Hyperbolic Discounting and Induced Informal Credit Transactions A

Hyperbolic Discounting and Induced Informal Credit Transactions: A Case of Debit Card Pawning in the Philippines Nobuhiko Fuwa, Waseda University Kei Kajisa, Aoyama Gakuin University Eduardo Lucio, University of Queensland Sharon Faye Piza, World Bank Yasuyuki Sawada, University of Tokyo 1

Preview 1. A significant proportion (1/3) of the factory workers in Laguna are hyperbolic (present-biased) discounters, while 1/5 are future-biased 2. Hyperbolic discounters and future-biased are more likely to borrow from ATM sangla, than are time-consistent discounters, and are likely naïve and sophisticated, respectively 3. Based on ‘luxury’ consumption behavior: present-biased men are naïve (so consume more of Jollibee hamburgers), while present-biased women may be sophisticated enough to eat less than do time-consistent women 2

ATM sangla (debit card pawning) as an induced institutional innovation 1 What is ATM sangla (pawning)? : informal lending with debit card (ATM card) for salary deposit as collateral Informal lending is very common & expanding in the Philippines 3

ATM sangla (debit card pawning) as an induced institutional innovation 2 An example of ‘induced institutional innovation’ (Hayami & Godo 2005), arising from unmet credit demand + new technology (ATM) Why is ATM sangla unique? Unlike existing collateralized loans (e. g. , physical collateral, group liability, savings deposit, etc. ) … Future income stream as collateral with an ATM technology Somewhat similar to ‘payday loan’ in the US, but with potentially unlimited(? ) cash access to the lender 4

ATM sangla (pawning) in the Philippines: The data 1: respondents characteristics 3 small scale factories in an industrial estate in Laguna province (south of Metro Manila) Respondent profile: 320 respondents (60% male) Average age: 30 yrs old, with 7 year work exp. Marital status: 53% married (49% with children) Education: One third with college degree, 45% with vocational or some collage training Average salary: Ph. P 15, 000/month (~ US$350) Type of employment: regular (72%); contractual (23%); probation(5%) • Mode of salary payment: bank deposit (99%); cash (1%) 5

ATM sangla (pawning) in the Philippines: The data 2: ATM sangla transactions Know ATM sangla: 297 (93%) l Ever utilized ATM sangla: 134 (42%): male: 33%; female: 48% l ATM sangla loan balance at the time of interview: 42 (13%) l Average loan: Ph. P 15, 220 (~average monthly salary), for 5. 2 months l Interest rates: 0~20%/month, average 3%/month (~ 40% PA) l Deduction from salary: Ph. P 2, 700 (34% of average (semimonthly) salary) l Purpose of borrowing: medical (21%), daily consumption (19%), education (16%), house (9%), etc. l Sources of ATM sangla: money lenders (54%), colleagues (21%), friends (16%) l 6

Distribution of Preference Type Based on hypothetical questions a la Ashraf, Karlan & Yin (2006) Count Hyperbolic/ Present-biased 110 34. 4% Share (in %) Consistent 139 Patient/ Futurebiased 71 Total 320 43. 4% 22. 2% 100% (Time) consistent discounters: current (now vs 1 mo) discount rate = future(6 mo vs 7 mo) discount rate Hyperbolic (present-biased) discounters: current (now vs 1 mo) discount rate > future(6 mo vs 7 mo) discount rate patient (future-biased) discounters: current (now vs 1 mo) discount rate < future(6 mo vs 7 mo) discount rate • Philippines: Hyperbolic: 28% ; Consistent 52% ; Patient 20% (Ashraf, Karlan & Yin, 2006) • India : Hyperbolic: 33% ; Consistent 57% ; Patient 10% (Bauer, et al. 2012) • US : Hyperbolic: 36% ; Consistent 55% ; Patient 9% (Meier & Sprenger 2010) Gender composition male (total=195) female (total=125) Hyperbolic 31. 3% 39. 2% Time consistent 45. 1% 40. 8% Patient 23. 6% 20. 0% 7

