The effect of corporate governance on sustainability disclosure

Social and environmental disclosure has been studied according to many theories: legitimacy theory; political economy theory and stakeholder theory. Those theories share many similarities and overlap with each other. In particular, according to legitimacy and stakeholder theories, a company is part of a broad social system in which the company and the society influence each other. As suggested by Gray et al. (1996), because of the overlap between stakeholder and legitimacy theory, to consider them as competing theories would be wrong. Deegan and Blomquist (2006: 349–350) point out that the main difference between the two theories is that “whilst legitimacy theory discusses the expectations of society in general, stakeholder theory provides a more refined resolution by referring to particular groups within society”. Although such similarities, stakeholder theory fits particularly well in our analysis because community influential board members serve as boundary spanners linking a company to specific stakeholders.

For example, an Italian listed company appointed as board member the past Pope’s spokesperson.

We did not consider disclosure on the website because the content analysis cannot be performed reliably and consistently given that we cannot track when the web-pages are published or updated.

Haniffa and Cooke (2005) note that the use of dichotomous procedure is considered a limitation, because it treats disclosure of one item as equal to a company that makes 50 disclosures and does not indicate how much emphasis is given to a particular content category. Nevertheless the advantage is that it gives coders less choice, thus being more reliable (Hackston and Milne 1996; Raffounier 1995). Moreover, the dichotomous coding allows us to gather the variety of information disclosed and it does not depend on how the sentence is constructed—which may be relevant for an international comparison. Examples of our items are: “Percentage of materials used that are wastes (processed or unprocessed) from sources external to the reporting organization”; “Total recycling and reuse of water”; “Greenhouse gas emissions”; “Employee benefits beyond those legally mandated”; “Standard injury, lost day, and absentee rates and number of work-related fatalities”, “Awards received relevant to social, ethical, and environmental performance”, etc. We will discuss further limitations of our disclosure index in the last paragraph.

Assume that company A presents along with some good practices some bad practices. The company could provide a great number of pieces of information on good practices avoiding completely to mention the bad practices, thus satisfying a specific group of stakeholders. Assume now that the company B provides a lower number of pieces of information but at the same time it covers all the topics. It is not clear which is the best disclosure behavior.

As the number of items increases the difference between a dichotomous procedure and the count of the number of sentences decrease. This explains why we include in the list of items also some suggested by Epstein and Birchard (2000).

The coding activity may suffer from two different consistency problems: i) across coders, ii) over time. Therefore, by having only one coder we are able to avoid the first problem, thus improving the overall reliability of the disclosure index.

In order to test whether relevant multicollinearity is affecting the results, we performed the Variance Inflator Factor (VIF). The largest value among all independent variables is often used as an indicator of the severity of multicollinearity (Neter et al. 1996). A maximum VIF value in excess of 10 is frequently taken as an indication that multicollinearity may be unduly influencing the least square estimate. In our case, the largest VIF is equal to 4.22, so multicollinearity among the predictor variables is not a problem.

Our results are robust to alternative statistical proxies. We measured size alternatively as the logarithm of sales, market value and asset and we obtain the same results. Results do not change when employing ROA as the performance measure. We have also run the regression using the number of community influentials instead of the proportion of CI and we obtain the same results.

Instead of presenting the results for all governance related variables, we show only the results about community influentials given that the other governance variables are not significantly correlated with corporate sustainability disclosure.

Since the results of the second equation are consistent over the different models, we do not report them.