Identification Strategy: A Field Experiment on Dynamic Incentives in Rural Credit Markets

1

Xavier GineWorld Bank

Jessica GoldbergU. Michigan

Dean YangU. Michigan

Identification Strategy:A Field Experiment on Dynamic

Incentives in Rural Credit Markets

Motivation

• Lending in low-income countries is difficult– Clients typically lack adequate collateral – Lenders have limited information about

creditworthiness of clients

• Problem is worse in agriculture as lenders cannot use microfinance mechanisms

• One key problem is lack of national ID system– Loan defaulters can avoid sanctions by using

different identities– Easier when multiple lenders operate in same area– Lenders respond by limiting the supply of credit

2

What we do

• Fingerprinting helps in future identification, in absence of a national ID system– Helps lenders identify past defaulters (within own

institution and potentially across banks)– Allows lenders to use dynamic incentives

• In this project, we ask: – What is the impact of fingerprinting on loan

repayment?– Is impact heterogeneous across borrower types?– What asymmetric information problems are being

reduced?

• Prospect: may raise lending profitability and encourage lenders to expand rural credit provision

3

Relevant aspects of loans provided

• Malawi Rural Finance Company (MRFC) provides loans to paprika farmers in central Malawi– Dowa, Dedza, Mchinji, Kasungu

• Collaboration with private paprika buyer, Cheetah Paprika Ltd.– Designed input package– Identified farmer groups– Forwarded loan repayment to lender before paying farmer

• Mean loan amount ~MK 17,000 (~US$120) for paprika seeds, fertilizer and chemicals– Farmers specifies loan size by deciding on 1 vs. 2 bags of CAN

fertilizer– Inputs provided in kind, not in cash– 15% deposit

• Formally joint liability, but individual liability in practice

4

Treatment and control groups

• Control group:– Educational module emphasizing importance of credit history

administered• Defaulters can be excluded from future loans• Reliable borrowers can get more and larger loans in future

• Treatment group: – Educational module on credit history (identical to module given to

control group) administered, plus:– Biometric fingerprint collected from all farmers as part of loan

application– Use of fingerprints for unique identification explained– Fingerprint identification demonstrated within group

5

Figure 1: Experimental Timeline

July 2007

August 2007

Sep. 30, 2008

Clubs organized

Baseline survey and fingerprinting begin

November 2007

Loans disbursed

Loans due

September 2007

Baseline survey and fingerprinting end

Follow-up survey

August2008

Fingerprinting

• Aug-Sep 2007

7

Demonstrating fingerprint identification

8

Theory: How do dynamic incentive vary by borrower type?

• Borrowers differ in the probability that production is successful (adverse selection)

• Borrowers can divert the loan amount instead of investing it in production (moral hazard)

• Lender offers a loan amount that can take on two values (depending on the number of fertilizer bags borrowed).

• Assumptions– Limited liability: lender can only seize value of cash crop

produced• No scope for strategic or opportunistic default• When inputs are diverted loan recovery is not possible

– Loan can always be repaid when lower loan amount is taken.

9

Theory: How do dynamic incentive vary by borrower type?

• Without biometric identification, borrowers can obtain a fresh loan even if they have defaulted in the past by simply using a different identity.

– Lenders are forced to offer the same one season contract every period

• When biometric technology is available, the lender has the ability to use dynamic incentives by denying credit to past defaulters.

– Borrowers face a tradeoff between diverting inputs (and jeopardizing chances of a loan in the future) versus ensuring repayment of the current loan (and securing a loan in the future)

10

Summary of theoretical predictions

• Repayment + 0

• Loan size - 0(adverse selection)

• Input diversion - 0(moral hazard)

11

Impact of dynamic incentive, by borrower type

“Worst” “Best”(low p) (high p)

