1. July 26, 2013 Liability Management : Autopilot Refinancing Jonathan Zinman Dartmouth College and IPA’s U.S. Household Finance Initiative Financial Products Innovation Fund Working Group Meeting

2. What’s This Talk About? Huge opportunity to add value by helping people manage their borrowing decisions Asset Management model for Liability Management Take a segment of Liability Management: “Personal Loan Shopper” Focus on slice of that segment: “Autopilot Refis”

3. Debt is Where the Money Is Huge market for asset management – Brokers, financial advisers/planners, private bankers, etc. Yet debt is where the money is for most Americans $13 trillion on household balance sheets – Many more people participate in debt market(s) than in, e.g., stock market Borrowing costs > asset yields – Higher stakes

4. Household Mismanage the Liability Side of Their Balance Sheets (Or at least evidence suggests that they do) Potential sources of “decision inefficiency” – Overspending – Misallocation  Credit cards  Student loans – Failure to take-up workouts (e.g., HAMP) – *Undershopping  Upfront contract choice (e.g., overpaying)  When refinancing, or not Money on table => Money to be made

5. Is There Really Money to Be Made? Focus just on overpaying. Take a household that: – Has $10k in credit card debt and pays +500bp – Has $20k in student loans and pays +200bp – Has $20k in auto loans and pays +100bp – Has $200k in mortgage and pays +100bp This household is paying >$3,000 in excess interest per year – Annuity value of $15,000-$100,000, depending on discount rate – Could get just a piece of this-- e.g., just auto loans– and be viable

6. A Segment of Liability Management: Personal Loan Shopping There is a growing market for this already. Examples: Problems with existing products: – Limited personalization (key for risk-based pricing) – Limited coverage: a single site/app/service rarely credibly covers all (low-cost) providers in the market – Biased coverage: nearly all business models are kickback-based – Limited/no coverage of refinancing Where there are gaps, there are opportunities!

7. A Slice of Personal Loan Shopping: Autopilot Refinancing Problem: People don’t refinance (auto, mortgage, student loans) when they should Sen. Gillibrand seeks automatic refinancing of student loans at 4% rate (Examiner) Sen. Gillibrand seeks automatic refinancing of student loans at 4% rate (Examiner) Agarwal et al on mortgages Why? Inattention post-origination Procrastination (fed up by complexity of option value)

8. Solution: Find a better decision point -> at origination Put refis on autopilot Allow borrower to set a refinancing rule at origination – Like a limit order in asset management Delegate mechanics to liability manager – Manager shops and negotiates on borrower’s behalf – Manager handles paperwork Maybe even signing off under power-of-attorney A Slice of Personal Loan Shopping: Autopilot Refinancing (Cont’d)

9. Making Autopilot Refis Work: Some Design Issues (These issues apply to loan shopping more broadly…) Getting personalized data on consumers – How much? Getting accurate data on offers Accurately mapping consumer characteristics into offers Moving money Fiduciary standards

10. Making Autopilot Refis Work: What’s the Business Model? My premise: get borrowers to pay for it – Kickback model dominates, so differentiate – Wouldn’t transparent pricing be better?  Branding  Regulatory constraints/threat  Signaling value proposition So how do we get there? – Issue for value-added services more generally

11. Making Autopilot Refis Work: Will borrowers pay? How? “People won’t pay out-of-pocket for value-added financial services… they expect freebies” – But people do pay out of pocket for credit report management, identity theft protection, tax prep, etc. Pricing Models that have worked in other contexts Upfront lump-sum  Natural for cash-out refinancing?  Otherwise paid by credit card? Monthly subscription  Teaser? Periodic “debt under management” fees

12. Pricing Autopilot Refis (and beyond): Unlocking borrower willingness-to-pay Behavioral research offers some insights… Willingness-to-pay depends a lot on framing, timing E.g., on marketing and choice architecture – Content – Timing of offer – Ease of take-up (on-ramping) Effectiveness can vary across contexts So R&D needed to successfully apply these insights to autopilot refis – To liability management more broadly – To financial product development even more broadly

13. Summing Up Huge missing market for liability management – Personal loan shopping is a big piece With autopilot refis a big piece of loan shopping Many technical challenges Some design/behavioral challenges, with the most important being business model – Unlocking willingness-to-pay My bet is that these challenges are solvable – Some will require R&D of the traditional variety – Others will require behavioral R&D, including the sort of piloting and RCTs (AB-testing) that is supported by IPA’s Ford-funded work

15. Extra Slides Follow

16. Unlocking Consumer Willingness-to- Pay: Marketing Content Informative appeals  How to frame cost savings?  How to signal quality/credibility? Emotional appeals  How to capture attention?  How to counter procrastination? Behavioral evidence suggests that: – Both informative and emotional can be effective – Context and customer-particulars matter => Optimizing content requires experimentation

17. Where do Consumers Go Wrong? (Product Design and Marketing) What leads to undershopping, overspending? Behavioral factors: self-control costs, limited attention, price misperceptions Confusion (financial illiteracy/innumeracy) Disutility from personal financial management (distaste, time costs) How to solve for these pain points? R&D