C C D E

C C D E C C D E

C C D E

C C

C

Figure 1: Distribution of of s Density 0 1 2 2 4 3 6 4 5 8 4 2 0 0 2 4.5 Fraction of Applications Approved 0.2.4.6.8 1 Figure 2: The Credit-Score Regression Discontinuity 4 2 0 2 4 Actual Approval Rate Predicted Approval Rate Figure 1 plots the distribu on of the credit score for rst- me payday loan applicants. The credit score has been rescaled. It equals the raw credit score minus the threshold for loan approval chosen by the lender, divided by the standard devia on of scores among this lender s rst- me applicants. We normalize by di erent standard devia ons for applica ons before and a er an August 2002 change in the scoring formula. The dashed line marks the threshold for loan approval; about 80% of rst- me applica ons are approved. Figure 2 plots the probability of approval for rst- me payday loan applicants as a func on of the credit score. Each point represents one of 100 quan les in the credit score. Points shown are at the medians of their quan les on the x-axis and at the means of their quan les on the y-axis. The predicted approval-rate func on plots the best- ng quar c polynomials on both sides of the credit score threshold. Source: Authors' calcula ons based on data from a na onal payday lending company and the Texas Bankruptcy Courts' PACER Database.!!!

Figure 3: One Year Bankruptcy Rates Figure 3a: Chapter 7 Bankruptcies Bankruptcy Probability (%) 0.5 1 1.5.5 0.5 Actual Bankruptcy Rate Predicted Bankruptcy Rate Figure 3b: Chapter 13 Bankruptcies Bankruptcy Probability (%) 0 1 2 3 4.5 0.5 Actual Bankruptcy Rate Predicted Bankruptcy Rate Figures 3a and 3b plot the e ect of payday loan access on chapter 7 and chapter 13 bankruptcy pe ons within the 12 months a er applicants rst payday loan applica on. Each point represents one of 200 bins. Points shown are at the medians of their bins on the x-axis and at the means of their bins on the y-axis. The predicted bankruptcy-rate func on plots the best- ng quar c polynomials on both sides of the creditscore threshold from a reduced-form regression. The polynomials are restricted to 0.5 standard devia ons above and below the passing credit score. Source: Authors' calcula ons based on data from a na onal payday lending company and the Texas Bankruptcy Courts' PACER Database.

Figure 4: Two Year Bankruptcy Rates Figure 4a: Chapter 7 Bankruptcies Bankruptcy Probability (%) 0.5 1 1.5 2.5 0.5 Actual Bankruptcy Rate Predicted Rate Figure 4b: Chapter 13 Bankruptcies Bankruptcy Probability (%) 0 2 4 6.5 0.5 Actual Bankruptcy Rate Predicted Rate Figures 4a and 4b plot the e ect of payday loan access on chapter 7 and chapter 13 bankruptcy pe ons within the 24 months a er applicants rst payday loan applica on. Each point represents one of 200 bins. Points shown are at the medians of their bins on the x-axis and at the means of their bins on the y-axis. The predicted bankruptcy-rate func on plots the best- ng quar c polynomials on both sides of the creditscore threshold from a reduced-form regression. The polynomials are restricted to 0.5 standard devia ons above and below the passing credit score. Source: Authors' calcula ons based on data from a na onal payday lending company and the Texas Bankruptcy Courts' PACER Database.

Figure 5: Effect of Payday Loans on Bankruptcy as a Function of Bandwidth Figure 5a: Chapter 7 Bankruptcy within 2 years Coefficient Estimates (Percentage Pts) 0.5 0 2 3 Bandwidth (standard deviations around the threshold) Coefficient Estimates (Percentage Pts) 0 2 4 0 0.5 2 Bandwidth (standard deviations around the threshold) Bandwidth.5 2 Bandwidth 0.5 2.5 Bandwidth 0.5 2 Bandwidth Figures 5a and 5b plot the reduced-form coe cient es mates from the main regressions as a func on of the window around the threshold in standard devia ons for chapter 7 and 13 bankruptcies respec vely. The line with open circles plots the point es mates for each regression varying the bandwidth around the credit-score threshold up to 2 standard devia ons. The solid lines represent +/--- two---standard errors of the es mates. Source: Authors' calcula ons based on data from a na onal payday lending company and the Texas Bankruptcy Courts' PACER Database.

