Financial Innovation and Borrowers: Evidence from Peer-to-Peer Lending

Transcription

1 Financial Innovation and Borrowers: Evidence from Peer-to-Peer Lending Tetyana Balyuk BdF-TSE Conference November 12, 2018

2 Research Question Motivation Motivation Imperfections in consumer credit market ($3.9 trillion in the U.S.) High interest rates, pooled pricing, credit rationing... The Economist:... for most borrowers, credit is scarce and costly Recent entry of financial technology (FinTech) companies Peer-to-peer (P2P) loan segment: $39.4 billion in the U.S.(Q1 2018) 6.7% of personal unsecured loan market $150 billion in P2P loans annually by 2025 (PwC) FinTech may reduce credit market inefficiencies Regulatory debate on the benefits of online lending for borrowers Research question: How does FinTech innovation impact credit provided by banks? Tetyana Balyuk Financial Innovation and Borrowers November 12, / 15

3 Research Question Contribution Contribution Literature on financial innovation (Boot and Thakor, 1997; Thakor, 2011; Hauswald and Marquez, 2003; Philippon, 2016) Mostly financial innovation in the product space Papers on technological innovation Focus on information acquisition and competition effects FinTech and P2P lending literature (Duarte et al., 2012; Ravina, 2012; Butler et al., 2016; Hertzberg et al., 2018) Mostly around determinants of funding on P2P lending platforms Recent contributions on FinTech and credit markets, market design This paper: Technology side of financial innovation focusing on P2P lending Key finding: FinTech innovation increases access to bank credit Tetyana Balyuk Financial Innovation and Borrowers November 12, / 15

4 Research Question Findings The Agenda 1 FinTech innovation causally increases access to bank credit Effect primarily comes from existing lenders Driver: borrowers who are ex ante more likely credit constrained (shorter credit histories, lower credit scores) 2 Cannot be explained by changes in demand or creditworthiness More access to bank credit for borrowers who shift away from bank debt and those who do not reduce their use of bank debt Credit scores do not improve 3 No evidence of higher delinquencies despite larger total debt Findings consistent with information spillovers in sequential lending Tetyana Balyuk Financial Innovation and Borrowers November 12, / 15

5 Data and Setting P2P Lending Peer-to-Peer (P2P) Lending How it works Direct matching of borrowers and lenders on-line Borrower On-line request: $12k Submits hard data Platform Verifies data APR: 6.4%-36% Investors To fund or not? Funding amount? The innovation of P2P lending New public market for unsecured consumer debt Innovation in how hard information is processed Fully-automated algorithm-based pricing and underwriting Technological shock to screening costs Tetyana Balyuk Financial Innovation and Borrowers November 12, / 15

6 Data and Setting Data Data and Institutional Details Data from Prosper Marketplace (Prosper) Anonymized application-level data Credit bureau data + borrower & loan characteristics Institutional setting 1 Prosper tracks repeat applicants (18% of sample) Allows constructing a panel used in empirical tests Robustness checks to rule out selection concerns 2 Loan amount and interest rate set prior to funding 3 Minimum threshold for the loan to fund: 70% Identification with regression discontinuity design (RDD) Sample: January 2011 to September 2015 Tetyana Balyuk Financial Innovation and Borrowers November 12, / 15

7 Data and Setting Data Who Borrows from P2P Lending Platforms? 30% Prosper's platform U.S. population 25% 20% 15% 10% 5% 0% [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] (N=445,502) Mean SD Median Min Max U.S. population Monthly income ($ 000) Mean: Employment (years) (per capita) Credit history (years) Home owner (1/0) Total debt ($ 000) Median: 60.4 Open accounts (household) Revolving accounts Credit card utilization 54.5% 27.0% 56.0% 0% 122% Tetyana Balyuk Financial Innovation and Borrowers November 12, / 15

8 Data and Setting Data Who Borrows from P2P Lending Platforms? 30% Prosper's platform U.S. population 25% 20% 15% 10% 5% 0% [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] (N=445,502) Mean SD Median Min Max U.S. population Monthly income ($ 000) Mean: Employment (years) (per capita) Credit history (years) Home owner (1/0) Total debt ($ 000) Median: 60.4 Open accounts (household) Revolving accounts Credit card utilization 54.5% 27.0% 56.0% 0% 122% Tetyana Balyuk Financial Innovation and Borrowers November 12, / 15

9 Data and Setting Data Who Borrows from P2P Lending Platforms? 30% Prosper's platform U.S. population 25% 20% 15% 10% 5% 0% [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] (N=445,502) Mean SD Median Min Max U.S. population Monthly income ($ 000) Mean: Employment (years) (per capita) Credit history (years) Home owner (1/0) Total debt ($ 000) Median: 60.4 Open accounts (household) Revolving accounts Credit card utilization 54.5% 27.0% 56.0% 0% 122% Tetyana Balyuk Financial Innovation and Borrowers November 12, / 15