Preference and Borrowing Behavior time consistent hyperbolic Has outstanding balance ATM sangla 8. 6% 16. 4%* bank, coops. , NGOs, MFIs 5. 0% 4. 2% gov't Fis (Soc. Sec. Sys. /Pag-ibig) 46. 0% 43. 7% pawnshop/private money lender 4. 3% 1. 8% relatives/friends/others 13. 6% 21. 8%* Outstanding balance (peso) [average across all obs. ] ATM sangla 1, 200 2, 400 bank, coops. , NGOs, MFIs 7, 080 280 gov't Fis (Soc. Sec. Sys. /Pag-ibig) 29, 960 30, 750 pawnshop/private money lender 530 330 relatives/friends/others 2, 320 1, 930 Outstanding balance (peso) [average across non-zero obs. only] ATM sangla 13, 950 14, 650 bank, coops. , NGOs, MFIs 140, 540 15, 200 gov't Fis (Soc. Sec. Sys. /Pag-ibig) 67, 170 61, 490 pawnshop/private money lender 12, 200 17, 900 relatives/friends/others 16, 940 8, 850* *** : significantly different from the time consistent at 1%; ** at 5%; *: 10% patient 16. 9%* 4. 2% 43. 7% 9. 9% 15. 5% 2, 550 870 13, 180 1, 780* 1, 130 15, 100 20, 670 30, 200 18, 100 7, 270* 8

ATM sangla and time-inconsistent discounters: empirical result 1 ATM sangla appears to expand credit access to hyperbolic discounters, but does it mean: more easy money, possibly leading to over-borrowing (a la Meier & Sprenger 2010)? , OR Can ATM sangla be a commitment device (a la Bauer, Chytilová, and Morduch, 2012)? 9

ATM sangla and time-inconsistent discounters: naïve or sophisticated? Testing for behavioral implications (following Bauer, Chytilová & Morduch, 2012) : Yi = b 0 + b 1 di 0 + b 2 Hi + b 3 Pi+ b 4 Xi + ei (1) Yi = b 0 + b 1 di 1 + b 2 Hi + b 3 Pi+ b 4 Xi + ei (2) • Yi: borrowing from ATM sangla, as well as from other sources: * dummy (1 if nonzero balance from a particular source) • di 0, di 1 : ‘current’ and ‘future’ discount rate dummy, respectively • Hi: hyperbolic discounter (dummy) • Pi: future biased/’patient’ (dummy) • Xi: borrower characteristics: age, age-squared, female (dummy), married (dummy), number of children, living with parents (dummy), regular status, salary amount, college level education (dummy), vocational education (dummy), loan application rejected (dummy) , number of years working 10 (dummy)

ATM sangla and time-inconsistent discounters: naïve or sophisticated? Testing for behavioral implications (following Bauer, Chytilová & Morduch, 2012) : Yi = b 0 + b 1 di 0 + b 2 Hi + b 3 Pi+ b 4 Xi + ei (1) Yi = b 0 + b 1 di 1 + b 2 Hi + b 3 Pi+ b 4 Xi + ei (2) di 0 : ‘current’ (now vs. 1 month) discount rate dummy; di 1 : ‘future’ (6 month vs. 7 month) discount rate dummy; • b 2 in (1): difference between hyperbolic vs high (consistent) discounter H 0: b 2 =0 in (1): naïve hyperbolic (or, sophisticated without commitment device) • b 2 in (2): difference between hyperbolic vs low (consistent) discounter H 0: b 2 =0 in (2): sophisticated with commitment device, b 2 ≠ 0 if naïve (Time) consistent discounters: current (now vs 1 mo) discount rate = future(6 mo vs 7 mo) discount rate Hyperbolic (present-biased) discounters: current (now vs 1 mo) discount rate > future(6 mo vs 7 mo) discount rate Patient (future-biased) discounters: current (now vs 1 mo) discount rate < future(6 mo vs 7 mo) discount rate 11

Reduced Form Equation with current [Eq (1)] vs. future [Eq (2)] discount rates controlled 1 (OLS) 　 Dep. Var = Have outstanding loan with (dummy): ATM Sangla banks, Coops NGOs & MFI Gov’ FIs Pawnshop/private Rels/Friends/ (SSS/Pag-ibig) Money Lender Others Eq (1) (current disc. rate controlled) Hyperbolic Preference Patient Preference Current Disc. Rate, Med. Discounter Current Disc. Rate, Low Discounter 0. 0477 (1. 00) 0. 0844* (1. 70) 0. 0608 (1. 22) 0. 0238 (0. 46) -0. 0263 (-0. 99) 0. 00618 (0. 21) 0. 0124 (0. 38) 0. 00481 (0. 15) 0. 0854 (1. 38) -0. 0282 (-0. 52) 0. 0147 (0. 25) 0. 00567 (0. 09) -0. 0406 (-1. 35) 0. 0525 (1. 32) 0. 0634* (1. 86) 0. 0113 (0. 36) 0. 132** (2. 33) -0. 0117 (-0. 22) -0. 0323 (-0. 61) -0. 0517 (-0. 94) 0. 0896 (1. 56) -0. 0287 (-0. 47) -0. 00236 (-0. 03) 0. 0523 (1. 01) 0. 00347 (0. 13) 0. 0303 (0. 67) 0. 0570 (1. 47) 0. 0590* (1. 79) 0. 101* (1. 83) 0. 0122 (0. 22) -0. 0570 (-1. 00) -0. 0924 (-1. 64) Eq (2) (future disc. rate controlled) Hyperbolic Preference Patient Preference Future Disc. Rate, Med. Discounter Future Disc. Rate, Low Discounter 0. 0801* (1. 69) 0. 0759 (1. 42) 0. 0232 (0. 44) 0. 0668 (1. 29) -0. 0151 (-0. 70) -0. 00104 (-0. 03) 0. 0188 (0. 49) 0. 00906 (0. 36) Additional covariates: age, age-sq, female, married, number of children, living with parents, regular employee, 12 salary, college level, vocational level, rejected for a loan, live in hometown, firm dummy, number of years employed

