Commercial Microfinance and Household Access to Finance

Transcription

1 Commercial Microfinance and Household Access to Finance Martin Brown with Benjamin Guin and Karolin Kirschenmann EBRD, 6 th September

2 Commercialization of microfinance Deposit taking Financial sustainability. but mission drift? Empirical evidence: commercial versus non-commercial MFI s In emerging markets the comparison between microfinance banks and (foreign-owned) retail banks is more relevant! 2

3 Research question & methodoloy To what extent does commercial microfinance expand households use of financial services?. beyond that of ordinary retail banks Study the expansion of the branch network of ProCredit Bank in South-East Europe between 2006 and Focus on locations where retail banks already operate in 2006 Examine whether the number and type of banked households change in regions where ProCredit opens a branch.. compared to where ProCredit does not open a branch 3

4 Data: Household use of bank services Life in Transition Survey, 2006 & 2010 Albania, Bulgaria, Macedonia, Serbia primary sampling units 20 households per survey wave Repeated cross-section Does any household member have a bank account? (1=yes, 0=no) Socioeconomic characteristics Geocodes of PSU match with indicator of local economic activity match with location of bank branches 4

5 Share of households with a bank account 0.8 «Does any household member have a bank account?» 1=yes, 0=no) Albania Bulgaria Macedonia Serbia 5

6 Household locations Bulgaria Primary Sampling Units (PSU) 2006 (50) 2010 (50) 6

7 Data: Bank locations BEPS 2012 Geo-codes of bank branches, 2006 & 2010 Manual check based on annual reports of banks Which banks? ProCredit 5 major retail banks 1 Placebo bank! independent of whether we observe households in 2006 or 2010 we know which banks were nearby in 2006 & 2010! 7

8 Bank branches Bulgaria - ProCredit branches in 2006 and 2010 branch in 2006 & 2010 (36) branch in 2010 only (51) branch in 2006 only (8) 8

9 Local economic activity Henderson et al. (AER 2012) show that satellite based nighttime light data are a good proxy for economic activity. We match each PSU location with annual nighttime light data for (source: 9

10 ProCredit branch expansion Bank close = PSU is within 5km of a bank branch Retail bank close in 2006 ProCredit close in 2006 No Yes No Yes In 100 PSU retail banks are present in 2006 and ProCredit is not ProCredit locates in 54 of these PSU between 2006 and 2010! 10

11 Hypotheses Location effect: Given the presence of a retail bank, ProCredit is more likely to open branches in regions with a higher share of low-income households. Volume effect: In regions where ProCredit opens a branch the share of households with a bank account increases more than in regions where ProCredit does not open a branch Composition effect: The increase in bank account use is stronger among low-income households than among high-income households. (model based on Mulligan & Sala-Martin, JPE 200) 11

12 Location effect: Univariate results ProCredit close in 2010 (n=54) ProCredit not close in 2010 (n=46) Difference (n=100) Share of low-income households ** (.04) Nightlight *** (2.84) Nightlight (0.584) Sample: 100 PSUs close to a retail bank in 2006 but not close to ProCredit in

13 Economic activity and branch location Sample: PSU where retail banks are present in 2006 and 2010 Nightlight year ProCredit close in 2006 ProCredit close in 2010 & not close in 2006 ProCredit not close in 2010 & not close in

14 Results: Volume effect Dependent variable: Household has a bank account (1=yes, 0=no) LITS Wave Difference Difference in Difference ProCredit close in (N=617) (N=440) 0.301*** (0.030) 0.085* (N=1057) (0.044) ProCredit not close in (N=398) (N=499) 0.216*** (N=1954) (0.032) (N=897) Sample: 1954 Households in PSUs located close to a retail bank in 2006 & 2010 but not close to ProCredit in

15 Multivariate Diff-in-Diff results Volume effect: Controlling for household characteristics (income) the ProCredit effect increases from 8.5 to 16.2 percentage points Controlling additionally for PSU characteristics (nightlight) the ProCredit effect increases from to 19.1 percentage points Composition effect ProCredit effect is estimated at 21 percentage points for lowest and medium income terciles ProCredit effects is estimated at 13 percentage points (not significant) for the highest income tercile. 15