Simple treatment vs. control comparison

12

Treatment Control Difference P-value of difference

Total borrowed (MK) 16,590 17,279 -688.37 0.23

Repayment by Sep. 30

Balance outstanding (MK) 2,262 3,652 -1389.70 0.11

% paid 88% 80% 8% 0.11

Eventual repayment

Balance outstanding (MK) 1,726 2,484 -758.79 0.33

% paid 90% 88% 3% 0.52

Measuring borrower type

• Theory predicts that impact of dynamic incentives will be heterogeneous according to borrower type (probability of success, p)– All effects are smaller the larger the borrower’s probability of

success

• Empirical implementation:– In model, loan repayment rate is monotonic in probability of

success– Take predicted loan repayment rate as proxy for probability of

success

• We create an index of how likely someone is to repay the loan (essentially a credit score):– Run regression of repayment rate on baseline observables, for

control group only– Then predict repayment rate for all borrowers– Determinants of repayment: Locality, age, gender, risk

indicators, performance on previous loans, income volatility, years of experience growing paprika

13

1%3%

6%

8%

3%

6%

9%

7%

18%

39%

0%

5%

10%

15%

20%

25%

30%

35%

40%

45%

less than 10%

10% to 20%

20% to 30%

30% to 40%

40% to 50%

50% to 60%

60% to 70%

70% to 80%

80% to 90%

more than 90%

Predicted Repayment for Loan Recipients

Predicted percentage repaid

Results: Loan Approval and Take-up

• Fingerprinting has no affect on:

– Probability that loans will be approved by credit officers

– Probability of taking a loan

• However, fingerprint does affect loan size.

– “Worst” clients take out smaller loans by MK 2,722 (roughly US$19) (p=0.13)

15

No impact on loan officer knowledge or behavior

16

Repayment: % of balance paid on-time

17

88%

79%

91% 93%

89%

26%

74%

92%

96%98%

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Worst 2nd quintile 3rd quintile 4th quintile Best

Fingerprinted

Control

Repayment: % of balance paid (eventual)

18

92%

83%

93% 94%92%

67%

77%

93%96%

99%

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Worst 2nd quintile 3rd quintile 4th quintile Best

Fingerprinted

Control

Repayment: balance, eventual (MK)

19

1,506

2,975

1,133 1,024

1,737

7,609

3,888

1,486

572

197

0

1,000

2,000

3,000

4,000

5,000

6,000

7,000

8,000

Worst 2nd quintile 3rd quintile 4th quintile Best

Fingerprinted

Control

Fraction of land allocated to paprika

20

19%

15%

21% 22%

23%

11%

16%

19%

21%

23%

0%

5%

10%

15%

20%

25%

Worst 2nd quintile 3rd quintile 4th quintile Best

Fingerprinted

Control

Market inputs used on paprika (MK)

21

9,600

8,381

9,858

8,088

8,874

2,503

4,911

11,803

11,262

12,378

0

2,000

4,000

6,000

8,000

10,000

12,000

14,000

Worst 2nd quintile 3rd quintile 4th quintile Best

Fingerprinted

Control

Summary of results

• An increase in the credibility of a lender’s dynamic incentive raises on-time and eventual loan repayment– As predicted by theory, effect is larger the worse the borrower’s

“type”– Effect for “worst” borrowers (lowest quintile of predicted

repayment) is dramatic: 32 pp increase in eventual repayment

• Evidence of reduction in asymmetric information problems for these “worst” borrowers– Less adverse selection (smaller loan sizes)– Less ex-ante moral hazard (greater input use in paprika farming)

– No strong evidence of reduction in ex-post moral hazard (no higher repayment conditional on loan size, income)

22

Cost-Benefit analysis

• Under conservative assumptions, benefit-cost ratio for lender is an attractive 2.27– MK 476 benefit vs. MK 209 cost per individual

fingerprinted

• Could be even more attractive with:– Passage of time, as threat becomes more credible– More cost-effective equipment package– Larger volume lower cost per fingerprint checked by

overseas vendor• E.g., if in context of credit bureau with other lenders

23

Conclusions / Points for Discussion

• Results suggests benefits of establishing cross-lender credit bureau, using fingerprints as unique identifier.– Common platform should be used

• Scale-up may face several potential challenges:– Not everyone can be enrolled (UK Passport Service Trial)– Accuracy– Individuals may have a negative attitude towards technology

– Biometric technology is not infallible

24