or perks for managers, especially for firms with low insider ownership. In our model, managers with low ownership stakes invest more in pet projects (which we will also call perks consumption) and are not at their first-best levels of capital investment. Their marginal cost to engaging in a dollar of perk spending is the after-corporate-tax marginal product of capital. In contrast, managers with high ownership stakes are closer to their first-best levels of investment, or already at their first-bests, so their marginal cost to a dollar of perk spending is paying dividends and getting the risk-free rate, which is of course lower than the after-corporate-tax marginal product of capital.

Taken together, we expect corporate social responsibility measures to fall on average after the dividend tax cut. More crucially for our identification of principal-agent issues, our model predicts that social responsibility measures for low insider ownership firms should fall more than the same measures for high insider ownership firms.

We use social responsibility measures from Kinder, Lydenberg and Domini (KLD) of S&P 500 firms between 1991 and 2008. These KLD goodness scores cover various attributes of corporate social responsibility behavior, including community relations, product characteristics, environmental impact, employee relations, and diversity. Firms with high KLD scores tend to have good community relations, no harmful products subject to litigation, no pollution or Environmental Protection Agency (EPA) violations, few labor strikes and a diverse board or set of top officers.

Consistent with the agency hypothesis, we find that firms’ goodness scores fell on average after the 2003 Dividend Tax Cut and that the scores of low insider ownership firms fell by relatively more. Specifically, we find that 9 percent more of low-ownership firms reduce their 2003-2004 goodness below their 2001-2002 average, relative to high-ownership firms, with 42 percent of total firms cutting goodness scores over this period. The average score reduction among low-ownership firms was 0.13 while the change in scores among high-ownership firms was negligible, representing an economically large 22% decline of the pre-tax-cut mean of 0.56 for the low ownership group.

A more dynamic analysis of the scores themselves reveals that average goodness was higher among low-ownership firms compared to high-ownership firms from 1996-2002. Beginning in 2003, the gap began to narrow and goodness among low-ownership firms even fell below the average goodness for high-ownership firms starting in 2004. Because comparing the levels of average goodness across portfolios could mask other unobserved heterogeneity, we use a differences-in-differences approach in an annual panel to tease out the effect of the tax cut. Because goodness scores fall for both groups before the tax cut, we also fit a trend using a four-year pre-tax-cut window (1999-2002) and four-year post-tax-cut window (2003-2006). We find that the goodness trend for the low-ownership group becomes even more negative than the trend for the high-ownership group after the dividend tax-cut. In other words, low-ownership firms cut goodness scores faster than high-ownership firms after the tax cut.

Our second quasi-experiment is motivated by the following approach to agency theory, which suggests that there are two substitutable ways to encourage managers to maximize firm value: (1) increase their incentives, which is the focus of our dividend tax cut analysis, or (2) increase monitoring through improved governance. Hence, we corroborate our first finding using a natural experiment with exogenous changes in firm governance. We exploit a regression-discontinuity experiment using close proxy contests regarding shareholder-initiated governance proposals. The identifying assumption is that close votes around the 50 percent passage cut-off are random in terms of whether a governance proposal gets passed and represent plausibly exogenous shocks to the monitoring of the firm. We build on earlier work, which finds that, while shareholder proposals are non-binding, close votes around the 50% cut-off lead to discontinuously substantial changes in the implementation of governance proposals (e.g., Ertimur, Ferri and Steuben, 2010 and Thomas and Cotter, 2007). Our analysis builds most closely on Cunat, Gine, and Guadalupe (2011), which uses a voteshare regression discontinuity approach to show that shareholder proposals increase firm value.

We find that firms in which shareholder proposals narrowly pass experience much slower growth in goodness scores than firms in which the proposals narrowly fail. Indeed, the economic magnitudes are sizeable. We find that firms that just failed these proposals experienced greater annual growth of KLD scores in the year of the vote those that just-passed: the difference in the change in KLD scores is around 2. Given that a standard deviation of the change in KLD scores is close to 2, this difference in growth rates for firms around the 50% cut-off is nearly one standard deviation of the left hand side variable. These effects are particularly strong for shareholder proposals which were voted on in 2009, the year after the financial crisis, when there were significantly more shareholder proposals and hence close votes. One interpretation for our strong results is that in the aftermath of the crisis, shareholders worried much more about agency problems and value extraction.

The full paper is available for download here.

Both comments and trackbacks are currently closed.