From private to public reputation in microfinance lending: Using a Credit Bureau

1
From private to public reputation in microfinance
lending Using a Credit Bureau

• Alain de Janvry Elisabeth Sadoulet, UC Berkeley
• Craig McIntosh, UC San Diego

2
Outline

• Microfinance revolution and credit bureau
• Three phases in development of microfinance
  industry
• Credit bureau information on group lending
• Guatemala case study Genesis (MFI) and Crediref
  (Credit Bureau)
• Natural experiment first phase (2002-04) -
  Selection effect
• Randomized sessions of information second phase
  (2004-05)

3
Microfinance revolution and credit bureau

• Microfinance revolution Access to loans for
  poor w/o collateral to pledge (Morduch)
• Time delay in transaction asymmetrical
  information
• Potential problem of adverse selection (AS)
  Hard for the MFI to select only good borrowers.
• Potential problem of moral hazard (MH)
  Temptation for borrowers to not give the highest
  priority to repayment of the loan
• --gt Need informal contract enforcement mechanism.

4

• Three phases in development of microfinance
  industry
• Phase 1. Monopolistic lending
• (MFI territories dont overlap)
• AS partly controlled by lending technology with
  very thorough investigation of potential
  borrowers by credit agents,
• and by group lending, where clients select each
  other under joint liability system.
• MH controlled by the perspective of getting
  larger loans and possibly better conditions if
  one repays well (dynamic incentives).

5

• Worked well for well managed MFI
• Gave access to credit to a large segment of the
  population
• For non-profit MFI with social objective,
  incorporation of new/poor clients possible with
  cross-subsidization.
• Yet
• Credit is expensive (expensive technology, and
  small loans)
• No possible graduation as information on
  reputation private to lender.

6
Phase 2. Competitive lending without information
sharing

• (multiple lenders, for profit and not for profit)
• Selection more difficult as clients may be hiding
  their borrowing from other sources (Rise of AS).
• Weakening of dynamic incentives, as clients may
  take several small loans from different lenders
  (Rise of MH).
• Harder to keep the better clients, need to give
  them better terms --gt lowers the possibility for
  cross-subsidization.
• --gt Rising late repayments and defaults

7

• Phase 3. Competitive lending with information
  sharing (positive Credit Bureau)
• Negative sharing of defaults only.
• Positive information on indebtedness as well.
• Hypotheses
• MFI selection AS declines with info sharing on
  client reputation and indebtedness.
• Borrower behavior MH declines, at least for
  those clients that are interested in having a
  good reputation outside of the MFI.

8

• Expected benefits
• For the MFI Increased efficiency of selection
  by credit agents. Increased repayment rates
  (selection and moral hazard).
• For individual borrower Graduation possible
  based on private reputation becoming public, and
  access to more/larger loans (credit ladder).
• For borrowers as a group lower costs of lending
• --gt better terms.

9
Credit bureau information on group lending

• CB reports total loan to group and group
  repayment performance in each individual record
  --gt Reflects strict joint liability rule.
• Hypotheses
• AS effect Increase group-selection incentive
• More effort to select others
• Bad groups eject bad members (voice and loyalty)
• Bad groups abandoned by good members (exit)
• MH effect Incentive to better monitor others and
  provide mutual insurance

10
Guatemala case study Genesis (MFI) and Crediref
(CB)

• (in close collaboration with Edgar Buccaro and
  Adela de Rizzo from Genesis Impresarial, and
  Tomás Rosada, from Landivar University)
• Creation of Crediref in 2002
• Staggered entry of the 39 branches of Genesis
  between March 2002 and January 2003.Clients not
  charged for Crediref consultation until July
  2004. Most did not know/understand Crediref
• Summer 2004 Organize sessions of information
  for groups, with randomization design

11

• Natural experiment first phase (2002-04)
• Most clients did not know/understand Crediref
• ? main effect from the use of Crediref by agents
  for the selection of new clients
• Staggered entry of the 39 branches (in a way that
  is not related to the performance of the clients)
• ? Allows identification of selection effect (from
  any other time effect)

