Upload
others
View
10
Download
0
Embed Size (px)
Citation preview
Online AppendixInformation Asymmetries in Consumer Credit Markets:
Evidence from Payday Lending
Will DobbieHarvard University
Paige Marta SkibaVanderbilt University
March 2013
Online Appendix Table 1Difference-in-Difference Estimates of the
Effect of Loan Amount on Default
(1) (2)Loan Amount −0.036∗ −0.035∗
(0.021) (0.022)Age −0.452∗∗∗
(0.033)Black −1.120
(1.834)Male 0.800
(1.551)Credit Score −0.040∗∗∗
(0.004)Checkings 0.000
(0.002)Home Owner −1.042
(1.617)Direct Deposit −0.322
(1.421)Garnishment 8.535
(5.932)Observations 10,279 10,279
Notes: This table reports difference-in-difference estimates of the impact of loan amount on de-fault. The sample consists of first-time payday-loan borrowers living in states offering paydayloans in $50 increments who are paid biweekly or semimonthly earning between $100 and $1100every two weeks. We instrument for loan size using a linear trend in income interacted with anindicator variable for living in Tennessee and being eligible for a $200 loan. All regressions controlfor month-, year-, and state-of-loan effects. The dependent variable is an indicator for bouncinga check on the first loan. Coefficients and robust standard errors are multiplied by 100. *** =significant at 1 percent level, ** = significant at 5 percent level, * = significant at 10 percent level.
1
Online Appendix Table 2OLS Estimates of Borrower Characteristics on Loan Choice
RD Sample RK Sample(1) (2) (3) (4) (5) (6)
Pay −0.074∗∗∗ −0.076∗∗∗ 0.020∗∗ −0.006∗∗∗ −0.006∗∗∗ 0.016∗∗∗
(0.002) (0.002) (0.010) (0.000) (0.000) (0.000)Age 0.152∗∗∗ 0.156∗∗∗ 0.026∗∗ 0.031∗∗∗
(0.038) (0.038) (0.011) (0.011)Black −8.281∗∗∗ −8.558∗∗∗ −0.838∗ −0.718
(2.968) (2.985) (0.472) (0.467)Male 0.894 1.238 1.187∗∗ 1.882∗∗∗
(2.739) (2.707) (0.491) (0.487)Credit Score −0.002 −0.003 −0.005∗∗∗ −0.002∗
(0.006) (0.006) (0.001) (0.001)Checkings 0.003∗ 0.003 0.001∗∗∗ 0.002∗∗∗
(0.002) (0.002) (0.000) (0.000)Home Owner 1.011 0.885 4.191∗∗∗ 4.371∗∗∗
(2.915) (2.914) (0.549) (0.545)Direct Deposit −0.167 −1.047 −2.260∗∗∗ −0.748∗∗
(2.183) (2.165) (0.383) (0.381)Garnishment −0.594 −0.958 −0.258 −0.641
(7.422) (7.368) (1.531) (1.516)Loan Eligibility −0.202∗∗∗ −0.094∗∗∗
(0.021) (0.002)R2 0.242 0.246 0.254 0.166 0.168 0.182Observations 9,473 9,473 9,473 130,025 130,025 130,025
Notes: This table reports OLS estimates of the cross-sectional correlation between borrower char-acteristics and loan choice. The regression discontinuity (RD) sample consists of first-time payday-loan borrowers living in states offering payday loans in $50 increments who are paid biweekly orsemimonthly earning between $100 and $1100 every two weeks. The regression kink (RK) sam-ple consists of first-time payday-loan borrowers living in states offering payday loans in $1 or $10increments who are paid biweekly or semimonthly earning more than $100 and within $1000 ofa kink point. The dependent variable is an indicator for choosing the largest loan the borrower iseligible for. All regressions control for month-, year-, and state-of-loan effects. Coefficients androbust standard errors are multiplied by 100. *** = significant at 1 percent level, ** = significantat 5 percent level, * = significant at 10 percent level.
