Jonathan Morduch New York University February 26, 2013 World Bank FPD Chief Economist s Talk

Transcription

1 Jonathan Morduch New York Unversty February 26, 2013 World Bank FPD Chef Economst s Talk

2 An emprcst s reflecton on theory Some lessons on emprcal foundatons of fnancal ncluson What does the data mply about theory?

3 2,500,000,000 adults Half the world Sources: Chaa, et al, Half the World s Unbanked. Ch. 2 Bankng the World. Based on Honohan. Demrgüç-Kunt and Klapper, Global Fndex.

4

5 Mllons of mcrofnance customers Mcrocredt Summt Campagn, total poorest

6 3 observatons 1. Inclusve fnance s stll too exclusve 2. Delverng mcrofnance s dffcult 3. Rsk s central

7 Inclusve fnance s stll too exclusve

8 A model of moral hazard Borrower s an entrepreneur Compettve market IC: Borrower needs ncentve to work hard Lmted lablty: Borrower has lmted assets for collateral. Rsky choces An optmal loan contract for an entrepreneur wth assets A allocates project clams x n the success (=1) and falure (=0) states. The entrepreneur keeps returns s, and the lender collects repayments x s R. The lender maxmzes the entrepreneur s expected returns to ensure hgh effort by borrowers: max E[ s p] s. t. s E[ x s p] I (1) E[ s p] E[ s q] B (2) x s x A for 0,1 (3) where the expected return when the entrepreneur works hard s E[ s p] ps1 (1 p) s0.

9 Implcaton Provdng optmal ncentves requres that borrower keeps lmted lablty rents. Snce no collateral, can t easly use stcks. So nstead need carrots to ensure borrower takes less rsky choces. May not be possble to do so under competton.

10 A model of moral hazard Borrower s an entrepreneur Compettve market IC: Borrower needs ncentve to work hard Lmted lablty: Borrower has lmted assets for collateral. Rsky choces An optmal loan contract for an entrepreneur wth assets A allocates project clams x n the success (=1) and falure (=0) states. The entrepreneur keeps returns s, and the lender collects repayments x s R. The lender maxmzes the entrepreneur s expected returns to ensure hgh effort by borrowers: max E[ s p] s. t. s E[ x s p] I (1) E[ s p] E[ s q] B (2) x s x A for 0,1 (3) where the expected return when the entrepreneur works hard s E[ s p] ps1 (1 p) s0.

11 Are most loans for busness nvestment? Evdence from Indonesa Don Johnston and Jonathan Morduch, The Unbanked: Evdence from Indonesa. World Bank Economc Revew 2008.

12 Are most loans for busness nvestment? Evdence from Mongola About half of all mcrocredt busness loans used for household ends: 1. Purchases of household assets such as rados and large domestc applances. 2. Payng down more expensve loans. 3. Help smooth seasonal ups and downs of consumpton. Attanaso, Orazo, Brtta Augsburg, Ralph De Haas, Emla Ftzsmons, Heke Harmgart (2011). Group lendng or ndvdual lendng? Evdence from a randomsed feld experment n Mongola. European Bank for Reconstructon and Development, workng paper. December.

13 Are most loans for busness nvestment? Evdence from Peru Dean Karlan and Jonathan Znman. "Lst Randomzaton for Senstve Behavor: An Applcaton for Measurng Use of Loan Proceeds. Journal of Development Economcs, May 2012.

14 Grameen Bank Innovaton: Loan top-ups Grameen Fnancal dares (Stuart Rutherford, ) Roughly half of loans are used n "productve" ways Ramna s use of top-ups Loan # Date Amount Use $83 Food and stocks 1 st Top up Aprl 2003 N/A Gran for comng monsoon season 2 nd Top up October 2003 $67 Funeral expenses 3 rd Top up May 2004 N/A Pay down prvate loan 4 th Top up December 2004 $75 Stocks of gran, medcal treatment 5 th Top up June 2005 $65 School fees, restock food Source: Collns, et al. Portfolos of the Poor. Prnceton, See also Collns, Gong wth the flow. Ch. 5 n Bankng the World.

