Moodys Analytics Risk Practitioner Conference 2014 Cyclical Loss - PowerPoint PPT Presentation

Moody’s Analytics Risk Practitioner Conference 2014 Cyclical Loss Volatility in Auto Lending Sebastian Ricketts Corporate Economist and Vice President of Economic Analysis

Disclaimer The presentation is intended for informational purposes only. The views expressed in this presentation are strictly those of the author. They do not necessarily represent the position of General Motors Financial Company, Inc. or its affiliates. General Motors Financial Company, Inc. does not endorse or approve, and assumes no responsibility for, the content, accuracy or completeness of the information presented. 2

3 Background

Company Overview  GM Financial is General Motors’ global captive finance company − Achieving growth in earning assets while increasing GM’s sales through expanded and competitive product offerings − Upon the completion of the acquisition of Ally’s international assets (China JV pending), GM Financial will have a global footprint that covers ≈ 80% of GM’s worldwide sales  GM Financial has over 20 years of operating history in North America and decades as GM’s captive in Europe and Latin America − Demonstrated expertise in originating, servicing and accessing capital markets to fund auto finance products − Both the North American and International management teams have led their respective operations through several economic and competitive cycles  GMF has earning assets of $37B, operations in 18 countries and offers auto finance products to approximately 14,000 dealers worldwide 4

Growth in GM Financial Canada Lease - Acquisition of U.S. Prime APR GM Financial U.S. Floorplan Canada Floorplan FinanciaLinx Launch Acquisition Launched Launched 2013 2014 2010 2011 2012 Canada International China U.S. Lease Subprime Acquisitions Acquisition Launched Launched 5 5

1. Cyclical Volatility across Credit Spectrum 2. Model and Methodology Agenda 3. Attributing Volatility 4. Where are we in the Cycle? 6

7 Volatility Cyclical

Data Set  Unique data that was derived from month end snapshots of the full Equifax U.S. consumer credit database − Includes all consumers with >=1 active account − Monthly history from July 2005 to current − Quarterly vintages at zip code level − Product Type (Loan and Lease) − Originator type (Captive. Peer, Bank, Finance Company, etc…) − Term (24/ 36/ 48 month etc..) − Data metrics # and $ (Active trades, Delinquency, C/O, Bankruptcy Filing, Closed Positive)  In this analysis we explore − All vintages and performance periods − 50 state aggregation − Originator type (Large lenders (Captive and Top 20 lenders)) − Product: Loan − Date metric: Charge off Units 8

Portfolio View - Aggregated on Calendar Date Calendar Date Data Source: Equifax 9 9

Vintage dimension Provides attribution of portfolio losses Data Source: Equifax 10 10

Vintage dimension Provides insight into trajectory of losses Will provide insight into future portfolio performance Vintages shifted to common starting point Data Source: Equifax 11 11

Vintage Dimension Loss rates across vintages and credit tier Example: The 2007Q3 vintage of 449 and below Bureau score loans; After 24 months (~2009Q3) from origination the vintage experienced roughly 20% losses from the initial pool. <449 680-719 Data Source: Equifax 12

Vintage dimension Provides basis for advanced analysis and modeling  Most “vintage analysis” refers to creating plots to gain intuition or utilize some simple algorithms to time out future losses − Vintage Plots by Time by Score Band by Term by LTV ……….. − Vintage Plots by Age by Score Band by Term by LTV ………… − Lifecycle Average and Time Average − But  Prone to high error rates when a portfolio is not static  Cannot capture economic cycle  Can try to capture vintage quality through segmentation to some extent  Low scalability and high management cost  Impractical for stress testing  There are modeling methods available allow for a deeper investigation of heteroskedasticity arising from “triangular” data sets. Best practices strive to incorporate explicitly − Economic dimension − Credit quality − Dynamic variation of originations for “what if” scenarios  These methods scale well and are low cost 13

14 Methodology Model and

Model of forces impacting losses - Product Specific - Age of Loan (a) LifeCycle (a) - Defaults on a 60 month car loan have a typical loss curve associated with product and segment - Origination Date Specific Vintage (v) - Cohort analysis (v) - Loans originated in the same period are subject to the reigning originations posture - Calendar Time (t) Economy (t) - 2007 recession impacted all vintages at different points in their lifecycle 15

Product Lifecycle The risk of going ‘bad’ depends on the age of the loan.  The lifecycle for a loan is the characteristic shape that describes  the timing of the events. Product attributes drive the lifecycle differences  16

Findings Across the Credit Spectrum Higher Cycle Volatility Non-Prime Prime Lower Cycle Volatility Prime volatility is significantly higher than Sub-Prime 17

Vintage Credit Quality Vintages have unique characteristics that affect default at origination  Show up as level shifts  Economic and Industry conditions at time of booking  Loan characteristics (score,ltv, dti, etc)  Changes in underwriting standards  18 Vintage Seasonality 

Vintage Quality Index Relative to Long Run Average (2007-2014) Non-Prime Prime • Non-prime retrenched significantly in • Prime did not retrench during the the crisis crisis • Best ever quality in and post the • Vintage Quality was negative as crisis as competition for loans was lenders flocked to prime in the crisis low to non-existent • As lenders have begun moving • Growth in subprime competition has downstream easing competition and seen normalization in credit quality good borrowers take advantage of low rates and an improving economy credit quality looks good 19

Originations across Credit Spectrum Subprime has yet to reach 2005 origination levels 20

Decomposition Economy The consumer environment impacts all loans, regardless of age or  vintage. The environment is composed of several factors.  Seasonality  Portfolio management  Macroeconomic environment  21

Economy Index Relative to Long Run Average (2007-2014) Non-Prime Prime • Non-prime: Contribution to losses • Prime Contribution to losses from the from the economy peaked in 2009 economy peaked in 2009 • Less bad impact through 2011 • Less bad impact through 2010 • Tailwind from economy in 2012 • Tailwind from economy in 2011 and 2013 and 2013 • Tailwind in 2014 appears to be • Tailwind in 2014 appears to be moving to neutral (consistent with LR moving to neutral (consistent with LR average impact) in 2015 average impact) in 2015 22

23 Attributing Volatility

Impact to Portfolio Deep subprime : 70-80% of volatility attributed to originations posture (economy) at time of booking Prime: Volatility in losses attributed roughly equally originations posture at Near-prime: time of booking and Volatility in losses the economic cycle attributed more to the economic cycle 24

Implications and Trends  Overall: − Models or analysis need to take into account different dynamics across the spectrum − Awareness of how your portfolio is affected can help navigate through a cycle  Economy: Today, across credit spectrum the economy continues to provide a tailwind to credit performance  Sub-Prime: high loss scenario would be more likely from a high growth scenario where originations policy becomes lax Effect on from economy is modest relative to originations posture (economy at time of  origination) Today, subprime vintage quality is neutral to a headwind to credit performance   Near-Prime: high loss scenario would be more likely from an economic downturn  Effect from economy is strong  Today, Near prime vintage quality has normalized to a neutral position  Prime: high loss scenario would be more likely from a combination of high growth where originations become more lax followed by an economic downturn  Effect on economy and originations posture appears to be 50/50ish  Today, Prime vintage quality remains a tailwind 25

Contact Info Sebastian Ricketts sebastian.ricketts@gmfinancial.com 817-302-7156 www.linkedin.com/in/sebastianpalaoricketts/ 26