150 4.4 Methodology Using separate subsamples, we conduct three different tests to test our three sets of hypotheses. Fc ri resdti, tt or attei sntghs yapsos ti hg ne se ids 1t o, wo en ec ot nr as nt rcuhcet. aHseurbe s wa me palree wi nhteerree ws t ee dc oi nm pt raarnecmh eusl tti hp al et rdei sc ceai vr edd aal l rtartai nn gc hoefs atthlaeta srte coeni ve elda r og en ea nr da t ionnge (s1m, 7a8l l3 )C RoAr . tFhorre et hri sa t ri ne ag ss o(n1, 9w6e) ft rr oa nmc ht he se rraetgerde sbsyi oFni tsc. hU an nf odr tausnma tael ll ye, r t ph ee er er (aDr eB Ra Sn oi nr sKuBf fRi cAi e) nt to neunma bbl ee rs ot af t Ri sMt i cBaSl ao nu ar l syas me spol ne tf ur ar nt hc ehre st or a2t ,e0d3 b5 yoFbist ec hr vaant ido an ss ,moaf l wl ChRi cAh ( 71 29 17 tt rr aa nn cc hh ee ss ) w. Tehr ies rrae tdeudc eb dy MS &oPo da yn’ ds aKnBdRDAB. TR hS ,e 528, 20 3b5y dMuoaol -drya’ tse adntdr aKnBc Rh Ae s, 5p0r 2e sbeyn St &t hPe afni rds tD sBuRbSs, aamn dp l2e 3u0s eb dy in our first model. To test hypothesis 1, we use an ordered logit model to estimate how the market share of large CRAs impact the rating quality of large and small CRAs. Our model specification to test our first hypothesis is: Rating Differencesjt = α 0 + α1 Market Share CRAjt + Tranche, Issuer and Market Controlsijt + ϵijt The data vary by year (t), deal (i), and security (j). We control for security-specific characteristics, credit rating and time-fixed effects. Second, to test hypothesis 2, we use our full sample of 4,190 tranches. Hence, we include all tranches rated by Moody’s, S&P, DBRS, and KBRA regardless of ws t ha ne tdhaer rd st hveayr yr ewc iet ihv ehdi g oh ne re oorr l omwuel tri pml ea rckreetdsi th ar ar et i no fg si t. sTcoo mt e ps te thi toowr s C(Rh Ay ps ’o rt ha tei sni gs
RkJQdWJsaXNoZXIy MTk4NDMw