ATM sangla and time-inconsistent discounters: naïve or sophisticated? Our results are in sharp contrast with Bauer, Chytilová & Morduch (2012)’s results from microfinance borrowers in India: female hyperbolic discounters are sophisticated and use microcredit as a commitment device Hyperbolic discounters among factory workers in Metro Manila are naïve, unlike female hyperbolic discounters in India (but similar to hyperbolic discounters in the US who accumulate credit card debts? ), Or, (Hyperbolic discounters in our sample may be sophisticated, but) ‘ATM Sangla’, unlike microfinance, does not function as a substitute for commitment savings, despite some similarities –e. g. , frequent and regular repayment: ßno group meeting? ßNot continuous borrowing ßLoan proceeds not used for investment purposes 13

ATM sangla and future-biased discounters? 1 ‘Future-biased’ discounters: borrow more from ATM sangla, compared to time-consistent discounters? ‘Future-biased’ have drawn relatively less attention But 10% ~ 20% have been identified as ‘future-biased’ Their behavior has often been found to be not significantly different from time-consistent discounters (e. g. , Ashraf et al, Bauer et al, Meier & Sprenger) Our findings are somewhat different: ‘future-biased’ discounters appear to borrow more from ATM sangla (than do time-consistent discounters) 14

ATM sangla and future-biased discounters? 2 ‘Future-biased’ discounters: borrow more from ATM sangla, compared to time-consistent discounters? (cont’d) There are some theories of ‘future-biased’: e. g. , “anticipal pleasure (or pain)” (= instantaneous pleasure from anticipating future consumption) plan to consume after some delay but then delay again when the planned moment of consumption arrives (Loewenstein 1987) can lead (naïve) future-biased to under-consumption, over -saving, under-borrowing? 15

ATM sangla and future-biased discounters? 3 ‘Future-biased’ discounters: borrow more from ATM sangla, compared to time-consistent discounters? (cont’d) Sophisticated future-biased? : “a sophisticated person is correctly pessimistic about her future behavior” “preemptive overcontrol”: sophistication effect can even outweigh the effect of time-inconsistent preferences “a sophisticated, present-biased person can save more than time-consistent (O’Donoghue & Rabin 1999) For the same reasons, sophisticated ‘future-biased’ discounters could borrow more from ATM sangla (than do time-consistent discounters)…? • Or, time-discounting profile could be S-shaped? ? (future-biased in very-short-run, present-biased in longer-run) 16

Additional results: time-consistency vs. consumption behavior (of luxury goods) ‘luxury goods’: use of smart phone; eating at Jollibee restaurants; accessing Facebook 17

Additional results: time-consistency vs. consumption behavior (of luxury goods) Women tend to eat more frequently at Jollibee hamburger restaurants than do men Present-biased men eat more (than time-consistent men) at Jollibee, but Present-biased women may be sophisticated enough to eat less (than timeconsistent women) at Jollibee Those with outstanding loans from ATM sangla eat less at Jolliebee, while those with loans from relatives/friends/company access Facebook less frequently Having loan balances has some negative income effects (intended or unintended? ) on some ‘luxury’ consumption? 　 Hyperbolic Preference Patient Preference Female* Hyperbolic Preference Has outstanding loans with ATM sangla Has outstanding loans with relative/friends/company Has outstanding loans with banks/cooperatives/NGOs Has outstanding loans with government Fis Has outstanding loans with pawnshop/private money lender Observations Adjusted R-squared Dep. Var = times in a month to go to Jollibee 0. 046( 0. 18) 0. 611 (1. 80)* 0. 730 (2. 14)* 0. 158 (0. 62) 0. 499 (2. 07)** 0. 158 (0. 62) 0. 898 (3. 31)*** -1. 278 (-2. 90)*** 317 0. 111 317 0. 132 0. 176 (0. 69) 0. 890 (3. 24)*** -1. 290 (2. 86)*** -0. 638 (-1. 97)** -0. 294 (1. 10) 0. 416 (0. 82) -0. 196 (-0. 62) 0. 450 (1. 22) 317 0. 134 Additional covariates: age, age-sq, female, married, number of children, living with parents, regular employee, salary, 18 college level, vocational level, rejected for a loan, live in hometown, firm dummy, number of years employed

