Auto Dealer Loan Intermediation: Consumer Behavior and Competitive - - PowerPoint PPT Presentation

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects

Andreas Grunewald (Bonn), Jonathan Lanning (Chicago Fed), David Low (CFPB), Tobias Salz (MIT) NYU Alumni Conference, September 7th

*The views expressed are those of the authors and do not necessarily reflect those of the Consumer Financial Protection Bureau, the Federal Reserve Bank of Chicago, the Federal Reserve System, or the United States.

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects

Most auto loans are intermediated by auto dealers.

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects

Most auto loans are intermediated by auto dealers. Research Question: How does loan intermediation affect consumers?

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects

Most auto loans are intermediated by auto dealers. Research Question: How does loan intermediation affect consumers? Dealers set price of vehicle and loan, potentially leading to a form of price discrimination.

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Motivation

Motivation

Bundling Loans and other Financial Products

Auto loan market is large: ◮ About $1 trillion, third-largest debt market in US Cars are typically bundled with loan: ◮ Around 85% of car loans in the US are intermediated by dealers. Bundling is important for dealers: ◮ 2011: > 50% of dealer profit from F&I department. Bundling w/ financial contracts common in other retail markets: ◮ Consumer durables with financing and warranties. ◮ Flights/hotels with travel insurance. ◮ New construction mortgages.

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Motivation

Project Overview

• 1. Describe market and dealers’ incentives.

− Vertical relationships between lenders and dealers.

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Motivation

Project Overview

• 1. Describe market and dealers’ incentives.

− Vertical relationships between lenders and dealers.

• 2. Use dealers’ incentives to study consumers’ price response

− Imposing only supply-side optimal behavior. − Derive individual-specific bounds on responsiveness.

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Motivation

Project Overview

• 1. Describe market and dealers’ incentives.

− Vertical relationships between lenders and dealers.

• 2. Use dealers’ incentives to study consumers’ price response

− Imposing only supply-side optimal behavior. − Derive individual-specific bounds on responsiveness.

• 3. Interpretation of our results

− Not driven by: taxes, default, prepayment, credit constraints/impatience

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Motivation

Project Overview

• 1. Describe market and dealers’ incentives.

− Vertical relationships between lenders and dealers.

• 2. Use dealers’ incentives to study consumers’ price response

− Imposing only supply-side optimal behavior. − Derive individual-specific bounds on responsiveness.

• 3. Interpretation of our results

− Not driven by: taxes, default, prepayment, credit constraints/impatience

• 4. Heterogeneity analysis

− Who exhibits smaller/larger wedges in responsiveness?

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Motivation

Project Overview

• 1. Describe market and dealers’ incentives.

− Vertical relationships between lenders and dealers.

• 2. Use dealers’ incentives to study consumers’ price response

− Imposing only supply-side optimal behavior. − Derive individual-specific bounds on responsiveness.

• 3. Interpretation of our results

− Not driven by: taxes, default, prepayment, credit constraints/impatience

• 4. Heterogeneity analysis

− Who exhibits smaller/larger wedges in responsiveness?

• 5. Counterfactual exercises

− Imposing demand + equilibrium model.

The Market

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects The Market

The Setting

The Typical Financing Process

• 1. Consumer chooses make and model of the car.
• 2. Consumer provides personal data; dealer checks credit.
• 3. Dealer collects “buy rates” from lenders through e.g. Dealer Track,

Route One, or Credit Union Direct Lending.

• 4. Dealer makes loan offer, including markup over buy rate.
• 5. Dealer receives payment (“dealer reserve”) from lender.

◮ Payment = (fixed payment) + (share of markup revenue) ◮ Average fixed payment is $137; average share is .66

78% of loans marked up. Average markup is 108 basis points.

Using Dealers’ Problem to Quantify Consumers’ Price Responsiveness

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Quantifying Consumers’ Price Responsiveness

Quantification

Intuition

Why do dealers mark up loans?

◮ Charging one extra dollar on the car yields a dollar. ◮ Charging one extra dollar on the loan yields 66 cents.

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Quantifying Consumers’ Price Responsiveness

Quantification

Intuition

Why do dealers mark up loans?