Figure 6: Effect of Payday Loan Access on Bankruptcy Over Time Percentage Points 0 1 2 3 4 5 Figure 6a: All Bankruptcies 0.5 1 1.5 2 2.5 Years Since First PDL Application Estimated coefficient Coefficient 2 se Coefficient + 2 se Percentage Points 1.5 0.5 1 Percentage Points 0 1 2 3 4 Figure 6b: Chapter 7 Bankruptcies 0.5 1 1.5 2 2.5 Years Since First PDL Application Figure 6c: Chapter 13 Bankruptcies Estimated coefficient Coefficient 2 se Coefficient + 2 se 0.5 1 1.5 2 2.5 Years Since First PDL Application Estimated coefficient Coefficient 2 se Coefficient + 2 se Figures 6a, 6b, and 6c plot the e ect of payday loan access on bankruptcy pe ons over me. The middle line represents the reduced-form e ect of First Applica on on Approved with a 0.5 standard devia on bandwidth around the threshold. The other lines are two-standard-error bands. The reduced-form regressions producing these es mates include quar c polynomials on both sides of the credit-score threshold, demographic controls, and dummies for month of rst applica on. Source: Authors' calcula ons based on data from a na onal payday lending company and the electronic records from Texas Bankruptcy Courts via PACER.

Figure 7: Bankruptcy Probability as a Function of the within One Year High-Market Share Locations Figure 7a: Chapter 7 Bankruptcy Filings Bankruptcy Probability (%) 0 1 2 3 4 Bankruptcy Probability (%) 0 1 2 3 4.5 0.5 Actual Bankruptcy Rate Predicted Bankruptcy Rate Figure 7b: Chapter 13 Bankruptcy Filings.5 0.5 Actual Bankruptcy Rate Predicted Bankruptcy Rate Figures 7a and 7b plot the effect of payday loan access on chapter 7 and 13 bankruptcy pe ons within the 12 months a er applicants rst payday loan applica on. The sample is restricted to payday loan outlets with 100% market share in their zip code. Each point represents one of 200 bins. The predicted bankruptcy-rate func on plots the best- ng quar c polynomials on both sides of the credit-score threshold. The polynomials are 0.5 standard devia ons above and below the passing credit score. Source: Authors calcula ons based on data from a na onal payday lending company and the Texas Bankruptcy Courts PACER Database.

Figure 8: Bankruptcy Probability as a Function of the within Two Years High-Market Share Locations Figure 8a: Chapter 7 Bankruptcy Filings Probability of Bankruptcy (%) 0 1 2 3 4.5 0.5 Actual Bankruptcy Rate Predicted Bankruptcy Rate Figure 8b: Chapter 13 Bankruptcy Filings Probability of Bankruptcy (%) 0 2 4 6 8 10.5 0.5 Actual Bankruptcy Rate Predicted Bankruptcy Rate Figures 8a and 8b plot the e ect of payday loan access on chapter 7 and 13 bankruptcy pe ons within the 24 months a er applicants rst payday loan applica on. The sample is restricted to payday loan outlets with 100% market share in their zip code. Each point represents one of 200 bins. The predicted bankruptcy-rate func on plots the best- ng quar c polynomials on both sides of the credit-score threshold. The polynomials are 0.5 standard devia ons above and below the passing credit score. Source: Authors calcula ons based on data from a na onal payday

Figure 9a: Subsequent Payday Loan Applications Number of payday loans within 2 years Dollars borrowed within 2 years 0 1000 2000 3000 4000 5000 Number payday loans within 2 years 0 5 10 15 20 4 2 0 2 4 Actual Application Rate Predicted Application Rate Figure 9b: Dollar Amount of Subsequent Payday Loans $ borrowed within 2 years 0 2 4 Actual Application Rate Predicted Application Rate Figures 9a and 9b plot the e ect of payday loan access on subsequent borrowing within 24 months of an applicants rst payday loan applica on. Points shown are the medians of their quan les on the x axis and at the mean of their quan les on the y-axis. The predicted line plots the best- ng quar c polynomials on both sides of the credit-score threshold. Figure 9a plots the e ect of payday loan access on the number of subsequent payday loan applica ons made. Figure 9b plots the dollar amount subsequently borrowed. Source: Authors calcula ons based on data from a na onal payday lending company and the Texas Bankruptcy Courts PACER Database.