ATM sangla (pawning) in the Philippines: conclusions 1 A significant proportion (1/3) of the factory workers in Laguna are hyperbolic (present-biased) discounters, while 1/5 are future-biased Hyperbolic discounters (and future-biased) are more likely to borrow from ATM sangla, than are time-consistent discounters Behavioral implication 1: those hyperbolic discounters are possibly naïve, rather than sophisticated (or, sophisticated but have no access to commitment device) the expansion of credit access (easy money) via ATM sangla for hyperbolic discounters in the Philippines may not be welfare enhancing (over-borrowing? ) Behavioral implication 2: Should we take ‘future-biased’ discounters (20%) a bit more seriously? 19

ATM sangla (pawning) in the Philippines: conclusions 2 Present-biased behavior and gender? less clear compared to the previous findings (e. g. , women = sophisticated hyperbolic discounters demanding commitment savings), but we find some evidence: in terms of eating at Jollibee hamburger restaurants, present-biased men are naïve (so consume more), while present-biased women may be sophisticated enough to eat less than do time-consistent women Are present-biased women relatively more aware (than present-biased men) of their own weakness? ? 20

Thank you for your attention! 21

ATM sangla (debit card pawning) as an induced institutional innovation 3 Our focus: Hyperbolic discounting (present-biased) and financial behavior on ATM sangla (pawning): Hypothesis I) commitment device for “sophisticated” hyperbolic discounters/present biased? e. g. , substitute for commitment savings; a la Bauer, Chytilová, and Morduch (2012)) Hypothesis II) additional temptation (over borrowing? ) for naïve hyperbolic discounters (or, financial innovation may reduce welfare by providing “too much liquidity”) 22

Identifying Hyperbolic Discounters (a la Ashraf, Karlan Yin 2006) HYPOTHETICAL QUESTIONS: We would like to ask you about the following hypothetical questions 7 -1. Would you prefer 200 pesos now or 250 pesos in one month? 7 -2. Would you prefer 200 pesos now or 300 pesos in one month? 7 -3. How much would we have to give you in one month for you to choose to wait? 1. Php 200 now 1. Php 250 in one month GO TO 7 -4 Php 200 now 2. Php 300 in one month GO TO 7 -4 2. Php ___________ 7 -4. Would you prefer 200 pesos on six 1. Php 200 in 6 months or 250 pesos in seven 2. Php 250 in 7 months GO TO 8 -1 months? 7 -5. Would you prefer 200 pesos in six 1. Php 200 in 6 months or 300 pesos in seven 2. Php 300 in 7 months GO TO 8 -1 months? Consistent : current (now vs 1 mo) discount rate = future(6 mo vs 7 mo) discount rate 7 -6. How much would we have to give you in seventh month for you to choose to Php ___________ Hyperbolic : current (now vs 1 mo) discount rate > future(6 mo vs 7 mo) discount rate wait? Patient : current (now vs 1 mo) discount rate < future(6 mo vs 7 mo) discount rate 23

Characteristics by Preference • • Proportion with higher education Married Average salary Proportion smart phone ownership Proportion Facebook account Frequency of Facebook access Frequency of Jollibee visits Time consistent Hyperbolic Patient 56. 8% 41. 8%** 43. 7%* 57. 6% P 17864 42. 7% P 13046** 57. 7% P 12794* 44. 6% 44. 5% 38. 0% 79. 9% 79. 1% 80. 3% 2. 44 2. 58 2. 00 2. 13 2. 09 2. 13 ***: significantly different from the time consistent at 1% **: significantly different from the time consistent at 5% *: significantly different from the time consistent at 10% 24