◮ Charging one extra dollar on the car yields a dollar. ◮ Charging one extra dollar on the loan yields 66 cents. ◮ Explanation: Some consumers respond less to finance charges.

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Quantifying Consumers’ Price Responsiveness

One-Period Model

Dealer’s Optimal Markup Choice

Consumer i: ◮ Down payment di, car price pi and interest rate ri. ◮ Disutility of pi is pi; disutility of finance charges x is Mi(x) ∈ C2. ◮ Requires utility ¯ ui to buy car. ◮ Can finance the car through the dealer or an outside lender.

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Quantifying Consumers’ Price Responsiveness

One-Period Model

Dealer’s Optimal Markup Choice

Consumer i: ◮ Down payment di, car price pi and interest rate ri. ◮ Disutility of pi is pi; disutility of finance charges x is Mi(x) ∈ C2. ◮ Requires utility ¯ ui to buy car. ◮ Can finance the car through the dealer or an outside lender. Dealers: ◮ Exogenous buy rate bi and costs for a car ci. ◮ Set pi and ri. ◮ Dealer reserve has slope α and intercept β.

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Quantifying Consumers’ Price Responsiveness

One-Period Model

Dealer’s Optimal Markup Choice

Constrained dealer’s maximization problem:

max

ri,pi

(pi − ci) + (pi − di) · (ri − bi) · α + β s.t. − pi − Mi((pi − di) · ri) ≥ ¯ ui − Mi((pi − di) · ri) ≥ −

• Mi((pi − di) · rL) · gi(rL) · drL − si,

ri ≥ bi, pi ≥ 0

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Quantifying Consumers’ Price Responsiveness

Propositions

Details and Proofs in Paper

• 1. Size & frequency of markups in data are inconsistent with M(x) = x

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Quantifying Consumers’ Price Responsiveness

Propositions

Details and Proofs in Paper

• 1. Size & frequency of markups in data are inconsistent with M(x) = x

∃ observable bounds on:

• 2. Marginal disutility of finance charges, M

′(r∗

i , p∗ i )

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Quantifying Consumers’ Price Responsiveness

Propositions

Details and Proofs in Paper

• 1. Size & frequency of markups in data are inconsistent with M(x) = x

∃ observable bounds on:

• 2. Marginal disutility of finance charges, M

′(r∗

i , p∗ i )

• 3a. Diff. btwn finance charges & disutility of finance charges, BO($)

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Quantifying Consumers’ Price Responsiveness

Propositions

Details and Proofs in Paper

• 1. Size & frequency of markups in data are inconsistent with M(x) = x

∃ observable bounds on:

• 2. Marginal disutility of finance charges, M

′(r∗

i , p∗ i )

• 3a. Diff. btwn finance charges & disutility of finance charges, BO($)
• 3b. Diff. btwn markup charges & disutility of markup charges, BM($)

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Quantifying Consumers’ Price Responsiveness

Results

Population Estimates

Table: Summary Statistics of Estimates

Variable Mean p10 p25 p50 p75 p90 M

′

i(·)

0.86 0.77 0.80 0.86 0.91 0.95 BO($) 380.12 105.71 186.81 324.33 510.73 721.56 BM($) 96.16 0.00 16.56 72.09 145.21 228.07

Note: Selected summary statistics of measures of consumers’ sensitivity to fi- nance charges. M ′

i(·) and BO i condition on positive markups. BM i

are derived for the full sample.

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Quantifying Consumers’ Price Responsiveness

Interpretation of our Results

Some Potential Explanations

Sales Tax: Do the calculations with sales tax τ. Default Risk: Consider only consumers with credit score above 720. Default risk approximately: 0.5%. Credit Constraints or Impatience Prepayment Risk

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Quantifying Consumers’ Price Responsiveness

How Consumer Impatience / Constraints...

...affect dealers’ trade-off

Auto loans have fixed payments that fully amortize. If total costs for a 72-month loan are $36,000, then consumer pays: ◮ $500 a month for 72 months, if pi is $1 and loan price is $35,999 ◮ $500 a month for 72 months, if pi is $35,999 and loan price is $1 Division of costs between car and loan has no effect on payment schedule! ⇒ Impatience/constraints do not affect pi/ri tradeoff

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Quantifying Consumers’ Price Responsiveness

How Prepayments...

