Impact Evaluation of Smokeless Chulha Programme Ganjam and

Report
Does Multiple Borrowing in
Microfinance Necessarily Mean
Over-borrowing?
Ratul Lahkar, IFMR
Viswanath Pingali, IIMA
Santadarshan Sadhu, CMF
February 11, 2013
Outline
•
•
•
•
Background & Motivation
Data & Empirical Analysis
Findings
Conclusion
Background
•Microfinance Institutions - instrument to fight
poverty
•Proliferation of commercial MFIs
–Easy access to credit: overborrowing
–Coercive/unethical collection practices
•Irresponsible lending?
•Irrational borrowing?
Background
•Does multiple borrowing necessarily lead to
overborrowing?
• Irrational borrowing: Does the availability of
credit, and not the necessity, that influences
borrowing decisions?
• Alternative:
– Explanation in which borrowers do not seek more
loans simply because more credit sources (like MFIs)
are available.
Background
• One such explanation that readily suggests
itself is the substitution of loans
– If microfinance is more preferable, then borrowers
tend to substitute microfinance loans for other
loans without necessarily increasing their loan
burden.
– However, since microcredit institutions ration the
amount of loan given to an individual, multiple
borrowing is inevitable for obtaining more credit.
Motivation
• Recent theoretical literature (Lahkar and
Pingali, 2012) provides another explanation
of multiple borrowing on the basis of
efficient risk management
Efficient Risk Management
• In joint liability setting there is always an inherent
risk of partner default, which increases the
expected loan burden of the borrower.
– In order to mitigate this risk, a borrower can divide the same
total loan into several small portions, and borrow each
portion with a completely different group from a different MFI
• This strategy enables a borrower to diversify the risk
of a single partner defaulting on a big loan into
several partners defaulting on smaller loans.
• For a risk averse individual, this is a welfare
improving measure.
Motivation
• The theoretical framework leads to hypotheses which
we can empirically investigate.
– First, to rule out overborrowing, we should find that an
increase in the number of formal lending agencies
should not lead to more borrowing
– Second, if the substitution hypothesis is true, we must
observe that people prefer microfinance loans to other
forms of loans available to them
– Third, even if there is no overborrowing there is
multiple borrowing in the form of multiple group
membership
Objective
• Test the key hypotheses using CMF’s Access to
Finance in AP data
• Hypotheses:
– Hypothesis 1: As number of formal credit agencies in the
village increases, average loan outstanding in the village
remains constant
– Hypothesis 2: As the number of formal credit agencies in
the village increases, average loan outstanding from the
formal credit agencies increases
– Hypothesis 3: As the number of microcredit institutions in
the village increases, average loan outstanding with the
microcredit institutions increases
Sample
• Survey conducted in June to November
2009 using a rigorous random sampling
methodology
Survey details:
– 8 districts (randomly selected from 22 districts of AP)
– 64 villages (8 villages randomly selected from each of these 8
districts)
– 1920 households (randomly selected from the 64 villages)
Overview of Borrowing
• Overall indebtedness is extremely high - 93% of
all rural households in AP are indebted to at least
one source including:
• Banks (State, Private)
• Self Help Group (SHG)
• Micro Finance Institutions (MFI)
Formal/Semi Formal
• Money lenders
• Friends and relatives (with and without interest)
• Employers
• Landlords
Informal
Borrowing Landscape
Multiple Borrowing
Multiple borrowing is extremely common
– 84% of households having two or more loans from any
source.
– Median of 4 loans outstanding per household
• Multiple borrowing is driven mainly by multiple
loans from informal sources
13
0%
5%
10%
15%
Multiple Borrowing
0
2
4
6
8
10
12
14
Total Number of Loan Outstanding
16
18
20
Source: Centre for Micro Finance, IFMR Research. "Access to Finance in Rural Andhra Pradesh 2010".
14
Multiple Borrowing by Active Clients of a
Given Source
90%
85%
80%
70%
60%
50%
40%
30%
30%
26%
16%
20%
10%
0%
Banks
Informal
SHG
MFI
15
Financing of household consumption, investment in agricultural activities
major purpose of loan usage.
Significant part of MFI and SHG loans is also used for repaying old debt.
Hypotheses to be tested
– Hypothesis 1: As number of formal credit
agencies in the village increases, average loan
outstanding in the village remains constant
– Hypothesis 2: As the number of formal credit
agencies in the village increases, average loan
outstanding from the formal credit agencies
increases
– Hypothesis 3: As the number of microcredit
institutions in the village increases, average loan
outstanding with the microcredit institutions
increases
Empirical Specification: Hypothesis 1
As number of formal credit agencies in the village increases,
average loan outstanding in the village remains constant
• Need to be able to show that as the total number of
formal credit agencies in the village increases, the
average total loan burden does not.