Time Preference and ATM Sangla Utilization time consistent hyperbolic Purpose of ATM sangla loan medical 5. 8% 12. 7%** education 7. 2% 5. 5% consumption 7. 2% 16. 4%** social 5. 0% 1. 8% Most recent ATM loan borrowing within 6 mo 27. 1% 47. 1%** more than 1 year ago 58. 3% 29. 4%*** Source of ATM loan private money lender 50. 0% 52. 9% co-workers 18. 8% 21. 6% friends 16. 7% 17. 7% Among those who have NOT used ATM sangla: potential source and intention Proportion intending to borrow private money lender co-workers friends *** patient 11. 3% 7. 0% 12. 7% 2. 8% 37. 1% 51. 4% 60. 0% 22. 9% 11. 4% of borrowing 30. 8% 40. 7% 36. 1% 42. 9% 17. 9% 10. 7% 12. 5%** 25. 0% 41. 7%*** 30. 8% 23. 1% : significantly different from the time consistent at 1%; ** at 5%; *: 10% 25

Reduced Form Equation of Borrowing 1 (OLS) Dep. Var = Have outstanding loan with (dummy): 　 　 ATM Sangla Banks, Cooperatives NGOs and MFI Government FIs (SSS/Pag-ibig) Pawnshop/ Private Money Lender Rels/Friends/ Company/ Others Hyperbolic Preference 0. 0749* (1. 71) -0. 0207 (-1. 07) 0. 0920* (1. 67) -0. 0126 (-0. 49) 0. 116** (2. 21) 0. 0884* (1. 76) 0. 00367 (0. 11) -0. 000125 (-0. 27) -0. 00385 (-0. 09) -0. 0747 (-1. 17) 0. 00247 (0. 09) -0. 0778* (-1. 89) 0. 132** (2. 44) -0. 0138** (-2. 21) 0. 0101 (0. 17) 0. 00515 (0. 09) 0. 237* (1. 68) 0. 00436 (0. 10) -0. 0645 (-0. 13) 317 0. 122 0. 00701 (0. 25) 0. 00201 (0. 15) -0. 0000314 (-0. 18) -0. 0305 (-1. 18) -0. 0275 (-0. 98) -0. 0000291 (-0. 00) -0. 0367* (-1. 93) 0. 0168 (1. 24) 0. 000211 (0. 05) -0. 0235 (-0. 75) -0. 0359 (-0. 97) 0. 0263 (0. 34) 0. 0125 (0. 64) 0. 0969 (0. 56) 317 0. 148 -0. 0272 (-0. 50) 0. 0544 (1. 36) -0. 000699 (-1. 21) 0. 0301 (0. 57) 0. 0654 (0. 97) 0. 0317 (1. 10) -0. 0507 (-0. 90) 0. 133** (2. 16) -0. 0183** (-2. 26) -0. 0668 (-0. 95) 0. 00532 (0. 08) 0. 150* (1. 67) -0. 0177 (-0. 31) -0. 868 (-1. 43) 317 0. 463 0. 0562 (1. 39) 0. 0169 (0. 71) -0. 000272 (-0. 90) 0. 0242 (0. 84) 0. 0710* (1. 87) -0. 0106 (-0. 54) -0. 0535** (-2. 21) 0. 0115 (0. 35) -0. 00364 (-1. 30) 0. 0471 (1. 30) 0. 0434 (1. 44) 0. 116 (1. 14) 0. 0158 (0. 51) -0. 292 (-0. 78) 317 0. 057 Patient Preference Age^2 Female Married Number of Chidren Living with Parents Regular Employee Amount of Salary Vocational Education At least 1 st year College degree Ever been rejected for a loan Hometown prov. same as Currrnt Residence prov. Constant 　 Observations Adjusted R-squared -0. 0153 (-0. 29) 0. 105*** (2. 65) -0. 00157*** (-3. 05) -0. 0560 (-1. 32) 0. 0598 (0. 92) 0. 0496* (1. 79) 0. 0190 (0. 35) -0. 0372 (-0. 56) -0. 00994 (-1. 56) -0. 0576 (-0. 85) -0. 0604 (-0. 97) 0. 225* (1. 90) 0. 0194 (0. 36) -1. 343** (-2. 17) 317 26 0. 100

ATM Sangla and Time-Inconsistent Discounters: empirical spec. 1 Base specification: Yi = b 0 + b 1 Hi + b 2 Pi + b 3 Xi + ei, (1) Yi: borrowing from ATM sangla, as well as from other sources: * dummy (1 if nonzero balance from a particular source) * amount of loan balance from a particular source Hi: hyperbolic discounter (dummy) Pi: future biased/’patient’ (dummy) Xi: borrower characteristics: age, age-squared, female (dummy), married (dummy), number of children, living with parents (dummy), regular status, salary amount, college level education (dummy), vocational education (dummy), loan application rejected (dummy) , number of years working (dummy) 27

