Securitization and Lending Competition

Abstract

Both securitization and remote lending have become much more extensive since the 1970s. We present a tractable theoretical model that links these two trends. DeMarzo and Duffie (Econometrica 1999) study the problem of a bank that wishes to securitize an exogenous portfolio of loans about which the bank has private information. We embed their model in a setting with lending competition. Rather than being exogenous, each bank’s loan portfolio is now the endogenous result of competition for heterogeneous loan applicants. There are two regions, each of which contains a single bank and a continuum of loan applicants. An applicant’s credit quality is unknown. For each applicant there is a private quality signal (“soft information”) that is seen only by her local bank as well as a public quality signal (a “credit score”) that is seen by both banks. Without securitization, banks lend only to local applicants because of a winner's curse. With securitization, in contrast, ignorance is bliss: the less a bank knows about its loans, the less of a lemons problem it faces in securitizing them. This enables banks to compete successfully for some remote applicants. Our model has several implications that have been confirmed by empirical studies. By encouraging banks to lend to remote applicants about whom they lack soft information, securitization raises the mean default risk. By leading banks to rely exclusively on public information for these applicants, securitization can also cause a deceptive rise in credit scores and other borrower observables. In our model, securitization has mixed effects on social welfare. It raises the supply of funding for worthwhile projects by connecting liquid investors with deserving loan applicants. However, it also leads to an inefficient loan allocation by giving banks an incentive to favor remote applicants with strong observables.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 21 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.