Online Appendix to “The Role of Social Media in the Capital Market: Evidence from Consumer Product Recalls”

We construct an indicator variable (HighScore) to aggregate and summarize four recall characteristics that we expect to be associated with higher overall recall costs, including greater loss of reputation and future sales as well as higher expected litigation costs.1 We employ the dichotomous variable, HighScore, because it is easy to interpret and because the relation between the individual recall characteristics and overall recall costs is not necessarily linear. In Footnote 26 of the paper, we state that the results are robust when we use Score (a simple summation of indicators across the four recall characteristics) instead of the indicator variable HighScore and when we disaggregate Score into its four components. We present a summary of these results (pvalues are in parentheses) in Table A1. All of the main inferences remain unchanged using either Score or its components. Specifically, the coefficient on RecallScale × SocialMedia remains positive and significant in the regression examining social media (H1); the coefficients on RecallScale × SocialMedia and RecallScale × SocialMedia × Interactive are positive and negative, respectively, in the regressions examining the interactivity of social media (H2); the coefficients on RecallScale × ATweetFirm and on RecallScale × ATweetOther are positive and negative, respectively, in regressions examining the level of firm involvement in and the amount of control exerted by other users over social media content (H3).