Reduced Form Equation of borrowing 2 (OLS) Dep. Var = Outstanding loan balance (in pesos) with: 　 　 ATM Sangla Banks, Cooperatives NGOs and MFI Government FIs (SSS/Pag-ibig) Pawnshop/ Private Money Lender Rels/Friends/ Company/ Others Hyperbolic Preference 1. 591 (1. 53) -0. 627 (-0. 83) 10. 43 (0. 66) 0. 0744 (0. 16) 0. 502 (0. 44) 1. 691 (1. 47) 0. 0247 (0. 05) -0. 00281 (-0. 41) 0. 114 (0. 14) -0. 0101 (-0. 01) -0. 249 (-0. 40) -1. 592** (-2. 10) 1. 874* (1. 81) -0. 157 (-1. 36) -0. 188 (-0. 17) 0. 377 (0. 44) 2. 864 (0. 74) 0. 166 (0. 17) -0. 423 (-0. 05) 317 0. 049 0. 130 (0. 13) -0. 573 (-0. 86) 0. 00997 (0. 91) -1. 541 (-0. 96) -0. 428 (-0. 42) -0. 546 (-0. 72) -1. 521 (-1. 31) -0. 355 (-0. 48) -0. 173 (-0. 65) -0. 0853 (-0. 09) 1. 911 (0. 78) 0. 929 (0. 60) 3. 054 (1. 01) 8. 478 (0. 82) 317 0. 860 -21. 96 (-1. 22) -39. 63 (-1. 05) 0. 712 (1. 14) -18. 24 (-0. 95) 40. 91 (1. 65) -33. 44 (-1. 46) -22. 54 (-0. 86) 19. 09 (1. 09) 5. 901 (1. 07) -10. 82 (-0. 56) 7. 009 (0. 38) 14. 28 (0. 61) 4. 165 (0. 16) 492. 7 (0. 96) 317 0. 122 1. 697* (1. 66) -0. 239 (-0. 72) 0. 00173 (0. 42) 0. 256 (0. 54) 0. 446 (0. 71) -0. 162 (-0. 35) -1. 215** (-2. 57) 1. 003 (1. 16) -0. 0153 (-0. 31) 0. 328 (0. 42) 0. 306 (0. 50) -0. 583 (-0. 33) 0. 567 (0. 80) 3. 475 (0. 70) 317 0. 006 -1. 568 (-1. 36) 0. 880 (1. 63) -0. 0153** (-2. 01) -0. 596 (-0. 91) 1. 999* (1. 76) 0. 412 (0. 89) 0. 00767 (0. 01) -0. 118 (-0. 12) 0. 0154 (0. 07) -1. 198 (-1. 14) -0. 000132 (-0. 00) 5. 454 (1. 32) 0. 267 (0. 26) -11. 32 (-1. 33) 317 0. 057 Patient Preference Age^2 Female Married Number of Chidren Living with Parents Regular Employee Amount of Salary Vocational Education At least 1 st year College degree Ever been rejected for a loan Hometown prov. same as Currrnt Residence prov. Constant 　 Observations Adjusted R-squared 28

Reduced Form Equation with current [Eq (1)] vs. future [Eq (2)] discount rates controlled 1 (OLS) 　 Dep. Var = Have outstanding loan with (dummy): ATM Sangla banks, Coops NGOs & MFI Hyperbolic Preference 0. 0749* (1. 71) -0. 0207 (-1. 07) 0. 0920* (1. 67) -0. 0126 (-0. 49) 0. 116** (2. 21) Patient Preference 0. 0884* (1. 76) 0. 00701 (0. 25) -0. 0272 (-0. 50) 0. 0562 (1. 39) -0. 0153 (-0. 29) 0. 0854 (1. 38) -0. 0282 (-0. 52) 0. 0147 (0. 25) 0. 00567 (0. 09) -0. 0406 (-1. 35) 0. 0525 (1. 32) 0. 0634* (1. 86) 0. 0113 (0. 36) 0. 132** (2. 33) -0. 0117 (-0. 22) -0. 0323 (-0. 61) -0. 0517 (-0. 94) 0. 0896 (1. 56) -0. 0287 (-0. 47) -0. 00236 (-0. 03) 0. 0523 (1. 01) 0. 00347 (0. 13) 0. 0303 (0. 67) 0. 0570 (1. 47) 0. 0590* (1. 79) 0. 101* (1. 83) 0. 0122 (0. 22) -0. 0570 (-1. 00) -0. 0924 (-1. 64) Eq (1) (current disc. rate controlled) Hyperbolic Preference Patient Preference Current Disc. Rate, Med. Discounter Current Disc. Rate, Low Discounter 0. 0477 (1. 00) 0. 0844* (1. 70) 0. 0608 (1. 22) 0. 0238 (0. 46) -0. 0263 (-0. 99) 0. 00618 (0. 21) 0. 0124 (0. 38) 0. 00481 (0. 15) Eq (2) (future disc. rate controlled) Hyperbolic Preference Patient Preference Future Disc. Rate, Med. Discounter Future Disc. Rate, Low Discounter 0. 0801* (1. 69) 0. 0759 (1. 42) 0. 0232 (0. 44) 0. 0668 (1. 29) -0. 0151 (-0. 70) -0. 00104 (-0. 03) 0. 0188 (0. 49) 0. 00906 (0. 36) Gov’ FIs Pawnshop/private Rels/Friends/ (SSS/Pag-ibig) Money Lender Others Additional covariates: age, age-sq, female, married, number of children, living with parents, regular employee, 29 salary, college level, vocational level, rejected for a loan, live in hometown, firm dummy, number of years employed