Figure 10: The Effect of Payday Loan Access Over Time Figure 10a: Number of Subsequent Payday Loan Applications Number of Subsequent Payday Loan Applications 0 2 4 6 0.5 1 1.5 2 Years Since First PDL Application Estimated coefficient Coefficient - 2 s.e. Coefficient + 2 s.e. Figure 10b: Amt ($) Subsequently Borrowed Amt ($) Subsequently Borrowed 0 500 1000 1500 2000 0.5 1 1.5 2 Years Since First PDL Application Estimated coefficient Coefficient - 2 s.e. Coefficient + 2 s.e. Figure 10c: Finance Charges ($) Paid Subsequently Finance Charges ($) Paid Subsequently 50 100 150 200 250 300 0.5 1 1.5 2 Years Since First PDL Application Estimated coefficient Coefficient - 2 s.e. Coefficient + 2 s.e. Figures 10a, 10b and 10c plot the number of subsequent applica on made at this company, the dollar amount borrowed, and the nance charges paid to this company, respec vely. The middle line represents the reduced-form es mated e ect of First Applica on Approved on subsequent behavior in the payday loan market. The other lines are two-standard-error bands. Regressions producing these es mates include quar c polynomials on both sides of the credit-score threshold, demographic controls, and dummies for month of rst applica on. Source: Authors calcula ons based on data from a na onal payday lending company and the Texas Bankruptcy Courts PACER Database.

Appendix Figure A1: First Stage Subsamples Fraction of Applications Approved 0.2.4.6.8 1 Black 1.5 0.5 1 Fraction of Applications Approved 0.2.4.6.8 1 Hispanic 1.5 0.5 1 Fraction of Applications Approved 0.2.4.6.8 1 White 1.5 0.5 1 Male Female Renters Fraction of Applications Approved 0.2.4.6.8 1 1.5 0.5 1 Fraction of Applications Approved 0.2.4.6.8 1 1.5 0.5 1 Fraction of Applications Approved 0.2.4.6.8 1 1.5 0.5 1 Homeowners Direct Depositors Non-Direct Depositors Fraction of Applications Approved 0.2.4.6.8 1 1.5 0.5 1 Fraction of Applications Approved 0.2.4.6.8 1 1.5 0.5 1 Fraction of Applications Approved 0.2.4.6.8 1 1.5 0.5 1 Figure A1 is the same as Figure 2 but shows the first-stage gures by subsamples. Source: Authors calcula ons based on data from a na onal payday lending company and the Texas Bankruptcy Courts PACER Database.

3 Direct Deposit Age Black direct_dep 0 age 30 35 40 45 0 black 0 Actual values Dual quartic fit Actual values Dual quartic fit Actual values Dual quartic fit tenure 0 4 6 8 Job Tenure hispanic Hispanic garnish Wages Garnished Actual values 0 Dual quartic fit Actual values 0 Dual quartic fit 0 Actual values 0 Dual quartic fit lbalance 0 3 4 5 Log(Checking Account Balance) Actual values 0 Dual quartic fit owns_home 0 Homeownership Actual values 0 Dual quartic fit nsf_count 0 4 6 8 NSFs Actual values 0 Dual quartic fit Log(Monthly Pay) Months at Current Residence Paid Biweekly lmpay 3 4 5 6 7 months_residence 40 60 80 biweekly Actual values 0 Dual quartic fit Actual values 0 Dual quartic fit Actual values 0 Dual quartic fit Figure A2 plots demographic variables against the credit score. Source and notes: Authors calcula ons based on data from a na onal payday lending company.

APPENDIX TABLE I * Significant at the 5 percent level * * Significant at the 1 percent level