We acknowledge at least three significant limitations associated with our research design
choices. First, our proxy for financial reporting quality is based on whether companies meet or beat
their consensus analyst earnings forecast, and our proxy undoubtedly measures this underlying
construct with noise (Prawitt et al. 2009). Specifically, we could observe that a company meets or
beats its consensus analyst earnings forecast if the company’s underlying economic performance
was truly in line with (or better than) expectations, or if the company’s management guided their
analyst forecasts down to a reasonable earnings expectation (Richardson, Teoh, and Wysocki
2004). Neither of these explanations for why a company meets or beats its consensus analyst
earnings forecast would be due to lower financial reporting quality. Thus, there is substantial noise
in our dependent variable. Second, we primarily examine one proxy for financial reporting quality.
Future research may wish to systematically examine whether our results hold if other measures,including the many alternative methods for calculating discretionary accruals, are used as proxies
for financial reporting quality. Future research may also wish to examine whether audit committee
equity incentives (option or otherwise) are negatively associated with more drastic measures of
financial reporting quality, such as financial statement fraud or restatements. Finally, we focus on
the monetary value of equity-based incentives to capture the total wealth effect of these incentives
around the time financial reporting quality is determined, as this should provide sharper analyses
than if we were to use an indicator variable for whether a director has option compensation or
holdings and/or a simple percentage of stock held by the director(s). However, this design choice
comes with a cost. As discussed in Appendices A through C, there is a substantial amount of handcollection
necessary to calculate our variables of interest. To reduce the cost of hand-collection, we
limit our sample to those instances right around the meet-or-beat benchmark—to those companies
that miss earnings by up to three cents or meet earnings by up to two cents. Although our results
should be most pronounced for these companies, admittedly this design choice is arbitrary and
limits the generalizability of our findings.