15 Global Fndex

16 Fnance and Hunger Stjn Claessens and Erc Fejen, ch. 7, Bankng the World 1% ncrease n prvate credt / GDP reduces undernutrton rate by % Broader evdence: Indrect mechansms may matter as much as drect.

17 Fungblty A busness loan dverted to consumpton could free up other resources for busness nvestment later. So could stll see a postve net mpact on busness and ncome. That s the pont: households need general use fnance In part for self-employment and n part for other prortes.

18 Takng consumer fnance serously Allows mcro-lenders to serve people wth jobs hosptal orderles, nannes, cooks, factory workers, drvers, agrcultural laborers, constructon workers, clerks, craft workers, and others. Allows mcro-lenders to (openly) meet the wder needs of entrepreneurs.

19 The Economc Impact of Expandng Access to Fnance n Mexco Mram Bruhn and Inessa Love Ch. 6, Bankng the World 2002: Grupo Elektra gets a bankng lcense Banco Azteca opens n 815 Elektra stores. 2007: The Ugly Sde of Mcrolendng, Busness Week. Muncpaltes wth Banco Azteca and other bank Branches

20 Installment lendng 53 52

21 A Banco Azteca Customer Blanca Guevara, lves n a workng-class neghborhood of Mexco Cty. 48-year-old wdow. Appled to buy a sewng machne for 2,174 pesos (about $210) on credt at Elektra. Guevara sad she could pay 41 pesos a week over 53 weeks, after a 141-peso down payment. But as an off-the-books house cleaner, she had no proof of ncome. Loan approved once mother-n-law agreed to back the loan. Source: Bruhn and Love, Ch. 6, Bankng the World

22 Grupo Elektra s loan portfolo (mllons of pesos) Grupo Elektra s Loan Portfolo wthn Mexco 12,000 10,000 8,000 6,000 4,000 2,000 - Source: Bruhn and Love, Ch. 6, Bankng the World Quarters

23 Coeffcents on Azteca Muncpaltes n Dec 2002*Quarter Dummes Effect of Banco Azteca Openng on Savngs Accounts 16,000 14,000 12,000 10,000 8,000 6,000 4,000 2, ,000-4,000 Quarter Source: Bruhn and Love, Ch. 6, Bankng the World

24 Average Infomal Busness Dummy Average of Informal Busnesses Owner Dummy for Muncpaltes wth and wthout Banco Azteca Quarter Muncpaltes wthout Azteca Muncpaltes wth Azteca Source: Bruhn and Love, Ch. 6, Bankng the World

25 Effect of Azteca Openng on Income Source: Bruhn and Love, Ch. 6, Bankng the World

26 Effect of Banco Azteca Openng on Wage Earners Source: Bruhn and Love, Ch. 6, Bankng the World

27 Effect of Banco Azteca Openng on Employment Source: Bruhn and Love, Ch. 6, Bankng the World

28 Sum so far Demand for general use funds Desre to turn small cash flows nto usefully large sums [Rutherford] Lenders lend aganst household cash flows Innovatve credt-worthness measures? E.g., Karlan, Mullanathan, and Robles on psychometrc measures, ch. 11, Bankng the World.

29 Delverng mcrofnance s dffcult

30 Lender as contract desgner, enforcer Borrower s an entrepreneur Compettve market IC: Borrower needs ncentve to work hard Lmted lablty: Borrower has lmted assets for collateral. Rsky choces An optmal loan contract for an entrepreneur wth assets A allocates project clams x n the success (=1) and falure (=0) states. The entrepreneur keeps returns s, and the lender collects repayments x s R. The lender maxmzes the entrepreneur s expected returns to ensure hgh effort by borrowers: max E[ s p] s. t. s E[ x s p] I (1) E[ s p] E[ s q] B (2) x s x A for 0,1 (3) where the expected return when the entrepreneur works hard s E[ s p] ps1 (1 p) s0.

