The intercept in model (3) is suppressed so that the coefficient loadings on the event dummy variables (PRE_KPMG, PRE_CENSURE, POST_KPMG, and POST_CENSURE)reflect the conditional mean log-odds rate of client loss for KPMG in the pre and post periods (PRE_KPMG and POST_KPMG, respectively), and for the censured audit firm in the pre and post periods (PRE_CENSURE and POST_CENSURE, respectively).