Reduced Form Equation with future (eq (2)) vs. current (eq (3)) discount rates controlled 2 (OLS) Eq (1)　 Dep. Var = Outstanding loan balance (in pesos) with: Hyperbolic Preference ATM Sangla 1. 591 (1. 53) banks, Coops NGOs & MFI -0. 627 (-0. 83) Gov’ FIs (SSS/Pag-ibig) 10. 43 (0. 66) Pawnshop/private Money Lender 0. 0744 (0. 16) Rels/Friends/ Others 0. 502 (0. 44) Patient Preference 1. 691 (1. 47) 0. 130 (0. 13) -21. 96 (-1. 22) 1. 697* (1. 66) -1. 568 (-1. 36) -4. 240 (-0. 24) -24. 83 (-1. 31) 31. 16 (1. 45) 33. 41* (1. 66) -0. 337 (-0. 64) 1. 653 (1. 63) 0. 958 (1. 33) -0. 158 (-0. 27) 0. 261 (0. 21) -1. 560 (-1. 35) 0. 642 (0. 57) -1. 136 (-1. 29) Eq (2) (current disc. Rate controlled) Hyperbolic Preference Patient Preference Future Disc. Rate, Med. Discounter Future Disc. Rate, Low Discounter 1. 180 (1. 13) 1. 653 (1. 46) 0. 972 (0. 96) -0. 364 (-0. 34) -2. 214 (-0. 96) -0. 143 (-0. 11) 3. 458 (0. 92) 2. 483 (0. 95) Eq (3) (future disc. Rate controlled) 1. 603 0. 642 21. 01 0. 283 0. 593 (1. 45) (0. 63) (1. 22) (0. 58) (0. 52) 1. 644 -1. 641 -36. 11 1. 334 -1. 539 Patient Preference (1. 35) (-0. 61) (-1. 32) (1. 20) (-1. 08) 0. 0746 4. 184 34. 17 0. 770 0. 124 Future Disc. Rate, Med. (0. 07) (0. 94) (1. 06) (1. 09) (0. 09) Discounter 0. 352 1. 497 5. 367 1. 073 -1. 671** Future Disc. Rate, Low (0. 32) (0. 93) (0. 25) (1. 64) (-2. 04) Discounter 30 Additional covariates: age, age-sq, female, married, number of children, living with parents, regular employee, salary, college level, vocational level, rejected for a loan, live in hometown, firm dummy, number of years employed Hyperbolic Preference

Outline ATM sangla (debit card pawning) in the Philippines, as an induced institutional innovation Initial findings on: Description hyperbolic discounters and the transactions Are hyperbolic (present-biased) discounters naïve or sophisticated? ‘future biased’ discounters and the transactions hyperbolic discounters and consumption behavior 31

Utilization of ATM Sangla 1 Do you know ATM Sangla ? Have you borrowed with ATM Sangla? When did you last borrow with ATM Sangla？ average term of borrowing Respondents who answered yes： 297（93%） Respondents who answered yes： 134（42%） （male: 41 (33% of total male); female: 93(48% of total female) Within the last 6 months： 50（37%） Between 6 months and 1 year ago： 23（17%） More than a year ago： 61（46%） Respondents with outstanding balance： 42 Average balance = PHP 14, 578. 88（42 responses; range = PHP 1500~PHP 47600） Individual money lender： 72（54%） colleagues： 28（21%） friends： 21（16%） neighbors ： 8（6%） relatives： 5（4%） 5. 2 months（134 respondents： range =1 week ~ 2 years） Average repayment amount (as share of total salary) average interest rate PHP 2, 702. 0（133 respondents：range PHP 350~ PHP 20, 000） 34. 4% (124 respondents：range = 0. 05%~100. 0%) 3. 02% per month（responses： 134） Average amount borrowed with ATM Sangla PHP 15, 220（134 respondents： range= PHP 1, 000 ~ PHP 100, 000）= equivalent to 1 month average salary（133 responses： range = 0. 07 ~5. 00 months） Outstanding ATM Sangla debt balance (as of the interview date) Sources of ATM Sangla borrowing 32

