The Federal Housing Finance Agency (FHFA) maintains a proprietary Mortgage Analytics Platform to support the Agency’s strategic plan. The objective of this white paper is to provide interested stakeholders with a detailed description of the platform, as it is one of the tools the FHFA uses in policy analysis. The distribution of this white paper is part of a larger effort to increase transparency on mortgage performance and the analytical tools used for policy analysis and evaluation within the FHFA.
The motivation to build the FHFA Mortgage Analytics Platform derived from the Agency’s need for an independent empirical view on multiple policy initiatives. Academic empirical studies may suffer from a lack of high quality data, while empirical work from inside the industry typically represents a specific view. The FHFA maintains several vendor platforms from which an independent view is possible, yet these platforms tend to be inflexible and opaque. The unique role of the FHFA as regulator and conservator necessitated platform flexibility and transparency to carry out its responsibilities.
The FHFA Mortgage Analytics Platform is maintained on a continuous basis; as such, the material herein represents the platform as of the publication date of this document. As resources permit, this document will be updated to reflect enhancements to the platform.
Partners Group provides capital for Taxfix, Litera
Partners Group Private Equity gained in May. The net asset value for Class I rose 3.5%, while the net asset value for Class A grew 3.4%. The total fund size increased to $5.6 billion. For the first five months of the year, Class A is down 4.4%, while Class I is down 4.2%. Q1 2020 Read More
FHFA Mortgage Analytics Platform Overview
The platform integrates econometric loan performance models, loan level data and external economic forecasts to project mortgage cash flows. This section offers an overview of the modules and their interconnections.
There are two sources of external inputs to the analytics platform: loan level data and economic forecasts. The economic forecasts include projections of house prices, interest rates and unemployment rates through the forecast horizon. Both vendor-supplied economic forecasts and FHFA projections of economic variables are stored in the economic forecast database. These economic forecasts cover a wide range of economic environments from baseline to highly optimistic to extremely stressful economic conditions. The economic forecast databases are quarterly.
The loan level data elements are the second source of external inputs; these include approximately thirty variables per loan comprising loan attributes and borrower characteristics. The platform projects mortgage performance from the loan’s current age to termination, including foreclosure alternatives and the resolution of real estate owned (REO). The platform applies projected probabilities of termination to performing loan balances such that a portion of the loan prepays, becomes delinquent and may resolve as a default each month. To simplify the discussion within this paper, when a loan is said to prepay (or default), only a portion of the loan is prepaying (or defaulting), not the whole loan. The components of the platform are summarized below and are described in greater detail in subsequent sections of this paper.
1. Performing Loan Module – the primary function of this module is to project monthly loan level prepayment and 90-day delinquency probabilities on performing and modified performing loans. Loans enter into this module if they are current, less than 90 days delinquent, or forecasted to cure from a delinquency during the simulation. The prepayment and delinquency equations are functions of borrower characteristics, loan characteristics, home values and other economic variables. Multiple pairs of prepayment and delinquency equations collectively cover several loan products and modified loans guaranteed or owned by the Enterprises.
2. Non-Performing Loan Module – the primary function of this module is to project lifetime outcomes for delinquent loans. Loans enter into this module if they are 90 to 180 days delinquent at the beginning of the projection, or if they are predicted to become delinquent within the performing loan module. The module outputs four mutually exclusive loan-specific probabilities each month: foreclosure completion (REO), voluntary prepayment, foreclosure alternative resolutions and re-performance (cure). The foreclosure alternative resolutions include deed-in-lieu of foreclosure, pre-foreclosure sale (short sale), and third party sale. A loan is defined as re-performing when all arrearages are paid and the cure is not due to a modification or restructuring. The models are a function of borrower characteristics, house prices and state legal structures. Unlike the performing loan module where multiple product level models are constructed, only one set of equations is estimated for non-performing loans.