2
Online Appendix Table 3Regression Discontinuity Tests of Quasi-Random Assignment
Polynomial Spline LinearCharacteristics (1) (2) (3)
Age 0.204 0.260 0.021(0.295) (0.289) (0.528)9443 9443 9443
Black 0.012 0.007 0.021(0.025) (0.027) (0.043)1316 1316 1316
Male 0.005 0.009 −0.038(0.027) (0.028) (0.044)1316 1316 1316
Credit Score −2.537 −4.855 −18.322(7.281) (7.357) (13.882)2165 2165 2165
Checkings 4.954 6.268 −17.160(16.305) (16.800) (32.300)
2274 2274 2274Home Owner 0.007 0.004 −0.042
(0.027) (0.028) (0.047)1160 1160 1160
Direct Deposit 0.000 0.004 −0.006(0.018) (0.018) (0.032)2350 2350 2350
Garnishment −0.009 −0.009 −0.006(0.010) (0.010) (0.015)1160 1160 1160
Density TestNbr. of Borrowers 2.208 2.309 −28.313
(1.929) (1.857) (24.044)100 100 10
Notes: This table reports tests of quasi-random assignment in our regression discontinuity design.The sample consists of first-time payday-loan borrowers living in states offering payday loans in$50 increments who are paid biweekly or semimonthly earning between $100 and $1100 every twoweeks. Column 1 controls for a seventh-order polynomial in net pay. Column 2 controls for a linearspline in net pay. Column 3 stacks data from each cutoff and controls for net pay using a linearregression interacted with the loan cutoff. Loan eligibility is the maximum loan size an individualis eligible for. All regressions control for month-, year-, and state-of-loan effects. Standard errorsare clustered by pay. Number of borrowers is defined using $10 bins in pay. See text for additionaldetails. *** = significant at 1 percent level, ** = significant at 5 percent level, * = significant at 10percent level.
3
Appenidx Table 4Regression Kink Tests of Quasi-Random Assignment
$300 $500Cutoff Cutoff
Characteristics (1) (2)Age −0.209∗ −0.142∗∗∗
(0.112) (0.036)33,164 96,631
Black – −0.013∗∗∗
(0.003)40,878
Male – 0.006∗∗
(0.003)40,878
Credit Score – −1.229∗
(0.738)91,261
Checkings – 0.540(2.421)
89,844Home Owner – 0.008∗∗
(0.003)34,133
Direct Deposit – −0.025∗∗∗
(0.002)91,790
Garnishment – 0.002∗
(0.001)34,133
Density TestNbr. of Borrowers −4.754 −1.546
(5.528) (4.490)61 77
Notes: This table reports tests of quasi-random assignment in our regression kink design. Thesample consists of first-time payday-loan borrowers living in states offering payday loans in $1or $10 increments who are paid biweekly or semimonthly earning more than $100 and within$1000 of a kink point. Loan cutoff is an indicator for eligibility for the largest loan available ina state. All regressions using baseline characteristics control pay and month-, year-, and state-of-loan effects. Standard errors are clustered by pay. Number of borrowers is defined using $10 binsin pay. Regressions using the number of borrowers control for a seventh-order polynomial in payinteracted with the loan cutoff. See text for additional details. *** = significant at 1 percent level,** = significant at 5 percent level, * = significant at 10 percent level.
