CSR Performance Proxies in Large-Sample Studies: 'Umbrella Advocates', Construct Clarity and the 'Validity Police'

Abstract

Corporate social responsibility (CSR) performance is multidimensional, and is often considered an “umbrella” construct (Gond & Crane, 2010). This multidimensionality has generated substantive concerns regarding the construct validity and reliability of common empirical proxies of CSR performance (CSRP); thereby attracting the “validity police” (Rowley & Berman, 2000; Gond & Crane, 2010). Although the breadth and depth of CSRP measurement continues to increase, its links to an underlying, agreed-upon theory remains in doubt. Meanwhile, a market for quantitative measurement of CSRP has grown dramatically. The major providers of CSR ratings are: MSCI (ESG STATS, formerly known as KLD); Thomson Reuters (ASSET4); and Sustainalytics (ESG Indicators). Almost all large-scale empirical studies dealing with CSR use only one of the three providers as the source for their CSRP proxies. The purpose of our study is to review the CSRP constructs and proxies employed in studies published in a select set of influential journals and working paper series, and to determine whether the results might be influenced by the selection of CSRP proxies. We situate our analysis within broader academic debates concerning the calibration of construct and proxy convergence and precision, which provides a basis for evaluating the empirical durability of CSRP constructs across proxy specifications. Using a combination of statistical techniques and interviews with professionals involved in the CSR ratings and investment industry, our analyses show the potential for empirical results to be significantly sensitive to proxy selection and provide qualitative evidence that helps explain the need for more precise and thorough construct and proxy development. We assert that this research is critically important because the accounting literature builds on significant results generated by different proxies for CSRP.