Utilization of ATM Sangla 2 Usage/purpose of most recent ATM Sangla borrowing Medical expenditure： 28（21%） Living expenses/consumption： 26(19%) Educational expenses： 21（16%） House repair： 12（9%） Social, religious expenses： 11（8%） Motor cycle purchase ： 5（4%） Debt repayment ： 4（3%） Other purposes ：appliances, personal emergencies, leisure, etc. （ total responses： 134） Average amount borrowed by Medical ：PHP 19, 393（responses： 28） usage/purpose Living expenses：PHP 9, 038（responses： 26） education：PHP 16, 476（responses： 21） house repair：PHP 15, 250（responses： 12） social, religious：PHP 17, 727（responses： 11） How much is the maximum amount Average = PHP 32, 954（responses: 316： range = PHP 500~ PHP that you think you can borrow? 100, 000） 33

characteristics of the respondents/factory workers Total number of respondents Sex of respondents Average Proportion of respondents who are married Those with children 320 （company A: 107; Comp. B: 78; Comp. C: 135） male： 195（61%）, female： 125（39%） 30. 0 168 out of 320 （53%） Living with parent(s) Type of employment 81（25%） Regular 229（72%） Probation　16（5%） Contractual　　75（23%） 6. 9 years Ave. number of years employed Level of schooling 157 out of 320（49%）（ave. number of children: 1. 96） High school grad or lower：　　　　　　　67（21%） Vocational schooling (undergrad or grad)：　　97（30%） College undergrad：　　　　　49　（15%） College graduate or higher：　　　　　　　107（33%） 34

Mode of salary payment, access to bank accounts, etc. Mode of salary payment Frequency of salary payment Average salary level（per half month） Bank deposit： 316（99%）; cash： 4（1%） Twice a month： 320（100%） PHP 7543（per half month）（317 responses; no answer = 3） Amount withdrawn on or the day after Pay day Average amount withdrawn: PHP 5583　 Average share of the above amount in total salary： 65% of total Average amount of own allowance PHP 3061 Average share of the above amount in total salary： 48% Own bank account other than Salary account (no interest) only： 234（73%） salary account? Own savings account： 82（26%） Own term-deposit account： 1（0. 4%） Own trust account： 1（0. 4%） Own current(checking) account； 2（0. 6%） Average amount left in salary PHP 52, 821（84 responses） account 35

Utilization of ATM Sangla 3 （Only for those who have never borrowed from ATM Sangla）Why have you not borrowed with ATM Sangla? （Only for those who have never borrowed from ATM Sangla）Do you have any intention to borrow from ATM Sangla in the future? No need: 141（76%） Don’t want to be in debt: 23（12%） High interest rate: 16（9%） Can borrow from relatives with no interest： 4（2%） others：likely to be denied of loan, don’t know a lender, etc. （ total responses: 186） Yes ： 65（35%）; No： 121（65%） 36

Time preference and financial behavior time consistent hyperbolic patient total debt (net) 21, 377 20, 487 4, 022 total debt (gross) 41, 086 35, 676 19, 522 total receivables 3, 658 2, 666 3, 832 bank balance 19, 708 15, 189 15, 501 relatives 31. 7% 38. 2% 40. 8% co-workers 21. 6% 27. 3% 31. 0% friends 22. 3% 30. 9% 23. 9% neighbors 13. 7% 9. 1% 9. 9% lending balance with: ***: significantly different from the time consistent at 1% different from the time consistent at 5% *: significantly different from the time consistent at 10% **: significantly 37

Rank of Informal Money Lender as a Source of Financing Type of Loan Rank Housing Loan Other Real Property Loan (Aside from Residence) 4 th out of 16 Vehicle Loan 5 th out of 10 Appliances/Equipment Loan 4 th out of 9 1 st out of 12 Philippines (% Share) 9. 6 NCR (% Share) 6. 8 AONCR (% Share) 15. 8 36. 5 4 8 28. 3 4. 8 14 47. 2 3. 4 5. 3 Source: 2009 Consumer Finance Survey, Bangko Sentral ng Pilipinas 38