• Regress average loan size in a village on the number
of formal credit agencies in the village and some
controls that influence the amount of loan taken
Empirical Specification: Hypothesis 1
• Use the following regression:
• Where ln(Li) represents natural log of average loan size
in the ith village, and FSC represents the count of formal
sources of credit in the village (including banks, MFIs,
SHPIs, chit agencies and cooperative societies) and X be
the vector comprising demographic & other characteristics
that influences average loan size
• For the first hypothesis to be true we must observe
that the estimated value of β1 is insignificant
Empirical Specification: Hypothesis 1
• Variables in X (controls):Several
demographics characteristics that
influence loan size in a village
–
–
–
–
Population
Per-capita irrigable land
Presence of Primary Health Care facility
Average number of times respondents in a given
village have had to incur unexpected expenditure
six months preceding the survey
– Distance to the nearest town
Results: Hypothesis 1
• β1 is insignificant: NO evidence of indiscriminate
borrowing
– Village average loan size does not depend on the number of formal
financial institutions in the village
• Controls having statistically significant effect:
– Average number of times a household incurred non-routine
expenditure in the village in six months prior to survey
• Controls not having significant effect:
– Per-capita irrigated land, presence or absence of primary
health care centres, population, distance to the nearest town
Empirical Specification: Hypothesis 2
As the number of formal credit agencies in the village
increases, average loan outstanding from the formal credit
agencies increases
• Need to be able to show that as the total number of
formal credit agencies in the village increases, the
average loan size from formal institutions increases
• Regress average loan outstanding from formal credit
agencies in a village on the number of formal credit
agencies in the village and other controls
Empirical Specification: Hypothesis 2
• Use the following regression:
• Where ln(FLi) represents natural log of average loan size from
formal institutions in the ith village, and FSC represents the
count of formal sources of credit in the village (including banks,
MFIs, SHPIs, chit agencies and cooperative societies) and X be
the vector of controls
• For the second hypothesis to be true we must observe
that the estimated value of γ1 is positive and significant
Results: Hypothesis 2
• 1 is significant and positive:
– If the number of formal credit institutions a village
has access to increases by one, then average
formal loan size increases by 3%
• Controls having statistically significant effect:
– Per-capita irrigated land
– Average number of times a household incurred non-routine
expenditure in the village in six months prior to survey
Combining Results:
Hypothsis1 & Hypothesis 2
• The overall loan burden of the village is not dependent on
the number of formal financial institutions; however, loan
from formal financial institutions is positively and
significantly dependent on number of formal institutions
the village has access to.
• As the accessibility of credit from formal sources
increases, people are tending to substitute formal sources
for informal sources.
• In other words, people seem to prefer formal sources of
credit over informal ones.
Empirical Specification: Hypothesis 3
As the number of microcredit institutions in the village
increases, average loan outstanding with the
microcredit institutions increases
• To show
– As the total number of microcredit agencies (MFI+SHPI) in
the village increases, the average loan outstanding with
microcredit institutions increases &
– Average loan outstanding with the microcredit institutions
increases faster than when compared to increase in formal
credit agencies
• Regress average loan outstanding from microcredit agencies in
a village on the number of microcredit agencies in the village
and other controls
Empirical Specification: Hypothesis 3
• Use the following regression:
• Where ln(MLi) represents natural log of average loan size from
microcredit institutions in the ith village, and MFI represents the
count of MFIs and SHPs in the village and X be the vector
comprising demographic characteristics that influences average
loan size
• For the second hypothesis to be true we must observe
that the estimated value of δ1 is positive and significant
Results: Hypothesis 3
• δ1 is significant and positive:
– As the number of microcredit institutions a village
has access to increases by one, then average
formal loan size increases by 11%
• Controls having statistically significant effect:
– Average number of times a household incurred non-routine
expenditure in the village in six months prior to survey has a
negative and statistically significant coefficient
• Seems to suggest that the villages where clients incur
greater non-routine expenditure obtain lesser amount of
loans through microcredit
• Effective screening of risky clients by JLG mechanism ?
Combining the Results….
• The overall loan size is independent of number of formal sources
of credit
• Loan size from formal sources of credit is positively affected by
number of formal sources of credit suggesting that with the
increase in the number of formal sources of credit, people tend to
make more use of such sources to meet their loan requirements.
