Freddie Mac · OTC:FMCC
Freddie Mac — Investor Day (April 2023)
27 slides from Freddie Mac's investor day presentation, Apr 20, 2023
Slide 1: Modeling's Role in Risk Mitigation
Modeling’s Role in Risk Mitigation Public Prepared by Freddie Mac © Freddie Mac
Slide 2: Today's Speaker
Today’s Speaker Michael Bradley Senior Vice President, Single-Family Modeling and Analytics Michael Bradley is responsible for setting the modeling and analytics strategy and positioning Freddie Mac to become the best credit guarantor. This is a critical role, since Freddie Mac relies on data-driven
Slide 3: Our Agenda
Our Agenda 1. Background 2. Overvalued Markets: How to Identify and for How Long 3. Freddie Mac’s AI/ML Journey 4. Closing Thoughts Public Document Title & Date Goes Here Prepared by Freddie Mac © Freddie Mac
Slide 4: Contextual Background
Contextual Background • The Three C’s • Four Lines of Defense • Formal Theories of Default Public Investor Day – April 20, 2023 Prepared by Freddie Mac © Freddie Mac
Slide 5: The Three C's of Mortgage Performance
Contextual Background The Three C’s of Mortgage Performance Credit Capacity Collateral Public Investor Day – April 20, 2023 Prepared by Freddie Mac © Freddie Mac
Slide 6: Four Lines of Defense
Contextual Background Four Lines of Defense Borrower Collateral Mortgage Insurance Capital Public Investor Day – April 20, 2023 Prepared by Freddie Mac © Freddie Mac
Slide 7: Trigger Theory
Contextual Background Trigger Theory • A mortgage can be viewed as a collection of financial options. • Think of a default as similar to a put option. Borrowers Borrowers • Default requires the combination of negative equity and a ‘life with negative who suffered event’, (e.g., job loss, divorce, eq
Slide 8: A variety of life events help drive delinquencies
Contextual Background A variety of life events help drive delinquencies Public Investor Day – April 20, 2023 Prepared by Freddie Mac © Freddie Mac
Slide 9: Adding home price drop drives credit losses
Contextual Background Adding home price drop drives credit losses Public Investor Day – April 20, 2023 Prepared by Freddie Mac © Freddie Mac
Slide 10: Overvalued Markets: How to Identify & for How Long
Overvalued Markets: How to Identify & for How Long Public Prepared by Freddie Mac © Freddie Mac
Slide 11: Standard Industry House Price Forecasting Model
Standard Industry House Price Forecasting Model • Standard House Price Forecasting Model in the industry is a two-stage error-correction model that: Predict the local long-run sustainable house- price level that can be supported by long-run o Uses local house-price drivers to supply and demand facto
Slide 12: One Example for Some MSA
One Example for Some MSA Public Investor Day – April 20, 2023 Prepared by Freddie Mac © Freddie Mac
Slide 13: Challenges of identifying an overvalued market
Challenges of identifying an overvalued market No definitive, universally accepted measure of market overvaluation; no observed equilibrium level of a market Different studies consider different factors: Even if one can determine the actual drivers of the • High price-to-income ratio equilibrium lev
Slide 14: Significant Variation - Different Model Views on Select MSAs
Significant Variation - Different Model Views on Select MSAs from 2022 Q2 Metro Model A Model B Model C Model D Model E Model F Atlanta Austin Baltimore Birmingham Boston Buffalo Chicago Dallas Detroit 50+% Houston Los Angeles 25%~50% Miami 10%~25% New York 0%~10% Phoenix -10%~0% San Francisco -25~1
Slide 15: Monte Carlo Simulation
Monte Carlo Simulation US Paths - Production 4 3.5 3 2.5 2 1.5 1 0.5 0 2020 2025 2030 2035 2040 2045 Public Investor Day – April 20, 2023 Prepared by Freddie Mac © Freddie Mac
Slide 16: Determining When an Overvalued Market will Decline is a Challenge
Determining When an Overvalued Market will Decline is a Challenge Not all overvalued markets will decline. An “overvalued” market can stay overvalued for extended periods of time. For example, • Miami had been called 15% overvalued in 2014 Q1 by one of the industry models. • However, Miami grew 132+
Slide 17: Market Overvaluation: Summary
Market Overvaluation: Summary • Determining whether a housing market is overvalued based on statistical models can be a challenging task. • It is equally difficult to determine how long a market can remain overvalued. • Better to monitor key indicators and economic trends to get a sense of where a p
Slide 18: Freddie Mac's AI/ML Journey: The Rationale & Examples
Freddie Mac’s AI/ML Journey: The Rationale & Examples Public Prepared by Freddie Mac © Freddie Mac
Slide 19: Benefits of Machine Learning
Benefits of Machine Learning • Our major models allow us to scale massively • Others are using AI in an attempt to disrupt our business model • Humans predictably make mistakes — “Thinking Fast and Slow” Public Investor Day – April 20, 2023 Prepared by Freddie Mac © Freddie Mac
Slide 20: Why We Need AI
Why We Need AI • Captures nonlinearities and interactions better than traditional methods, which maps to improved business outcomes. • Has potential to greatly improve outcomes (including expanding access to credit) for consumers, businesses and other stakeholders across wide range of applications a
Slide 21: LPA and Fair Lending Testing
LPA and Fair Lending Testing Fair Lending testing of LPA uses innovative techniques to identify Less Discriminatory Alternatives Less Discriminatory Alternatives (LDA): • Alternative model specifications that lower impact on protected classes without a material degradation in performance. If found,
Slide 22: LDA Search Results from GAN – An Example
LDA Search Results from GAN – An Example Hypothetical Example – Actual Results Not Shown • The graph shows impact/accuracy tradeoff for alternatives created from the GAN process. • Currently in use for LPA fair lending testing. By greatly expanding the search relative to previous methods, the GAN al
Slide 23: Asset Income Modeler (AIM) Helps Manage Risks
Asset Income Modeler (AIM) Helps Manage Risks Using new auto-labeling techniques to build future versions of our models increases accuracy and speed. Continually Managing Risk: • Removes subjectivity & manual errors by applying consistent rules to all loans. • Receives data directly from the source
Slide 24: Condition & Quality Analysis
Condition & Quality Analysis Applying AI/MI algorithms to develop solution for property condition and quality assessment Understanding the type of characteristics to determine the overall property quality and condition are needed to understand market expectations and how that influences value. A Tal
Slide 25: Conclusion
Conclusion • Freddie Mac is harnessing the power of AI and ML to drive faster, more accurate decisioning, manage risk effectively and create value, liquidity and stability within the housing ecosystem. • Freddie Mac’s investment in AI/ML and research of new applications is enhancing our understandin
Slide 26: Questions?
Questions? Public Prepared by Freddie Mac © Freddie Mac
Slide 27: Thank You
Thank You Public Prepared by Freddie Mac © Freddie Mac