16 Robustness tests Thresholds for bank proximity: 10km instead of 5km Placebo bank Similar size of the branch network in 2006 as ProCredit. Similar expansion of the branch network as ProCredit. Subsidiary of a foreign bank Dependent variable placebo tests Car ownership Computer at home Internet connection 16

17 Summary and conclusion ProCredit opens branches in areas with high economic activity... but also a high share of low-income households New ProCredit branches are associated with a substantial increase in the share of banked households This effect is more pronounced among low-income than high income households This microfinance bank expands the frontier of finance 17

18 Extra slides 18

19 Volume Effect: Multivariate Dependant variable Household has a bank account (1=yes) LITS wave *** [0.032] [0.072] [0.080] ProCredit close in [0.031] [0.066] [0.057] LITS wave 2010 *ProCredit close in * 0.162* 0.191** [0.044] [0.094] [0.088] Household controls NO YES YES PSU-level controls NO NO YES Country fixed effects NO NO YES Observations 1,954 1,954 1,954 R-squared Method OLS OLS OLS Household: controls: Expenditures, Self employed, Wage income, Female, University degree, Household size, Age, Language, Muslim, Car, Computer, Mobile phone, Internet PSU-level controls: Nightlight, number of retail banks, share of low-income households, share of middle-income households, 19

20 Composition effect Dependant variable Household has a bank account (1=yes) Low-income households Middle-income households High-income households LITS wave [0.085] [0.111] [0.102] ProCredit close in [0.055] [0.061] [0.088] LITS wave 2010* 0.212** 0.208** ProCredit close in 2010 [0.099] [0.103] [0.114] Household controls YES YES YES PSU-level controls YES YES YES Country fixed effects YES YES YES Observations R-squared Method OLS OLS OLS 20

21 Commercial microfinance vs. Retail banking ( ) a functioning and inclusive financial system makes a contribution to a country's development. Stimulating and accelerating such processes takes time. It requires stable institutions whose owners take a long-term view and whose staff are well-trained and can win customers' trust. This is precisely the focus of our efforts. Source: ( ) We seek long-term customer relationships. We provide contemporary financial services to our customers meeting the highest professional standards and effectively satisfying the customers' needs. We seek to be seen as a friendly and constructive partner for our customers and we are pro-active and quick in delivering our services. Source: 21

22 Appendix Banks in Albania Ranking (branches) Bank Branches in 2006 (EBRD data) Branches in 2010 (EBRD data) Branches current (EBRD data) 1 Raiffeisen Bank Albania Banka Kombetare Tregtare Sh.s. n/a n/a 59 3 Tirana Bank SA-Banka e Tiranes Sha n/a n/a 53 4 Credins Bank Sh.A Alpha Bank - Albania (AB) n/a n/a 43 6 Banka Popullore Sh.A (Societe Generale Albania) ProCredit Bank (Albania) Sh.A Intesa Sanpaolo Bank Albania Union Bank Sh.a National Bank of Greece Emporiki Bank - Albania SA n/a n/a Veneto Banka sh.a. (VB) International Commercial Bank-Banka Tregtare Nderkombetare

23 Appendix Banks in Bulgaria Ranking (branches) Bank Branches in 2006 Branches in 2010 (EBRD data) (EBRD data) 1 UniCredit Bulbank AD n/a n/a United Bulgarian Bank - UBB Eurobank EFG Bulgaria AD (Postbank) n/a n/a Raiffeisenbank (Bulgaria) EAD Societe Generale Expressbank NLB Banka Sofia AD Investbank Plc n/a n/a 117 Branches current (EBRD data) Allianz Bank Bulgaria AD-CB Allianz Bulgaria AD CIBANK PLC n/a n/a Alpha Bank Municipal Bank Plc Piraeus Bank Bulgaria AD International Asset Bank AD ProCredit Bank (Bulgaria) AD MKB Unionbank AD n/a n/a D Commerce Bank AD n/a n/a Corporate Commercial Bank AD Tokuda Bank Teximbank-PEB Texim AD n/a n/a Emporiki Bank - Bulgaria EAD n/a n/a

