UC Berkeley

This dissertation consists of three chapters on mortgage funding and risk management.

The U.S mortgage market is very concentrated. In 2006, the top 40 lenders were responsible

for the origination of 96 percent of all mortgages (Stanton et al. (2014)). These large lenders

originated about 60 percent of the mortgages through the wholesale channel, delegating parts

of the origination process to third party agents such as mortgage brokers and correspondent

lenders. I show that the type of agent selected by the wholesale lender could crowd out

local banks, who often act as correspondents and rely on these wholesale lenders for funding.

I also show that this crowding out has spillover eects. As local banks decreased their

presence in the county, they have also reduced other types of lending. As a result, their local

communities showed less growth in small businesses.

The second chapter discusses the perils of warehouse lending. Warehouse lending is an

important part of the U.S. mortgage market because a large fraction of mortgage origination,

both pre-crisis and currently, is carried out by non-depositories who are reliant on warehouse

facilities to fund their mortgage origination activity. After the passage of The Bankruptcy

Abuse Prevention Act (BAPCPA), in April 2005, most warehouse facilities were structured

as Master Repurchase Agreements (MRAs). BAPCPA re-dened the mortgage loans held

as collateral on the warehouse lines (the newly originated mortgages) as repo thus exempting

them from automatic stay upon the bankruptcy of the mortgage originator (the repo

seller). We consider the eect of the growth of MRAs for funding mortgage originations on

the performance of the mortgage originators (repo sellers) and warehouse lenders (repo buyers).

We nd that mortgage originators (repo sellers), that used MRAs to fund their loans,

originated mortgages of lower quality and that these originators were more likely to declare

bankruptcy. Symmetrically, we nd that the warehouse lenders (repo buyers) experienced a

sharper increase in mortgage charge-os and non-performing mortgages than non-warehouse

lenders, even though the quality of the retail and wholesale mortgages that they originated

were comparable to the quality of mortgages originated by non warehouse lenders. This

negative outcome for warehouse lenders arose from the exemption of the mortgage repo

collateral from automatic stay, since under BAPCPA the poor quality assets of bankrupt counter parties, the mortgage originators, became consolidated on the warehouse lender balance

sheets. Thus, the consolidated loans from the bankrupt counter parties generated an

important component of the deterioration in the warehouse lenders' mortgage

In the third chapter, we propose an empirical duration measure for the stock of U.S.

Agency MBS that appears to be less prone to model risk than measures such as the Barclays

Eective Duration measure. We nd that this measure does not appear to have a strong

eect on the 12-month excess returns of ten-year Treasuries as would be expected if shocks

to MBS duration lead to commensurate shocks to the quantity of interest rate risk borne

by professional bond investors (see, Hanson, 2014; Malkhozov et al., 2016). Given this

negative reduced form result, we then explore the mortgage and treasury hedging activities

of the primary MBS investors such as commercial banks, insurance companies, the agencies,

the Federal Reserve Bank, Mutual Funds, and foreign investors. We nd that the only

investors that may follow the models of Hanson (2014) and Malkhozov et al. (2016) are

foreign investors in Switzerland and the United Kingdom and life insurance rms. Life

insurance rm market share has declined over the period, dropping below 10% since 1996

and reaching 4% in 2016. Furthermore, Switzerland and the United Kingdom are not major

participants in the US Treasury market. Of the investors we are not able to study, hedge

funds and pensions/retirement funds are the two investor groups that may trade along the

Hanson (2014) and Malkhozov et al. (2016) models. However, although these two investor

groups held almost 25% of the Agency MBS market (including households and non prot

organizations) in the late 1990s, post crisis their share has fallen below 10%