4
Online Appendix Table 5Regression Discontinuity Falsification Test of the First Stage
Polynomial Linear Spline Local Linear(1) (2) (3) (4) (5) (6)
Loan Cutoff 2.750 2.755 2.864 2.860 1.657 1.790(2.074) (2.063) (2.074) (2.064) (1.042) (1.167)
Age 0.149∗∗∗ 0.149∗∗∗ 0.142∗∗∗
(0.026) (0.026) (0.026)Black 1.233 1.234 1.166
(1.021) (1.021) (0.994)Male −1.893∗ −1.890∗ −2.028∗
(1.116) (1.116) (1.116)Credit Score −0.005 −0.005 −0.006∗∗
(0.003) (0.003) (0.003)Checkings 0.006∗∗∗ 0.006∗∗∗ 0.006∗∗∗
(0.001) (0.001) (0.001)Home Owner 9.371∗∗∗ 9.362∗∗∗ 9.408∗∗∗
(1.264) (1.263) (1.261)Direct Deposit 0.710 0.719 0.375
(0.823) (0.823) (0.751)Garnishment −2.675 −2.662 −2.651
(3.371) (3.370) (3.368)Observations 101,026 101,026 101,026 101,026 101,026 101,026
Notes: This table reports regression discontinuity first-stage estimates in a sample of states whereno effect is expected. The sample consists of first-time payday-loan borrowers living in statesoffering payday loans in $1 or $10 increments who are paid biweekly or semimonthly earningbetween $100 and $1100 every two weeks. Columns 1-2 control for a seventh-order polynomial innet pay. Columns 3-4 control for a linear spline in net pay. Columns 5-6 stack data from each cutoffand control for net pay using a linear regression interacted with the loan cutoff. The dependentvariable is the dollar amount of the borrower’s first loan. Loan eligibility is the maximum loan sizean individual is eligible for. All regressions control for month-, year-, and state-of-loan effects.Columns 5 and 6 also control for cutoff fixed effects. Standard errors are clustered by pay. *** =significant at 1 percent level, ** = significant at 5 percent level, * = significant at 10 percent level.
5
Online Appendix Table 6Regression Discontinuity Falsification Test of the Main Results
Polynomial Linear Spline Local Linear(1) (2) (3) (4) (5) (6)
Loan Amount 0.127 0.110 0.121 0.091 0.014 0.021(0.158) (0.131) (0.128) (0.126) (0.025) (0.023)
Age −0.307∗∗∗ −0.304∗∗∗ −0.301∗∗∗
(0.022) (0.021) (0.020)Black 2.966∗∗∗ 2.990∗∗∗ 3.111∗∗∗
(0.344) (0.338) (0.291)Male 2.401∗∗∗ 2.361∗∗∗ 2.121∗∗∗
(0.416) (0.406) (0.381)Credit Score −0.019∗∗∗ −0.019∗∗∗ −0.016∗∗∗
(0.001) (0.001) (0.001)Checkings −0.002∗ −0.001∗ −0.001∗∗
(0.001) (0.001) (0.001)Home Owner −2.321∗ −2.143∗ −1.728∗∗∗
(1.288) (1.240) (0.503)Direct Deposit −1.920∗∗∗ −1.907∗∗∗ −1.433∗∗∗
(0.305) (0.302) (0.423)Garnishment 0.878 0.839 0.939
(1.288) (1.275) (1.258)Observations 101,026 101,026 101,026 101,026 101,026 101,026
Notes: This table reports regression discontinuity estimates of loan amount on default in a sampleof states where no effect is expected. The sample consists of first-time payday-loan borrowers liv-ing in states offering payday loans in $1 or $10 increments who are paid biweekly or semimonthlyearning between $100 and $1100 every two weeks. Columns 1-2 control for a seventh-order poly-nomial in net pay. Columns 3-4 control for a linear spline in net pay. Columns 5-6 stack datafrom each cutoff and control for net pay using a linear regression interacted with the loan cutoff.The dependent variable is an indicator for bouncing a check on the first loan. All regressions in-strument for loan amount using loan eligibility and control for month-, year-, and state-of-loaneffects. Columns 5 and 6 also control for cutoff fixed effects. Standard errors are clustered by pay.Coefficients and standard errors are multiplied by 100. *** = significant at 1 percent level, ** =significant at 5 percent level, * = significant at 10 percent level.