10 Data and Setting Data Who Borrows from P2P Lending Platforms? 30% Prosper's platform U.S. population 25% 20% 15% 10% 5% 0% [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] (N=445,502) Mean SD Median Min Max U.S. population Monthly income ($ 000) Mean: Employment (years) (per capita) Credit history (years) Home owner (1/0) Total debt ($ 000) Median: 60.4 Open accounts (household) Revolving accounts Credit card utilization 54.5% 27.0% 56.0% 0% 122% Tetyana Balyuk Financial Innovation and Borrowers November 12, / 15

11 Results Access to Credit Access to Bank Credit: OLS with Borrower FE Y ist = βp2p loan is,t 1 + X ist ζ + α i + γ st + u ist Dependent variable: Revolver limits ($ 000) Revolving accounts (#) (1) (2) (3) (4) (5) (6) (7) (8) P2P loan t *** 1.02*** 0.78*** 0.82*** 0.11*** 0.11*** 0.12*** 0.12*** Economic magnitude 2.5% 2.6% 2.0% 2.1% 1.3% 1.3% 1.5% 1.5% Borrower controls FICO bins ScoreX bins Borrower FE State & year FE State-year FE Observations 126, , , , , , , ,156 Adj. within R Results: FinTech associated with more access to bank credit (2.6% in limits) Size of effect more than half of effect of home ownership Effect primarily comes from existing lenders (only 1.3% in accounts) Tetyana Balyuk Financial Innovation and Borrowers November 12, / 15

12 Results Access to Credit Endogeneity and Identification Endogeneity: Unobservable borrower risk Regression discontinuity design (RDD) Probability of loan origination P2P loan origination Percent funded Local fuzzy RDD around 70%±10% Causality for marginally-funded borrowers Robusteness checks of internal validity P2P loan t 1 Above t *** (8.01) Distance (-0.64) Distance Above t (-0.48) Distance (-0.76) Distance 2 Above t (0.98) Control for lagged outcomes State & year FE Observations 810 Adj. R Tetyana Balyuk Financial Innovation and Borrowers November 12, / 15

13 Results Access to Credit Access to Bank Credit: Conditional Expectation Revolver limits ($'000) Revolver limits Revolving accounts (#) Revolving accounts Percent funded Percent funded Revolver limits are higher for borrowers above the threshold Not all borrowers above the threshold receive P2P loans No fixed effects or other controls Tetyana Balyuk Financial Innovation and Borrowers November 12, / 15

14 Results Access to Credit Access to Bank Credit: RDD P2P loan is,t 1 = φ 1 Above is,t 1 + φ 2 f (D is,t 1 ) + φ 3 f (D is,t 1 )Above is,t 1 + φ 4 Y is,t 1 + γ s + δ t + ω ist Y ist = β 1 P2P loan is,t 1 + β 2 f (D is,t 1 ) + β 3 f (D is,t 1 )Above is,t 1 + β 4 Y is,t 1 + γ s + δ t + ɛ ist (1st stage) (2nd stage) Dependent variable: Revolver limits ($ 000) Revolving accounts (#) 10% bandwidth (1) (2) (3) (4) (5) (6) (7) (8) P2P loan t *** 46.6*** 46.0*** 38.4** (3.28) (3.14) (3.08) (2.00) (-0.89) (-0.80) (-0.71) (-0.33) Economic magnitude 114.4% 104.2% 102.8% 85.8% -5.7% -4.5% -3.9% -2.1% Distance Distance Above t 1 Distance 2 Distance 2 Above t 1 Revolver limits t 1 Revolving accounts t 1 State & year FE Observations Adjusted R FinTech lending causes an increase in access to bank credit Much larger magnitude: LATE for marginally-funded borrowers Little evidence of increase in the number of accounts in RDD Tetyana Balyuk Financial Innovation and Borrowers November 12, / 15

15 Results Access to Credit Access to Bank Credit: Donut RDD P2P loan is,t 1 = φ 1 Above is,t 1 + φ 2 f (D is,t 1 ) + φ 3 f (D is,t 1 )Above is,t 1 + φ 4 Y is,t 1 + γ s + δ t + ω ist Y ist = β 1 P2P loan is,t 1 + β 2 f (D is,t 1 ) + β 3 f (D is,t 1 )Above is,t 1 + β 4 Y is,t 1 + γ s + δ t + ɛ ist (1st stage) (2nd stage) Dependent variable: Revolver limits ($ 000) Revolving accounts (#) 10% bandwidth (1) (2) (3) (4) (5) (6) (7) (8) P2P loan t *** 38.6*** 39.1*** 25.3* (3.03) (2.92) (2.92) (1.73) (-0.00) (0.09) (0.09) (0.73) Economic magnitude 93.5% 88.0% 89.2% 57.7% -0.02% 0.4% 0.5% 3.9% Distance Distance Above t 1 Distance 2 Distance 2 Above t 1 Revolver limits t 1 Revolving accounts t 1 State & year FE Observations Adjusted R Similar results qualitatively with somewhat smaller magnitudes No effect on access to credit in placebo RDD Tetyana Balyuk Financial Innovation and Borrowers November 12, / 15

