Post-innovation CSR Performance and Firm Value

While we ignore agency problems as drivers of CSR activities in formulating our hypothesis, we do control for firms’ agency issues in many ways in our empirical analysis.

For robustness, we also replicate Tables 3 and 6 (next subsection) by correcting the standard errors for one-way (by firm) and two-way (by firm and time) clustering. Our conclusions remain unaffected following these corrections.

Our main tests employ Tobin’s Q as the main dependent variable. For robustness, we replicate Table 6 using average risk-adjusted returns (i.e., returns in excess of the predictions of the Fama–French three-factor model, in which the risk-free rate is the 1-month T-bill yield) for the 12 and 24 months following the observation of CSR. We find that these returns generally decrease with a firm’s level of innovation but increase with the level of innovation for socially responsible firms, which suggests that innovative firms that engage in CSR as strategic choice reap higher valuation post-innovation. A univariate analysis shows innovative high-CSR (CSR_N >0) firms earn approximately 6.36 % (in 12 months) and 9.62 % (in 24 months) more post-innovation compared with low-CSR (CSR_N <0) firms.

The P _IN dataset was obtained from Professor Stephen Brown’s website.

We conduct several additional tests. First, we use the number of citations per patent and patents scaled by R&D expenses as proxies for innovation. Second, we employ a strong innovation dummy for firms with more than the industry–year average number of patents. Third, in selecting our sample we exclude firms that are not in an industry that reported at least one registered patent per firm–year. Fourth, we repeat our analysis using a CSR index that includes corporate governance qualitative area scores (originally excluded). Fifth, certain items are not covered by KLD for the entire sample period, which might potentially cause the index to endogenously change over time. To mitigate this issue, we estimate industry–year-adjusted CSR_N by subtracting the industry CSR_N medians. Our results continue to remain practically as expected in all these tests.