31 Grameen Bank, Bangladesh

32 Cost per Borrower Innovaton to reduce cost per customer % 100% 200% 300% 400% 500% 600% 700% 800% Avg Loan Balance per borrower/gni per capta Cull, Demrgüç-Kunt, and Morduch, Journal of Economc Perspectves, 2009

33 OpEx / Loan Portfolo But small transacton szes mean hgh cost per unt transacted Graph 1 60% 40% 20% 0% 0% 100% 200% 300% 400% 500% 600% 700% 800% -20% Avg Loan Balance per borrower/gni per capta Cull, Demrgüç-Kunt, and Morduch, Journal of Economc Perspectves, 2009

34 Yeld on gross portfolo (nomnal) Response: rase prces on the low-end 60% 50% 40% 30% 20% 10% 0% 0% 100% 200% 300% 400% 500% 600% 700% 800% Avg Loan Balance per borrower/gni per capta Cull, Demrgüç-Kunt, and Morduch, Journal of Economc Perspectves, 2009

35 Two layers of moral hazard Investors Wll lenders be dlgent and effcent? Lenders Wll customers be prudent? Customers

36 A model of moral hazard LENDER Borrower s an entrepreneur Compettve market IC: Borrower LENDER needs ncentve to work hard Lmted lablty: Borrower LENDER has lmted assets for collateral. Rsky choces An optmal loan contract for an entrepreneur wth assets A allocates project clams x n the success (=1) and falure (=0) states. The entrepreneur keeps returns s, and the lender collects repayments x s R. The lender maxmzes the entrepreneur s expected returns to ensure hgh effort by borrowers: max E[ s p] s. t. s E[ x s p] I (1) E[ s p] E[ s q] B (2) x s x A for 0,1 (3) where the expected return when the entrepreneur works hard s E[ s p] ps1 (1 p) s0.

37 Implcatons Even f solve lender-borrower moral hazard, stll have a fundamental ncentve ssue Management of small-scale transactons, rsk, etc. Lmted lablty rents for lenders Implcatons for leverage

38 Leverage Insttutons lendng to customers wth more collateral can acheve greater leverage (all else equal) A dollar to nsttutons on the hgh end may yeld a greater multpler A dollar to nsttutons on the lowest end may brng them nto exstence Assets (collateral) of borrowers Source: Connng and Morduch, Annual Revew of Fnancal Economcs, 2011.

39 Vllage Savng and Loan Assocaton, Tanzana

40 Make rsk more central

41 A model of moral hazard Borrower s an entrepreneur Compettve market IC: Borrower needs ncentve to work hard Lmted lablty: Borrower has lmted assets for collateral. Rsky choces An optmal loan contract for an entrepreneur wth assets A allocates project clams x n the success (=1) and falure (=0) states. The entrepreneur keeps returns s, and the lender collects repayments x s R. The lender maxmzes the entrepreneur s expected returns to ensure hgh effort by borrowers: max E[ s p] s. t. s E[ x s p] I (1) E[ s p] E[ s q] B (2) x s x A for 0,1 (3) where the expected return when the entrepreneur works hard s E[ s p] ps1 (1 p) s0.

42 Moral hazard n credt markets Ingredents 1. Borrower s actons are not observed. 2. Borrower needs fnancng for a rsky ncome-generaton project. Expected return s greater than opportunty cost. So gettng the loan ncreases productve effcency. 3. Borrower lacks collateral so can t bear all the rsk herself. 4. Nonnsurable rsk means lender can t determne borrower s actons ex post. 5. Compettve market, wthout subsdy.

43 Cape Town, South Afrca Funerals affected 81% of households n South Afrca. Fre and loss of property affected 19% of Dary households n Bangladesh Illness affected 50% of Dary households n Bangladesh and 42% of Dary households n Inda Source: Portfolos of the Poor

44 Ex post moral hazard One theoretcal depcton: Take the money and run A better depcton? Crss happened and my prorty s now my famly s welfare. Not clearly a moral hazard problem, nor a market falure. Stll banks wll be reluctant to lend. Need for flexblty Collateral may be a lablty for households and welfare maxmzaton Need data on trggers of default

45 Bankng the world Very large gaps n fnancal access Could use much bgger deas And to step out of conceptual boxes Entrepreneural captal? Emphasze money management, rsk management Need an economcs of fnancal organzatons at low-end of the market