6
Online Appendix Figure 1ARegression Discontinuity Tests of Quasi-Random Assignment
Baseline Characteristics
Age
25
30
35
40
Age
200 400 600 800 1000Pay
25
30
35
40
Age
200 400 600 800 1000Pay
-2-1
01
Resid
ualiz
ed A
ge
-50 -25 0 25 50Pay
A. Polynomial B. Linear Spline C. Local Linear
Fraction Black
.4.6
.81
Black
200 400 600 800 1000Pay
.4.6
.81
Black
200 400 600 800 1000Pay
-.05
0.0
5R
esid
ualiz
ed B
lack
-50 -25 0 25 50Pay
A. Polynomial B. Linear Spline C. Local Linear
Fraction Male
0.2
.4.6
.8Male
200 400 600 800 1000Pay
0.2
.4.6
.8Male
200 400 600 800 1000Pay
-.3
-.2
-.1
0.1
Resid
ualiz
ed M
ale
-50 -25 0 25 50Pay
A. Polynomial B. Linear Spline C. Local Linear
Credit Score
400
450
500
550
600
Cre
dit S
core
200 400 600 800 1000Pay
400
450
500
550
600
Cre
dit S
core
200 400 600 800 1000Pay
050
100
Resid
ualiz
ed C
redit S
core
-50 -25 0 25 50Pay
A. Polynomial B. Linear Spline C. Local Linear
7
Checking Balance0
200
400
600
800
1000
Checkings
200 400 600 800 1000Pay
0200
400
600
800
1000
Checkings
200 400 600 800 1000Pay
-200
-100
0100
200
Resid
ualiz
ed C
heckin
gs
-50 -25 0 25 50Pay
A. Polynomial B. Linear Spline C. Local Linear
Home Ownership
0.2
.4.6
.8H
om
e O
wner
200 400 600 800 1000Pay
0.2
.4.6
.8H
om
e O
wner
200 400 600 800 1000Pay
-.05
0.0
5.1
.15
Resid
ualiz
ed H
om
e O
wner
-50 -25 0 25 50Pay
A. Polynomial B. Linear Spline C. Local Linear
Direct Deposit
.1.2
.3.4
.5.6
Direct D
eposit
200 400 600 800 1000Pay
.1.2
.3.4
.5.6
Direct D
eposit
200 400 600 800 1000Pay
-.06
-.04
-.02
0.0
2.0
4R
esid
ualiz
ed D
irect D
eposit
-50 -25 0 25 50Pay
A. Polynomial B. Linear Spline C. Local Linear
Garnishment
0.1
.2.3
Garn
ishm
ent F
lag
200 400 600 800 1000Pay
0.1
.2.3
Garn
ishm
ent F
lag
200 400 600 800 1000Pay
-.02
0.0
2.0
4R
esid
ualiz
ed G
arn
ishm
ent F
lag
-50 -25 0 25 50Pay
A. Polynomial B. Linear Spline C. Local Linear
8
Notes: These figures plot baseline characteristics and biweekly pay for first-time payday borrowersin our regression discontinuity sample. The sample consists of borrowers living in states offeringpayday loans in $50 increments who are paid biweekly or semimonthly between $100 and $1100.The smoothed line in the first column of figures controls for a seventh-order polynomial in net pay.The second column controls for a linear spline in net pay. The third column stacks data from eachcutoff and controls for net pay using a linear regression and a linear regression interacted with theloan cutoff. See text for additional details.
9
Online Appendix Figure 1BRegression Discontinuity Tests of Quasi-Random Assignment
Number of Observations
A. Polynomial B. Linear Spline
02
04
06
08
0N
um
be
r o
f B
orr
ow
ers
200 400 600 800 1000Pay
02
04
06
08
0N
um
be
r o
f B
orr
ow
ers
200 400 600 800 1000Pay
C. Local Linear
26
02
80
30
03
20
34
0N
um
be
r o
f B
orr
ow
ers
-50 -25 0 25 50Pay
Notes: These figures plot the number of borrowers and biweekly pay for first-time payday bor-rowers in our regression discontinuity sample. The sample consists of borrowers living in statesoffering payday loans in $50 increments who are paid biweekly or semimonthly between $100 and$1100. The smoothed line in the first figure controls for a seventh-order polynomial in net pay.The second figure controls for a linear spline in net pay. The third figure stacks data from eachcutoff and controls for net pay using a linear regression and a linear regression interacted with theloan cutoff. See text for additional details.