16 Results Access to Credit Access to Bank Credit: Heterogenous Effects Sensitivity to length of credit history Revolver limits ($ 000) Revolving accounts (#) (1) (2) (3) (4) (5) (6) (7) (8) Credit hist. P2P loan t *** -0.13*** -0.14*** -0.14*** *** *** *** *** P2P loan t *** 3.50*** 3.36*** 3.39*** 0.22*** 0.22*** 0.25*** 0.25*** Controls: Borrower characteristics: (1) (8); FICO bins: (1) (2) & (5) (6); ScoreX bins: (3) (4) & (7) (8) Fixed Effects: Borrower: (1) (8); State & year: (1), (3), (5) & (7); State-year: (2), (4), (6) & (8) Sample split by credit score (FICO) Revolver limits ($'000) 10% 8% *** 1.45*** 6% 4% % % -2% % -6% % -10% FICO [$<$660] FICO [ ] FICO [ ] FICO [ ] FICO [ ] FICO [ ] FICO [ ] FICO [ ] FICO [ ] Information or screening costs? Increase in access to credit larger for borrowers with: Shorter credit histories Lower credit scores (5.21% ) Ex ante higher likelihood to be credit constrained Tetyana Balyuk Financial Innovation and Borrowers November 12, / 15

17 Results Access to Credit Access to Bank Credit: Key Findings Findings so far... 1 Banks increase access to credit for borrowers who get P2P loans 2 Effect primarily comes from existing lenders 3 Effect stronger for consumers who are more likely credit-rationed Inconsistent with negative externalities from higher default risk Competition and adverse selection theories (Broecker, 1990) Sequential banking theory (Bizer and DeMarzo, 1992) Consistent with certification and information spillovers Information processing and aggregation theories (Allen and Gale, 1998) Theories of information spillovers (Welch, 1992, Sufi, 2009) Tetyana Balyuk Financial Innovation and Borrowers November 12, / 15

18 Results Credit Demand Explained by Borrowing Patterns or Creditworthiness? OLS with Borrower FE Dependent variable: Revolving balance ($ 000) FICO score ScoreX score (1) (2) (3) (4) P2P loan t *** -2.26*** 8.78*** -3.65*** Economic magnitude -7.6% -11.7% 1.3% -0.5% Controls: Borrower characteristics: (1) (4); FICO bins: (1); ScoreX bins: (2) Fixed Effects: Borrower: (1) (4); State-year: (1) (4) Donut RDD RDD model: Linear Quadratic Linear Quadratic + splines + splines Revolving balance ($ 000) Coefficient Magnitude 11.1% 12.1% 12.8% 8.9% FICO score Coefficient Magnitude 0.7% 0.7% 0.3% 0.2% ScoreX score Coefficient * -29.7* Magnitude -3.3% -3.6% -3.8% -4.2% Tetyana Balyuk Financial Innovation and Borrowers November 12, / 15

19 Results Credit Demand Information or Transaction Costs? Sample split by credit score (FICO) 5% 0% Revolving balance ($'000) -5% -10% -15% -20% -25% -0.65*** -0.62** -1.01*** -1.90*** -2.79*** -2.47*** -2.25*** -3.75*** % FICO [$<$660] FICO [ ] FICO [ ] FICO [ ] FICO [ ] FICO [ ] FICO [ ] FICO [ ] FICO [ ] Debt refinancing larger for borrowers with higher credit scores Consumers with higher pre-existing costs of pooling Consistent with improved accuracy of processing hard information Tetyana Balyuk Financial Innovation and Borrowers November 12, / 15

20 Results Delinquency Excess Borrowing and Delinquencies OLS with Borrower FE Donut RDD Results: Dependent variable: Total debt ($ 000) Delinquency (1/0) (1) (2) (3) (4) P2P loan t *** 5.55*** * *** Economic magnitude 4.5% 3.6% -3.2% -8.6% RDD model: Linear Quadratic Linear Quadratic + splines + splines Total debt ($ 000) Coefficient Magnitude 3.5% 3.1% 2.9% 15.7% Delinquency (1/0) Coefficient Magnitude -11.6% -22.6% -23.2% 34.1% No evidence of higher delinquency rates despite somewhat higher debt Inconsistent with overborrowing (Laibson, 1997) Tetyana Balyuk Financial Innovation and Borrowers November 12, / 15

21 Summary Findings Summary of Findings 1 Banks increase access to credit following a P2P loan Effect primarily comes from existing lenders Driver: borrowers who are ex ante more likely credit-rationed 2 Effect cannot be explained by changes in demand or creditworthiness P2P borrowers refinance expensive debt on average Marginally-funded borrowers do not change borrowing BUT: access to bank credit increases for both groups Little evidence that credit scores improve 3 No evidence of increase in delinquency rates despite higher debt Consistent with certification and information spillovers Key Takeaway: FinTech lending increases access to credit even from existing lenders Tetyana Balyuk Financial Innovation and Borrowers November 12, / 15