Five Bridges is the result of years of development of data, models, skills and expertise, now harnessed to add clarity to today’s very challenging mortgage and housing markets.
We began developing our central thesis over a decade ago, while working on commercial mortgage loans and securities. We believed that local economic conditions influence the performance of residential mortgage loans as strongly as they do commercial mortgage loans. We were struck by the conspicuous absence of market studies of residential mortgage loans, which are essential to evaluating investments in or financing of commercial properties.
So the question was asked, “Does location matter in residential real estate performance? Do changes in local economies lead to different credit and prepayment performance of otherwise similar borrowers and loan products in different metropolitan areas?”
Prior to the beginning of this decade, the answers to these questions affected only a handful of investors. The government-sponsored enterprises (GSEs) were the primary insurers or holders of residential mortgage credit risk that banks and thrift institutions chose not to retain on their balance sheets. While the non-agency mortgage market existed, it was miniscule in comparison to the much larger agency mortgage market.
A major factor in the growth of the non-agency mortgage market was the revision of the financial statements of Fannie Mae and Freddie Mac and the resulting loss of focus on their central mission of insuring conventional, first-lien mortgage loans. Rapid home price appreciation, historically tight yield spreads on RMBS assets and extremely deep and fluid pools of capital for mortgage lending, securitization and balance sheet leveraging all helped to fuel the rapid growth of the non-agency (RMBS and ABS) securitization. Investors, for the first time, held a significant portion of residential mortgage credit risk via non-agency mortgage securities. In light of the benign credit performance of the early 2000s, participants did not see the value of developing more transparent and robust credit modeling capabilities.
In order to answer our question about location, we first needed high quality, geographically comprehensive data about loan performance. Unfortunately, clean and comprehensive data on residential mortgage loan performance data did not exist. Leveraging a long-standing relationship with mortgage data provider LoanPerformance, we sponsored the development of the largest and most comprehensive publicly available databases of mortgages in existence.
With data in place, we constructed extensive time series on residential mortgage credit and prepayment performance that became the backbone of our modeling efforts. We developed econometric models of residential loan performance, by product type, for each of the 363 metropolitan statistical areas (“MSAs”) in mid-2004. When evaluating mortgage performance, zip code level aggregation is not broad enough to capture the distance between where most borrowers live and work, while state level aggregation is too broad and obscures important differences in local economies.
The results absolutely supported our core belief and encouraged us to explore the systematic deficiencies in the interwoven process of originating, securitizing, rating and trading of mortgage related assets. Based on further reviews of servicing platforms, origination models, securitization methods, the rating agency process and investor demand, we quickly discovered widespread systematic and idiosyncratic risk. Our results compelled us to become even more granular in our work’s development.
In 2005, we began to develop forwarding looking loan level models that were based on the traditional metrics of borrower risk (FICO, LTV, DTI, etc.) but were augmented by our models of local economic conditions and their causal relation to performance.
In 2006, we began a process to harness the massive amounts of data we manage and make it available through a highly scalable and efficient front end platform. Over time, we have developed numerous improvements to that platform.
Every month, our models are updated and refreshed to incorporate the latest mortgage data and economic information available on loans and securities. Importantly, we have nearly three years of out-of-sample results that help us make critical updates and improvements quickly.
At Five Bridges, our team averages two decades of mortgage market experience covering every discipline from origination, servicing, workouts, loss mitigation, structuring, research, trading, portfolio optimization and risk management.
Our platform is not a response to the current crisis, but one that adds transparency, depth, and independent thought to the critical evaluation of mortgage and housing related assets.