10
Online Appendix Figure 2ARegression Kink Results
Test of Quasi-Random Assignment
Fraction Black Fraction Male
.3.4
.5.6
.7B
lack
200 400 600 800 1000 1200 1400 1600 1800 2000Pay
$500 Cap
.2.4
.6.8
Ma
le
200 400 600 800 1000 1200 1400 1600 1800 2000Pay
$500 Cap
Credit Score Checking Balance
40
04
50
50
05
50
Cre
dit S
co
re
200 400 600 800 1000 1200 1400 1600 1800 2000Pay
$500 Cap
10
02
00
30
04
00
50
06
00
Ch
eckin
gs
200 400 600 800 1000 1200 1400 1600 1800 2000Pay
$500 Cap
Home Ownership Direct Deposit
.05
.1.1
5.2
Ho
me
Ow
ne
r
200 400 600 800 1000 1200 1400 1600 1800 2000Pay
$500 Cap
.2.3
.4.5
.6D
ire
ct
De
po
sit
200 400 600 800 1000 1200 1400 1600 1800 2000Pay
$500 Cap
11
Garnishment Age
0.0
05
.01
.01
5.0
2G
arn
ish
me
nt
Fla
g
200 400 600 800 1000 1200 1400 1600 1800 2000Pay
$500 Cap
30
35
40
45
Ag
e
200 400 600 800 1000 1200 1400 1600 1800 2000Pay
$300 Cap $500 Cap
Notes: These figures plots average baseline characteristics and biweekly pay for first-time paydayborrowers in our regression kink sample. The sample consists of borrowers living in states offeringpayday loans in $1 or $10 increments who are paid biweekly or semimonthly and earning morethan $100 and within $1000 of a kink point. The smoothed line controls for pay interacted withbeing eligible for the maximum loan size in a state. Age is the only baseline characteristic availablefor states with a $300 cap. See text for additional details.
12
Online Appendix Figure 2BRegression Kink Results
Test of Quasi-Random Assignment
01000
2000
3000
4000
Num
ber
of B
orr
ow
ers
200 400 600 800 1000 1200 1400 1600 1800 2000Pay
$300 Cap $500 Cap
Notes: This figure plots the number of borrowers and biweekly pay for first-time payday borrowersin our regression kink sample. The sample consists of borrowers living in states offering paydayloans in $1 or $10 increments who are paid biweekly or semimonthly and earning more than $100and within $1000 of a kink point. The smoothed line controls for a seventh-order polynomial inpay interacted with being eligible for the maximum loan size in a state. See text for additionaldetails.
13
Online Appendix Figure 3Regression Discontinuity Falsification Test of First Stage
A. Polynomial B. Linear Spline
50
10
01
50
20
02
50
30
0
Lo
an
Am
ou
nt
200 400 600 800 1000
Pay
50
10
01
50
20
02
50
30
0
Lo
an
Am
ou
nt
200 400 600 800 1000
Pay
C. Local Lineaer
-2-1
01
2R
esid
ua
lize
d L
oa
n A
mo
un
t
-50 -25 0 25 50
Pay Relative to Loan Eligibility
Notes: These figures plot average loan size and biweekly pay for first-time payday borrowers ina sample of states where no effect is expected. The sample consists of borrowers living in statesoffering payday loans in $1 or $10 increments who are paid biweekly or semimonthly between$100 and $1100. The smoothed line in Figure A controls for a seventh-order polynomial in netpay. Figure B controls for a linear spline in net pay. Figure C stacks data from each cutoff andcontrols for net pay using a linear regression and a linear regression interacted with the loan cutoff.See text for additional details.
14
Online Appendix Figure 4Regression Discontinuity Falsification Test of Main Results
A. Polynomial B. Linear Spline
.1.1
5.2
.25
Fra
ctio
n D
efa
ult
200 400 600 800 1000
Pay
.1.1
5.2
.25
Fra
ctio
n D
efa
ult
200 400 600 800 1000
Pay
C. Local Linear
-.0
4-.
02
0.0
2.0
4R
esid
ua
lize
d F
ractio
n D
efa
ult
-50 -25 0 25 50
Pay Relative to Loan Eligibility
Notes: These figures plot average default and biweekly pay for first-time payday borrowers in asample of states where no effect is expected. The sample consists of borrowers living in statesoffering payday loans in $1 or $10 increments who are paid biweekly or semimonthly between$100 and $1100. The smoothed line in Figure A controls for a seventh-order polynomial in netpay. Figure B controls for a linear spline in net pay. Figure C stacks data from each cutoff andcontrols for net pay using a linear regression and a linear regression interacted with the loan cutoff.See text for additional details.
15