Pro-Poor Growth & Microfinance: Some Related Evidence, and a

advertisement
Pro-Poor Growth & Microfinance:
Some Related Evidence,
and a Research Agenda
Dean S. Karlan
Princeton University,
M.I.T. Poverty Action Lab
Jonathan Zinman
FRBNY*
World Bank
April 21st, 2005
* Views expressed are those of the authors and do not necessarily represent those of the Federal Reserve
System or the Federal Reserve Bank of New York.
Some Key Questions, &
Overview of Talk
I. Can microfinance be used to promote
pro-poor growth?
II. If it can, how?
Talk today:
1. Outline research questions we need to
answer to help address I. and II.
2. Outline related Karlan-Zinman field
experiments and findings
Research, Microfinance, and
Pro-Poor Growth
Some research findings we need to help answer the big questions:
1. How do the poor make (financial) decisions?
–
Do people make the “right” decisions?
2. How do financial markets work, and not work, in terms of bringing
together capital and productive opportunities (broadly defined)?
–
–
If there are financial constraints, what underlying frictions cause them?
What is the nature of financial constraints?
3. How large are marginal returns, broadly defined, to
borrowing/investing?
–
–
Private returns
Social returns
4. If 1-3 motivate interventions, which ones are most effective?
–
–
Optimal design ex-ante
Evaluation ex-post
Set of Research Questions #1:
How do the Poor Make Decisions?
• Response to incentives
• Response to intertemporal tradeoffs
• Importance, or lack thereof, of
“behavioral”/“psychological” factors, of
bounded rationality
– Do folks make the “right” decision?
Set of Research Questions #2:
How do financial markets work, or not?
•
•
•
Lots of theory (e.g., on adverse selection and
moral hazard)
Lots of practice
Little clean evidence on specific failures
–
–
Even best work on the finance-growth nexus is very
reduced-form, looks at symptoms of financial
frictions rather than diagnosing specific problems
Particularly true of information asymmetries
•
•
Chiappori and Salanie (2000 survey article)
Nobel Committee citation for 2001 Prize
Set of Research Questions #3:
What are the marginal borrower/investor’s
returns?
• The trillion-dollar “impact” question
– Has microfinance delivered on its promise?
• Again, theory and practice far ahead of evidence
• Keys to getting better answers here:
– Defining and measuring impacts broadly
– Measuring impacts cleanly (methodology)
– Benchmarking any impacts against alternative (social)
investments
• I.e., can’t ignore opportunity cost of allocating resources to
microfinance
Set of Research Questions #4:
Interventions
• If basic research (the “R” in “R&D”)
produces evidence that favors intervention
in microfinance markets, what next?
• The “D”, and the “E”
– “D”evelop and “D”esign Interventions
– “E”valuate
“Market Field Experiments”
• Answering Questions #1-#4 is difficult
– Identifying causality
– Identifying deep economic parameters of interest
• What we’ve been doing:
– Designing “market field experiments” meant to identify
deep parameters
– Finding financial institutions willing to implement
randomized-control designs as part of their day-today operations
– Working with institutions to implement experimental
protocols subject to operational constraints
– This type of partnership between academics and
firms is novel, especially in a market setting
Interplay Between Field Experiments &
Other Methodologies
• Field Experiments not a panacea, but
complement to other methodologies:
• Strengths:
– Clean evidence derived from “gold standard”
methodology of behavioral sciences
– Large stakes
– Natural setting
• Weaknesses:
– Expensive
– Less control than, e.g., lab
– External Validity
New Evidence on Questions #1-#4 from
Karlan-Zinman Field Experiments
• Experiment #1: Randomize interest rates and
marketing strategies offered by South African
consumer lender
• Quick background:
– “Cash loan” market providing term loans (modal 4
months) at 12% per month
– Targets working poor
– Market sprung up to replace moneylenders following
usury deregulation
– Dominated by for-profit lenders
Experiment #1: Design Overview
• Randomize marketing strategies
• Randomize interest rates along 3 different dimensions:
– Single dimension sufficient for deriving demand curves for
consumer credit
– Multiple dimensions needed to identify and disentangle whether
adverse selection and moral hazard needed in this market
• “Offer rate” advertised on direct mailers sent to 60,000 former
clients
– Offer rate is generally =< Lender’s standard rate
• “Contract rate” revealed to clients only after the come in to apply,
hence revealing demand to borrow at their offer rate
– Contract rate always =< offer rate
• “Dynamic repayment incentive”
– All randomizations conditional on observable risk
Identifying Info Asymmetries:
Basic Intuition Behind the Design
High Contract Rate Low Contract Rate
High Offer Rate
Low Offer Rate
N/A
Adverse Selection
Moral Hazard / Repayment Burden
What Have we Learned from
Interest Rate Randomizations?
Re: Question #1 (Decision-Making)
• Intertemporal tradeoffs: these borrowers are
price-elastic on average, but:
– Demand curves are relatively flat (contra recent
evidence from US showing price elasticities > |1|
– Elasticity is decreasing in income
– Female borrowers are more elastic than males
– They are more elastic with respect term (a la
Attanasio, Goldberg & Kyriadzidou 2004)
– See KZ 2005 on Demand Curves and Credit
Constraints (new draft soon)
What Have we Learned from
Interest Rate Randomizations?
Re: Question 2. How financial markets work:
• Evidence that both adverse selection and moral
hazard matter:
– But surprising pattern by gender: only female
borrowers exhibit adverse selection, only male
borrowers moral hazard
• Not necessarily gender per se
– Effects are large where present
• 20% of defaults
– Effects are consistent with “relationships” mitigating
information problems
– But: functional form (power) issues
Project #1:
Marketing Randomizations
Evidence on Question #1 (Decision-Making)
• See Bertrand, Karlan, Mullainathan, Shafir, and Zinman (2005)
• Direct mailers included randomly assigned marketing “treatments”
motivated by (lab) findings from psychology
• Treatments manipulated how loan offer was “cued” and “framed”
• Examples:
–
–
–
–
Deadlines
More v. less information
Photos
Suggestions
• Predictions:
– Psych/Behavioral Economics: These treatments will affect demand.
(But how much?)
– Neoclassical Economics: treatments irrelevant
Marketing Randomizations:
Novelty
• What’s unique here compared to lab
findings, and similar marketing field
experiments
– Real stakes
– Commodity (i.e., not a branded product)
– Consumers familiar with product (borrowed
before)
– Marketing effects “priced”/scaled vis a vis
interest rate elasticity
Marketing Experiment:
Findings and Lessons
• Many treatments do matter
• But was hard to predict ex-ante (from lab,
theory) which would work in our setting
• Are psychologists right that context matters
much, and consequently that it’s difficult to
create general theories of consumer choice (and
human behavior more generally)?
• Consider framing effects when designing and
marketing programs (Question #4)
Project #2: A new experiment
Re: Question #3. Marginal returns, and the billiondollar impacts question.
Design:
• Work with lenders to randomly assign loans to marginal
applicants who would normally be rejected
– South Africa, Philippines
– Consumer loans, commercial loans
• Follow up 6-months later with household surveys to
measure impacts
– On households (wide range of proxies for well-being)
– On micro-businesses
• Then compare outcomes (and inputs) of those who
randomly got loans (the “derationed”) and those who
stayed rejected (the “rationed”)
Take-Aways
• Microfinance’s role, if any, in promoting
pro-poor growth depends on answers to
several questions on which we still lack
convincing evidence
• Market field experiments can help answer
these questions
• Field experimentation can then feed back
into other, complementary methodologies
Download