83 Our model specifications are as follows: The data vary by year (t), deal (i) and security (j). We control for security-specific characteristics, issuer-fixed effects and time-fixed effects. We denote pre- and post- crisis years through the dummy variable Post, which we interact with our CLOs deal complexity explanatory variables (Tranche Count, Log Tranche Size, Capital Allocation). We do so as credit rating risk is found to be countercyclical: CRAs are more likely to issue less-accurate ratings during boom periods (see e.g., Bar-Isaac & Shapiro, 2013; Bolton et al., 2012; Dilly & Mählmann, 2016; He et al., 2012). 3.3.3 Variable Construction and Summary Statistics 3.3.3.1 Dependent Variables Tt raabnl ceh3e .s1 ,wPiat hn eol nAe rdeips oc lrot ss esdu mr amt i na gr y f rsot amt i setiitchse fro rMtohoed tyo’ st aol rs aSm& Pp,l ea. nWd et riannc cl uhde es with ratings from both CRAs. The dependent variable of model (1), Number of Chapter 3 - Security Design and Credit Rating Risk
RkJQdWJsaXNoZXIy MTk4NDMw