The arti­cle deve­lops the hypothe­sis that con­ven­ti­ons can be key to the deve­lop­ment of mar­kets for pre­vi­ously untra­ded finan­ci­al assets. It argues that the absen­ce of his­to­ri­cal data on the per­for­man­ce of assets may indu­ce agents to form beli­efs in ways that pre­clu­de the for­ma­ti­on of mar­kets for cer­tain finan­ci­al ins­tru­ments. Con­ver­sely, in the absen­ce of tra­ding, the data on the basis of whi­ch agents could revi­se their beli­efs will not be pro­du­ced, giving rise to a feed­back loop that hin­ders the pro­cess of mar­ket-for­ma­ti­on. The arti­cle con­tends that one of the ways of over­co­ming this sta­te is the deve­lop­ment of con­ven­ti­ons about the dis­tri­bu­ti­on of futu­re payoffs. It then shows that the emer­gen­ce of the con­ven­ti­ons that cre­dit rating agen­ci­es could reli­a­bly mea­su­re the ris­ki­ness of non-pri­me resi­den­ti­al mort­ga­ge-secu­ri­ti­es (RMBSs), and that the FICO sco­re could reli­a­bly give a quan­ti­ta­ti­ve expres­si­on of the cre­ditworthi­ness of indi­vi­du­al bor­rowers was a cru­ci­al ena­bling con­di­ti­on of the for­ma­ti­on of a mar­ket for RMBSs in the Uni­ted States.

Keywords: con­ven­ti­o­nal valu­a­ti­on, pro­ba­bi­lity, mort­ga­ge-bac­ked secu­ri­ti­es, cre­dit rating agen­ci­es, cre­dit scores.

JEL codes: B29, G11, G12.