...affect dealers’ trade-off

Prepayment risk means markups: ◮ Lower cost for consumers... ◮ But lower benefit for dealers, who bear “early” prepayment risk Theoretical implications unclear, and depend on: ◮ How consumers trade off total costs vs monthly payments ◮ May focus more on monthly payments (Argyle et al. (2019)) ◮ Whether dealers or consumers better predict prepayment ◮ Conditional correlations between early & late prepayment risk, etc. Empirically, higher prepayment risk predicts smaller valuation differences! Prepayment risk does not explain our results.

Heterogeneity Analysis

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Heterogeneity Analysis

Heterogeneity Analysis

Consumer and County Characteristics

Table: Analysis of Buyer Characteristics, ACS Variables, and Prepayment Probability on Estimated Bounds Overall Bound BO

i

Markup Bound BM

i

(1) (2) (3) (4) (5) (6) Log Monthly Income 8.977∗∗∗

• 10.02∗∗∗
• 9.340∗∗∗

2.005∗∗∗

• 2.326∗∗∗
• 1.813∗∗∗

(0.307) (0.244) (0.244) (0.134) (0.130) (0.130) Credit score, 100 points

• 75.66∗∗∗
• 30.68∗∗∗
• 30.58∗∗∗
• 14.22∗∗∗
• 3.942∗∗∗
• 3.562∗∗∗

(0.531) (0.371) (0.377) (0.207) (0.195) (0.196) Mileage, Tens of Thousands

• 23.72∗∗∗

4.553∗∗∗ 4.524∗∗∗

• 7.196∗∗∗
• 0.846∗∗∗
• 0.827∗∗∗

(0.091) (0.078) (0.077) (0.038) (0.038) (0.038) New Car 12.67∗∗∗

• 7.092∗∗∗
• 6.891∗∗∗
• 5.292∗∗∗
• 9.167∗∗∗
• 8.509∗∗∗

(0.520) (0.390) (0.390) (0.238) (0.227) (0.228) Log Loan Amount 384.4∗∗∗ 381.4∗∗∗ 84.11∗∗∗ 76.28∗∗∗ (0.924) (1.253) (0.309) (0.401) Average years of education

• 3.807∗∗∗
• 0.981∗∗∗

(0.335) (0.210) Fraction with internet access

• 9.523∗∗∗
• 8.801∗∗∗

(1.482) (0.918) Estimated prepayment probability

• 24.16∗∗∗
• 69.71∗∗∗

(6.432) (1.722) Fixed Effects Lender Yes Yes Yes Yes Yes Yes Model Yes Yes Yes Yes Yes Yes State Yes Yes Yes Yes Yes Yes

Note: The table shows estimates from an OLS regression.

Full Equilibrium Model

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Full Equilibrium Model

Full Model Setup

Overview

Model: ◮ BLP Differentiated Product Bertrand. ◮ Dealers set prices and interest rates for each model j. ◮ Lender competition for loans (d, j): first price auction. ◮ Recover lender cost (modification of Guerre et al. (2001)). ◮ Convex functional form, Mi(x), estimated separately. Estimation: ◮ Comprehensive market share data from AutoCount. ◮ Data from 28 states and 687 counties. ◮ Market defined as county. On average 7 dealers per county. ◮ Travel distances g(d, i) to dealers from Google Maps API.

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Full Equilibrium Model

Correct model of competition?

Alternative Models

Model a posted price market? ◮ Ignoring price variation. ◮ Ignore outside lender competition, see Allen et al. (2013). ◮ Bundle of a homogenous product and very idiosyncratic product. Why not a search framework? ◮ Model of search + intermediation, Salz (2017). ◮ We estimate 2/82/14/2% of buyers have 0/1/2/3+ dealer inquiries

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Full Equilibrium Model

Counterfactuals

Two Experiments

No Wedge:

◮ Consumers treat charges the same, M(x) = x.

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Full Equilibrium Model

Counterfactuals

Two Experiments

No Wedge:

◮ Consumers treat charges the same, M(x) = x.