• Loan size from microcredit institutions seems to increase faster
with the increase in number of such institutions than loan size
from formal credit sources with the increase in number of formal
sources of credit (11% with microcredit institutions as compared
to 3%)
– Even within the formal sources, borrowers seem to prefer microcredit.
MFIs and Multiple Borrowing
• Test whether borrowers resort to multiple
borrowing as the number of MFIs in a village
increases
– How?
• Measure the prevalence of multiple borrowing
by the total number of joint liability groups a
resident of the village is a member of
• Find correlation between number of MFIs in the
village and average number of groups a
resident of the village is a part of.
Result: Correlation of MFIs and number of
group membership
Correlation between total number of MFI in
the village and average number of JLG
memberships of a household
Correlation
Co-efficient
t-stat for
significance of
correlation
0.6678
6.89
• The number of MFIs present in a village and the number of
groups a borrower is a part of are positively correlated, and
that correlation is statistically significant
• Supports the hypothesis of the incidence of multiple
borrowing in the presence of multiple MFIs in the village
MFIs and Multiple Groups
• Two possible explanations
– Multiple group membership necessary to
circumvent the credit rationing imposed by
microcredit institutions
– Multiple borrowing to efficient (partners default)
risk management
Conclusions
• No evidence of indiscriminate borrowing:
– Increase in number of lending agencies need not necessarily
mean an increase in the amount of loan size in a village.
• Substitution of informal sources of credit by
formal sources when access to credit from more
organized sources is available.
• Preference for microcredit over loans from other
sources available to them
• As the number of microcredit institutions increase in a
locality, people tend to associate themselves with
more and more groups.
Thank You
Non-Routine Expenditures
Top 5 Non-routine Expenditures
Non-routine Expenditure
Share of Households which
Incurred Major Expenditure on
Item in past 6 Months
Health
36%
Festival or special event aside from
11%
marriage
Marriage
11%
Buy agricultural machinery or
10%
inputs
Home
7%
improvement/repair/construction
Any non-routine expenditure
64%
38
Non-Routine Expenditure: Source of Funding
Top 5 Non-routine Expenditures
Source of Funding Non-routine Share of Households which
Expenditure
Incurred Major Expenditure on
Item in past 6 Months
Loan from friends/relatives
Own income or savings
43%
29%
Loan from moneylender
Loan from landlord
13%
11%
Loan from MFI/SHG
6%
39
Districts Selected for Surveying
Share of poor
from NSSO
Poverty
Stratum
MFI penetration
MFI stratum
Adjusted MFI Stratum
Final
Stratum
Selected for
Surveying?
Medak
9.3
Not so poor
11.3
High penetration
High penetration
1
YES
Nalgonda
5.4
Not so poor
14.5
High penetration
High penetration
1
YES
East Godavari
3.3
Not so poor
12.5
High penetration
High penetration
1
NO
West Godavari
4.4
Not so poor
12.3
High penetration
High penetration
1
NO
Krishna
2.8
Not so poor
18.7
High penetration
High penetration
1
NA
Guntur
3.9
Not so poor
13.2
High penetration
High penetration
1
NO
Vizianagaram
4.7
Not so poor
4.7
Low penetration
Low penetration
2
YES
Cuddapah
5.4
Not so poor
9.9
High penetration
Low penetration
2
YES
Karimnagar
7.2
Not so poor
5.5
Low penetration
Low penetration
2
NO
Warangal
0.9
Not so poor
6.1
Low penetration
Low penetration
2
NO
Srikakulam
6.0
Not so poor
4.4
Low penetration
Low penetration
2
NO
Nizamabad
23.1
Poor
9.1
High penetration
High penetration
3
YES
Visakhapatnam
18.9
Poor
10.6
High penetration
High penetration
3
YES
Khammam
13.1
Poor
10.1
High penetration
High penetration
3
NO
Nellore
14.1
Poor
10.9
High penetration
High penetration
3
NO
Kurnool
24.6
Poor
8.6
Low penetration
High penetration
3
NO
Mahbubnagar
11.8
Poor
2.9
Low penetration
Low penetration
4
YES
Prakasam
9.9
Poor
7.7
Low penetration
Low penetration
4
YES
Adilabad
26.1
Poor
4.0
Low penetration
Low penetration
4
NO
Rangareddi
10.9
Poor
6.0
Low penetration
Low penetration
4
NO
Anantapur
20.2
Poor
4.1
Low penetration
Low penetration
4
NO
Chittoor
15.9
Poor
8.4
Low penetration
Low penetration
4
NO
District
40

similar documents