“…To address the potential endogeneity problem in the estimation model, we took several measures. First, in order to alleviate the potential reverse causality, we lagged our independent variable, moderator and all control variables for one period [67,68], i.e., the data on independent variable (i.e., brand equity), moderator (i.e., industrial competition) and controls variables were from 2004 to 2015, while data on mediators (i.e., analysts' recommendations) and dependent variable (i.e., firm sustainable performance) were from 2005 to 2016. Second, we employed a hackman two-stage model [69] to account for the sample selection bias.…”