No Discretion:

◮ Dealers take interest rates as given. ◮ Lenders bid in second price auction. ◮ Interest rate set by second lowest bidder. ◮ Dealer offer optimal downstream car price conditional on rate. ◮ Effects:

• 1. Lenders have less information ⇒ less price discrimination.
• 2. Contractual (double marginalization).

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Full Equilibrium Model

Counterfactuals

Two Different Experiments

Outcome No No Dealer Measure Baseline Wedge ∆% Discretion ∆% Total Price (p · (1 + r)) ($ 1000) 27181.0 26874.1

• 1.13

26906.4

• 1.29

Car Price ($ 1000) 24139.0 26683.9 10.54 23897.0

• 1.26

Interest Rate (r) 0.134 0.017

• 87.14

0.12

• 13.58
• Cons. Surplus (ˆ

ρ) ($ Billion) 36.5 35.95

• 1.38

38.0 4.12

• Cons. Surplus (ρ = 0) ($ Billion)

32.3 35.75 10.63 34.0 5.14

• Prod. Surplus ($ Billion)

2.755 2.206

• 26.6

2.19

• 26.11

Note: This table shows counterfactual outcomes for two different scenarios. In scenario No Wedge M(x) = x. In scenario No Dealer Discretion lenders set interest rates directly and dealers compete downstream in prices taking them as given. All numbers are averages across all markets, which ac- cording to our definition are counties.

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Full Equilibrium Model

Counterfactuals

Two Different Experiments

Outcome No No Dealer Measure Baseline Wedge ∆% Discretion ∆% Total Price (p · (1 + r)) ($ 1000) 27181.0 26874.1

• 1.13

26906.4

• 1.29

Car Price ($ 1000) 24139.0 26683.9 10.54 23897.0

• 1.26

Interest Rate (r) 0.134 0.017

• 87.14

0.12

• 13.58
• Cons. Surplus (ˆ

ρ) ($ Billion) 36.5 35.95

• 1.38

38.0 4.12

• Cons. Surplus (ρ = 0) ($ Billion)

32.3 35.75 10.63 34.0 5.14

• Prod. Surplus ($ Billion)

2.755 2.206

• 26.6

2.19

• 26.11

Note: This table shows counterfactual outcomes for two different scenarios. In scenario No Wedge M(x) = x. In scenario No Dealer Discretion lenders set interest rates directly and dealers compete downstream in prices taking them as given. All numbers are averages across all markets, which ac- cording to our definition are counties.

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Full Equilibrium Model

Counterfactuals

Two Different Experiments

Outcome No No Dealer Measure Baseline Wedge ∆% Discretion ∆% Total Price (p · (1 + r)) ($ 1000) 27181.0 26874.1

• 1.13

26906.4

• 1.29

Car Price ($ 1000) 24139.0 26683.9 10.54 23897.0

• 1.26

Interest Rate (r) 0.134 0.017

• 87.14

0.12

• 13.58
• Cons. Surplus (ˆ

ρ) ($ Billion) 36.5 35.95

• 1.38

38.0 4.12

• Cons. Surplus (ρ = 0) ($ Billion)

32.3 35.75 10.63 34.0 5.14

• Prod. Surplus ($ Billion)

2.755 2.206

• 26.6

2.19

• 26.11

Note: This table shows counterfactual outcomes for two different scenarios. In scenario No Wedge M(x) = x. In scenario No Dealer Discretion lenders set interest rates directly and dealers compete downstream in prices taking them as given. All numbers are averages across all markets, which ac- cording to our definition are counties.

Auto Dealer Loan Intermediation: Consumer Behavior and Competitive Effects Conclusion

Summary

Summary:

◮ Use contracts to quantify buyers’ disutility for loan vs car ◮ Average disutility from finance charges at least $360 less than cost ◮ Difference between disutility and cost larger for consumers with lower income, credit scores, education, internet access ◮ Total prices would be substantially lower if disutility equaled cost. +$3.5 billion annual consumer surplus. ◮ No Dealer Discretion leads to large increases in consumer welfare.

Additional Counterfactual:

◮ How do credit shocks affect the market?

Thank you!