Abstract
Not long after the formation of the United Nations Global Compact (UNGC) in 2000, two opposing theories emerged regarding its efficacy and why organizations continue to join UNGC. The critics take the position that because it has low barriers to entry and no enforcement of compliance, it attracts organizations with low CSR performance who merely want to enhance their reputations. The advocates reject these arguments because of their belief in the purpose of the UNGC, to offer a platform for learning and improvement, especially for under-resourced organizations. Haack, Martignoni, and Schoeneborn have offered a conceptual framework that has the potential to bridge the differences between these two opposing theoretical positions by suggesting that CSR can be adopted ceremonially under conditions of opacity and evolve to substantive adoption over time as transparency increases. In this study, we use the Haack et al. conceptual framework to empirically test this proposition by investigating U.S. corporations that have joined UNGC. We expand the analysis to examine the motivations for ceremonially adopting CSR. Our results support the conditions proposed by Haack et al., and we emphasize the importance of organizational learning to achieve substantive adoption of CSR practices over time.
Notes
1 AGE represents the age of the firm from its Initial Public Offering; LOGEMP is a size measure based on the log of the number of employees at the time the firm joined the global compact; and DSALES is also a size measure operationalized as a dichotomous variable split at the median sales volume for firms in the sample.
2 We were unable to find useful firm age data – for example, the firms considered include Ford, incorporated in 1903, but along with other auto companies it instituted large structural changes in 2008; one could reasonably argue that it is faulty to characterize Ford as a 115-year-old firm.
3 The choice of Logit or Probit for estimating a dichotomous dependent variable model is largely one of preference; the results rarely differ. The Logit model assumes the random error of the model is distributed following a logistic probability distribution; Probit assumes the error is normally distributed, while the logistic is somewhat thicker in the tails (see, for instance, http://visionlab.harvard.edu/Members/Anne/Math/Logistic_vs_Gaussian.html).
4 A variance inflation factor analysis was performed and, while none were very high, the VIF for sales, 7.8, was of sufficient concern that we produced Model II.
Additional information
Notes on contributors
James Barrese
James Barrese is a professor of risk management and insurance at the Maurice R. Greenburg School of Risk Management, Insurance and Actuarial Science at The Peter J. Tobin College of Business, St. John’s University in Manhattan, New York. His research interests include risk mitigation, the legal and ethical consequences of risk-based decisions, and the impact of risk in a global context.
Cynthia Phillips
Cynthia Phillips is an associate professor of accounting at The Peter J. Tobin College of Business, St. John’s University in New York. Her research interests include not-for-profit accounting, governance and social issues in corporations and not-for-profits, including higher education, and accounting pedagogy.
Victoria Shoaf
Victoria Shoaf is an associate professor of accounting at The Peter J. Tobin College of Business, St. John’s University in New York. Her research focuses on U.S. and international accounting standards, corporate social responsibility and business ethics, and accounting pedagogy.