Doctor of Philosophy (PhD)



Document Type



This dissertation studies how the affiliation between different entities in securitization process make different market outcomes, and how this estimation of affiliation effects is susceptible to limitations in securitized residential mortgage data. Three essays constitute the dissertation. The first essay illustrates the limitations and potential bias in the loan-level trustee data. Substantial amounts of loan attributes and risk factors are missing. The patterns of data omissions dramatically vary across different risk factors, sponsors, and trustees, and over the time. Missing of one risk factor is in general positively correlated with missing of the other. Omissions of loan attributes are systematically associated with intermediate level of ex-post default risk. These findings suggest that if any data is sliced and diced based on the availability of loan attributes, the sample for default regression model may not be random. The second essay examines how default risk is associated with the affiliation between the loan provider and the sponsor. The identity of loan provider is, however, selectively disclosed for riskier mortgages. Without consideration of this selective disclosure, the affiliation is seemingly linked to higher ex-ante and ex-post default risk. In contrast, if the affiliation is correctly calculated by backfilling loan provider identity, or if the sample selection problem is explicitly addressed, then loan providers cherry-pick mortgages with better ex-ante risk characteristics for their affiliated securitizations. Also, with more complete sample where missing and erroneous loan provider identities are backfilled and corrected, the affiliation between the loan provider and the sponsor is shown to significantly decrease the likelihood of default. The third essay examines why sponsors are concerned about the performance of mortgages even after they are securitized and sold to investors in the form of bonds. Without any empirical tests, previous studies assume that sponsors have “skin in the game” because they retain the certificates backed by the residual tranches. However, I show that sponsors with their own servicing platform increase their servicing quality even after the most junior tranche has dried up. This result implies reputational concerns may make sponsors care about performance of securitized mortgages.



Document Availability at the Time of Submission

Release the entire work immediately for access worldwide.

Committee Chair

Mason, Joseph