12

• Results
• On individual loans performance
• Strong decline in late repayments among
  individual borrowers (but not of default per se)
• Strong increase in the percentage of borrowers
  that become long-term borrowers (take a second
  loan)
• Strong increase in the loan growth of the repeat
  borrowers
• Initial anti-poor selection followed by reentry
  of good poor
• --gt Large performance gain by selected poor

13

• Results
• On group loans performance
• No measurable effect, despite a relatively high
  use of Crediref in certain branches
• On credit agents activity
• Very large increase in the number of new
  borrowers the credit agents enlisted per month
• Both individual clients and group clients
• Increase in the size of the first loan to new
  clients (richer clients)
• No change in volume of lending to old clients (as
  expected)

14
Loans of clients selected with Crediref
Better
Long-term
15
Aligning the entry dates of the different branches
Proportion of first loans with any late repayment
Before using Crediref
After
16
Dramatic improvement of performance among the
poorest clients
Poorest clients
Less poor clients
17
New clients only selection
18
Larger loans richer clients?
19

• Randomized sessions of information
• second phase (2004-05)
• Sessions of information on Credit Bureau for
  group members
• Sessions emphasized who is sharing information,
  what information is being shared, and how this
  can impact the clients
• Cost of bad public reputation General loss of
  access to future loans
• Benefit from good public reputation Access to
  outside loans (graduation, more loans).
• 163 Communal Banks and 207 Solidarity Groups in
  sessions (5,000 clients), randomly chosen

20

• Two types of group lending
• Communal Banks
• 5-30 members, no collateral or cosigning, pure
  joint liability, very small loans, members
  typically held credit constrained to ensure
  dynamic incentives to repay in spite of
  competition (limited access to other loans w/o
  CB).
• Solidarity Groups
• 2-4 members, some collateral, always use
  cosigners, much weaker joint liability, larger
  loans, borrowers are likely to be unconstrained
  and receiving a product which is more similar to
  what a bank would offer (but more expensive).
• --gt Expect different results analyze separately

21
(No Transcript)
22
Results

• On the performance of Genesis loans
• Within a loan cycle, i.e., with a given group -
  Change in behavior
• In successive loans, with recomposition of the
  groups (selection among themselves)
• On outside borrowing (observed in Crediref
  information)
• Amount of borrowing
• Repayment performance

23
Impact of information within a loan cycle (MH
only)
Large drop in delinquency among SGs for final
payments
No effect on CBs
24
Impact of information across loan cycles (MH
and group AS)
Again, big drop in delinquency for SGs,
none for CBs
25
Who is a problem member? SG
onlyDivorced, female, younger clients
26
AS at work restructuring the group before the
next loanNew clients Less divorced, and less
femaleDropouts Large share of women
27
Changes in outside borrowing
No effect for SG
CB increase in number starting outside loans by
47 and in number of loans by 33 more than the
control groups.
28
Heterogeneity among CB members
The more experienced improved their repayment
performance
The less experienced increased their outside
borrowing .
but their repayments worsened
29
Conclusions Objective and approach

• Objective Credit bureaus are reputation-sharing
  institutions introduced to facilitate selection
  and sustain repayment performance for MFI lending
  under asymmetrical information and competition.
  How does it happen? Who benefits?
• Approach We used a combination of
• Natural experiment with staggered entry (first
  stage) to analyze the role of selection by agents
• Randomized sessions of information to clients of
  group lending (second stage) to analyze the
  clients response
• Administrative data on client records in MFI and
  Credit Bureau

30
Conclusions Results

• MFI gained efficiency through use of Credit
  Bureau in client selection (improved performance
  and increased portfolio per agent)
• Response of SG members to information improve
  repayment performance on Genesis loans, no effect
  on outside loans
• Response of CB members heterogeneous
• No effect on Genesis loans
• Good clients take more loans outside
• Less experienced take more but worsen performance