24 Appendix Banks in Macedonia Ranking (branches) Bank Branches in 2006 (EBRD data) Branches in 2010 Branches current (EBRD data) (EBRD data) 1 Stopanska Banka AD Skopje n/a n/a 65 2 Komercijalna Banka AD Skopje NLB Tutunska banka AD Skopje Procredit Bank AD Skopje Sparkasse Bank AD Skopje n/a n/a 31 6 TTK Banka AD Skopje Ohridska Banka AD Ohrid Centralna Kooperativna Banka AD Skopje n/a n/a 28 9 Halkbank AD Skopje n/a n/a Eurostandard Bank AD Skopje n/a n/a Kapital Banka AD Skopje Ziraat Bank AD Skopje Macedonian Bank for Development Promotion AD Skopje

25 Appendix Banks in Serbia Ranking (branches) Bank Branches in 2006 Branches in 2010 (EBRD data) (EBRD data) 1 Komercijalna Banka A.D. Beograd Banca Intesa ad Beograd Alpha Bank Srbija AD n/a n/a Vojvodjanska Banka n/a n/a 119 Branches current (EBRD data) Eurobank EFG Stedionica AD Beograd Societe Generale Banka Srbija ad n/a n/a Metals Banka Ad Novi Sad Raiffeisenbank a.d Credit Agricole Srbija a.d. Novi Sad n/a n/a UniCredit Bank Serbia JSC-UniCredit Bank Srbija a.d 11 KBC Banka ad n/a n/a AIK Banka ad Nis n/a n/a ProCredit Bank Serbia Postal Savings Bank J.S.C.-Postanske Stedionice, a.d. n/a n/a NLB BANKA n/a n/a Credy banka AD n/a n/a

26 Appendix MFIs in Albania Bank Gross Loan Portfolio Active borrowers ProCredit Bank - ALB 229,379,635 BESA 47,280,837 22,894 NOA - ALB 39,773,373 15,266 ASC Union 35,547,681 15,559 FAF-DC 21,585,482 4,225 VisionFund Albania 2,045,369 2,381 Note: Credins bank left out Note: Mixmarket 26

27 Placebo bank Location effect Placebo bank not close in 2006 Characteristics of PSU Nightlight 2006 Placebo bank close in 2010 (n=71) Placebo bank not close in 2010 (n=59) Difference (n=130) *** (2.602) D.Nightlight (1.088) Share of Low income households (0.035) Sample: PSU that are close to a Retail bank branch in 2006 &

28 Placebo bank - Volume effect ProCredit close in 2010 ProCredit not close in Difference (N=690) (N=608) (N=676) (N=540) Difference in Difference (DiD) 0.219*** (0.026) ** (N=1366) (0.038) 0.303*** (N=2514) (0.028) (N=1148) Sample: PSUs close to Retail banks in 2006 & 2010 and not close to Placebo bank in

29 Placebo bank Composition effect Subsample PSUs close to Retail banks in 2006 & 2010 and not close to Placebo bank in 2006 Low income households Middle income households High income households Dependant variable Account Account Account Account Account Account Time 0.354*** 0.311*** *** ** [0.085] [0.069] [0.087] [0.088] [0.085] [0.081] ProCredit close in [0.065] [0.063] [0.057] [0.063] [0.078] [0.073] Time*ProCredit close in * [0.109] [0.076] [0.095] [0.087] [0.110] [0.097] Household Controls YES YES YES YES YES YES Average PSU characteristics NO YES NO YES NO YES Nightlight NO YES NO YES NO YES Change & number of RBs NO YES NO YES NO YES Country fixed effects NO YES NO YES NO YES Observations R-squared Method OLS OLS OLS OLS OLS OLS 29

30 Bank branch expansion, ProCredit Branches Retail Bank Branches

31 Volume effect: Identification PSU s close to Retail bank, but not close to ProCredit in 2006 P(Account) ProCredit close in 2010 ProCredit close in 2010 Effect of interest ProCredit not close in 2010 ProCredit not close in 2010 PSU observed in 2006 PSU observed in